Neo-Davidsonian Metaphysics: From the True to the Good

Wheeler: Neo-Davidsonian Metaphysics: From the True to the Good

Introduction page 1

Introduction

I Re-launching the Davidson Program

Donald Davidson’s program, as I encountered it in the late

1960s, was inspiring and exciting. It appeared that an account of

logical forms, coupled with a semantics that eschewed

metaphysics, would soon solve or dissolve many philosophical

problems. Davidson’s thesis was that the first step to dealing

with a philosophical problem was getting the semantics straight.

For instance, the first step in answering the question “what

things are good” is finding out the logical form of sentences

using “good.” The hope was that this kind of ground-clearing

would enable real progress on the problems that did not, like the

“problem of predication” or whether to believe in sparse or

abundant properties, disappear with a proper understanding of

semantics.

Much of Davidson’s thinking was profoundly anti-

metaphysical, and the Davidsonian program was likewise anti-

metaphysical. Predication has no better account than “`is a frog’

is true of an object just in case the object is a frog.” Truth is

not correspondence to anything. Davidson’s account of meaning

extended Quine’s ideas about radical translation to a theory of

interpretation and an account of meaning without a metaphysics of

meanings. Much of Davidson’s work continued the anti-

essentialist, anti-metaphysical logical positivist tradition,


Introduction page 2

albeit without the dogmas, without the empiricism, and by and

large without the scientism.

Davidson himself implemented only a few parts of this

program, and only suggested lines of thought for other parts of

the program. His work on events, causation, adverbs, and the

mind-body problem were actual concrete applications of his

semantics. The project of finding logical forms for constructions

of natural languages was being taken up most effectively by

linguistic semanticists following the alternative path of

Montague.1 This side of recursive truth-conditional semantics

was, from a Quinean point of view, completely shameless in

invoking possible worlds, exotic functions, and the like.2 The

linguistic semanticists by and large accepted notions of

presupposition, lexical meanings, and other notions which a

Davidsonian would eschew. Considering the scope of the original

program, relatively little work was carried out trying to

implement the Davidsonian program in the austere form that it

began.

1 A few of his students produced Davidson-inspired accounts of adjectives, proper names, and quantifiers. See Burge (1972, 1973, 1974) Wallace (1971, 1972) and Wheeler (1972, 1974, 1978). The vast majority of writing connected with Davidson from philosophers, as one would expect, consisted of claiming that there are flaws in Davidson’simplementation of the program rather than carrying out aspects of the program. 2 Several years ago, I gave a graduate seminar for which the texts wereLarsen and Segal’s (1995) and Heim and Kratzer’s (1998). The conclusion reached by the seminar is that these came close to being notational variants.


Introduction page 3

Davidson thus left many important topics untouched. Davidson

followed Quine in not trying to give a semantics for modalities.3

Unlike Quine, though, Davidson cannot just claim that there are

no necessary truths. As long as some sentences using “necessary,”

“possible” and the like are true, these words must make some

contribution to the truth-conditions of sentences. Other

questions which the Davidson program in principle had to give a

semantics were never addressed either by Davidson or his

acolytes. For instance, the question of what propositions,

properties, and facts might be has to have an answer, given that

there are true affirmative sentences using those terms. Even

though Davidson has shown that properties play no role in

understanding predication, that propositions as meanings of

sentences are not necessary in semantics, and that facts play no

role in providing an entity in the world for a sentence to

correspond to in order to be true, still there are truths using

those count-nouns whose truth-conditions need to be given some

sort of account.

Davidson had views about the objectivity of values and

ethical notions, which appeared in scattered articles. The

philosophical field of ethics, though, was never one of his main

3 Davidson gives hints and suggestions. Davidson (1968) suggest the beginnings of a semantics for modalities, but “modality” is not mentioned by name. Davidson (1970a) suggests what a semantics for “ought” should look like, in the course of discussing a problem in thephilosophy of mind and action.


Introduction page 4

interests. His earliest work was on decision-theory, and this was

a continuing interest and topic of his writing. The project of

connecting the theory of preference with a theory of what is good

or what a person should do never got done.

In sum, many parts of a completed Davidsonian program were

never even begun. Part of the explanation is to be found in the

resurgence of realism. Sometime after 1970, the majority, of

philosophers abandoned the whole logical positivist, anti-

metaphysical attitude toward philosophical problems that had

culminated in Quine and then in nearly pure form, in Davidson’s

program.

The very intuitive arguments of Kripke (1980) convinced even

Davidsonians that appeals to natures made intuitive sense and had

to be right. In my own case, I became convinced that Aristotle’s

Metaphysics, especially Zeta, Eta, and Theta, were almost exactly

right. There had to be natures of things and the de re

necessities that would be the consequences of such natures. So,

at least some Davidsonian disciples, and certainly very many

philosophers who might have been attracted to the program, lost

faith.

Now, in 2013, metaphysics is a thoroughly respectable field,

with very intelligent philosophers arguing pro and con about

whether truth is an explanatory property, what mysterious bonding

joins universals to particulars to yield facts, whether Ferraris

are entities which persist or perdure, and the like. For


Introduction page 5

instance, the question is asked in the philosophy of mind whether

property dualism or substance dualism are equivalent or whether one

is preferable to the other.(Lycan, forthcoming). The Quinean-

Davidsonian view that property versus substance talk is

misdirected talk about predicates is ignored.

Davidsonism and pursuit of the Davidson program has thus

become a distinctly minority view. Davidson scholarship, while a

burgeoning literature, has become mostly exposition of the

details of his actual writings on various topics, rather than an

effort to carry out the program. I think this is a very large

mistake, and that what has happened is that philosophers have

forgotten Davidson’s basic insights about truth, predication, and

interpretation. An example of this forgetting is the continued

proliferation of “logics” for the various modal predicates and

tenses.4 Quine’s notion of logic as pure structure and Davidson’s

minimalist conception of semantics would put these “logics” in

their proper place as theories of the truth-conditions of

predicates.

A couple of decades ago, I began to realize that the

original program was not a dead end, but had been abandoned for

inadequate reasons. A kind of forgetting had taken place,

4 Admittedly, both Davidson and Quine took it as obvious that tenses involved an extra argument-place for times. But that supposition is not essential to their basic anti-essentialist, anti-metaphysical view, as we will show in Chapter 7.


Introduction page 6

analogous to the forgetting that Kripke5 rectified. The

philosophical orthodoxies before Kripke’s 1970 lectures are a

good example of the kind of forgetfulness that can occur in

philosophy. As many people have realized, the intuitions Kripke

was appealing to are essentially the same as those Aristotle was

appealing to in Metaphysics Zeta, Eta and Theta. Aristotle was

responding to the challenge of Heraclitus against continuants as

well as the inadequate defense of common sense from Plato.

Aristotle’s distinction between essence and accident was a

defense of common sense. Kripke appealed to essentially the same

intuitions. In Kripke’s case, of course, the frame of discussion

was names and the conditions for their application. Very

different philosophical environments generated very similar

accounts of what it takes for this person to be the same over

time or in different circumstances. To remind ourselves how much

things have changed, remember that Quine (1953b) took the

necessity for appealing to Aristotelian essences to be a decisive

reason to reject the third grade of modal involvement.

Aristotle’s insights about the necessities implicit in the

idea of a lasting being were forgotten because Aristotle took the

medium-sized objects of everyday life to be also the primary

5 Kripke was not alone in implementing this resurrection. Modal logicians such as Hintikka, Marcus and others had views which likewisewere close to recapturing the Aristotelian insights via thinking aboutnaming. But the single most effective event was the series of lecturesin the Woodrow Wilson School in 1970.


Introduction page 7

terms in which scientific explanation takes place. When It became

clear that scientific explanations required something other than

the objects of ordinary life, corpuscles or atoms rather than men

and earth, all of Aristotle was abandoned, the insights along with

the short-comings.

Something similar happened in a shorter time frame with

Davidson’s thought. The main apparent defect in Davidson’s

program was his rejection of realistic metaphysics, the idea that

nature is itself “divided at natural joints.” It seemed to

almost everyone that a realistic metaphysics was required in

order to accommodate natural Aristotelian Kripkean intuitions, so

that Davidson’s views could not be right. As with the rejection

of Aristotle in the 17th century, the rejection of Davidson’s

program threw out the good with the inadequate. Just as

Aristotle’s central insights about the conditions for continuants

remained valid, so, it seems to me are Davidson’s basic anti-

metaphysical ideas about semantics, predication, and

interpretation. In addition some of the core ideas in Davidson’s

implementation of the program are in fact part of the correct

account, even if the exact version Davidson proposed is mistaken.

There were other reasons philosophers ceased to take the

Davidson program as worth pursuing. Some of the abandonment of

the program was due to excessive focus on the particular

implementations Davidson himself made of his central ideas rather

than on the fruitful central ideas themselves. Some of Davidson’s


Introduction page 8

particular implementations and applications of his program have

come under attack and have rightly been judged unacceptable. His

elegant solution to the mind-body problem, anomalous monism, is

flawed by his residual scientism. Davidson’s views about

anomalous monism, events, and causation were more extreme than

they needed to be, and implausible. But that does not mean that

the fundamental ideas were mistaken. The idea that the mind-body

problem is about different systems of predicates is still a good

one, and basically right. The vast literature on anomalous monism

and event-identity, though, largely abandoned Davidson’s idea

that mis-matching predicate-systems rather than properties were

the key to understanding. The central idea in Davidson (1967b and

1967c), that problems of logical form can be solved by

quantifying over something, applies whether or not those

somethings are limited to events.

Another reason philosophers abandoned the program was that

Davidson paid relatively little attention to developments in

linguistic semantics. Thus some of his semantical views seem

quaint. He never abandoned the treatment of quantifiers as

operators. He continued to think that the truth-functional

conditional was all you need to say about “if” even though his

own work (Davidson 1970b) showed that that could not be the case

with conditional “ought” sentences. But the conception of

quantification as operators on open sentences and “if” as a

truth-function was inessential for Davidson. Nothing important


Introduction page 9

changes if quantification is regarded as primarily set-theoretic

and a better theory of “if” is part of a truth-definitional

theory. More generally, the very great progress in linguistics on

a variety of topics can, I believe, can be incorporated into an

essentially Davidsonian conception of semantics. Part of this

incorporation is accomplished just by enforcing Davidson’s very

austere notion of what a semantics should be and reclassifying

the parts of these linguistic theories that do more than

semantics requires as theories associated with predicates.

Davidson’s idea was that enough structure should be assigned to a

kind of sentence so that a speaker with finite learning capacity

could understand an infinity of sentences. To assign more

structure, for instance to suppose that “is a horse” has a

structural element “is an animal,” so that the inference from

“Stewball was a horse” to “Stewball was an animal” is formal goes

beyond anything that motivates assigning logical forms in the

first place. Quine’s idea that logic is indifference to which

particular predicates occur, and that everything else is theory,

has been lost if semantics does more than it has to.

But the most important reason the program was abandoned is

the first one. Philosophers took it as obvious that Davidson had

no way of accommodating intuitions about the persistence of

objects or counterfactuals about things that could have happened

to Aristotle and Ben Franklin, given his denial of given natures

in Davidson (1974). It seemed that some kind of objective


Introduction page 10

privilege had to distinguish some predicates from others. A

natural understanding was that correspondence to a real division

in nature was the basis of that privilege.6 If a “given” domain

of objects is required in order to accommodate some truths, that

completely undermines Davidson’s talk of predicates rather than

properties and so abandons the entire Quinean-Davidsonian

perspective on how to think of questions about what there is.

I argue on the contrary that Davidson’s views on the given

are entirely compatible with Kripkean and Aristotelian intuitions

about what Aristotle might or might not have turned out to do for

a living and every other “metaphysical” intuition. Davidson,

given his view that most of what people believe is true, has to

have some explanation of these intuitions about the continued

existence of medium-sized objects which makes them come out true.

Unlike Quine, who is willing to jettison “common sense” in favor

of physics, Davidson, at least in principle, has to accommodate

common sense. On a number of topics I supply what I think

Davidson ought to have said about essentialism, properties, the

mind-body problem, properties, and facts. If we discount the

residual scientism that Davidson seems to have inherited from

Quine, and focus on what someone with Davidson’s views who

6 My 1975 essay applying the Sorites argument was an argument that, if the real objects were those which corresponded to joints in nature, the real objects would be micro-particles. Since the objects of ordinary experience have no determinate relation to micro-particles, they are strictly not real beings.



completed parts of his program and accommodated the truths of

common sense might have written, we get the beginnings of the

implementation of a program which is Davidsonian, but which

differs in a number of respects from the implementation begun by

the actual Davidson, while inspired and shaped by Davidson’s work

on these topics. I think I have found what Davidson should have

said about why a person can gain weight and continue to exist but

not be made into sausage and continue to exist. Chapters 2 and 3

construct an essentialism which is relative to predicates and

which treats entities as posits required for thinking rather than

articulations of reality.

So, in brief, this book is an attempt to re-launch the

Davidson program, adhering to his fundamental insights about

semantics, interpretation and predication while differing in many

respects from the implementation that Davidson actually produced,

accommodating essentialist intuitions and dealing with issues,

such as modality, temporal continuants, sorites arguments,

ethical concepts, properties, and facts that Davidson never got

around to addressing. The book offers the outlines of a

Davidsonian account of these topics in some cases, such as the

modals, and

This book consists of ten chapters on a wide variety of

topics. After the first three chapters, which introduce the

essentials of Davidsonian semantics, his account of

interpretation, and a view I call “relative essentialism,” the



rest of the chapters are outlines of how to continue the

Davidsonian program. Not every topic about modality, the

philosophy of time, or ethics is dealt with. Rather, the intent

is to show how the central Davidsonian insights about

predication, interpretation, and the assignment of logical forms

can illuminate, solve or dissolve a variety of problems of

semantics, ethics, and metaphysics. The conclusion in most of the

chapters is that metaphysical theories are unmotivated. There is

no need or role for a science which would supplement natural

science and common sense about the nature of material objects.

The accounts of physics and the natural sciences and the common-

sense concepts about medium-sized objects, organisms, and persons

exhaust the real questions. If you are interested in the real

nature of time or matter, for instance, ask the physicists. If

you wish to know what it takes for Joe to be a frog, there is a

simple answer. “Joe is a frog” is true if and only if Joe is a

frog. If you want a more detailed answer, consult biologists and

read some natural history. That there might be more fundamental

questions, deeper than biology, about what has to obtain for the

predicate “frog” to be true of an object is a mistake. There just

is no “problem of predication.” The mystery of how there can be

entities, namely properties, which are the same even though

multiply located, disappears when we take seriously the idea in



Davidson’s (2005)7 that there is nothing illuminating to say

about predication beyond statements such as “`Is a frog’ is true

of an object just in case that object is a frog.”

The last two chapters are not so much a rejection of

metaphysics as a presentation of ways a Davidsonian approach

could clarify ethical discussion. An understanding of logical

form would allow ethical enquiry to get somewhere, rather than

arguing about intuitions. As in the previous chapters, it would

be expected that very much excellent work would be, slightly

reformulated, a part of an adequate theory.

II The Chapters

Chapter 1 is a discussion and development of Davidsonian

principles of semantics as I understand them, focusing on the

topics that will come up in succeeding chapters. The crucial

parts of Davidson’s view for my purposes are his account of truth

and predication, his conception of interpretation as

rationalization, and his conception of semantics as distinct from

theory. I defend the idea that a disquotational truth-definition

is the only proper semantics. Whatever else we know or do not

know about what the conditions are for being a pig, we can be 7 I should mention that Davidson held this view of predication long before 2005. I called him during the early 2000s and asked him what hewas working on. He told me that he was writing an account of predication which for decades he had thought was completely obvious. He said it had only recently dawned on him that not everyone knew thatpredication was no mystery at all. The view is also implicit in Davidson (1967a), but “implicit” does not entail “understood by most readers.”



sure that “is a pig” is true of an object if and only if that

object is a pig. Semantical theories which go beyond disquotation

for syntactically simple predicates are properly understood as

theories of those objects. Given this austere semantics,

interpretation plays a major role in how we understand one

another.

Chapter 2 is an appreciation and critique of absolute

essentialism, the idea that the world is intrinsically divided up

“at the joints” into entities and properties of those entities.

It focuses on Aristotle, the philosopher who realized that you

cannot posit a kind of object without supposing objective

necessities about it, and Kripke, who revived Aristotelian

essentialism. It consists of two main arguments: The first

argument is that many kinds of entity which we would be reluctant

to regard as natural kinds have many of the features of natural

kinds. Terms for them are “rigid designators” and their

extensions are not determined by conceptual contents.rigid

designation of kinds. There are things Pepsi and bouillon could

be and cannot be which are independent of a person’s conception

of them. The second argument is that, from an objective point of

view, there is very little reason to take, for instance “has

atomic number 79” as an essential property of a stuff. The stuff

necessarily has atomic number 79 given that it is gold, but that

is not an obvious “given.” The stuff itself could be the same

stuff and not be gold. The mere change of a single quark from up



to down would make an atom be one or another element. Nothing is

special about “elements.”

Chapter 3 presents a view of natures, properties, and beings

that accommodates intuitions about persistence without supposing

that nature itself is intrinsically articulated. It is a view

that Davidson could have accepted which accommodates Kripkean8

intuitions about objects while respecting Davidson’s view that

there is no “given” domain of beings, but only different

predicate systems. The basic problem for a Davidsonian

accommodation of the insights Kripke presents is that Davidson is

committed to a view on which there is no “given’ domain of

objects. Using the material from Chapter 2, I argue for what I

call “relative essentialism,” the view that the articulation of

reality into beings and properties is a requirement for thought,

given that thought must represent sub-sentential formal logical

relations among sentences. Thus, in a somewhat Kantian way,

positing beings and properties is a precondition for thought. I

argue that the happy fact that reality seems to come in beings

which have properties is not our conformity to a given

articulated world, but rather our doing, in some sense of “us”

and “do.” On this understanding, Quine and Davidson are both

8 These insights accord very well with Aristotle’s views in Metaphysics Zeta, Eta and Theta, as noted above. I might remark autobiographicallythat after hearing Kripke’s lectures, and remembering the Aristotle seminar I took from Terry Penner, I took classical Greek for three years so that I could read Aristotle and understand the secondary literature.



Kantians, in the sense that they regard the positing of beings as

impositions from us, rather than a reflection of an articulation

in nature itself.

Positing a kind of object, though, means positing

persistence and identity conditions, which generate modal truths

about those objects. I argue that Davidson and Quine are

implicitly committed to a kind of at least probabilistic

essentialism. Given that we posit all sorts of objects, there can

be distinct physical objects, each with a kind of essence,

occupying the same space at the same time. This does not mean

either that there are alternative conceptual schemes or that

there are not really chairs and electrons. It does mean that

kinds of entity should be thought of on the model of units of

length. No one would deny that a hundred yards separates the two

goal lines on a football field. But few would assert that the

field itself is naturally and intrinsically divided into one-yard

chunks.

That agents themselves are also posits rather than a given

part of the articulation of reality is not a problem. Given that

being an agent and thinking of oneself as an agent within the

intentional family of concepts is a condition of being a language

user, there after all really are for us no global alternative

conceptual schemes.

Chapter 4 extends the results of Chapter 3 to the category

of events. It argues that distinct events can occur in the same



place at the same time, just as distinct physical objects can

occupy the same space at the same time. Events of various kinds

have (relaive) essences, and distinct events of different kinds

can occur at the same place and time, even though they have the

same causes and effects. This result is applied to Davidson’s

(1970a) argument for anomalous monism. The view of the relation

of the mental to the physical I propose could be called innocuous

dualism.

A discussion of causation urges that facts and states, which

are distinct from events, can be referents of verb-phrases and

also causes. The core of Davidson’s analysis of adverbs is that

something is quantified over to which many predicates are

ascribed. This core is independent of the question of what kind

of entity is quantified over. Facts and states, properly

construed as metaphysically inert constructions out of things

said, can perfectly well be quantified over and can perfectly

well be arguments of a “cause” predicate. What facts and states

are is the topic of Chapter 6.

Chapter 5 is a Davidsonian account of modality and

conditionals. Chapters 2, 3, and 4 make extensive use of modal

predicates, and an account is called for. The first sections make

the obvious adaptation of Davidson’s (1968) account of indirect

discourse by treating modals as predicates of demonstrated things

said. While this would be adequate for some special modalities,

it will not work for modalities in general. The situation is akin



to that with the quantifiers—the extreme quantifiers “all” and

“some” can be treated as generalized truth functions, but many

others cannot. In particular, conditional modalities, such as

“ought” and “probably” cannot be treated as a modal applied to a

conditional. An account of modality must therefore also be an

account of conditionals. I defend the default view of

conditionals that “if” is univocal and that “then” is an

independent word, so that it is a mistake to treat “if..then” as

a single semantic unit. The account of conditionals developed

treats “if” as something akin to punctuation indicating first

arguments of two-place predicates of things said. In this respect

the theory of conditionals is akin to Kratzer’s (2012) account,

minus possible worlds and a few other things. The modals “ought”

and “should” get further development in Chapter 10.

Chapter 6 is a discussion of properties, propositions and

facts. While Davidsonian semantics gives these three sorts of

entities no role whatsoever in semantics, they surely exist,

given that there are true sentences about them. This chapter

supplies an account of what these objects are that makes truths

about them come out true. On this account, they are useful posits

constructed from utterances, but with no role in semantics. As

with the previous chapter, this answer takes as the basis for all

three kinds of entities Davidson’s (1968) account of indirect

discourse.



Chapter 7 is an illustration of how Davidson’s account of

predication and truth, combined with a generalization of the

treatment of modality in Chapter 5 and the semantically and

metaphysically inert character of Davidsonian properties and

facts discussed in Chapter 6, can dissolves metaphysical puzzles.

Two ancient metaphysical problems about time are the problem of

future contingents and Heraclitus’ problem of accidental

intrinsics. This chapter argues that those puzzles have no force

whatsoever without truth-maker semantics and an ontologically and

semantically weighty conception of properties and facts. No

exciting new logic is called for to allow an open future, and no

problem arises from thinking that the very same person is now

older than she was.

Chapter 8 is a discussion and resolution of the Sorites

paradox, which has been a topic of previous chapters and a long-

standing puzzle to me. The Sorites raises a grave difficulty for

monistic essentialism, since, if natures of things owe to reality

being divided at the joints, and there is only one set of joints,

entities which are not reducible to basic entities have no

natures. The Sorites thus threatens the reality of medium-sized

objects, since it shows that medium-sized objects and their

properties cannot be reduced to complexes of micro-particles and

their properties. Given that the real joints, if there were such,

would be at the micro-particle level, there seems to be no



metaphysical basis for thinking medium-sized objects can have

natures and so be objects at all.

The Sorites is a main reason to be a relative essentialist

rather than a monistic essentialist. The struggles to be a “cut-

at-the-joints” realist and have real medium-sized objects and

children evaporate with relative essentialism. With relative

essentialism, many objects overlap without getting in each

other’s way, in something like the way feet, cubits and meters

co-exist on the football field. One of the features of the

ontological permissiveness allowed by relative essentialism is

that most kinds of objects and properties will of course lack sharp

definitions in other terms. While the metaphysical issue of how

medium-sized objects of the lived world co-exist with micro-

particles does not arise for relative essentialism, the logical

problem of whether a bivalent logic is appropriate given that it

is in principle indeterminable in some cases whether an entity is

a pig or not, does.

This chapter therefore offers a kind of pragmatic solution

to the problem of what logical system is appropriate for

understanding human language. I argue that there is a sense in

which every predicate is completely determinate and that

bivalence can be accepted on practical grounds.

Chapter 9 is an account of the semantics of sentences using

“good.” Davidson said in class in 1967 (Wheeler 2012) that the

first thing an ethical theory should look for is an account of



the logical form of sentences using “good” and “ought.” Meta-

ethical writing almost never does this. Some of the things

writers say about “good” contradict basic constraints on what the

truth-conditions of sentences using “good” could be. Almost all

of the ethical writing about what we ought to do is oblivious to

facts about logical form which Davidson (1970b) pointed out long

ago.

Whereas values are always part of Davidson’s understanding

of interpretation, since he always characterizes “charity of

interpretation” in terms of treating others as “believers in the

true and lovers of the good,” most discussions of Davidsonian

interpretation focus exclusively on agreement in belief. This and

the next chapter correct that.

“Good” is a member of a subset of the comparative adjectives

with the following features: Such adjectives create intensional

contexts, require the inference from “good F” to “F” and have

comparatives that are apparently relative to a kind. You can be a

better tennis player than I am, but not a better pool player.

Nothing like this is possible with normal comparative adjectives

such as “tall.” If you are a taller basketball player than I am,

you are a taller anything.

The logical form of the broader category of comparative

adjectives is itself a topic that has generated hundreds of

accounts, a few of which I discuss. Most of those accounts are

incompatible with a Davidsonian disquotational semantics. The



chapter develops a Davidsonian account of comparative adjectives,

adapting the idea of quantifying over something in Davidson

(1967b) and extending that account to cover the odd kind of

adjective that “good” is. It offers the beginnings of a theory of

“good.”

Chapter 10 is an account of “ought.”This chapter first shows

that “ought” and “obligation” are completely different modal

notions with different logical characteristics, and diagnoses the

confusion most ethical theorists have had since at least the time

of Kant. I argue that, while obligations are an important aspect

of ethics, “ought” is the fundamental modality of ethical

reasoning.

The theory of the truth-conditions of applications of “good”

suggested at the end of Chapter 9 connects “good” with “ought.” A

good bagel is one you ought to want if you want a bagel. “Ought”

was one topic of Chapter 6, on modalities. Whereas that chapter

was concerned only with the logical form of sentences using

“ought,” this chapter develops a theory of what it takes for a

conditional “ought” predicate to apply to a pair of propositions,

understood as “things said.”

The theory articulates “ought” as a chain of conditional

probabilities; that is “If P then probably Q” sentences. Those

conditional probabilities, in the case of “ought”-sentences about

agents, are supported by Davidsonian principles of

interpretation. Those principles are normative, in a broad sense.



Briefly, if correct interpretation maximizes agreement in beliefs

and desires, treating the other agent as a believer of truths and

a valuer of the good, then, just as most of the beliefs we hold

about pigs are attributed to the other when we interpret one of

the other’s predicates as being true of an object if and only if

“is a pig” is, so with the valuations we hold regarding pigs. In

the simplest cases, this will mean that if a term is correctly

interpreted as meaning “chair,” it will serve interests in the

life of the other that chairs serve for us.

Moral and prudential uses of “ought” get analogous

treatment. The prudential “ought” applies our concept of rational

agency to conditional “ought”-sentences with desires as

antecedents. The moral “ought” takes as antecedent “is an agent.”

The ethical theory that comes out of Chapter 10 differs from

those current in at least three ways. First, “good” and “ought”

are univocal. There is no special moral sense of these words.

Rather, moral uses arise from different first arguments of

conditional “ought”-sentences. Second, many ethical questions are

absolutely undeterminable, since the predicates that generate the

meaning of applications of “good” are vague predicates without

definitions in other terms. Third, the model of ethical reasoning

is induction rather than deduction. “Ought” is taken as the

fundamental ethical modality. General principles about what one

ought to do can only be guidelines “Obligation” and related

predicates are a very distinct family of modals whose logic is



deduction. Thus principles, for a Davidsonian, can only be

guidelines.

III Thanks

A number of people have given me hours and hours of their

time, reading drafts of various chapters, saving me from gaffes,

telling me about important literature I had missed and making

very useful suggestions and criticisms. Bill Lycan, John Troyer,

Lionel Shapiro, Marcus Rossberg, and Crawford Elder read various

chapters and made very helpful comments and suggestions.

Distaste for the views expressed did not deter them from helping

this enemy of truth convince others of his mistaken views. Mike

Lynch read the prospectus and this chapter. His sage advice is

responsible for the impression this introduction gives that this

would be an interesting book to actually read. The publisher’s

readers, especially the least enthusiastic of them, made the book

much better than it would have been.

Over the past fifteen years, the assembled graduate students

and faculty at our department’s weekly Wednesday Brown Bags have

heard versions of many of these chapters and have given me much

assistance in understanding how anyone could not immediately be

convinced that what I am saying is correct. The Logic Group at

the University of Connecticut heard early versions of two of the

more technical chapters, giving me much helpful feedback. In that

group, Magdalena and Stefan Kaufmann’s willingness to take my way



of dealing with modalities seriously convinced me that my views

might not be entirely crazy.

Two of my graduate students over the past fifteen years have

written dissertations showing how parts of the Davidson program

have not been shown inadequate after all. Daniel Blair (2003) and

Nilanjan Bhowmick (2012), using their vast knowledge of the

linguistics literature, have bolstered my confidence that the

Davidson program is not dead, but in a kind of suspended

animation, thus making it possible to hope that a relaunching is

possible.

Chapter 1: Davidsonian truth and its consequences

“The beginning of wisdom is the realization that sameness is

always relative to a predicate.” Donald Davidson9

Davidson’s views on semantics and metaphysics strike many

philosophers as paradoxical. Views in semantics and metaphysics

that are typically paired are not paired for Davidson. Davidson’s

account of reference is externalist, but he denies that there are

“joints” in nature. The objective world supplies the application-

conditions for predicates, but the world itself has no intrinsic

articulation and reality is not a given domain of beings and

kinds of beings. Whether A is the same object as B is relative to

9 Davidson, Donald, remark in an APA symposium somewhere on the West

Coast in the 1990’s.



a description. Such a combination of views seems implausible and

unsatisfactory on first encounter.

Davidson’s minimalist semantic views themselves strike many

philosophers as inadequate.10 His semantics and account of truth

and predication gives no account of what makes sentences true and

gives no analysis of the meanings of individual predicates.

Truth-conditions of sentences such as “Fred is a frog”, for

instance, are given by trivial-sounding formulations such as

“`Fred is a frog’ is true if and only if Fred is a frog.”

Davidson gives no account of what it takes for “is a frog” to be

true of an object other than that the object is a frog.

This book argues that Davidson’s views about semantics and

interpretation are by and large right. It further argues that

Davidson’s ideas make some attractive views about metaphysics and

philosophy of language possible. Some of these views go beyond

anything Davidson explicitly endorsed. Some of these views are

contrary to what Davidson himself held. In these last cases, this

book argues that Davidson is implicitly committed to some views

he did not hold.

10 In a conversation with Stephen Neale, who agrees with Davidson on many points, some years ago, his reaction to my suggestion that axiomsconnecting “all” and “same” were not strictly part of semantics, his response was that this would made semantics too uninformative. Semantics that is informative beyond what is needed to explain how finite learners can understand an infinity of sentences is too informative.



This chapter will lay out some central features of

Davidson’s account of truth, reference, and the nature of

semantics. The next two chapters will describe the metaphysics

that seems to me to fit best with that semantics, as well as with

other Davidsonian views. Many of the features of Davidson’s

metaphysical theory can be seen as consequences of taking

seriously his view that an adequate semantics for the meaning of

sentences is given by a Tarskian truth-definition.

I Truth-definition and semantics

a) What a semantics formally is

To treat semantics as a Tarskian truth-definition is to

conceive of a semantics for a language as having two components:

1) There is a list of primitive vocabulary clauses of the form

“F is satisfied by an object a if and only if Ga” where “F” is a

name of the primitive vocabulary item in the language being

interpreted and “G” is a predicate in the interpreting language.

A further Davidsonian restriction, inherited from Quine, is that,

if the interpreting language is the same as the interpreted

language, then if “F” names a syntactically primitive predicate,

then “G” is a primitive predicate. There is no predicate

decomposition in the semantics.11 This denial of predicate

decomposition has consequences for actual semantic theories, 11 In truth-definitions between languages, the necessity for a decompositionalsemantics is the result of there not being a simple predicate in one language whose extension matches a given predicate in the interpreting language. Saussure’s example of “sheep”/”mutton” //”mouton” is a good illustration. English has no word whose extension matches “mouton.”



since an adequate semantics must account for all formally valid

inferences. So, a Davidsonian is committed to there being non-

decompositional semantic accounts of constructions which seem to

cry out for decompositional analysis. Chapter 9, treating “good”

as a special type of comparative adjective, shows that this

approach can produce an empirically adequate semantics for the

comparative adjective construction, a major case in which almost

every current linguistic analysis requires predicate

decomposition.

2) The semantics is recursive. A recursive semantics is a set of

instructions such that for every compounding device in the syntax

there is a clause in the semantics which determines the

satisfaction-conditions of the compound formed by that device

from the satisfaction-conditions of the components compounded.

The “syntax” for such a semantics can either be a rendering of

the sentences in logical form, or in the surface structure of the

language. This presupposes that there is a systematic way of

getting from surface sentences to logical forms and vice versa.

This book will not attempt that syntactic project. It will

suffice for our purposes to have assigned a logical form which

captures the structure required for the inferences which must be

formal.

When the meta-language in which the truth-definition/

semantics is done is the same as the language for which a

semantics is being given, the meanings of sentences will be given



by apparently trivial remarks such as, famously, “`Snow is white’

is true if and only if snow is white.”12 Such remarks are t-

sentences, theorems of the truth-theory for the individual’s

language being interpreted at a time.

As Davidson points out, in order for a truth-definition to

apply to sentences with indexicals and demonstratives, the theory

strictly has to apply to utterances or inscribings by individual

language-users at particular times. Thus the sentences which the

syntax generates and the semantics interprets will almost all be

possible but not actual. As we argue below, this should be no

more troubling than the fact that most of the consequences of

physical laws are non-actual events. Truth-definitions are

empirical theories, just as Newton’s laws are. Newton’s laws

yield an infinity of conditionals about what velocities an object

would have in given conditions. Likewise, a truth-definition

yields an infinity of conditionals about what the truth-

conditions of an utterance would be if a speaker uttered it.

It is important that according to a Tarskian truth-

definition, truth is not a relation at all, but a one-place

predicate of sentences. So, truth is not a relation to facts,

12 As Davidson observes (1967a) p.35, it is actually very difficult to find a theory that will generate that particular sentence while also dealing adequately with the predicate “snow,” since there are tricky phenomena about how such mass terms work. Briefly, classical logic is designed to accommodate count-nouns, whose reference can be supposed to be classes. What “snow” refers to is an unresolved problem.



states of affairs, or other alleged “truth-makers.”13 If Tarski’s

account says the whole truth about truth, it is not any kind of

correspondence theory at all. Truth is not a relation of anything

to anything else. Truth is a one-place predicate of expressions.

Tarski’s account is of course compatible with there being

entities such as facts, etc., whose existence is correlated with

the predicate “is true” being true of a sentence. It is even

compatible with there being truth-makers. But nothing in the

formalism itself requires anything but a one-place truth-

predicate.14

Since truth is not a relation, some conundra about truths

and reality that lead philosophers astray are not puzzling. For

example, consider the argument: “Since there were no humans when

tyrannosaurs were alive, there were no sentences then. But then

`There are tyrannosaurs’ would not have been true then, since the

sentence did not exist then. So, eternal objects such as

propositions must be the truth-bearers.” This argument has no

appearance of plausibility if “is true” is a one-place predicate

13 Given that a sentence is true, there will be relations between that sentence and facts and states of affairs. In Chapter 6, facts and states of affairs will be analyzed as entities constructed from true sentences. Truth itself, though, is still a one-place predicate. 14 Tarski’s actual account, (Tarski 1933) in effect agrees with Frege that all true sentences have the same referent, as do all false sentences. The reference of an open sentence is the set of sequences that satisfy it. By Tarski’s account, a true sentence is satisfied by all sequences; a false sentence by no sequences. Given that satisfaction is a reference relation, this more or less interprets theTrue as the universal set and the False as the null set.



of sentences. If “is true” is a predicate of this utterance, that

is of this inscription of this sentence as produced now by this

speaker, then it is true then if and only if there were

tyrannosaurs then. If “is true” is a predicate of sentences

understood as possible utterances on occasions, sentences about

the existence of tyrannosaurs are true and were true. While it is

correct that there were no sentences in existence then, it is not

correct that “There are tyrannosaurs” was not true then. “It is

55,000,000 BCE” is a sentence of ours which was true for a year

back then.

The case is analogous to puzzles about the fact that

Aristotle might not have been called “Aristotle.” There is no

possible world, to use the illuminating metaphor, in which

speakers can truly say “Aristotle is not named Àristotle.’”

Nevertheless, there are worlds in which “Aristotle is not named

Àristotle’,” our sentence said in “this world” but describing

that world, is true. “Aristotle is not named Àristotle’” is true

in those worlds because the person we designate as “Aristotle” is

otherwise designated in those worlds. In the same way, in the

distant past there were no speakers who could have said “There

are tyrannosaurs” but our sentence “There are tyrannosaurs” was

true. In both the temporal and the modal cases we are

“evaluating” our sentence in another framework, either another

possible world or another time. Our English sentence “There are

tyrannosaurs” was true a while ago but that does not mean that



our sentence existed in those bygone days. Happily, “There are

tyrannosaurs” is not true now, but was true then.15

For Davidson, meanings are given by such truth-theory

clauses as “`is a frog’ is true of an object just in case that

object is a frog.” Notions such as that the meaning of a term is

its place in a web of belief, or that the meaning of a term is

its role in inferences are not exactly wrong. Such remarks about

meaning give accounts of the evidence germane to arriving at a

truth-definition clause in the process of constructing a truth

definition from what a person says in what circumstances. But

evidence for meaning is not the same as meaning. Acceptance by the

speaker of the inferences that follow from Fred’s being a frog is

of course good evidence that his sentence means that Fred is a

frog. Likewise, if we have evidence that the speaker’s beliefs

expressed using the term we are contemplating interpreting as

“frog” by and large match our beliefs about frogs, that will

15 As we will see in a Chapter 7, if tenses are construed as modal-like, then puzzles about truth-values of future-tense sentences, fatalism, and the like turn out to be akin to modal puzzles. If truth is understood as Davidson understands it, there is no reason to suppose that sentences about the future require the existence now of truth-makers for those sentences to correspond to. From a Davidsonian perspective, philosophers of time should let physicists find out what time is. Nothing of importance is forthcoming a priori.

Likewise, as we will see in Chapter 6, thinking of properties as“things said” might seem to run afoul of the fact that no-one was saying anything when the tyrannosaurs, with their fearsome properties,stomped about. On the Tarskian conception which Davidson adopts, none of these exotic consequences follow.



likewise be very good evidence in favor of interpreting the term

as “frog,” that is, assigning to “frog” the predicate clause,

“`is a frog’ is true of an object A just in case A is a frog.”

A Davidsonian semantics for a natural language requires that

the sentences of the language be interpreted as having a

recursive structure so that for every syntactic compounding

device there is a semantic instruction as to how semantic values

of results of that compounding device depend on semantic values

of its syntactic components. Thus a Davidsonian semantics is a

theory of logical form. For every sentence of the natural

language, there is a representation of that sentence in logical

form, that is, as a syntactic structure of predicates, such that

the algorithm generating truth-conditions generates the right

truth-conditions. “The right truth conditions” requires that

truths be interpreted as truths and that truth-preserving

inferences which must be formal are logically valid. Thus finding

a semantics for a natural language is finding a way of assigning

logical forms to the sentences of that language such that the

true sentences in the language are the true sentences in the

formal language, and such that truth-preservation patterns which

cannot be matters of particular information are explained as

cases of logical consequence.

A Davidsonian semantics is a counterfactual-supporting

empirical theory about the sentences of a person’s language. If

we have a theory which says that “is a frog” is true of an object



if and only if it is a frog, then the theory predicts that if the

person were to say “Fred is a frog,” that sentence would be true

just in case Fred is a frog. We find out whether that “`Fred is a

frog’ is true if and only if Fred is a frog” is true by applying

Davidsonian principles of interpretation. If our theory has the

person being interpreted asserting sentences that are false, that

is evidence against the theory. We discuss below how

interpretation assigns truth-values to sentences and thus allows

a truth-theory to be tested.

In arguing that much of Davidson’s view follows from “taking

seriously” the idea that a semantics is a Tarskian truth-

definition, I mean supposing that such a semantics is indeed

adequate and correct and working out what other things would have

to be true if that were the case. For example, if a predicate-

clause such as “`is a frog’ is true of A if and only if A is a

frog” is an adequate semantics for the predicate “is a frog,”

then an adequate semantics does not analyze the meanings of

individual lexical items. Thus any account of a language-user’s

understanding of individual lexical items must be by some other

device than giving explications or definitions in other terms.

Davidson’s semantics is also autonomous, if a Tarskian

truth-definition is a complete and sufficient account of meaning.

On some applications of Tarski’s ideas16 meaning is given 16 Deflationists such as Horwich, (1998) for instance, take meaning to be a notion independent of truth. So, “is true” for Horwich can be a fairly trivial device for generalizing. If meaning itself is



independently and it is possible to think of the truth-predicate

as a convenience useful for stating generalities. But if the

meaning of an expression is its Tarskian truth-conditions, as

understood by Davidson, giving the meaning of another’s sentence

presupposes that one uses and understands one’s own sentences. A

truth-definitional theorem uses a clause after the biconditional,

giving the meaning of the mentioned sentence before the

biconditional.

The explanation of how a speaker learns a first language,

the language in which the speaker interprets, is not part of

semantics. Davidson uses the notion of triangulation among

speakers in a common world, described below, to provide the basic

framework of an account of what it is to acquire a first

language. Triangulation, the process by which a speaker, in the

presence of another speaker in a common world, learns to

communicate about that common world by seeing what is said in

what circumstances, is not a part of semantics proper. Strictly

speaking, truth-conditions, truth, and meaning are analyzed

circularly, so that there is no analytical link to anything

outside the circle. Meaning is truth-conditions; truth-conditions

presuppose that the meaning of the clause after the biconditional

is understood. There is no formal link to anything outside the

circle. There is evidence for correct and incorrect assignments

understood in terms of truth, then this generalization cannot exhaust “the nature of truth.”



of truth-conditions, but truth and meaning do not reduce to some

other kind of phenomena.

If giving a t-sentence is all there is to giving the meaning

of any sentence, then truth is primitive. Any analysis of truth

would connect truth with something outside of the circle of

truth-meaning, and truth-conditions. So, given that complete and

adequate statements of truth-conditions are t-sentences, truth is

not to be reduced to anything else. There is nothing in the

account about correspondence to facts, the end of inquiry, or

coherence with a consistent set of beliefs.17

Just as “true” is primitive according to the theory, so is

“true of.” The predicate-clauses that constitute the bulk of the

truth-definition completely characterize the predication-

relation. That is, the semantics has nothing to say about what it

takes for a predicate to apply to an object beyond formulations

like “`Is a frog’ is true of Fred if and only if Fred is a frog.”

There is no appeal to properties, participation, set-membership,

or causal relations between entities and speakers.18 As Davidson

17 See Davidson, 1996.18 Davidson is aware that his account of predication is essentially what Aristotle arrived at, at least for the case of essential predication. He points out that the problem with to understanding predication as a relation of one entity to another goes back to Plato’s Parmenides. See Davidson (2005).

Aristotle addresses the issue of whether predication must involveForms or properties at least twice. In Metaphysics Z17 (1041b 12-31) he points out that, if what makes a thing one being is a further entity, then the same problem of uniting that further entity with the



argues, the theory has nothing more to say because there is

nothing more to say that is in the province of philosophy.

What is it for a semantics to be “complete and adequate?”

There is, after all, a lot more to be said about the truth-

conditions of “Fred is a frog.” If Fred is a frog, then Fred must

be an amphibian, Fred must live on a planet on which life have

evolved, Fred must be bilaterally symmetrical, and so forth. In

the same way, there is a lot more to be said about what it takes

for “is a frog” to be true of an object. Once we understand “is a

frog,” there is a lot to be learned about what it is true of.19

My Davidsonian view is that there is a natural scientific account

of what the world must be like for particular predicates such as

“is a frog” to be true of objects, and there is the semantic

components to be unified arises, so one would have a regress. In Metaphysics B (1001a29-33) Aristotle points out that if Being and One are natures, then Parmenides’ conclusion would follow. So Aristotle recognizes that two kinds of predications, essential predicates and so-called “categorical” predicates, cannot be treated as relations to other entities. Davidson in effect extends these observations to the conclusion that predication never needs to be construed as a relation of a subject to another entity, but can always be treated as primitive.

Another way of putting this is that Aristotle recognizes that predication cannot always be treated as attachment of a property to a subject, and Davidson generalizes to the conclusion that predication never needs to be so treated. “Same” is ambiguous for both Davidson and Aristotle. “Same kind” is just different from “same individual.”19 As we will see below, “understanding” is a somewhat tricky and vagueterm. Nevertheless, semantics presupposes that some terms are understood, and a Davidsonian semantics is nothing but a systematic explanation of the truth conditions of a language in terms the interpreter understands.



account.20 However, besides these two kinds of truth-conditions,

there is no third set of truth conditions, coming from some

deeper and more fundamental science, metaphysics, that adds

something important to answers to the question of what it is for

“Fred is a frog” to be true.

b) “Externalism”

In both the analysis of truth and that of reference, the

theory is externalist. The conditions for both truth and truth-of

are conditions of the world, not of the language-user. But no

notion other than truth itself or a use of a predicate occurs in

the analysis of either relation. As we will work out in detail

below, this means that the theory has neither a “content-fitting”

account nor a “causal” account of reference.

An externalist account of reference might seem to require a

given, that is, natural, articulation of reality into beings (for

referents of singular terms) and kinds of beings (for predicates)

independent of the beliefs of speakers. The externalist, it might

seem, must suppose that the world consists of a given domain of

beings and kinds of beings.

20 The semantic account we defend below is somewhat akin to Plato’s simple aitia in the Phaedo St. 99d-102a. There explanation of why a thing is, for instance, a frog is the safe and secure hypothesis that the thing instantiates froghood. Likewise the truth-conditions for “Joe is a frog” are given by the safe and secure “`Joe is a frog’ is true if and only if Joe is a frog.” In both cases, the answer is not as trivial as it may seem



Another view that seems to follow from externalism is that

names directly refer to their bearers, unmediated by any

concepts. Since names do not refer by having a meaning which fits

an individual, i.e. by an individual matching an individual

concept, it might appear that reference to individuals must be

some kind of direct demonstration. Thus it must be possible to

refer to individuals directly by demonstratives such as “this” or

names such as “Fred.” According to the usual externalism which is

coupled with the view that reality is intrinsically articulated

into beings, since there is only one entity in the region

demonstrated, pointing by itself can achieve reference to that

entity.

On both of these topics, Davidson maintains his externalism

while saying exactly the opposite, as illustrated in the aphorism

that opens this chapter. Davidson denies a given articulation of

reality into beings. On this topic, he is not even a normal

relativist, since he directly attacks the idea of a given domain

of beings sortable in various ways as incoherent.21 Without a

given domain of beings, there is little hope of making sense of

direct reference. If sameness is relative to a choice of

predicates, then there is no such thing as one entity remaining

the same as itself over time except relative to being brought

under a concept. But without such identity-conditions, there is

no notion of an object. One way of putting the difference between

21 Davidson (1970).



Davidson’s view and a direct reference view is that on a direct

reference view, “is an entity” is a count-noun that delivers a

number from a given region. For Davidson “is an entity,” while it

has the syntactic form of a count-noun, does not provide any

basis for individuation. How the combination of externalism and

the denial of a natural, given articulation of reality is

possible and in fact the most reasonable view is the main topic

of Chapters 2 and 3 below.22

It could seem that a Davidsonian semantics does not deliver

much, and that what it fails to deliver is crucial to the

adequacy of a semantics. In particular, an account of predication

that does not connect it either with the contents of concepts or

with designating an item as of a kind in a given articulation of

the world may seem hopeless. Davidson, as an externalist about

reference who does not suppose a given domain of beings to be

latched onto by the singular terms and predicates of a language,

provides no determination of or constraint on truth either from

inside the speaker or from the outside world. So nothing seems to

constrain truth or the application of a predicate. The questions

then naturally arise “What does make a sentence true?” “What

makes a predicate apply to an object?”23 The answer implicit in 22 The topics of how names refer is not one that Davidson ever addresses in a

thorough way. In Davidson (1993) he does suggest what his account would be.

This will be discussed below.23 This question of the apparent lack of constraint on a sentence being true

by anything external to the sentence was brought home to me by Martin



t-sentences and t-clauses is that nothing makes a sentence true

and nothing brings it about that a predicate applies to an object.

For Davidson, truth is a primitive with connections with, but not

a definition in terms of, notions of the intentional scheme—the

terms we apply in talking about beliefs, desires, actions, and

intentions. The same holds for “true of.”

I argue below that this laconic semantics is a good thing,

and gives the right answers in the right way. I also think that

some of the so-called basic questions are pseudo-questions.

Metaphysical answers to what predication is, for instance, do

little but re-phrase “is true of” in terms that leave one no

better off than the disquotational t-clause.24 For the most part,

I will construct accounts that fit the above conditions on a

Davidsonian semantical theory and also cohere with other views

Davidson held. So, this book will argue that these apparent

problems are actually positive features. I will argue that the

Prinzhorn, oral communication, Vienna, 2005.24 Davidson endorses regress arguments in (1967a). Here is one version, adapted from his class in 1967: Suppose that “Theaetetus flies” is to be understood as “Theaetetus participates in Flying.” Participation is a predicate. So, the real structure would be “Participation obtains between Theaetetus and Flying.” And so on. The argument of course goes back to Plato’s Parmenides. A similar regress structure causes difficulty for explaining the relation that binds an individual and a universal to form a fact, since the binding relation must itself be a universal. A further intuitive exposition of the problem puts it as articulating the difference between a set and something being true. Joe and froghood both exist whether ornot Joe is a frog. Joe, froghood, and participation all exist wheterh or not Joe participates in froghood. Davidson takes these ancient regress arguments to show something important. I agree.



Davidsonian position makes possible a “metaphysical” view that

accommodates Kripkean intuitions about naming and modalities

while retaining Davidson’s rejection of a given articulation of

the world into Beings.

The resulting ontological view, which I call “relative

essentialism”, was never espoused by the actual Davidson or the

actual Quine. However, it seems to me to flow naturally from

their perspective, and to be in fact something to which they are

implicitly committed. The view accommodates an overlapping

plurality of kinds of real objects governed by laws of various

degrees of strength. Cows, tables, gluons and 2007 Impalas all,

as we would hope, gain admission to Being. What is missing,

metaphysically, is the idea that there is one articulation into

beings that is “given” in nature and is foundational for the

others. In a way, this is to say that the theory does not have

metaphysics, in the pre-Kantian sense that is now so much the

fashion. The view has the trivial cost that countless un-law-

governed kinds of beings exist as well, so to speak, burbling

harmlessly alongside the ones we care about. I will argue in

Chapter 3that the theory arrived at is a very plausible and

satisfying one. Once a Davidsonian semantics is understood, it

allows the possibility of positions that previously seemed

desirable but impossible.

If metaphysics is the history of the uncovering of Being, as

Heidegger (2000) argues, the present essay is anti-metaphysical.



Being is about as mysterious a notion as notions get. To make

“being” the centerpiece of one’s thought is, I will argue, a

mistake. Beings are posits required in order that we can make

formal inferences which depend on a sub-sentential structure.

Beings are posited in order to think and speak.

Some discussions that call themselves “metaphysical” will

make sense on the present account. Discussions of how various

kinds of beings are related, whether laws of one kind of object

are explained by laws about another, and so forth are untouched

by the idea that there is no single given natural articulation of

the world. Likewise, they are untouched by the rejection of

properties and facts as fundamental explainers.

On the other hand, questions about whether the world is one

or many, whether only organisms are really beings, whether the

world is “gunk” or atomic, what the identity conditions of

properties are, and how properties attach to individuals will

seem exactly like metaphysics seemed to the logical positivists—

pseudo-questions. This is not to say that they are of no

interest. They are interesting in the way that providing a

general theory of how to cast horoscopes for star systems with an

arbitrarily large number of planets would be interesting. They

are pure conceptual artwork.

c) Theories of Reference

If the format “`is a frog’ is true of Fred if and only if

Fred is a frog” is an adequate homophonic semantics for



predication and the format “`Fred’ names Fred if and only if Fred

is a Fred”25 is an adequate semantics of the name-relation, then

every standard account of reference is likely to be mistaken. If

truth is primitive, then so is “true of,” and “names” (if these

relations are different and a fortiori if they are the same.) One

way to see what is different about Davidson’s view of reference

is to contrast it with two general ways that reference has been

theorized about, and show that Davidson’s view follows neither of

them. The two general formats are internalist “fit” accounts

which take reference to be some kind of relation between a

conception and the world, and externalist “causal” accounts which

take reference to be some kind of natural relation between

expressions and their origins. Davidson’s theory of predication

and reference is neither of these.

1) Internalist “fit” accounts take reference to be fixed by a

match of some sort between the features of the word or underlying

concept and the features of the object or kind referred to. A

classic example, of which modern versions are variants, is the

ideas of Locke or Condillac. More modern versions are the

description theories scouted by Kripke (1970) and Frege’s (1892)

senses, from which there is a function to a referent.

“Fit” accounts that invoke Platonic Forms, language-

transcendent intensions, or Fregean senses are not genuine

25 I argue below that Davidson in effect assumed that something like

Burge’s(1973) account of names was correct.



options for Davidson. Davidson, along withy Quine and many, many

others, seeks an account of language that accounts for language

as a natural phenomenon, the result of natural interactions of

human beings with their environments. Correct application

according to a naturalistically acceptable internalist theory

depends on how the term is used. In an internalist theory of the

acceptable naturalist kind, something like the intension of a

term is constructed by projecting from actual applications in the

language community.

Since an account of correct application must allow the

possibility of incorrect application, the language community must

be a plurality of speakers. If natural kinds are not available as

constraints on extensions of human language predicates, any

psychologically possible projection of a sequence of applications

of a predicate to new cases would be as acceptable an

interpretation as any other. If there are no natural kinds, there

is no reason to select one projection over another. An

internalist account of the extension of predicates cannot appeal

to “joints in nature.” So, if we allow a single idiolect to be a

language-community, that is, allow that there could be private

languages, and determine partial intensions of terms by seeing

when terms were applied, there would be no possibility of

incorrect application. Since any psychologically possible pattern

of linguistic response to an environment would be a candidate for

a partial intension, no response a speaker could make could fail



to fit an acceptable projection of a term to new cases. Only with

a plurality of speakers, so that correctness can be fixed by some

kind of majority rule, would it be possible for a speaker to

wrongly apply a term. Thus the natural idea for a naturalistic

internalist philosopher of language is a conception of language

on which language is essentially social in the sense that there

are rules and norms of a public language required for a speaker

or writer to say something that could be true, since only such

norms can supply the possibility of an utterance being false.

On this kind of internalist picture, of which the later

Wittgenstein (1953) is an example, only approximations of a

Fregean intension are generated. A real Fregean intension for a

term F would sort every possible object into either the F-pile or

the not-F pile. But from no finite collection of a society’s

applications to objects will such a sorting be determined. An

internalist will thus have three sets associated with every

predicate: The set to which the predicate applies; the set to

which its negation applies, and the set to which neither the

predicate nor the negation applies. There will always be multiple

ways of projecting, that is, bringing new elements into the first

set, from any such collection. So, Lewis’ “semantic indecision”26

is built into any theory of internal contents.

On such internalist accounts, any necessities about objects

would be analytic truths—if the term fits the object, the object

26 Lewis, David, (1986), page 213.



must have the features represented in the term. Since the

necessary truths and the truths about possibility rest on

intensions associated with terms, such necessities are de dicto.

Historically, many versions of internalist theories held that all

necessity was analyticity—to be explained by the containment of

one concept in another, so that necessary truths were at bottom

logical truths.27

In fact, though, such theories typically seem to be

committed to an underlying essentialism requiring some other

account of reference. Identifying the objects responded to and

the properties of those objects whose possession qualifies an

object as of one kind or another presupposes a domain of objects

and properties already given to the language-users. The job of

the language-users is to sort the given objects into groups

corresponding to their predicates according to their possession

of the given properties. But sorting objects into categories

presupposes already having identified those objects. If the

objects are given prior to the application of those concepts

whose intensions determine their correct application, then those

objects, as given, cannot without regress be picked out by

concepts whose intensions determine correct application.

27 Carnap’s (1956) explains such containment as a matter of rules of language, as an explication of the classical metaphor of “containment.” “Containment” is, I think, literal in Plato’s thinking in the Sophist. See my (2010).



That is, to have an array of objects to sort, some

conditions on when an object exists and when there is one object

rather than more than one is presupposed. But that generates

objective de re necessities about, for instance, physical objects

as such. Thus, apparently, internalist accounts of reference

presuppose another externalist account of “object as such.” To

conceive of a language as a division of the world into categories

is to suppose that the world comes divided into beings already. A

genuinely pure internalism would find no basis for resemblance

that does not beg the question or presuppose a given. The only

pure internalism I know of in philosophy is Nelson Goodman’s.

Goodman had thought through this issue decades ago and

characterizes assignments of predicates to objects as “decrees.”

That is, as the basis for applying the same predicate again is

not some given articulation, but rather a decree which

establishes an articulation.28

b) Externalist “causal” accounts take reference to be a relation

established, not by any match between intrinsic features of a

language-user’s words or concepts and objects, but rather by some

kind of causal interaction between the speaker and a given array

of beings and properties in the world. The relation can be

evolutionary history, initial baptism, or “made this dent in the

wax tablet.”29 Reference is fixed by a relation that does not

28 Goodman, Nelson, (1966) p.134.



depend on intrinsic features of the word or any internal state of

the language-user.

Such realist “causal” theories seem30 to rely on a given

segmentation of the world into beings and kinds of beings. From

that given segmentation, both objects to be named and the

predicates applied to those objects are fixed by the given

segmentation into beings and kinds of being. Nature has joints

and is carved at its joints31 by a predicate system that can

express truths. Predicates true of anything name kinds, and names

that refer designate entities that are elements of those kinds.

The natural segmentation is reflected in objective necessities,

which for a naturalist are laws of nature. These necessities

yield essences, objective conditions on what it takes to be an

entity of a given kind. As we will see in Chapter 8 on the

sorites, the “necessities” have to be seriously weakened in order

to accommodate the existence of my Chevy Impala.

29 Plato suggests such a theory in the Theaetetus St. 190e-195b, but does

not take it up, rather treating the dent as referring by resemblance.30 In fact, as we will see below, for many cases a Davidsonian can agree that a causal relation is required in cases of ostension of a physical object. For Davidson, causal relations do not require a privileged segmentation. In general though, reference would not involve causal relations and would certainly not be defined in terms of causal relations. Kripke (1970) does not think he has a “causal theory of reference,” only that causal relations are relevant in lots of cases. Perhaps not for the cube root of 2.31 Plato Phaedrus 265d-266a.



Externalist realist theories thus presuppose a given domain

of beings sorted by nature into kinds, manifest in natural laws,

whose natures individuate those objects. In human languages,

reference of names and correct application of predicates is

secured by linkage to those entities and their kinds. The

necessities thus underwritten are de re absolute necessities.

As the next chapter will examine in detail, the great

systematic advantage of a monistic causal realistic semantics is

that it reflects the very intuitive Aristotelian distinctions

that outline the conditions for a world of objects and

properties. The essence-accident distinction, positing a

difference between changes that are alterations and those that

are extinctions, in effect is the format for a world of objects

that undergo change while surviving some of them. The basic

contrast between the object itself and a feature of that object,

part of any common-sense realist metaphysics, is encoded in a

realist causal semantics. A descriptive metaphysics of the world

we live in seems to require the picture that this semantics

reflects: given objects divided into given kinds.32

32 Even revisionist metaphysical accounts which abandon the objects of ordinary life use the same Aristotelian format, with different primarysubstances. An ontology of space-time points, for example, takes each point to be necessarily related to every other point, so that its identity and persistence conditions are fixed by their very nature. For ontologies with other-than permanent beings, the essence-accident distinction is more complex.



Modal judgments such as that Fred can get a better job while

remaining the same person, that this rabbit could have eaten

different carrots but comes to an end when stewed, and that this

stuff, since it’s water, has to freeze at -100 Celsius are among

the intuitively obvious truths that make up the normal person’s

picture of the world. Accepting de re modal sentences is part of

the normal view of the world, and is constitutive of the view

that there is a world of objects and agents to be treasured and

dealt with.

Given that Davidson holds that interpretation presupposes

that much of what any interpretable agent believes is true, many

of these de re modal beliefs must be true, according to his own

theory. It has to be the case that people can get haircuts and

survive, and that pedicures are not fatal. Chapter 3 will show

how a Davidsonian can countenance such de re modal truths without

supposing that the world is in itself naturally divided along

joints; that there is a privileged segmentation into objects and

properties.

II Davidson’s externalism

Davidson’s semantics is externalist. The application-

conditions for a predicate or name are given by formulas that

only mention objects in the world. Predicate-clauses and clauses

giving the references of singular terms mention only objects in

the world. There is nothing in the clauses of the semantics about

features of the concepts or words that must fit anything, except



to the extent that the requirement that “is a frog” be true only

of entities that are frogs is “fit.” Truth is not a relation of

anything to anything else. Truth is a one-place predicate of

expressions.

Davidson denies that there is a privileged “real”

segmentation of the world into beings and properties. Nothing

about reference to real kinds occurs in the truth-definitional

formulae. In addition, the aphorism that heads this chapter,

ruling out the absolute sameness that a privileged segmentation

would warrant, is well-supported by Davidson’s writings33 and by

his Quinean anti-essentialist heritage.

Most of what most people say, on Davidsonian interpretive

principles,34 is true. So, a consequence of taking reference to

be completely defined by truth-definitional formulas is a

proliferation of kinds, as it were. Given that most of what most

people say is true, the “ontology” to which Davidson is

apparently committed includes everything about which there are

true affirmative sentences in any culture, at any time, in any

language. Since there is no privileged segmentation into objects

and kinds of objects, there is much latitude in which sets of

salient objects are extensions of predicates. Terms are correctly

applied to a wide variety of overlapping objects. Thus,

extensions are very much shaped by usage—what people say when can

33 See for instance Davidson (1974.)34 Davidson (1973).



yield sets of true-of-objects predicates that vary between

cultures and within cultures over time.

It is a misunderstanding of Davidson’s aphorism that “The

beginning of wisdom is the realization that sameness is always

relative to a predicate,” to say that this account of reference

and predication takes linguistic items to be foundational. Given

Davidson’s understanding of the conditions of application of a

predicate, the aphorism is about objects, primarily, not about

words. Since what the predicates are true of is given by the

schema “`is a frog’ is true of A if and only if A is a frog,”

Davidson could have said “The beginning of wisdom is the

realization that sameness is always relative to a kind of

object.” This would be true to his view, but would lend itself

to a joints-in-nature monistic realist misinterpretation. Talk of

“predicates” emphasizes that the objects are not a given

privileged domain of beings. It is true, as will be discussed

below, that human practices, including linguistic practices, in

some sense select and in a way create the objects and properties

we talk about and think in terms of.35

Among the various objects which are elements of extensions

of predicates there are of course many distinctions that could be

made. Some posits are useful for science, some for art, some for

industry, some for personal relations. But there is no clear 35 It will be noted that the humans and their practices are not elements of a privileged segmentation either. I’ll get to this. Bear in mind that not being privileged does not imply not being.



sense Davidson can give to the question which predicates pick out

the real Beings.36 It could turn out that some families of

predicates can be reduced to others in a law-like way, and so

would be dispensable. The Quinean scientistic option of

privileging the predicates of physics and junking the less

orderly predicates is unmotivated for Davidson, and incompatible

with Davidson’s expansive conception of charity. On one

significant issue, though, the present essay will diverge from

Davidson. As we will see in Chapter 5, on events, the aphorism is

strictly incompatible with anomalous monism.

There are some prima facie puzzling consequences of the view

I’ve ascribed to Davidson. First, it would appear that different

physical objects can completely fill the same space at a time,

perhaps raising a question about their solidity. On the natural

construal of Davidson’s view, since there are overlapping kinds

and so overlapping objects, there is no particular object that

36 It is puzzling to me that many of my colleagues seem to have such clear conceptions of Being that they are willing to deny that there really are shadows, and the like. I have no such conceptions beyond the Quinean one in “On What There Is,” (1950) which I take to be a kind of jocular response to some of his contemporaries. I recommend reading a few chapters of Paul Weiss’ (1938) or (1958) in order to join in Quine’s bemusement with the idea that Being is a Big Question.Even Quine, though, took Being seriously, to the extent that he thought there was some advantage to having fewer beings in one’s ontology. I take this to be an aspect of his scientism, his insistencethat scientific terms were the only worthwhile kinds of terms in whichto think of the world, and that scientific theories were the only worthwhile picture of knowledge.



will be demonstrated by a demonstrative. Thus direct reference to

objects seems to be impossible. Apparent de re modalities, such

as “This has to be H2O” would only be relative to a predicate or

non-exclusive kind of stuff, not a real de re modality. In

Chapter 3 these ontological consequences will be either

vindicated or shown not to be genuine consequences.

Second, as noted above, and verified in conversations with

many colleagues, this account can seem to be too thin to have

content. The account of truth-conditions as given by t-sentences

appears to be analytic, not empirical. T-sentences seem to tell

us very little about what it takes for a sentence to be true.

Third, the account also seems not to accommodate strong

intuitions about modality. There seem to be true de re modal

sentences. But without a given domain of objects with their

kinds, it might seem that such sentences would not have a ground

in reality. How can there be de re modal sentences unless the res

are given prior to the language use? More generally, how can

there be an externalism, where independent objects fix the

reference of terms, unless there are beings that are given

independently of our terms, concepts, and thinking?

The next section will argue that t-sentences are the only

general and accurate way of giving truth-conditions. Chapter 2

will argue that there are in fact no absolute de re modalities,

even though there are de re modalities. The modalities depend on

the objects, not the contents of the concepts behind the words.



Chapter 3 will argue that objects, including ourselves, are in

some sense our “posits”, but are none the less real, and not mere

projections of our thinking. In some very loose sense, Hegel’s

(1807) identification of objects and consciousness is vindicated

on Davidson’s account. Chapter 3 will conclude with a discussion

of how apparent direct reference takes place without a given

domain of objects. Chapter 4 extends the results of Chapter 3 to

the special case of events and to the relationship of the mental

and the physical. Anomalous monism will turn out to be

incompatible with this neo-Davidsonian conception of Being.

“Innocuous dualism” might be the term for the resulting view.

Chapter 5 discusses what a (neo)-Davidsonian theory of modality

and conditionals might be and how the particular neo-Davidsonism

writing this book proposes to understand the relative de re

modalities of Chapter 2.

Chapters 6, 7 and 8 will discuss the consequences of this

neo-Davidsonism for metaphysics and meta-ethics. Questions about

what properties and propositions are, how medium-sized objects

fit with micro-particles given the sorites, future contingents,

and questions whether objects endure or perdure will look very

different from a Davidsonian perspective. Chapters 9 and 10

present what I think is the most important consequence of

thinking in a Davidsonian way. A Davidsonian semantics for

“good,” and a Davidsonian theory of “good” and “ought” based on

Davidsonian ideas about interpretation yields a conception of



ethics that differs a great deal from most of what in current in

both ethics and meta-ethics.

III Davidsonian t-sentences

a) ostension and understanding

For Davidson, ostensive language-learning of a first

language takes place by triangulation. “Triangulation” is

Davidson’s term for the general idea that language-learning takes

place between two subjects who recognize each other, each other’s

responses, and a common world. Triangulation has three elements:

First, there is a speaker-teacher in an environment uttering a

word with some object salient in that environment. Second there

is a hearer-learner observing the environment, the speaker, and

the word, and interpreting that utterance. Third, there is the

world common to both the speaker and the hearer. The hearer

interprets the speaker’s words as true of some salient object in

their common world.37 Davidson, in characterizing this

triangulation and its consequences, is careful to say that the

sameness of object and sameness of response by the language

learner and teacher are samenesses for them, not in some sense of

sameness that transcends any predicate scheme. When the

samenesses for the teacher are also samenesses for the learner,

the beginnings of acquiring a term has occurred. In ostensive

learning, the learning is learning to identify elements of the

extension of the term.

37 See Davidson (1992).



Triangulation is a nascent case of radical interpretation.

The language-learner has to take the situation to be one of

interpretation of the behavior of the language-teacher. Thus the

language-learner must have the pre-conceptual beginnings of a

theory of other minds—the learner has to recognize the teacher as

an agent, doing things for reasons. Thus saying the same word in

the presence of the same object must mean that the word has

something to do with the object, and that the teacher is speaking

in the presence of the learner for some purpose, namely

communication.

“Triangulation” is not intended to explain anything about

the psychological conditions for original language acquisition.

Rather, it sets out the general conditions that logically must

obtain for a speaker to acquire language. The core idea is that a

language-user has acquired a concept of truth, which requires a

distinction between truth and falsity. If there is nothing

distinguishing correct application of a term from an application

one is disposed to make, there is no purchase for truth, because

there is no purchase for falsity. A learner can see that the

teacher utters “dog” when a Saint Bernard is present and when a

Chihuahua is present. The learner has acquiesced in this part of

the “conceptual scheme” of the teacher.

Davidson’s “triangulation” story is a variant of

Wittgenstein’s (1953)38 “Private language Argument” that language

38 Wittgenstein (1953) section 243 and following.



must be social. To be a language user requires that the speaker

be in a language-community. However, the language community need

not use the same words for the same kinds of things. The

“language community” for Davidson need only recognize a common

world in the sense that the samenesses for the learner are the

same as the samenesses for a teacher. There is no requirement

that the words be the same. While the learner may choose to utter

“chien” in the presence of dogs, the learner, assuming that the

teacher is communicating truths, will apply “chien” to Chihuahuas

and Saint Bernards as well, even though their perceptual

presentations are very different. That is still acquiescing in

the “conceptual scheme,” if not in the language of the other.

In spite of their differences about whether a common

language rather than a common world is required for language to

take place, Davidson and Wittgenstein are articulating the same

distinction, that between two things looking the same as opposed

to two things being the same, that is between appearance and

truth. Without another agent, there is only looking the same.

There being an objective world is exactly this difference. When a

person recognizes this difference, the person has the notion of

truth.

In fact, for this kind of ascent to the conceptual to take

place, a language-learner must be pre-programmed to learn

language and to understand other agents as agents. Just as

important, the speakers have to be prepared by evolution to posit



objects in roughly the same way. Medium-sized animals must be

salient objects for both of them, for instance.

Once the learner has a repertoire of ostensively learned

predicates, the learner can learn various connections among

predicates, can store images, etc., in a store of information

associated with the lexical item. The learner in effect acquires

a personal repertoire of material when coming to understand a

word. In certain special cases, such as scientific terms, the

learner acquires a theory that constitutes a real Web of Belief,

where the connections among terms become theorems. That is, for

words such as “gluon” or “isosceles,” the material lends itself

to the kind of propositional organization and systematic

connection with other words that makes “theory” an appropriate

characterization of the source of a person’s competence with a

term. While Quine was primarily interested in scientific

language, the theory idea is not very persuasive in the case of

other lexical items, such as “John.” An insight such as that of

Augustin Rayo, (forthcoming) briefly discussed below, is a more

appropriate picture.

b) What t-sentences are

A t-sentence pairs a mentioned or cited sentence with a

clause that is used to give the truth-conditions of the mentioned

sentence. (And likewise for predicate clauses in a truth-

definition.) Truth-conditions are given in a language that is

understood already, so that it can be used to say what the world



has to be like for the sentence to be true. So we are dealing

with the language of a particular person at a time, an idiolect,

not a language as a social construct. A Davidsonian is aware that

even if I don’t know what “amygdule” means, I know that in the

mouth of someone who can use “amygdule,” ‘ “Fred is an amygdule”

is true if and only if Fred is an amygdule’ is true just in

virtue of my knowing that “amygdule” is a count-noun. Without

some connection to usage, “amygdule” is not part of my language

(in Davidson’s idiolect sense) so I can’t use it. I can only cite

it or mention it. Homophonic t-sentences, unless their component

words are being used, are themselves not being used at all. Since

they contain a citation in a position where a use is required, an

apparent homophonic t-sentence with a cited word is not a t-

sentence at all.

For “use” to be literally about a speaker’s speaking or a

writer’s writing, most of the t-sentences the semantics generates

have to be counterfactuals about what the truth-conditions of

possible alternative speech-acts by this person speaking this

idiolect now would be. As noted above, a truth-theory for a

person at a time generates counterfactuals about what the truth-

conditions of an infinite number of possible utterances and

inscriptions would be. This constitutes an empirical theory about

which utterances would be truths under what conditions. Put

another way, for a word to be used, it has to be part of the

speaker’s language. So, a term can only be used if it is



understood. So, what can a Davidsonian say about the conditions

for a word being understood? It should be clear on reflection

that both “use” and “understand” admit of degrees and have vague

borderlines. In consequence, the “use-mention” distinction is

clearly a vague borderline notion, to be discussed in Chapter 8.

If I don’t understand French and I say “Jacques said that

les neiges d’antan sont disparues,” I haven’t really done

indirect discourse in Davidson’s analysis, but rather a kind of

citation or quotation. . Likewise, when a spy listens in on

discussions of nuclear physicists and reports to his superiors

what the physicists said, but doesn’t know anything about gluons,

neutrons or quarks. When he reports to his spy-master “Hashem

said that the mass of a neutron is mostly gluon energy, not

intrinsic quark mass,” he is not using all the words in the

“that” clause. All he knows about gluons, neutrons, and quarks,

is that they are things physicists talk about, and that is not

enough for those terms to be part of his language. He is like an

illiterate transporting a text.

A speaker needs to know “enough” about the extension of a

predicate in order to use the predicate. The spy knows something

about gluons and the predicate “is a gluon” when he knows that

gluons were what physicists were talking about yesterday. That

seems inadequate to constitute “understanding” and so inadequate

to qualify his utterance of the word as “use.” And so, if this

speaker utters, “`Fred is a gluon’ is true if and only if Fred is



a gluon,” he has not given the meaning, according to my

Davidsonism. But what is sufficient for understanding? Do I have

to be able to cite the equations that determine the range of the

strong force?

As Putnam39 pointed out some decades ago, many of the terms

of a person’s language, like “is an elm” in Putnam’s idiolect,

have only a weak connection to the rest of what we know, and are

not cases where we can directly detect items of the extension. We

refer to things by referring to the experts who know about the

things. What experts say is evidence in the same way that cloud-

chamber tracks are evidence. In both situations, we have a kind

of indirect access to an extension. In Putnam’s examples, though,

Putnam knows that elms are trees, and knows something about

trees. There is a difference between my knowing for crossword

purposes that a gnu is a kind of antelope without being able to

identify one, and my having learned to fill in (correctly)

“haggadic” when the crossword puzzle clue is “non-halakhic

midrash” when I can’t say anything about either “haggadic,”

“halakhic,” or “midrash.” I knew what antelopes are, roughly,

even though I didn’t know that they are members of the genus

Connochaietes, of the family Bovidae, of the order Artiodactyla.

Since a Davidsonian is an externalist, the resulting grab-

bag that constitutes understanding a term is a collection40 of

39 Putnam, Hilary (1975). 40 See Rayo (forthcoming).



bits of knowledge, not a collection of bits of belief. My

misconceptions about quarks are not additions to my understanding

of the term and ability to use it. If we take a person’s concept

to be the person’s opinions and dispositions to identify

associated with a kind of thing, our understanding of a term

generally differs from our concept expressed by the term.

With a few exceptions, for instance mathematical concepts

and those precise terms where there are sharp laws connecting

predicates of different families, our understanding of the terms

we use falls short of necessary and sufficient conditions. I know

that cows are domestic mammals with horns and are used for milk,

and can identify cows reasonably well. But sorites arguments and

peculiar non-standard cows may baffle me. In particular,

sequences of cow-like animals constructed by ingenious subtle

changes in a decomposition sorites may leave me unable to

determine whether an entity is a cow or not. The possibility of

a genetic sorites, where DNA is shifted unit by unit, starting

from an animal that is clearly a cow and ending with an animal

that is clearly a capybara, seems to show that the very notion

that there are lawlike necessary and sufficient conditions for

being a cow is dubious. In such an imagined continuum, when does

an organism cease to be a cow? Regardless of such difficulties in

determining the truth-conditions of “Denise is a cow” in

helpfully other terms, I can be very confident that “is a cow”



applies to an object if and only if it is a cow and that “Denise

is a cow” is true if and only if Denise is a cow.

T-sentences are thus the only reliable device for giving the

meaning of a term in my language or in another’s. I can

understand a term without knowing everything about its extension,

or even very much about its extension. Only in rare circumstances

can I give a definition of a term by giving necessary and

sufficient conditions using other predicates I understand. So,

the homophonic t-sentence is my only complete and accurate

account of the meaning. Given that I understand the term so that

my utterance or inscription is a use, my t-sentence legitimately

and completely accurately gives the meaning in terms available to

me. So, my utterance may be analytic in my idiolect, and perhaps

analytic in the idiolect of anyone who is using the words in the

sentence as I do, but this analyticity does not make the t-

sentence trivial. So, t-sentences and t-clauses are as good and

informative as a semantics can be.

III Understanding without Lexical Semantics

This section tries to assuage some worries about whether a

Davidsonian semantics, as described above, can account for what

appear to be analytic truths and objective necessary truths. At

several points there will have to be promissory notes about

theories to be argued for in later chapters. In particular, I

will be developing a different conception of being in later

chapters. I will be arguing that there being no privileged



divisions into Beings and Kinds does not mean that there are no

objects. Given that there are objects, any explanations in terms

of natures of those objects that explain anything on the

“realist” absolute model will also explain the same thing on a

model on which indefinitely many “objects” are real. Briefly,

only some of those potential “objects” will have anything

approximating laws about them. As I will argue, “joints in

nature” do nothing that is not accomplished by posited objects

about which there are truths and laws.

This section will first explain my understanding of

Davidson’s notion of interpretation, and then briefly discuss

analytic truths and the notion of the lexical. Necessary truths

will be a topic of the next few chapters.

a) Interpretation

On a Davidsonian account of language understanding,

“interpretation” does much of the work assigned to semantics in

some other kinds of theories. Interpretation of speech or text

acts is a special case of action interpretation. In action

interpretation, we apply the core components of our theories of

agenthood as probabilistic constraints on what this presumed

agent is doing. These core components constitute what can be

called “the intentional scheme.” “The intentional scheme” is the

rough explanatory theory in terms of belief, desire, intention,

that is central to humans' conceptions of themselves as agents.

We can try to articulate this rough theory as basic decision-



theoretic principles supplemented by guidelines about maximizing

agreement, along with other guidelines about agents. Since the

agent with whom we are best acquainted is ourselves, such

principles generally amount to “maximization of agreement”

principles. Another term for the “intentional scheme” would be

“rationalization” the interpretation of events as actions by an

agent. There is reason to think that the intentional scheme is

built into humans (and perhaps other organisms) innately, in much

the way that the “enduring physical object” scheme is.

For much the same reason that there is no algorithm for

theory-change in the face of new data,41 so there can be no

algorithm for interpretation of people’s speech acts.

Interpretation requires coming up with explanations of behavior

that involve new concepts and new phenomena. No algorithm from

1950 would yield the understanding of a person applying thumbs to

a small box that someone is texting on a cell-phone. In addition,

interpretation requires understanding new metaphors and other

figures, mis-statements, malapropisms,42 and other novel

linguistic phenomena. Since interpretation is always of

particular situations in which novel collections of information

are relevant, interpretation is guided by rules of thumb and

general guidelines, but is not a theory in any technical sense.

41 Carl Hempel (1965) page 6, argues that since scientific theories contain new, non-observational terms, there can be no algorithm for getting from observations to explanatory theories.42 Davidson (1986)



Anything and everything can be relevant to the question of what

to make of a speech act.

Examples of the application of interpretation are ubiquitous

in Davidson’s thinking. Davidsonian interpretation is a kind of

generalized Griceanism.43 Davidson understands metaphors, for

instance, as utterances that are obviously not true or relevant

to the situation, and so must be intended for something else.

“The outer door will be locked whenever all employees have left

the building” could be understood either as a generalization or

as an instruction, depending on whether the audience is fellow

burglars or employees. On every interpretive occasion, the

immediate environment, various pieces of information about the

speaker, intonation-patterns, and so forth affect how the

utterance should be understood. If we don’t think of “context” as

a set of parameters from which there is a function to a referent,

we can say that interpretation is always in and relative to a

context.

Interpretation is thus required, not only in arriving at a

truth-theory for a particular speaker, but also in understanding

utterances of those for whom you already have arrived at the

clauses of a truth-theory. Each speech event or writing event is

43 A lot of detailed work has been done on exactly what goes into interpretation in conversation. Some of this work extends Stalnaker’s notion of “context set,” other work extends Grice’s (1989) work and Lewis’(1979) notion of ‘keeping score.” See recent work by Craige Roberts, for instance her (1996).



in a different context, and each such context is relevant to what

the speaker intends to communicate.

The objects being interpreted for Davidson are essentially

individual actions, whether speech-acts or writing acts. A

speaker presents a sentence with a truth-value for some purpose.

The rhetorical force of the utterance is part of the expression

of that purpose. If a person wants to communicate dissatisfaction

with my performance, one way to so do is to say “Good job.”

Interpretation as communicating dissatisfaction supposes that the

person knows that I know that it was not a good job, and knows

that I know that he knows that it was not a good job, and is

being sarcastic. Likewise, when a person calls his guinea pig an

eggplant, we know he knows that no guinea pigs are vegetables of

any kind, and so we conclude that his utterance is meant to point

out something about how dumpy the little animal is.

b) Analytic truths and the lexical

One characteristic of a Davidsonian semantics, inherited

from Quine, is that there are no analytic truths. Every

consequence of “is a frog” being true of a thing beyond

disquotation is information about the extension, not part of the

meaning. Analytic truths traditionally would be those truths

which follow from the content of a concept, where a concept is

construed as a non-linguistic or trans-linguistic bearer of

meaning expressible in a language. From the perspective of a

recursive semantics, the denial of analytic truths is the denial



that the semantics need contain any predicate clauses other than

the disquotational ones described above. For instance, there are

no clauses in the semantics of the form “`is a dog’ is true of an

entity a only if a is an animal.” That dogs are animals is not a

semantic fact, but a fact about dogs.

It certainly does seem that being an animal is part of what

it means to be a dog. Davidson and Quine can explain the feeling

that certain connections are “built into the meaning” of certain

predicates. While the explanation of “analyticity” as meaning-

containment is unavailable to a theorist who gives truth-

conditions with t-clauses, a Davidsonian need not deny that there

are sentences that can only be interpreted reasonably if they are

interpreted as committing the speaker to other sentences. If a

person’s use of “taller than” does not in general abide by

transitivity, the person cannot mean “taller than” by “taller

than,” but may perhaps mean “looks taller than.” Obvious truths

are of course preserved in interpretation, but they are preserved

because they are obvious and true, not because they are part of

meanings.

The problem cases for a semantics without predicate

decomposition are ones where inferences which must be formal seem

to require that the semantics analyze some predicates in terms of

others. The construction which seems to require such predicate

decomposition is the comparative adjectives, words such as “tall”

and “long.” Briefly, in order to get both “John is taller than



Fred” and “John is two feet tall” to use the same “tall,” it

appears that some hidden reference to dimensions or some such

device is required in the semantics itself.

In fact, a very plausible theory is possible which abides by

Davidsonian strictures. One of the comparative adjectives is

“good,” as Davidson mentions in “Truth and Meaning.” A proper

account of the Nature of the Good will, following Davidson’s

ideas, starts with an account of logical form. An account of

“good” as a component of an account of comparative adjectives, is

Chapter 9. Modal constructions, another case where something

beyond predicates, set abstraction, and truth-functions might

seem to be required, are given a Davidsonian treatment in Chapter

5.

Davidson said relatively little about how lexical items

should be treated. He says a great deal about interpretation—how

one understands what someone else says or writes. Such

interpretation has to put in one’s own terms what the other

person said or wrote. This raises the question, how those “own

terms” are to be conceived. Beyond proposing a semantics in which

analysis or decomposition of lexical items is absent, Davidson

does not propose an account of what understanding a lexical item

is. An account such as that of Agustin Rayo (forthcoming) fills

this gap in a satisfying way that comports with much that

Davidson does say. So, in the more general case, a Davidsonian

should follow Rayo and say that the speaker acquires what Rayo



calls a “grab-bag”-- a collection of images, examples, truths,

and so forth relevant to what the word applies to when the

speaker comes to understand a word. Such an account does not

constitute a “semantics” for lexical items so much as a program

for a natural history of how people understand words.

A word like “John,” for instance, may have within it a

collection of faces, a dim memory of “en arche ein ho logos,”

something about the Magna Carta, and a host of other material

that will vary from person to person. Typically, the language

user has enough of this repertoire to fix a referent with the

help of context, and hopes to communicate to the hearer what this

referent is. When does the learner understand a given lexical

item? It seems clear that there is no good answer beyond “when

the learner is able to use the word.” While the semantics is in a

way quite clear, the application to cases is vague.

c) The individuation of lexical items

Word are different from basic44 predicates. Words are

socially transmitted entities whereas basic predicates are those

same words occurring in the left-hand side of clauses in a truth-

definition for an individual’s idiolect. The same public word can

correspond to two or more predicates. Also, the same word may be

one predicate in one idiolect and more than one in another.

44 “Basic” here means “requiring its own clause in a truth-definition. While there are clear connections between “authoritative,” “authorial,” “authorities” and “author” these words should correspond to distinct predicates in most people’s idiolect.



Predicates are words functioning as elements of idiolects. A

truth-definition has a clause for each distinct predicate.

Idiolects of the same “language” may differ in their predicates.

David Kaplan (1990) is something Davidsonians should take to

heart. Even though a common language is not essential to a

language, a common language is very much part of actual

communication, not to mention literature, writing, and the very

idea of a text.45 The key idea is that the identity of a word is

determined, not by identity of inscription, but by the relational

idea of being a copy of the right kind. “Is the same word as” is

vague, since the line of transmission can be confused. 46

Whether two occurrences of the same word express the same

predicate is sometimes indeterminable both over time and at a

given time. The English word “swim” derives from two distinct

roots, one meaning “dizzy” and the other meaning “propel oneself

through the water.” For most current speakers, “swim” probably is

a single predicate, but speakers use the predicate metaphorically

in expressions like “You’ll be out of here so fast it will make

your head swim,” even though there are in fact two distinct

words, historically, which have coalesced for most speakers. The

words and predicates of a language are in large part the result

45 Tolhurst and Wheeler,(1979), proposed a similar idea about the identity of texts. 46 Spelling is no guide to word-identity. “Entrance” is clearly twowords, as pronunciation indicates. “Number” is both the comparative of“numb” and a count-noun.



of successive common metaphors becoming what a word means. I have

published several essays on this topic and its implications for

semantical theories. 47

The lexical entities that are relevant to the thesis that

there is no lexical decomposition are the predicates. For

example, “more,” is clearly a single English word. In the

appendix on comparative adjectives, the English word “more” is

treated as a compound of “much” and the comparative particle.

This is different from treating “dog” as a compound of something

and “animal,” since there is some mark in the word itself that it

is a comparative.

d) Davidson on names

One thesis of this book, developed in subsequent chapters,

is that there is no direct reference except to times and places

by way of demonstratives. Without a privileged partition of the

world into individuals via count-terms, “this” cannot pick out a

particular physical object and a name without semantic content

cannot pick out an individual. Davidson never offers a theory of

names. The closest he comes to doing so is in Davidson (1993)

where he discusses the desiderata for a theory and criticizes

causal theories. Davidson agrees with Derrida (1977) that so-

called marginal cases of naming cannot be ignored in constructing

a semantical theory. So, names in fiction and names that are used

for the first time in a fiction to designate real people cannot

47 Wheeler (2007, 2011).



be set aside while a theory is constructed of “normal”

occurrences of names.

The difference between proper names used in fiction and

proper names used to talk about real people cannot be semantic,

so the difference must be in the speech act, that is, what is

done with the sentence with its truth-conditions. The difference

must also be that something about the writing- or speech-action

is a pretense. An account that fits these desiderata is the

following: What is pretended is the demonstration that

accompanies the proper name. An intended fictional use of a name

involves pretending to demonstrate an object. It is not exactly a

failed demonstration, any more than a hyperbole is a failed

assertion. So, something like Burge’s (1973) account of proper

names seems to be what Davidson envisages. Proper names are

general terms of a special sort whose use is accompanied by a

demonstrative.

One version of this idea would be that “Theaetetus flies”

has the form, “(That) Theaetetus (man) flies.” In symbols this

would be, “Ex(That x /\ Theaetetus x /\ Man x /\ flies x).”

“That” is a predicate that characterizes what a speaker is

demonstrating or has in mind. Its predicate clause is “`That’ is

true of an object a if and only if that is a.” The “man” is a

count-noun disambiguating what sort of object is being

demonstrated. The necessity for such a count-noun will be

discussed in Chapter 3.



The idea in application to names is that the lack of an

article indicates a demonstrative. With an article, we would have

sentences such as “One of the Thomas Smiths in the phone book

must be the one we need to contact.” The absence of an article is

interpreted as a demonstrative.

e) Indeterminability of interpretation

Quine48 and Davidson49 argue for one basic kind of

indeterminacy of interpretation. In trying to get maximal fit

between one’s beliefs and desires and those of the person being

interpreted, an interpreter sometimes has the choice between

assigning a mistake in beliefs and desires (the person wants

something more than she should or believes something more

confidently than she should) or a difference in meaning. When a

person in one’s communicative community says something bizarre,

but is still taken to be an agent, one always has this choice.

Sometimes the choice is easy. When my colleagues at the fishing

club discuss prostrate cancer, I can interpret them on the

hypothesis that they are using words differently. Other times the

question is not so clear.

A kind of case I have discussed is metaphor, as mentioned

above. It can be indeterminable whether a word is being used with

metaphorical intent or whether the word has more than one literal

meaning. In my idiolect, “The Giants crushed the Bears” may or

48 Quine (1960) Chapter 2.49 Davidson (1979).



may not be metaphorical. I cannot tell from the inside whether I

have two predicates “crush” or just one which is employed

metaphorically.

Interpretation is an optimization problem with several

factors. When all the evidence does not determine a single

optimum, there is indeterminability. Since the notion of meaning

for a Davidsonian is essentially defined in terms of correct

interpretation, what a truth-definition says, there is no outside

spiritual state to check against an interpretation. Quine and

Davidson treat such cases as indeterminate. As I argue in Chapter

8, what is established is indeterminability, which is compatible

with determinacy. Briefly, it is open to Davidson and Quine to

posit bivalence for practical reasons. If one treats “\/p (p or

not-p)” as a harmless practical principle, then one is committed

to apparent indeterminacies having determinate, but in principle

indeterminable truth-values.

IV What does a Davidsonian semantical theory look like?

Logical structure, for Quine, reveals the inference-patterns

that do not depend on what predicates occur in the respective

sentences, but only on the pattern of occurrence of distinct

predicates. That is, Quine gives an account of form, as in

“formal logic.” “Logical form” for a Davidsonian abides by the

same idea. Semantics is to look for recursive structure and leave

general truths that depend on what a particular predicate is to

the theories speakers have using the predicate. So, “triangles



have three sides” will be a truth of geometry, not of logic. This

principle is a consequence of denying the analytic-synthetic

distinction. If there are no non-logically true sentences true in

virtue of meaning, and semantics is a theory of meaning, then

there is no decomposition of predicates in a proper semantics.

A contemporary Davidsonian application of this principle

will yield an even more austere semantics than Davidson

envisaged. For Quine and Davidson, logical structure included

everything through first-order quantification theory, at least.

So, quantifiers and quantificational patterns were part of

structure. That can no longer be the case.

Contemporary linguists and philosophers of language who take

linguistics seriously pretty universally agree with what I

(ignorantly) thought was an original and rather bizarre discovery

about quantifiers in “Attributives and their Modifiers”(1972) and

“Quantification in English”(1978). My idea was motivated by

considerations about comparative adjectives and intensifiers,

whereas people like Mostowski (1957) were interested in a logic

as such and Montague (1973) was interested in a general account

of logical form that preserved English syntax. Montague’s work

became the norm in linguistic semantics, culminating in Barwise

and Cooper’s “Generalized Quantifiers and Natural Language”

(1981).

The basic idea is that what first-order logic treats as

operators on open sentences are really two-place predicates of



collections (sets or aggregates.) In brief, the standard

“Aristotelian” or “Fregean” quantifiers are special cases of a

large family of atomic and compound predicates of collections

(collections of stuff or individuals), which includes “many,”

“much,” “most,” “five,” “few,” “so many that I have a hard time

remembering the exact number,” and the like. To take a simple

case, “Some frogs are green” says that the green frogs, i.e. the

intersection of the frogs and the green, is some, i.e. not empty.

“Five frogs are green” says that the intersection is a five-set.

“Some furniture is stylish” says that a portion of the furniture

overlaps with the stylish.

The linguistic account that results treats a quantified

sentence as a subject-predicate sentence, with a determiner-

phrase and a verb phrase. What appears to be the subject, say,

“all frogs” or “some frogs” is indeed the subject. “Frogs” is the

restrictor, by which is meant that “frogs” determines the

collection that the verb-phrase characterizes. “All”, “many” or

“few,” for instance, is the quantifier, a sub-species of

determiner, and the whole subject is a determiner-phrase. The

linguistic account of the quantifiers has a good explanation of

why the predicate calculus account of quantification, which binds

a single variable, works. As Heim and Kratzer observe,50 “Vx(Fx

->Gx)” and “Ex(Fx /\Gx)” as canonical representations of “All

frogs are green” and “Some frogs are green” basically paraphrase

50 Heim and Kratzer (1998) p. 191.



quantified sentences using the determiner-phrases “everything” and

“something” (“every thing” and “some thing”). The determiners

would just be the bare “E” and “V”. Quantification-theory works

as well as it does because for these special quantified

expressions there is an equivalent that talks about the

relationship between “the entities” and a subset of them. So,

Vx(Fx->Gx) says that the set of things is the set such that Fx

only if Gx. Ex(Fx /\Gx) says that the set of things x such that

Fx and Gx is a subset of the set of things, i.e. that the

intersection is not null. There is, of course, an algorithm for

finding paraphrases of “Three frogs are green” and “At least

11,603 frogs are green.” But it is grossly counterintuitive to

suppose that the logical form of “At least 11,603 frogs are

green” actually has the hundreds of thousands of clauses that the

paraphrase algorithm requires.

There is a strict proof51 that no such first-order

paraphrase works for a number of quantifiers, such as “finitely

many” and the like. These could be added as primitive predicates

of collections. However, for some quantifiers, for instance

“most” interpreted as “more than half,” there is a strict proof

(Rescher 1964) that there cannot be a representation of that

relation between sets as a relation between “the entities” and a

single set. That is, “Most frogs are green” can’t be expressed by

51 Barwise and Cooper (1981).



anything you can say about most things. So addition of primitives

is of no help.

So we can treat quantifiers as predicates with special rules

governing their application. Being predicates doesn’t make them

mere predicates. Rather, the quantifiers are like identity. Many

quantifiers are important and significant predicates. This result

re-draws the line between structure and theory. If we chose, we

could add some more special rules. Besides the bachelor rule, and

the rule for “all,” we could have the Dog rule, that licenses

inferences from “A is a dog” to “A is a domestic animal of a

species sometimes kept as pets but eaten in some parts of the

world.” More plausibly, all classical mathematics could be

treated as part of logic. In effect, this would be to reinstate

logicism, as I understand it, since, if the quantifiers are part

of logic, the rest of set-theory as well as the mereology that

seems to be presupposed in quantifiers with mass-terms, may as

well be.

Notice that the quantifiers are not all susceptible to

precise theories. “Many,” for example, seems to depend on the two

classes, and to have a vague borderline. “Many frogs weigh over

three pounds” is false if there are only 12,000 such frogs;

whereas “Many democracies have collapsed into dictatorships” is

true even though fewer than 1000 have. “Many” is a topic of

Chapter 9.



In practice, in the rest of the book except for Chapter 9, I

will retain the traditional account of the quantifiers as

operators on open sentences, for readability, familiarity, and

convenience.

Meanwhile, consider what this account of the quantifiers

does to logical form and to logic. If we remove the quantifiers

from the domain of structure and put it into predicates with

special principles (parts of theories), then we have a conception

of logic as essentially just truth-functional logic, set

abstraction, and predicates. Everything else that some would like

to call “logic” consists of mathematical and other theories about

particular predicates. The truths about the membership predicate,

the part-whole predicate, and the identity predicate will be

especially important, but not sui generis. Note that of course

truth-functional logic is itself a kind of mathematics. The

truths about the connection between “not all frogs are green” and

“some frogs are not green” will be mathematical or set-

theoretical truths, not strictly speaking logical truths in the

sense of truths that rely solely on structure.52

52Note that with this result, Quine’s (1970, p.68) complaint that second order logic is just “set theory in sheep’s clothing” turns out to be a bit unfair. First-order logic is set-theory in disguise. Second-order logic’s “defects” (incompleteness) are due solely to Cantor’s results about the uncountability of the set of subsets of an infinite countable set.



What about logical form? Logical form will be a description

of the sentence’s predicate-structure (predicates and their

arguments) and truth-functional composition. I think it is

important to separate semantics from theory. Perhaps another level

of separation would be to separate some fundamental quasi-

mathematical predicates’ theories from less important predicates’

theories as well. Perhaps it would be useful to have a special

status for our theory of “more” that accounts for the

transitivity of so many apparent two-place relations. Likewise

for the identity predicate.

But if semantics is supposed to be an account of the

competency of a language-user which will account for the

possibility of that language-user understanding an infinity of

sentences on the basis of a finite amount of learned information,

then an austere Davidsonian semantics delivers that. Everything

else is information.

The next two chapters, Chapter 2 and Chapter 3, will argue

the following theses:

1) Some kind of essentialism is required if common sense truths

are to be preserved without paraphrase. Aristotle is the model

for such theories.

2) Davidson is implicitly committed to the de re necessary truths

that essentialism is supposed to explain, and in fact committed

to some sort of essentialism



3) While essentialism is correct, it does not follow that there

is a single articulation of the world into entities and

properties. Rather, a kind of “relative essentialism” is

defensible.

4) Relative essentialism solves the difficulties with medium-

sized objects that arise when one wants to endorse de re

necessary truths both about medium-sized objects and about the

entities which have a stronger claim to grounding necessities,

the micro-particles of physics.

Chapter 2: Against absolute essentialism

Davidson’s account of truth and predication seems to remove

one reason to think metaphysical theories of the world are

necessary in addition to scientific theories. As we will see in

later chapters, his way of thinking about truth and predication

also dissolves some other puzzles that have occupied metaphysics.

However, the most important intuitive truths that seem to require

metaphysical theories are intuitions that there are necessities

about medium-sized objects. Necessities about medium-sized

objects seem to require a single privileged division of the world

into objects and kinds. Such a privileged division raises puzzles

about how those necessities fit with the necessities about the

objects of physics.

Given that interpretation starts with the medium-sized

objects of common sense, and given that Davidson is committed to



the truth of most common-sense beliefs, a Davidsonian needs to

accommodate these modal intuitions about medium-sized objects. A

Davidsonian must also offer some account of how puzzles about the

relationship of the objects of our lived world to the objects of

physics. This and the next chapter will set out a version of

essentialism, the idea that there are de re necessities about

objects that are due to their nature, that will fit with

Davidson’s aphorism, “The beginning of wisdom is the realization

that sameness is always relative to a predicate,” that headed

Chapter 1. Chapter 8 will discuss the sorites, which I understand

to illustrate the main difficulty with the view that the world is

divided into objects and properties.

This chapter is an exposition and appreciation of

essentialism and a critique of absolute essentialism. Absolute

essentialism I understand to be the view that reality consists of

a single given array of beings whose natures determine the

necessary truths. The critique focuses on the “single given” part

of this idea. A subsidiary target is the idea that predication

must involve truth-makers, states of affairs that consist of a

property attaching to an individual. The next chapter will

develop an essentialism that does not suppose that there is a

single given array of beings. What a Davidsonian can say about

the nature of properties will be one topic of Chapter 6.

Following Davidson and Aristotle, I take the default view in

philosophy to be that there are enduring objects which undergo



change while remaining the same thing. This is to treat as a very

last resort views which supply paraphrases supplying “what is

really the case when we say that you have existed for sixty-eight

years.” The only reason to accept such paraphrases rather than

what is transparently the obvious truth with its obvious truth-

conditions would be irresolvable problems in the default view.

This book argues that there are no such irresolvable problems, if

one follows Davidsonian thinking. Essentialism in some form is

part of common sense.

I Historical prelude: origins of philosophical essentialism

It is important to see what difficulties essentialism

resolves. Much of the essentialist picture has to be preserved,

according to any account that accommodates common sense. A

Davidsonian account must accommodate most of common sense, given

that Davidsonian interpretation starts from the idea that people

believe truths. One simple way to see what essentialism is good

for is to see how it arose. As we will see, Aristotle’s thinking

about what is is in many ways the fundamental essentialist

thinking. Here is a broad-brush narrative of the origin of the

core of pre-critical53 metaphysics, by which I mean the kind of

metaphysics that takes the Beings to be a given domain.

53 I use “pre-critical” not as a temporal predicate but as a label for metaphysical thinking which entirely abandons Kant’s idea that human thinking and knowing capacities shape what beingsthere are for us and supposes that “things in themselves” are given to be theorized about.



Heraclitus was understood by Plato54 and Aristotle55 to have

argued that nothing survives change, and that since change is

ubiquitous and continuous in the physical world, that nothing

endures. The basic argument is quite simple: If something

changes, it is not the same as it was. But everything is the same

as itself. So it is impossible for an entity to change and remain

the same thing. Nothing lasts.

Plato both argues that the conclusion is wrong and

challenges the argument. His challenge to the argument

distinguishes between intrinsic change and relational change.56

While intrinsic changes do result in distinct objects, changes in

relational features need not. Plato, like David Lewis some years

later, takes the “problem of temporary intrinsics”57 seriously.

His metaphysical system, at some points in his thinking, can be

construed as an account which allows some continuing existents

while acknowledging that change destroys beings. The world of

Forms is an array of Beings that are unchanging intrinsically,

while undergoing constant change in relations, given that

participation is a relation. Plato’s accommodation of Heraclitus’

54 Plato, Sophist St. 179c-181b.55 Aristotle, Metaphysics Γ5 1010a7-16; Γ7 1012a25-29.56 Plato, Theaetetus St. 154c-155c.57 David Lewis, (1986) page 202. Lewis identifies “…a problem of

temporary intrinsics, which is the traditional problem of

change.”



argument of course leaves out physical objects as beings that can

survive change.

Parmenides begins with another truism: Being is. Something

exists. Non-being isn’t. Nothingness does not exist. If there is

no non-being, then alleged descriptions of what is the case that

make reference to what is not cannot be correct descriptions of

reality. That is, no negations can correspond to facts about what

is the case. Among the alleged parts of reality that are ruled

out by Parmenides’ truisms are distinctness and change.

Distinctness requires that one entity not be another. Change

requires that what was not now is. Since Being is, and there is

no distinctness, what exists is a single, changeless entity. This

part of Parmenides’ view can be thought of as adopting more or

less the same thoughts about identity and difference as

Heraclitus, but with the denial of Heraclitus’ conclusion as a

premise. Since Beings exist, anything incompatible with being

cannot be part of reality.

Plato deals with Parmenides’ arguments in two ways. First,

he argues in the Sophist that Difference is actually a positive

feature, and not to be analyzed in terms of non-existence.58

Second, he shows in the Parmenides59 that, strictly speaking, since

Oneness and Being are distinct things, there could not be a being

58 Plato, Sophist59 Plato, Parmenides, Hypothesis I



which was actually one, if having many features entailed

plurality. It would have two features.

Plato seems to accept Parmenides’ idea that being a being is

a nature. For Plato, all of what come to be characterized as

“categorical” properties, for instance, being a particular, being

a property, being one, and being a being, are natures. As we will

see below, some modern metaphysicians try to avoid some

difficulties by claiming that “categorical” properties are a

special case that need not be subject to the arguments for their

theories of predication.

Plato’s responses to Heraclitus and Parmenides leave him

with a picture of the world that is radically different from what

most people would say the world is like. The real entities in the

world are natures, which he calls Forms. The entities which most

people would take to be lasting objects surviving change are in

fact not such at all. Everything is the physical world is really

flux, constantly changing beings that exist only for an instant,

more or less as David Lewis’ physical beings do.

Aristotle has different responses to Heraclitus and

Parmenides which shape his entire metaphysics. He finds Plato’s

rescue of continuing beings inadequate, since clear-cut cases of

continuing beings, namely organisms, do not turn out to really

continue. Aristotle is a kind of early Davidsonian in his view



that much of what most people believe is correct.60 His response

to Heraclitus can be thought of as adding a further distinction

to Plato’s distinction between intrinsic and relational

properties. Among the intrinsic properties of a thing are those

it cannot lose while remaining the same thing, its essential

properties, and those it can lose while continuing to be the same

thing, its accidents. The essence of an entity, what it would be

to be an entity of that kind, amounts to something like the

collection of its essential properties.61

The relationship between an entity and its essence, though,

cannot be something attaching to a subject. An entity would not

be that subject at all without that nature, so there is not a

single subject at all to which a nature could attach.62 So the

model of essential predication cannot be that of a subject to

which a nature or any other kind of feature attaches. Since

60 Aristotle’s support for such a view is rather different from Davidson’s. He takes it as clear that people are by nature organisms that other things being equal tend to know things aboutthe world. Thus he begins most of his treatises with a survey of what his predecessors have said, since much of what his predecessors have said is likely to be true, and a starting pointfor his investigation. The first line of Aristotle’s Metaphysics A1 is “All men are by nature such as to know.”( 980a21)61 Not every collection of properties qualifies as an essence, but onlythose which characterize real substances. Aristotle discusses whether being a pale man could be the essence of something, so that as a man became tan, something would cease to exist and something else would come into existence in Metaphysics Z4. He takes it as obvious that “is apale man” is just not a substance.62 Metaphysics Z 17 1041b12-31.



entity-constituting features make the entity be an entity at all,

the entity-constituting natures themselves cannot be construed as

attachments. Thus predication is primitive at least in some

cases. Predication cannot always be analyzed as one thing joining

with another. If predication were always an entity attaching

(e.g. a property-instance) to another entity, we would have a

regress.

If predication were construed as always a relation between

entities, then the “is the same as” relation that obtains between

two frogs, identity in kind, would be analyzable in terms of

sameness as numerical identity. Two frogs are the same if there

is a thing they both share, a universal. For Aristotle, at least

in the case of essential predication, “this is the same as this”

is ambiguous between this being the same entity as this and this

being the same in kind as this.

Aristotle thus has the conceptual equipment to hold a

completely Davidsonian account of predication. Predication is

primitive, and irreducible to entities in relation. This is a

result that had been bruited since the Third Man arguments in the

Parmenides.63 In fact, though, because Aristotle is reluctant to

deny existence to anything that can be said to be, he continues

63 Plato, Parmenides, St. 131c-132b is the basic regress. Aristotle refers

repeatedly to “the third man,”



to recognize property-particulars and universals as beings “in a

sense.”64

Besides entity-constituting predicates, two other predicates

cannot be construed as natures of any kind. These are “is a

being” and “is one thing.” Aristotle construes Parmenides as

holding that Being and Oneness are the essence of everything that

exists. Aristotle’s response to Parmenides is to deny that being

a being or being one are natures at all. One feature of

Parmenides’ view is the truism that anything either is or is not.

Anything, i.e. any feature, that would distinguish one being from

another would, if real, be a Being. Aristotle ascribes a regress

argument to Parmenides with the conclusion that there is only one

thing. Supposing that there are beings, anything distinguishing

one being from another would itself be a being. But if that being

distinguishing the two beings were something different from just

being, that difference would itself be a being, by the truism “is

or is not.” Since any being is one, the same argument works for

being one. In effect, if being a being is a kind of object, then

there cannot be distinct beings—everything would be one and this

one is being, as Aristotle says.65 64 In fact, Aristotle seems to treat composites of form and the particular

matter an entity happens to have, as well as composites of substances and

accidents, as entities that are distinct from the substance itself. Fred with

the matter he has has different modal properties from Fred himself. Fred

himself could have different matter; Fred plus his matter could not. 65 Metaphysics B4 1001a27-33.



Rather than accepting Parmenides’ conclusion, Aristotle

takes this as a reductio ad absurdum of the idea that being a

being is a substance-determiner, since there is obviously more

than one thing. Since being one is coextensive with being a

being, Aristotle takes the same result to apply to oneness.

Neither of these things said about objects can be treated as

entity-constituters. Still less could they be treated as

accidents of the things they are true of. This has consequences

for the conception of entities, for Aristotle. There is really no

such kind as an entity—rather every entity-constituter is itself

the “what is would be to be” of any entity it is true of.

Thus Aristotle seems to agree with Davidson’s aphorism that

sameness is relative to a kind, inasmuch as he holds that “is a

being” is not a term that divides reference, so that “same being

as” requires specification of “same what.” “Is a being” is a

pseudo-count-noun. On the other hand, Aristotle agrees with

contemporary monistic essentialist realists that “is a substance”

does divide reference, since each kind of substance has an

essence which individuates elements of the kind. So, “same

individual” is not relative to a kind.

II Contemporary theories of lasting objects

This section argues that the basic framework of Aristotle’s

essence-accident distinction is imbedded in contemporary

metaphysical theories that agree with common sense in holding

that some physical objects undergo changes while remaining the



same object.66 The imbedding may be subtle, but every theory

which countenances objects which survive change makes a

distinction, implicit or explicit, between essential, entity-

constituting features and features an entity can lose.

There are two ways to conceive of individuals that yields a

unique given domain of beings:

1) If a theorist takes “is a being” or “is one thing” to be

features that themselves individuate, the individuals are the

objects that are beings or ones. Plato seems to have held such a

theory.67 An entity being a single thing just is its sharing in

the One. David Lewis likewise takes “being a being” to be

unproblematic. Such theorists take “is a thing” or “is a being”

to be an individuating feature. Such realists appear to have no

need for being-constituting natures other than Oneness and Being.

They also do not directly accommodate lasting entities. Plato and

Lewis both take the “problem of temporary intrinsics” very

seriously. The book will say little about such theories, which

only indirectly accommodate the common sense conception of the

world as consisting of lasting objects which undergo changes.

66 I will not be discussing issues about endurance versus

perdurance. I argue in Chapter 7 below that the “metaphysics of

time” more or less disappears from a Davidsonian perspective.67 See Wheeler (2010).



It is worth seeing that both Plato and Lewis do in fact take

some objects to have constituting natures. I think it is clear

that Plato takes natures to be their own constituting natures.

Aristotle certainly implies this in Metaphysics Z6.68 Plato not

only needs a plurality of beings but needs a plurality of

natures. I argue in Wheeler (2010) that he constructs all such

natures out of Oneness and Difference.

The case of David Lewis is perhaps less obvious. Lewis

treats properties as sets of individuals. So, he identifies the

property of being a frog with the set of frogs in all possible

worlds. What is the connection between the set and its members?

One would think first, that the connection is necessary, that

this set by its very nature has these items as members. But the

set is a separate being from its members. Since its members are

scattered through the possible worlds, the necessity of the link

between a set and its members cannot be explicated by what is the

case in every possible world, since there is no possible world

that contains all the items in the set. The same is true of the

bond between unit sets and what they are unit sets of.69 So there

68 Metaphysics Z6,1030a27-32. “As for things stated by themselves, is it

necessary for them to be the same as their essences? For example this would be

the case if some substances exist, like the Ideas posited by some thinkers,

prior to which [no] other substances or natures exist. For if Good Itself were

distinct from the essence of Good…” 69 Thus, at least on this score, Lewis’s (1991) account of classes leaves

classes an exception to his usual denial of necessary connection among



is a least one kind of entity with a constituting essence, namely

sets. Sets and their members are a case of a necessary connection

between distinct existences.70 Lewis is an Aristotelian about a

central necessary connection.

2) If a theorist takes entities to have constituting natures, a

single thing’s being one thing derives from its being constituted

by some features that constitute it. That is, the relation

between, for instance a man and manhood cannot be attachment of a

feature to an entity, but rather manhood is the what it is to be

of the thing. The thing has a nature and is therefore an entity.

The features that can thus constitute entities are the natural

kinds.

This conception generates modal truths about such

constituted objects. Given that an entity is a man, then it has

what it takes to be a man—perhaps rationality, animality, and a

propensity to grasp natures of entities. Later philosophers have

different opinions about what the natural kinds are. Descartes

(1644), for example, takes there to be just two kinds of entities

—extended and mental substance, each with the features that

constitute them as what they are. Armstrong, as we will see

distinct existences. 70 The membership relation and the relation between a unit set and its member

would both have to be natural relations. That is, since there are an infinity

of grembership relations between a natural property and other objects, the

special relation membership has to be itself distinguished.



below, takes the kinds of beings constituted by natures to be

particulars and universals.

Essential natures have an important role in tying beings to

something determinable. Without some necessary truths about

beings other than that they are beings, lasting beings would be

untrackable. Beings about which there were no necessary truths,

for instance, would be beings that might or might not have the

conditional property “if at location x at time t, will be at some

location within 108 kilometers of x at t+1 second.” No

restrictions on the kinds of changes such beings could undergo

would be anything other than contingent. So, the hypothesis that

beings are sailing through our landscape picking up the

properties that seem to apply to us and our chairs would be as

well-confirmed as the hypothesis that we and our chairs are

themselves beings.71 On the other hand, individual beings might

be moving randomly every second from location to location, but

happily picking up most of the features of the previous

inhabitant of the locations they find themselves in. I have

called this argument the “Wind of Being” argument, which I take

71 A metaphysician who rejects lasting beings on other grounds, will of course be unmoved by this consideration. A Heraclitean such as David Lewis, who views physical objects as sums of instantaneous objects, will be unmoved by the impossibility of tracking continuants.



to be an expansion of Aristotle’s remark about matter in

Metaphysics Z.72

Any materialism which recognizes lasting physical

particulars is committed to essential features—features without

which the beings would not be physical. That is, if “obedience”

to physical laws is what it is to be physical, then physical

objects as such have essences. In the example mentioned above,

the law that nothing exceeds the speed of light amounts to a

necessary conditional feature about where an object can be from

moment to moment. So, any “realism” that takes the domain of

being to consist of a given array of physical objects is

committed to essential properties. The issue among such realists

is what the objects are that have them.

Both Plato and Aristotle have anthropocentric views about

nature. For Plato, there is a natural affinity between our minds

and the structure of nature. For Aristotle, human nature itself

is naturally disposed to form concepts that correspond to the

objective kinds in nature.73 The medium sized-objects that

populate the lived world are fundamental objects in Aristotelian

science. For Aristotle, then the harmony of the mind’s kinds, the

72 Metaphysics Z3 1029a26ff: “From what has been said, it follows that matter is a substance. But this is impossible, for to be separable anda this is thought to belong most of all to a substance.73 See Aristotle, Posterior Analytics 99b31 and following. Humans are constituted so that after enough experience they will have the essencein mind. He supplies only a metaphor as to how this happens.



common sense world of organisms and medium-sized objects, with

the kinds in nature can be taken for granted.

When a philosopher, acknowledging the authority of

developing natural science, no longer takes the fundamental

objects of science to be things like squirrels and daffodils, but

rather quarks, tau neutrinos, and the Higgs boson, the connection

between the articulation of nature and the articulations medium-

sized humans apply to their world is much more tenuous. As will

be discussed in Chapter 8, sorites arguments show a serious

difficulty for a monistic essentialist who wishes to have both a

single privileged given articulation in nature and a metaphysical

support for the reality of the medium-sized objects, including

ourselves, that we treasure.74

Part of Aristotle’s idea that animals and plants are

fundamental objects is that such entities are fundamental objects

of science. For a substance, there is exactly one nature

constituting it, and so there is exactly one set of existence and

extinction conditions for that whole substance. The “whole

substance” is all the matter making that substance up. So, any

part of the substance is not itself a single thing. No substance

contains another as proper part.

If a part survived the destruction of the whole of which is

was a part, then there would be no ontological point to having

74 The basic issue was raised in Aristotle’s time by Eubulides. See Wheeler (1983).



the whole be a substance at all. The reality would be what the

atomists maintained, that only the atoms are really beings.

Aristotle agrees with common sense against Lewis’ mereological

liberalism that chess sets and piles of sand are not further

beings, but are nothing over and above their component parts.

Thus a kind of Ockham’s Razor counts in favor of saying that “No

substance is composed of substances which exist in actuality…”75

Aristotle’s argument that parts of substances are not

themselves substances presupposes that there is exactly one count

of how many things are in a place. “…The double line is composed

of two halves, but these latter exist potentially; for the

actuality of each would make them separate.”76 Another

consideration is that otherwise, an entity would have two

essences, which would mean two conditions for continuing to

exist. If an oxygen atom is an entity, and it is part of a larger

entity, for instance a person, then what happens when the person

is killed? If being a person is an entity-constituter, Aristotle

holds that it constitutes the being of whatever it physically

encompasses. So the “what it would be to be” of the whole mass of

a person is being a person. So, every part of the person ceases

to be what it was when the person perishes. So, the oxygen atom

cannot have been an actual being itself, since if it met the

conditions for being an oxygen atom, and those were sufficient

75 Metaphysics Z 13, 1039a4.76 Metaphysic Z13 1039 a7.



for being an entity, it would survive the death of the person,

which it cannot, if “is a person” determines the “what it is to

be” of the entire organism.

A premise of this argument is that a given filler of

physical space is exactly one thing.77 Parts cannot be full-

fledged entities if the composites of which they are parts are

also full-fledged entities. Aristotle is thus driven to the

thesis that parts of objects are potential entities rather than

genuine entities. This premise that the filler of a given region

can be only one thing runs into serious difficulties once it

begins to turn out that the medium-sized objects of the lived

world, such as people and plants, are not the beings about which

there are the best physical laws. Natures of things, their

essences, are supposed to be what explains why natural phenomena

occur as they do. As science began to find out that explanatory

laws of basic physics did not take medium-sized object as

fundamental, but rather corpuscles, later atoms, and now quarks

as the fundamental beings whose natures explain what happens, the

idea that medium sized objects were fundamental became

implausible. Given that conclusion, and a more or less

Aristotelian picture of substances and properties, it began to

seem implausible that the medium-sized objects really had

essential natures at all. If, following Aristotle, we think of

77 There is actually some tension in Aristotle here, since he thinks that the person with her actual accidents and the person as an entity which could have had a different accident are distinct yet occupy the same space.



natures as the basis for physical laws, and so natural necessity,

there is not really room for medium-sized objects to have genuine

natures unless medium-sized objects are reducible to complexes of

fundamental objects, so that their natures would be thus

derivative.

It is clear, though, that there is no systematic reduction

of medium-sized objects to the current candidates for fundamental

objects. A device for showing this is the Sorites argument, which

can be run either as an argument about properties, showing that a

predicate of medium-sized objects has no definition in terms of

micro-particle properties, or as an argument about decomposition,

showing that no aggregate of micro-particles systematically

amounts to a medium-sized object. The argument of an early paper

in twentieth-century writing on the sorites, my (1975) was

precisely this. If predication is construed as properties

attaching to subjects, and properties either attach or do not

attach, then there are no medium-sized objects. Put another way,

if to be an entity is to have a nature that determines the

conditions for continuing to exist, and the laws of nature

reflect natures, and the firmest exceptionless laws are about

micro-particles, then unless medium-sized objects are

systematically and lawfully related to composites of micro-

particles, so that natures of medium-sized objects are determined

by the natures of the constituent micro-particles, medium-sized

objects would not have natures and would not be genuine objects.



It is obvious that medium-sized objects, including people,

exist. So, something is amiss. Typical metaphysical responses to

this dilemma use the notion of “levels of being,” postulating

different kinds of real properties for different levels and

allowing the objects of science as well as the objects of micro-

physics to have constituting natures.78

I argue that such theories follow the Aristotelian

conception of natures in making a very intuitive mistake. In

various ways, such accounts abandon Aristotle’s argument that

parts of substances cannot be substances as well as the

substances of which they are parts, but nonetheless take to heart

the core intuition behind Aristotle’s argument. The presumption

is that there cannot be more than one material being in the same

space at the same time. As I will argue in the next chapter, “is

a being” is not a genuine count-noun and multiple physical beings

can occupy the same space. Beings are posits, not givens.

The view proposed in the next chapter is that there is no

unmediated reference to stuffs or individuals. Neither “this” nor

“this stuff” refer. “De re” modal truths are always relative to a78 I will not discuss revisionary accounts such as David Armstrong’s, beyond noting that his theory gives no account of intuitive modal truths about what could or could not have been the case about, for instance, particular humans. Given my limitation to non-revisionary theories, the competitors to a Davidsonian theory are those that try to retain such intuitive modal truths while acknowledging that the laws of physics are more fundamental than any natural laws that support the intuitivetruths.



kind of object or stuff, rather than fixed by a division in the

nature of things. There are objects, kinds of objects, and many

distinctions to be made among those kinds in terms of interest,

importance, law-likeness, and so forth.

My account can happily appropriate the accounts of “levels”

theorists. The considerations that persuade some monistic

realists that, for instance, artifacts are real, even though

their natures are “looser” and their laws much weaker than the

determinate natures of micro-particles with their strong laws

would also show that artifacts are real on the present account.

If there are truths about objects, and they are posited as

objects at all, there are necessities about those objects which

qualify them as subjects at all, namely perhaps loose but law-

like existence and persistence conditions. If such necessities

are objective, those necessities are laws of nature. Nothing

about uniqueness of segmentation is required for there to be

natures. No “bottom fundamental level” is required, as long as we

are willing to accept something that does the job of laws, causal

powers, or natures as required for real objects.

The standard monistic view makes sameness absolute—not

relative to a predicate or to a way of articulating objects and

properties. The conception proposed in the next chapter denies

that there is absolute sameness, but retains objective

necessities, necessities that are not matters of conceptual or

verbal content, and so are not “de dicto.” The objects are real,



and have conditions for continuing to exist and conditions for

when there is one object or many objects.

One argument for a single privileged segmentation is

apparent direct reference. Absolute essentialism supports direct

reference. By “direct reference” is meant reference that is

unmediated by any kind term. That is, “This has to be H2O” is

true in virtue of there being a privileged predicate true of the

stuff, which predicate determines what the stuff is in itself,

that is, not relative to a way of characterizing it. Since there

is only one kind of stuff the stuff really is, “this” by itself

suffices to fix reference. The alternative account of

essentialism offered in Chapter 3 will treat apparent direct

reference as only apparent.

Why is the resulting account not itself a realism? It is, in

a way. There is a single, objective world independent of human

beings’ thoughts. What is missing is the realist idea that the

distinction between the real beings we posit and beings that

could be posited consistently with that reality is a significant

difference, which only a privileged articulation into the real

beings and properties can explain. The next chapter will provide

an alternative deflationary conception of Being. Roughly and

briefly, the idea is that beings and properties are, as it were,

metrics of reality in that way that meters and feet are metrics

for spatial dimensions. .

II Modal features of stuffs and individuals



This section argues that there are no absolute “de re”

necessities about kinds of stuff. It first argues that some of

the famous intuitive examples of de re necessary truths have

counter-examples if they are construed absolutely, i.e. as about

“the stuff itself.” The stuffs in question could lack the alleged

necessary feature, even though the modality is de re.

The section then develops arguments that there are de re modal

truths about many things to which most people would be reluctant

to assign absolute essences, or to think of as reflecting “joints

in nature.” These implausible natural kinds of objects have all

the features that seem to distinguish rigid designation--namely,

essential properties, and reference that is not determined by

speakers’ conceptions. Such objects and kinds, of course, are

exactly those that realist defenders of “cultural kinds” go to

exquisite lengths to defend. The difference in the neo-

Davidsonian answers I will propose is that, with a somewhat

deflationary account of Being, a great deal less exquisiteness is

demanded to retain the common-sense idea that tables, chairs, and

billiard-cues are real.

In accordance with Davidsonian semantics, the objects in question

are not determined by conceptual content of the predicates true

of them, but are of course determined by the reference of the

predicates that single them out. They are not determined by

agreement with the single segmentation of the World into Beings

and Kinds, either. On the other hand, since they are entities, it



is a matter of fact whether they exist or not after undergoing

various changes. So there are necessary truths about them and

they have essences. Being naturalists, Davidsonians interpret

“there are necessary truths” as “there are natural laws.”

a) Kripke’s persuasive examples

In 1970, when I attended Saul Kripke’s lectures that became

Naming and Necessity, I was completely entranced. They were the most

compelling lectures I have ever heard. For about a decade,

Kripkean-Aristotelian-essentialist revivalism seemed right. Over

the past several decades the infatuation has faded and I have

come to my Davidsonian senses. Kripke is right, but so is

Davidson. This section illustrates why, while Kripke’s intuitive

examples do indeed support essentialism, they do not necessarily

support monistic Aristotelian essentialism.

1) Water and H2O 79

Is water necessarily H2O?80 Water has to be H2O given that

its water, but that stuff itself need not have been water. So, is

this necessarily H2O? Here is one way, among many, that this very

stuff could be something other than water, and (of course also and

thus) something other than H2O: The components of any molecule of

H2O could have been configured as an atom of Carbon, an atom of

79 Kripke, Saul, (1980) page 128. See also Putnam, Hilary(1973). 80 Being predominantly H2O is presumably necessary for being water but not sufficient. Tea is not water, nor is Pepsi or broth, but those other substances may have as much predominance of water in their composition as the water in Mirror Lake, an algae-ridden pond next to my building..



Helium, and two atoms of Hydrogen. There are exactly the protons,

neutrons, and electrons in H2O as in the aggregate C + He + H2,

in the most common isotopes. As it happens, this stuff is in

water-configuration, but it did not have to be. Being water is an

accident of the stuff, which might have been a hydrocarbon-helium

mixture.

A metaphysical monistic realist need not follow Kripke at this

macro-level, and may not think water is a genuine kind of stuff.

Perhaps the stuff itself is a mixture of entities that are

essentially protons, neutrons, etc. However, we know that protons

and neutrons are not simples. Protons consist of two up quarks

and a down quark; neutrons of two down quarks and an up quark.

But now we need to intuit whether the decay of a down quark into

an up quark is an essential change in a truly fundamental entity

or a change of configuration.81 No end to compositional

complexity is obvious. We could hope that there is a “bottom

level” of entities it is worth positing that yields the real

kinds of stuffs. There is no reason to think there is such a

bottom level and very good reason to think not, as Jonathan

Schaffer has convincingly argued.82

81

? At this level, the very notion of “stuff” is of questionable application. The actual quarks in a neutron, for instance, contribute less than two percent to the mass of the neutron, the rest being due to the energy of the intrinsically massless gluons binding the quarks.82

? Schaffer, Jonathan (2003).



Of course, water is still necessarily H2O. But this

necessity is a relative, conditional necessity. “Since this is

water, it must be H2O.” The necessity that holds for the “this”

that is demonstrated in “this stuff” is relative to the stuff

being conceptualized or posited as water. There is no such thing

as “this stuff” tout court. This water coincides with many

different kinds of stuff, relative to each of which different

necessities hold.

2) The ice table83

a) Could that very table in the Woodrow Wilson School have

been made of ice? Kripke famously answered “No.” That table was

made of wood, so its identity as that substance (in Aristotle’s

sense) depended on its having this matter, to wit, wood. But

Kripke’s answer is wrong. The wood is protons, neutrons, and

electrons arranged in organic molecules, by and large. Those

components could have been organized so that they were largely

water, and, given that one wanted a table, ice would be the form

that would be imposed. The stuff, construed differently, need not

have been wood and could indeed have been ice.

Following Aristotle, we could say that the table is not now

potentially ice, meaning that the natural processes by which the

matter was re-organized into ice would take more than one step,

in the same way that he would claim that grass is not potentially

shoes, even though the grass can become a cow and the cow’s hide

83 Kripke, Saul (1980), pages 113-114.



can become shoes.84 The stuff itself, however, now shaped in

wooden-table-configuration, could have been ice. Also, the table

itself, the same stuff in nearly the same configuration (given

differences in the positions of protons, etc.) could have been

ice.

b) Kripke might have intended to say, not that this object

has to be wood to be this object, but that it couldn’t be the

same table if it were made of different stuff, however

characterized. This could be correct, relative to the item being

a table. Tables may be identified by the stuff they are made of.

But many entities are such that they could have had different

stuff. You and I, for instance, would be made of different

material if we had lived in different places and ingested

different doughnuts. The “metaphysical” claim that a table is

necessarily constituted by the “matter” it is in fact constituted

by is relative to the entity being construed as a table. There

are, as I will argue in the next chapter, equally ontologically

acceptable kinds of beings of which this table would be an

instance such that their identity and survival conditions mirror

those of organisms, or differ from the “table” standards in other

ways that allow multiple possibilities about which electrons,

etc., constitute them.

The inferiority of such alternative kinds (for us, at least)

is not ontological, but practical. Such alternative kinds are

84 See Aristotle’s Metaphysics Θ.



useless, but that does not undermine their reality. If

uselessness proved unreality, we could prove that Associate

Provosts do not exist.

3) Gold and other elements

3a) Gold85

Gold has atomic number 79. If something is gold, it

necessarily has that atomic number, and this necessity was

empirically discovered. However, this is not an a posteriori

necessity about this stuff but rather is an a posteriori necessity

about the conditions for this stuff being gold. If one of its 79

protons had instead been a neutron it would have been a platinum

ion instead of a gold atom. At a finer-grained level, if one of

the up quarks in one proton in the gold atom had been a down

quark, the whole atom would have been a platinum ion. On the

other hand, if one of the gold atom’s down quarks had been an up

quark, the atom would have been an ion of an isotope of mercury.

Gold has isotopes of atomic weights ranging from 169 to 205.

Platinum has isotopes with atomic weights ranging from 162 to

202. Given that the mass of a proton is roughly 1836 times the

mass of an electron, the gold isotopes with atomic weights in the

190s have more in common with platinum isotopes in the 190s than

with gold isotopes in the 170s.

A gold isotope that decays by electron capture, 195Au, picks

up an electron that turns a proton into a neutron, yielding a

85 Kripke, Saul. (1980), pp. 116 ff.



nearly identical platinum atom of the same atomic weight

differing from the parent gold atom by one part in roughly

358,000. That seems to me like a pretty minor modification,

objectively. There are some interests that make that difference a

significant one, just as there are some interests that make a

first edition of an historically significant work with title page

intact and advertisements at the end many times more valuable for

a book collector than a first edition without that title-page and

advertisements.

As to the other features of gold, you would be ill-advised

to gild your steeple with gold if it is anything other than 197Au,

since the other isotopes have half-lives ranging from 186 days

for 195Au down to 30 micro-seconds, with the vast majority in the

micro-second range. Likewise, investors should be cautious unless

they really wanted platinum.

3b) Elements in general

Chemical elements are often held to be natural kinds that

yield absolute de re necessities. Perhaps this is because many

elements occur in nature, generally persist through change, and

so on. This sub-section argues that there is no reason to treat

necessities about elements as such as absolute de re necessities,

primarily because elements are composites with no special claim,

qua element, to special status. There are certainly necessary

truths about oxygen and carbon. But those necessary truths are

truths about a given chunk of stuff only given that that stuff is



in, for instance, oxygen or carbon configuration.86 There is

little reason to think that chemical elements yield absolute de

re necessities. Those necessary consequences of oxygen or carbon

configuration are often very important, but that importance is

chemical, not metaphysical.

Stuffs consisting entirely of one element might be better

construed as kinds of aggregates of more basic components. The

status of chemical elements then should be, metaphysically, akin

to other entities made up of components. My tinker-toy set can be

a variety of things. When I have built a house, and then decide

to make it into a barn or a bridge, I add or remove or re-arrange

some components. Something comes to exist and something else

ceases to exist at each stage, perhaps an art object, a house, or

a construction we can enter in the abstract-tinker-toy-art show.

In none of these cases is some privileged entity created or

destroyed, although in every case an entity is destroyed. This is

not to deny that there are necessities about tinker-toy

constructions. A tinker-toy house has to have a roof, for

instance. A tinker-toy car has to have wheels. Given that this is

a tinker toy car, it has to have wheels.

86“Configuration” should not be read as saying that there are given items that

are so configured. The ontological thesis of this book, argued in the next

chapter, is that every kind of object is a posit, including quarks and whatever

smaller entities we may have reason to posit as composing quarks.



Uranium, with its 14 isotopes and various dispositions of

those isotopes to become other elements, undermines the idea that

there is something metaphysically special about being a chemical

element, because different isotopes behave very differently. The

destruction of the Uranium atom means something has been

destroyed, but the destruction is akin to the destruction of a

tinker-toy bridge. No special metaphysical destruction is

evident. Being Uranium is, from the perspective of its

components, an accidental relational feature of a collection of

neutrons, protons, and electrons.

The metaphysical importance of elements as such completely

vanishes when we consider some recently created trans-uranium

elements, which have very fleeting existences. Given that the

half-life of the most stable isotope of element 112 is 0.3

milliseconds, continuing to be element 112 hardly seems

significant. Even the relatively long-lived isotope 265 of

element 108, Hassium, (half-life 2 milliseconds) does not seem to

warrant special metaphysical treatment. Given that it is Hassium,

this tiny chunk must have 108 protons. If there were sense to be

made of “the stuff itself,” being Hassium would be a transient

accident of the stuff itself.

None of these kinds of stuff are “what the stuff itself” is.

It is doubtful that there is any sense to be made of “the stuff

itself” absolutely understood. There are de re necessities about

all of them, but there are different necessities depending on how



the stuff is described. Those are still necessities about the

stuff, but relative to a description.

At the atomic level. It is clear that there is not one

system of kinds of atoms into which they naturally divide.

Elements, that is, atoms sorted by how many protons they have,

are one way of sorting atoms which has a lot of utility most of

the time. But they are not the natural kinds. Some of them are

not natural kinds at all, in any normal sense of “natural.”

The de re necessities about stuffs are de re necessities

relative to a description. Given that this stuff is gold, it has

to have 79 protons. Given that it has atomic weight 195, it could

be gold or platinum, but it has to have 585 quarks.

b) “De re” necessities nobody thinks metaphysical

The following are examples of stuff and individuals about

which there are clearly de re necessities, but obviously not

kinds given by nature’s joints.

1) Pepsi

Pepsi is not on anyone’s list of natural kinds of stuff. The

present sub-section argues that there are true de re modalities

about Pepsi. This is thus an argument against interpreting de re

grammatical constructions as expressing absolute “metaphysical”

necessities. Pepsi is certainly a kind of stuff. What determines

that we are referring to Pepsi when we order a Pepsi by saying

“Pepsi” though, is not a matter of our concepts, but rather a



matter of what is there in the world.87 For a Davidsonian, the

application-conditions for being Pepsi have to be worldly: “Is

Pepsi” is true of some stuff A just in case A is indeed Pepsi.

Nothing about concepts is relevant here.

People who succeed in referring to Pepsi may have radical

misconceptions of what it takes to be Pepsi. Pepsi is in fact the

only cola sold at Burger King, but Burger King could have signed

a contract with Coca Cola. A person’s concept of Pepsi might be

only that Pepsi is the cola sold at Burger King. What people know

about Pepsi might be contingent features of Pepsi. Thus Pepsi

shares a feature of proper names. “Pepsi” names a stuff, and its

referent is independent of the content of the speaker’s concept.

In the counterfactual situation of different negotiations in

which Burger King contracts with Coca Cola, “Pepsi” still refers

to this stuff and not to Coke. Pepsi is a rigid designator—it

refers to Pepsi in “worlds in which Burger King negotiates

otherwise,” so to speak. That is, worlds in which Burger King

87 Some questions arise about whether it is necessary for Pepsi to be made by Pepsico. Can there be pirate batches of Pepsi? Could Pepsi be a Coca Cola product? A worry for the Pepsi-theorist might be what would be the case iF Coke had bought out Pepsi in the 1980s but continued the brand, but modified the recipe.

Can there be home-made Pepsi? I’m not sure. Pepsi may be an institutional kind, but some corporate products seem to have home-madeversions. When I was a child, my family had little disposable income, so, having played Monopoly at friends’ houses, I made my own set to play with my younger siblings. Was that a real Monopoly set, a pirateset, or a fake set?



signs a contract with Coke are not worlds in which Pepsi is Coke,

to use the familiar metaphor.88

Pepsi can vary chemically somewhat even on Earth, perhaps in

different markets. Precisely how far from the chemical

constitution of this glass of Pepsi a liquid can be and still be

Pepsi is vague. But, while some counterfactuals about Pepsi are

unclear, it is very clear that if a ginger ale distributed by

Pepsico had been called “Pepsi” it would not have been Pepsi.

Pepsi is a cola, and necessarily so. So, “This couldn’t have

tasted like ginger ale,” is a true grammatically de re

construction about a glass of Pepsi. This is not to deny that the

“very stuff” could have been re-configured atomically to taste

like ginger ale.

So, there are things Pepsi could be and things it could not

be. So, Pepsi has modal properties and modal endurance, just as

water and gold do. But this doesn’t mean that this glass of Pepsi

in itself, this “very stuff,” has to be Pepsi. Surely in this

case, without imagining subatomic rearrangement, the ingredients

did not have to be put together in a Pepsi configuration. Being

Pepsi is something quantities of stuff retain under some

counterfactual situations but not under others. Necessities about

88 Pepsi need not have been called “Pepsi.” Just as there are worlds in

which Aristotle is not called “Aristotle,” so there are worlds in

which Pepsi is called “Pipsy” or even “Fred.”



Pepsi are objective relative necessities, but not conceptually-

grounded necessities.

2) Bouillon

Pepsi may be in some respects an “institutional” kind, in

that it may require the existence of Pepsico, and so

corporations, capitalism, etc. Bouillon has no such corporate

ties. So, the essence of bouillon would be laid out in

propositions such as “Bouillon cannot be creamy,” and “Bouillon

can have more or less salt.” Thus we would get the particular

claims, “This could have had more salt,” and “This would not have

been bouillon if you had put in milk and a lot of barley.”

Quantities of bouillon are typically quantities of saline

solution. Saline solution has to contain salt, but bouillon does

not. Suppose your restaurant makes bouillon by buying hospital

surplus saline solution and boiling animal remains in it. The

restaurant also offers salt-free bouillon, which is made by

electrically removing the sodium and chlorine ions from regular

bouillon. When I am presented with a bowl of bouillon I ordered,

and thought about ordering the salt-free kind for health reasons,

I might say, “This could have been salt free.” What I said is

true of the bouillon before me; but false of the bowl of saline

solution before me. My “this” does not refer to the stuff itself,

but to the bouillon. Such grammatically de re modal utterances

should be understood as relative to a predicate or kind of



object, in much the way that “He’s tall” is understood relative

to a predicate. I develop this suggestion below.

There is no chemical definition of bouillon, although there

clearly are some chemical necessities about it. For instance,

sulfur cannot be the predominant ingredient. The “stuff itself,”

given the reflections of the previous section, need not contain

water, and need not contain hydrogen and oxygen. The quarks might

be differently arranged. We will see below that Pepsi and

bouillon need not contain water even when they are Pepsi and

bouillon.

As is the case with Pepsi, a person’s concept of bouillon

may not capture the essence of bouillon. A non-soup-savvy person

might have a conception of bouillon consisting of just “a kind of

soup Campbell’s sells.” That person is referring to bouillon even

though that is a contingent feature of bouillon. She has heard of

bouillon, knows it’s a kind of soup, has seen it on shelves in

the supermarket, but doesn’t know much more than that.

“Bouillabaisse” was that way for me before I googled it to get

the spelling right and saw that it is a kind of fish stew.

Bouillabaisse necessarily contains fish even though my concept of

it had nothing about fish in it. Many people have that relation

to certain color-words. I know that chartreuse and puce are

colors, but I have no idea what general color either one is a

shade of.



3) Pepsi and bouillon on Twin Earth89

Consider these substances on Twin Earth. On Earth, both

Pepsi and bouillon contain water. If Pepsi establishes an

intergalactic branch on Twin Earth, the franchise operators will

use XYZ to make Pepsi there. Bouillon on Twin Earth also contains

XYZ. If this is right, chemical composition is not crucial for

either of these stuffs. Perhaps our intuitions about water on

Twin Earth should be re-examined, given that Pepsi and bouillon

retain their natures with change of chemistry. Of course, some

chemical compositions are incompatible with being bouillon.

Nothing consisting entirely of elements with atomic numbers

greater than 95 could be either bouillon or Pepsi.

There are truths about Pepsi and bouillon that rest on “what

it would be to be Pepsi” and “what it would be to be bouillon,”

to use Aristotle’s ‘to ti ein einai” explication. In the

relative, conditional sense to be explained below, Pepsi and

bouillon have essences. That is, if a substance is bouillon, it

has to have amino acids. If a substance is Pepsi, it can’t

consist predominantly of lead. These necessities are necessities

about the truth-of-conditions of predicates, not about concepts

and not about “stuffs in themselves apart from all predicates.”

Pepsi and bouillon surely are not divided from other stuffs at

joints in the cosmic structure. Pepsi and bouillon are kinds of

89 Putnam, Hilary (1973).



stuff, but it is implausible to think that nature itself divides

matter into the Pepsi and not-Pepsi.

Is water any different from Pepsi? Water is a more natural

kind than Pepsi or bouillon for us. We and our fellow organisms

on this planet can get along without Pepsi, but not without

water. The naturalness of the kind water, though, rests on our

being organisms which require water in order to live, not on some

metaphysical distinction of Being. In a broad sense of “culture,”

the “culture of carbon-based organisms,” water is a cultural

kind, just as Pepsi and bouillon are.

4) bushes and other medium-sized individuals

Colas and soups are stuffs, and so not paradigm individuals.

Bushes and trees, though, are discrete organisms and so full-

fledged Aristotelian substances. They have much the same kind of

de re features as persons and electrons, even though they are

rejected by botany as kinds of organisms. “Is a bush” is not a

scientific botanical classification of plants. We can be

confident that botanists will not discover a special dna profile

that characterizes all and only bushes. Nevertheless, bushes are

real and there are de re necessary truths about those objects.

Bushes are necessarily mostly not over one hundred feet tall, for

instance. There could be a freak azalea that grew to over one

hundred feet, but that just shows that Aristotle was right to

speak of natures as being “always or for the most part.”

Aristotle knew of two-headed cows. These necessary truths are not



truths that rest on concepts. Hilary Putnam may know about bushes

only that they are the short woody plants in front of buildings,

and be under the necessarily mistaken impression that young trees

are bushes until they mature. The same things apply to “is a

tree.” There are de re necessities about trees that Aristotle did

not know, for instance that they are carbon-based organisms.

Among inorganic individuals, rocks, stones, and other terms

of divided reference, are like trees and bushes, perfectly good

predicates true of objects. There are some natural laws about

rocks and stones, but those count nouns are not part of any

natural science.

5) The above examples: Why no frogs?

What are missing from the above apparent counter-examples to

absolute monistic essentialism are paradigm Aristotelian

substances, things like organisms. For organisms, and perhaps for

some other objects, there do not seem to be intuitive or easily

comprehended alternative descriptions relative to which different

necessities obtain about them. A theory which denies that there

are absolute natural kinds must give some kind of explanation of

the truth lying behind Aristotle’s intuition, which he shares

with Van Inwagen,90 that organisms are especially clear cases of

substances. As Aristotle implies in saying that in some sense

every genuine substance is identical with its essence,91

90 Peter van Inwagen, (1990)91 Metaphysics Z6



organisms are independent of their matter. That is while an

organism has to have some matter, there is no particular matter a

particular organism has to have, unlike the case with artifacts.

Organisms are special.

I certainly agree that organisms are special. I love being

one. Why do these considerations not convince me that nature

itself selects frogs and squirrels as natural kinds? Let me

review several considerations, some of which explain away these

intuitions and some of which attack squirrels and frogs.

a) One most compelling reason that kinds of organisms cannot

reflect a unique set of joints in nature is that the species

which define them are vague and loose. One way to illustrate this

would be by looking at actual biological research on speciation,

which treat it as a gradual process. Another route to the same

conclusion is to reflect that there are continua the steps of

which are one-gene genetic variations between, for instance,

capybaras and cows. That is, to use a sorites argument. There are

explanations of why in fact the distribution of animals is

clumpy, with lots of cows and capybaras but few intermediates,

but those explanations have to do with reproduction, not natural

joints.

b) Being an organism itself is not so clear in some cases.

Viruses are questionable. The individuation of organisms in

sometimes not intuitive at all. Portuguese Man-o-Wars differ from

humans in having larger components as their specialized parts—



multi-celled organisms rather than cells. The thirty-acre fungus

in Wisconsin is allegedly one single organism rather than Siamese

twins.

c) Organisms will of course be salient to us because hundreds of

millions of years of evolution have fostered the detection and

tracking of predators and prey—medium-sized objects like us. It

certainly behooves us to posit organisms with edges, and that is

precisely what we can thank our ancestors for. That salience to

us, though, does not require nature’s laws to articulate the

universe into the organisms and the non-organisms any more than

numerous other possible articulations.

II Natures as Explanations of Modal Truths

How can there be law-like relations without a privileged

segmentation? As far as I can see, the intuitions that lead one

to connect laws, natures, and necessity from a monistic realist

perspective will do the same on a perspective, to be laid out in

the next chapter, on which there is no segmentation in nature at

all.

From Plato’s Phaedo 92 on, formal explanations have explained

necessary connections by appealing to the natures of things. The

theory being developed here continues that tradition. The present

theory does not suppose, though, that there is a unique division

of reality into beings with their natures. Here is an analogy:

suppose space has some determinate geometry, say Euclidean. The

92 Plato, Phaedo, St.100c3-c8.



theorems of that geometry determine relations between the lengths

of a hypotenuse and the lengths of the side of a triangle. But

the units do not matter. Feet, meters, ells, cubits, and

innumerable other unposited bits of space are themselves

constrained by the same structure. Space itself, though, has no

special units. Any substitution of another stable unit-term for

“meter” in “The nature of a right triangle with two shorter sides

three meters by four meters determines that the third side is

five meters long” will be a truth.93

The precision of the determination depends on the

suitability of the unit. If we had units that were less suitable,

the results would be probabilistic and approximate. If, for

instance, we measured in paces by stepping off distances, the

theorem would only give us approximate predictions about the

number of paces in the hypotenuse of a triangle with sides three

thousand paces and four thousand paces. Paces are not so good for

detailed work. This is not to say that there are no paces, nor

that paces are a bad unit. Sometimes nothing better is available,

many times paces are convenient and good enough. Cubits, hands,

and ells likewise will give very approximate results. Reliable

physical standards for units such that units “have the same size”

so that theorems could be expected to actually apply in detail

have only recently been achieved.94 Given the density of the

number line and the inability of physical measurement to

93 This metric analogy was used by Davidson (1977) p. 224-5.



distinguish rational lengths from irrational ones, no complete

accuracy is even possible.

In an extreme case, say a unit based on the meter that

assigned successive units a randomly- assigned real number of

meters between one and a thousand, the theorems of geometry would

tell us nothing about relations among sides and hypotenuses.

Usable units would be somewhere between the meter and these

bizarre units. There would be enough relation to the underlying

geometry to make the units useful in practical matters. Which

practical matters, of course, would depend. You would not want

your surgeon to measure your stents in thousandths of a pace. It

would be (doubly) foolish to fight a duel with pistols and

require that the standard meter replace paces.

From the perspective of the present chapter, there is no

question about the reality of the units. They are all real

legitimate divisions of space. The space is there, so any chunks

of it are there. They equally share in Being. What should be

obvious is that none of them are the “genuine” or “given” units

of space. We don’t think of space as having joints at all. The

94 “Since 1983 the meter has been defined as the distance light travels in avacuum in exactly 1⁄299,792,458th of a second (17th CGPM, Resolution 1). Thisdefinition of the meter makes the length of the meter depend on the durationof the second; by definition the speed of light is now exactly 299,792,458 meters per second.” (www.sizes.com/units/meter.htm)

I am a bit worried. If a successor physics were to discover that lightgets tired after travelling a real long distance, this definition might turnout to be pace-like.

http://www.sizes.com/units/meter.htm

http://www.sizes.com/units/cgpm.htm



science of geometry, of course, has its own count-nouns. Lengths

and volumes provide enough articulation to state the general

truths about geometry without supplementary positing of units of

length or volume.

The laws of nature generally, from this perspective, differ from

those of geometry in that something akin to the units are

required for the very statement of the laws. That is, the laws

themselves require expression in terms of beings, whether

particles or fields. It is as if geometry could only be stated in

terms of ells, meters, paces, or some other physical unit. Call

the underlying necessity Ananke. The underlying Ananke that made

the geometries have any success at all in predicting hypotenuses

would be the same for the different units, but the quality and

precision of the laws as stated in varying units would differ a

lot. That may be what happens when you concretize geometry and

add energy.

Ananke may be the way to think of necessity. Strictly

speaking, talk of “laws” misstates what is under discussion. Laws

are linguistic expressions. There being a law is a necessity

sentence in some language being true. Since nothing makes a

sentence true, there is no thing or fact that is a law’s being

true. “Is a law” is a characterization of sentences, so that we

can describe what it is about the sentence that makes it

necessary, but this predicate adds nothing to “necessary.” For a



variety of reasons discussed in Chapter 5,95 I think that

necessity and possibility are primitive, which is to say that

there is no reduction of these concepts to non-modal ones. A

theory of the modals will be a theory of how various modals are

related. For the moment, think of natural laws, making predicate-

families more or less law-abiding and useful, as the expression

of Ananke.

The details of how to think of laws and modality from the

Davidsonian perspective will be discussed in Chapter 5. How to

think of Being is the topic of a section in the next chapter.

object alternatives. But that’s just Chapter 3: Natures,

Necessity, and Relative Essentialism

This chapter argues for a kind of essentialism that is

compatible with Davidson’s views, as I understand them. Davidson

is committed to the truth of various modal judgments about what

has to be the case just in virtue of the fact that these are

obvious truths. It is clear that these truths are de re modal

truths. So, “That cow has to be a ruminant” is true but “That cow

has to go `moo’” is false even though the concept of the very

95 Briefly: Reduction of modals to part-whole necessity a la “possible worlds”is like reduction of predication to the single predicate “participates” or “isan element of.” Just as you still have unreduced predication for Plato or a set-theorist, you have primitive modal truth even on a realist possible world view. The same is true for other accounts. So one may as well accept primitivemodality to start with..



young person speaking is such that the salient, defining feature

for “cow” is “Cows go `moo.’”

I Davidson’s commitment to essentialism

In this section I argue that Davidson is implicitly

committed to a kind of essentialism. Since this is a conclusion

Davidson himself would have annoyedly rejected, I will defend it

at some length.

Some kind of essentialism is manifest in Davidson (1974),

page 192: “The notion of organization applies only to

pluralities. But whatever plurality we take experience to consist

in—events like losing a button or stubbing a toe, having a

sensation of warmth or hearing an oboe—we will have to

individuate according to familiar principles.” Davidson argues

that the idea of a domain of objects to be sorted by a

“conceptual scheme” presupposes that objects of some sort have

already been posited as something. His point is that the objects

to be grouped into alternative extensions must already be

conceptualized as some kind of thing in order to jointly

constitute a plurality of distinct entities. There are no

pluralities without count-nouns individuating them. So, he

rejects as incoherent the idea that conceptual schemes sort a

pre-conceptual “given” in possibly different ways.

Davidson’s conception of individuation is Aristotelian.

According to Aristotle, for some predicates, there is no thing to

which that predicate is attached. According to Davidson, for some



predicates, there is no thing which can be assigned to an

extension. Both Aristotle and Davidson hold that some count-nouns

are entity-constituters, predicates which bring a principle of

individuation and so say what it is to be that individual. For

every object, there is some general term such that that general

term gives the “what it would be to be,” the “to ti en einai” of

that object. That is, for any object there is some feature such

that that object would not be itself without that feature. That

is what “principles of individuation do.” They provide conditions

for when an object exists, how many objects there are, and when

an object still exists. The passage quoted above is thus a

Davidsonian version of Aristotle’s idea that the essence of an

object cannot be an attachment to that object, but must rather be

what constitutes that object as an object with objective

persistence conditions. In Davidsonian terms, Aristotle agrees

that there is no given, in the sense of a plurality of beings

prior to substances with their natures.96

The very large difference is that, for Aristotle, the

predication is done by nature, as it were. Conceptualization, for

Aristotle, is the mind’s adequation to the given structure of the

world. Davidson, on the other hand, is more a Kantian. Like Kant,

Davidson’s frames the issue in epistemological terms, holding

96 We argued in Chapter 2 that even David Lewis is committed to Aristotelian

essentialism about some kind of objects.



that there is no full-fledged cognition without

conceptualization.97

Davidson is also a pluralist, at least about what substances

and attributes there are. An important text in this regard is his

response to the Goodman “grue-bleen” paradox.98 Whereas many

philosophers have taken the problem to be one of finding

something about “green” such that things really are green but not

grue, Davidson points out99 that “grue” is perfectly fine, if

applied to other objects. Briefly, (switching examples, as

Davidson does for literary effect) supposing that emeroses are

emeralds if examined before 2020 and roses thereafter, then if

“All roses are red” and “all emeralds are green” are law-like, so

is “All emeroses are gred.” Given that there are emeralds and

roses, there is little reason to deny the existence of emeroses

and gred things. There are truths about them.

Davidson would have been reluctant to call his view any

version of “Aristotelian essentialism,” just as Quine was

reluctant to so characterize his view.100 Quine and Davidson of

course treat essentialism as the view that entities have

necessary features regardless of how they are characterized. They

97 Kant, (1785). A 51.98 Goodman, Nelson. (1955). 99 Davidson, Donald. (1966).100 Quine (1953b) page173-174.



are thinking of monistic essentialism, the idea that nature

itself is divided into beings and properties.

Quine, however, does have a given—stimulations, the impacts

on sense-organs, which he takes to be the objective material

relative to which we posit various kinds of objects. So, for

Quine, posits are at a kind of remove from actual reality. For

Davidson, since there is no given, the objects with their

individuating conditions are epistemologically basic. Their

relation to what it is to be one of them is like Aristotle’s—not

to be understood as something applying to something else.

The obvious question is how essentialism comports with

Davidson’s remark that sameness is always relative to a

predicate. There is a tension between Davidson’s apparent

rejection of essentialism and maintaining the Davidsonian view

that to have a domain of beings requires conceptualization as

something. If conceptualization is basic, each being must have a

“what it would be to be,” some predicate true of it apart from

which it would not be that thing. Since ascribing a“what it would

be to be” is ascribing an essence, then the conceptualization

that could yield a domain of objects is an assignment of an

essence.

The obvious resolution of this tension out for a Davidsonian

is to articulate a version of essentialism that allows there to

be constituting features of objects and so necessary truths about

objects while denying that there is any sameness except relative



to a predicate. The general idea is that, given Davidson’s

conception of reference as discussed in Chapter 1, the

necessities about an object relative to that object being a

particular kind of object are not conceptual, but de re. In

Chapter 1 I argued that Davidson is neither a causal theorist nor

a resemblance theorist about extensions of predicates. Only the

truth-condition clause for the predicate gives the meaning. If

Davidson were to allow, which he should, that different objects

can share the same space at the same time, then there would be

necessary truths about each of them. However, there would not be

a single entity to be pointed to in that space about which those

necessities were true. So, both sameness and necessity would be

relative to which kind of thing is taken to be the occupant of

the space at that time. This will mean that some cases which

Davidson thinks of as the same thing under different descriptions

are actually distinct things. I argue in the next chapter that

that modification of Davidson’s application of his view is an

improvement.

This chapter develops a way of thinking about entities on

which, while samenesses of entities are relative to predicates,

entities are still constituted by those predicates being true of

them. It will also be clear how there can be de re necessary

truths about such entities. I claim that Davidson is an

essentialist malgre lui, but that is because, in denying a given,

Davidson’s conception of entities seems to me to be like



Aristotle’s, minus Aristotle’s idea that there is only one set of

predicates that say what things are in themselves.

In spite of Quine’s rejection of Aristotelian essentialism,

the debt of the theory to be presented to Quine should be

evident. Quine’s term for this conceptualization is “positing.”

He thinks of posits as not thoroughly objective, but rather as

underdetermined sortings of sensory inputs. This chapter will

retain the same term, “posit,” but understand it to mean the kind

of direct access101 contact with objects Davidson gets by dropping

the notion of a given. Without the contrast to a given, calling

something a “posit” does not diminish its claim to objectivity

and reality.

For Davidson, positing is generally not optional and usually

not voluntary. Davidson is a partial subscriber to Quine’s

dictum, “no entity without identity.”102 Where Quine means that a

positer has to specify the identity-conditions of a posited

object, Davidson need only require that there be objective

identity and persistence conditions. Davidson can agree with

Kripke and Wittgenstein that there are very few definitions

outside of mathematics which give necessary and sufficient

conditions for anything. If positing dogs is acceptable, then

101 Davidson (1974), p. 198: “In giving up the dualism of scheme and world, we

do not give up the world, but re-establish unmediated touch with the familiar

objects whose antics make our sentences and opinions true or false.”102 Quine (1969) page 23.



positing entities without knowing necessary and sufficient

conditions is acceptable.

For useful posits, some necessary truths will be

consequences of a given kind of object being posited. While it is

true that there will not in general be a definition of a term in

other terms, a useful posit will have some law-like connections

with other terms. We can be quite confident in the truth of

sentences like, “Cows cannot be deciduous” “Cows have to be warm-

blooded,” and so forth even if we are not cow experts. If we

understand a term, and the referent of the term is a useful

posit, there will be true modal remarks about those objects—what

they can and cannot be.

Many of the modal truths associated with medium-sized

objects will be weaker than necessity.103 For some natural kind

terms, all the special modal truths about a given kind may be

weaker than necessity. Such conditions would be probabilistic, in

the way that Aristotle’s natural laws, which are “always or for

the most part”104 are. Aristotle’s phrase for such weaker

necessities, “always or for the most part,” I understand to be a

claim of necessary conditional probability. If it is the nature

of a cow to be “always or for the most part” single-headed, then

that likelihood is a necessary feature of a cow. Such a statement

103 Of course, in a sense that will not produce useful essential truths,every posit allows specification in terms of necessary and sufficient conditions. “Is a cow” is true of an object just in case it is a cow.104 Aristotle, Physics, B5 196b10; Metaphysics Epsilon 2 1027b30ff.



of necessary conditions would be, for example, “Necessarily, if

an animal is a cow, then it very probably has only one head.”

Such necessary conditional probabilities are sufficient for

tracking entities.

We can, of course, arrive at posits that have non-

probabilistic essences, or whose necessary conditional

probabilities are higher, or are more useful for some purposes.

For instance, we have supplemented Earth, Air, Fire, and Water as

elements with chemical kinds. But that supplementation is

compatible with continuing to posit earth, air, fire and water,

which we of course do. We have earth-moving machines, water

parks, air-quality indices, and fire departments.

A posit will generally not entail necessary and sufficient

conditions in other terms for the existence, persistence, and

identity of a posited object. A posit of animals as a kind of

being can take place without any speaker knowing what exactly it

takes for an organism to be an animal, as long as the posit has

useful enough connections with other predicates. However, any

language-user who understands the term will know some necessary

probabilistic necessary conditions and some necessary

probabilistic sufficient conditions. In general, then, positing a

kind of entity generates necessary truths.

II The Miracle of Beings and Properties

This section argues for a kind of Kantianism about general

terms and singular terms. It is an interpretation of Davidson’s



view on positing objects. In effect, I understand Quine, and

Davidson following him to have proposed an alternative conception

of Being. Beings are in a certain sense artifacts rather than

givens. This is a Kantian interpretation of Quine’s

characterization of entities as “posits” interpreted in

Davidsonian given-less terms.

I start from this consideration: We cannot be rational

agents and thinkers unless we can make well-founded inferences

which are not truth-functionally valid. A rational being must be

able to reason from, for instance, “All frogs are green” and “Joe

is a frog” to “Joe is green.” Such inferences are only possible

in a language with sub-sentential structure, that is, equivalents

to singular terms and general terms. So, any thinking being must

think in terms of objects as referents of singular terms, and

properties, as general terms true of such objects.

On one conception, the anthropocentric one, our minds are

suited to nature by a natural law about us. For Plato, there is

an affinity of soul to the structure of reality.105 For Aristotle,

the intellect is a faculty which gets forms of natural kinds

without their matter. Nature is articulated into beings and

properties and the articulation our thought and speech, happily,

matches the articulation of the world. The modern version of the

adequation of our minds to the joints in the world makes our

minds conform to nature via an account involving evolution. But

105 I take this to be an interpretation of anamnesis.



only very imperfect laws will result from this approach, given

the vast difference between the equations that describe what

happens in space-time and the crude, exceptionful generalizations

available about macro-objects. It is difficult to see how any

version of “natural joints” could correspond to the vague

extensions of the predicates true of medium-sized objects that

the sorites argument makes salient. As discussed in Chapter 2,

this is the dilemma that post-Aristotelian essentialists have had

to deal with since it became clear that the objects of the

sciences with the strongest laws are not organisms and not the

objects we organisms deal with in our lives. I regard this as the

main crisis of contemporary metaphysics.

Given that we correctly make strong modal claims about

medium-sized objects, some kind of basis in the natures of things

must be constructed to support such claims. Many philosophers106

who quite reasonably wish to make their metaphysics accommodate

common sense have proposed a variety of accounts of how the

objects of everyday life fit with an underlying world of micro-

particles. I do not intend to try to refute their views or to

show that no such view could possibly work.

On the alternative Kantian107 conception, there is no

miraculous match between the structure of language and a special 106 See, for instance Elder (2005), Millikan (1984). 107 By “Kantian” I mean that the accommodation is from our side rather from the side of the world. We impose an articulation on what is not intrinsically articulated.



ontology of beings and properties. Rather, we impose a structure

of singular terms and predicates as a condition of thinking and

using a language. This is not to say that we misrepresent

anything in doing so, or that what our sentences are about is

“phenomenal” rather than noumenal. Just as any text must be in

some font or other, any thinkable thoughts and speakable

discourses must have general terms and singular terms. From this

point of view, the idea that nature itself is divided up into

beings and properties to correspond to our predicates is like the

idea that only Times New Roman gets English right.

I call this view of ontology “Kantian” because of the strong

similarity between Kant’s approach to metaphysical questions and

what is being proposed here. Kant thought that many metaphysical

questions were mistakenly taken to be questions about a given,

knower-dependent reality. He had arguments designed to show that

those issues were really about what knowers have to be like. In a

similar fashion, the present view of ontology takes the

articulation of the world into beings and properties to be an

artifact of the necessity of thinking in a language that allows

inferences that depend on sub-sentential structure, rather than a

feature of the world itself, intrinsic to nature. This Kantian

version of Davidson, like Kant, takes the structure of logic to

be the structure of the world as we can think it, but with a

better conception of the structure of logic.



Quine’s and Davidson’s analogy is that to suppose there is a

given articulation in nature would be like supposing that space

comes intrinsically divided into meters, and time comes segmented

into seconds. It is true that in order to apply mathematics to

space and time, which application is a necessity for the

possibility of formulating decent physical laws, some unit must

be posited. Likewise, in order to think about the world, the

thinking must be in an articulated language. To deny that space

is itself articulated in terms of meters is not to deny that

meters are real.108 Likewise, to deny that nature divides the

world into chairs, squirrels, and hadrons is not to deny that

chairs, squirrels, and hadrons are real things. Just as space in

not given in meters, so the physical world of objects is not

given in chairs, squirrels, and hadrons.

The general character of the posits and predicates we in

fact employ is by and large determined by our billions-of-years-

long sequence of ancestors. Most of our posits are inheritances,

rather than voluntary creations.109 By evolution’s mechanisms, we

have come to apply medium-sized object predicates to ourselves

108 Measures of quantities are peculiar items, about which a lot could be said. Schwartzchild and Wilkinson (2002) call them “classifiers,” but do not seem to say whether they refer to entities. My view is thatthey denote odd entities, since they take definite and indefinite articles. “An extra inch would make him tall.”109 Quine’s (1969) is an early expression of this idea about the acclimation of our posits with the workings of the physical world.



and our surroundings. Since “is an agent” and “is a language” are

part of our inheritance, and apply to us and to our thought, we

are unlikely to coherently imagine a fundamentally different

alternative, a radically different set of predicates, that

language-using agents could be interpreted as having. On the other

hand, we perfectly well envision regional alternatives, and over

time our predicates change—we did not talk about gluons even

seventy-five years ago.

To review: We could not think unless we could recognize

formal inferential relations among truth-functionally simple

sentences. Any language that allows logical relationships among

truth-functionally simple sentences must have singular terms and

something equivalent to predicates. A semantics for such a

language of course assigns objects as semantic values of singular

terms and is at liberty to assign properties or sets as semantic

values of predicates. A natural question is, “How does it happen

that reality happens to be divided into objects and properties

just as it must be in order for us to think about it?” It is

possible that the way the world is in itself, is divided into objects

and properties, so that the world just happens to conform to the

requirement of thought and language adequate for describing that

world. It might be just lucky that nature is divided into

properties and objects, which are just what is needed to be the

semantic values of singular terms and predicates. Or it might be

part of the Divine plan.



The two real possibilities for this coincidence seem to be:

a) We have evolved the singular term/ general term scheme as an

adaptation to the world’s being naturally divided into objects

and properties.

b) The singular term/ general term scheme is an artifact of the

subjects who are thinking and speaking. Subjects who think and

speak have to posit objects and properties in order to be able to

think and speak. These subjects are of course also posits, as we

will discuss below.

The next few paragraphs argue against possibility a). If

there were objective divisions in nature, the most likely

divisions in nature would have only a loose connection to the

objects we have evolved to be and to notice. We impose the

singular term/general term scheme independently of any pressure

from the natural divisions of the world, if such there be.

As naturalists, Davidsonians take necessities to be grounded

in natural laws. The “what it is to be” of an electron is

constituted by the laws about electrons. If our use of the

object-feature scheme were itself shaped by nature, in the way

that, plausibly, the particular predicates we use are shaped by

our evolutionary history,110 then the divisions natural for us 110 Quine’s(1969) argument gives the best explanation of the happy match between our predicates and the world we live in. Billions of years of evolutionary history has indeed shaped our conceptions to the phenomena recognition of which is of supreme practical importance. Such an evolutionary argument, though, is ill-equipped to explain how we have come to have the object-property scheme itself. Our starting point, surely, would have nothing to do with nature’s fundamental joints.



would be the divisions in nature revealed by the most excellent

natural laws. However, we have good evidence that the best

candidates for reflecting the fundamental divisions in nature are

micro-particles and fields, relative to which the medium-sized

objects of everyday life, including ourselves, are not definable

or really connected by any good laws. Most of our predicates have

only the loosest connection to the micro-particle laws that would

be the most accurate indications of nature’s intrinsic structure,

if there were such an intrinsic structure.

So, we posit properties and beings whether or not our posits

are dictated by nature’s laws. The explanation of why we impose

the predicate-singular term format cannot be that that is how

nature is. If there were intrinsically fundamental objects in the

world, they would be micro-particles and fields, relative to

which the objects of our “lived world” are not connected in a

law-like way which would allow reducibility. The beings that

include us organisms and are the focus of Darwinian concern of

organisms are not the fundamental objects. If it had not been for

the features of ribonucleic and deoxyribonucleic acid that

brought about complexes that can reproduce, and the happy

accident that there are environments in which such complicated

compounds can occur, terms for organisms and medium-sized objects

would have no application whatsoever.

So, it’s hard to see how nature’s being divided into beings

and properties would actually explain how it came about that



nature is structured in a way that allows it to be intelligible

to us. The divisions into properties and beings that matter to

organisms, including the ones that identify those very organisms,

seem not to be a product of nature, but rather a very sloppy

product of those very sloppy products, organisms, themselves.

RNA and DNA, in the right environment, lead to complexes

that respond to complexes differentially. In the longish run,

selection leads to some of the complexes having languages whose

referents, so to speak, are the sloppy objects recognition of

which account for their existence. Organisms and medium-sized

objects, that is, are best construed as cultural objects of the

advanced carbon-based replicator culture. We make our objects. In

Hegel’s sense, our objects are us.

In any case, nature doesn’t seem to be doing any work on the

object-and-property front that really accounts for the nice

correlation between language being possible and nature’s actually

having a particular articulation into beings and properties. That

is, even granting that nature itself is intrinsically organized

into beings and properties, that would not explain our

accommodation to that fact.

The striking truth is that we have to think in terms of

beings and properties whether or not nature dictates those

properties and beings. We for two thousand years at least thought

that the fundamental kinds of stuff are earth, air, fire and

water. We still think those are kinds of stuff, of course, since



there are truths about, for instance earth and earth-moving

machines. We can by and large sort stuffs according to those

predicates, and it is useful to have them. Their usefulness means

that there are some law-like at least approximate generalizations

about them. However, we don’t think that nature divides the world

into Beings such that this division is central. We do not think

that Earth is a natural kind.111

III) Alternative Positing and Alternative Conceptual Schemes

Following Goodman’s (1955), and the literature that

followed, we tend to think of alternatives to our predicates as

re-sorting of objects and stuff we already posit. This may

sometimes be the case, perhaps in biological classification and

perhaps other topics. In the general case, the situation with

different predicate schemes is nothing like that.

We share enough of Aristotle’s beliefs to understand him

rather well. But what re-sorting of the extensions of Aristotle’s

terms would yield magnetic fields, electrons, or neutrons? These

111 The Quinean-Kantian conception that all beings and features are posits should not be confused with the idea that the universeis gunk, i.e. infinitely divisible stuff with no intrinsic articulation. The gunk hypothesis is a view about how many beingsthere really are; the present theory is the claim that any objects and properties whatsoever are posits. There is no number which numbers the beings, just as there is no number which gives the distance between Grand Central Station and the Empire State Building. It all depends on whether the measure is paces, meters,or cubits.



entities are very likely to be preserved in future theories, and

it seems reasonable to conclude that there are indeed truths

about them. But no re-sorting of Aristotle’s entities will give

the extension of “is an electron” even though electrons are

ubiquitous as components of the entities Aristotle shares with

us. We can reasonably suppose that if scientific progress

continues, there will be truths discovered using predicates that

are to us as “gluon” would be to Aristotle—not constructible by

sorting the extensions of our predicates. Note that this is not

to say that those predicates are inexpressible in English or that

“gluon” is inexpressible in classical Greek.

The very idea of positing objects is that those objects are

not generally available to the thinker without the positing. This

does not mean that, for instance, electrons did not exist before

humans started to talk about them. We are using our language to

correctly describe the way the world was. Since our utterances

are true, electrons existed in classical Greece. It is of course

the case that those truths did not exist then. 112

Positing for Davidson cannot be an internalist stipulation

of conditions. Quine’s conception of a posit often seems to

presuppose a “fit” account of reference. The objects we posit are

112 Over-reaction to the consideration that the beings are posited is what leads people like Latour and Woolgar (1979) to say that bacteria didn’t exist before Pasteur, and that electrons didn’t exist before Thomson’s 1897 experiments. That would be idealism, which the present theory is not.



determined by the positing. The necessary truths about posited

objects are as it were stipulated. For Davidson, that cannot be

the correct account. On Davidson’s externalist view, we could be

wrong about important features of the objects we posit.

Extensions are not fixed by intrinsic features of our concepts.

For Davidson, to say that we posit an object of kind A is to say

that we utter things that it is reasonable to interpret as

referring to As, and that that interpretation is not forced by

the narrow choices made available by the intrinsic articulation

of nature.

Objects and properties are posits, but “posit” should not be

understood as active, in most cases. The predicates we use are

not generally chosen, but rather inherited from our ancestors

along the lines Quine (1969) laid out. On Davidsonian

interpretive grounds, most of what we say with our evolved

predicates is true. The fact that “tall” or “earth” do not reduce

to something precise and scientific doesn’t make sentences using

“tall” and “earth” false, it just makes them less useful for some

purposes. That is, given the Davidsonian account of truth, as

long as there are true attributions of the predicates, the

objects picked out by predicates in an imperfect predicate-system

are real. If “There are three chairs in the room” is true, then

there are three chairs in the room. So chairs are real, even

though there is no criterion in the terms of particle physics or

any other predicate family for when “is a chair” applies to a



collection of particles. Given that connection to the common

world is how language gets underway, the vast majority of the

things people try to talk about are there to be talked about.

IV) Alternative Conceptual Schemes?

The relative essentialism proposed here, which denies that

nature itself is divided into objects, does not mean that there

are alternative “conceptual schemes” in the sense of radically

different systems of predicates. Without the picture of something

given to be conceptualized in various ways, it is hard to make

sense of global alternatives even as abstract objects. If there

were sense to be made of such alternatives, they would not be

alternative predicates. Predicates are parts of languages and

languages are used by agents. “Is a language” is a predicate in

our system of predicates, along with “speaker,” “thinks,” and the

whole intentional framework. That intentional framework treats us

as special medium-sized objects to whom those predicates apply.

So, we think of ourselves as such medium-sized objects dealing

with medium-sized objects. So, for us, while there can be

substantial additions and modifications of our theories, when we

stop believing in the Tooth Fairy and start to wonder about

gravitons, global conceptual change, with mostly new objects and

mostly new predicates, abandoning the intentional framework, is

out of the question. If we did not have the intentional scheme,

we would not be language-users of any kind. The same applies to

any other language-users. So the idea of a global alternative



predicate system is incoherent. If the predicates were very

different, they wouldn’t be predicates.

So, even though the predicates and singular terms we posit

are posits, there are no alternatives for language-users, just

because there are no truths without language-users, and there are

no “users” without the intentional framework. But the lack of

coherently imaginable alternatives does not mean that the nature

conforms in itself to the positing we do.

Perhaps in some ontological sense there are an indefinitely

large number of objects in the same place and time span that one

of our objects inhabits. Even if that could be made sense of, the

vast majority of such additional objects are not possible posits

for language-users, and so not really possible posits, and so not

really possible alternative predicates, except in some wholly

abstract way.

V) Local variations

Davidson’s remark that sameness is always relative to a

predicate, while it does not countenance global variants of the

“conceptual scheme” kind, does allow that different people and

cultures can think differently.

Davidson’s account of meaning and language-learning

tolerates much variation within the framework of interpretable

predicates. A language-learner will learn that this object is a

cow. What “is a cow” is true of, though, the extension of the

term, is something that is not immediately apparent. Whether, for



instance, a capybara is in the extension of “is a cow” or not is

not evident for a while. Different possible language-communities,

that is, different possible variants of application of English

predicates, will have different conditions required for “is a

cow” to be true of an individual. Because truth requires the

possibility of falsehood, the correct interpretation of the

extensions of predicates is founded on, but not determined by,

the application-practices of the culture within which the

language-learner learns. This does not mean that the culture as a

whole cannot make mistakes. The members of that same culture may

also apply “cow” to deceased ancestors on the basis of religious

opinions. In that case, even though there is an appropriate set,

an interpreter (or skeptical language-learner later in life) may

reasonably hold that what everyone else thinks is just mistaken.

So, we certainly sometimes would be justified in

interpretations that, as it were, “divide up the world

differently.” This kind of alternative is indeed just re-sorting

of entities we have already posited. We allow that different

cultures can divide emotions differently. It is well known that

different cultures divide colors differently. But as a general

rule, assigning strange extensions to other speakers goes against

charity.113 The correct application of a predicate is determined

113 The term and concept “charity” is due to Neil Wilson (1959).



by the format, “`F’ is true of an object A if and only if A is

F.” 114

VI Multiple objects in the same place at the same time

We have already seen examples of overlapping objects in the

previous chapter. There are other possible predicates true of

objects that occupy the same space-time region filled by Fred the

frog that have different identity and sameness conditions. Fred’s

material components will survive Fred’s being squashed whereas

Fred will not. So, “This has to be an amphibian” will be true of

the frog, but not true of the squashed animal matter. And that

predicate, “is squashed animal matter” is only an easily

available, comprehensible alternative we already can formulate.

Familiar artificial objects, such as space-time worms and the

114 If there are no entities of which “F” is true, the term does not have an extension. How, then, can “phlogiston” be deemed not to refer at all? In interpretation, we take into account the connections of the term being interpreted with other terms of thespeaker’s language. The advocates of phlogiston held a number of views about when phlogiston was present and what de-phlogistonated air was. Since application behavior is relevant, but not constitutive of extension, we can reasonably conclude that this predicate does not apply to anything, even though thereis a kind that fits all of the users’ actual application-behavior. Such a conclusion rests on a substantial amount of agreement between us and the phlogiston-theorist. We have to understand the phlogiston-theorist by interpreting him as largelybelieving truths in order to see that he is mistaken in positing phlogiston.



Quinean ontological alternatives apparently supply more cases of

multiple objects occupying the same space and time.

Relative essentialism treats all such objects as equally

real, and as distinct objects. If two things differ in their

modal properties, they are different. This is the major deviation

of the relative essentialism presented above and Davidson’s

actual views. Davidson treats what I would regard as distinct

objects as two descriptions of the same object.

It is true that in many of cases of complete coincidence,

there seems to be only one thing. The distinct things I would

recognize are filling the same space at a given time, and perhaps

for their entire careers. So, the statue and the material, for

instance, may plausibly be thought to be one thing, as Aristotle

thought. Davidson seems to agree, even though much of his thought

would seem to indicate the opposite—if sameness is relative to a

predicate, then multiple descriptions of the same thing will at

least sometimes be descriptions of different things. Where this

monism seems to lead to very unintuitive results is when the

topic turns to events, as we will see in the next chapter.

Davidson actually is committed to distinct objects filling

the same space at the same time, if his remarks about emeroses

are taken to be an endorsement of “Emeroses are gred” as a

lawlike truth, and so about real objects. Emeroses are distinct

from emeralds, but overlap with all emeralds that have been

examined before 2020, as well as with very many other entities.



So, at least in some cases, Davidson countenances the possibility

that distinct objects occupy the same space at the same time.

VII How do we fit in?

There is a prima facie problem of how to think of thinking,

speaking, positing subjects (like us) existing in such a world.

We ourselves, the language-users and object-positers, are among

the objects we posit. Some account of self-conscious objects,

objects who are also subjects, is demanded. Such objects are

self-positers. This will sound paradoxical if “positing” is

identified with “making up” or “constructing a convention.” That

is not the thesis here. The present theory does not deny that

human beings are real and does not say that humans or any other

objects other than things like ground-rule doubles are

conventional. There is no obvious reason why, among the objects

that are posited, human beings are not themselves posited. In

fact, it would appear from animal studies that self-consciousness

is rather late among the positings, when the organism has a lot

of conceptual or proto-conceptual equipment. Nevertheless perhaps

a story is needed.

So the next few paragraphs are a sketch, details of which

would be filled out by virtually any respectable biological story

about how humans emerge from the slime, of how self-positers

could happen. The key is that “positing” need not be construed as

voluntary choice, but rather as evolved discrimination. Positing

or perhaps proto-positing, is not limited to self-conscious or



even conscious beings. The positing of self-conscious beings

should be seen as a special case of the more general phenomenon

that organisms can come to respond to their environment and to

each other. As far as I can see, at no point in the story do we

need to appeal to a given articulation of the world into beings

and properties. There is an articulation appealed to, of course,

namely the one we medium-sized organisms have evolved and the one

which includes us self-representing, self-positing positers.

In an environment with medium-sized objects interacting with

one another, it can happen (in the right environment, given DNA

or the like) that some medium-sized objects reproduce similar

medium-sized objects. Given differential success in such

reproduction, it can come to matter that such medium-sized

objects respond differentially to medium-sized objects. Patterns

of responses that increase the probability of reproduction, when

those patterns are themselves reproduced, lead to greater

probabilities of reproduction. It can turn out to be advantageous

for such organisms to respond to their own responses, and to

represent their responses. They can come to have a language that

embodies a medium-sized object predicate scheme. Of course, such

medium-sized objects will think of themselves as medium-sized

objects. Given that their entire language is built on medium-

sized objects and their properties, they could hardly do anything

else.



While such a scheme is indeed natural for such objects in

such specialized and rare environments, that does not mean that

nature selected that scheme. It means that, in this specialized

environment, this scheme works well enough to be selected. The

positing is automatic, up to the point at which some of these

organisms are so self-conscious that they are conscious of their

scheme and the possibilities of alternatives.

In this story, it is of course important that the items in

the story be real objects. At no point, though, is there anything

that requires a single domain of objects given by nature. Any

story of how humans come to be objects and to recognize objects

on a “realist,” unique- given- domain- of- beings story will do

just as well according to a story on which the objects are real,

but not naturally given, as feet are real but not naturally given

units of length.

The usefulness of the different posits for different

purposes has to be based on something. Clearly, there have to be

laws of nature that apply to any useful posited object. As

discussed in the last section of the previous chapter, laws of

nature need not presuppose any particular articulation of the

world into objects.

VIII Names and Demonstratives: Direct Reference



An advantage of the realist conception of given beings is

that realism coupled with a causal account of reference115

explains the apparent phenomenon that we are able to indicate

beings directly. We are able to refer to things by their names

without mentioning a kind to which they belong. A use of “Fred”

designates Fred the human. An explanation of the fact that we do

not have to specify “human” is that the human is the only

candidate object. Likewise when I hold a stiletto up and say

“This is what you need.” The stiletto is referred to, but nothing

is selecting it. It must be already selected. If there is no

privileged division in nature, it would seem that neither bare

naming nor bare demonstratives could succeed.

a) Names

An account of names that seems to accord with Davidson’s views

treats names, when used as names of individuals, as predicates

with a concealed demonstrative. This demonstrative itself only

demonstrates relative to some general term, filled-in in

interpretation. That is, just as the interpretation of “John is

tall” in Chapter 9 interprets the sentence as “John is tall for a

man/ basketball player/ Medieval monk,”116 so the interpretation

115 Nothing, by the way, prevents Davidson from saying that, by and large, whena name is learned by ostension, the referent of the name is the object which causes the joint response. A causal story does not require a unique articulation of reality into beings.116 The account sketched here descends from Peter Geach’s (1962). Kripke (1970), footnote 58, pp.115-116, criticizes Geach’s account because it seems to make the count-noun associated with the name part of the meaning of the name. The present account does not follow Geach in that respect.



of “Aristotle is intelligent” interprets the demonstrative

element of the name as relative to some count noun, in this case.

The general term relative to which the demonstrative is

interpreted is not part of the meaning.

Necessities about Aristotle are then necessities that rest on the

theory of organisms or men. Aristotle (this man) in

counterfactual situations is still Aristotle this man. While

there are other naming intentions that could have been present

when Aristotle was named (this batch of baby-flesh, this space-

time region, etc.) those intentions are biologically unlikely,

and the interpreter is unlikely to couple the demonstrative part

of the name with such entities. Organisms that exist only because

of reproduction and survive primarily by dealing with other

organisms must take account of organisms. Organisms whose young

require care will have evolved so as to identify their young as

continuing objects. While organisms labeling organisms is normal

and natural, that naturalness is biological, not metaphysical.

A name’s containing a demonstrative element that refers relative

to a general term means that the person using a name generally

has some count-noun in mind, even if the hearer cannot identify

which count-noun. We could think of the general term as filling

an argument place that is left unpronounced, but I am inclined to

spoken or too obvious to mention. The treatment would thus be

formally like the unspoken general term presupposed on many

accounts of the attributive construction of comparative



adjectives. The general term is not part of the “meaning” of the

name, which, as always, is given by the appropriate truth-

definition clause. “`Is John” is true of an entity if and only if

that entity is John.” “John is a man” is not analytic.

The general term can be of greater or less specificity, and have

more or fewer consequences. “Professor,” “human,” “protoplasm,”

“organism” “stuff” and “thing” would be a some alternatives.

“Thing” and “stuff” would qualify as a general terms at all only

as distinguished from “shadow,” “illusion” and perhaps

“hologram.” “Entity” on the present account would seem not to be

a general term that could strictly provide a referent. “Entity”

or “being” could be interpreted, perhaps, as a disjunction of the

kinds of entity that the speaker and hearer posit. So, “What the

heck is that?” with a pointing may succeed in referring, even

though the speaker has no intention to refer to what is there

under any general term.

Most of our intuitions about “initial baptism” work as well

on the present account as on accounts that presuppose a

privileged articulation. Let us consider a simplified naming

event, by Balboa at the western shore of Central America. Suppose

Balboa says, “Pacific,” gesturing towards the west, and then,

overcome by the rigors of his recent journey, collapses and dies.

Balboa’s men then split into two groups, one headed north and the

other south. Balboa’s speech act may have been a description, but

let us suppose he intended to be naming. Balboa himself may have



had in mind the Pacific Ocean, San Miguel Bay, or the Isla del

Rey among plausible objects intuitively deserving of a name. His

men, though, never find out what object he has in mind. When

they, with the intention of following their former leader, use

“Pacific” as a name, they are interpreting by assigning a coupled

count-noun. The two groups may interpret differently, and maps

drawn by one group may label San Miguel Bay “Pacific Bay” while

maps drawn by the other group label Isla del Ray “Pacific

Island.” Thus, even if Balboa himself had had “ocean” in mind

during the baptism, that is not part of the meaning of the name.

The interpretation of what kind of thing it is that the name

names determines what the necessary truths about the item are,

according to the interpreter. The authority, though, does not

necessarily rest with the baptizer. If Balboa intended “island”

and his followers interpret “bay”, “Pacific (bay)” is a

misinterpretation of Balboa, but a correct interpretation of the

name in the idiolect of the followers.117

b) Demonstratives and Pure Indexicals

Balboa in the above example of naming, used a demonstrative,

“this” or “that” coupled with a general term. How do “this” and

“that” work? Simple demonstratives, on the present account, refer

only relative to a general term. The model is Jeffrey King’s

117 The speaker’s intention does not necessarily determine that anything is named, even by the speaker himself. If Balboa mistakes the expanse of water for a mirage on a great desert, and intends to name that desert, he has not succeeded.



(2001) account of complex demonstratives. Balboa at the edge of

the Pacific saying “This is rather attractive” might have had the

bay, the beach, water in general, or the ocean in mind, among

other objects before him. Roughly, his demonstration is

successful when a case of the intended count-noun is before him.

Demonstratives locate the case of a general term relative to

the speaker’s location and the time of the utterance. “This

(frog)” will be “the frog here now that I’m pointing to,”

roughly. Given that the reference is to the case of the count-

noun with the indicated relation to the speaker, the continued

reference to that individual in counterfactual situations (the

firmness of the demonstration) can be explained by the theory

associated with the general term.118 The de re necessities that

are true of the kind coupled with the demonstrative determine

what counterfactuals are about that object. But don’t we need

direct reference to the time, place and speaker? Indeed.

The one place where there is direct reference is with words

like “here”, “now” and “I”—the indexicals, which directly refer

to the place, the time or the speaker of the utterance or text.

The tenses also locate events by reference to the speaker’s

present. Why don’t we need count nouns here? Different

explanations are required. In the case of space and time (“here”

and “now”) no one supposes that there is a natural segmentation 118 In some cases, it can be indeterminable exactly what “that” refers to. After an enjoyable time, your partner says, “Let’s do that again.” Neither younor your partner may know exactly what event-type is being referred to.



into units. While there are count-nouns, there are no illusions

of privileged count-nouns. Anyone knows that, while there are feet

and meters, so that 100 meters is roughly 300 feet, nature itself

does not segment space into either meters or feet. Likewise,

everyone knows that years, seconds, hours, et alia, are, while

real, not naturally selected units. So, the vague, contextually

specified locations in time and space that “here” and “now” refer

to don’t need a general term because there is no illusion of

natural articulation into units. The reference of “now” can range

from billions of years (“Now that eukaryotes are on the scene…”)

to microseconds. Likewise with “here.” We talk about our local

cluster of galaxies as well as a place on a DNA chain.

How can a speaker refer to herself in order to locate “here”

and “now” without using a count noun? Given the she is a speaker,

she can only take herself to be an agent. So the count-noun for

an utterance of “me” or “I” is an agent, a medium-sized object

conscious of itself as such. There are of course indefinitely

many entities in the place from which the utterance emerges.

However, only the agent is speaking.

d) Why does there seem to be direct reference?

The counter-examples to absolute essentialism in the

previous chapter focused on stuffs. Stuffs are easier to

construct counterexamples for because with stuffs, there are

alternatives available for us to candidate metaphysically natural

kinds. As discussed in Chapter 2, this is not generally the case



with count nouns that determine individuals. Individuals such as

Aristotle and R2D2 are intuitively more plausible examples of

entities that have metaphysically significant de re properties

because, as it were, there is nothing else we can conceive of

that they could be. Van Inwagen (1990) and Aristotle are correct

to take organisms as the best case of Beings. An organism’s

organized transient relation to its matter means that, for

medium-sized119 organisms at least, individuation can only

practically be done in one way. An organism has few intuitive

alternatives. Organisms of roughly our size yield an articulation

that would be difficult for organisms of our size to track as any

other kind of medium-sized objects otherwise. Furthermore,

hundreds of millions of years of evolution have made organisms

salient to us. So any alternatives, other than “this plant stuff”

or “this meat” leave us with nothing recognizable at all. The

intuition that people and frogs are metaphysically special is

explained by the unavailability for us of alternative medium-sized

objects, but that’s just us.

Chapter 4: Kinds of Events

One important generalization of relative essentialism is

that it applies to all sorts of beings, not just to physical

objects and stuffs. In particular, the view applies to events.

“Is an event” is a general category of being, not a particular

119 As discussed in Chapter 2, fungi that spread below ground over hundreds of feet and Portugese Man O’ Wars may be intuitive counterexamples.



kind of entity. Just as there is no single formula that specifies

identity conditions for all physical objects, so there is no

criterion for identity of events as such. Furthermore, just as

there can be multiple physical objects in the same space at a

time, so there can be multiple events with the same causes and

same effects. Relative essentialism can allow that we posit many

sorts of events with different identity conditions. Distinct

events may coincide by having the same causes and effects, just

as distinct physical objects may coincide by occupying the same

space at the same time.

This chapter will make this generalization plausible. I will

first give some examples which make the relative essentialist

view of events plausible, then discuss the application of this

view to Davidson’s anomalous monism. My view departs in several

respects both from Davidson’s view of events and also from his

anomalous monism. Rather than an anomalous monism my view of the

mind-body relationship is an innocuous dualism.

I) events in Davidsonian semantics

Davidson’s brilliant account (1967b) of adverbial

modification quantified over events. Here is a sketch of the

problem it solved. “John buttered the toast with a knife at

midnight” clearly entails “John buttered the toast with

something,” “John did something with a knife,” and a variety of

other things. If the prepositional clauses are treated as



arguments of a single buttering-predicate, there would be two

choices:

a) An all-purpose buttering predicate would have to have every

possible variety of prepositional phrase, so that “John buttered

the bread” would actually use the same six-place predicate as in

“John buttered the bread with butter, with a knife, in the

closet, after midnight” but with three of the places

existentially quantified. The inference from “John buttered the

bread with fresh butter, with a knife, in the closet, after

midnight” to “John buttered the bread,” would then be existential

generalization, from B(John, bread, butter, knife, closet,

midnight) to ExEyEzEw(John, bread,x,y,z,w).

b) There would be several buttering predicates with a variety of

arguments-places for instruments, objects, locations, and so

forth. For each pair of such predicates, there would be special

rules connecting them. For instance, the four-place predicate “x

buttered y with z in w” would need connection to the three-place-

predicate “x buttered in w” by a generalization,

\/x\/y\/z\/w(Fxyzw ->Gxyzw). There would be nothing formal

connecting the buttering predicates, which would be logically

homonyms.

Neither of these solutions is satisfactory.120 One problem

with a) is that there seems to be no obvious limit to the number

120 Kenny (1963) called this problem the “problem of variable polyadicity.”



of prepositional clauses that could be attached so setting a

number of predicate-places adequate for every possible buttering-

complex looks hopeless. It is difficult to come up with a formula

for adding indefinitely more places, however. So a better

argument is the following: “John lectured through a megaphone”

entails “John lectured.” But lecturing need not be through

anything. So, the inference from “John lectured” to “There is

something through which John lectured” is not valid. So, all-

purpose predicates with every argument place that could be needed

will not work, since those argument places can generate incorrect

consequences for many verbs.

As for b), there are too many connections required among the

sets of predicates to make it plausible that we learn special

rules for each of them. Between any two combinations, for

instance, “John buttered something with a knife in the closet”

and each of “John buttered something” and “John did something in

the closet” and “John buttered something with a knife” there will

be special information required that if the first is true, so is

the other. Obviously, though, we somehow have all such

connections figured out in advance of confronting any particular

such inference. If we understand “John butters,” we understand

“John butters in the closet,” and likewise for the other

connections. These inferences cannot be matters of information

about the special subject matters of buttering, knives, and



toast, so they must be truth-preserving in virtue of their form.

So the connections must be formal.

Davidson’s insight is that if we treat events as objects

about which a number of things can be said, all these inferences

fall out as simple cases of logical consequence. The theory is

simple and extensional. The logical form assigned to “John

buttered the toast with a knife at midnight” is “Ee(A(j,e) /\

O(t,e) /\ I(k,e) /\ T(m,e)),” and the form assigned to “John

buttered the toast with something” , is “ExEe(A(j,e) /\ I(x,e).”

So, the form assigned to “If John buttered the toast with a knife

at midnight, then John buttered the toast with something” is

“Ee(A(j,e) /\ O(t,e) /\ I(k,e) /\ T(m,e)) -> ExEe(A(j,e) /\

I(x,e)),” a logical truth. Each of the prepositional clauses is

treated as just another conjunct in the conjunction. These

connections, then, all turn out to be logical truths of the form

((A /\ B) ->A) or (Fa ->ExFx). Thus, the problem of variable

polyadicity is solved in a stroke. In addition, predicates with

expletive subjects are given a natural logical form. “It is

raining” is just “Ex(Rx).” It will be true of buttering that it

must be done with something, and not true of lecturing that it

must be done through something. These truths will indeed be

pieces of information special to predicates, and support for the

inference from “John buttered the toast” to “John buttered the

toast with something” will come from such special information

about buttering.



Variations on Davidson’s idea have come to be the main

theory of adverbs among linguists.121 This quantified conjunction

is a completely extensional construction. For Davidson, events

are just a kind of object, and quantification over them makes

light work of accounting for truth-conditional connections among

adverbial constructions.

Causal sentences are another application of the idea of

events as objects. Apparent connectives such as “because” in “The

thunderstorm formed because two air masses collided” cannot be

genuine sentence connectives, since the sentence seems to be

referentially transparent.122 An event analysis treats “because” 121 See Kratzer (forthcoming), Schein (1993), and Larson (1995).

122 By the Slingshot argument, any referentially transparent sentence-containing context is truth-functional. Briefly, following Quine (1953b), page 161-162, suppose we have a sentence- connective, perhaps B for “because”, as in “John hit Fred because John was mad.” Suppose further that the context created by the connective permits substitution of co-referring expressions, and that set-theoretically equivalent sentences have the same referent. That is, suppose that if Bpq and r results from q by substitution of some co-referring expression within q, then Bpr. Then the proof that this makes the connective truth-functional, so that if Bpq and s has the same truth-value as q, then Bps, is as follows:(1) Bpq

(2) Bp (x|{x=x /\ q}= x|{x=x}) [from (1) by set-theoretic equivalenceto q](3) Bp (x|{x=x /\ s}= x|{x=x}) [from (2) by substitution of the co-referential expression,

x|{x=x /\ s}for x|{x=x /\ p}](4) Bps [by set-theoretical equivalence of s to the second argument of Bp (x|{x=x /\ s}= x|{x=x}) in (3)



as a two-place relation between events, so that the above

sentence would become, “ExEy(x is the formation of the

thunderstorm /\ y is the collision of two air masses /\ x caused

y).”

The least important aspect of the Davidsonian strategy, it

seems to me, is the limitation to quantifying over events

exclusively. Quantification over something which can have

multiple features is what makes this strategy useful and have

wide application.123 As I will argue in this chapter and in

Chapter 6, states of things are different from events that happen

to things. Some verbs certainly appear to be predications about

states. I will argue that facts are states of the world. Facts

appear to be causes some of the time. Although, given the

different views of ontology Chapter 6 will present, I have a

rather different view about causal sentences and event sentences

from Bennett (1988), his discussion of nominals in chapter 1 of

that book should convince any theorist that there is more to

adverbial modification and nominals than events.

The important part of the Davidsonian idea is that of

quantifying over a kind of object about which many things can be

said. There are numerous kinds of objects with which this can be

done. Allowing quantification over a variety of different objects

can provide an adequate account of adverbial modification.

123 In Chapter 9, a completely different application of the same strategy will

be applied to comparative adjectives.



Briefly, different verbs and different verbal tenses are

different, and talk about different entities—some states, some

events. “I have been a university employee since 1970” does not

report an event, but something else. With the relaxed view about

ontology that relative essentialism and the constructions of

Chapter 6 make possible, there is no reason whatsoever to try to

keep one’s ontology simple. Every kind of object is real. If some

of the objects we posit are redundant, that does not

II) Different coinciding events

Given that events are entities, as Davidson’s analyses make

very plausible, we can ask how general the category “event” is.

On my view, “is an event” is a predicate of the same level of

generality as “is a physical object.” Being an event is not

something that can be expected to have a general specification in

terms of necessary and sufficient conditions for A being the same

event as B, any more than there is a general specification of the

conditions for A being the same physical object as B.

There are lots of kinds of physical objects that we posit,

and indefinitely many more kinds that we could posit. Physical

objects range from micro-particles to automobiles to organisms to

galaxies. For each kind of physical object, there are conditions

for when an object is of that kind and for when there is one or

more than one. These conditions may be different for many

different kinds of physical object. Being the same shadow is

different from being the same flock and both are different from



being the same oxygen atom. Sometimes the conditions are very

vague, as with medium-sized physical objects; other times they

are quite specific, as with micro-particles. Tracking quantities

of stuff which overlap with individuals is different from

tracking the individuals. Different kinds of composites have

different identity conditions relative to their components. Cars

can have their fluids changed while remaining the same, whereas

bottles of wine cannot. It would be very difficult to come up

with a criterion stating necessary and sufficient conditions for

being the same physical object which would accord with our

positings.

Exactly the same is true of kinds of events. The identity-

conditions for a Mass are different from the identity conditions

for a wedding. The conditions for events A and B being parts of

the same war are different from the conditions for events C and D

being parts of the same battle. Just as in the case of physical

objects, smaller events are typically parts of larger events. A

transubstantiation of some wafers and wine is part of a Mass.

Just as the micro-macro contrast among kinds of objects and the

contrast between mereological sums of micro-particles and medium-

sized objects yields different but sometimes-coinciding physical

objects, so the contrast between sums of micro-events and macro-

events likewise yields different but sometimes coinciding events.

Events are entities which have modal features. My trip to

Ohio last year would have taken much longer if I had taken I-80



instead of the Pennsylvania Turnpike. If we think of that trip as

consisting of a sequence of sub-trips from mile-post to mile-

post, there is the mereological sum of those subtrips. That

mereological sum coincides exactly with my trip to Ohio from I-84

t0 the Merritt, to the Cross Bronx Expressway, to the Tappan Zee,

to I-278. To I 78 to I 81 to I 76 to I 70. However, since

mereological sums are defined by their elements, that sum could

not have included events taking place in Scranton and Wilkes-

Barre, on I-81 along the I-80 route. But my trip to Ohio via I-80

would have passed through Scranton and Wilkes-Barre.

Now of course it is possible to say that this apparent

singular term, “my trip to Ohio” in fact doesn’t refer to a

single entity that could have been otherwise at all, since it is

a definite description, like “the President of the US.” Just as

the President of the US is not an entity that could have had

another person as occupant, it could be argued, so my trip to

Ohio is not an entity that could have consisted of other roads

taken and longer rest stops. Those possibilities are really

alternative trips, just as John Cain is an alternative President,

rather than something this President could have been. So it could

be argued that in fact what we are contemplating are alternative

trips to Ohio, not other ways this trip to Ohio could have been.

So consider an example with a named event which has a

stronger intuitive claim to being something which could have been

different. The Battle of Stalingrad was a major (for humans)



large-scale event. “The Battle of Stalingrad” is not a definite

description, since there had been battles there before, notably

after the revolution, and since, while the original focus of the

fighting was in Stalingrad, much of the fighting took place

outside the city, it is not a definite description which refers

to a battle at that city. In addition to the event called “The

Battle of Stalingrad” there also occurred the mereological sum of

the one-second-interval-length events124 that were parts of the

battle125 and took place in the region around and including

Stalingrad in the winter of 1942-43. That mereological sum, like

Fred’s space-time worm, was fixed by the components it in fact

had. Coinciding with but not identical with that mereological

sum, as it happens, is the Battle of Stalingrad. The Battle

could have gone very differently, in detail and in large.

Different explosions could have taken place, different tactical

decisions might have been made, and so forth. The Battle of

124 The component events can be taken to be relative to any “level.” The firing

of a cylinder in a T-34 motor itself, relative to the individual molecules’

motions, could have been different. A particular charge, relative to the

footsteps of the personnel involved could have been different. There may be no

“given” level of basic events from which all others are constructed, but worms

can be made from any kinds of events as long as the mereological sum of them

has the same causes and the same effects as the Battle could have been the

same event, while different.125 “Part of the battle” is somewhat vague. We do not want to include as part

of the Battle of Stalingrad a snowflake drifting down from the sky.



Stalingrad, having explosions and decisions as parts, was a mixed

compound of mental and physical events.

The worm and the actual Battle have many coinciding parts.

Particular chemical events involving nitrate molecules during

explosions, which could not have gone otherwise, may be parts of

both events. Certainly particular firings of cylinders in a

particular T-34 tank are at least coinciding parts of both

entities. Particular decisions on the part of participants might

be shared. However, some parts of the Battle, particular sub-

battles, are parts only of the Battle, not of the worm. Here is

why:

The Battle of Stalingrad was very large, and had components

that were themselves on a scale that would be vast by the

standards of the Western front. Operatsiya Uran, the Russian

encirclement of Paulhus’ 6th army, involved an estimated

1,100,000 personnel, 804 tanks, 13,400 artillery pieces and over

1,000 aircraft. This component of the battle, which took place

well outside the city, might have gone differently. So, while the

coinciding Operatsiya Uran worm is part of the Battle worm, and

many of its parts are also parts of the Battle, the event

Operatsiya Uran is not part of the Battle worm. It could have

been otherwise, but no part of the Battle worm126 could have been

otherwise.

126 At any level of analysis other than “major sub-operation,” the worm will

not have this encirclement as a component.



The Battle of Stalingrad, while it is in a sense “nothing

over and above” the space-time worm consisting of the sum of the

sum of the events at each of its moments, has different parts

from that sum because it and its parts have different modal

features. Battles are a kind of event, with their own, admittedly

vague, existence and identity conditions.

Battles are among the few kinds of events that typically

have names. 9/11, Ali-Frazier, wars, the Exodus, the

Resurrection, the Crash of 1929, the South Sea Bubble and others

terms for events are clearly names rather than descriptions.

Whatever kinds of events these were are therefore nameable

entities. So, at least catastrophes, boxing matches, migrations,

miracles, and economic disasters are events such that that very

event could have gone otherwise. Counterfactuals about how other

kinds of events could have occurred should lead us to think that

all sorts of kinds of events have their own identity conditions,

and so are the sorts of things that have to be some ways, but

could have been different in other ways from how they in fact

were. Many events could have been otherwise, and these are very

plausibly alternative ways those very events could have gone.

Further examples may assist the reader in accepting the idea

that events are substance-like in this respect. Consider “John’s

wedding would have been more amusing if he had worn tennis shoes

and couldn’t effectively stomp the glasses.” We are not

speculating about how some other wedding might have gone, but



about this wedding. Or, consider a basketball game in which a

last-second three-pointer is attempted, but misses, resulting in

a loss by two points. “That game” in “We would have won that game

if Amare had connected on that last three-pointer” obviously

refers to that game, not to another game that would have turned

out to have been played instead.

Kinds of events have essential features. It is impossible

to score a touchdown in a basketball game. Weddings require more

than one organism participating. Earthquakes cannot take place in

empty space. Having essential features, however, does not mean

having exact essential features, any more than it does for

physical objects. Just as it is vague exactly what the limits are

for being a chair when one is dismantled chip by chip or being

the same boat after replacement of planks, so it may be

indeterminable what things could be different about an event

while it was the same event. Some differences, of course, are

clear. If I in fact marry Pam, but might have recruited Joan at

the last minute when Pam came to her senses and realized what she

was getting into, that alteration in the plans would bring about

a different wedding, rather than being another way this wedding

could have gone. Even though I would keep the same venue,

caterer, and string quartet, it would seem that another wedding

had taken place, rather than that one wedding had gone

differently.



A sorites can be constructed for weddings. There are often

slight delays getting the ceremony under way. If there had been a

delay of a minute, that would seem to be a case where this wedding

which in fact started at 2:00pm, instead started at 2:01pm. But a

wedding that is delayed for a month would seem to be a different

wedding. Clearly, “The wedding of Sam and Pam might have happened

in the late nineties, after they both had had disappointing first

marriages” is talking about a different wedding than the event in

1966, when they were both packing up their stuff after

graduation. How much difference, and what kinds of difference

there must be between one scenario and another for there to be a

different wedding rather than another wedding of the same people

is indeterminate. As we mentioned before, and will discuss in

detail in the chapter on sorites arguments, we should drop

Quine’s dictum “no entity without identity” unless we are willing

to give up the existence of the objects of ordinary life we

cherish—our cars, houses, spouses, and children. With Quine’s

dictum, we would be left with mathematical objects and micro-

particles, at best. For Quine and St. Anthony with their tastes

for desert landscapes this may be acceptable, but for someone

reluctant to abandon organisms and fast cars it is not.

Multiple events can coincide in the way that multiple

physical objects do. If Natty Bumppo alerts the Iroquois by

stepping on a twig, Natty performs two events by one physical

movement, on my account. The two events coincide, in that they



have the same causes and same effects, but that same stepping, if

the Iroquois had been out of earshot, would not have been an

alerting. A condition of being an alerting is that someone is

alerted, but that is not a condition of being a twig-stepping.

So, in a sense, Natty doesn’t have to do two things since doing

one is also doing the other, because the things coincide.

Notice that coinciding for events is sufficient for having

the same causes and the same effects, just as coinciding is

sufficient for two physical objects to be involved in the same

causes and the same effects. Just as whatever happens to Fred

happens to his space-time worm, so whatever brings about some

part of my trip to Ohio brings about the event-worm corresponding

to it as well.

III Analyses of adverbials with multiple coinciding events

Recognizing that pulling a trigger is a different event from

murdering Fred even though there is only one displacement of a

body in space that is both events often complicates analysis.

Single events become multiple events which coincide. Sometimes

adverbial modifiers are true of one coinciding event but not the

other. The great advantage is that apparent counter-examples to

the whole idea of handling adverbial modification by quantifying

over something are handled smoothly.

The simplest cases which illustrate the difference between

my view of events and Davidson’s are sentences using “by.” “By”

phrases generally indicate events which coincide but are not



identical. We will use such constructions below in our

discussion of Davidson (1970). Consider an example:

“John deeply offended Bill by slowly winking.” On the Davidsonian

analysis, this is quantification over a single event which has

multiple features. Simplifying somewhat by ignoring the

difficulty with the comparative adjectives “deep” and “slow,” the

analysis would be: “There is an event x of which John is the

agent, x is an offending of Bill, x is deep, x is a winking, and

x is slow.” One drawback of this analysis, and a long-standing

criticism, is that the same event can be slow and not slow. Bill

may have quickly taken offence while the wink was slow.

On the multiple coinciding events picture, there are two

events which coincide, one of which is a slow winking, the other

of which is a deep offending of Bill, and the agent of both which

is John. So, in symbols, “ExEy (Agent (x,John)/\Agent(y,John)/\

Coincides (x,y) /\ Winking x /\ Slow x /\ Offending y /\ Deep

y /\ Patient (y,Bill)). “Coinciding” captures the idea that

under one description, a purely physical one whose principle of

individuation is “having the same causes and effects,” John did

only one thing. To regard coinciding events as the same is among

other things to take the physical, non-intentional

characterizations of events as the fundamental items of which

every other characterization is really just another description.

That seems to me to be what Davidson, against his own better



judgment elsewhere, for instance in Davidson (1974), is doing in

identifying disparate events.

Notice that the analyses I propose are completely

extensional, although that characterization begs the question, to

a degree. In proposing that there are multiple events rather than

multiple descriptions of the same event, we are preserving

extensionality by multiplying entities. This issue can get

confused, because some predicates of events are intentional. This

is because acting is the propositional attitude “making true”.127

The things that are made true are propositions, which I argue in

Chapter 6 are essentially linguistic.

If John makes it true that he is married to Sheila, then

when that action takes place he is married to Sheila. The

inference from “John made it true that he is married to Sheila”

to “John was married to Sheila” need not be part of logical form,

but can rather be information special to the predicate “makes 127 The following account is adapted from various of Davidson’s remarks. In his

1967 class (see Wheeler 2012), discussing propositional attitudes, he remarked

that you could conceive of the relationship between desire and action as like

the relationship between belief and knowledge. Knowledge has, as it were,

internal and external truth-conditions. Knowledge that p is belief that p that

comes about in the right way from p being true. Doing (intentionally) likewise

has internal and external truth conditions. Doing p is a desire that p

bringing about p being true in the right way. Briefly, we can think of

intentional action as the propositional attitude “makes-true.” The items that

are made true are, as I will argue in Chapter 6, propositions construed as

things said.



true,” if we decide that verbs of action have the formal

component “makes true.” The same would apply to “knows”

“realizes” and other factives. In Chapter 6 we discuss factives

and the facts that they talk about. Facts will be innocuous.

Davidson is correct to note that the intentional is not a

class of events128.The intentional classifies actions differently

from the classification we get if we treat different events as

different. When John intentionally marries Sheila, that is the

same event, a marrying, as John’s marrying the woman who would

cause him more grief than any other human being. Presumably John

did not intentionally marry the woman who would make him so

miserable. What is intentional and what is not is a division

within kinds of events. Two distinct but coinciding events can be

done intentionally. John may marry Sheila intentionally and

acquire a half-interest in Montague Farm intentionally, that

being the motive for this disastrous marriage. But acquiring a

half-interest in Montague Farm could have been done by a simple

cash transaction, where marrying Sheila could not, given the

American legal system.

IV) The mental, the physical, and innocuous dualism

1) Davidson’s anomalous monism

Davidson famously argued for a view, anomalous monism, that

mental and physical events are identical, but characterized under

predicate systems such that there could be no general systematic

128 Davidson (1978).



law-like relation between them. To call a mental event “mental”

means that that event is characterized under a mental

description. That same event also, Davidson argues, has a

physical description, and so is a physical event. There being no

law-like relations between the mental and the physical is a

feature, not of the events themselves, but of the events as

described.

In outline form, Davidson’s argument for the above view is

as follows:

1) There are no law-like relations between mental events and

physical events.

2) There are causal relations between mental events and physical

events.

3) Causal relations between a pair of events require that there

be a description of both events such that a conditional with

those two events as antecedent and consequent is an instance of

an exceptionless law.

4) Only physical laws can cover all items in the world, so the

only exceptionless laws are laws about events under a physical

description.

5) Mental events are identical with physical events.

Let us first establish premise 1): The lack of law-like

relation between these systems of predicates is due to there

being different sets of constraints structuring the assignment of

physical predicates and mental predicates. The constraints on



mental predicates are what I have called above “the intentional

scheme.” Principles of interpretation, for instance the

maximization principles discussed in Chapter 1, determine which

mental predicates get applied. A proposed interpretation of a

person as deciding to bring about a simple contradiction is

almost ruled out as the possible content of a decision. The

application of physical predicates, on the other hand, is

constrained by principles like the transitivity of physical

comparatives. If we measure three events, and find that the first

measures as having a higher energy output than the second, the

second measures as having a higher energy output than the third,

but the third measures as having a higher energy output than the

first, we do not conclude that transitivity of “has a higher

energy output” has broken down but rather that something is wrong

with our measurements. Our insisting on maximizing consistency of

belief, on the one hand and maintaining transitivity of “has a

higher energy output than” on the other hand, constrain our

applications of mental and physical predicates respectively.

Since these constraints are distinct and unrelated, there is no

reason whatsoever to suppose that they will sort events into

systematically related sets of groups. Thus there will be no

physical predicates that match any mental predicate in a way that

would allow a conditional with a physical predicate as antecedent

and the physical predicate coextensive with the mental predicate

which conditional is an instance of a physical law.



Of course, given the finitude of human existence, there is a

physical predicate of brain-events that exactly matches the

extension of “decides to buy a 2007 Chevrolet Impala.” That

physical predicate is a very long disjunction of the brain

states, physically described, of all those over human history

while they were deciding to buy that car. However, there is

nothing about this weird predicate that would tell anyone whether

a new brain state was a decision to buy a 2007 Impala or not. The

identity does not support counterfactuals. If there had been

another person deciding to buy a 2007 Impala, there is no reason

to think that the predicate would be true of that person.

Premises 2) and 3) connect this result with the conclusion.

Not being a Berkeleyan, Davidson takes it as obvious that some

physical events cause mental events and that some mental events

cause physical events. Decisions cause physical events and

perception and other physical impacts cause mental events. The

thesis that causal relations require exceptionless laws can be

defended in many ways, but may be stronger than Davidson needs. A

weaker thesis is that the “real cause” of a phenomenon is the

explanation that uses the strongest laws. The strength of a law

is, given an instance of the law, the conditional probability of

the consequent of the instance relative to the truth of its

antecedent. The limit of such conditional probabilities is one,

the exceptionless laws.



Now, it may happen, and apparently does happen, that the

exceptionless laws do not determine any particular phenomena, but

rather determine exceptionless probability-distributions for such

phenomena. That seems to be the case with quantum mechanics. When

a particle’s position is predicted probabilistically by the

exceptionless law, we need not say that its position is uncaused.

My view, articulated below, is that “cause” is really a

predicate applicable to explanation, which is a relation between

sentences, or descriptions of events. We don’t find the term

“cause” for instance, in the equations of physics.129 What we find

is equations which say how things are and have to be—Laws of

Nature. The laws of quantum physics, and the laws of contemporary

physics generally, are the strongest laws we have. The best

explanations of everything that happens, in principle, would be

explanations in terms of the predicate system of the best

physics. As we noticed earlier, and will dwell on later, the

sorites argument shows that taking this to be a reason to rule

out other causes and other beings rules out not only the medium-

sized objects of our world, but ourselves as well.

Within and between different families of predicates there

are stronger and weaker laws. Physical laws about medium-sized

objects are weak relative to micro-particle laws. Stronger laws

are those that give better explanations. One explanation is

129 This is a point made repeatedly in UCONN departmental colloquia by Scott

Lehmann.



better than another for a given event just in case the laws that

explanation appeals to assign the event a higher probability.

Premise 4), the ubiquity of the physical predicate system,

is an unargued-for premise. The physical predicate system has a

generality that the intentional system lacks. Every event in the

world has a physical description, but the vast majority of truths

about the world are not truths about agents and their mental

events and states, so are not expressible in mental terms, except

trivially.130 The intentional system’s event predicates are only

true of rational agents and their effects. Mental events, because

they are sometimes caused by and cause events that are out of

their domain, necessarily have weaker laws. However close the

relation between one thought and another, an intervening

explosion in the vicinity of the thinker will break that

connection.

So, the intentional system’s predicates have weaker laws

than the physical system’s predicates. By premises 3) and 4), any

causal relations between mental and physical events must be

instances of exceptionless laws. Thus, the mental events must

also have physical descriptions, under which a conditional with

the causing item being the antecedent and the caused item the

consequent is an instance of an exceptionless law. So mental

events are physical events.

130 “If Fred believes p, Fred is right” makes everything mental.



It should be clear how my view differs. If the condition for

causal relations is that if A causes B, then some description of

some event coinciding with A is the antecedent of a conditional

of which some description of some event coinciding with B is the

consequent is an instance of a law, then, while almost everything

Davidson argues is true, the conclusion about identity does not

follow. Briefly, if the requirement for causal relations is

coincidence rather than identity, then are really two kinds of

events here. Dualism is correct.

Now, it is difficult to read Davidson’s (1970) not get the

impression that, at bottom, Davidson thinks that what is really

going on are the physical events, and that the mental events are

really just a necessary but less ontologically illuminating

perspective on them. That is, Davidson’s view sounds very much

like that of Daniel Dennett (1968) that the intentional scheme is

just another way of talking about the physical world.

One way to see that there is a real problem in Davidson’s

view is to bear in mind his remark that sameness is always

relative to predicates. If mental events are just physical events

differently described, Davidson must suppose that there is a

domain of events which can be described either with mental or

with physical predicates. But some predicates or others must

provide the sameness for such a domain of entities that underlie

both the mental and the physical. So, the picture in Davidson

(1970) is very much like the picture he criticizes in Davidson



(1974). There Davidson argues that it is a mistake, the “third

dogma of empiricism,” to think of the world as a given domain of

objects to be sorted by predicate systems. If that is the case,

then it is another instance of the third dogma to think of the

mental and the physical as two sets of descriptions either of

physical events or of neutral “given” events.

On the neo-Davidsonian view I propose, Davidson should have

said something like the above about Tom’s decision to marry

Allison in relation to coinciding brain events as we say about

Tom himself in relation to the coinciding worm. That very

deciding event, that single entity, could have been slightly

longer, could have involved slightly different guesses about

Allison’s wealth, and so forth. These possibilities are not

possibilities of the brain event that coincides with the

decision. Brain events have different principles of individuation

and persistence. That is, different modal predicates are true of

brain events and mental events. So, the mental event is not

identical with the brain event, even though mental events

supervene on physical events.

From the present perspective, Davidson’s arguments against

psycho-physical laws are converted to arguments that mental

events have different modal features and different components

from the brain events that coincide with them. So mental events

are distinct events. Every part of a mental event is a mental

event, just as every part of a physical event is a physical



event. Tom’s deciding to marry Allison has components such as his

weighing her wealth, weighing her beauty, and so forth, each of

which could have been longer, could have involved other thoughts

about her monopolizing the bathroom in the morning that did not

occur to him but could have. Given that the conditions for when

the corresponding brain events would have been the same

individual brain events are different from the conditions for

when a given individual decision or realization, or components

thereof, would have been the same or a different decision or

realization, as Davidson’s arguments about the impossibility of

psycho-physical laws seem to show, brain events are different

from physical events.

So my proposal is a dualism. But it is a dualism of the same

sort as the dualism of statues and lumps, or battles and battle-

worms. The dualism is trivial rather than metaphysical. Non-

trivial monism and dualism only make sense given a monistic

conception of ontology. If beings are posits which allow us to

think about the world, then the dualism of the mental and the

physical is innocuous.

So perhaps the question arises as to why Davidson did not in

fact adopt this view. My hypothesis is that Davidson retains some

Quinean scientism in endorsing anomalous monism rather than

innocuous dualism. Everything he says in “On the Very Idea of a

Conceptual Scheme” (Davidson 1974) should lead him to reject the

metaphysical primacy of the physical. His idea that any array of



objects already presupposes conceptualization might have

indicated to him that he was committed to a version relative

essentialism.

d) The Grain of Events

Davidson’s analyses of adverbials and causal sentences in

terms of events, as we noted, treats events as entities to be

quantified over in extensional contexts. However, there are some

difficulties with the idea that events are always the same

entities that are just re-described in these analyses. When an

agent does something by doing something else, and there is only a

single action, there is also, according to Davidson, a single

event. The sequence of “by” clauses in “Susan murdered Fred by

firing the gun by pulling the trigger” must refer to a single

event--when Susan pulls the trigger, she also fires the gun and

murders Fred. Davidson is forced to say that Susan’s pulling the

trigger is the same event as her murdering Fred. But the trigger-

pulling might have been non-fatal, whereas the murdering could

not have been a murdering while being non-fatal. Also, it seems

that the murdering takes more time than the pulling of the

trigger, and is not completed until Fred dies, whereas pulling

the trigger is over in a fraction of a second. In an already-used

example, Natty Bumppo both steps on a twig and alerts the

Iroquois. It seems clear that these are two different things done

by the same physical movement, neither of them intentional, in

this case.



Such considerations lead to theories according to which

events are entities like propositions and properties, and have

“grain,” either fine or coarse. Various constructions have been

proposed out of properties, times, and subjects to be what

properties really are. The present account accommodates

intuitions that alerting is a different event from stepping on a

twig while remaining extensional. When we give the logical form

of a sentence with adverbial prepositional clauses, the result is

exactly the same as in Davidson’s analyses. It is the same

individual event that is being characterized in various ways.

The difference will show up in sentences about actions and

their consequences. Sometimes, when the consequences are really

(according to me) the same event described differently, the

analysis will be the same as Davidson’s. When Susan pulls the

trigger, Susan does what Roy Roger’s did when confronting bad

guys. So, substitution preserves truth, since both descriptions

are descriptions of the same kind of act, pulling a trigger.

But when Susan pulls the trigger, one of the things she does

is what Cain did to Abel, but another thing she did is not. Cain

presumably used some other device than a firearm to murder Abel.

Some of what Davidson regards as redescriptions are indeed

redescriptions; others are descriptions of different events. In

particular, in most if not all cases where one thing is done by

doing something else, multiple events are done by a single



motion. When Natty alerts the Iroquois by stepping on a twig,

there are two things he does, neither intentional.

These different things Natty does coincide, in the sense

that they have the same causes and the same effects, given that

we include the event itself as among the effects of that event.

That is, when Susan pulls the trigger, she also murders Fred.

Pulling the trigger includes among its effects Fred’s death.

Murdering Fred, on the other hand, directly includes Fred’s

death.

Notice that the definition of coincidence as having the same

causes and same effects is the same as Davidson’s definition of

“same event.” So, the effect of discriminating events by

different modal features captures the intuitions about the

distinctness of pulling a trigger from murdering and of stepping

on a twig from alerting. However, there is no departure from

extensionality. The tightening of conditions for substitution of

terms is only apparent. Co-referential terms are still

intersubstitutable preserving truth; it is just that which terms

are co-referential is conceived differently.

This is not the same at all as treating events as having

properties as components, and will generally give different

results. Carrying out a famous Davidsonian action is a different

property from buttering bread, but it is not another kind of

event. Events have accidental features as well as essential

features. Just as the statue and the mass of bronze both have the



same location, because location is an accident of each of them,

so the stepping and the alerting are both reported in the

Colonial Gazette.

e) Events, states, and causes

One worry about the Davidsonian analysis of adverbial

expressions as quantifications over events is that there seems to

be no end to replacing predicates that are apparently predicates

of physical objects with predicates of events. Consider the

analysis of “John buttered some toast with a knife” as “Ee(Be /\

Agent(John, e)/\ Ex(Toast x /\ Object(x, e) /\Ey (Knife y /\

With(y,e).” The worry would be that being toast is an enduring

state of bread and being a knife is an enduring state of a knife.

Enduring states may appear to be rather quiet events.

Furthermore, events themselves seem to have states. The battle

may be in a state of relative quiescence, for instance.

There is a large difference, though, between states and

events. States are states of individuals, whereas events happen

to individuals. The same verb may in some aspects refer to an

event, and in other aspects refer to a property of an object.

Roughly speaking, verb forms with imperfective aspect refer to

events. On the other hand, verb forms with perfective aspect

refer to states, in effect properties of the objects involved in

the event the verb describes. So there is a difference between

“John kisses Susan” and “Susan has been kissed by John.” The

kissing is an event, which takes time. Having been kissed by John



is a state of Susan which endures as long as Susan does, well

past the kiss.

Having been kissed by John, of course has something to do

with a kissing event of which John is the agent. “Has been kissed

by John” is an open sentence true of Susan which determines a

property. The state of having been kissed by John is an entity,

but an entity very much like a property. Properties are discussed

in Chapter 6. It is true that, for every event a that happens to

any object b, there is a temporary state of that object,

“undergoing a.” But being such that one is undergoing a is

different from being a. The close relation between events and the

corresponding states of the objects involved in them seems to

have led to a mistaken identification.

So, in logical form, predicates are generally true of the

objects they seem to be true of. Being a frog is not an event,

but a state. Just as we could paraphrase “Joe is a frog” as “Joe

has the property of being a frog,” so we could also paraphrase

“Joe is a frog” as “Joe is in the state of being a frog.” In

Chapter 6, we show how such paraphrase is innocuous.

f) Causes

Davidson famously held that only events are causes. A

difficulty often pointed out is that “The fact that the building

was full of flammable materials caused the fire to spread

quickly” seems true. Davidson, since he has a convincing argument

that facts are not truth-makers, rejects facts as causes. Since



“The building’s being full of flammable material caused the fire

to spread quickly” seems fine, there is also a reason to

assimilate states to events, given the assumption that states are

some kind of slow happening.

As I will argue in Chapter 6, facts are innocuous states of

the world. States, as I will argue in Chapter 6, are properties

of individuals. So, facts and states are entities that can be

quantified over. I argue that both states and facts are

constructed from sentences, and so are intensional entities. So,

Davidson’s analysis of causal sentences as two-place predications

works perfectly well when more kinds of entities are allowed to

enter the relation. My view is that Davidson’s basic insight,

that causal sentences are quantifications over some kind of

object, is untouched by expanding the entities that can be

quantified over to facts and states of objects. Davidson’s basic

idea, that “cause” is a two-place relation, can be preserved.

Furthermore, “because” sentences, which seem to be intensional

contexts, will turn out to be perfectly fine, because they are

relational predications some of the relata of which may be

intensional entities. In Chapter 6, once I have set out how a

Davidsonian can accommodate properties and facts, I will return

to this topic and show how Davidson’s idea works with

uncontroversial and innocuous intensional entities, propositions

and facts.



The view here conflicts with the idea that causation is the

“Cement of the Universe.”131 If one were a monistic realist, the

idea that a central metaphysical relation could be so

promiscuous, involving sometimes events, sometimes pairs of

abstract object such as facts and events, sometimes states and

facts and events, would be quite incredible. I do not think that

“cause” is the cement of the universe. If we look to physics, for

instance, as the natural science with the most rigorous and

precise laws, we find no mention of “cause” whatsoever.132 What we

find are equations, which can be forced into universally

quantified conditionals. These conditionals, if true, are laws of

nature.

How does the necessitation by laws of nature occur without

“is a cause” being part of physics? For a Davidsonian, the answer

will turn out to be straightforward. “Is a law” is a predicate of

propositions. It is a necessary truth that for all propositions

x, if “is a law” is true of x, then x is true. Only if necessity

requires some analysis beyond the disquotational “necessarily a”

is true if and only if it is necessary that a, will appeal to

some extra ingredient, causation, be required in order to

understand the remark. The next chapter will give a Davidsonian

131 The phrase is of course borrowed from John Mackie’s excellent (1980). 132 My colleague Scott Lehmann has urged this point in departmental colloquia

many times.



analysis of modalities according to which “is necessary” is

treated as a primitive predicate of propositions.

f) Events and Substances

Part of the resistance to an ontology of events comes from

the fact that they are not substances, intuitively. Technically,

if they have essential features and accidental ones, as I have

argued, they are substances in Aristotle’s sense. But Aristotle,

along with many later thinkers, takes continuing physical objects

to be ontologically basic. If one had to choose one kind of

object to be primary objects, perhaps Aristotle’s is the right

choice. When someone asks “What exists?,” medium-sized physical

objects, especially organisms, are the first to spring to mind.

Other candidates, such as Plato’s choice, numbers, and the

Ionians’ choices, various kinds of stuff, would be acknowledged

to exist, but would not be the obvious choice of examples of

beings.133

Aristotle was puzzled how it could be that so many diverse

kinds of thing were all correctly called “beings,” things that

are. Aristotle observes that if “is a being” is a single nature

things have, you would get Parmenides’ result that what is is

one. Aristotle’s solution to the apparent problem of multiple

homonymy is his famous “focal meaning” account, according to

133 This description of Aristotle in relation to Plato and the Ionians is

borrowed from a Richard Rorty handout in the late 1960s entitled “Substances

and Substrata,” which I cannot find in any bibliography of his published work.



which being substance is the primary sense of being and other

things said to be are said to be in virtue of some relation to

substance. So, Aristotle is able to saying that “is a being” does

after all have an extension, namely the substances.

Aristotle and many others engaged in metaphysics134 thus take

events to be something to be analyzed in terms of the possession

of a property by a substance or substances for a time, or other

constructions which make physical objects basic. It is true that

very often, at least, there will be a sentence about physical

substances which will “say the same thing” as a sentence about

events. “Saying the same thing” for many purposes says little

about the ontology, unless there is a reason to have less

ontology.

Quine’s idea, that to be is to be the value of a variable,135

I take to be a version of the view of this book, that beings are

posits necessary for a recursive syntax and semantics for

thinkers who can make formal inferences that depend on sub-

sentential structure. From Quine’s point of view, as I understand

it, beings are not given, but are rather posited to allow us to

think with a recursive syntax and semantics. We have to have

singular terms and predicates in order to think and speak with a

134 Jonathan Bennett’s wonderful book, Events, presents a theory that in many

ways, except for its assumption of monistic realism, agrees with the views

presented here.135 Quine (1948), page 15.



recursive syntax and semantics which can discern logical

relations that depend on sub-sentential structure. In order to do

so, we posit beings. Since Quine is wedded to the idea of a kind

of primacy for natural science, the posits he takes most

seriously are the ones that lend themselves to scientific

purposes. Without such scientism, recognizing that different

posits need not get in one another’s way, we can recognize that

humans posit many kinds of beings, some useful for some purposes,

some useful for other purposes.

Chapter 5 Modals and Conditionals

I What is the problem with modaIs?

This chapter and the next three chapters illustrate how

Davidson’s basic semantics of predication, combined with various

other views of his, dissolve or solve metaphysical questions.

This chapter, on modals and conditionals, applies Davidson’s

analysis of indirect discourse, together with his conception of

what a semantics is as opposed to what a theory of a subject

matter is, to the semantics of modality. As it happens, a

Davidsonian account of modality requires a Davidsonian account of

conditionals. The account of conditionals developed in the

present chapter differs quite a bit from anything Davidson ever

said.136

136 Davidson’s view of conditionals in 1967 (Wheeler 2012) was that they are

truth-functions, following Grice’s ideas. I do not believe he ever changed his

mind. The deviations from apparent truth-functionality would be explained by



That there is a Davidsonian account of modality is quite

central to the project of this book, since we have helped

ourselves to lots of modal constructions in characterizing

relative essentialism. This free use of modal notions supposes

that modal notions are not a special problem for a Davidsonian.

My project in this chapter is to present the outlines of an

account of modal sentences that conforms to the principles of

Davidsonian semantics and accommodates the data at least as well

as other semantic theories. While a thorough account of the

intricacies of modal constructions in English is well beyond the

scope of this chapter, the outlines of how a Davidsonian account

would go will at least cover some basic modal predicates and show

how the theory works.

Davidson himself only indirectly dealt with the semantics of

modality. Davidson (1968) suggested that his idea that “that” is

a demonstrative pointing to an utterance that is not strictly

part of the sentence generalizes to psychological verbs, but did

not even mention modalities. Davidson (1970b) showed how “ought”

is akin to “probably.” Once again, even though ethical words,

“good,” “right,” and “ought” are among the examples of concepts

whose logical grammar must be understood before any worthwhile

philosophical analysis is possible in Davidson (1967) no further

development of an account of “ought” occurs in Davidson’s

interpretation. As I argue below, that almost works, but not quite.



writing. Chapters 9 and 10 fill out one way a Davidsonian could

proceed.

This chapter will pursue Davidson’s suggestions in Davidson

(1968) and (1970b) and develop a Davidsonian account of

modalities, including “ought.” The last two chapters of this

book, on Davidsonian ethics, will show how Davidson’s insight

about “ought,” combined with his conception of interpretation,

yields the outlines of an ethical theory.

Modality is one area where most semantical theories

typically go far beyond truth-preserving recursive disquotation.

It is also an area where some metaphysicians have appealed to

very exotic entities. Why this is the case seems to go rather far

back in the history of modern philosophy.

The empiricists, especially David Hume, worked out the

consequences of the view that everything we know is based either

on sense-perceptions or on the analysis of concepts. Among other

things, necessity and possibility, briefly, do not have

corresponding sense impressions. Thus since we do not sense such

connections, such connections were not part of the given from

which we construct knowledge of the world. The dominant tradition

in analytic philosophy follows a basically Humean empiricism,

that there are no necessary connections between distinct

existences. For much of the twentieth century, analytic

philosophers conceived of concept formation as operating on a

given domain of objects of some kind, so that a system of



concepts was a kind of sorting into groups of this given domain.

There were alternatives, but the evidential basis was the same

for all conceptual schemes. Necessity and possibility were no

part of this given. Thus necessity and possibility were suspect

notions—some special analysis was needed, either in terms of

conceptual containment137 or by a semantics that would treat the

special relationships of modal concepts as falling out of

quantification over possible worlds.

For a Davidsonian, there is no reason to treat modal

concepts any differently from others, except for the fact that

they create non-extensional contexts. Non-extensionality is the

only puzzle about logical form. Davidson (1974) rejects the

conception of a “given” domain of objects. Without a given, there

is no reason to treat necessity and possibility differently from

other predicates, as long as there is something for them to be

true of. Given an ontology of Fregean senses, for instance, “is

necessary” might be just a predicate of such senses.138 The

semantics would demand nothing else, since there is no special

mystery about necessity not being among the given data, since

nothing is among the given data.

137 For instance, Carnap’s (1947), which treats conceptual containment as

linguistic containment.138 This was not Frege’s view of necessity, but the position is available to

someone with his ontology.



Kant (1781) tried to ground necessity as epistemological by

analyzing necessities as impositions of the mind. In a way, the

relative essentialism of Chapter 3 is a continuation of the

Kantian strategy. Most of the early twentieth century analytic

philosophers until 1970 followed Kant insofar as they accepted

the idea that modal truths depended on humans and their thought

and language rather than a feature of the world itself. The

majority view was that necessity was analyticity, if anything.

However, once the rediscovery of Aristotelian essentialism

gathered steam after 1970 with Kripke’s (1980), it still seemed

to philosophers that some kind of special formal analysis of

modality was required. Even though it was now respectable to

treat necessity and possibility as features of the world,

modality was somehow still suspect, requiring some special

reduction to something more acceptable.

II Possible worlds

Since the 1950s, a number of philosophers constructed

theories of various modalities, especially necessity and

possibility, noting the strong analogy between the pair

“possible” and “necessary” and the pair “some” and “all.”139

139 Notice that, for my kind of Davidsonian, the relationship between “all” and “some” is part of theory, not of logical form. A theory of thetruth-conditions of modal predicates will of course take note of the resemblances among the all/some pair, the obligatory/permitted pair, and the necessary/possible pair.



Modalities could be construed as quantifications over possible

worlds. Given that possible worlds are some kind of beings, this

would treat modalities extensionally. This idea of possible

worlds goes back to Leibniz, the idea of quantifying over them

had been part of Carnap’s (1947), and has been vigorously pursued

ever since.

The project among philosophers of language and

metaphysicians has been to construct an account of modality that

analyzes modal sentences as some kind of complex non-modal

quantification over possible worlds. Roughly, what has to be or

could be is some kind of construction out of what merely is.

Most, like Stalnaker (2003), treat the construction as a kind of

heuristic device, and do not think that the construction is a

reduction at all. I am happy to understand such analyses as

heuristic devices which illuminate what it takes for modal

predicates to apply to a proposition or to relate two

propositions.

Others take the philosophical project to be an ontological

reduction of what must be to what is. The lesson from Aristotle

that the notion of being can be no more basic than that of

necessity seems to be forgotten. To suppose that being a being is

clear whereas being necessary is obscure forgets that any beings

that can be counted require a principle of individuation, and

thus a distinction between what they can be and what they cannot



be. In the long footnote below,140 I argue that both David Lewis

and David Armstrong, who have two very different ways of trying

to understand modalities without appealing to primitive modality,

in fact fail to reduce modality to what merely is. In fact, both

are committed to Aristotelian essentialism as an explanation of a

modal truth about at least one kind of entity.140? Lewis (1986) and Armstrong (1989) both accept Hume’s Principle, that there are no necessary connections between distinct existences. This means that anything can go with anything, and that essential connections cannot be primitive. Briefly, Lewis treats possible worldsas mereological sums of beings and Armstrong explains possibility as recombination of beings that are. For Armstrong, the fundamental kindsof beings are particulars, universals, and facts that are the result of particulars instantiating universals. As it happens, Armstrong and Lewis are already committed to primitive modalities. In Armstrong’s case, “is a particular” and “is a universal” determine important modal truths about particulars and universals, the features without which the entities would not be what they are. Only a particular can be an instance of a universal; only a universal can have instances. These are necessities about particulars and universals. Armstrong has two basic kinds of substances, in Aristotle’s terms. But the natures of universals and particulars whichaccount for these modal characteristics require special treatment. Particulars by their very nature can instantiate universals but cannotbe instantiated. Symmetric necessities hold about universals. These necessities cannot be explained by possibilities of re-arranginguniversals and particulars. No recombination seems applicable. So, thetruth-maker for “a is necessarily a particular, not a universal,” mustbe the particular a itself. This is exactly what Aristotle says about every substance-determiner. Aristotle just adopts that theory of truth-making for every substance-constituting nature. Lewis avails himself of necessities about the part-whole relation, butthose could perhaps be construed as part of logic, and are certainly not independent existences. Lewis most telling appeal to necessity that cannot be understood by appeal to possible worlds is in his account of sets. A set is a distinct existence from its members. Lewis’ set theory (1991) limits set-abstraction to unit sets and their



Theories quantifying over possible worlds, construed as

proposals about the semantics of modal sentences, have spawned a

variety of proposals for the semantics of various constructions,

yielding deontic logics, alethic logics, temporal logics, et al.

What the target sentences have in common is an apparent breakdown

in extensionality. On the Davidsonian141 view I propose, these

“logics” are theories about the extensions of modal and other

predicates. They are not part of the semantics at all, any more

than biological science is part of the semantics of “horse” and

elements. Sets and elements must be independent existences in order todistinguish sets from mereological sums. Lewis must posit a necessary bond between a unit set and the entity of which it is the unit set. There is no possibility of explaining this necessity in terms of possible worlds, since the sets of most interest are the sets he identifies with properties, which have elements in many possible worlds. Lewis himself characterizes the bond between a unit set and its element as “mysterious” (Lewis 1991, chapter 2). A necessary feature, in fact an entity-constituting feature, of each singleton is that it is the singleton of the entity of which it is the singleton. Likewise, a necessary feature of every entity is that it bears the membership relation to its singleton. Singletons are strange substances, perhaps, but substances none the less. Individuals actually have haecceities, in the sense that there is a feature each of them has, namely having this entity as their unit set that no otherindividual has. In short, Lewis and Armstrong fail to reduce modality to brute fact and fail to eliminate Aristotelian appeal to natures as the basis of necessity. At least some of their beings have Aristotelian essences. Their accomplishment is to reduce modality to a very few kinds of instances of primitive modality and a lot of ontological equipment.

141 Davidson (1967) page 32, as well as his remarks in class in 1967 (Wheeler

2012) are the source of my view of “logics.”



“plankton.” On the conception of the modalities I will defend,

modals are predicates of “things said,” or propositions, about

which much more below in Chapter 6.

Since the opposing view of the vast majority is that modals

are quantifications, perhaps we can step back a moment and

consider what modal words seem intuitively to be. Many modals

seem to be basically adjectives, with, of course, adverbial and

nominal forms. “Necessary” does not seem to be characterizing a

quantity of anything. “Probability” seems to be a quantity of which

there can be more or less. We have “necessary,” “necessarily,”

and “necessity.” Many modals seem to have comparatives. “More

likely than not, we will have snow this winter,” “We are much

more likely to have snow in January than in July,” and the like.

One is tempted to treat modals as a special case of comparative

adjectives, as discussed in Chapter 9 below.142 I succumb to that

temptation below.

A further difficulty with treating possible worlds as the

entities over which we quantify in understanding modals is Lewis’

(1986) argument that only his realist conception of possible

worlds as concrete can do justice to modal intuitions without

appealing to modality as primitive. I find his argument 142 Of course, quantifiers are also adjectives, many of which have

comparatives, so nothing rules out the possibility that the modal adjectives

are really quantifiers true of sets of possible worlds. But the possibility

should occur to us that less exciting entities may be the objects to which

these predicates apply.



persuasive. I also find his conception of possible worlds

impossible to believe. Therefore, an assignment of logical form

that treats modals as predications and accepts that “necessary”

is as primitive as “exists” deserves at least a try.

III Parataxis

Of the constructions in which extensionality fails, such as

psychological verbs and modals, the least problematic is

quotation. In quotation, it is clear that one is talking about

words as opposed to what words designate. Thus, of course, it is

not substitution of co-referential terms when one substitutes

“The square root of twenty-five” for “five” in “`Five’ has four

letters.” A naturalistically-inclined philosopher would be

attracted, as Carnap (1947 and 1937) was, to the idea of a

linguistic version of Frege’s conception of a reference-shift to

explain substitution-failure. That is, one could explain why

substitution of co-referential terms does not preserve truth-

value by positing a reference-shift to something linguistic

rather than to an exotic abstract entity like a Fregean sense.

The fundamental difficulty with such theories is that, if the

quotation or simulacrum of quotation in fact does its job by

referring to a linguistic item, the sentence would appear to be

talking about a linguistic item. But sentences about necessity

and belief do not seem to be talking about language at all.

Church’s (1950) translation objection raised this question

about a sophisticated version of this linguistic approach to



belief-contexts, Carnap’s (1947) account. Carnap needed to

accommodate the fact that a person can believe one of a pair of

logically equivalents and not believe the other. So, “Galileo

believed that the Earth moves” became “Galileo accepted a

sentence in some language L intensionally isomorphic to `The

Earth moves’ as an English sentence.” Church’s objection,

briefly, was that the translation of the analysis of the belief-

sentence into another language would carry along the translation

of that specification of the language relative to which the signs

were to be interpreted. That is, since “ `The Earth moves’ as an

English sentence” refers to an English sentence, and translation

at least preserves reference, the translation into Italian would

also refer to the English sentence. But it does not.

a) “On Saying That”

Davidson (1968) proposed a semantics for non-extensional

contexts that retains the idea that non-extensional contexts are

to be understood in terms of the linguistic while avoiding the

Church translation objection. Davidson also avoids the ad hoc

device of postulating that certain sentential contexts bring

about systematic shifts of reference. This second desideratum he

called “semantic innocence,” (1968, p.68) by which he means that,

prima facie, we should let words mean what they usually mean.

Davidson’s account invokes interpreted linguistic items, but these

interpretations of linguistic items are not parts of the sentence

for which the semantics is being given. Thus the translation into



another language need not contain names of those linguistic

items. A further difference from Carnap is that the linguistic

items are not conceived of as sequences of marks which have

meaning only relative to an interpretation, but as speech- or

writing-acts, which have a particular speaker or writing at a

particular time, and so come with a meaning.

Davidson’s idea is to treat the “that” in “Galileo said that

the earth moves” as a demonstrative, so that the analysis would

be, “Galileo said that. The earth moves.” “The earth moves” is a

sentence with truth-conditions presented by the speaker for a

particular purpose, that is, to present what it is that Galileo

said. The speaker is neither endorsing nor not endorsing the

sentence.

The idea in Davidson’s (1968) is akin to Davidson’s (1967)

thesis about truth-conditions generated by a well-founded truth-

theory giving the meaning of sentences. Davidson (1967) needs no

ontology of meanings, but rather specifies the two-place

predicate “means.” In the same way, there is no entity which is

the content of an utterance, but what the utterance says can be

demonstrated by producing an utterance that says the same, that

is, an acceptable interpretation of the utterance. Another way of

putting this is that the “that” clause, “that the Earth moves”

does not refer at all, because it is not a constituent, just as “the

present king of France” is not a constituent of “The present king

of France is bald” according to Russell. Yet another way to put



this: The “content” of the “that”-clause is no more a semantic

constituent of the sentence than my facial grimace is of “When

Fred tasted her Tofu Broccoli Surprise he went like that.”

What does it take for “the earth moves” to be what Galileo

said? Davidson supposes a relation between utterances he calls

“samesaying” that holds, in this case, between utterances. Two

people are samesayers if their utterances say the same thing. The

“things” are just the utterances and inscription-acts that are

bound by the samesaying relation. Samesaying is supposed to

capture something like the following: A’s and B’s utterances a

and b make A and B samesayers just in case a is a good

interpretation of b by A.

There are several things to note about saying what another

person said:

First, it is vague how close the match of truth-conditions

has to be. In many circumstances, an interpretation presented

need not even have the same truth-conditions as the actual

utterance being interpreted. If Fred says, “I’m going to drive to

Willimantic this afternoon” and I tell Bill “Fred said he would

be in Willimantic before evening,” I have correctly reported what

Fred said, even though some ways of being in Willimantic before

evening do not involve driving. “Before evening” would be true if

Fred arrived in the morning.

It is indeterminable exactly which things I utter say the

same as what another person said or wrote. In the same way, and



for the same reason, it is indeterminable whether two people have

exactly the same belief, and indeterminable whether one property

is the same as another. As we will examine in the case of

property-identity, the conditions that might be specified for

saying the same thing parallel conditions that have been proposed

for identity of properties.

Some contexts indeed require something close to citing the

exact words of Fred, except for replacing his demonstratives with

yours. If Fred said “I’m going to kill Susan,” then when I am on

the witness stand, I cannot say “Fred said he was very angry at

Susan” but I should say “Fred said that he was going to kill

Susan.”

Second, as the above example makes clear, the

interpretations that are given as what a person said are not

translations, but interpretations. We reporters say the same thing by

saying what Fred said from our point of view, as it were,

changing the demonstratives to say the same thing from our

position. So, the references of demonstratives and pronouns are

shifted, as well as tenses. Fred’s utterance was made at a

particular time and place. The reporter’s time and place may be

quite different. To say the same thing, the demonstratives Fred

used have to be replaced by demonstratives that demonstrate the

same things from where and when and who the reporter is.

Third, there is no problem with demonstratives that contain

other demonstrations, once one bears in mind how tenses and



references of demonstratives shift from context to context.

Imagine a scenario where I am reenacting, for the entertainment

of Fred’s detractors, Fred’s surprise at seeing the new façade of

Manchester Hall. In front of the building, I start out saying “He

went, like…” I step back, as if stunned, and say, “Holy cow,

that’s a disaster!” indicating the façade. My demonstration was a

demonstration of a demonstration. The interpretation of my

performance could be something like “Sam said that Fred said that

the renovation was aesthetically unsatisfying.”

Fourth, the overt demonstrative is optional. “John said he

would be here” is only stylistically different from “John said

that he would be here.” Many applications of Davidson’s idea that

I make in other chapters have no overt demonstrative. In

particular, the applications of Davidson’s idea in Chapter 9,

which deals with words like “good,” which create intentional

contexts, involve no overt demonstrative.

It is important that the notion of samesaying is not part of

the semantics, but rather part of the explication or theory of

the application of the predicate “says.” For someone to say an

utterance that is correctly presented by the that-clause,

samesaying has to obtain between the utterance and the content of

the that-clause. In the same way, it is not part of the semantics

of “run” that if “John runs” is true, then “John moves” is true.

Davidson’s account of indirect discourse has been attacked

on many fronts. Some of these objections seem to me to rest on



misapprehensions about what “samesaying” is. What we have

observed above, that indirect discourse changes demonstratives

and temporal indices, is raised as an objection to the theory. If

“the same thing said” is taken to be “the same words, translated”

then the objection would be good. But Davidson is talking about

utterances, not sentence-types. And he is talking about

interpretation, not translation. Others seem to be the result of

lack of imagination. Lepore and Ludwig’s (2007) Chapter 11

catalogs the objections and responds to them in ways I endorse

with only trivial reservations. Blair (2003) is a book-length

defense of Davidson’s basic idea.

Ian Rumfitt’s (1993) suggests some modifications of

Davidson’s ideas to accommodate some apparent difficulties. One

of these suggestions is that we not focus on Davidson’s

suggestion that “that” in the complementizer position is really

the same word as “that” the demonstrative. If the theory is to be

a general account of indirect discourse, modals, and

propositional attitudes, it has to generalize beyond English. As

Schiffer (1987, page 125) points out, other languages have

different words in the two roles. The view has to be, rather that

“that” in the “complementizer” position just is a demonstrative,

and that “que” in French is likewise a demonstrative.

Another of Rumfitt’s suggestions is that what Davidson needs

is something like interpreted logical forms as demonstrated

objects. Otherwise, ambiguous sentences, such as “John said the



shooting of the hunters was atrocious” will not convey any

particular thing that was said. That is, Rumfitt’s suggestion is

that the demonstration must be to something more fine-grained

than utterances as strings of words, since such strings can

correspond to more than one thing said.

I think this suggestion overlooks a distinction between

“says that” and other applications of Davidson’s ideas. It makes

perfect sense to say “John said that his daughter goes to a

pretty little girls’ school, but I’m not sure what he meant,”

whereas it makes no sense to say “John believes that his daughter

goes to a pretty little girls’ school, but I’m not sure what he

believes. Propositional attitudes and modalities indeed require

that one’s demonstrated utterance be interpreted in the logical

form in which it is meant. When one is reported speech, however,

there are at least two ways one can be understood: One could mean

to be reporting the intended meaning of what is said, which is to

treat “said that” as exactly like “believes that,” or one could

be reporting, in one’s own words, the actual speech production.

Indirect discourse has a kind of ambiguity that other

applications of Davidson’s idea lack. When one reports a belief,

one’s utterance has to be understood as one meant it, in order to

be reporting a single belief. What someone said, though, can be

understood as either their actual words, or their words as used,

that is, with a particular logical form. “Things said” that are

objects of attitudes and subjects of modal predicates are



sentences as meant, that is sentences with their logical forms. A

sentence is only used when it has a particular logical form.

So, there are several possible things said by an utterance

of a sentence such as “My daughter attends a pretty little girls’

school.” I can say”John said that his daughter goes to a pretty

little girls’ school, but I don’t know what he meant.”

b) Generalization of Davidson’s idea

Davidson (1968) is an account of indirect discourse. I will

make three generalizations of Davidson’s idea. The first, which

he briefly mentioned in Davidson (1968), is to other categories

of intensional propositional containment. The generalization to

modalities will be the focus of this chapter. The extension to

psychological verbs and mixed cases of verbs whose truth-

conditions involve both psychological states and states of the

world, such as knowledge and action, is a topic this book will

not explore, but the general idea of how that project would be

pursued should be evident.

The second generalization of Davidson’s idea is to extend

the objects of demonstrations to open sentences and other sub-

sentential things that can be said rather than just clauses with

truth-values.143 There are failures of extensionality which do not

involve clausal complements. This can happen in ways that are

143 Rumfitt (1993) suggests this as a way of accommodating quantification into

indirect discourse. The idea has much broader application, as we will see.



really quotation. Quine’s example was “Giorgione was so called

because of his size.” In this sentence, “Giorgione” is being both

used and mentioned. Chapter 9 on “good” and other comparative

adjectives uses demonstration of things said that are not

quotation-like, but rather like indirect discourse. This

extension will account for the apparent intensionality of “good,”

“expert” and “skillful,” for instance, to be discussed in Chapter

9.

As Rumfitt notes, this second generalization addresses one

of the oft-cited objections to Davidson’s paratactic account of

indirect discourse and is worth some elaboration, since it gives

a Davidsonian an account of quantifying into belief contexts. The

sentence, “Every boy said he would eventually get married”

appears to be a difficulty for this demonstrative account. If

open sentences are among the things said that can be

demonstrated, this can be rendered “Every boy said of himself

(that). X would eventually be married.” Likewise “Every boy said

he loved his mother” can be rendered, “Every boy said of himself

(that). Loves his mother. (There is a unique x such that x is the

mother of y and y loves x).

The extension to “quantifying-in” to psychological verbs is

clear. “There is someone whom Ortcutt believes to be a spy” is

just “There is someone of whom Ortcutt believes that. Is a spy.”

“Ortcutt believes there are spies,” on the other hand, is

“Ortcutt believes that. There are spies.” Of course, a



Davidsonian account will not say much about the precise

conditions under which this will be true rather that the form

“Ortcutt believes that. There is someone who is a spy.” I very

much doubt that there are precise conditions for that difference

between “having someone in mind” and not.

The third generalization of Davidson’s idea interprets

“things said” in a Davidsonian way and applies it to Aristotle’s

locution “legomena.” Chapter 6 on properties and propositions

will give a language-derived, innocuous theory of properties,

propositions, and facts. Like the modal predicates, the account

of these items will be an extension of Davidson (1968). Briefly,

properties are things said that are open sentences, propositions

are just things said, and facts are states of the world being

such that things said are true.

Starting with indirect discourse has a great advantage for a

Davidsonian theory. Metaphysical disputes about how “fine-

grained” properties and propositions should be become

observations about the various standards we have for accurate

indirect discourse. Suppose Fred says “I want to permanently

solve my Susan problem.” If I am reporting what Fred said when I

report this to Susan, I can say “Fred said he intends to kill

you” and be accurately saying what Fred said. In court, however,

under cross-examination by Fred’s lawyer, I have to say, still

accurately, “Fred said that he wanted to permanently solve his

Susan problem,” changing the pronouns and verb tense. Indirect



discourse makes it obvious what is going on when we can correctly

characterize a person’s beliefs, say, in more or less general

terms. Indirect discourse is a model of why it is that various

property-theorists have different conceptions of what it takes

for a property F to be the same property as G, as we will see in

Chapter 6. Likewise, identity-conditions for propositions and

facts are questions of standards in place.

IV The logical form of modals

The account of the modals I will offer has four parts:

First, I show how simple modalities can be given a Davidsonian

semantics. Second, I point out some difficulties that arise from

the fact that many “conditional” modalities seem not to be a

modal applied to a conditional. Third, I show that, apart from

modal contexts, “if”-sentences can be interpreted as truth-

functions. Fourth and finally, I offer an account of modals that

treat the “if” component as in effect semantically neutral. “If”

is basically punctuation for a two-place relation among things

said.

a) Simple modals

The simple, traditional modalities, such as “It is necessary

that two is even” and “It is possible that more rain will fall

next year” would fall easily under Davidson’s paratactic account.

They will be something like “This is necessary. Two is even” and

“This is possible. More rain will fall next year.” The “it” is an

expletive that provides a subject for the sentence, like the “it”



in “it is raining.” What a demonstrated sentence says is

characterized with an adjective. Simple, partial, inadequate

theories about what is said about such things said might be “Is a

truth of logic” or “Follows from physical laws and present

conditions.” What precisely a thing-said has to be in order for

the predicate “is necessary” to be true of it may not have an

account in other terms—few predicates of interest are strictly

definable outside of mathematics. However, the normal

disquotational predicate clause will be accurate, if somewhat

unilluminating: “X ҇ ‘is necessary’ is true if and only if what

‘X’ says is necessary.” So, “It is necessary that” is true just

in case that is necessary. Depending on what “that” demonstrates,

the sentence will be true or false.

“Is necessary” and “is possible” are adjectives, on this

account, and should be treated in the way adjectives are treated.

Given the account of adjectives in Chapter 9, the kinds of

necessity and possibility can be straightforwardly treated as

adverbs. “That is logically possible,” “physically possible,” and

the like will have the form, “There is some necessity x and x is

of that and x is logical/ physical.” Just as with other

comparative adjectives, the idea is to think of “quantities.” It

may be objected that necessity and possibility do not admit of

degree. That is a part of the theory of necessity and possibility,



not the semantics.144 This peculiar feature has some important

consequences, as we will see below. Many other modalities, for

instance, “probable” do admit of degrees. For the purposes of

this chapter, though, we will not go into how to treat these

modal predicates as comparative adjectives.

Nothing in the semantics will account for the equivalence of

“It is not necessary that two is even” and “It is possible that

two is not even.” That equivalence, along with the accounts of

what it takes for the predicate “necessary” and “possible” to be

true of things said, are theories involving those predicates. It is

not out of the question that the best such theory will posit

possible worlds, nor is it out of the question that the best such

theory will analyze such predicates in terms of physical laws and

their consequences, moral laws and their consequences, and the

like. None of those theories need be construed as part of the

meaning of the predicates. The semantics will deliver only the

disappointing “`It is necessary that two is even’ is true if and

only if it is necessary that two is even.” We have a simple

predication.

In practice, however, such simple predications need to be

interpreted. When someone says, “It is impossible for you to get

from Boston to Manhattan in less than an hour,” some

interpretation is needed if what was said is to be interpreted as

144 I’m not sure that in ordinary speech, necessity and possibility do not admit of degrees. “To succeed in life, it is more necessary to have influential friends than it is to work hard.”



true. An F-16 can make it in fifteen minutes. Somehow, conditions

are imported into this “impossible” sentence. 145 One obvious

idea is to take the demonstrated sentence to be the second

argument of a modal “if”-sentence, with the antecedent implicit

and given by context. Another obvious idea is to suppose that

context, supplied by interpretation, allows sentences which are

not literally true to be understood as saying something true. Any

standard account of modality using possible worlds has

essentially the same difficulty coming up with an account of how

we actually understand modals. I will adopt both obvious

strategies, in a way.

The modalities central to relative essentialism, as it

happens, are the traditional ones, namely necessity and

possibility. “Necessity” is of course the nominal formed from the

predicate; “necessary” is the adjectival form, and “necessarily”

the adverbial form of this predicate. Given that modal predicates

are predicates, what are they true of? “Necessary” applies to

very many kinds of things, all of them derived from linguistic

items. There are necessary truths, necessary facts, necessary

propositions, and necessarily applying properties. As we will see

in Chapter 6, Davidsonian accounts of properties, propositions

145 Other readings are possible. If my colleagues all have Lamborghinis, Ferraris, and Aston Martins, so that they all could, at two in the morning, perhaps, make the trip in question, then they might, mockingly, with emphasis on the “you” say the same thing as a way of pointing out that I drive a Prius.



and facts are innocuous generalizations of Davidson (1968.) No

point that I can see is served by regimentation. In general,

there are numerous ways of saying the same thing. “It is a

necessary truth that p,” “P is necessarily true,” and “The fact

that p is a necessity” all come to the same thing, since

propositions, states, and facts are all entities that derive from

the same things said.

b) Necessary properties and propositions

If Fred is necessarily rational, our understanding, by

Chapter 3, is that Fred (the human) is necessarily rational.

Relative essentialism endorses such attributions of necessary

properties. If properties are construed as things said, and

things said are demonstrated items strictly outside the sentence,

then the “necessarily” will be true of the state of Fred that he

has the property, that is, of Fred’s being rational. The analysis

will be, “There is a state x of Fred (the human) and x is Fred’s

being that and x is necessary. Is rational.” The utterance

demonstrates a thing said, in this case an open sentence. The

failure of truth-preservation under substitution of, for

instance, “has the trait Carnap most often mentions as a

necessary one” is explained by “necessarily” being true of

“rational” being true of Fred, a state. Facts, that is, as we

will argue in Chapter 6, states of the world, can be necessary.

The “Fred” position is extensional, since “is rational” is the

demonstrated thing said, not “Fred is human.” So, “The (human)



jerk who offended everyone at Barbara’s wedding is necessarily

rational” will be true just in case “Fred (the human) is

necessarily rational” is true, given that Fred is the culprit.

However, “is necessarily rational” will not be true of Fred the

complex of micro-particles.

“It is necessary that Fred is human” is ambiguous between

“That is necessary. Fred (the human) is rational.” and “Fred (the

human) is such that he is necessarily that. Is rational.” The

first is de dicto. It says that the proposition that Fred (the

human) is rational is necessary. The second is de re. The second

says that Fred is in a state, being rational, necessarily.

V A fly in the ointment

Among the modalities, “necessary” and “possibly” are

exceptional cases akin to “all” and “some” among the quantifiers.

They are, as it were, extreme cases and that is why they can be

represented as one-place predicates and their conditionals can be

predicates scoping over conditionals. The wide scope reading of

“If P then necessarily Q” can be represented as just “It is

necessary that if P, then Q,” and the wide-scope reading of “If

P, then possibly Q” can be represented as “It is possible that P

and Q.” From the perspective of a general account of the

semantics of modalities, though, these representations are

misleading in the same way that the representation of “All frogs

are green” as “\/x(Fx -> Gx)” is misleading. That is, the

situation is very much like what is the case with “all” and



“some.” Because they do not admit of degrees, “all” and “some”

allow paraphrases using truth-functions. As Heim and Kratzer

(1998) page 191 point out, it is a kind of misleading accident

that “all” and “some” can be treated as one-place quantifications

of truth-functions. “All frogs are green” comes out to have the

same truth-conditions as the universal quantification of the

truth-functional conditional and “Some frogs are green” is the

existential quantification of the conjunction. As discussed in

Chapter 1, other quantifiers are not so simple. Necessity and

possibility are likewise extreme limiting cases of modalities,

and only for this reason can conditional necessity and

possibility be one-place predicates of single things said,

whether truth-functions or other conditionals.

The list of obvious modal predicates is very long, just as

the list of quantifiers is very long. “Can,” “has to”, “may,”

“should,” “probably,” and “ought” only begin the list. The lesson

of the quantifiers applies to modals. Rescher (1964) showed that

no first-order paraphrase of “most” in terms of anything about a

truth-functional conditional preserves truth-conditions. David

Lewis (1976) establishes for modalities something similar to

Rescher’s result about quantifiers. Where “If A then probably B”

is understood as the ordinary language claim that the conditional

probability of B relative to A is high, there is no way to get

“If A then probably B” to be a compound of a one-place modality

applied to any kind of conditional. Lewis proves that conditional



probability is not the probability of any reasonable

conditional.146 Trying to replace the truth-functional

conditional with something whose truth-conditions are more suited

to the English conditionals won’t work.

Hempel (1960) showed long ago that conditional probabilities

do not detach. From “If A, then probably B” and “A” you cannot

infer “B.” Davidson (1970b) argued for the same result in the

case of “ought.” In the case of “ought,” it is pretty clear, as I

will argue in Chapter 10, that it cannot be a modality applied to

a conditional. “If you want to succeed, you ought to study” is

unlikely to amount to some conditional that ought to hold. As I

will argue in Chapter 10, the idea that the conditional “ought”

is “prima facie obligation” is as misguided as the idea that

probability is conditional necessity.

I argue that most modalities explicitly or implicitly come

with an “if” clause. Only for special cases like “necessary” does

it turn out that a conditional modality comes out to be the same

as the modality applied to a conditional. Briefly, if the “if” in

“If A, then probably B” is not the probability of a conditional,

then “if” must be doing nothing other than signaling the first

146 If, as we will see below, “If A then B” is itself a

conditional probability claim, then naked conditionals seem not

to have truth-values.



argument of a semantically primitive two-place modal or otherwise

intentional147 predicate.

The consequence for a Davidsonian is that many conditional

modalities are generally primitive two-place predicates. Since

predicates have a determinate number of places, modalities must

be either two-place or one-place. If conditional probability and

conditional “ought”-sentences are not a probability-predicate or

an “ought”-predicate applied to a conditional, then, if these are

predicates at all, the “probability” and “ought” predicates in

categorical and conditional sentences are the same. In that case,

the conditional modalities are two-place predicates and the

apparently categorical one-place modality has a covert argument

place.148

The choices for a theory seem to be the following:

1) Treat the modalities as quantifications over something (e.g.

possible worlds), and treat “if”-clauses as restrictors of

quantifiers. This requires that “if”s in indicative conditionals

have an occult quantifier and that modal predicates be

quantifications. “If then probably” might, for instance, use

“most.” “More likely than not” might use the quantifier “more

than half.” When we think about the ingenuity that would be

147 As we will discuss in Chapter 10, at least some propositional attitudes are

two-place propositional attitudes. Conditional desire, for instance appears to

be like conditional probability. See McDaniel and Bradley (2008).148 This is exactly what my (1974) and McDaniel and Bradley (2008) do.



required to come up with appropriate quantifiers and appropriate

analyses, the apparatus required looks much more like a theory of

something than an assignment of logical forms. If the theory were

about logical forms, then every modal sentence would entail the

existence of the special, to my mind exotic, objects over which

we are quantifying.

2) Treat “if…then…ought”, “if…then…necessarily” and the like as

unstructured primitives. The difficulty with this option is that

“if…then…” in these constructions would have nothing to do with

“if…then…” in contexts without modals and “ought” and “probably”

would be homonyms, depending on whether they occurred with an

“if…then…” or not.

3) Treat “ought” and “probably” as always two-place, treat “if”

as not really having a semantic role at all in these cases,149 say

something about “then,” and treat apparent categorical “ought”s

and “probably”s as having a hidden first argument clause.

On reflection over decades, it seems to me that the best

idea is 3). Before exploring this idea, though, it would be good

to say something about conditionals to make the “no semantic

role” account plausible. Hence the next section, on

conditionals.

149 The third approach, as the reader will see, will result in a theory very

much like that of Kratzer (2012) Chapter 4, minus possible worlds and other

features. The basic idea will be that “if” does almost nothing semantically..



VI Conditionals

The thesis of this section is that “if” has almost no

semantic role, except to distinguish the first argument from the

second argument of two-place relations among things said, i.e.

the things demonstrated in clauses. The truth-functional

conditional will turn out to be a kind of null modal, applicable

when no modal or propositional attitude expression is present.

That is, all “if” sentences have the form “If modal that, that.”

This section will begin with a discussion of some apparent

difficulties with treating the English conditional as a truth-

function. I will then briefly sketch the sorts of alternatives

that have been proposed. There is an excellent book, Jonathan

Bennett’s (2003), which examines these alternatives in detail.

For decades, there have been sporadic attempts to defend the

idea that the English conditional is a truth-function. There are

some difficulties with the position.

a) (Some of the) Problems with the English conditional being

truth-functional:

1) Truth-functionally true, but not true?

Some truth-functionally true sentences seem to be false,

while others seem to be true. “If Wheeler is our Dean, then the

Earth will open up and swallow his enemies” sounds false, while

“If Wheeler is our Dean, then the Philosophy Department will have

many new appointments,” is true. Likewise, “If a comet crashes

into Storrs tonight, then Phil 2211 will meet as usual,” seems to



be false, while “If a comet crashes into Storrs tonight, then

Phil 2211 will not meet as usual,” seems to be true. Of course,

all four sentences are true, on a truth-functional account.

2) Contraposition:

Contraposition is logical equivalence-preserving, according

to the truth-functional account of the conditional, but this

seems to be falsified by many examples. For instance, “If Bush is

elected, it won’t be by a wide margin” might be true while “If

Bush is elected by a wide margin, he will not be elected” is

absurd.

3) Antecedent strengthening:

Suppose that my garden is doing reasonably well, but an inch

of rain would make it flourish. Then “If it rains today, my

garden will prosper” seems to be true. But “If it rains today and

a comet crashes on Willington, my garden will prosper” seems

false. Yet the second sentence is a logical consequence of the

first.

4) Denied conditionals:

When someone says, “If a comet hits Storrs, it will be God’s

justice” and you say “No way,” you are not saying that a comet

will hit Storrs. But that is an implication of the negation of

the speaker’s sentence says, understood truth-functionally.

b) “Connection”:

The basic problem with the truth-functional conditional is

that it requires no connection between antecedent and consequent



in order to be true. There is general agreement among theorists

that a conditional makes some kind of claim about a real

connection. The antecedent has something to do with making the

consequent more likely, at least. A precise expression of this

idea is that we usually assert a conditionals “If P, then Q” only

when we take the conditional probability of Q given P (Pr(Q/P))

to be high. The intuitively appealing idea is the “Ramsey test”150

which is, “Add `P’ to your beliefs; see how that affects or ought

to affect belief in Q.”151

Two strategies emerge when this intuition is accommodated:

First, a theorist can try to separate truth from assertibility,

and explain the feeling that connection is required as some

pragmatic effect. Second, a theorist can drop the idea that

conditionals have truth-values, and instead argue that

conditionals have assertibility-values. This second strategy

seems to its proponents to be forced on them because they

strongly hold the intuition that “If P then Q” is only assertible

when the conditional probability of Q given P is high enough and

150 Ramsey (1929), footnote page 247.151 One problem with the conditional probability analysis of

conditionals is that there is already a straightforward way of

saying that the conditional probability of Q given P is high,

namely the normal, wide-scope reading of “If P, then probably Q.”



higher than the probability of Q by itself and are unsatisfied

with attempts to separate truth from assertibility.

The first strategy for defending the truth-functional

conditional was proposed by Grice (1967). His notion of

conversational implicature is the idea that maxims of

conversation shape interpretation. A maxim he proposed was

“assert the stronger claim.” The idea was that just as it is

misleading to say “A or B” when you know that A, so it is

misleading to say “If A, then B” when you know either than A is

false or that B is true. What you have said is true, but violates

conversational norms. But the only condition in which you would

know or believe “either not-A or B” without knowing or believing

either not-A or B is when you know there is a connection between

A and B.

Grice seemed to think that a series of such rules of

conversation would explain the lack of connection between truth

and assertibility that thinking that the English conditional is a

truth-function would require. But there are counter-examples to

this rule about saying the stronger.

If you measure “strength” as degree of belief, there are

acceptable and assertible conditionals whose strength is not

significantly greater than the consequent’s strength. “If a comet

strikes my classroom today during class, this will be the last

meeting of the semester,” is assertible and presumably true,

while “If a comet strikes my classroom today during class, we



will still have the final exam here,” is not assertible and

false. Yet in both cases, the probability of the conditional

being true is only infinitesimally larger than the probability of

the consequent being true.

Frank Jackson (1979) proposes the notion of “robustness”

which he construes as an addition to the “meaning” of “if…then…”,

akin to “but”’s addition to “and.” “Robustness” is a relation

between information and propositions-for-a-person. Two sentences

can be equally strongly believed, but one is robust with respect

to some information while the other is not. So why assert P->Q

when we are almost as sure of Q as we are of P->Q? We want to

indicate that (-P v Q) is still something we believe even if Q is

false. Numerous problems arise if robustness is taken to be

“conventionally implied” by “if…then.” In brief, for a variety of

reasons, the implication of connection between antecedent and

consequent of conditionals does not seem to work like “but” or

“nevertheless.”

The second alternative, abandoning the idea that

conditionals have truth-values, has prominent adherents.152 The

thesis is that a conditional is assertible just in case the

probability of the consequent, given the antecedent, is high.

Given Lewis’ proof that no conditional can be true just in case

the conditional probability of antecedent given consequent is

152 For instance Edgington (1995) .



high, the view has to be that conditionals do not have truth-

values. What do they do, then?

Conditional probability is (usually) an evidential relation,

the relation a sentence bears to another sentence given

background knowledge. So, conditionals express one’s confidence

in q given p, and are appropriate when one has high confidence in

q given p, but do not assert that one has that confidence. This

means that conditionals are subjective—a conditional that is

assertible by you may not be so by me. We all have different

background knowledge, so conditional probabilities vary from

person to person.

An illustration and apparent confirmation of this

perspectival feature of conditionals is Alan Gibbard’s (1981)

“Riverboat” situation, where two conditionals, “If he called,

then he won” and “If he called then he lost” both seem to be fine

and if either true or false, both true. If conditionals are

accepted in virtue of high conditional probabilities of the

consequent given the antecedent, conditionals are subjective.

Generally, if the antecedent is false, there will be adequate

evidence for both the truth-functional conditionals (A->C) and

(A-> -C), since this is just evidence that neither (A&C) nor (A &

-C) are both true.

c) Truth-functions after all?

Notice that in Gibbard’s example, there is no contradiction

between the two conditionals, if they are read truth-



functionally. Gibbard in fact has a proof that any conditional

that satisfies three very reasonable conditions is a truth-

function, if the conditionals are propositions with truth-values.

The reasonable conditions are that (p ->(q->r)) is logically

equivalent to ((p/\q) ->r);153 that the truth-functional

conditional is at least never false when the conditional is true,

i.e. that no conditional (p->q) is true when p is true and q is

false; and that if q logically follows from p, then p ->q is also

a logical truth.

Given the apparent failure of attempts to explain the

separation of assertibility from truth that interpreting

conditionals as truth-functions requires, the conclusion seems to

be that indicative conditionals do not express propositions with

truth-values. This is about as radical an account as there could

be. For instance, it would require a non-truth-conditional

semantics,154 since truth-conditions will not be applicable to 153 That this is reasonable does not mean it cannot be denied at a price. See

Vann McGee (1985).154 Theorists might not be disturbed by this, since they might think that

questions and commands require a non-truth-conditional semantics anyhow. But

questions and commands have an obvious Davidsonian account: Commands are

sentences presented for the purpose of having them be made true. “Yes-no”

questions are sentences presented for the purpose of having the hearer tell

you its truth-value. “Wh”-questions are open sentences presented for the

purpose of having the audience supply a satisfier for the open sentence. The

grammatical categories imperative and interrogative are dealt with in Davidson

(1979), to which I have nothing to add.



conditionals. It is therefore worth trying to mount another

defense of the truth-functional conditional as correct semantics

for sentences using “if.”

I need to highlight two features of the account I am

proposing:

First, one of the Davidsonian strictures on semantical

theories, which he shares with Derrida,155 is that it is bad

procedure to exclude “etiolated” or “marginal” uses when

constructing a theory of a word or kind of expression. The

clearest expression of this view is in his discussion of the

theory of naming, is in Davidson (1993). There Davidson

criticizes “causal” and “baptismal” theories of naming because

they fail to account for fictional characters and fail to deal

with cases where there is clearly naming, but where it is

pretended that “the names are changed to protect the innocent,”

while the author and reader know who is being named.

Second, there is no reason whatsoever to think that “if” and

“then” somehow combine to make a single unit in logical form.

Prima facie, it would be amazing if “if” and “then” were a

semantical unit, since, as we will see, “if” occurs in contexts

where “then” cannot accompany it without seriously distorting the

communication, and “then” obviously occurs in many environments

which have no connection with anything conditional. The account

155 Derrida makes an exactly similar point against J. L. Austin’s theory of

speech-acts in Derrida (1977).



to follow will treat “if” as meaning “if” and “then” as meaning

“then,” a resumptive pronoun referring back to the event in the

previous clause and placing its clause “after” it, in some sense.

The Davidsonian stricture on ignoring marginal cases (1993)

implies that a semantics for “if” should seek a single, unified

account of sentences using “if.” If we take this project

seriously, the account has to cover all the actual uses of “if,”

not just the alleged “central” ones or the “conditional” ones

that happen to interest us. Several kinds of uses of “if” have

generally been treated as marginal, etiolated, and definitely not

the “if” we philosophers are theorizing about.

First, there are the “non-conditional” conditionals. For

example, consider “There’s some beer in the cooler, if you want

one,” “If you must know, Fred and I have been seeing one another

for months,” “George doesn’t love Susan now, if he ever did,” and

“Fred had a pressing engagement, if you know what I mean.” This

is only the beginning of a long list of kinds of uses of “if”

that seem to have nothing to do with connection or conditions.

Note, by the way, for future reference, that these “if”s do not

have an accompanying “then.” Note also that, since the “if” seems

to be just an add-on to the “consequent” clause, they are all

truth-functionally acceptable. The only controversial case would

be the case where both “antecedent” and “consequent” are false.

If I say “Fred had a pressing engagement, if you know what I

mean” and Fred has nothing on his calendar and you have no idea



what I’m insinuating, there is some inclination to say that the

whole sentence is false. However, there is also the inclination

to say that, as it happens, the sentence is true, just as “If

Fred is a scholar, I’m a monkey’s uncle” is true, as discussed

below. It is of course also true that it is difficult to imagine

a communication situation in which a speaker knowing that the

antecedent was false and the consequent was false, would non-

misleadingly say the sentence.

Note that the non-conditional conditionals can be understood

as connection-implying conditionals. If there is a beer fairy in

my house, who accommodates my guests by telepathically reading

their desires, and then magically satisfying them, but my guest

Fred does not know this, “There is beer is the fridge if you want

one” can be meant as a connection-implying use of “if.” In that

case, “then” would be appropriate, but not required, for reasons

we will discuss below. Or, adapting one of Lycan’s (2001, page

191) examples: “Frijhof Boeger will be at the conference, if I’ve

spelled it right.” It could be that there is some connection

between spelling the name right and the guy showing up, but that

is so unlikely that we don’t ascribe such a belief.

Another, smaller category of problem conditional is so-

called “joke” conditionals, such as “If Bush is a statesman, then

I’m a monkey’s uncle.” These are generally acknowledged to be

truth-functions, but excludable as marginal, rather than central.

They are, after all, jokes. But how they can even work as jokes



requires explanation. Someone who takes Davidson (1993), not to

mention Derrida (1967) seriously will be suspicious at this

exclusion of the “marginal.” The idea that semantics is like

physics, so that simple cases (balls rolling down inclined

planes) will provide the laws to which “marginal” cases can

eventually be assimilated, is suspect. There is no reason to

think that the relationship between a beer-in-the-fridge

conditional and a causal conditional is like that between balls

rolling down inclined planes and trees falling in a dense forest.

What a defender of the truth-functional conditional needs is

an account which explains the disconnection between assertibility

and truth for many conditionals, while accommodating all the uses

of “if.”

d) Separating truth from assertibility

Grice’s (1967) and Jackson’s (1979) proposed explanations of

why truth-functional conditionals could lead an interpreter to

expect that the speaker thought there was a connection between

antecedent and consequent. They looked for principles of

interpretation that would handle all the cases. Lists of

principles seem not to be adequate. Davidson thought of

interpretation, not as an algorithm or set of principles to be

applied to cases, but as rationalization. Communication-

situations may vary in many ways. The factors that may be

relevant in understanding what another has said are difficult to

enumerate. This does not mean that there is nothing systematic to



say about how we go about interpreting another, but rather that

we should not expect there to be an algorithm governed by strict

principles..

A parallel to Davidsonian interpretation is induction. A

great deal of successful theorizing has been done about how

conditional probabilities should change in the light of new

information. There is the whole field of statistics and

probability theory. But no one expects to have an inductive

logic, in the sense of an algorithm that determines for arbitrary

pairs of sets of sentences S and individual sentences P, the

conditional probability of P given S.

Davidsonian interpretation, in the case of conditionals, is

a generalized and extended use of the strategies of Grice and

Jackson. An interpreter is presented with an “if”-sentence, and

asks “Why is he saying that?” The interpretive task is to

understand why the person presents a given sentence with those

truth-conditions. The “why” is not a part of the meaning, but a

part of interpretation of actions of truth-value-presentations.

For Davidson, these two sides of interpretation are simultaneous.

Given that we have an hypothesis about the reason the person is

presenting an utterance, we arrive at truth-conditions. A gesture

toward a buffalo accompanied by an utterance suggests that the

truth-conditions of the utterance have something to do with a

buffalo. Given that we have a good idea what the truth-conditions

of an utterance are, we can hypothesize about the person’s



reasons and the force of the utterance. When a teammate says

“Good shot” after my airball, the utterance is likely to have

been intended sarcastically.

Davidson’s idea is that you do not need a system of rules.

Rather, you need to suppose that the other is a rational agent.

So when a person presents, for instance, a disjunction, it is

often reasonable to suppose that he doesn’t know which disjunct

is true, given that the person is trying to communicate. In the

same way, most of the time, if someone asks me “Do you know where

the Budds building is?” and I say “Yes,” I’m being obtuse or

philosophical. Other times, I may be being quizzed by someone who

doubts that I know my way around campus. One way to be obnoxious

is to respond to “Can you reach the salt?” with “Yes.” “Can you

reach the salt?” though, said by my physical therapist, when I am

recovering from shoulder surgery, could be a question about how

my therapy is progressing. We don’t need “conventions” here, just

normal skills of interpretation.

Many interpretation-scenarios are repeated again and again,

and interpretation becomes close to automatic. Such routine

interpretation can feel like part of the meaning of what is said.

The interpreted message is part of the speaker meant to convey by

the speech-act. The intended interpretation, though, is not the

same truth-conditions of the sentence uttered for that

communicative purpose.



Accounts of various problem examples then amount to

explanations of why a hearer would interpret the presentation of

a truth-function in a given way, and why a speaker could expect

the hearer to do so. “Conversational implicature” and

“robustness” explanations will be accurate accounts of the

reasoning for many such explanations.

On the present account, most of the work done by “semantics”

for conditionals is to be assigned to interpretation—a person

presents a truth-value and the interpreter has to figure out why

the person presented an utterance with those truth-conditions.

Interpretation ascribes a reason for the person’s remark. Given

that the presenter wants the interpreter to get it right, and the

interpreter knows this, much of the understanding that takes

place is extra-semantic.

Interpretation explains problem 4) above, why denials of

conditionals do not seem to work truth-functionally. Denials are

not always of the truth-conditions of what is said, as Jackson

points out. If you say “I believe it will rain,” and I reply “No

way,” I am not casting doubt on your accuracy in reporting your

mental state. Denials can be of what is “signaled,” to use

Jackson’s term. In the same way, when someone says, “If a comet

hits Storrs tomorrow, it will be God’s justice” and you say “No

way,” you are not saying that a comet will hit Storrs. You are

denying the implied connection, which implication interpretation



explains.156 The sentence is almost certainly true, but the

implied connection almost certainly does not exist, given the

piety of people in Storrs.

e) Tense and conditionals

A not-always noticed feature of proposed counterexamples to

the thesis that the English conditional is a truth-function is

that problem indicative conditionals seem always to be future

“tense.” The English “will” is in fact a modal auxiliary, but the

same holds for French, which actually has a future tense. In

languages with still-functioning moods, modal distinctions are

represented by subjunctives, optatives, and the like.

My account conjectures that the future tense, the past

tense, and tenses other than the null present tense are at least

akin to modals. The various tense logics make this abundantly

clear.157 The future tense speaks of what is not (yet) in

existence. The future is at least akin to the possible, the might

be, and such modal notions. My account proposes that “if” is

basically a marker of two-place modal and tense predicates of

things said. An occurrence of “if” says “first argument of a

modal or future tense predicate coming up.” When there is no

modal or future tense there, but the marking takes place, the

“if” is the null-modal, the truth-functional conditional, to

156 Ramsey (1929) page 239 makes the same point about denied conditionals. 157 See Burgess (2009) chapter 2 for a nice treatment of that field.



which interpretive strategies will need to be employed,

typically. When there is a “then” to indicate that some kind of

sequence is intended, the “if” still marks a pure truth-

functional conditional, an add-on to a clause, as in “non-

conditional conditionals,” and the “then” adds the idea of a

sequence. The claim of sequence between the arguments then

suggests, on interpretive grounds, that the first argument has

something to do with the second.

On the present account, “will” is a two-place predicate of

things said. Something similar happens with “will” as happens

with “probably” and “ought.” In “If A, then probably B,” there is

a “wide scope” reading and a “narrow scope” reading. I put these

characterizations in scare quotes because they presuppose that in

“If...then…probably” sentences something is either governing a

conditional or its consequent. On the present account, that is

not what is going on. The two readings are a “probably” modality

with first argument being the apparent antecedent and second

argument being the apparent consequent, and a truth-functional

conditional with first argument being the antecedent and second

argument being a “categorical” probability. “Categorical” or

“absolute” probabilities I take to be something like “all things

considered” probabilities, which are epistemic, like “all things

considered” “ought”-sentences.158 158 As I will argue in Chapter 9, “ought” is primarily an epistemic modality, calculating what it is reasonable to believe, want and do. The theory of “ought” I will argue for as an alternative to



So, just as we can say “John is probably in New Hampshire by

now,” taking the tacit antecedent to be contextually given or

common knowledge, so we can say “It will rain tomorrow” taking

the tacit antecedent to be contextually given or common

knowledge. A pure prediction would be like an absolute modal or

like an unconditional desire—a two-place relation with a

tautology as the first argument. Pure predictions that it will

rain just say that relative to anything, it will rain. In the

same way, apparent unconditional desires are desires relative to

anything; and pure necessity claims are claims that relative to

anything, this is necessary.

We can see the difference the future tense makes by

replacing the future reference in our earlier examples with

present or past:

1) With the future, “If Wheeler is our Dean, then the Earth will

open up and swallow his enemies” sounds false, while “If Wheeler

is our Dean, then the Philosophy Department will have many new

appointments,” sounds true. But both of the pair “If Wheeler is

our Dean, then the Earth opened up and swallowed his enemies” and

“If Wheeler is our Dean, then the Philosophy Department has many

new appointments,” seem strange, but not false. Note that, the

“metaphysical” theories about “ethically accessible worlds” and such will be a rationality-based account. If we took “categorical” probability to be about what is true, all probabilities would be either one or zero.



actual temporal understanding sense of the antecedent, when the

consequent is future, may be itself is future, as well as the

temporal sense of the consequent. This should lead us to think

that “will” is two-place, as I argue below that it is.

2) Contraposition examples are less than persuasive to begin

with. Consider “If Bush is elected, it will not be by a wide

margin.” Notice that “it” is anaphoric with something, namely an

election. So, the logical form of this original sentence, taking

“B” to be “is an election which Bush wins” and “W” to be “is a

wide-margin election,” would be either \/x( Bx -> -Wx) or \/x(Bx

-> (Bx/\-Wx)). The contrapositives of neither of these are

paradoxical. “\/x( Wx ->-Bx)” would say that if it was a large-

margin win election, then Bush doesn’t win it. And “\/x(- (Bx/\-

Wx)->-Bx)” says that if either Bush doesn’t win or the election

is won by a large margin, then Bush doesn’t win.

3) Antecedent strengthening again depends on the future modality.

Suppose that my garden is doing reasonably well, but an inch of

rain would make it flourish. Then “If it rains today, my garden

will prosper” seems to be true. But “If it rains today and a

comet crashes on Willington, my garden will prosper” seems false.

Yet the second sentence is a logical consequence of the first. If

“will” is a two-place modal predicate, then this paradox would be

no more troubling than the paradox that if Fred is only twenty,

he probably lives at least another forty years, whereas it is

very doubtful that if Fred is twenty and is facing a firing



squad, he probably lives at least another forty years. The “will”

seems to be at least akin to a modal that predicts the

“consequent” relative to the “antecedent,” more or less in the

way that “if…then…probably…” assesses the “consequent” relative

to the information in the “antecedent.” In both cases, all the

work is being done by the modal or the tense. The “if” is doing

nothing special, except indicating the first argument of a two-

place predicate of things said. “If” is like the “either” in

“either…or” in this respect.159

f) “Then”

It should seem obvious that “then” is an independent

component of the alleged “if…then…” connective. “Then” has

independent occurrences outside conditional contexts, and seems

to be a demonstrative akin to “there.” So, we have sentences

like, “First sauté the garlic, then add the basil,” and the like.

“Then” indicates a place in a sequence, whether a temporal

sequence, a sequence of argument steps, or the trivial fact that

the second clause is second. Consider the following conversation:

“Bill is coming, along with several friends.” “Then we had better

buy another case of beer.” The “then” indicates the next place in

an inferential sequence. The “sequence” that “then” indicates may

be pretty minimal. When I am asked, “Who is in your department?”

I may say, “Well, there’s Lionel, Baxter, Lynch, Paul and Bontly

on that side of the hall. And then there’s JC, Hallie, Marcus,

159 “Men…de” is something like this in Greek—that is, punctuation for clauses..



Elder and me. And on the first floor there’s Bill and Austen.”

The sequence is just sequence in this list. This kind of trivial

sequence seems to be a way of marking that the current utterance

is a continuation of the previous utterance.

When “then” occurs with an “if” some kind of sequence is

implied, which interpretation must supply. An “if”-sentence such

as “If you want a beer, there’s some in the fridge,” will, as

noted above, require a different interpretation if “then” is

inserted, giving us, “If you want a beer, then there’s some in

the fridge.” With the “then,” the sentence seems to imply some

kind of sequence with your desire as first element and the

presence of beer in the fridge as second. What kind of sequence

could this be? There are lots of possibilities, all of them

unlikely, and most of them involving some kind of causal relation

between the desire and the presence.160

“Sequence” can be any kind of ordering. In “If an integer is

prime and even, then it is two,” the sequence is inferential, but

not causal or temporal. This “then” is acceptable even though

there is in fact no priority of one clause over the other

inferentially.

160 Suppose someone says “If you want a beer, then there’s some in the fridge” knowing that nothing beer-fairy like is the case. In this case, the “then” is the trivial “second place in the sequence of clauses.” Since the “then” is pointless, the sentence is odd and misleading, but not false.



“Then” is optional in conditional remarks using “if.” “If an

integer is prime and even, it is two,” is fine. “If you open the

window, you cool off the room” and “If you open the window, then

you cool off the room” say much the same. The presence of a

“then” indicates some kind of sequence, perhaps the trivial

sequence of being the second thing you said. But the sequence

may be supposed even without the “then.”

g) Interpreting “if…then…” sentences

Besides Davidsonian rationalizing interpretation, by which I

understand a kind of generalized Grice and Jackson, using “if”

without a modal and with a “then” strongly demands an

interpretation according to which there is a non-trivial sequence

between the first and the second. “Then,” that is, strengthens

the interpretive inclination to ascribe to the speaker a claim

that the truth of the antecedent has something to do with whether

or not the consequent is true.

h) Counterfactual conditionals

In Chapter 1 I characterized truth-definitions as

“supporting counterfactuals.” More generally, a mark of the

natural laws that provide essences is that they support

counterfactuals. So, I should say something about

“counterfactual” conditionals. According to my account of

conditionals, there is no special counterfactual “if…then…” that

is distinguished from the indicative conditional. Rather, such

markers as “would,” the past tense of “will,” in the consequent



and either a past or past perfect in the antecedent indicate one

or more “counterfactual” modal relations between things said. I

do not have anything like a theory of the truth-conditions of

such modal predicates. A sufficient condition is that the

conditional be an instance of a law. However “is a law” seems to

me to be just another way of saying that the second argument is

necessary, given the first.

I don’t have a theory about what to say about “If Caesar had

been in charge, he would have used catapults/atom bombs,”161 nor

about “If three were even there would be two even primes.” About

the general run of counterfactual conditionals which seem to be

dependent on what the shared assumptions of the conversation are,

I can observe that the same features affect the truth-conditions

or interpretations of “necessary,” “can,” and “impossible.” My

inclination would be to treat the actual utterance “You cannot

physically get from here to Willimantic in less than five

minutes” as literally false, given F-16s, but interpreted with

“background” as antecedent so as to be understood as true.

Virtually every theory of counterfactuals does something

similar.162

161 Quine (1960) page 222.162 See Iatridou (2000), Dehghani, Iliev, and Kaufman (2012), for instance. A Davidsonian should be happy to applaud linguistic work on the truth-conditionsof counterfactuals, given that a Davidsonian semantics has nothing to say about the truth-conditions of the “counterfactual” or “subjunctive” modal predicate. These are interesting and plausible theories about the truth-conditions of the predicate.



VII Modality and “if…then”

Here is the conclusion: Modals are two-place predicates of

things said. Many modals have pretty good theories about what it

takes for that modality to obtain. “If” marks the first argument

of the modality. A conditional modality is just a modality with

both of its arguments explicit.

Since what “if” does is mark the antecedent of a two-place

modal, what is going on when there is no modal? An “if” with no

modal, that is, just connecting two clauses, marks clauses in the

“null” modality. The null modality is the truth-functional

conditional. This explains why the various “non-conditional

conditionals” are still truth-functionally all right. It also

explains why “then” is very odd in such uses of “if.”

“Probably,” “ought,” and other modals are not further modals

on top of the null modal. Rather they are the modal whose first

argument is being marked by “if”. For such modals, typically,

“then” is appropriate because there is some relation, expressed

by the modal, which makes the first argument have something to do

with the second. For a few modals, the extreme modal predicates,

it will in fact work out well to pretend that the conditional

modality is the categorical necessity of the null modal applied

to its arguments. But that is an accident. Even they are in fact

two-place modals which actually have nothing semantically to do

with the truth-functional modality. The satisfaction-conditions

of the particular predicates make it turn out that they could be



understood as a one-place categorical modal predicate applied to

a null modal.

The semantics for modalities can thus be very simple: Each

modality is a one- or two-place predicate of things said.

Chapter 6: Properties, Propositions, and Facts

As we saw in Chapter 1, Davidson does not employ the

metaphysical notion of “meaning” in giving his account of

linguistic meaning. Propositions, as understood by philosophers

to be the entities expressed by sentences in use, thus have no

part in his account of language. Similarly, he does not analyze

predication in terms of an object instantiating a property. The

notion of a property as a component of facts is of course ruled

out by his arguments that sentences do not refer. Likewise, he

rejects facts as referents or truth-makers for sentences. He

accepts a proof, whose avoidance requires major constraints on an

underlying logic, that there is at most one fact, if facts are

truth-makers for sentences.

In consequence, Davidson never develops a semantics for

sentences using these concepts. There is no Davidsonian account

of properties, propositions or facts. However, any Davidsonian is

committed to saying something about what such objects are. A

Davidsonian is committed to the idea that most of what most

people say is true. There are truisms about properties,



propositions and facts. It is true that my brothers and I share

many properties. It is a truism that if my brothers and I are all

bald, there is a property we share. It’s true that that’s a fact.

There are at least some propositions you and I agree on, for

instance, that the present year is after the seventeenth century.

Modalities, as we noted in Chapter 5, often seem to be

predications about properties, propositions, and facts. So,

properties, propositions and facts must all be something,

according to Davidson.

So, a Davidsonian philosopher has two projects:

a) Find something innocuous that property-talk, proposition-talk,

and fact-talk could be about. Find what it is that we posit, when

we posit properties, propositions and facts.

b) Argue that these innocuous somethings are all that the data

support, that they have no metaphysical role in making

predication possible, supplying meanings for sentences, or

supplying truth-makers for sentences, and that the truths a

Davidsonian has to acknowledge about properties, propositions,

and facts are truths about them.

I) General Thesis and objections

Since the accounts of these three kinds of objects will be

so similar, I will sketch the general kind of account, discuss

objections to this kind of account that could be or have been

raised, and then go on to say something about each of these kinds

of objects. Briefly, the Davidsonian ontology for properties,



propositions and facts is that they are the same quasi-linguistic

posits he proposed in Davidson (1968), the items appealed to in

the account of modalities in Chapter 6. Briefly, again,

properties are “ascribables,” “things that can be said” which are

open sentences rather than sentences. Propositions and facts are

posits of sentential “things said.” Subject-predicate

propositions are ascriptions of ascribables. Facts are states of

the world that some ascription is true. Facts have some peculiar

problems, but are posits of the same innocuousness as properties

and propositions.

Different views about the identity of properties,

propositions, and facts coincide very nicely with different

standards for “saying the same thing” in indirect discourse. Thus

intuitions and examples supporting various metaphysical theses

about sparseness of properties are accommodated by the theory.

Someone taking a “fine-grained” approach to properties,

propositions or facts is someone requiring a different standard

for accurately representing “what Fred said.” Whether “Aristotle

is a Greek” attributes the same or a different property or

expresses the same or a different proposition or states the same

of a different fact from “Aristotle is a Hellene” depends on the

purposes at hand. There is considerable latitude in judging

reports of what someone has said as true or false: I’ve asked

Fred to cover a class and ask a colleague to pester Fred to get a

response. Fred may have actually said, “What a pain. But I owe



Wheeler one. So I guess I’ll do it.” The colleague’s true report

of what Fred said may be “Fred said OK.”

Identity conditions in various contexts for things said

match the variety of identity conditions for various “grains”

ascribed to properties, propositions, and facts. “Lois denied

that Superman is Superman” is a false report of her words “Clark

isn’t Superman,” Marcus’ report that I said I would give a talk

in a fortnight is true even if I said “I will do it in two weeks”

even if I think a fortnight is a period of four days. However,

Thales said that everything is basically water, but he didn’t say

that everything is basically di-hydrogen oxide.

All three posits in effect start with Aristotle’s

characterization of predications as “legomena,” things said. If

we expand the things said to include all the things sayable, we

need an account tying things said to language, but not to any

particular language. Davidson’s(1968) account treats the “that”

as a demonstrative, and regards the content of the subordinate

clause as an item demonstrated in the sentence, but not part of

the sentence. Just as when I say, “When I tried her Banana Tofu

Surprise I couldn’t help doing this,” my facial expression is not

part of the sentence, so when I say “Pascal said that it is

rational to be Catholic,” “It is rational to be Catholic” is not

part of the sentence. What makes what Pascal said and what I

wrote above the same ascription is that each is an acceptable

interpretation of the other. So my sentence and Pascal’s sentence



ascribe the same property to being a Catholic and express the

same proposition, by most standards of saying the same thing.

A description of what a proposition is in other terms would

be the following: A proposition, relative to a standard of same-

saying, is a collection akin to an equivalence class of possible

utterances. The relation is “same-saying.” “Says the same as” is

not quite an equivalence relation, since it is intransitive. A

picture of these classes in a space of possible utterances would

be a set of overlapping circles, each centered on an utterance.

How far from a given possible utterance another possible

utterance would have to be to be saying the same thing depends on

the purposes at hand, as discussed in Chapter 5.

A further reason a proposition would not be an equivalence

class is that the same-saying relation is sometimes

indeterminable due to indeterminability of interpretation. An

illustration of how it can be indeterminable whether two

utterances express the same proposition, even given a single

standard of “saying the same thing” might be the following:

Suppose Euclid and Dedekind both say, “Two is the only even prime

number.” Euclid takes the numbers to be the natural numbers.

Dedekind takes the numbers to include the reals, etc. They are

positing different sets of objects with their terms “number.” On

the other hand, the only numbers that are prime are natural

numbers. Are they thinking the same thing, if a proposition is a

content of thought? Or take Saussure’s(1977, page 115-116)



example of “sheep” and “mouton.” Francois says “Charles est un

mouton,” Frank says “Charles is a sheep.” As Saussure points out,

“mouton” has a different extension from “sheep,” covering mutton

as well as intact animals. However, with the indefinite article,

it can only apply to an animal, since it is being treated as a

count noun.

Possible utterances are (mostly) abstract objects generated

by truth-definitions for idiolects. Since a truth-definition

supports counterfactuals about what possible utterances would

have meant if they have been uttered, these possible utterances

reflect those counterfactuals. It should be obvious how this

account applies to properties and facts. Properties are things

ascribable, open-sentence-like entities, while a fact is a state

of the world that a thing said is true.III Objections and replies to the general thesis

The objections to follow are primarily ones that have been generated

from my colleagues, since there are no duplicates of precisely the theory I

propose here to which the literature has responded, at least that I know of.

a) Objection 1: These theories treat the world as linguistic.

No. That “is a frog” is true of an object says a lot about the object.

If the entity is such “is a frog” is true of it, then it’s a frog, hops,

cheeps in the spring, etc. Joe being a frog has consequences that are not

linguistic consequences, but consequences about what a part of the world is

like. It is true that these consequences must be reported in language, but

that does not make them linguistic. The proposition that Joe is a frog, if

true, likewise has consequences for the world, as does the fact that Joe is a

frog being a fact.



In each case, by the present account, the truth-definition says how the

world has to be for a property to be instantiated by an object, what the world

has to be like for a proposition to be true, or for a fact to be a fact. The

whole point of the truth-definition is that the world being a certain way is

the condition of a linguistic item having a semantic property or relation.

b) Objection 2: There are more propositions, properties and facts than there

are things people have said.

On the present theory, every possible sentential utterance-type

expresses a proposition. Possible utterance-types are abstract objects of a

familiar sort, the sentences and predicate-phrases generated by the syntax of

an idiolect. It is fairly easy to come up with propositions that have never

been expressed and properties that have never been ascribed. A syntax, coupled

with a recursive semantics, will, after a few iterations of a familiar

construction, produce sentences, i.e. possible utterances, that have never

been uttered, certainly not by someone speaking your idiolect. A truth-

definition for an idiolect generates what the truth-conditions of an infinity

of utterances would be. Truth-definitions support counterfactuals.

c) Objection 3: Propositions, properties, and facts are prior to

language.

This is an objection one hears a lot. “Since there were no

humans when tyrannosaurs were alive, there were no sentences

then. But then `Tyrannosaurs are fierce’ would not have been true

then, since the sentence did not exist then. But it was true. The

property of being fierce, likewise, was something tyrannosaurs

had, even though there were no things being said about them. So,

eternal objects such as propositions and properties must be the

truth-bearers and that in virtue of which things have features.”

As I briefly mentioned in Chapter 1, this argument would has



little appearance of plausibility if “is true” is conceived of as

a one-place predicate of sentences and “is true of” is a relation

between predicates and entities. If “is true” is a predicate of

sentences, sentences about the existence of tyrannosaurs are

true. If “is true of” is a relation between predicates and

things, then my predicate “is fierce” is true of tyrannosaurs.”

While it is correct that there were no sentences in existence

then, it is not correct that “There are tyrannosaurs” was not

true then. “It is 1,000,000,000 BCE” is a sentence of ours which

was true for a year back then. The case is analogous to puzzles about the fact that Aristotle might not

have been called “Aristotle.” There may be no possible world in which speakers

can truly say “Aristotle is not named Àristotle.’” Nevertheless, there are

worlds in which “Aristotle is not named Àristotle’” is true. “Aristotle is

not named Àristotle’” is true in those worlds even though that utterance or

inscription, interpreted from the point of view of that world, that is, with

that world’s references for terms would not be true. The utterance is mine and

is to be interpreted with this world’s references.

In the same way, in the distant past there were no speakers saying

“There are tyrannosaurs” but my inscription or utterance, “There are

tyrannosaurs” was true. In both cases we are “evaluating” our utterance

relative to another framework, either another possible way things might be or

another time. Our English sentence “There are tyrannosaurs” was true a while

ago but that does not mean that our sentence existed in those bygone days.

d) Objection 4: Expressions of a language are enumerable, but the true

propositions are not.

Consider any real number, say pi. For every other real number n, either

it is greater than pi or smaller than pi. In general, \/n\/m (n≠m -> (n>m v



n<m)) . Here are 2א things true of n, i.e. properties of n, and so

propositions about n, for every real number n. But the expressions of English

are denumerable. The real numbers are not. So propositions can’t be just

things constructed out of linguistic expressions. Analogous arguments for non-

denumerable pluralities of properties and facts run the same way, with only

slight adaptation of wording.

Response d1) There are a non-denumerable infinity of expressions

in English.

This is an argument made by Postal and Langendoen (1986):

Consider a grammar for a fragment of English S: S is the smallest

set such that: 1) “Cabernet is a wine” ε S. 2) If X ε S then “I

believe that” ͡ ͡ X ε S. S is clearly isomorphic with the positive

integers, and so denumerably infinite. S contains only sentences

of finite length. Add one more syntactic rule to expand S, namely

3) If X and Y ε S, then X ͡ “and” ͡ Y ε S.

From 3), if X is a sequence (an ordered -tuple) of elements

of S, then there is a conjunction X’, also in S, the order of

whose elements follows the order of X. S thus has the cardinality

א א . This set, after reaching cardinality 2א as it were, then

takes those elements as items to be grouped into subsets, and so

on. Rule 3) creates a set of expressions for every subset of

elements of S. S has elements of infinite length, and in fact an

infinity of them. Some of the conjunctions (an infinity) are such

that each conjunct is of infinite length.Consider the conditional whose antecedent is some conjunction of the

totality of subsets of S (a conjunction whose characters have the cardinality of



the continuum) and whose consequent is “Cabernet is a wine.” It has to be a

logical truth, since every model on which the antecedent is true is one on

which the consequent is true. This is so, even though the antecedent has

length 2 א.

It may be objected that such strings are not sentences, because they

cannot be produced by humans. Whether they are sentences or not, they

certainly are linguistic objects. So, there is an expression representing the

distance between √2 and π. Here is a finite segment of that expression: “The

distance between √2 and π is 1.7273790…..” The same holds for all the other

reals. Call a d-property a property of being a certain real distance less than

a given number on the number line. So, you can say things like “For any d-

property D, if Dn and Dm, then n=m.” You can quantify over properties and

relations. There are enough properties, that is, ascribables.

Response d2: The possible linguistic expressions are not clearly denumerable

The theory does not claim that, for every proposition and every

idiolect, there is an expression in that idiolect which expresses that

proposition. Aristotle, for instance, could not express the proposition that

protons are composed of quarks. Rather, the theory claims that for every

proposition, there is a possible idiolect and a possible expression of that

idiolect which expresses that proposition. Properties are sets of possible

open sentences, propositions are sets of possible closed sentences.

What are the possible languages? We have an expression for pi, so we can

say things about pi. For every real number, there is a possible idiolect with

an expression for that number. Even if we require that the language be

English, there are possible idiolects in which pi refers to some other real

number. Suppose the speaker has heard the expression “pi” but has no idea that

it has anything important to do with circles. He thinks it’s an irrational

number whose first three thousand digits are 3.1415…….. By reasonable

principles of interpretation, we could assign any of a non-denumerable

infinity of real numbers as what that speaker means by “pi.”



Objection e) There are properties we are aware of that the ancient Greeks

could not ascribe to things. There were propositions and facts they could not

express. So, we should expect that there are properties future people will

ascribe to things that we cannot ascribe. Possible people might ascribe still

other properties and express still other propositions. If properties are

ascriptions, and propositions and facts are things said, how could they have

been there before there were people able to ascribe them or express them?

Reply e1) Those properties did apply to things classical times, just as

propositions about protons were true in classical times.This is a form of the

same objection above, that there were true propositions before there were

language-users. Of course. We are describing classical times and ascribing.

Reply e2) Supposing that rational beings continue to exist and improve their

science, properties we cannot ascribe are truly ascribable. If the ascriptions

those future people in fact say are true, those properties exist now. If they

say things that are true, but not practically useful to us or them, yes. If

they are silly, yes. (That is, “If what they said is in fact said, it is

said.”)

Reply e3) Everything that could be ascribed, whether truly or not, by possible

language-users can harmless qualify as a property. Here is an analogy: Feet,

ells, paces, meters, and the like are all acceptable units, and I’m happy to

quantify over them all. For certain purposes, meters are much better than

paces. 163 What about units that could be used?

Consider the meter. In1984, the Geneva Conference on

Weights and Measures defined the meter as the distance light

travels in a vacuum in 1/299,792,458th of a second as

measured by a cesium-133 atomic clock. Change any one of the 163 Why are paces relatively crummy for physics? It could happen, even for a single pacer, that 1/290th of a distance was more paces than 1/289th of that same distance, measured again. The same problem arises for hands, cat-noseleathers and paws.



digits of the denominator of that fraction, and you get a new

unit. The French decided to make the meter equal to 1/107 of

what they took to be the distance from the equator to the

North Pole, but lots of other choices were possible. So,

there are an unlimited number of units just as good as

meters, with other rational values for the number of seconds.

Any physical laws that can be expressed in the metric system

can be expressed just as well with these measures. There are,

of course, also a lot of relatively crummy (for physics)

possible units.Are those units, as it were, out there waiting to be discovered? We do

quantify over meters, but that does not show that there are joints in space a

meter apart. If the French had not been so fixated on the base-10 system, and

had instead used the Babylonian system or a hexadecimal system, physics would

be untouched. But the units of those unused metrics are only already there in

space in an innocuous sense that should not make us think that space is

crowded.

In the same way, from the neo-Davidsonian perspective on ontology, there

are indefinitely many kinds of being that could be posited. Just as there are

an infinity of ties for first in a physically useful metric, so there are very

many alternative posits that will supply singular terms and general terms for

articulating the world in equally effective ways, for any conception of

“effective.” Except “useful for us.”

As discussed in Chapter 3, since we are posits ourselves, and posit

ourselves as agents and language-users, the posits we can actually use must

retain those units. Radically alternative posits are not posits of language

users, so are not possible posits at all. But that says nothing about nature’s

joints and a great deal about us.



II) Properties

There is a sense in which Davidson rejects properties.

Davidson rejects properties as metaphysical explainers of

predication. As discussed in Chapter 1, according to Davidson,

“`Is a frog’ is true of a if and only if a is a frog” is all

there is to say. Appeal to properties does not illuminate

predication. On the other hand, Davidson must acknowledge that

frogs are instances of the property of being a frog. Either there

are properties or else all positive ascriptions of the term

“property” are false. So, a property is something.

Whatever properties are, whether metaphysical components of

facts or what, with a subject, makes it true that an item has a

property, there is at least one kind of object that at least

encompasses properties. A kind of entity whose existence everyone

has to accept is “something ascribable to objects.” A

metaphysician like Armstrong may not think that every ascribable

is a property, but it is hard to see how ascribables could be

nothing at all. Some of the ascribables are complicated things.

Some of the ascribables are even impossible things, such as being

the guy who found an integral solution for n greater than two to

an equation of the form an = bn + cn , where a b and c are

integers.

The ascribables are abundant. Anything can be ascribed to an

object—any adjective phrase of any complexity can be used to

ascribe something to an object. So, whatever a philosopher’s



metaphysical equipment, she has to also supply some kind of thing

for ascriptions to ascribe.

An ascribable cannot be a particular linguistic item, since

Francois and Fritz can ascribe the property of being a frog

without speaking English. However, they are somehow connected

with language, since the predicative expressions of a language

exactly coincide with the things ascribable by that language.

The departure of the account of properties from the accounts

of propositions and facts is the extension of Davidson’s (1968)

idea to non-clauses. Davidson’s idea can be extended to a number

of predicates which seem to create non-extensional contexts. The

comparative adjectives “good,” “expert,” and “skillful,” for

instance, as discussed in the Chapter 9, have the peculiarity

that when co-extensional predicates are interchanged, truth-value

may not be preserved. Syncategorematic predicates behave oddly as

well. Even though Gucci purses are the only high-end purses

manufactured by a company founded in Florence in 1921, a fake

Gucci purse is not a fake high-end purse manufactured by a

company founded in Florence in 1921.

Constructions which generate expressions for terms for

properties out of open sentences, such as “F-ness,” “F-ity,” “F-

itude” and the general and flexible construction “being -F”

demonstrate an utterance, the thing said, do so without an overt

demonstrative. Expressions such as “the property of baldness”,



“the quality of mercy” and the like are then count-nouns in

apposition with these expressions.

Notice that some of these nominalizations make explicit a

kind of mass-noun understanding of the application of these

predicates. “Tallness,” “aptitude,” and “stupidity” admit of

degrees, by which I understand that there are quantities of such

features. That is, “John is tall” can be paraphrased as “John has

tallness,” that is, John has the abstract property of being tall,

or as “John has some tallness,” where “tallness” now is

functioning as something like a mass-noun. This mass-noun is

actually the primary way of understanding comparative adjectives,

as we will argue in Chapter 9. Tallness, as a quantity of which

there can be more or less, is a convenient kind of entity over

which to quantify.

The lack of an overt demonstrative should not trouble us.

Thus, just as “John believes the Godhead is a Trinity” is

perfectly good without the demonstrative “that,” so the predicate

“good” demonstrates a thing said by being followed by the mention

of a predicate. So, Chapter 9 will argue that “Fred is a good

nail-pounder” is something close to “Fred is a good that.`Nail-

pounder.’” Since Fred can be a good nail-pounder without being a

good carpenter, the comparative adjective “good” cannot be being

applied to a set. The same holds for “expert,” “adept,” and



other words.164 “The property of being…” and “the property of

being a…” likewise create contexts where an expression strictly

outside the sentence is demonstrated.

Properties, qualities, attributes and the like are posits

with very loose identity-conditions-- quasi-equivalence-classes

of sentence-types in idiolects in circumstances. “What was said”

reported on the witness stand has different standards than in

normal conversations, for instance. The quasi-equivalence

relation is, as usual, “samesays.” (This is now theory and not

semantics.)

That “samesays” is intransitive should not be disturbing.

The relation “is the same sentence-type as” is also intransitive.

Two physically identical marks may be of different types,

depending on the intentions of the maker. In general, relations

that are usable may have what would seem to be anomalies. For

instance, “larger than” seems to be a well-ordering relation

among physical objects. For any two objects, A and B, either A is

larger than B or B is larger than A or they are equally large.

Rigel has a diameter equal to 74 solar radii; Mira’s is 320 solar

radii. But Rigel has a mass equal to 18 solar masses, whereas

Mira is 1.8 solar masses. Which is larger? Of course, for

purposes of physics, we refine “larger than” into the two

relations: “has greater mass than” and “has more volume than.”

164 “Better than” and “more expert than,” unlike “taller than,” appear to be relative to a class. Chapter 9 will explain this.



But sentences like “Fred is larger than Bill” are still sometimes

true. And there is such a thing as largeness. Analogous remarks

would apply to many of the two-place predicates we apply in

ordinary talk. Just as a physicist can propose sharp relations as substitutes for her

purposes for ordinary terms, so a metaphysician can propose sharp criteria for

“same thing said.”165 Other refinements a theorist might propose are a notion

of property restricted to things said that are predicates of true physical

laws, or some other condition treating some ascribables as real properties and

others not. A metaphysician could claim some subset corresponds to causally

efficacious objects. Whatever well-defined notion a theorist adopts for

metaphysical purposes, such as constructing a theory of predication or giving

a metaphysics, that theorist must accept as well objects corresponding to the

generous notion of “thing said”, since, whether or not those things are all

properties in some strict sense, they are things—things believed of objects,

for instance.

On the account I propose, the properties just are the ascribables. So

properties have nothing to do with any metaphysical account of predication.

How could a quasi-equivalence class of abstract objects make anything be an

amphibian or have a negative charge? Every predicate clause has the form “`F’

is true of an object a iff F’a,” where “F’” is a predicate of the interpreting165 Some choices: 1) A predicate F says the same thing as a predicate G iff \/x(Fx<->Gx) is logically true.2) A predicate F says the same thing as a predicate G iff \/x(Fx<->Gx)is true in virtue of physical laws.3) A predicate F says the same thing as a predicate G iff \/x(Fx<->Gx)is true.4) A predicate F says the same thing as a predicate G iff \/xPr(Gx given Fx) and \/xPr(Fx given Gx) are both high. 5) A predicate F says the same thing as a predicate G if and only if F=G.



language and “F” of the language being interpreted. That is all there is to

say about predication. On Davidson’s account of predication, the regress

difficulties with treating predication as a relation holding between the

property and the object beyond the relation “is truly ascribed to,” vanish.

There’s nothing more basic than “is true of.”

Other questions evaporate or change their form as well. Without having

properties as metaphysical components of things being true, what could be the

import of questions about whether substance dualism or property dualism in the

philosophy of mind was correct? From a Davidsonian point of view, these are

discussions about predicates.

IV Propositions and Facts

At a first glance, it might seem that a Davidsonian should

say that propositions are things said and that facts are true

propositions. This will not quite work, as we will see. Like

“property,” “proposition” and “fact” are count-nouns that can

occur in apposition to something indicated by “that.”

a) “That”-clauses and other nominalizations of sentences

A “that” clause is usually treated as one among several

devices by which closed sentences are nominalized, that is,

turned into referring expressions, nouns, as it were. So, for

instance, infinitive phrases can do the same work as “that”-

clauses. “Fred knows that Joe is a criminal” and “Fred knows Joe

to be a criminal” say the same thing. Likewise, it is difficult

to find a difference between “Fred believes that Joe is a



criminal” and “Fred believes Joe to be a criminal.”166 In the

infinitive construction, there is no overt demonstrative.

One of the ways in which Davidson’s treatment differs from

others is that the “that”-clause as a whole does not designate

anything, since it is not a whole. The “that” is a demonstrative

referring to something that is not strictly part of the sentence,

but is rather an utterance which indicates something. What

exactly the “that” indicates depends.

“That” is multi-purpose—the demonstrative may refer to any

of a number of objects that are around. So, I might make a

gesture and say “Take that.” In response to a body-part donation

request, “that” might demonstrate my forearm. In response to an

insult, “that” might demonstrate a punch involving my forearm.

“That” can be interpreted as demonstrating entities in many

categories, while directly demonstrating a performance. For

example, the “that” in “That was published 40 years ago,” “That

has a dirty cover,” and “That is a crummy edition,” refers to a

work, to an instance of a work, and to an edition of a work.

166 When Joe is made part of the sentence, as in “Joe is believed by Fred to be a criminal” we have the problem of how to deal with “quantifying In” (Kaplan 1968). This is analogous to the problem of quantifying into “said to be” contexts. There is a difference between “John is said to be corrupt” and “The embezzler of the Wilpond fortuneis said to be corrupt.” If all that has been said is, “Whoever embezzled the Wilpond fortune is corrupt,” and John is the man, we cannot say that John is said to be corrupt. But if John is the Wilpond’s accountant, and the authorities say, “The Wilpond’s accountant is corrupt,” then John has been said to be corrupt.



In the cases at hand, as we’ll see, “that” always

demonstrates an utterance, some immediately following words. But

how those words are to be interpreted will depend on much else.

“That” always demonstrates an utterance, but that utterance can

be taken to determine various entities, such as propositions or

facts. The “that” perhaps indicates a proposition. Perhaps a

fact. The contents of “that” clauses don’t refer at all. That is,

while “that” refers to the succeeding utterance, the utterance

itself does not directly refer to any kind of entity. Sentences

do not refer. But presenting a sentence as what Fred believes or

knows or is happy about can indicate some entity to be the second

argument of a relation of which Fred is first argument. “That”

will sometimes be interpreted as referring to an entity

determined by the contained utterance.

On a Davidsonian account, many other nominalizations are

treated in the same way as “that”-clauses. Just as the overt

demonstrative is optional in “Willard believes Ortcutt is a spy,”

so it is omitted in “Willard believes Ortcutt to be a spy.” A

Davidsonian semantics treats all “nominalizations” as

demonstrated expressions. So, infinitive phrases that complete

propositional attitude verbs are likewise demonstrated in the

speech-act. Why those expressions are being demonstrated varies

from case to case. Various objects are what the indicating

speaker intends.



In addition to nominals that are complements of

psychological verbs and adjectives such as “true,” there are

nominals which designate events and states. The destruction of

the Roman Empire, the Roman Empire’s having been destroyed, and

the destroying of the Roman Empire all are nominalizations of

“the Roman Empire was destroyed.” Such nominals seem to be

descriptions of events or states. The Roman Empire’s having been

destroyed is a state of the world that came into existence, has

endured for over a millennium and will last forever. The

destruction of the Roman Empire and the destroying of the Roman

Empire are events that have duration, but are past. Such nominals

are relevant to the theory of facts, below.

Do all propositional attitudes, such as “wish,” “imagine,”

“think” state a relation between an individual and a proposition?

Only in some cases can “proposition” occur in apposition with the

nominal. “I think the proposition that I have a Ferrari” and “I

imagine the proposition that I have a Ferrari” are both at least

very odd. “I fear the proposition that a bully will be elected,”

seems especially odd, since the proposition is already in

existence. We would like to say that what is feared is that,

namely a bully being elected, a possible state. Perhaps we should

say that what are feared are threats. What we can say is that in

all such cases, the “that” indicates a set of same-saying

possible utterances of which the demonstrated utterance is a

token. Propositions are also the directly demonstrated objects of



linguistic attitudes and predicates, such as “agrees,” “says,”

“declares,” and the like. In “John agrees that Fred is a frog,”

the “that” points to a thing said.167

I do not have a theory about why some predicates allow

appositives and other predicates do not. I also do not have a

theory of what the various kinds of objects with propositional

content might be, if we were positing special objects of fear and

hope, for instance. The variety of English “propositional

attitude” predicates is much too large to discuss them all here,

since the questions of this chapter are primarily about ontology.

However, one kind of difference among complement-taking verbs is

salient. I turn to that difference now.

b) Factives

Among the expressions that are completed by “that”-clauses,

some create intensional contexts but still allow the inference to

the content of the “that” clause. Such constructions, the

factives, seem not to be talking about propositions at all. For

167 The question may arise about what sorts of objects “wh” complements take, as in “Fred wonders whether it will rain” and “Fred wonders who is at the door.” These seem to be propositions and properties, respectively. For a Davidsonian, “yes-no” questions are presentations of sentences with the purpose of finding out their truth-values, if they are real rather than rhetorical questions. “Wh”-questions are presentations of open sentences with the usual purpose of finding out an argument that makes the resulting sentence true. Wondering whether Fred is a frog is an attitude toward the proposition that Fred is a frog of wishing to know its truth-value, roughly. Wondering who is at the door is wondering what the property of being at the door is true of.



instance, knowing a proposition is not knowing that a proposition

is true. Before anyone knew whether Fermat’s Last Theorem was

true, every philosopher needing an example of a sentence that was

either necessarily true or necessarily false knew the proposition

that there are no integral roots of an + bn =cn for n>2, where a,

b and c are also integers. Now, some mathematicians at least know

that there are no integral roots of an + bn =cn for n>2, where a,

b and c are also integers.

“Knows that,” “is happy that,” and numerous other

constructions support the inference from the sentence to the

content of the “that”-clause. Some but not all of these

constructions happily combine with one or both of the count-nouns

“fact” and “state of affairs.” 168 “John knows the fact that Fred

is a frog,” is fine as is “John is aware of the fact that Fred is

a frog.” “John is happy the fact that Fred is a frog,” though, is

distinctly odd.

While some kinds of nominalization may yield propositions or

facts, other nominalizations yield facts rather than

propositions. “John’s being a frog amused Fred” is only true when

John is a frog. With “believes” the expression can only be “Fred

believed in John’s being a frog.”

c) Facts

168 Another count-noun that occurs with “that”-clauses, “case,” in the pleonastic “It is the case that Fred is a frog,” does not seem to occur with such verbs.



I will begin by discussing some reasons why the obvious

idea that facts are true propositions will not quite work. First,

there are some anomalies when we substitute “true proposition”

for “fact” in the case of many verbs. We discussed “know” above.

Second, facts seem to be causes in many cases. The fact that the

wire was not grounded caused the fire. In “The fire started

because the wire was ungrounded” a state of affairs seems to be

cited as a cause. In “That the wire was ungrounded explains why

the fire started” it might be thought that the true proposition

“the wire was ungrounded” will do as well. But it is not the true

proposition, but rather the proposition’s being true that really

explains. Somehow, a proposition’s being true must be an entity

of some kind.

The account I will propose is roughly that facts are states

of the world. The fact that Fred is a frog is the world’s being

in a state such that Fred is a frog.

There are several desiderata a Davidsonian account of facts

should meet:

1) There should be more than one fact. Some pair of true

propositions should correspond to distinct facts. That is,

whatever facts are, they should somehow prevent the Slingshot

argument from being applied. In effect, this means that

substitution of co-extensive terms in sentences reporting facts

need not preserve truth-value.



2) Facts should not involve mysterious non-relational bonds169

holding among their components. Facts are not combinations of

entities which account for predication or which are named by true

sentences.

3) There ought to be a fact corresponding to every true

proposition. In particular, a theory should accommodate negative

facts and general facts. It is certainly a fact that there were

dinosaurs in Connecticut and that Wheeler is not a dinosaur. A

Davidsonian theory of facts accommodates such facts effortlessly.

This desideratum requires some defense. The defense is a

follows: If “know” is a factive, that is, a relation to a fact,

then for any pair of facts A and B such that it is possible for a

person to know that A and not know that B, then the fact that A

is different from the fact that B. So the one-to-one correlation

between facts and true propositions seems secure.

4) True propositions should correspond to facts, in the sense

that there is a 1-1 function from true propositions to facts. So,

if someone wants to be a “correspondence theorist” about truth,

169 Non-relational bonds are required in order to stop the Third Man Argument. If the instantiation relation between an individual and a property is a relation, then there is a fact that this relation obtains between the individual and the property. Otherwise, there would be nothing distinguishing the mere co-existence of the three entities from the individual instantiation the property. But the obtaining of that relation is a three-place relation. The quadruple of that relation, instantiation, the individual, and the property exists whether or not the instantiation relation obtains between individual and the property. And so on.



that is fine. This is the sense of “correspond” which motivates

Tarski to say that his account of truth is a correspondence

theory. It is obviously not a correspondence theory in the

metaphysical sense, that true predication needs an explication in

other terms.

5) If the fact that Fred is annoyed exists, it should have to be

true that Fred is annoyed.

6) Facts should not be entities that supply the metaphysical

ground for sentences or propositions being true. “`Fred is a

frog’ is true if and only if Fred is a frog” supplies an adequate

metaphysical ground.

Here’s a simple theory that meets these conditions, and is

available to anyone: Facts are states of the world. That is, a

given fact is a property of the world as a whole. It is a fact

that Fred is a frog just in case the world is such that Fred is a

frog. Facts are specified by the complements of “the fact that…”

expressions, and are the referents of factives. The “content

clause” specifies the state of the world. Roughly, the world is

such that the demonstrated proposition is true. More exactly,

that clauses of factives demonstrate states.

There is a distinct fact for every true proposition. To see

that this is that case, notice first that at least some of these

states must be distinct, since different states have different

modal properties. The fact that Fred is a frog could exist

without the fact that Fred is green existing. So, those two



states cannot be identical. One could perhaps argue that

necessarily equivalent sentences should determine the same

states. It might seem plausible, that is, that there is only one

state of the world corresponding to “Two is even” and “The cube

root of eight is even.” But given that “know” is a factive, if it

is possible for anyone to know the first but not know the second,

they are distinct facts. If belief is required for knowledge,

then since two propositions are distinct if one can be believed

but not the other, these states of the world exactly correspond

to the true propositions.

The states that are facts might seem to be strange objects.

It is hard to see why. Given that the universe is such that it is

true that a transit of Venus took place in 2012, why is that not

a state of the universe?. It might be asked whether such a state

is a physical state. Given the fact about the transit, there are

clearly physical states of Venus, the Sun, and the Earth. Also,

there are physical events in which they are participants. For

every subject-predicate truth about a physical object, there are

states of that object or events that happen to that object.

“Being green” is a state of Fred, the green frog. But “Fred is a

green frog” does not refer—sentences are not referring

expressions. However, there is a state of the universe, that the

proposition that Fred is green is true. That’s a physical fact.

There are, of course, non-physical facts, such as that 11 is the

cube root of 1331.



States are generally conditions of smaller entities than the

whole of existence. They are a kind of entity intimately

connected with events. For any event, there will generally be

states of the participants in that event. When Fred becomes

green, there is a time when Fred is in the state of becoming

green. When it is true that Fred is green, there exists a

temporary corresponding state of Fred, his being green. Once Fred

becomes green, he will be in a state of having been green as long

as he lives, whatever his future color history. States of

entities typically have an onset, a period of existence, and an

end-point. There are, of course, eternal states of eternal

entities. The number two’s being prime, for example, has always

been a state of that number.

Facts are states of the whole world. Some states of the

whole world, the facts about Fred,170 are temporary, some come

into existence and stay in existence from then on, and some are

eternal. The fact that Fred is green lasts as long as Fred is

green. The fact that Fred has been green continues in existence

forever once it comes into existence. Given that it is acceptable

to posit a tenseless “is,” the fact that Fred is green on October

4, 1974 is an eternal fact, a fact that has always existed. We

will see in the next chapter that there are no consequences for

170 Being “about Fred” is not a notion about which much can be said beyond something like “Fred is the grammatical subject.” That “Joe hitFred” is about Joe, while “Fred was hit by Joe” is about Fred seems tome to show that aboutness is not a semantical concept.



the Philosophy of Time from the existence of such eternal facts,

since these states amount to nothing more than the truth of the

content proposition.

At this point, one could wonder how, since facts are

essentially constructs out of linguistic items, there could be

facts which have existed from eternity. After all, on many

accounts of the universe, there have not been language users from

eternity. Once again, this worry is an artifact of truth-maker

theories. Let us consider the facts about the universe of twelve

billion years ago, rather than eternity, which has its own

issues. Many of our sentences were true at that time, not to

mention sentences in other possible languages which say the same

thing. For a fact to exist at a time is just for the proposition

expressed by the sentence to be true. Facts as we have

constructed them do not make anything true. Rather, they are made

to exist by what is the case, the true sentences, ultimately.

It is perhaps worth noting that what one would take to be

the natural interpretation of “states of affairs,” the

alternative term often used for “facts,” agrees with my

understanding of facts. What are the “affairs” of which a

particular state of affairs is a state? The “affairs,” I take it,

are things in general, the world. So, a state of affairs is a

state of the world, the way the world is. If facts and states of

affairs are the same things, then the fact that Fred is a frog is

the state of the world that the proposition that Fred is a frog



is true. The focus of so many thinkers on “truth-makers” of

various kinds to explain why truths are true may explain why this

conception of facts has remained hidden despite its overtness.

Since facts are states, and states, by the considerations in

Chapter 4, are often the relata of causal relations, facts can be

causes. So, sentences like “The fact that the building was full

of flammable material caused the fire to spread quickly” are

perfectly fine, as is “The building’s being full of flammable

material causes the fire to spread quickly.” States of

individuals can be causes and states of the world can also be

causes. A neo-Davidsonian has more than just events available to

be arguments of a two-place “cause”-relation and for describing

adverbial modification.

Facts will, at least prima facie, be as numerous as the true

propositions. They are distinguished from one another by being

derived from distinct propositions. Those distinct propositions

are given in what follow the “that,” the demonstrated utterance

or inscription. The content clause specifies a state of the world

in exactly the way that the content clause of “Joan is worried

that John will be early” specifies Joan’s worry.171 171 If a metaphysician wishes to have a conception of facts that identifies what are prima facie different facts as the same, the identity conditions of such postulated facts will be as tight as the identity conditions for “saying the same thing.” That is, whether the fact that Fred is below Susan is the same as the fact that Susan is above Fred will be dependent on context in the same way that the question whether in saying that Susan is above Fred you were also saying that Fred was below Susan depends on what standards are in



So how do such facts meet the desiderata?

1) Such states of the world are distinct entities if there are

distinct propositions, which there are.

2) Since facts are specified by the complement of a “that”-

clause, there are as many facts as there are utterances that say

different things. That’s more than one.

3) Since the facts don’t have components at all, are not “made”

of properties and particulars, no bonding or anything else is

required, except that there is a possible language which

expresses that fact.

4) Facts that are puzzling from the Tractarian, combination of

entities with properties conception are not puzzling at all.

Negative facts are states of the world specified by negative

utterances. General facts are states of the world specified by

general sentences. Future facts are states of the world as well,

about which more in the next chapter. Briefly, the world is

either now in a state such that “Some humans will be alive in

4000CE” is true or not. If we posit a tenseless “is,” the

universe is also either in a state such that “Some human is alive

in 4000CE” is true or not.

5) A specified state of the world will only exist when the

specifying utterance is true. The state corresponding to “Fred is

a frog” is “the state that Fred is a frog,” where the “that” is

place.



in apposition with “state.” This guarantees that all and only

true utterances/sentences correspond to facts. 6) It will also be true that the existence of a fact guarantees that the

corresponding sentence is true. However, “The fact that Fred is green makes it

true that Fred is green” should be understood in the way that “Fred’s being a

grandfather makes him a father” is understood. Fred can be a grandfather only

if Fred is a father, but the “metaphysical ground” of Fred’s being a father is

not Fred’s being a grandfather. So, “correspondence” and “making true” don’t

have anything to do with the theory of truth.

On this theory, the “identity conditions” of facts are the

same as identity-conditions of states. So, the identity-

conditions are not so clear, in many cases. Just as the identity

conditions for “same-saying” are context- and interest-dependent,

so the identity condition of states of the world specified by

“same-saying” are context- and interest-dependent.

7) To the extent that utterances and possible utterances, that

is, ascriptions, are acceptable and innocuous, facts are

acceptable and innocuous. Everyone, Davidsonian or not, can help

themselves to these facts.

Facts, according t construed in this way are not referents

of sentences and they are not truth-makers. Rather truths make

them.

Chapter 7: Future Contingents and Temporary Intrinsics

This chapter is an illustration of metaphysical questions

being dissolved by the Davidsonian truth-maker-free account of



truth and predication. Whereas a Davidsonian is committed to

there being something to say about modalities and the nature of

propositions, properties and facts, puzzles about time are

different. One puzzle about time seems to be purely an artifact

of the truth-maker theory of truth. Another puzzle, about the

possibility of enduring objects, seems to rest on taking

properties to be metaphysical attachments rather than things said

of objects.

The passage of time has posed metaphysical difficulties for

metaphysicians for a long time. This chapter will discuss two of

those difficulties, the problem of future contingencies, which

leads some to non-classical logics analogous to those inspired by

the sorites, and the problem of temporary intrinsics, which calls

into question the possibility of objects undergoing change while

continuing to be the same object. The first problem is discussed

in Aristotle’s De Interpretatione, Chapter 9. The second problem

arose in response to Heraclitus’ position, as understood by Plato

and Aristotle, that nothing can survive change. Arguably,

Heraclitus’ problem shaped important parts of the metaphysical

systems of these two thinkers, and so of the rest of Western

philosophy, as we discussed in Chapter 2. Recently, new

Heracliteans have emerged with radical reconceptions of the

nature of reality designed to accommodate the Heraclitean

argument. This chapter argues that radical solutions to these two

problems rest on misconceiving the semantics of predication and



of tense, and that a neo-Davidsonian approach to truth,

essentialism, being and semantics can solve both problems by

essentially endorsing common sense.

Many of the disputes in the philosophy of time are nearly

incomprehensible from the point of view of at least this

Davidsonian. The arguments among philosophers responding to

McTaggart’s (1908) by advocating for either the A-series or the

B-series, that is those favoring the objectivity of “now” and

those denying it, are about whether there are properties that are

expressed by tense. Given the previous chapter’s exposition of

properties, the answer will be trivially “yes,” but in a way that

neither side would deny. It may be that the claim of the B-theory

advocates is that indexicals like “now” and predicates like

“present”, “future” and “past” are like “phlogiston,” just

mistaken posits which are never truly applied. It is hard to see

how this last suggestion could be the case, since inferiority of

some predicates to other predicates covering the same ground is

not generally a disqualification for being true of things. Given

that “larger than” is an acceptable predicate even though it does

not well-order physical objects, and given that “is earth” is

true of objects even though chemical kinds are well established,

it is difficult to see how temporal indexicals could be

disqualified. That is, suppose it can be shown that a system

without the indexicals but using dates instead is better suited

to many purposes. That “greater in mass” and “greater in volume”



are better dimensions that “larger than,” does not show that

nothing is larger than anything. In the same way, if reference to

times by dates rather than temporal indexicals is more useful in

physics, that does not show that the indexicals cannot be used to

say truths. Without the idea that there is exactly one way the

world is articulated, it is hard to get this dispute going.

Other disputes, those among Eternalists, Presentists, and

those like Tooley(1997) who regard reality as growing as time

marches on, concern the reality of the past, present, and future.

If “a is real” means “There are true positive sentences

quantifying over a,” then it is difficult to see why past and

future objects are not real. It is true of future objects, of

course, that, since they are future, there are no causal links

between us and them, so that we cannot refer to them. On the

other hand, with that deflationary characterization of “real,”

absent a truth-maker theory of truth, there is nothing in the

reality of future objects to support the mysterious idea that

future objects are, as it were, waiting for the present to get to

them. Since I do not know how to engage such views of time and

its passage, given my views about properties and reality, I will

not discuss the ontology of time. I will let the physicists

advise me.

This does not mean that there are no issues about time. A

Davidsonian would not deny that there are some deep mysteries

about the nature of time. There are current cosmological debates



about whether time, space, or both space and time are fundamental

or rather arise from some more basic phenomena. While much of the

discussion is speculative, the speculation is grounded in

empirically-supported theories of what equations characterize the

world. To think that philosophical inquiries into the nature of

time, independent of the physics of space and time, could uncover

substantive truths that might constrain what reality can be like

is expecting a lot from evolution. Why should we expect that

organisms would evolve intuitions from their experience of

duration that would plumb the essence of temporality? That seems

no more likely than the idea that we would intuitively know that

energy and matter are at bottom the same or come up with the

periodic table a priori on the basis of our experience with

material objects.

The idea that there is some special subject matter,

metaphysics, with a domain of problems that must be solved before

an adequate physics is possible is not only incompatible with the

Quinean view that there is a continuum of abstraction from direct

empirical sense experience, but prima facie absurd. A Davidsonian

inherits the Quinean perspective, with an admixture of

Wittgenstein’s respect for natural language. The Davidsonian

project could be viewed as one kind of what Strawson (1959)

called “descriptive metaphysics.” Other projects, revisionary

metaphysics and revisionary logics, find incoherencies in the

judgments of “ordinary language” and propose metaphysical



solutions that solve those alleged problems. The point of view of

this book is that such alleged problems arise from the

metaphysical point of view taken by revisionists, that there is

exactly one privileged partition of the world into objects and

properties, and that a truth-maker semantics of predication

reveals the metaphysical structure of the world.

On the contrary, according to the present neo-Davidsonian

position, the semantics of natural languages makes few

commitments about the issues that metaphysicians are concerned

about. Those issues are the province of investigators who try to

find out what the world is like by observing it rather than

thinking about how it must be.

The arguments to follow are against fatalism, alternative

logics for time, four-dimensionalism, eternalism, presentism, and

so forth. But they are not arguments that these views are

incoherent. The attack is on their motivation. Such views begin

with alleged difficulties in what we might call the “common

sense” conception of time and tense which difficulties the

revisionary account addresses. The arguments below are that the

alleged difficulties with the natural common-sense conception of

what things are like are bogus. The difficulties are artifacts of

accepting truth-maker theories of truth and predication or taking

properties to be metaphysical attachments to beings.

For Davidson, any alleged difficulties with the natural

common-sense conception of the world have to be squared with the



idea that understanding a person at all requires assigning truth

to most of their fundamental beliefs. If there are deep

incoherencies in the natural conception of the world, then either

most of our beliefs are in fact false and we cannot understand

ourselves, or what we really mean is something other than what we

seem to be saying. But if what we really mean is something other

than what we seem to be saying, that is what we are saying.

So, the burden of proof is on someone who claims we are

mistaken in the common opinions that we can be correct when we

guess about future events and that objects last over time. If it

is in fact false that objects can undergo changes and be the same

object, huge swaths of our beliefs are mistaken. Powerful reasons

indeed would be therefore needed to show this. I doubt that there

are such reasons. This chapter surveys the reasons that have been

given and shows them to be inadequate to their task.

I Future contingents

Sentences about the future sometimes have undeterminable

truth-values. “October 14, 3932 will be a Friday” can be known to

be true.172 However, a sentence such as “One of Wheeler’s

descendants will be elected President of the US” is not known to

be true and not known to be false. Philosophers have argued that 172 It may not be absolutely certain. If time ceases to exist before that date

comes around, it may turn out to be false. One could also worry about the

possibility that the Earth encounters a very large asteroid and disintegrates

before 3932 or that its rate of rotation slows significantly enough so that

periods of sun and shadow last a week. What is the date then?



such sentences have no truth-values and provide reasons to modify

our logic in order to accommodate their apparent lack of a

determinate truth-value. Other philosophers accept bivalence and

take the future to be now a determinate reality, albeit a future

reality, and enduring objects to be four-dimensional. Here are

the only two reasons I have encountered to think that future-

tense sentences with indeterminable truth-values raise special

problems:

Ia) The truth-maker argument

The first thought is that if it is in fact true that one of

my descendants will be elected president, the truth-maker making

it true must exist.173 We think that this event might not happen.

But if the sentence has a determinate truth-value, and has it

now, something must make it true or make its negation true. The

truth-maker corresponding to the truth must exist for the

relation to hold.

If truth is a relation to the world, then the relevant parts

of the world have to be there now for truth to be here now.

Here's an analogy: Suppose I am now a grandfather. Being a

grandfather now requires the existence now of my grandchildren.

The grandchildren must be real now if I am now a grandfather. In173 See Richard Taylor (1957) page 3, “The first assumption is a correspondence theory of truth, the minimum requirement of which is that in the case of any true proposition asserting some predicate of a particular individual, there is (tenselessly) a fact consisting of that individual having that predicate.”



the same way, if truth is a relation to truth-making components

of the world then the corresponding truth-makers in the world

have to be there now for truth to be present now. It is

uncontroversial that a sentence that turns out to be true will be

true. If a sentence is true now, though, the future fact that

makes it true must already be in existence, albeit in a future

location. So in fact, if truth is a relation, either the future

is now determined by some existing truth-maker or sentences about

the future are neither true nor false. One might object that,

say, a table setting is already there, but can be changed. But,

if the truth-makers themselves are truth-makers for future-tense

sentences, and they now exist, then changing them would make them

not only not exist, but never have existed.

The argument for the case of the truth-value True as set out

in premises would run as follows:

1) Truth is a relation between a sentence (or proposition or

other truth-bearer) and a truth-maker.

2) For the truth relation to obtain at a time, both the sentence

and the truth-maker must exist at that time.

3) Either:

a) Truth-makers for sentences which will come true now exist, and

those sentences are now true; or

b) Sentences which will come true are now neither true nor false.



This argument gives yet another reason to follow Davidson

and others in junking truth-makers, treating truth as a one-place

predicate, and taking “’It will rain tomorrow’ is true if and

only if it will rain tomorrow” as a complete and adequate

expression of the truth-conditions of “It will rain tomorrow.”

Without truth-makers and without truth consisting of a relation

of sentences to such truth-makers, this truth-maker consideration

for taking sentences about the future to be truth-valueless

collapses. One of the premises, premise 1) is false.

Ib) The range of the quantifier

The form of “One of Wheeler’s descendants will be elected

President of the US” would appear to be (with “D” the “is a

descendent of” relation and “P” the “is elected president of the

US) something like “Ǝx( Dxw /\ Px)” or perhaps “ƎeƎx(Dxw /\ Ee

…)”. That is, the sentence would appear to be a quantification

over future individuals or future events.

Ib1) Future objects?

It looks like we are quantifying over future objects. If the

future is fixed in the sense that the future truth-maker, though

future, is now available to be referred to, future objects

available for reference now would be determinate individuals who

happen to be located at a space-time point whose temporal

coordinate is in the future. Future objects being available for

our reference is a main attraction of the four-dimensionalist

conception of enduring objects. If the future is not fixed,



future objects are inhabitants of select future possible worlds

which, as it happens, turn out also to be actual worlds. Of

course I have very many distinct possible descendents who get

elected, even if it never happens that one of my descendents is

elected. So, the quantifiers have to range over possible future

objects, and somehow select the actual future objects, the ones

that actually come into existence. If there are now many ways

things could turn out but only one way things do turn out, the

truth-conditions of quantified sentences about the future (if

they are either true or false) require possible future objects to

be available now, and a special feature of the possible objects

that become actual.

What a Davidsonian should be interested in is the semantics

of tense. This puts the Davidsonian into the camp of those who

“take tense seriously.”174 What is characterized as “temporal

logic” are various attempts to systematize the strong

resemblances between the tenses and the modal predicates. My

Davidsonian appropriation of such attempts is as follows: I like

almost all of them, and am happy to adapt them by treating modal

operators as predicates of things said. Most temporal logics seem

to be very interesting theories of the truth-conditions of

predicates of things said. My only difference from any such

theories is that I think they are theories and not proposed

174 A pioneer in this regard is, of course, Arthur Prior, in numerous

publications, for instance, Prior (1968)



logics or semantics. Reading the literature on the varieties of

tense, we find there are various difficulties and counter-

examples to theories. I am happy to have the specialists on

English tenses figure out a theory of how these predicates are

inter-related. A Davidsonian can appropriate tense logic

seamlessly, treating operators as predicates of things said, as

usual. That is, the same reinterpretation of modal logic that

turned quantifications over possible worlds into predicates of

things said turns tense-operators into predicates, and leaves the

law-like relations among those predicates to theorists, rather

than semanticists.

I should mention perhaps just one way in which a Davidsonian

appropriation dodges certain issues. The relation between “John

is happy” and “There will be a time at which John was happy,”

from the point of view of this book, is not logical consequence,

but a theorem of the theory of a modal-like temporal predicate.

The past component of “was” is a predicate, not a part of

structure. In the same way, the automatic inferences that seem to

be part of logic from the point of view of temporal logicians are

from the present point of view, the results of knowledge of laws

connecting the temporal modal predicates.

One well-worked out way to conceive of the future along

these lines while allowing that the future is not yet determinate



is Storrs McCall’s(1976) conception of branching worlds.175 The

picture very roughly is that the future consists of infinite

branching physically possible worlds, where each moment in the

future corresponds to a node from which a number of physically

possible branches proceed. As the present moves into the future,

all but one of the branches from the node at which the present is

located are eliminated, and the past includes only the one

remaining node. The future at any moment consists of a vast

number of possible ways things could turn out, all but one of

which get eliminated from reality.176 Each of these ways is, as it

were, populated by possible beings.

This way of picturing time is a version of a realist

conception of possible worlds, restricted to what is physically

possible and further restricted by a particular starting point.

The future of a given moment is a set of possible worlds all of

which share a common past, and so have all past and present

objects as common parts. The possible worlds are the possible

paths through the tree. So a single possible world is a sum of

successive moments. Depending on what sort of physical state the

initial point is, various alternatives exist at every moment. All

but one of these is “eliminated” at each moment. “Eliminated”

means “become merely formerly possible.” Things that could happen

175 This is an early one of many papers McCall has written touching on this topic. 176 On McCall’s conception, these ways have different probabilities, dictated

by the truths of quantum mechanics.



become things that could have happened. Thus “elimination” would

seem to be more “demotion,” since things that could have happened

could still available to be referred to.

On McCall’s conception of possible worlds, the same possible

individual inhabits many possible worlds. Your actual course

through time is one of several of your possible courses through

time. McCall thus manages to capture the Kripkean intuitions that

alternative histories are histories of individuals with a

realist conception of worlds. So, if a particular individual gets

elected under many different circumstances (as will certainly be

the case) those distinct possible worlds will be ones in which

that individual is elected.177 The passage of time from the past into

the present and into the future makes all of these objects real,

but demotes alternative presents and futures deriving from

demoted paths. At any point, there is only one actual past and

present, but all the alternative pasts and presents, which had

once been accessible, are no longer accessible.

McCall’s branching theory of the passage of time has

branching individuals each of which inhabits (is a part of) many

possible worlds. At any time, individuals have branching futures,

but pasts that consist of exactly one actual branch and lots of

once-but-now-possible paths. As time passes, all but one of an

177 There are other, distinct descendants who could be elected. When

those individuals either do not come into existence or die without

having been elected, they still could have been elected.



individual’s branches at a moment become former branches. At the

end of its existence, a possible individual will consist of a

single path.

For many purposes one could upgrade “extinction” to the

modal feature “used to be possible.” An individual’s paths that

used to be possible will still be (extinct) possibilities that it

once had. At any point in time when a possible individual exists,

though, it will have many possible futures, all but one of which

will eventually be “used to be possible” futures. McCall’s

individuals, since they are elements of branching paths

This would yield two grades of possibility: The still-

possible worlds include the world that will eventually be

selected at the end of time, but the population of still-possible

worlds changes as time passes. In fact that is what it is for

time to pass. At the end of time, the population of still-

possible worlds has been pared down to one. But the population of

once-or-still possible worlds is constant. The possible worlds in

the once-or-still possible sense at the end of time will be

single collective paths through the tree, but those sums of

objects share components with many other paths. So possible

worlds have shared parts.

So the branching individuals, as we would hope, undergo

change as time passes. Consider an actual object on its deathbed.

It is still a branched object, as it has always been. Many of its

branches have changed from still-possible to used-to-be-possible.



So it has, now, these modal properties. It, the total branched

individual, is at every point an element of many possible worlds.

The possible worlds of which it is a part are themselves branched

objects, where the branching starts at the point of origin of the

branched individual.178

If we consider the whole branched individual, everything

that happens to that individual on any branch is something that

could have happened to the individual. Many of these complicated

modal features will be chains of conditionals—if I had dropped

out of high school and become a safe-cracker and been caught and

served time, then I could have a tattoo. The important point for

my purposes is that all of these modal features are features of

the actual individual. Looking ahead, McCall’s picture of an

enduring individual gives a model of what I call below the “CV”

of an individual.

McCall’s way of picturing time makes the quantification

easy. We are quantifying over possible individuals and, in the

quantification, requiring that they be in the selected world.

“Selected” means “turns out to be actual.” Of course, many, many

178 An option for a David Lewis charmed by McCall’s picture would be to treat “actual”, “still-possible” and “used-to-be-possible” as indexicals as Lewis does, so that each path is actual but onlyfrom its point of view. That is, the erasure of alternatives as one moved through time would be perspectival. Then the branched individuals would be a single branched object with no objective distinctions among the branches about which is more real.



future possible objects are descendents of mine who are elected

president. The quantification will come out true if one of those

possible people is at some point an element of every possible

future. So, “A descendent of Wheeler will be elected president”

is “There is a possible object which is a descendent of Wheeler

and is in a selected world and is elected president in that

selected world.”

I think McCall’s theory of time might be correct. It seems

to capture the flow of modalities that characterizes the passage

of time. However, it is hard for me to believe that this theory

is imbedded in the semantics of tense. On the other hand, just as

(as I would advocate) the connection between “Not all frogs are

green” and “Some frogs are not green” is not strictly part of

semantics, but a part of set-theory, so with the tenses. The

inference from “Fred was a frog” to “`Fred is a frog’ was once

true” will be a theorem of a theory, not a logical truth.

1b2) Ontology-free tenses and modality

As the reader will remember from Chapter 5, on modality,

there is at least one non-reductive alternative to possible world

semantics. Here is the analogous view applied to tenses: The

tenses should be treated as akin to modals, and treated in the

unilluminating (perhaps) way that a Davidsonian can treat

modality.179 McCall and others are correct to see a strong analogy

179 The future tense in English is a modal. In languages with future tense verb

forms, my claim is that the tenses are modal-like, in that they are, like



between tense and modality. McCall has a very plausible theory of

what it takes for temporal (and other) modal sentences to be

true. But McCall’s account is part of physics (or, if you insist,

metaphysics, since his essays on how to accommodate special

relativity with quantum mechanics are compatible with a great

many particular physical theories). Just as it is not part of

semantics that a truth-condition of “Zoe is a Labrador retriever”

is that Zoe is a dog, so the existence of possible worlds is not

part of the semantics of “It will rain tomorrow.” In my terms,

McCall’s account is a theory of tenses. The formal agreement I

have with McCall is that he recognizes that tenses are modal-

like. He happens to have a particularly interesting theory of

these quasi-modals, but that is not really relevant to the

semantics.

My proposal is that, just as a Davidsonian can treat

accounts of the modals as parts of theory and not part of

semantics by taking modals to be semantically primitive

predicates of things said, so can a Davidsonian account of tenses

re-phrase any tense-logic180 as an account of the truth-conditions

of predicates. Among the tense-logics that deal with the

multitude of of detailed differences within and among the ways

modals, predicates of things said. 180 Following Davidson (1967a) page 32 I regard tense-logics and not really

being logics at all.



languages deal with temporal passing, I do not have a preferred

candidate.

So, if we think of tenses as akin to Davidsonian modals, the

semantics tells us relatively little. Tenses are predicates of

things said, just as “possibly” and “necessarily” are.

Davidsonian primitivism applies to tense predicates as well as to

modals. This means that the apparent quantification in “One of my

descendants will be elected president” does not have the form of

a quantification, but rather is a modal predicate applied to a

quantified sentence. “There is someone who might celebrate New

Year’s Eve in 2300” differs from “Someone might celebrate New

Year’s Eve in 2300” in that the first is a quantification-in,

claiming that there is someone who will live that long, while in

the second the quantifier has narrow scope. In the same way, the

“someone” who may be only a future individual in “Someone will

celebrate New Year’s Eve in 2300” has narrow scope inside the

thing said characterized by the temporal predicate. So,

schematically, the respective forms are: “Ǝx( Will that. Cx)” and

“Will that. (ƎxCx).”181

181 There are significant differences between our talk about past objects and

our talk about future ones. We can refer to past objects using names because

there can be a causal chain between us and past individuals. The chains

connecting us to future objects go in the wrong direction. So, even if I know

that my great grandchild will be named Samuel C. Wheeler VI, my reference to

that future individual will be via description, “The great-grandchild of

SCWIII who will be named SCWVI.” That description could turn out to be true



To the question, “What do temporal predicates say?” the

Davidsonian answer is like the answer to “What does ‘is a frog’

say?” The answer in that case would be “’Is a frog’ says of an

entity that it is a frog.” In the same way, the future tense and

past tense say of an item with truth-conditions that it is future

or that it is past. If our language had other kinds of primitive

temporal predicates, as Hebrew does, for instance, the truth-

definition clauses would be similar.

Quantifications over future objects, then, give no reason to

regard truths about the future as problematic. The truths about

the future involving quantification over not-yet-existent objects

do not require that the future be now determinable. Even if we

insist that future possible objects exist in order to be

quantified over, there is still no argument, apart from one that

presupposes truth-makers, for supposing that future tense

sentences either have some exotic truth-value or are now fixed by

a currently existing fact. McCall’s theory allows future-tense

sentences to be true or false, apart from any commitment to

truth-makers.

Another question is whether the present tense, as well as

other tenses, is something like a modal. If the present tense is

modal-like, it is a kind of null-modal that delivers the truth-

conditions of the predicated sentence as its truth-conditions. A

of various possible individuals, depending on the procreation behavior of my

descendents.



former roommate of mine always prefaced his remarks by “It is the

case that,” a kind of inert modality. Thus “It is the case that

it is raining” is true if and only if it is raining. In some

sense, the present account treats the present tense as a sort of

surrogate for the naked truth-conditions.182

To defend the “deflationism:” McCall’s theory gives a nice

account of why it is I cannot avoid a high school diploma even

though I once could have. On my account of temporal predicates,

being purely semantic, there is nothing much to say about this

other than that’s the way it is. The passage of time is reflected

in “can”s becoming “could have”s and sometimes “could have but

now can’t”s. Modal primitivism, as discussed in Chapter 6, is a

natural consequence of rejecting the analytic-synthetic

distinction and adopting the paratactic conception of intentional

contexts. McCall-type accounts, being essentially modal-like

accounts, strongly support the idea that whatever account one

182

? An option is to think of tenses as predicates qualifying a sentence in the “timeless present,” regarding the timeless present as the naked sentence. A timeless naked sentence “Fred ispresident” would be true sometimes and false sometimes. The tense-predicates would indicate when the sentence was true relative to a demonstrated time. “Two is now an even prime” would be odd, but true, as would “Two was an even prime” and “Twohad been an even prime before the French Revolution” and other such remarks. The difficulty would be that most naked sentences would not have truth-values except relative to a time of evaluation. In effect, this would treat tenses as fillers of argument-places.



gives of modals, there is an analogous account of tense. But the

semantics of tense is a great deal simpler than a theory of how

the tenses work together and what the law-like relations are

among various tenses.

Just as with modals, on the present account, the semantics

doesn’t tell you what possibility and necessity are, so with

tense. Temporal primitivism is an account along the same lines as

modal primitivism. A semantics of time is not a theory of time.

It is an account of what the predicates are, but not an account

of what it takes for any particular temporal predicate to apply

to a sentence. That is an important job, but it is not a job that

is part of semantics.

II) Enduring objects

IIa) History

Why should there be a problem about enduring objects?

Heraclitus, as noted, seems to be the first to have argued183 that

there is some problem with enduring objects. His argument is

simple: If something changes, it becomes different. If it is

different, it is not the same. But every object is the same as

itself.

183 It is not very clear that the historical figure Heraclitus argued for his

theses rather than regarding them as insights to be transmitted by

illuminating aphorisms. It is clear that from at least Plato on, he was taken

to have a view for which arguments could be given that needed a response.

Heraclitus and Heracliteans are addressed both in the Theaetetus and the Sophist.



One narrative of classical philosophy takes Plato and

Aristotle to respond to the Heraclitean argument as they

understood it. Plato can be read as responding to this argument

by distinguishing intrinsic from relational changes.184 Thus

objects which undergo no intrinsic changes could endure. The Form

of the Good, for instance, can endure even though people only

intermittently think about it.

Aristotle can be read as distinguishing among intrinsic

changes those that bring about a different substance from those

that modify a continuing substance. Accidental changes are those

that an entity survives; essential changes result in the entity

ceasing to exist. What remains after an accidental change is the

same entity to which the change happened. What remains after an

essential change is something other than the entity to which the

change happened.

Aristotle’s can avoid Heraclitus’ argument by claiming that

there are two senses of “is the same as”—the “same” can mean

“numerically the same” or “the same in some feature.” But even if

there are two senses, they are necessarily related. If A is

numerically identical to B, then every feature of A is also a

feature of B. Thus Aristotle’s account of “is the same substance

as” does not really solve the problem of how the same thing can

undergo change and remain the same. Aristotle in fact allows that

the same thing is simultaneously in some sense several things.

184 Theaetetus 155a3 ff.



The person-with-accidents is different from the person considered

as a soul, for instance.185

Leibniz is a Heraclitean with a metaphysics. His principle,

that if A = B, then anything true of A is true of B, seems to

entail that nothing can undergo any kind of change whatsoever, if

“anything true of” is taken to mean “any open sentence.” He

accommodates this apparent truth by eliminating time, a course

which is approximated by many of his recent acolytes. If “weighs

over 150 pounds” is true of the adult and not true of the baby,

then the baby cannot be the same as the adult. If “same person”

is just a special case of “same,” then, apparently, the baby

cannot even be the same person as the adult. Enduring objects are

eliminated by fiat on such a view.

Lewis, like Leibniz, accepts Heraclitus’ argument, with some

qualification explained below. His solution to the “problem of

temporary intrinsics,”186 is to accept Heraclitus’ conclusion that

there are no lasting objects. However, for Lewis, there are

objects—they are instantaneous or of atomic duration. 187 Whereas

185 Metaphysics Z. Aristotle has the resources to answer Leibniz’ version of Heraclitus’ challenge, as Jeffrey Brower’s (2010) shows. Aristotle’s solution invokes “qua” objects, which are more expensive than necessary, according to the CV theory I propose below. 186 Lewis (1986) page 202.187 Heraclitus and the other ancients saw no difference between an object that

existed for no time whatsoever and no Being at all. It is puzzling how four-

dimensionalists hold that there is a determinate number of such instantaneous



Heraclitus portrays people’s view that there are lasting objects

as an illusion engendered by the regularities induced by the

logos, Lewis regards lasting objects as composites of

instantaneous objects. The main substantial differences on this

topic are that Lewis fills in detail, is rather less aphoristic,

and writes in English rather than Greek.

IIb) Was I ever a four year old?

If Leibniz’ Law means that anything that is truly said of A

can be truly said of B, then no tensed predicates apply to

anything. Tensed remarks about the blond cutie (“He is four years

old”) would be false about the curmudgeonly adult. But the number

two is likewise in trouble, serenely unchanging though it appears

to be. It is now being used as an example by Wheeler. That remark

will not be true of it this evening. This is of course a

relational feature, but that shows that even Plato and Lewis will

have to construe Leibniz’ Law as not about true predications, but

about features in some restricted sense, in which truths about

temporary relations do not count. Otherwise, even eternal objects

such as the number two and the set of possible worlds will not

beings at a given instant. Like Heraclitus, they must think that patterns of

“natural” features at locations separate out beings from gerrymandered worms.

But this is to read off the instantaneous objects from the pattern, rather

than constructing the patterns from the instants. Heraclitus would acknowledge

that there are “natural properties” in this sense—the Logos brings it about

that such illusory things as people, chairs, stars, etc., emerge and can be

relied on.



survive change. Unless temporal remarks are reconstrued, either

there are no objects which endure, or tensed remarks are never

true.

In ascribing features from the truth of tensed predications,

Plato and Lewis restrict truths about the changing relational

predicates true of an object to expressions of relations between

objects, properties and times. Leibniz’ Law applies only to

intrinsic properties of objects, since nothing about the object itself

has changed when it enters into a different external relation at

a different date. “Two has not been used by Wheeler” becomes

“There is a time x at which two is not used by Wheeler at x.” The

tensed predicate “was used by Wheeler” is disqualified as a

counterexample to Leibniz Law. Thus Lewis’ formulation is the

problem of temporary intrinsics rather than the problem of

temporarily true predicates.

The tense-as-like-modals account can agree with Leibniz’

Law. It is surely correct that if A is identical to B, then any

feature of A is also a feature of B. Like Lewis and Plato, we have

to be careful about what a feature is. The present account will

agree in many respects with Plato and Lewis.

Lewis in effect maps tensed sentences onto untensed two-

place predications of times which “say the same thing.” So, “Fred

will be happy” is mapped onto “There is a time future from now at

which Fred is happy.” We can do the same thing. We can take a

feature of an object to be the de-tensed and dated content of a



present, future, or past tense truth about the object. A

feature, that is, is what is ascribed by a dated predication. So,

the features of an individual A that has undergone changes in any

predicates true of it will be in essence a list of pairs of

predicates and dates. If we take the total of the contents of

tensed sentences that are true of an individual, we have what we

might call the individual’s Curriculum Vitae. The CVs of Sam in

1949 and Sam in 2012 are identical. Of course, given the view of

future contingents described in the first section, there are

truths stateable about Sam in 1948 that could not have been known

and were not determined by anything at that time to be truths.

Truths do not need truth-makers. However, the very same dated

truths, as well as all the quantified truths ascribing temporary

accidental intrinsic (=one-place predicates), are true of me now

and were true of me in 1948.

There is no reason not to take the future, past and present

truths about an enduring entity as the features that of course

are the same by Leibniz’ Law. If it is admissible for Lewis, it

is admissible for a Davidsonian. That the contents of those

truths are expressed with different modal-like predicates is just

part of what it is to be an enduring entity that changes over

time.188

188 Remember, we are talking about Sam in 1948, not Sam-in-1948. While there is no reason to deny that there are lives, stages of lives and instants of lives of people, and perhaps no reason to deny exotic compounds of people-plus-times, there is likewise no reason to suppose



An objection at this point might be that what Leibniz and

others have in mind by Leibniz’ Law is not predicates being true

of objects, but rather objects having properties. Wheeler is now

someone who has taught Philosophy 5301. It would indeed be

difficult to find that property, in the sense of a component

attaching as a dependent particular, by carefully examining the

four-year old. Given a metaphysical view that predication

ascribes such components to the substances to which they are

subordinate, there is indeed a real problem about how objects at

different times could be the strictly the same. To be a proper

endurantist, you seem to need to posit properties of the four-

year-old that are already future instantiations of properties.

Or something. Despairing, you could reasonably be driven to four-

dimensionalism or some other such view.

These considerations seem to me to be very good reasons not

to think of Leibniz’ Law in metaphysical terms, and to constitute

yet another reason to agree with Davidson’s conception of the

metaphysics (or non-metaphysics) of predication: That what it

takes for “There have been frogs” to be true is that there have

been frogs. There are, of course, properties of things, but we

that people are compounds of such entities. By the perspective on “ontology” from Chapter 3, people and other enduring objects are single entities with spatial parts and temporal spans. They have partsand exist for spans, but, while they may coincide with collections of spans or parts, they are not identical to those sums or sets. None of the objects we posit need get in any of the others’ ways.



have seen in Chapter 6 that such properties are metaphysically

innocuous and inert. Metaphysics only causes trouble. Who needs

it?

It is important to see how this account of enduring objects

fits with the relative essentialism of Chapter 3. In Chapter 3,

we argued that ascribing modal features is part of positing kinds

of objects. Distinct modal predicates being true of persons and

their worms in fact is what distinguishes people from their

worms. Modal truths about an enduring object change as rapidly as

non-modal truths. Those temporal modal truths are replaced by

other modal truths. In 1948, it was true of me that I might drop

out of high school. That is no longer something I might do. It is

now necessary that I went to Carleton College, but it was not

necessary before. My CV will therefore not include these

particular passing modal truths. But the CV will include more

complex modal truths, such as that it was possible for me to drop

out of high school until early June of 1962 but not thereafter.

That is, corresponding to every change in modal truths, there is

a dated truth about that change. The complete CV will catalogue

the modal changes that McCall’s branching trees model.

The CV of an enduring object will be very different from the

CV of the object’s corresponding worm. Supposing my coinciding

space-time worm is worth positing, it will have an entirely

different set of modal truths. At no time will it be possible

that my worm went anywhere other than Princeton. My worm could



not have visited Cooperstown, NY. The worm is a particular path

through the tree, so its possibilities are limited to futures

after the stage at which I have died and left my worm to fend for

itself. That is, the worm is a mereological sum of possible

object-stages, so it did not have alternatives during its

duration.189

It is worth seeing how this distinguishes among coinciding

enduring objects, such as the statue and the lump. Suppose a

statue made of a mass of gold comes into existence instantly by a

vapor-condensation process and a few years later is completely

converted into energy instantly, so that the statue and the lump

have exactly the same CV, restricted to actual features. Modal

features, such as “could have been an ingot on June 22” will be

true of the mass of gold but not of the statue. The two enduring

objects both survive changes, but some of the changes they

survive are different. The lump will have the potential to become

an ingot throughout its career, where the statue will not. The

lump will have the complex feature “possibly an ingot on June 22

until June 22 and not possibly an ingot on June 22 after June 22”

where the statue will not. The modal CVs of the lump and the

statue are different, and so they are different objects. 189 It would be possible to adapt a Storrs McCall story to a kind of worm whose

only difference from a person was that it was a collection of stages. We would

then have an entity that could have different parts, like my book collection,

which is not a single object, but has always had modal properties. It is

something other than a mereological sum.



So, “endurantism,” the thesis that objects such as people

and monuments endure through time as the same object, scarcely

qualifies as an “ism” at all. The alleged difficulties that

Heraclitus and others find with the thesis are self-inflicted

binds that metaphysical misunderstanding of predication brings

about. A Davidsonian understanding of predication and modality

shows the “paradox” of surviving change to be a paradox only

given unmotivated and unjustified theses about the ontology of

predication and felt but not real necessity of reducing the modal

to the non-modal.

Chapter 8:The Sorites and Davidsonian Innocuous Epistemicism190

The sorites argument and the ontological problem it raises with

essentialism was the core of my argument in Chapters 2 and 3 to

abandon monistic essentialism, the idea that reality is

intrinsically articulated into beings and properties. If there

were precise and systematic delineations in micro-particle terms

of the objects of ordinary life, we could happily be neo-

Aristotelians, regarding the medium-sized objects and organisms

he took to be basic substances as respectable entities, but

entities having compound natures derived from more the more

fundamental natures of the micro-particles into which nature was

articulated.

190 An early version of some of this chapter was presented at the conference on

“Davidson and Pragmatism” at the University of Zurich, 27-28 April, 2010.



According to the relative essentialist account of ontology,

vagueness could seem to be a feature of objects themselves, not

only of language. The objects and properties posited are by and

large informally posited by whole cultures or organic lineages.

Precise definition relative to other posits is much too much to

expect from such posits. For the vast majority of such posits, it

is quite unlikely that anything in the positing would provide

necessary and sufficient conditions for an entity being of the

kind posited in terms of any other posits. Furthermore, for

perceptual predicates, those primarily directly applied by

acquiring involuntary perceptual beliefs, it is very unlikely

that those perceptual dispositions will divide possible cases

into those which elicit the belief that the predicate is true in

a case and those which elicit the disposition to believe the

negation of the predicate is true.

The view I will argue for is that the objects themselves are

quite precise, and have sharp borders, in a sense. It will always

be true that “is red” is true of an object if and only if it is

red, even though at some point in a continuous change from red to

orange, speakers cannot say whether or not a color-patch is red.

Likewise, everything is either a chair or not a chair. This

chapter will argue for a kind of innocuous epistemicism without

truth-makers.



Predicates subject to the sorites, that is, almost all

predicates, divide into two kinds: First, there are a few one-

dimensional predicates, such as color-words and comparative

adjectives of dimension, such as “tall” and “warm.” Such features

can be placed on a continuum and ordered, so that for any two

items, either the first is taller than the second, or vice versa,

or they are equally tall.

The vast majority of writing on the sorites has focused on this

first kind, the one-dimensional cases. In fact, the complicated

“dimension” baldness is often treated as if it were solely a

matter of number of hairs on a head, rather than some kind of

combination of numbers and distribution and location (consider

Ben Franklin with his wide expanse of bare scalp, but dense hair

at the edge). The illusion is thus generated that the sorites can

be fixed by dealing with the arbitrariness of any dividing line.

In the first kind of problem predicate, since we have a continuum

along a single dimension, we can imagine that the difficulty is

that of coming up with a demarcation. All we need to do is decide

where to draw the line. Since the language-community and our

language-learning practices have not established such a line,

unlike the case with “adult” or “is old enough to genuinely

consent to sex,” the natural suggestion is that there are a

variety of acceptable demarcations, yielding supervaluation.

Another idea is that we really use a logic with tolerance for

either “is both tall and not tall” or “is both tall and not tall”



can be assigned to language users which will accommodate these

borderline cases while preventing speakers from inferring that

every sentence is true.

Second, there are predicates such as “bald,” “chair,” “human

being,” “nice”191 and virtually every count-noun. In such cases,

there is not only more than one dimension, but it is often

difficult to see how the question could be put in terms of any

number of anything that would seem to be any number of

dimensions. There is nothing close to a well-ordering in the

borderline area. I am quite certain that a collection of objects

such that it is difficult to say whether they are chairs could

not be ordered in any helpful way. There will of course be

objects such that one is clearly more a chair than the other.

For predicates like “chair” or “person” it is hard to see how any

kind of precisification would go, since there are so many ways a

thing could be defective. A good example of a predicate to which

it has proven to be difficult to apply “acceptable”

precisifications is the topic of Wheeler (1980). There the

predicate under discussion is “has a right to” and the relevant

sorites step is “is not ethically significantly different from.”

A chain of cases goes from a person’s right to move their body to

unlimited property rights. But no precisification of “has a right

191 “Nice” is Rosanna Keefe’s (1998) example in a paper making some of these

points.



to” is uncontroversial. Rights and their limits are a hotly

contested topic in political theory. With some exceptions, no

theorist thinks that another theorist’s precisification is

acceptable. If “acceptable” means more than “acceptable to me,”

then “precisification” and supervaluation will not be effective

solace here.

I have always taken the fundamental question raised by the

sorites to be metaphysical rather than logical.192 The most

important lesson of the sorites is that the natures of the

objects of ordinary life and their properties cannot be complexes

of the natures of the objects that the sciences tell us about.

The central motivation for the relative essentialism developed in

Chapters 2 and 3 is to save the reality of medium-sized objects

and their features from the hegemony of physics. Relative

essentialism allows medium-sized objects to have natures. It does

not have to describe how such natures relate to the basic natures

of reality, because it denies that there are any basic natures.

The sorites makes the problem and solution clear.

In Wheeler (1975), I suggested that the sorites posed a serious

dilemma for monistic essentialism, a view I was then considering.

If the world consists of given individuals and properties, and

predication is understood as a property being instantiated in an 192 Eubulides, the inventor or purveyer of the paradox, likewise thought the paradox was primarily of metaphysical significance. He took it, along with hisother paradoxes, to be a demonstration that Parmenides was right. See Wheeler (1983). As Peter Seuren (2005) also notes, Eubulides’ paradoxes set the agendafor twentieth and twenty-first century semantics and philosophy of language.



individual, then for any property and any being, that property is

either instantiated or not. If Aristotle were right that medium-

sized objects are the basic entities of the sub-lunary world, in

the sense that the fundamental necessities are necessities about

people, and other organisms, then the sorites might be a

problem,193 but not a threat to the existence of the objects of

ordinary life. Aristotle did not regard heaps as substances in

any case, so acknowledging that heaps are not well-defined beings

might have been acceptable.

The sorites becomes more than a logical puzzle, though, when it

develops that the strongest laws are not about medium-sized

objects and organisms, but about micro-particles.. Given that an

Aristotelian understanding of necessity is natural necessity,

having the strongest laws means being more fundamental. So, given

the naturalism shared by Aristotle and most modern

metaphysicians, this would mean that the beings with essences are

the micro-particles, and any beings with any sort of essence

would have to derive that essence from micro-particle components.

As discussed in Chapter 2, the major problem in descriptive

metaphysics is accounting for how objects which are not basic to

science and are not reducible to objects that are basic to

science, can still have essential features. Relative essentialism

193 Historically, it appears that Eubulides, the apparent inventor of the sorites, was attacking Aristotle. See Jon Moline (1969), as well as Wheeler (1983.)



is a solution to that problem along Davidsonian-Quinean lines.

Relative essentialism solves the problem by having a multitude of

independent essences and natures. Posits and their natures need

not correspond to notional “joints.”

The logical problem raised by the sorites, however, remains even

when the metaphysical dilemma it poses is neutralized by relative

essentialism. How can bivalent logic, which seems so useful and

true, be correct in light of the apparent indeterminacy the

sorites seems to reveal? A borderline chair seems to be neither a

chair nor a non-chair. Through several failed attempts, beginning

in 1973,194 I tried to find an intelligent response to the

sorites, which seemed to me to pose an intractable problem for

any philosopher of almost any persuasion. This chapter is my

latest attempt.

I What is the sorites paradox about?

The sorites paradox was originally presented as the inconsistency

of the propositions that a) a single grain is not a heap, that b)

no addition of a single grain to a non-heap turns that non-heap

into a heap, and that c) a million grains together is a heap.

Sorites sequences can be constructed for “is a chair,” “is a

person,” “delicately,” “crushes,” and almost every predicate

outside of mathematics and the hard sciences.

194 At the 1973 APA Eastern meeting I presented “A Solution to Wang’s Paradox,”a nihilist “solution,” which became Wheeler (1975). See also Wheeler (1979), (1986), (1991).



Since the early 1970’s the literature on the sorites has grown to

a vast size,195 so that a discussion of even the major lines along

which philosophers have sought a solution is out of the question

here.196

There is reason to think197 that Eubulides, the originator of the

sorites, took the sorites to be making a metaphysical point. The

medium-sized objects and pluralities of the common-sense world do

not make logical sense, so believing in their existence is

incoherent. The theory here presented agrees with Eubulides that

the central problem of the sorites is metaphysical rather than

logical. The present theory relies on neo-Davidsonian views of

predication, kinds, and truth.

There are basically three responses to the sorites, given that

nobody thinks a single grain is a heap:

1) Deny that the sentences are inconsistent, when properly

understood, or that the inconsistency has serious consequences.

195 In the early 1970’s a philosopher could read everything that had been written on the sorites problem in the Twentieth century in a leisurely afternoon, with time for a nap. I did.196 Excellent guides to this literature include Williamson (1994) and Keefe and

Smith (1997).197 See Wheeler, (1983). I argue that the point of the sorites, as well as of the other paradoxes Eubulides constructs, is to argue that Parmenides is right. Interestingly, after millennia, the paradoxes attributed to Eubulides are at the center of the philosophy of logic and the philosophy of language. Besides the Sorites, Eubulides proposed paradoxes involving presupposition, intentional contexts, andthe liar paradox.



This is the solution of the supervaluationists and the logic

adjusters.

2) Deny proposition b), and claim that at every point in the

progression, a collection either is a heap or is not a heap. The

problem is just that for a large number of cases no answer seems

objectively correct. One version of this position is the

epistemicist solution; we will propose another version below.

3) Deny proposition c), and claim that in fact there are no

heaps, as the argument shows. This is the nihilist198 solution.

Each of these strategies has had recent advocates. The first

strategy has generated the largest literature.

The present theory is akin to an epistemicist account. It turns

on metaphysical views, borrowed or adapted from Davidson, about

language, predication, kinds, and truth. Being adapted from

Davidson, the theory is naturalistic, and eschews language-

transcendent concepts.199 From a Davidsonian or Quinean point of

view, of course, there can be no epistemicism resting on

inadequate grasp of concepts. Language is to be understood as 198 “Nihilism” became the label for views that deny that sentences aboutmedium-sized objects and their properties are true. Wheeler (1975, 1979) and Unger (1979) proposed this view in the 1970s. Unger proposed“nihilism” as the label for the view, claiming that even more things than Wheeler thought fell under the sorites argument. More recently, Ted Sider and David Braun (2007) and Kirk Ludwig and Greg Ray (2002) have reached similar conclusions.199 The presentation below owes its inspiration to Vann McGee’s conceptualization of the sorites problem in terms of the relationship of usageand extension, in his presentation at the conference on Truth, University of Connecticut May 15-17, 2009.



grounded in responses of organisms to their surroundings in a

social environment. There is no chance whatsoever that concepts,

as internal contents of predicates, could be constructed out of

such material which would divide all possible objects into those

which satisfied the predicate and those which did not.

The Davidsonian proposal I will suggest is akin to an

epistemicist account in claiming truth-values for borderline

cases, but unlike an epistemicist claim, does not diagnose the

problem as some kind of limitation to our knowledge.

II Usage and Extension

As discussed in Chapter 1, Davidson supposes neither that

“concept-fitting” nor selection by privileged kinds that actually

exist determines what a predicate is true of. Only the truth-

condition clause gives the meaning of a predicate. One way to

articulate Davidson’s view is to raise the question of what the

relationship is between a term’s extension and usage.

What is the relation between the pattern of application of a

predicate within a culture and the extension of that predicate?

There are two standard general conceptions of this relation,

discussed in Chapter 1, while I will briefly review: The “meaning

is use” conception takes the content of a predicate to be

determined by some version of “what people say when,” and takes

the extension of a predicate to be a function of its content. The

monistic essentialist conception rests on natural segmentations



in the world as extensions of predicates. This section sketches

both familiar views and outlines their difficulties with the

sorites.

a) usage determines extension views

The obvious naturalistic way to assign meaning or content to a

predicate is to take the meaning of a predicate to be a function

of “use” or “usage,” and to take the extension of the predicate

to be determined by its meaning. What the members of a culture

say when fixes all there can be to the content of a predicate,

and therefore all there can be to fixing what entities the

predicate is true of. The extension of a predicate, then, is a

projection from actual applications to all possible applications.

The use/usage-is-meaning view of extensions leaves extensions,

and so truth, indeterminable. For most of our predicates, no

amount of actual application-behavior will fix a single

projection that selects a single extension within the collection

of possible objects for which the question whether the predicate

applies might arise.200 No matter how many data-points one has

derived from applications of “is a table” no particular complete

extension, from among the infinity of psychologically projectable

sets that include those data-points, is selected. Thus there are

possible (and actual) entities such that it is indeterminable

200 This is the point emphasized by Vann McGee in his presentation

mentioned above.



whether that entity falls within the extension of, for instance,

“table.”

b) monistic essentialist views

For a monistic essentialist, there is a privileged segmentation

of the world into kinds. This segmentation is reflected in laws

connecting kinds in the segmentation. These natural laws give the

essences of the kinds that are the extensions of terms. The laws

may be strict or may be Aristotle’s “always or for the most part”

laws.

Language-learning proceeds by acquaintance with such natural

kinds, which brands a given kind with a term. Alternatively, a

variety of stories about how evolution has equipped us to get at

the kinds or how our language-faculty has as its proper function

designating the right kinds are told. A metaphysical realist

conception of language and its relation to the extensions of

terms is externalist. The patterns of application of the

predicate, whether in the individual or in the society as a

whole, do not determine the extension. Usage has to have some

relation to extension in order for the reference-fixing to occur,

but that relation can be minimal.

This kind of externalism allows that, for properties and kinds of

objects that are governed by strict laws, there are no genuine

borderline cases—either a difficult-to-characterize entity is in

the extension or not. A metaphysical realist conception explains



the divergence between what one’s language-teachers teach and the

truth by appealing to a natural division in nature which selects

some extensions as appropriate extensions for predicates.

c) Difficulties with the sorites

c1) extension is a function of use theories

For the meaning is usage theorist who takes extensions to be

determined by meaning, the sorites argument shows that for almost

every predicate, the meaning or sense of the predicate, if

resting on what people say when, does not determine an extension

even in familiar and often-encountered cases. Nothing about my

culture’s history of verbal behavior defines an extension for

“chair,” in the sense of sorting the possible objects into the

chairs and the non-chairs. The set of extensions that accord with

actual usage is insufficiently restricted to sort even the

actually available objects into the chairs and the non-chairs. An

account of meaning as resting on usage rather that nature seems

to condemn the usage-as-meaning theorist to incomplete meanings,

to multiple truth-values, precisifications, and the like.

c2) monistic essentialist

The monistic essentialist solution seems to address the problem

of determining a single extension to a term by assigning that job

to nature. However, monistic essentialism has difficulty in

application to medium-sized object predicates. If we interpret

necessity naturalistically, and treat natural kinds as determined



by natural laws, then a segmentation into natural kinds requires

that the natural kinds be the subjects of laws. Natural kinds are

supposed to have essences that yield necessary truths about when

they apply. For natural objects, those essences are expressed in

natural laws. But the laws about medium-sized objects and

organisms at best admit exceptions. They are loose relative to

the laws of physics, chemistry, or even cell biology. The sorites

illustrates that, for instance, even though we have a necessary,

perhaps probabilistic, relation of some kind between being a tall

man and having an adequate height in meters, there is no lawlike

relation that would determine what that height in meters must be.

As we have noted in Chapter 2, the basic idea of monistic

essentialist accounts of reference, that reference is fixed by

divisions in nature, seems to run afoul of the sorites, at least

on the assumption that the medium-sized objects of ordinary life

exist and so have essences.201 Apart from some quite unusual

predicates,202 kinds of medium-sized objects and their properties

are intuitively not completely determined by a privileged

201 For the arguments that if natural kinds are taken seriously as expressed by necessary truths, then any alleged kinds whose application cannot be determined by natural laws would not be real kinds, and so would not supply essences for objects, see Wheeler (1975). 202 Black holes, unlike most big objects, have a natural law-governed delineation, the Schwarzchild radius, defined by the escape velocity of a particle being beyond the speed of light. So black holes are precisely definedentities, as far as I can discern. I cannot think of other examples of macro-objects such that there is a precise point at which their borders begin.



segmentation in nature. At best, nature sets parameters within

which wide variation is possible. For objects such as tables and

turtles, if we imagine a particle-by-particle dismantling, there

appears to be no objective line at which the entity in question

ceases to fall under the extension of “is a table” or “is a

turtle.”

A monistic essentialist who does not suppose that there is an

objective answer to questions about extensions of medium-sized

object count-nouns and properties becomes a usage-theorist about

all the predicates of ordinary life. In effect, supervaluation

abandons metaphysical realist selections of a privileged scheme,

and supposes that acceptable precisifications are acceptable

extensions for problem terms. “Acceptable” is then given by

usage.

Monistic essentialists thus are faced with awkward choices about

what to say about medium-sized objects and predicates of them. On

the one hand, it would be nice if there really were tables and

people. On the other hand, the whole idea that having an essence,

i.e. having objective existence and extinction conditions, is

required for reality is undermined by medium-sized object-

predicates and their vagueness with respect to other families of

predicates.

d) Davidson

d1) externalism without natural kinds



As discussed in Chapter 1, Davidson is an externalist about

extensions, but does not believe in a privileged, given

segmentation. Thus, for Davidson, all kind predicates that

actually apply to objects, i.e. all predicates for which there

are affirmative true sentences, are ontologically on a par. Of

course there are electrons; of course there are tables, but there

are no leprechauns. For Davidson, we learn to apply terms by

triangulation, coming to call an object salient both to us and to

another by the same term. Davidson characterizes this

triangulation and its consequences as follows:

“Ostensive learning works first and best with whole sentences, in

practice often represented by what for the experienced speaker

are single names, common nouns, adjectives, and adverbs (`Mama’,

‘Man’, ‘Come’, ‘Good’, ‘Careful’). The child who has no more is

still a pragmatist. Once some grammar is in hand, however,

separately learned parts can be assembled in new ways, and truth

separates from the merely useful or approved. The references of

names, the extensions of predicates, the combinatorial devices

themselves, are in the hands of teachers and society; truth is

not.” (Davidson 2005, page 15)

Since there is no privileged segmentation into objects and

kinds of objects, there is much latitude in what groups of

salient objects are correctly called by the same term. Thus,





cultures and within cultures over time. Language is a human

creation, and human language-behavior shapes the ontology that

can be assigned to a culture. That is, given that most of what

people say is true, patterns of labeling behavior will shape the

extensions an interpreter assigns to a predicate.

According to our relative essentialism, which I am treating

as Davidson’s implicit view, an ontology is an imposition or

positing of a set of predicates for putting the world into

subject-predicate form, in order to allow logical relations among

truth-functionally simple sentences. As we discussed in Chapter

3, “positing” should be thought of as by and large automatic,

coming from biological inheritance and culture. Only rarely do we

actively posit such things as gravitons and gluons and sit-coms.

Since there is no privileged segmentation, whatever can become

salient to humans is a possible partial extension of a single

predicate. But of course for the reasons given in Chapter 1 in

discussing the usage account of extension determination, no

amount of such training or decision-making about what to say when

will yield a single extension for an individual or for a culture.

That is, there is no projection from any finite amount of

identification of elements of the set to any particular set.

Since Davidson explicitly denies a privileged segmentation, the

difficulty is even more transparent.

For Davidson, though, extensions are not functions of the

content, in the sense of “content” that would be a projection



from usage. As emphasized in Chapter 1, the meaning of a

predicate is its truth-conditions. Extensions are given by the

deflationary-sounding formula, “`Is a dog’ is true of an object A

if and only if A is a dog.” Just as truth-conditions or meanings

of sentences are given in homophonic translation, so satisfaction

or truth-of conditions, that is, extensions, are given in

likewise apparently unilluminating form. So, even though meanings

and extensions are learned by imitating usage, the meaning is not

identical with the usage. Furthermore, for the reasons given

above that usage cannot select a single set on the basis of a

finite number of occasions of use, extensions, while not

independent of usage, could not be determined by usage. Usage is

evidence for meaning when we are interpreting, but meaning itself

is given completely by predicate-clauses in a truth-definition.

For Davidson, most of what people say using these divergent

predicates is true. For a Davidsonian, then, there are an

indefinitely large number of overlapping natural kinds, as it

were. All the distinct predicate-systems of different cultures

and all the overlapping kinds are correct posits, as long as

there are truths of the form “ExFx.” Given Davidson’s

externalism, a culture’s divergences from another culture means

that each culture’s entities are by and large real.203 Objects

overlap without being reducible one to the other. Diverse objects

203 Gods, demons, auras, Guardian Angels and such can be exceptions. Davidson’spoint is that you could only come to think the other is mistaken by thinking the other is correct about a lot of other judgments.



can co-exist and overlap without getting in one another’s way.

Most importantly, diverse overlapping families of predicates can

overlap within a single culture, so that the distinct objects can

coincide.

d2) extensions

As we discussed in Chapter 1, the combination of externalism

about reference and denial of a privileged segmentation means

that Davidson can have objective extensions without supposing

that there is a single division of the world into kinds, and so

without having to rescue weakly-law-governed entities from the

threat of not being really part of what is. For Davidson, all

kinds are on a par ontologically. Some kinds are connected to

other kinds by very good laws; others by not so very good, “for

the most part” laws or by the kind of very vague generalization

that would tell us that tables have to have a fair amount of

matter.

Here, then, is my Davidsonian view about extensions and usage:

1) Predicates have extensions. The union of the extension of a

predicate and the extension of its negation is the universal set.

There is an answer, “yes” or “no” as to whether a given predicate

F is true of a given object A, for any A. Given that the meaning

of a predicate is given by its truth-condition clause, a

predicate’s meaning in a sense trivially “fits” what it is true

of.



2) Although of course learning a language is finding out about

the extensions of predicates by learning to detect elements of

their extension either directly or indirectly. We both learn to

detect perceptually when a predicate F applies and we learn F’s

connections to other predicates, and so can apply F by knowing

what else is true in the case at hand. But no finite amount of

observation or training or collation of our culture’s

application-practices will prepare us to put every candidate for

“member of the extension” of a predicate, at least in general.204

That is, when we encounter or consider objects, always under some

description, and so with some posited nature, we are not equipped

in general to assign them either to F or to not-F.

3) Learning the extension of a predicate is only rarely learning

necessary and sufficient conditions using other predicates that

determines when a given predicate applies to a given object.

4) Extensions of predicates are not generally determinable by

determining extensions of other predicates, but are determinate.

That is, it can be in principle impossible to determine what the 204 For the unusual cases where we can give the extension in other terms, such as “prime” and “composite” applied to natural numbers, membership in one or the other of the extensions is not determinable relative to some predicate families applying to numbers. “The number of planets” was once thought to be a prime, for instance. “Prime” and “composite” are only determinable relative to some ways of specifying numbers. This is masked by the fact that there is a procedure, in principle, for converting number-description of the form “the number of F’s” to a description for which there is an algorithm for determining whether a number is prime.



extension of a predicate is for the general case that includes

every possible object.

Thesis 1) commits a Davidsonian to bivalence. One-dimensional

borderline cases of a predicate P are cases where one cannot tell

directly whether P is true of the case, P is known to depend on

nothing but truths from predicates in family Q, and all the

relevant evidence from Q predicates is in. Given that there are

borderline cases where nothing can indicate to us whether a

predicate truly applies or not, there are sentences that are

determinately true or false when those sentences’ truth-values

are not determinable. Not being determinable does not mean not

being determinate.

Davidson and Quine would both call such situations

“indeterminacy.” Throughout this book, I have called such

situations “indeterminability.” So, I have spoken of

“indeterminability of interpretation” for instance. Why should we

think that determinacy requires determinability and that

therefore indeterminability is indeterminacy? The basis for this

view seems to be descended from the Verification Theory of

Meaning, via the Verification Theory of Meaningfulness. The idea

has been that, unless there is a possible procedure for detecting

whether a predicate applies in a case, it does not make sense to

suppose that the question whether the predicate applies or not is

not a question of fact. But the verification theory of

meaningfulness can only be obvious if one thinks that there is a



given level of experience or something relative to which all

questions are decided. Davidson argued against any such given in

Davidson (1974). There being no other terms relative to which a

given term is determinable would only imply indeterminacy if

there were other terms that were intrinsically more basic.

If the meaning of a predicate is given by its predicate clause,

and there is no given articulation of the world into basic beings

relative to which all other beings must be understood, then there

are no given terms, either. So there being some situations in

which it is in principle impossible to determine truth-values of

predications does not mean that those predicates have no truth-

values.

Here is an example of an absolutely indeterminable predicate

whose truth or falsity of a given object is determinate. Consider

the infinite family of mathematical predicates P1, P2,… where P1

is “=2 if the continuum hypothesis is true, = 4 otherwise,” P2

is “= 3 if the continuum hypothesis is true, = 6 otherwise”, and

where Pn is “= the nth prime if the continuum hypothesis is true

and = the nth composite otherwise.” Relative to any of these

predicates, “prime” is indeterminable. From the information that

the number of my first cousins is P5, nothing can determine

whether the whether the number of my first cousins in 11 or 10.

It can be determined that it is either the 5th prime, 11, or the

fifth composite, 10. So, it given that I have 10 first cousins,

“The number of my first cousins is P5” is either true or false,



but absolutely indeterminable. But, supposing that Platonism is

coherent, that this predicate is always determinate but never

determinable is also coherent. So there is no necessary

connection between determinacy and determinability.

It is of course true that predicates which have very little

connection to any other predicates and are rarely directly

determinable by observation, so that predications using them can

only rarely can be determined to be true, will be useless, and so

in a sense meaningless. But it is hard to see what argument could

be given that determinate truth-value requires determinability,

absent a ground-level of basic predications that are given by

nature.

So, a Davidsonian can be an epistemicist without supposing hidden

facts or states of affairs making such sentences true. There are

hidden facts in the notional sense described in Chapter 6, of

course, but the “inaccessibility” of such facts is not a defect

in our knowing abilities. For Davidson, truth is primitive.

Sentences are not “made true” by anything. For a variety of

reasons,205 there are no entities corresponding to true sentences.

Since there are no truth-makers, and truth is primitive, truths

205 Davidson is an advocate of the Slingshot, examined in detail by Stephen Neale (2001). In any case, facts and other concrete correlatesof sentences are suspect on many grounds. There is the problem of binding, which goes back to Plato’s Parmenides, there is an implausible population of negative and general facts, and so on.



do not have particular chunks of the world or particular

environmental incidents to fasten onto.

So, all terms have perfectly clear extensions, given by the

predicate-clause in the truth-definition.

III Theoretically justified bivalence and Davidsonian pragmatism

A Davidsonian, following Quine, can allow that some truths,

typically true standing sentences and their instances, are

reasonably held to be true in virtue of theoretical

considerations. Useful analogies abound in mathematics. Neither

usage nor intuition would suggest that among the sub-groups of my

siblings there is the null set. Yet it is a theorem that the null

set is a subset of every set. We need that theorem to be true in

order to retain such principles as that everything that is a

member of the subset is a member of the superset. Likewise, the

idea that for any number, raising it to the zeroth power yields

one as value fits no intuitions about what “raising to the zeroth

power” means, since the notion is intuitively meaningless. This

truth is true because, among other things, n to the mth divided

by n to the pth equals n to the m-minus-pth. In both cases,

theory demands that a sentence be true. Accepting that truth is

harmless in both cases.

Davidson can say the same thing about the thesis of bivalence for

sentences using medium-sized object predicates. In cases in which

“there is no fact of the matter” there is no importance to the

matter either. Just as we run into no practical difficulties in



treating five to the zeroth power as equal to one, so we have no

practical difficulties in treating borderline cases as being true

or false, but it does not matter which.

For the reasons outlined below, namely the lack of sharp laws

connecting families of medium-sized object predicates and the

lack of perceptual determinations, it would be expected that

cases would arise where a predicate such as “is a tall man” could

neither be applied nor denied on the basis of inspection, even

though a predicate from another family, “is 1.78 meters tall”

could be, and whether an individual is tall depends on nothing

more than how tall the individual is. Without correlates of

sentences, i.e. facts or states of affairs, the phenomena of

vagueness are either cases of incomplete connection among

heterogeneous predicates or cases where perception does not

determine an answer and nothing else is relevant. But since all

it takes for “Joe is a tall man” to be true is that Joe be a tall

man, “is a tall man” has determinate truth-conditions.

In the case at hand, sorites arguments, the Davidsonian answer

would be that of course classical excluded middle holds for

sentences about medium-sized objects. There turn out to be good

reasons why, for many sentences, truth is not in principle

determinable. Among those sentences are predications which usage

does not fix as true or false. So, for example, at every point on

a continuum forced march, there is an objective answer as to



whether the entity is a tall man, even though there is in

principle no way to tell.

Just as in the cases of mathematical truths accepted for

theoretical reasons, accepting bivalence in non-semantic206

predications is harmless. As long as we do not suppose that there

are occult facts or other truth-makers, the claim that every

sentence is true or false, so that “Fred is tall” is true or

false, conflicts with no other claims we should wish to make. The

view is an epistemicism insofar as it assigns unknowable truth-

values to borderline cases. This acceptance of bivalence requires

no miracles nor does it require concepts to which we have

imperfect access. It just requires that we treat truth,

extensions, and meaning as completely given by the relevant

clauses in a truth-definition. Meaning is truth-conditions, where

that means that beyond “`Fred is a dog’ is true if and only if

Fred is a dog,” there is nothing illuminating and accurate and

general to say.

IV What is vagueness without truth-makers?

A presumption of the sorites is that the situation where a

man is 1.78 meters tall demands an answer about whether he is a

tall man or not. Why is an answer expected? Briefly, we have a

206 Given the existence of the semantic paradoxes, the general claim of bivalence, that every sentence is true or false, may be false. But therestricted claim that sentences attributing predicates to medium-sizedobjects seems to have everything to recommend it.



datum, a fact before us, specified by “is 1.78 meters tall.” In a

world in which that fact exists, it must, it seems, be the case

either that the man is a tall man or that he is not a tall man.

But suppose, as Davidson does, that there are no entities

corresponding to true sentences, i.e. no facts or states of

affairs. Then the truth-conditions of “Fred is 1.78 meters tall”

are just that Fred is 1.78 meters tall. The problem with “Fred is

a tall man” not being determinable is just that from “Fred is

1.78 meters tall” neither “Fred is a tall man” nor its negation

follows by law. So, one can infer neither the sentence nor its

negation from the “datum.” The point is that the datum is not a

given, but is already-conceptualized. The datum is a truth. So

why should it be surprising that neither “Fred is a tall man” nor

“Fred is not a tall man” can be derived by laws from another

sentence, “Fred is 1.78 meters tall?”

What gives rise to puzzlement is the truth that there is nothing

to Fred’s being tall other than how tall Fred is. So, we think

there ought to be a law and there isn’t. For well-behaved

predicates like “tall,” 207 there are indeed some lawlike truths

207 “Tall” is well-behaved in many ways. One feature of being one-dimensional, unlike “nice” and “bald,” is that one-dimensional predicates have a non-vague comparative. For many concepts there is apparent indeterminacy not only for the attributive construction, but also for the comparative construction. It can seem indeterminate whichof two men, if either, is balder than or nicer than or more obnoxious than the other. With count-nouns in relation to particle-complexes the situation is even worse. No list of dimensions is forthcoming. Of two table-like objects, no laws determine which is more a table than



connecting “n meters tall” “taller than” and “tall man.” For

instance, if Joe is 1.8 meters tall and is a tall man, then any

man taller than 1.8 meters tall is a tall man. Height is the only

relevant dimension for “tall”, which makes it a favorite among

sorites theorists. But even though the family of predicates “is n

units tall” has lots of connection with the family of predicates

“is a tall F”, the “is a tall F” family does not reduce to the

“is n meters tall” family. So, in many cases, we can know the

truth of a sentence about Fred using a member of one family

without knowing the truth of a sentence about Fred using the

other family.

Without truth-makers as correlates of sentences, “borderline

cases” are just sentences whose truth-values are not determinable

directly and are not determinable from the truth-values of other

sentences about the same object. A borderline case can arise if

there is a true sentence using one kind of predicate such that

there is no strict definitional necessary connection to the

truth-value of a sentence using another kind of predicate, but

where the two kinds of predicate are necessarily related. Since

height in meters is related to whether an individual is a tall

man, so that a man having a given height in meters is sometimes

obviously also a case of being a tall man, a sort of paradox

arises if we assume that because the truth-value of one

the other. Only the vaguest “laws” connect material count-nouns with particle-complexes.



characterization of the situation is determinable, so must the

other be, that is, that there ought to be a law. So, “John is

1.78 meters tall” may be known to be true, while “John is a tall

man” may not be known to be either true or false..

If we do not suppose that there are privileged ways of

characterizing objects, then indeterminability arises from one

of two sources: First, we get indeterminability when only

perception is relevant to whether P, and perception just does not

divide cases into ones that elicit belief in P and those that

elicit belief in not-P. Second, we get indeterminability whenever

two families of predicates apply to some of the same entities,

have some necessary relationship, but lack precise laws relating

them. Vagueness arises from relations between predicate-families.

A predicate is vague if there are no non-trivial strict laws

connecting it with predicates from another family of predicates

applying to items in its extension.

We could define a “precise” predicate as one such that there is a

law-like relation connecting predicates from one family of

predicates true of an object with predicates from another. So,

“electron” is precise because of a law-like connection to

predicates of mass and charge. All and only electrons have that

particular mass and charge. A single non-trivial law-like

connection to another predicate family applying to the same

objects suffices. If we demand that preciseness and clear

essences require that a predicate is determined by any



characterization of an object, then no object will have an

essence, because no predicate is such that it is always

determinable whether it applies to an object under any

description. There are always predicates like the P-predicates

above.

V What happened to the paradox?

Very briefly, with a correct understanding of truth and

predication, a rejection of a “given” in which nature applies a

special predicate system to make beings real, and the realization

that therefore the truth that “`is a tall man’ applies to an

object a just in case a is a tall man,” gives a criterion for

application, there is no paradox.

For many theorists, the difficulty with this harmless

Davidsonian solution is that there is no obvious way to

generalize it to cover the semantic paradoxes. No such solution

is available for the semantic paradoxes, because they directly

generate contradictions. Thus the “indeterminable but

determinate” solution above will not work. . Thus the above

metaphysical take on the sorites abandons the quest for what

McGee has called the “Holy grail,”208 a single theory that will

handle the sorites, the liar, the sea battle, etc.

I think this quest is very much like the quest for the Holy Grail

in being directed at something that probably does not exist. The

sorites “paradox” is a phenomenon that is more or less bound to

208 At the conference mentioned in previous footnotes.



occur when either: 1) We have a language with perceptual

predicates; or 2) We have a language with many families of

predicates whose extensions overlap, have truth-conditions

related to one another, but are not precisely definitionally

related. Relative to those other predicates in the language,

applications of the predicate are sometime indeterminable.

Applications of “tall” are sometimes indeterminable in relation

to “is n meters high.” Without another family of predicates

covering the same extension with some less-than-definitional

relation to the first, there is no indeterminacy, but only lack

of knowledge. We would have “`Fred is tall’ is true if and only

if Fred is tall,” and not know whether Fred is tall, but no

sorites paradox. Some attributions of “Fred is tall” would be

like speculations about details of the past—determinate but

unknowable. There is no paradox except relative to intuitions

that there ought to be a law determining the application of one

predicate in situations described by the other predicate. To

summarize: the sorites depends essentially on relationships among

predicates or a predicate’s relation to perception. No particular

predicates are “vague” except relative to some other predicates

or to perceptual training.

The semantic paradoxes involve only a single predicate. They

arise when a semantic predicate refers to semantic predications,

either by self-reference or by quantifying over items with

semantic properties. Intuitively something about sentences saying



semantic things about sentences or semantic terms applying to

semantic terms gives rise to paradox. The cases are not alike at

all. From Chapter 7, the “future contingents” part of the Grail

is also illusory. Seeking the Grail might still be fun, I guess,

but you don’t get to ride horses anymore, just type and think.

Chapter 8:The Sorites and Davidsonian Innocuous Epistemicism209

The sorites argument and the ontological problem it raises with

essentialism was the core of my argument in Chapters 2 and 3 to

abandon monistic essentialism, the idea that reality is

intrinsically articulated into beings and properties. If there

were precise and systematic delineations in micro-particle terms

of the objects of ordinary life, we could happily be neo-

Aristotelians, regarding the medium-sized objects and organisms

he took to be basic substances as respectable entities, but

entities having compound natures derived from more the more

fundamental natures of the micro-particles into which nature was

articulated.

According to the relative essentialist account of ontology,

vagueness could seem to be a feature of objects themselves, not

only of language. The objects and properties posited are by and

large informally posited by whole cultures or organic lineages.

209 An early version of some of this chapter was presented at the conference on

“Davidson and Pragmatism” at the University of Zurich, 27-28 April, 2010.



Precise definition relative to other posits is much too much to

expect from such posits. For the vast majority of such posits, it

is quite unlikely that anything in the positing would provide

necessary and sufficient conditions for an entity being of the

kind posited in terms of any other posits. Furthermore, for

perceptual predicates, those primarily directly applied by

acquiring involuntary perceptual beliefs, it is very unlikely

that those perceptual dispositions will divide possible cases

into those which elicit the belief that the predicate is true in

a case and those which elicit the disposition to believe the

negation of the predicate is true.

The view I will argue for is that the objects themselves are

quite precise, and have sharp borders, in a sense. It will always

be true that “is red” is true of an object if and only if it is

red, even though at some point in a continuous change from red to

orange, speakers cannot say whether or not a color-patch is red.

Likewise, everything is either a chair or not a chair. This

chapter will argue for a kind of innocuous epistemicism without

truth-makers.

Predicates subject to the sorites, that is, almost all

predicates, divide into two kinds: First, there are a few one-

dimensional predicates, such as color-words and comparative

adjectives of dimension, such as “tall” and “warm.” Such features

can be placed on a continuum and ordered, so that for any two



items, either the first is taller than the second, or vice versa,

or they are equally tall.

The vast majority of writing on the sorites has focused on this

first kind, the one-dimensional cases. In fact, the complicated

“dimension” baldness is often treated as if it were solely a

matter of number of hairs on a head, rather than some kind of

combination of numbers and distribution and location (consider

Ben Franklin with his wide expanse of bare scalp, but dense hair

at the edge). The illusion is thus generated that the sorites can

be fixed by dealing with the arbitrariness of any dividing line.

In the first kind of problem predicate, since we have a continuum

along a single dimension, we can imagine that the difficulty is

that of coming up with a demarcation. All we need to do is decide

where to draw the line. Since the language-community and our

language-learning practices have not established such a line,

unlike the case with “adult” or “is old enough to genuinely

consent to sex,” the natural suggestion is that there are a

variety of acceptable demarcations, yielding supervaluation.

Another idea is that we really use a logic with tolerance for

either “is both tall and not tall” or “is both tall and not tall”

can be assigned to language users which will accommodate these

borderline cases while preventing speakers from inferring that

every sentence is true.



Second, there are predicates such as “bald,” “chair,” “human

being,” “nice”210 and virtually every count-noun. In such cases,

there is not only more than one dimension, but it is often

difficult to see how the question could be put in terms of any

number of anything that would seem to be any number of

dimensions. There is nothing close to a well-ordering in the

borderline area. I am quite certain that a collection of objects

such that it is difficult to say whether they are chairs could

not be ordered in any helpful way. There will of course be

objects such that one is clearly more a chair than the other.

For predicates like “chair” or “person” it is hard to see how any

kind of precisification would go, since there are so many ways a

thing could be defective. A good example of a predicate to which

it has proven to be difficult to apply “acceptable”

precisifications is the topic of Wheeler (1980). There the

predicate under discussion is “has a right to” and the relevant

sorites step is “is not ethically significantly different from.”

A chain of cases goes from a person’s right to move their body to

unlimited property rights. But no precisification of “has a right

to” is uncontroversial. Rights and their limits are a hotly

contested topic in political theory. With some exceptions, no

theorist thinks that another theorist’s precisification is

acceptable. If “acceptable” means more than “acceptable to me,”

210 “Nice” is Rosanna Keefe’s (1998) example in a paper making some of these

points.



then “precisification” and supervaluation will not be effective

solace here.

I have always taken the fundamental question raised by the

sorites to be metaphysical rather than logical.211 The most

important lesson of the sorites is that the natures of the

objects of ordinary life and their properties cannot be complexes

of the natures of the objects that the sciences tell us about.

The central motivation for the relative essentialism developed in

Chapters 2 and 3 is to save the reality of medium-sized objects

and their features from the hegemony of physics. Relative

essentialism allows medium-sized objects to have natures. It does

not have to describe how such natures relate to the basic natures

of reality, because it denies that there are any basic natures.

The sorites makes the problem and solution clear.

In Wheeler (1975), I suggested that the sorites posed a serious

dilemma for monistic essentialism, a view I was then considering.

If the world consists of given individuals and properties, and

predication is understood as a property being instantiated in an

individual, then for any property and any being, that property is

either instantiated or not. If Aristotle were right that medium-

sized objects are the basic entities of the sub-lunary world, in

the sense that the fundamental necessities are necessities about 211 Eubulides, the inventor or purveyer of the paradox, likewise thought the paradox was primarily of metaphysical significance. He took it, along with hisother paradoxes, to be a demonstration that Parmenides was right. See Wheeler (1983). As Peter Seuren (2005) also notes, Eubulides’ paradoxes set the agendafor twentieth and twenty-first century semantics and philosophy of language.



people, and other organisms, then the sorites might be a

problem,212 but not a threat to the existence of the objects of

ordinary life. Aristotle did not regard heaps as substances in

any case, so acknowledging that heaps are not well-defined beings

might have been acceptable.

The sorites becomes more than a logical puzzle, though, when it

develops that the strongest laws are not about medium-sized

objects and organisms, but about micro-particles.. Given that an

Aristotelian understanding of necessity is natural necessity,

having the strongest laws means being more fundamental. So, given

the naturalism shared by Aristotle and most modern

metaphysicians, this would mean that the beings with essences are

the micro-particles, and any beings with any sort of essence

would have to derive that essence from micro-particle components.

As discussed in Chapter 2, the major problem in descriptive

metaphysics is accounting for how objects which are not basic to

science and are not reducible to objects that are basic to

science, can still have essential features. Relative essentialism

is a solution to that problem along Davidsonian-Quinean lines.

Relative essentialism solves the problem by having a multitude of

independent essences and natures. Posits and their natures need

not correspond to notional “joints.”

212 Historically, it appears that Eubulides, the apparent inventor of the sorites, was attacking Aristotle. See Jon Moline (1969), as well as Wheeler (1983.)



The logical problem raised by the sorites, however, remains even

when the metaphysical dilemma it poses is neutralized by relative

essentialism. How can bivalent logic, which seems so useful and

true, be correct in light of the apparent indeterminacy the

sorites seems to reveal? A borderline chair seems to be neither a

chair nor a non-chair. Through several failed attempts, beginning

in 1973,213 I tried to find an intelligent response to the

sorites, which seemed to me to pose an intractable problem for

any philosopher of almost any persuasion. This chapter is my

latest attempt.

I What is the sorites paradox about?

The sorites paradox was originally presented as the inconsistency

of the propositions that a) a single grain is not a heap, that b)

no addition of a single grain to a non-heap turns that non-heap

into a heap, and that c) a million grains together is a heap.

Sorites sequences can be constructed for “is a chair,” “is a

person,” “delicately,” “crushes,” and almost every predicate

outside of mathematics and the hard sciences.

Since the early 1970’s the literature on the sorites has grown to

a vast size,214 so that a discussion of even the major lines along

213 At the 1973 APA Eastern meeting I presented “A Solution to Wang’s Paradox,”a nihilist “solution,” which became Wheeler (1975). See also Wheeler (1979), (1986), (1991).214 In the early 1970’s a philosopher could read everything that had been written on the sorites problem in the Twentieth century in a leisurely afternoon, with time for a nap. I did.



which philosophers have sought a solution is out of the question

here.215

There is reason to think216 that Eubulides, the originator of the

sorites, took the sorites to be making a metaphysical point. The

medium-sized objects and pluralities of the common-sense world do

not make logical sense, so believing in their existence is

incoherent. The theory here presented agrees with Eubulides that

the central problem of the sorites is metaphysical rather than

logical. The present theory relies on neo-Davidsonian views of

predication, kinds, and truth.

There are basically three responses to the sorites, given that

nobody thinks a single grain is a heap:

1) Deny that the sentences are inconsistent, when properly

understood, or that the inconsistency has serious consequences.

This is the solution of the supervaluationists and the logic

adjusters.

2) Deny proposition b), and claim that at every point in the

progression, a collection either is a heap or is not a heap. The

problem is just that for a large number of cases no answer seems 215 Excellent guides to this literature include Williamson (1994) and Keefe and

Smith (1997).216 See Wheeler, (1983). I argue that the point of the sorites, as well as of the other paradoxes Eubulides constructs, is to argue that Parmenides is right. Interestingly, after millennia, the paradoxes attributed to Eubulides are at the center of the philosophy of logic and the philosophy of language. Besides the Sorites, Eubulides proposed paradoxes involving presupposition, intentional contexts, andthe liar paradox.



objectively correct. One version of this position is the

epistemicist solution; we will propose another version below.

3) Deny proposition c), and claim that in fact there are no

heaps, as the argument shows. This is the nihilist217 solution.

Each of these strategies has had recent advocates. The first

strategy has generated the largest literature.

The present theory is akin to an epistemicist account. It turns

on metaphysical views, borrowed or adapted from Davidson, about

language, predication, kinds, and truth. Being adapted from

Davidson, the theory is naturalistic, and eschews language-

transcendent concepts.218 From a Davidsonian or Quinean point of

view, of course, there can be no epistemicism resting on

inadequate grasp of concepts. Language is to be understood as

grounded in responses of organisms to their surroundings in a

social environment. There is no chance whatsoever that concepts,

as internal contents of predicates, could be constructed out of

such material which would divide all possible objects into those

which satisfied the predicate and those which did not. 217 “Nihilism” became the label for views that deny that sentences aboutmedium-sized objects and their properties are true. Wheeler (1975, 1979) and Unger (1979) proposed this view in the 1970s. Unger proposed“nihilism” as the label for the view, claiming that even more things than Wheeler thought fell under the sorites argument. More recently, Ted Sider and David Braun (2007) and Kirk Ludwig and Greg Ray (2002) have reached similar conclusions.218 The presentation below owes its inspiration to Vann McGee’s conceptualization of the sorites problem in terms of the relationship of usageand extension, in his presentation at the conference on Truth, University of Connecticut May 15-17, 2009.



The Davidsonian proposal I will suggest is akin to an

epistemicist account in claiming truth-values for borderline

cases, but unlike an epistemicist claim, does not diagnose the

problem as some kind of limitation to our knowledge.

II Usage and Extension

As discussed in Chapter 1, Davidson supposes neither that

“concept-fitting” nor selection by privileged kinds that actually

exist determines what a predicate is true of. Only the truth-

condition clause gives the meaning of a predicate. One way to

articulate Davidson’s view is to raise the question of what the

relationship is between a term’s extension and usage.

What is the relation between the pattern of application of a

predicate within a culture and the extension of that predicate?

There are two standard general conceptions of this relation,

discussed in Chapter 1, while I will briefly review: The “meaning

is use” conception takes the content of a predicate to be

determined by some version of “what people say when,” and takes

the extension of a predicate to be a function of its content. The

monistic essentialist conception rests on natural segmentations

in the world as extensions of predicates. This section sketches

both familiar views and outlines their difficulties with the

sorites.

a) usage determines extension views



The obvious naturalistic way to assign meaning or content to a

predicate is to take the meaning of a predicate to be a function

of “use” or “usage,” and to take the extension of the predicate

to be determined by its meaning. What the members of a culture

say when fixes all there can be to the content of a predicate,

and therefore all there can be to fixing what entities the

predicate is true of. The extension of a predicate, then, is a

projection from actual applications to all possible applications.

The use/usage-is-meaning view of extensions leaves extensions,

and so truth, indeterminable. For most of our predicates, no

amount of actual application-behavior will fix a single

projection that selects a single extension within the collection

of possible objects for which the question whether the predicate

applies might arise.219 No matter how many data-points one has

derived from applications of “is a table” no particular complete

extension, from among the infinity of psychologically projectable

sets that include those data-points, is selected. Thus there are

possible (and actual) entities such that it is indeterminable

whether that entity falls within the extension of, for instance,

“table.”

b) monistic essentialist views

219 This is the point emphasized by Vann McGee in his presentation

mentioned above.



For a monistic essentialist, there is a privileged segmentation

of the world into kinds. This segmentation is reflected in laws

connecting kinds in the segmentation. These natural laws give the

essences of the kinds that are the extensions of terms. The laws

may be strict or may be Aristotle’s “always or for the most part”

laws.

Language-learning proceeds by acquaintance with such natural

kinds, which brands a given kind with a term. Alternatively, a

variety of stories about how evolution has equipped us to get at

the kinds or how our language-faculty has as its proper function

designating the right kinds are told. A metaphysical realist

conception of language and its relation to the extensions of

terms is externalist. The patterns of application of the

predicate, whether in the individual or in the society as a

whole, do not determine the extension. Usage has to have some

relation to extension in order for the reference-fixing to occur,

but that relation can be minimal.

This kind of externalism allows that, for properties and kinds of

objects that are governed by strict laws, there are no genuine

borderline cases—either a difficult-to-characterize entity is in

the extension or not. A metaphysical realist conception explains

the divergence between what one’s language-teachers teach and the

truth by appealing to a natural division in nature which selects

some extensions as appropriate extensions for predicates.



c) Difficulties with the sorites

c1) extension is a function of use theories

For the meaning is usage theorist who takes extensions to be

determined by meaning, the sorites argument shows that for almost

every predicate, the meaning or sense of the predicate, if

resting on what people say when, does not determine an extension

even in familiar and often-encountered cases. Nothing about my

culture’s history of verbal behavior defines an extension for

“chair,” in the sense of sorting the possible objects into the

chairs and the non-chairs. The set of extensions that accord with

actual usage is insufficiently restricted to sort even the

actually available objects into the chairs and the non-chairs. An

account of meaning as resting on usage rather that nature seems

to condemn the usage-as-meaning theorist to incomplete meanings,

to multiple truth-values, precisifications, and the like.

c2) monistic essentialist

The monistic essentialist solution seems to address the problem

of determining a single extension to a term by assigning that job

to nature. However, monistic essentialism has difficulty in

application to medium-sized object predicates. If we interpret

necessity naturalistically, and treat natural kinds as determined

by natural laws, then a segmentation into natural kinds requires

that the natural kinds be the subjects of laws. Natural kinds are

supposed to have essences that yield necessary truths about when

they apply. For natural objects, those essences are expressed in



natural laws. But the laws about medium-sized objects and

organisms at best admit exceptions. They are loose relative to

the laws of physics, chemistry, or even cell biology. The sorites

illustrates that, for instance, even though we have a necessary,

perhaps probabilistic, relation of some kind between being a tall

man and having an adequate height in meters, there is no lawlike

relation that would determine what that height in meters must be.

As we have noted in Chapter 2, the basic idea of monistic

essentialist accounts of reference, that reference is fixed by

divisions in nature, seems to run afoul of the sorites, at least

on the assumption that the medium-sized objects of ordinary life

exist and so have essences.220 Apart from some quite unusual

predicates,221 kinds of medium-sized objects and their properties

are intuitively not completely determined by a privileged

segmentation in nature. At best, nature sets parameters within

which wide variation is possible. For objects such as tables and

turtles, if we imagine a particle-by-particle dismantling, there

appears to be no objective line at which the entity in question

220 For the arguments that if natural kinds are taken seriously as expressed by necessary truths, then any alleged kinds whose application cannot be determined by natural laws would not be real kinds, and so would not supply essences for objects, see Wheeler (1975). 221 Black holes, unlike most big objects, have a natural law-governed delineation, the Schwarzchild radius, defined by the escape velocity of a particle being beyond the speed of light. So black holes are precisely definedentities, as far as I can discern. I cannot think of other examples of macro-objects such that there is a precise point at which their borders begin.



ceases to fall under the extension of “is a table” or “is a

turtle.”

A monistic essentialist who does not suppose that there is an

objective answer to questions about extensions of medium-sized

object count-nouns and properties becomes a usage-theorist about

all the predicates of ordinary life. In effect, supervaluation

abandons metaphysical realist selections of a privileged scheme,

and supposes that acceptable precisifications are acceptable

extensions for problem terms. “Acceptable” is then given by

usage.

Monistic essentialists thus are faced with awkward choices about

what to say about medium-sized objects and predicates of them. On

the one hand, it would be nice if there really were tables and

people. On the other hand, the whole idea that having an essence,

i.e. having objective existence and extinction conditions, is

required for reality is undermined by medium-sized object-

predicates and their vagueness with respect to other families of

predicates.

d) Davidson

d1) externalism without natural kinds

As discussed in Chapter 1, Davidson is an externalist about

extensions, but does not believe in a privileged, given

segmentation. Thus, for Davidson, all kind predicates that

actually apply to objects, i.e. all predicates for which there

are affirmative true sentences, are ontologically on a par. Of



course there are electrons; of course there are tables, but there

are no leprechauns. For Davidson, we learn to apply terms by

triangulation, coming to call an object salient both to us and to

another by the same term. Davidson characterizes this

triangulation and its consequences as follows:

“Ostensive learning works first and best with whole sentences, in

practice often represented by what for the experienced speaker

are single names, common nouns, adjectives, and adverbs (`Mama’,

‘Man’, ‘Come’, ‘Good’, ‘Careful’). The child who has no more is

still a pragmatist. Once some grammar is in hand, however,

separately learned parts can be assembled in new ways, and truth

separates from the merely useful or approved. The references of

names, the extensions of predicates, the combinatorial devices

themselves, are in the hands of teachers and society; truth is

not.” (Davidson 2005, page 15)

Since there is no privileged segmentation into objects and

kinds of objects, there is much latitude in what groups of

salient objects are correctly called by the same term. Thus,



cultures and within cultures over time. Language is a human

creation, and human language-behavior shapes the ontology that

can be assigned to a culture. That is, given that most of what

people say is true, patterns of labeling behavior will shape the

extensions an interpreter assigns to a predicate.



According to our relative essentialism, which I am treating

as Davidson’s implicit view, an ontology is an imposition or

positing of a set of predicates for putting the world into

subject-predicate form, in order to allow logical relations among

truth-functionally simple sentences. As we discussed in Chapter

3, “positing” should be thought of as by and large automatic,

coming from biological inheritance and culture. Only rarely do we

actively posit such things as gravitons and gluons and sit-coms.

Since there is no privileged segmentation, whatever can become

salient to humans is a possible partial extension of a single

predicate. But of course for the reasons given in Chapter 1 in

discussing the usage account of extension determination, no

amount of such training or decision-making about what to say when

will yield a single extension for an individual or for a culture.

That is, there is no projection from any finite amount of

identification of elements of the set to any particular set.

Since Davidson explicitly denies a privileged segmentation, the

difficulty is even more transparent.

For Davidson, though, extensions are not functions of the

content, in the sense of “content” that would be a projection

from usage. As emphasized in Chapter 1, the meaning of a

predicate is its truth-conditions. Extensions are given by the

deflationary-sounding formula, “`Is a dog’ is true of an object A

if and only if A is a dog.” Just as truth-conditions or meanings

of sentences are given in homophonic translation, so satisfaction



or truth-of conditions, that is, extensions, are given in

likewise apparently unilluminating form. So, even though meanings

and extensions are learned by imitating usage, the meaning is not

identical with the usage. Furthermore, for the reasons given

above that usage cannot select a single set on the basis of a

finite number of occasions of use, extensions, while not

independent of usage, could not be determined by usage. Usage is

evidence for meaning when we are interpreting, but meaning itself

is given completely by predicate-clauses in a truth-definition.

For Davidson, most of what people say using these divergent

predicates is true. For a Davidsonian, then, there are an

indefinitely large number of overlapping natural kinds, as it

were. All the distinct predicate-systems of different cultures

and all the overlapping kinds are correct posits, as long as

there are truths of the form “ExFx.” Given Davidson’s

externalism, a culture’s divergences from another culture means

that each culture’s entities are by and large real.222 Objects

overlap without being reducible one to the other. Diverse objects

can co-exist and overlap without getting in one another’s way.

Most importantly, diverse overlapping families of predicates can

overlap within a single culture, so that the distinct objects can

coincide.

d2) extensions

222 Gods, demons, auras, Guardian Angels and such can be exceptions. Davidson’spoint is that you could only come to think the other is mistaken by thinking the other is correct about a lot of other judgments.



As we discussed in Chapter 1, the combination of externalism

about reference and denial of a privileged segmentation means

that Davidson can have objective extensions without supposing

that there is a single division of the world into kinds, and so

without having to rescue weakly-law-governed entities from the

threat of not being really part of what is. For Davidson, all

kinds are on a par ontologically. Some kinds are connected to

other kinds by very good laws; others by not so very good, “for

the most part” laws or by the kind of very vague generalization

that would tell us that tables have to have a fair amount of

matter.

Here, then, is my Davidsonian view about extensions and usage:

1) Predicates have extensions. The union of the extension of a

predicate and the extension of its negation is the universal set.

There is an answer, “yes” or “no” as to whether a given predicate

F is true of a given object A, for any A. Given that the meaning

of a predicate is given by its truth-condition clause, a

predicate’s meaning in a sense trivially “fits” what it is true

of.

2) Although of course learning a language is finding out about

the extensions of predicates by learning to detect elements of

their extension either directly or indirectly. We both learn to

detect perceptually when a predicate F applies and we learn F’s

connections to other predicates, and so can apply F by knowing

what else is true in the case at hand. But no finite amount of



observation or training or collation of our culture’s

application-practices will prepare us to put every candidate for

“member of the extension” of a predicate, at least in general.223

That is, when we encounter or consider objects, always under some

description, and so with some posited nature, we are not equipped

in general to assign them either to F or to not-F.

3) Learning the extension of a predicate is only rarely learning

necessary and sufficient conditions using other predicates that

determines when a given predicate applies to a given object.

4) Extensions of predicates are not generally determinable by

determining extensions of other predicates, but are determinate.

That is, it can be in principle impossible to determine what the

extension of a predicate is for the general case that includes

every possible object.

Thesis 1) commits a Davidsonian to bivalence. One-dimensional

borderline cases of a predicate P are cases where one cannot tell

directly whether P is true of the case, P is known to depend on

nothing but truths from predicates in family Q, and all the 223 For the unusual cases where we can give the extension in other terms, such as “prime” and “composite” applied to natural numbers, membership in one or the other of the extensions is not determinable relative to some predicate families applying to numbers. “The number of planets” was once thought to be a prime, for instance. “Prime” and “composite” are only determinable relative to some ways of specifying numbers. This is masked by the fact that there is a procedure, in principle, for converting number-description of the form “the number of F’s” to a description for which there is an algorithm for determining whether a number is prime.



relevant evidence from Q predicates is in. Given that there are

borderline cases where nothing can indicate to us whether a

predicate truly applies or not, there are sentences that are

determinately true or false when those sentences’ truth-values

are not determinable. Not being determinable does not mean not

being determinate.

Davidson and Quine would both call such situations

“indeterminacy.” Throughout this book, I have called such

situations “indeterminability.” So, I have spoken of

“indeterminability of interpretation” for instance. Why should we

think that determinacy requires determinability and that

therefore indeterminability is indeterminacy? The basis for this

view seems to be descended from the Verification Theory of

Meaning, via the Verification Theory of Meaningfulness. The idea

has been that, unless there is a possible procedure for detecting

whether a predicate applies in a case, it does not make sense to

suppose that the question whether the predicate applies or not is

not a question of fact. But the verification theory of

meaningfulness can only be obvious if one thinks that there is a

given level of experience or something relative to which all

questions are decided. Davidson argued against any such given in

Davidson (1974). There being no other terms relative to which a

given term is determinable would only imply indeterminacy if

there were other terms that were intrinsically more basic.



If the meaning of a predicate is given by its predicate clause,

and there is no given articulation of the world into basic beings

relative to which all other beings must be understood, then there

are no given terms, either. So there being some situations in

which it is in principle impossible to determine truth-values of

predications does not mean that those predicates have no truth-

values.

Here is an example of an absolutely indeterminable predicate

whose truth or falsity of a given object is determinate. Consider

the infinite family of mathematical predicates P1, P2,… where P1

is “=2 if the continuum hypothesis is true, = 4 otherwise,” P2

is “= 3 if the continuum hypothesis is true, = 6 otherwise”, and

where Pn is “= the nth prime if the continuum hypothesis is true

and = the nth composite otherwise.” Relative to any of these

predicates, “prime” is indeterminable. From the information that

the number of my first cousins is P5, nothing can determine

whether the whether the number of my first cousins in 11 or 10.

It can be determined that it is either the 5th prime, 11, or the

fifth composite, 10. So, it given that I have 10 first cousins,

“The number of my first cousins is P5” is either true or false,

but absolutely indeterminable. But, supposing that Platonism is

coherent, that this predicate is always determinate but never

determinable is also coherent. So there is no necessary

connection between determinacy and determinability.



It is of course true that predicates which have very little

connection to any other predicates and are rarely directly

determinable by observation, so that predications using them can

only rarely can be determined to be true, will be useless, and so

in a sense meaningless. But it is hard to see what argument could

be given that determinate truth-value requires determinability,

absent a ground-level of basic predications that are given by

nature.

So, a Davidsonian can be an epistemicist without supposing hidden

facts or states of affairs making such sentences true. There are

hidden facts in the notional sense described in Chapter 6, of

course, but the “inaccessibility” of such facts is not a defect

in our knowing abilities. For Davidson, truth is primitive.

Sentences are not “made true” by anything. For a variety of

reasons,224 there are no entities corresponding to true sentences.

Since there are no truth-makers, and truth is primitive, truths

do not have particular chunks of the world or particular

environmental incidents to fasten onto.

So, all terms have perfectly clear extensions, given by the

predicate-clause in the truth-definition.

III Theoretically justified bivalence and Davidsonian pragmatism

224 Davidson is an advocate of the Slingshot, examined in detail by Stephen Neale (2001). In any case, facts and other concrete correlatesof sentences are suspect on many grounds. There is the problem of binding, which goes back to Plato’s Parmenides, there is an implausible population of negative and general facts, and so on.



A Davidsonian, following Quine, can allow that some truths,

typically true standing sentences and their instances, are

reasonably held to be true in virtue of theoretical

considerations. Useful analogies abound in mathematics. Neither

usage nor intuition would suggest that among the sub-groups of my

siblings there is the null set. Yet it is a theorem that the null

set is a subset of every set. We need that theorem to be true in

order to retain such principles as that everything that is a

member of the subset is a member of the superset. Likewise, the

idea that for any number, raising it to the zeroth power yields

one as value fits no intuitions about what “raising to the zeroth

power” means, since the notion is intuitively meaningless. This

truth is true because, among other things, n to the mth divided

by n to the pth equals n to the m-minus-pth. In both cases,

theory demands that a sentence be true. Accepting that truth is

harmless in both cases.

Davidson can say the same thing about the thesis of bivalence for

sentences using medium-sized object predicates. In cases in which

“there is no fact of the matter” there is no importance to the

matter either. Just as we run into no practical difficulties in

treating five to the zeroth power as equal to one, so we have no

practical difficulties in treating borderline cases as being true

or false, but it does not matter which.

For the reasons outlined below, namely the lack of sharp laws

connecting families of medium-sized object predicates and the



lack of perceptual determinations, it would be expected that

cases would arise where a predicate such as “is a tall man” could

neither be applied nor denied on the basis of inspection, even

though a predicate from another family, “is 1.78 meters tall”

could be, and whether an individual is tall depends on nothing

more than how tall the individual is. Without correlates of

sentences, i.e. facts or states of affairs, the phenomena of

vagueness are either cases of incomplete connection among

heterogeneous predicates or cases where perception does not

determine an answer and nothing else is relevant. But since all

it takes for “Joe is a tall man” to be true is that Joe be a tall

man, “is a tall man” has determinate truth-conditions.

In the case at hand, sorites arguments, the Davidsonian answer

would be that of course classical excluded middle holds for

sentences about medium-sized objects. There turn out to be good

reasons why, for many sentences, truth is not in principle

determinable. Among those sentences are predications which usage

does not fix as true or false. So, for example, at every point on

a continuum forced march, there is an objective answer as to

whether the entity is a tall man, even though there is in

principle no way to tell.

Just as in the cases of mathematical truths accepted for

theoretical reasons, accepting bivalence in non-semantic225 225 Given the existence of the semantic paradoxes, the general claim of bivalence, that every sentence is true or false, may be false. But therestricted claim that sentences attributing predicates to medium-sized



predications is harmless. As long as we do not suppose that there

are occult facts or other truth-makers, the claim that every

sentence is true or false, so that “Fred is tall” is true or

false, conflicts with no other claims we should wish to make. The

view is an epistemicism insofar as it assigns unknowable truth-

values to borderline cases. This acceptance of bivalence requires

no miracles nor does it require concepts to which we have

imperfect access. It just requires that we treat truth,

extensions, and meaning as completely given by the relevant

clauses in a truth-definition. Meaning is truth-conditions, where

that means that beyond “`Fred is a dog’ is true if and only if

Fred is a dog,” there is nothing illuminating and accurate and

general to say.

IV What is vagueness without truth-makers?

A presumption of the sorites is that the situation where a

man is 1.78 meters tall demands an answer about whether he is a

tall man or not. Why is an answer expected? Briefly, we have a

datum, a fact before us, specified by “is 1.78 meters tall.” In a

world in which that fact exists, it must, it seems, be the case

either that the man is a tall man or that he is not a tall man.

But suppose, as Davidson does, that there are no entities

corresponding to true sentences, i.e. no facts or states of

affairs. Then the truth-conditions of “Fred is 1.78 meters tall”

objects seems to have everything to recommend it.



are just that Fred is 1.78 meters tall. The problem with “Fred is

a tall man” not being determinable is just that from “Fred is

1.78 meters tall” neither “Fred is a tall man” nor its negation

follows by law. So, one can infer neither the sentence nor its

negation from the “datum.” The point is that the datum is not a

given, but is already-conceptualized. The datum is a truth. So

why should it be surprising that neither “Fred is a tall man” nor

“Fred is not a tall man” can be derived by laws from another

sentence, “Fred is 1.78 meters tall?”

What gives rise to puzzlement is the truth that there is nothing

to Fred’s being tall other than how tall Fred is. So, we think

there ought to be a law and there isn’t. For well-behaved

predicates like “tall,” 226 there are indeed some lawlike truths

connecting “n meters tall” “taller than” and “tall man.” For

instance, if Joe is 1.8 meters tall and is a tall man, then any

man taller than 1.8 meters tall is a tall man. Height is the only

relevant dimension for “tall”, which makes it a favorite among

sorites theorists. But even though the family of predicates “is n226 “Tall” is well-behaved in many ways. One feature of being one-dimensional, unlike “nice” and “bald,” is that one-dimensional predicates have a non-vague comparative. For many concepts there is apparent indeterminacy not only for the attributive construction, but also for the comparative construction. It can seem indeterminate whichof two men, if either, is balder than or nicer than or more obnoxious than the other. With count-nouns in relation to particle-complexes the situation is even worse. No list of dimensions is forthcoming. Of two table-like objects, no laws determine which is more a table than the other. Only the vaguest “laws” connect material count-nouns with particle-complexes.



units tall” has lots of connection with the family of predicates

“is a tall F”, the “is a tall F” family does not reduce to the

“is n meters tall” family. So, in many cases, we can know the

truth of a sentence about Fred using a member of one family

without knowing the truth of a sentence about Fred using the

other family.

Without truth-makers as correlates of sentences, “borderline

cases” are just sentences whose truth-values are not determinable

directly and are not determinable from the truth-values of other

sentences about the same object. A borderline case can arise if

there is a true sentence using one kind of predicate such that

there is no strict definitional necessary connection to the

truth-value of a sentence using another kind of predicate, but

where the two kinds of predicate are necessarily related. Since

height in meters is related to whether an individual is a tall

man, so that a man having a given height in meters is sometimes

obviously also a case of being a tall man, a sort of paradox

arises if we assume that because the truth-value of one

characterization of the situation is determinable, so must the

other be, that is, that there ought to be a law. So, “John is

1.78 meters tall” may be known to be true, while “John is a tall

man” may not be known to be either true or false..

If we do not suppose that there are privileged ways of

characterizing objects, then indeterminability arises from one

of two sources: First, we get indeterminability when only



perception is relevant to whether P, and perception just does not

divide cases into ones that elicit belief in P and those that

elicit belief in not-P. Second, we get indeterminability whenever

two families of predicates apply to some of the same entities,

have some necessary relationship, but lack precise laws relating

them. Vagueness arises from relations between predicate-families.

A predicate is vague if there are no non-trivial strict laws

connecting it with predicates from another family of predicates

applying to items in its extension.

We could define a “precise” predicate as one such that there is a

law-like relation connecting predicates from one family of

predicates true of an object with predicates from another. So,

“electron” is precise because of a law-like connection to

predicates of mass and charge. All and only electrons have that

particular mass and charge. A single non-trivial law-like

connection to another predicate family applying to the same

objects suffices. If we demand that preciseness and clear

essences require that a predicate is determined by any

characterization of an object, then no object will have an

essence, because no predicate is such that it is always

determinable whether it applies to an object under any

description. There are always predicates like the P-predicates

above.

V What happened to the paradox?



Very briefly, with a correct understanding of truth and

predication, a rejection of a “given” in which nature applies a

special predicate system to make beings real, and the realization

that therefore the truth that “`is a tall man’ applies to an

object a just in case a is a tall man,” gives a criterion for

application, there is no paradox.

For many theorists, the difficulty with this harmless

Davidsonian solution is that there is no obvious way to

generalize it to cover the semantic paradoxes. No such solution

is available for the semantic paradoxes, because they directly

generate contradictions. Thus the “indeterminable but

determinate” solution above will not work. . Thus the above

metaphysical take on the sorites abandons the quest for what

McGee has called the “Holy grail,”227 a single theory that will

handle the sorites, the liar, the sea battle, etc.

I think this quest is very much like the quest for the Holy Grail

in being directed at something that probably does not exist. The

sorites “paradox” is a phenomenon that is more or less bound to

occur when either: 1) We have a language with perceptual

predicates; or 2) We have a language with many families of

predicates whose extensions overlap, have truth-conditions

related to one another, but are not precisely definitionally

related. Relative to those other predicates in the language,

applications of the predicate are sometime indeterminable.

227 At the conference mentioned in previous footnotes.



Applications of “tall” are sometimes indeterminable in relation

to “is n meters high.” Without another family of predicates

covering the same extension with some less-than-definitional

relation to the first, there is no indeterminacy, but only lack

of knowledge. We would have “`Fred is tall’ is true if and only

if Fred is tall,” and not know whether Fred is tall, but no

sorites paradox. Some attributions of “Fred is tall” would be

like speculations about details of the past—determinate but

unknowable. There is no paradox except relative to intuitions

that there ought to be a law determining the application of one

predicate in situations described by the other predicate. To

summarize: the sorites depends essentially on relationships among

predicates or a predicate’s relation to perception. No particular

predicates are “vague” except relative to some other predicates

or to perceptual training.

The semantic paradoxes involve only a single predicate. They

arise when a semantic predicate refers to semantic predications,

either by self-reference or by quantifying over items with

semantic properties. Intuitively something about sentences saying

semantic things about sentences or semantic terms applying to

semantic terms gives rise to paradox. The cases are not alike at

all. From Chapter 7, the “future contingents” part of the Grail

is also illusory. Seeking the Grail might still be fun, I guess,

but you don’t get to ride horses anymore, just type and think.



Chapter 10 What We Ought to Do

“The ethical thing to do is to tell your wife. But I don’t think

you should”228

The reader will have noticed that the theory of what “good”

means in Chapter 9 employs the notion “ought.” This may seem to

be small progress in understanding ethical notions, since “ought”

is at least as puzzling and controversial as “good.” This chapter

will articulate and defend a Davidsonian account of “ought” and

its relation to ethics.

In brief outline, the account is that “ought,” rather than

“obligation,” is the fundamental ethical notion. “Ought,” as

Davidson (1970b) argued, is akin to “probably” in its logic. If

we understand “probably” to be a modality based on rationally

acquiring degrees of belief in the light of evidence, then we

should investigate the possibility that “ought” likewise has

something to do with rational expectations. “Ought” in its

application to cases other than human actions, indeed seems to be

true of a pair of things said just in case there is a chain of

true conditional probably-claims connecting the first argument to

the second. What ought to happen is what would happen if what

would probably happen always happened.

228 Advice columnist Amy Dickenson, (Ask Amy), Chicago Tribune, TT 500, 435 N. Michigan Avenue, Chicago, IL, 60611. (July 17, 2006)



Such an account can apply to prudential and moral “ought”-

sentences because principles of interpretation, as maximization

principles constitutive of agency, are both probabilistic and

normative. Davidson holds that interpretation must treat the

other as “a believer in the true and desirer of the good.”

(Davidson 1970a, p.222) Because this maxim of interpretation is

to maximize these features, the inference from a person being an

agent to believing a truth or desiring a good is probabilistic.

Interpretation embodies probabilistic inferences which are, of

course normative—this is what an agent ought to be. Every

principle of interpretation is probabilistic, given that

interpretation is holistic. So, the Davidsonian account of

“ought” will be based on the features that interpretation

maximizes. So, what one ought to believe and desire, and thus

what one ought to do is what reflects maximal belief in the true

and desire for the good. To the degree and extent that one’s

conception of the true and the good differs from another’s, the

truth about what one ought to do will be indeterminate. We would

thus expect that, while many ethical questions have objective

answers, many others do not.

A Davidsonian account of ethics starts with the notion of a

rational agent, the object of interpretation. “Good,” “ought,”

and other basic notions are predicates derived from the

“intentional scheme.” Questions about the objectivity of

judgments of what we ought to do that challenge the existence of



properties in nature such as “goodness” or “wrongness” have a

difficult time getting a purchase on this kind of theory. Given

the analysis of “good” in the previous chapter, and given the

account of properties in Chapter 6, the fact that goodness is not

likely to be a physically significant property of any kind says

little about the possibility that sentences using “good” and

“ought” are true. Given the view of ontology in Chapter 3 along

with the innocuous dualism of Chapter 4, the lack of systematic

fit between the predicates of science and ethical concepts is no

more a problem than the lack of systematic fit between medium-

sized object predicates and micro-predicates. They are different

things.

A critique such as Mackey’s (1977) presupposes a

metaphysical realist view about the world. On such a view, it

would be absolutely remarkable if there were a property picked

out by “good” or “wrong.” Such properties would be indeed queer.

On a view of properties as metaphysically inert, argued in

Chapter 6, and the pluralist “ontology” one arrives at by

treating beings and properties as imposed in order to think about

the world, this queerness just amounts to the lack of systematic

law-like connection between any properties of the medium-sized

object and organism world and the predicates of basic physics.

This account of what we ought to do in a situation will be

akin to an account of what we ought to believe, given some

evidence. It will not be an algorithm, there will not be a



metaphysical property “belief-worthiness” that reasonably

believed propositions have, but there will be right and wrong

applications of the predicate, based on the intentional scheme of

agent-interpretation. The difference is that action involves

desires as well as beliefs. Agent-interpretation is always

directed at actions, including speech actions. The following

account will be an extension of Davidson’s conception of rational

agency.

The first section of this chapter discusses the difference

between “ought” -sentences and “must” or “obligation”-sentences.

This difference has been ignored or understood badly by ethical

theorists, at least since Kant. As mentioned briefly in Chapter

5, these are two distinct modals with very different properties.

The second section develops a Davidsonian theory of “ought”-

sentences, and how such a theory applies to rational agents.

a) “Must,” obligation, necessity, prima facie obligation, and

“ought”

a1) “Must” and necessity

There are a number of distinct pairs of modals, including

“obliged/permitted,” “must/may,” and “has to/can” which have a

logic akin to necessity. While each has peculiarities of its

own,229 they can be treated together.

Consider these arguments:

229 Only people can be obligated, for instance, while anything that goesup must come down.



1) “If you are Sophomore, you must take Sophomore English.

You are a Sophomore.

You must take sophomore English.”

Compare the above with:

2) “If you are a Sophomore, then you are necessarily a Sophomore

You are a Sophomore.

You are necessarily a Sophomore.”

Whereas the first argument is always truth-preserving, the

second argument is only sometimes truth-preserving unless the “is

necessarily” has narrow scope in the first premise, in which case

it is valid. Otherwise, argument 2) is an instance of the

Megarian fallacy. The conditions under which it would be a sound

argument would be that people are essentially Sophomores.

The distinction between wide and narrow scope does not apply

with “must” in argument 1) as it does with “must” in argument 2).

Consider argument 2) above in conjunction with:

3) “If you failed Freshman English, you must not take Sophomore

English.

You failed Freshman English.

You must not take Sophomore English.”

Taking the first premises of 1) and 3) to be simultaneously

true, a student could have two obligations which are in conflict.

The obligations expressed are requirements. If the authorities

allow those who have failed Freshman English to become Sophomores



anyhow, the rules from which the obligations are derived are

incoherent.

Conditional obligations are conditionals whose antecedents’

truth requires the obligatoriness of the consequent. On a wide-

scope reading of conditional necessity, the truth of the

antecedent requires the truth of the consequent, not its

necessity. There is nothing as it were halfway between the

obligatory and the permissible, as truth is between the necessary

and the possible.

As formulators of deontic logic have observed for over half

a century, the pairs “is obligatory”/ “is permissible” and

“must”/”may” behave in many ways like “necessarily” and

“possibly.” Just as “not possibly not” is equivalent to

“necessarily,” so “not permissible not” is equivalent to

“obligatory.” The equivalences are similar for other pairs of

predicates and their negations. But, as the examples above show,

the “obligatory”/”permissible” and “must”/”may” pairs and the

“necessary/ possible” pair are not quite alike.

One way to see the difference is to suppose the conditional

in both conditional obligation and conditional necessity is a

truth-function.230 Then one could say that the contrast is as

follows: Conditional necessity with wide scope just says that a

230 Other conditionals are at least like the truth-function in being false when the antecedent if true and the consequent false. So other conditionals are at least disjunctions.



disjunction is necessarily true. Thus, it has to be that if one

disjunct is false the other is true.

Conditional obligation, on the other hand, says that the

conditional being true is obligatory. The modal has scope over a

conditional. So, it is your obligation that the conditional be

true and so you must make the conditional true. So, if the first

disjunct is false, you must make the other one true. So the

modality applies to the consequent if the antecedent is true.

Thus, in the case of “is necessary,” the scope of the modal

predicate, whether it applies to the conditional or just to the

consequent, matters. In the case of obligation, while there is a

syntactic difference between the modal predicate applying to the

conditional and just to the consequent, the two scopes are

equivalent in practice.

“Necessity,” “possibility,” “is obligatory” “is

permissible,” “must” “can”, etcetera, are one-place predicates of

things said. Very plausibly, conditional necessity and

conditional obligation are just ordinary truth-functional

conditionals either falling under a modal predicate or having

consequents which fall under a modal predicate. That is, these

modalities are conditionals combined with a modal.

a2) The truth-conditions of “must”

It has seemed to many people that “must” is ambiguous

between a physical and a moral sense. “What goes up must come

down” is incompatible with “Fred went up but did not come down,”



whereas “One must pay what one owes” is compatible with “Fred

owed Bill and did not pay him.” A Davidsonian will seek an

account according to which “must” means “must,” a single

predicate applied to something said. Here is what I propose: “Must” is a predicate whose truth-conditions amount to “is a

consequence of F being true,” where “F” can be just any sort of set of

sentences. The set of sentences need not be described by listing the

sentences. “The laws of nature” and “the laws of obligation” specify sets of

sentences, but not ones most of us could list. In the case of a sentence like

“Since it’s Tuesday, this must be Belgium,” “F” is the schedule of the tour.

Notice that “must p” is sometimes weaker in force than “p” alone. A

sentence that follows from some other sentences will not necessarily be true

if the generalizations from which you are deducing are only probably true.

You’re waiting for the arrival of Fred, and you hear a familiar motor in the

distance. “That must be Fred” may be true231 even when it is not Fred, but

someone else driving Fred’s car. The set of generalizations from which you

deduced that this arrival must be Fred were presumably only probabilistic. If

A is a consequence of a set of sentences B, A may be true or false, unless the

sentences in B are true. Only relative to being a consequence of true

sentences does the content of the consequent of the conditional follow. In the

case of the tourist above, the “must” relies on schedule-notations like,

“Monday: Amsterdam. Tuesday: Brussels,” and the like. Those notations’ truth

entail that if it’s Tuesday, they are in Belgium. If the tourist’s bus has

been hijacked without her knowledge, and she is in fact still in Holland, what

231 This needs some defense. When it turns out not to be Fred, we may say “No, it’s Sheila.” This is much like when someone asks what the weather will be and I say “NOAA says it will rain.” The report “No, itwas dry all day” does not say that my report was wrong, but the that probabilistic claim of NOAA was mistaken.



she said is true. On the other hand, if in the same circumstances she had said

“Since it’s Tuesday, this is Belgium” she would have been mistaken.

In many applications of “must” the conclusion of what must be is not

even probably true. In “If John owes Fred money, he must pay him” the “F” is a

system of universal descriptions of peoples’ actions in situations, roughly

set of the universal generalizations which describe people fulfilling their

obligations. “Everyone pays what they owe” would be a member of the set. The

“must” relative to these universal generalizations gives only weak support to

the conditional being true. People often fail to pay what they owe. The

probability may approach zero when the “must” is something like “Fred must

treat all person’s interests as equally important.”

If the contextually referred-to sets of sentences are true, then “must”

will agree in truth-value with “necessarily,” and what must happen does

happen. So, when the “F” is whatever the laws of nature are, or “the present

circumstances,” one is warranted in inferring “q” from “If p then must q” and

“p.” The requirements of nature and reality get enforced, as it were. Thus we

get the appearance of two “senses” of “must.” The “must” of “What goes up,

must come down,” does indeed support the inference from “Fred went up” to

“Fred went down,” but the inference is truth-preserving because the laws of

nature and the facts about reality, whatever they are, are true.

“F” is understood contextually, but is of course subject to

interpretation. If Bill says “You must to pay Fred, because you promised to,”

we could suppose that Bill is under the impression that the laws of nature

require fulfillment of promises. Since that is unlikely, it is probably not

the requirements Bill means.

a3) Anankastic conditionals

A much-discussed kind of “must”-sentence is what are now called

“anankastic” conditionals,232 which are conditionals with a clause about an

232 See for instance, von Fintel and Iatradou (2005) and Kaufman and Schwager

(2009).



agent’s desire in the antecedent. We will encounter this kind of sentence

below in the discussion of Kant’s “hypothetical imperatives.”

In much of the literature, “anankastic” covers such conditionals with

“must” and with “ought.” On my view, as we will presently see, “If you want a

Big Mac, you ought to go to MacDonalds” and “If you want a Big Mac, you must

go to MacDonalds” are two very different conditionals.

The difficulty with such sentences is that they have two readings, only

one of which seems to fit the above account. Suppose heroin is very bad for

you and that Fred is the only person who supplies heroin to users. Then “If

you want heroin, you must contact Fred,” and “If you want heroin, you must get

into a rehab program” both seem to be true. The first sentence tells you what

the requirement is, given the facts, for getting the object of your desire.

The second says what, given that you have the desire, you must do, according

to some requirements. Only the second reading accords with the form ascribed

above to conditionals with wide-scope “must” in the consequent. Nothing

requires that that a person do what she wants to do.

My account of the first case is that the “must” really applies to “You

get a Big Mac, only if you go to McDonalds.” The facts, namely that Big Macs

are available only at McDonalds, require this conditional to be true. The

normal case is that one’s wants are also in one’s interest and in accord with

any requirements one may recognize. In those cases, what one must do, given

that one fulfills this want, is what is required to fulfill that want. Of

course, strictly speaking, a desire does not obligate a person to do anything,

unless it is a desire that must be changed or satisfied. Morality demands that

desires to torture people be changed, since that is a bad desire. But, in

other cases, the sentences are not literally true.

The fact that you want a Big Mac does not require you to go to

MacDonald’s. However, normally, when your wants are acceptable, what you must

do given a want, all things considered, is, and given the facts, is the only

way to satisfy that want. This point has been obscured by the identification



of “must” and “ought.” As we will see, there are things you ought to do, if

you want a Big Mac. As it happens, when there is only one place to get a Big

Mac, what you ought to do and what you must do will agree. The agreement in

this kind of case will lead Kant to think of conditional “must” and “ought”

sentences as imperatives and as synonymous.

In the general case, though, anankastic conditionals with

“ought” in the consequent say that the content of the consequent

is optimal given the circumstances. For example, that you ought to

take I-84 and connect with the Mass Pike to get from Hartford to

Boston does not imply that there is no other way. I-91 to Route

20, and taking a town road in Sturbridge to Route 9, will get you

to Boston, but is quite a bit longer. However, if there are

construction delays on the Mass Pike, or if you want to enjoy the

scenery in the Brookfields and have lunch in Worcester, the

longer route might be the one you ought to take. Relative to

other circumstances, other actions may be optimal. While

anankastic conditionals with “must” could be construed as

hypothetical imperatives, such conditionals with “ought” cannot

be. They are recommendations. We will discuss below how

anankastic conditionals with “ought” can justify the analysis of

“This is a good hammer” as approximately, “If you want a hammer,

you ought to want this.”

a3) “Ought”

The logical patterning of conditional “ought”-sentences is

quite different both from conditional “must” sentences and from



conditional necessity. Both “If it is her birthday, you ought to

kiss Susan” and “If Susan has the stomach flu, you ought not to

kiss Susan” may both be true while Susan has the stomach flu on

her birthday. With “must” or “is obligatory” replacing “ought” in

these sentences, the situation would be incoherent. In the

situation where Susan has stomach flu on her birthday, there are

not two things you ought to do, but only one, whatever you ought

to do given that Susan has the stomach flu on her birthday, which

is presumably not kiss her, most of the time.

What you ought to do given that Susan has stomach flu on her

birthday, though, does not definitively show that you ought not

to kiss Susan. While “If Susan has the stomach flu on her

birthday, you ought not to kiss her” is true it may also be true

that “If kissing Susan on her birthday and she has stomach flu

will awaken her from the evil queen’s spell, you ought to kiss

her.” In that case, presumably, you should take the risk of

catching the flu, the fact that it is Susan’s birthday paling in

significance in the context of the evil queen’s spell.

As Davidson (1970b) building on Hempel (1960) pointed out,

this logical patterning is very like that of conditional

probability claims. Given that “If A then probably B,” with

“probably” having wide scope, says the same as the more technical

“The conditional probability of B given A is high,” it can be

true both that “If Joe is in the NRA, he is probably a

Republican” and that “If Joe is for legalizing recreational



drugs, he is probably not a Republican” while Joe is an NRA

member who is in favor of legalization. Conditional probability

is a relation between the antecedent and the consequent. As noted

in Chapter 5, there is a proof that conditional probability

cannot be the probability of any conditional. That is, unlike

conditional necessity and conditional obligation, “if…then

probably…” cannot be a modal applied to a conditional. Of course

“If A then probably B,” with “probably” having narrow scope can

be a conditional with a “categorical” or “all things considered”

“probably” sentence as consequent.

“Probably” I understand to be primarily about inductive

support, and so reasonable degree of belief, relative to some

evidence. That is, it is a modality dealing with rationality. The

concept of the objective likelihood of an event, as in quantum

mechanics, I take to be a kind of extension of this notion

If the logic of “ought” is like that of conditional

probability, then there is no mystery whatsoever about how these

“ought”-sentences can all be true. Just as, relative to different

evidence, the conditional probability of an event can be

different, so relative to different considerations, what you

ought to do can be different.

The inferences that proceed via “ought”-sentences would

likewise be inductive. In general, no matter how high the

conditional probability of conclusion C given evidence (p1…pn),

the conditional probability of C given (p1…pn, pn+1) may be quite



low. No matter what ironclad evidence one has that Fred is the

murderer (Fred hated Bill, was there when Bill died, had

threatened Bill, etc.) when it emerges that Fred died before the

murder took place, the case against Fred collapses. In

induction, additional premises can weaken a strong argument for a

given conclusion. In deduction, if an argument from premises (p1…

pn) to conclusion C is valid, the argument from any number of

premises that include (p1…pn) is also valid. If every model on

which p1 through pn is true is also one on which C is true, then

a fortiori any model on which p1 through pn, pn+1…pn+m is true is

one on which C is true.

Because “ought” has nothing close to the Kolmogorov axioms

as constraints on any possible application of an “ought”-

predicate to a conditional, there is no rigorous proof that a

conditional “ought”-sentence could not be an “ought” predicate

applied to a conditional, i.e. that the conditional “ought”

cannot be the “oughtness” of some conditional. However, the

examples from Davidson (1970b), as well as the analysis below of

conditional “oughts” as conditional probabilities should make the

thesis plausible. If the analysis is correct, the proof for

conditional probability will be a proof for “ought.” To summarize: There are important logical differences between “ought”-

sentences and

“must”-sentences. Obligations, the entailment relations between

sets of sentences and the content of the “must”-clause, remain



whatever else turns out to be the case. On the other hand,

conclusions about what one ought to do can change in the light of

additional information. As the examples illustrate, a strongly-

supported “ought” can be undermined by additional information.

There is every reason to treat these modalities as distinct, and

not to use “obliged to” and “ought” as synonyms. Nevertheless,

the continuing tradition in ethics treats these notions as

essentially the same. It is worthwhile to speculate on why that

has been the case.

a4) Assimilating “ought” to “must”: KantThe assimilation, against all the evidence, goes back at least to Kant.

Given Kant’s prominence in shaping thinking about ethics, it is worth seeing

how this assimilation takes place. Kant’s Groundwork of the Metaphysics of Morals

contains a very influential discussion of the logic of “ought”-sentences,

identifying them with “imperatives” and distinguishing between “hypothetical”

and “categorical” imperatives. This is of course a great work, but it has a

blind spot about “must” and “ought.”

Kant’s discussion in the opening passages is designed to show the accord

of Kant’s theory with peoples’ ordinary, pre-theoretical moral and semantic

intuitions concerning “ought.” Kant is implicitly arguing that the categorical

“ought” is a categorical “must” by asserting that the conditional “ought” is

really a conditional “must.”

However, the examples Kant gives of “hypothetical

imperatives” do not have the characteristics he ascribes to such

imperatives. Kant’s examples fit the logic of “ought” rather than



“must.”233 Such an “ought,” if Kant had taken it to be the core

notion of ethical reasoning, would yield an ethics very different

from the one Kant thought was obviously right.

As discussed above, conditional “ought”-sentences with a

desire-clause in the antecedent are not necessitations. So the

233 Consider the following passages: (The paragraph numbering is based on the Academie Edition of Kant’s works (1911), Volume Four. The translation is mostly James Ellington’s (1983). 1) 413 “All imperatives are expressed by an ought…” The German says “Sollen,” not “Mussen.” “Ought” and “should” are the appropriate translations. “Must” would be the appropriate modal in imperatives, so an implication is that “ought” and “must” are synonymous. 2) 414 (my emphasis) “Now all imperatives command either hypothetically or categorically. The former represent the practical necessity of a possible action as means (The translator inserts a definite article,giving “as a means.” Neither a definite nor an indefinite article precedes “means” (Mittel) in the above passage. However, the logic of “practical necessity” and congruence with the passages on 417 below would seem to require “the,” if an article were needed. So the insertion of “a” by the translator is uncalled for.) ……Every practicallaw represents a possible action as good and hence as necessary……Now if the action would be good merely as [a] means to something else, so is the imperative hypothetical.”3) “the only means,” “the means,” and “the sole means”

The passages below (with my emphasis) show that Kant’s conceptionis that “ought” means “must” and implies that willing an end requires willing a particular means. Given an end, there is a single means thatmust be willed also.3a) 417 “But when I know that the proposed result can come about only by means of such an action, …”3b) 417 “Whoever wills the end wills the means that are indispensably necessary to his actions and that are in his power.”3c) 418 “…whoever wills the end also wills…the sole means thereto which are in his power.”



“ought” in such sentences is not a “must.”234 Of course, as noted

above also, there are anankastic “must” sentences which are

appropriate when the desire itself must be dealt with or there is

only one way to satisfy it.

There are some consequences of Kant’s being convinced that

“ought” means “must.” Kant’s categorical imperative, using “must”

and therefore having a deductive logic, allows an agent to

determine what must be done given limited information. Once an

Necessitation would follow from “sole” “the” and “only.” “Must” would indeed be appropriate and the sense of “ought” in “hypothetical imperatives,” since the proposed end would indeed select exactly one means.

The examples of hypothetical imperatives Kant gives in this exposition are as follows:415: “The prescriptions needed by a doctor in order to make his patient thoroughly healthy and by a poisoner in order to make sure of killing his victim….”417: “Mathematics teaches us …that in order to bisect a line accordingto a sure principle I must from each of its extremities draw arcs suchthat they intersect.”

These two examples obviously fail to fit Kant’s characterization of the “hypothetical imperative.” Kant’s examples both mention means that have alternatives and so are not necessitated. He uses “must” butthe examples would only support an “ought,” in the circumstance that these are optimal solutions. There are many ways to poison someone andmany procedures for bisecting a line. So, these are not cases where “the only means,” “the means,” or “the sole means” is being given for satisfying a particular desire, so “must” is inappropriate. Only Kant’s willingness to treat “must” and “ought” as synonyms allows these examples to seem plausible.

234 In fact, there is a weak implicature from “ought” to the negation of“must.” “If you want to major in Philosophy you ought to take 211” is misleading if 211 is in fact required. “Ought” weakly implies that youhave a choice rather than a requirement, unlike “must.”



agent knows that an action would be a lie, for instance, its

moral character is determined. New information about the

circumstances in which a person is tempted to lie are irrelevant,

just as, if X implies Y, X conjoined with any other sentences

whatsoever also implies Y. If new information could change an

ethical evaluation, a good will would not protect you from doing

the wrong thing. Circumstances that an agent was not aware of

could make the negation of a now-justified “ought”-judgment true.

You could, with the best of intentions, maxims, and care, do the

wrong thing. If “ought” is like conditional probability, this

would be the case. A person always acts on less-than-complete

information, and if missing information could change the

judgment, action would always be possibly wrong.

What motivates Kant’s assimilation of “ought” to “must”?

Kant knew the answer he expected to get about the logical

structure of a moral theory. Kant thought he knew that a correct

ethical theory would be a system of laws. Such a system would

provide a person with the possibility of never knowingly doing

the wrong thing. If morality is such that a human can be sure of

doing the right thing given limited information, it must be based

on principles whose application to a situation determine the

moral quality of the action. The idea that any wrong action is

one that could in principle be avoided by sufficient care and



obedience is reflected in the New Testament235 as well as in the

Hebrew Bible.236

If new information could undermine an ethical conclusion,

then, since there is always the possibility of finding out more

about a given situation, no action would be morally secure. The

great advantage of an ethics with “must” is this moral security.

“Ought”-sentences as they really are offer no such security.

a5) “prima facie obligation”

Continuing in the same tradition as Kant, but being

sensitive to the fact that true “ought”-principles can be in

conflict, W. D. Ross (1930) constructed the idea of “prima facie

obligation.” The idea is that although “ought” is the same

concept as “must” or “is obligated,” “ought”-principles and

“obligation”-principles are often “prima facie” principles,

principles that apply “other things being equal.”

This idea, while it allows for an ethical theory more in

accord with common sense, gets the linguistic facts completely

wrong, and so blocks any progress toward having a correct ethical

theory.

Let me illustrate what is wrong with the notion of “prima

facie obligation” by an analogy. One could hold that all 235

? “Be Ye therefore Perfect, even as your Father which is in heaven is

Perfect,” (Matthew 5:43-48)236 “For this Torah I enjoin on you today is not beyond your strength orbeyond your reach. It is not in heaven.” (Deuteronomy 30:11-12)



connections among events are necessary connections. The

principles which express those connections, however, are

sometimes prima facie principles. Suppose we take the following

sentences to be literally true:

1) “If Fred is an NRA member [N], he has to be a

conservative[C].”

2) “If Fred is for legalized recreational drugs[R], he can’t be a

conservative[C].”

Suppose, as is possible, that Fred is both an NRA member and for

legalizing drugs. Then these modal conditionals are true even

though the de-modalized conditionals are not. That is, taking the

truth-functional conditional to be the minimal truth-condition

for any conditional, and assuming that if something is necessary

then it’s true, we get [(N/\R) /\ (-N v C) /\ (-R v -C)], which

is contradictory.

A theory akin to the theory of prima facie obligation that

allows 1) and 2) to be simultaneously true would be the

following: “Necessary” and its synonyms “must be” “gotta be” and

the like, have a “prima facie” sense as well as an “absolute”

sense. A conditional can be prima facie necessary without its

truth-functional correspondent being true. Some uses or senses of

“necessity” express “prima facie” necessities.

Calling the necessity “prima facie” means that there is some

connection between the states of affairs mentioned in the

antecedent and the consequent, and that other things being equal,



that connection is manifest and operative. Thus, prima facie

necessity is equivalent to the “real connection” sense of

“probably.” “If Fred is an NRA member, he is probably a

conservative” is an alternative to the “prima facie” use of

necessity. So “probably,” another “connection” term, is

assimilated to necessity. Probability or prima facie necessity is

just a lower degree of absolute necessity.

Such a theory would be completely misguided. The obviously

right thing to say is that sentences 1) and 2) are hyperbolic.

Yet a precisely analogous theory seized the minds of many ethical

thinkers. The logic of “ought” was treated as the logic of

obligation, which indeed behaves like traditional necessity. The

unhelpful fudge of “prima facie obligation” was supposed to fit

such a theory to some intuitive facts.

Fundamentally different things are assimilated by treating

probability as a weakened form of necessity. “Probably” is

primarily epistemic; whereas “necessarily” is ontological.

“Probably” is supported by reasons to believe that something is

the case, and governed by norms of what it is rational to

believe, given other beliefs. “Necessity,” on the other hand,

necessitates. Reasons making something necessary are not norms

for rational beliefs, but laws of nature, mathematics, or (if I

am wrong) metaphysics. So reasons making something necessary



cannot be counter-balanced. 237 Roughly, “probably” is the

modality for inductive inference, “necessary” the modality for

deductive inference. A strong inductive argument can be weakened

by the addition of further information. You can have very

compelling evidence that Fred is the murderer, only to have it

overwhelmed by a newly-discovered videotape. On the other hand,

if a deductive argument is valid, then any additional information

cannot affect its validity.

b) The theory of “ought”

b1) “Ought” and conditional probability

Just as is the case with “good” and every other predicate, a

Davidsonian theory of the truth-conditions of “ought” should

treat homonymy as a last resort. The default position is that

“ought” means “ought.” The basic semantics for “ought”-sentences

should be the same for “If you turn the key, the car ought to

start,” “If you want a nice meal, you ought to go to Cavey’s” and

“If you are an agent, you ought to treat people with respect.”

The logical form of “ought”-sentences tells us something

about them. As discussed in Chapter 5, “ought”-sentences are two-

237 This feature of necessity corresponds to the fact about “must” thatreasonably unmet obligations remain. Even though you should ignore some obligations sometimes, you still have to compensate those you letdown. If you’re deciding who to ask to the prom, and end up asking Joan, you don’t have a debt to Jane, even though there were reasons for asking her as well. But if you invite Joan, and then stand her up because your mother takes ill, you have to apologize, even though you did what you ought. The relation between what we ought to do and obligations is discussed below.



place predicates of things said. So, “ought” is some kind of

relation between things said, propositions. The following is a

theory about what it takes for the “ought”-predicate to apply to a

pair of things said, not an account of logical form.

The first thing to notice about conditional “ought”-

sentences is that there are many which say almost the same thing

as the corresponding conditional probability sentence. “If Fred

hit that log in the right spot, it ought to have split with one

blow” and “If Fred hit that log in the right spot, it probably

split with one blow” seem at least very close. They both have the

feature that antecedent-strengthening does not in general

preserve truth. “If Fred hit that log in the right spot and it

had a knot running through it, it ought to have split with one

blow” does not follow and does not follow with “probably”

substituted for “ought.” Thus, also, neither conditional

detaches. From the truth of the “Fred hit that log in the right

spot” “It probably split/ought to have split with one blow” does

not follow.

An hypothesis worth pursuing is that “ought” is a

compounding of probability. “Ought” has the sense of “ideally

will” that “probably” doesn’t. Consider the following story: For

those of us who graduated from middle-school in the 1950s, for

any pair of three-digit numbers, if we were asked to find their

product, we would probably be successful, remembering to move the

interim result to the left for the tens and the hundreds places,



and so on. The same thing holds with long division, subtraction,

and so on. When we did our income taxes by hand in ancient days,

there were many such calculations to do to find out how much we

owed the IRS. The probability that we would get the right result

in the end was very low, due to the multiplicative laws about

probabilities. So, routinely, we would have to double- and triple

check our arithmetic. If there was a 90% chance of our getting

any particular calculation correct on one try, there was only

about a 35% chance of getting ten right answers in a row. If the

IRS required ten calculations, then, we would probably not get

the result we ought to have gotten.

A natural idea, developed in Wheeler (1974), is this: The

result we ought to get is the result we would get if, for each calculation, we got what we

would probably get. The idea is that “ought” is a kind of chain of

conditional probabilities. The sense of ideality present in

“ought” sentences but not in probability sentences is captured by

this idea of a chain. In very short chains of probabilities, both

the “ought”-sentence and the “probability” sentence are likely to

have the same truth-value. In very long chains, the agreement in

truth-values will be low, unless the probabilities are very high.

I would probably not get the result I ought to get. The intuitive

idea of this chain of conditional probabilities is rather hard to

specify in detail.238

238 See Wheeler (1974). There, the chain of conditional probabilities isconstructed so that the antecedents become more and more complex conjunctions. Roughly, “If you have multiplication problem A, you will



b2) “Ought”-sentences about rational agents

A chain of conditional probability sentences connects with

human beings and what they ought to believe, desire, and do

because the principle of rational agent interpretation are all

maximization-principles. Other things being equal, if I believe

p, the person I’m interpreting believes p. Other things being

equal, if I desire p, the person I’m interpreting also desires p.

A Davidsonian ethics starts from the claim that mostly correct

desires and mostly true beliefs are constitutive of being an

agent.

Thus, a Davidsonian theory of “ought” for rational agents is

akin in some ways to Kant’s ethics. Kant attempts to derive the

probably calculate B. If you have calculated A and B, you will probably calculate C. If you have calculated A, B, and C, you will probably calculate D.” And so on. The chain idea is that, at each step, if a calculator has gotten this far, it is probable that the calculator will successfully get to the next result. Of course, after a while the cumulative conditional probability of the speaker having gotten to this point, that is, having correctly calculated the first, say twenty long division and multiplication problems, is very low. Butgiven that the calculator passed the eighth grade, for each calculation, the probability is high.

When you try to see how these chains can go, you have to block various side-tracks by various ad hoc devices. You also have to block infinite sequences at various points. The project of coming up with a precise definition of “ought” in terms of “probably” is something likethe project of trying to define “x knows that p.” Whatever conditions one puts in, there is probably a clever counter-example. I take this lack of a sharp definition in other terms to be a characteristic of all but a very few notions that we perfectly happily use in negotiating our way through the world.



normativity of morality from the normativity of action. Kant239

argues that the very concept of acting on purpose requires that a

perfectly rational agent do only what has a coherent general

principle. Thus acting for reasons consistently requires acting

in a way that everyone could act. Kant argues that meeting this

minimal necessary consistency condition for rationality is also

sufficient for doing the right thing. Kant, of course, construes

“coherent general principle” as an exceptionless universally

quantified principle.The other Kantian route from rationality to morality argues that it is

irrational to take one’s own reasons as an objective motive for action but notto take others’ reasons as similarly objective. A Davidsonian ethics, as developed below, will adapt an analogous route in order to connect the prudential with the moral “ought.”

One core idea in Donald Davidson’s philosophy suggests a way to revive the part of Kant’s project that derives ethics from agency and rationality. Interpretation operates under probabilistic constraints constituting the concept of agency. In order to be understood as an agent at all, a speaker or actor must hold mostly true beliefs, must by and large value things that it isreasonable to value, and must by and large make reasonable inferences. That is, agency itself is constituted by probabilistic concepts. To interpret another entity as an agent is to treat that entity as governed by probabilistic notions. “The principles of rationality” are just these probabilistic constraints that we impose on interpretation.

Davidson is profoundly anti-Humean in his conception of desires and their rationality. For Davidson, having mostly correct desires, that is, desiring the good, is constitutive of being a rational agent. The Humean alternative, that, as far as rationality goes, anything can be an object of desire, would make interpretation of actions impossible. Any action could be taken to express any set of beliefs whatsoever if the intentional system imposed no constraints on the contents of desires.

A version of the private language argument will illustrate this point. If rationality consisted only in believing the true and revising beliefs correctly when new evidence arises, then there could be no evidence that

239 This summary of Kant focuses exclusively on arguments in Kant

(1785).



someone believed a falsehood. The sole maximization constraint would be “interpret the Other in a way that maximizes agreement.” Given appropriate assignment of desires, whatever the person said or did would reflect only truebeliefs. If the person says something that one would take to be mistaken, suchas “Venus is always at a uniform distance from the Earth” an adjustment in thespeakers desires, for instance to test us or mislead us, allows the interpreter to ascribe the correct beliefs about the distance from the Earth to Venus. What appear to be disastrous choices, for instance betting all hers assets on red when in fact black is what the roulette wheel ends up at, is compatible with the actor’s true belief that black will win by supposing, as would be required, that the actor wants to lose all her assets. Her dismayed sobbing, while perhaps not an intentional action, will have nothing to do withthe interpretation of her beliefs. On the Humean conception, no action would give any indication whatsoever that the person thinks any future event will happen. Absent constraints on the assignment of desires, no words or actions can indicate false beliefs. The person can be assigned true beliefs about any outcome no matter what happens. The Wittgensteinian point is that if no alternative behavior would manifest a false belief, no behavior can manifest a true belief. Given that interpretation is possible, that we can assign beliefsand desires on the basis of what is said and done, constraints on the contentsof desires are required. As I argue below desires are true and false, depending of whether their object is a good, given the circumstances.240

Actions, including speech-actions, are explained by belief and desire, the core concepts of ‘the intentional scheme.” In general, any behavior can beinterpreted as an action expressing a given set of beliefs, if there are no constraints on desires; and any behavior can be interpreted as expressing any desire, if there are no constraints on belief. Both kinds of constraints are essential to applying the intentional scheme.

Davidson, viewing interpretation holistically, holds that by and large we and the other must share beliefs and desires. While there are desires it isrational to have, namely the ones agreement with which we maximize in interpretation, there is no particular desire that a rational agent must have. The advantage of Davidson’s approach is that it allows us to treat those with some irrational desires as still agents. Agents with some irrational desires must still be subject to rational understanding. It will be possible to assignirrational desires to an agent given a background of rational desires. Davidson’s holistic conception of “constitutive” accommodates the fact that we

240 A Davidsonian should happily agree with McDaniels and Bradley’s (2008) that desires are themselves two-place relations between propositions. Rationality of desires will be partly a matter of appropriateness. A desire to take an umbrella if rain is likely makes sense, whereas a desire to take an umbrella if one is wearing black socks may not.



must use the intentional system to understand behavior as action at all. A Davidsonian account of rationalization of misguided actions based on irrational desires would be akin to Davidson’s account of weakness of the will(1970b).

Normativity is derived from conditional probabilities relative to agencythat come from principles of interpretation. Put another way, interpretation itself is governed by norms of rational agency. The principles of agent interpretation, after all, can easily be construed as “ought” sentences –what beliefs and desires we ought to ascribe to a person, and that the person oughtto have. My Davidsonian strategy is to show how ethical concepts derive from this normativity implicit in interpretation of an agent.

One of the features of “ought”-sentences, as is the case with other modals, is that they are almost always understood relative to contextual parameters. Different “backgrounds” yield different truth-values. To illustrate this, consider the following case: Suppose a thief is fleeing the cops. If he turns to the left, he enters a warren of alleys where he is sure to escape. If he turns to the right, he will run right into his pursuers, be captured, serve his time, be rehabilitated, and become a better and happier person. Which way ought he to turn? It would seem that there are reasons to give both answers. Relative to his actual goals, he ought to turn left. Relative to his real interests, though, being removed from a life of crime will be better for him. A third understanding of “ought” would note that, in stressful pursuit situations, he tends to go left, so that would be the reasonable prediction about what he ought to do in this situation. So, what ought he to do?

There are two kinds of relativity in conditional “ought” sentences: The first, which we can call “kinds of considerations,” is sometimes expressed by an adverb-phrase, such as “morally” or “prudentially” and is sometimes contextual, such as “relative to the aims of chess.” Illustrations of this kind of relativity are especially clear with narrow sets of considerations. For instance, there is a general principle of the “logical” “ought” that one ought to believe the logical consequences of what one believes. There are obvious counter-examples. If Fred believes that Susan is honest and that Susanhas been embezzling his funds, he should not conclude the conjunction and conclude that honest women sometimes embezzle, but rather ought to give up oneof his premises. This kind of relativization, relativity to considerations, would allow that the principle is true relative to logical considerations only, but may be false relative to broader considerations. Another example is a player in a chess game who can force mate in three with a bishop sacrifice at f7, while every other move loses. Relative to chess considerations, the player ought to move bishop f7 check. But her opponent is the dictator, who takes defeat very badly. So, prudentially, bearing in mind her own personal welfare, she should not go bishop to f7check.” Of course, relative to considerations of honor and standing up for her people against the dictator,



perhaps she should bring about mate in three and bravely face the consequences.

This relativity to considerations distinguishes what people have called the various “senses” of “ought.” An utterer in a context can intend one or another set of considerations, and different sets of considerations would yield different truth-values for the “ought”-sentence. “Ought” is relative to conversational context in much the way other modalities are.

In this kind of relativity, the apparent “absolute” “ought” is “all of these things considered.” It is still a conditional “ought” with the effective “antecedent” restricted to the special subset of the considerations that wouldbe relevant to a genuinely general “all things considered.”

The other kind of relativity is to the content of the “antecedent” of a conditional “ought.” So, “If you turn the key sharply, it ought to start” and “If Fred wants to please Susan, he should ask about her daughter” both specifybeginning points of chains of probability-sentences. The truth-conditions of such claims is of course determined by the information available in the situation. So, “If Fred wants to please Susan, he should ask about her daughter” recommends one way of pleasing Susan. Part of the background information that may make it true is that Fred, while he could give Susan his Lamborghini, and this would also please Susan, perhaps even more than asking about her daughter, would find the rewards not worth the cost. So, generally, when a conditional “ought” sentence has an antecedent, it is still a question whether the considerations that should be in the “antecedent” are a restrictedsubset or absolutely everything.

If “ought”-sentences are essentially like “if…then…probably” sentences, then their “categorical” form is implicitly relativized. That is, since “If…then…ought” sentences are not some conditional applied to a categorical “ought”-sentence, then, since a predicate has a given number of places, and the categorical and conditional “ought”s are not homonyms, the categorical “ought” must be actually the two-place conditional “ought.” So, “It will probably rain,” and “We ought not to lie” are both represented more accuratelyas having a implicit “all things considered” or “given all available information” “antecedent.”

The basic idea of the theory analyzing “ought” as a chain of conditionalprobabilities is that complex practical inferences can be broken down into a series of simple steps each of which an agent would probably get right. In an “ought”-sentence with a desire in the antecedent, the chain will typically be a series of probabilistic connections between truths about how the object of desire241 can be achieved and beliefs in those truths. The connection between truth and belief is probabilistic, as is the connection between desire for A, the belief that if C then B, the belief that if B then A, and a desire to 241 Since “desire” sentences are themselves conditionals, a thorough account of the form of “ought” sentences with a desire in the antecedent would be more complicated.



bring about C. Aristotle’s (1980, 1981a) picture, where a syllogism used practically transmits desire from conclusion to premises and used theoretically transmits belief from premises to conclusion, is basically right. The difference is that Davidson treats ethical thinking on the model ofmodes of non-deductive inference,242 so the inferences are not usually formallyvalid, but at best inductively strong. So there is no algorithm for the “calculation.” Each of the smaller steps is one that a person would probably make, just in virtue of being a rational agent.

Of course there are other conditional probability chains that derive from putting still other information about the agent besides desires into the first argument of the “ought” sentence. Other chains yield what a person oughtto do given evil wants, idiosyncratic wants, peculiar dispositions to react tokinds of situations, or flaws in competence. Knowing that Fred gets very nervous about exams, we can say, minutes before the exam, “Fred ought to be nearly catatonic about now,” even though that is not in Fred’s interests. In brief, there are a large variety of “ought”-sentences that will be true of a person in a situation.

On this account, the interpretation of an action will, other things being equal, maximize the rationality of the person being interpreted. Sometimes, contrary to what Davidson argued against Hempel, a person’s rationality can be part of the explanation of an action. While a person must be a rational agent of some degree of rationality in order to act or be interpreted at all, clearly some people are more skilled at various aspects ofrationality—people who are especially susceptible to being moved by arguments,or especially able to work through complex sets of considerations. That everyone is rational in the sense that they qualify as an agent does not mean that exceptional rationality cannot be appealed to as part of the explanation of an action.

In the same way, if a person acts reasonably in a situation where most would be moved by non-rational considerations, the person’s rationality may bea partial explanation of the action—why did she not panic? Michael Smith’s (2009) critique of Davidson’s remarks about the place of rationality in explaining action can be accepted by Davidsonians, since they leave the basic Davidsonian conception, that rationality is constitutive of agency, intact. The theory I am presenting of what a person ought to do in fact takes the rationality of an agent who does what she ought to be far beyond the rationality that would be required for a person to be an agent at all.

242 “Inductive logic” is the measure of the degree of support that a setof sentences A gives to a conclusion B. Hempel (1965, pages 5-6) showed that there cannot be an algorithm for inductive logic. Hempel’sdemonstration uses the fact that theory-formation can require new concepts.



c) Why a good hammer is one you ought to want if you want a hammerAt the end of Chapter 9, I said that a justification in terms of the

theory of “ought” would be given for the account of the truth-conditions of “good F” that “A is a good F” amounts to “If you want an F, you ought to want this.” Interpretation maximizes agreement in beliefs and desires. The evidencefor an interpretation of a predicate H as “framing hammer” will then be that the predicate has a similar place in the other’s network of beliefs and desires that “hammer” does in ours. So, given that “H” is correctly interpreted as “framing hammer,” for any feature desired by people in our culture in a framing hammer will be a feature probably desired by people in the foreign culture for objects of which “H” is true. These probabilities then give the basis for an “ought.” At least as important as physical shape will be typical uses, that is typical sekings of thing of which the term is true in order to satisfy various desires. So, for instance, the foreign term for hammer-shaped objects that are primarily carried in spring fertility ceremonies is probably not best interpreted as “hammer.”

On this account, the person understanding “good framing hammer” need have no idea what physical properties are desirable in a framing hammer in order to understand the expression. The resulting theory of the truth-conditions of “good” is then not very new or surprising—it is just a version of the “functional role” account of “good F” popular since Aristotle. The virtue of the present theory is that this truism about “good” falls out of thetheory of interpretation and the theory of “ought.”d) The Ethical “Ought”

What about the “ought”s that are at the center of philosophical interest, the “ought”s about what person ethically should do, just in virtue of being a person in a given situation? The present theory would treat this “ought” as an “ought” that abstracts from all idiosyncratic desires. As we will note below, it also abstracts from all partiality. That is, just relative to being a rational agent in the given situation, what “ought” sentences wouldbe true? An agent who always believed what it is rational to believe, from ourpoint of view, and who always desired the good, from our point of view, would be an agent who did what she ought, from our point of view. The important feature of the chain of conditional probabilities for the agent who does what she objectively ought is that they are all relative just to being an agent. Anyrational agent, insofar as she is a rational agent, ought to do the same thing, in her situation.

The ethical “ought” abstracts from more complex first arguments, such asbeing an agent who wants a Big Mac, or who reasons defectively in certain circumstances, or who is Italian or who wants to go to medical school. Relative to these features and desires, different actions would make sense. Relative just to being an agent at all, though, the chain is constructed from desires and beliefs everyone probably has, just in virtue of being an agent. So, ethical reasoning arrives at “ought”-sentences by constructing what a



person would do if that person always believed, desired and did what, on principles of reasonable interpretation, any person would probably believe, desire and do at every step of the decision-process. The “ought” thus constructed is an impartial, all-things-considered “ought.” It might, in retrospect, differ from what the optimal thing to do would have been, since itis based on what it is rational to do given the information available and worth getting.243

A kind of universalist ethical “ought” is thus constructible from the Davidsonian account of “ought.” A person who counts everyone’s interests as equally weighty, that is, is partial to no one, ought to act in a way that at least comes close to ideal ethical action. There are two questions to raise about this conception of what a person really ought to do: First, what is the relation between this conception and the dictates of “morality” as a system ofprinciples of obligation? We argued above that what we ought to do will not ingeneral correspond to outputs of systems of obligations. However, we would expect or at least hope that much of the time a person ethically ought to do what she is obligated to do. The next section will discuss that relationship, according to the present theory. The second question is whether a person ought, all things considered, to do what she ethically ought to do, as constructed above. That is, while we can construct an “ought” that abstracts from natural partialities to oneself, one’s family, tribe, fellow citizen, or whatever, it is not obvious that a person ever or usually ought to so abstract. Whether this is the case will, as I will argue, depend on three things: First can the rationality of taking other’s interests seriously be justified? Second, can desires be more or less accurate? Third, how much partiality, if any, is rational? Differential care about one’s own children versus someone else’s, seems to everyone, except perhaps Kant, to be somethingethically justified. The third problem seems to me the most intractable. I have no idea.e) The Ethical “Ought,” Moral Obligation, and what we ought to desire

From a Davidsonian perspective, many systems of rights and obligations have a great deal of bearing on what one ought to do. Other things being equal, one should do what one is genuinely obligated to do. On the other hand,there are systems of obligations and duties that have little relationship to what one ought to do ethically, such as the system of duties and obligations that obtain between slaves and their masters. These duties and obligations areones that an individual often has prudential reasons to obey, but they are notones that have moral standing. That is, if it were not for the special power relations obtaining when one is a slave, one should not subordinate one’s willto this other person.

243 For different kinds of decisions, there are different criteria for things you should find out before deciding, as well as different things you should have known.



The principles of moral obligations and rights that Kantians such as Scanlon, Korsgaard and others have constructed are, from a Davidsonian perspective, principles that would be good guidelines to what one ought to do.The arguments that one’s duties to others have a rational basis are very persuasive. Generally speaking, that is, acting in accordance with rules of obligation and rights would correspond to doing what one ought to do, all things considered, if it could be shown that these promises and commitments were ones that in fact should be respected.

The possibility of an argument for the claim that the principles of moral systems of obligation and the impartial, ethical “ought” constructed above actually have a bearing on what a person should do, all things considered, seems to me to rest on two arguments:

The first argument would show that it is irrational to take one’s desires as reasons for action and not take other peoples’ desires as equally reasons. Kant’s (1785) project was at least in part this project. Somewhat more recently, Thomas Nagel’s (1970) tried to show this by showing that being motivated by interests of one’s future self requires being motivated by interests as such.

Second, since the results of the first argument would at best produce lines of reasoning a person could respond to by changing beliefs and desires, a further argument is needed to show that the reasoned desires required in order to be even somewhat impartial in taking other people seriously are ones a person ought to have. Briefly, a Davidsonian needs to find an account of the“desire” component of reasoning to action that meets Scanlon’s (1998) conception of desires as reason-sensitive attitudes.

McDaniel and Bradley (2008) have shown that the very semantics of desiremakes desire a “reason-sensitive” attitude. Apparent “categorical” desires are, by their very form, relative to considerations, in this case, all considerations. I have always thought that Kant’s and Nagel’s arguments did infact yield the desired conclusion, that it is irrational not to take other peoples’ interests into account. It seems to me clear also, that Davidson’s conception of desire fits perfectly with Scanlon’s (1998, chapter 1) conception of desires as attitudes for which reasons can be given.

There are two ways in which reasons can be given for desires. First, in conditional desires, the connection between the condition and the desire is subject to rational evaluation. If Fred wants to bring an umbrella if rain is predicted, that makes sense. The opposing view, of course, is that desires, unlike beliefs, do not correspond to anything objective. People differ in preferences about which there is nothing like “true” or “false.” There just isno contrast between desire and anything else that is analogous to the contrastbetween “I believe that A” and “A is true.” Unlike differences in belief, differences in desires are said to reflect nothing but internal differences inpersons. The claim is that there is nothing like triangulation for desires andno common value-world to which both an interpreter and the interpretee are



related when an individual expresses a preference. So the interpretation of actions lacks one of the groundings in a common world that interpretation of belief has. This claim is just mistaken, since desires are constrained in interpretation just as beliefs are.

A Davidsonian treats ethics as objective, and holds that ethical sentences usually have truth-values. Differences in preferences are differences in beliefs about what is good. In-principle irresolvability are beascribed to indeterminability of interpretation, according to a Davidsonian, and is no different from indeterminability about which beliefs to ascribe to another.

Truth-values of ascriptions of “good,” that is, desires or pro-attitudesin general, are no less objective than truth-values of sentences about houses and stones. A rational person is aware that her judgments may be idiosyncraticand should be checked for correctness by seeing what others believe. Just as we do not in general identify our beliefs with the truth, so we should not identify our desires with the good “for us.” There is no more a coherent notion of “good for me” as an explication of “good” there is a coherent notionof “true for me” as an explication of “true.” Given our past experience, we know that some of our beliefs are likely to be mistaken and that some of our desires are ill-considered. We realize that we are less than perfectly reasonable, even though we hold, of each particular view, calculation, and valuation, that it is reasonable. If interpretation maximizes desire for the good, and sentences using “good” have truth-values,

The decision-theoretic tradition has taken preferences to be just brute facts about which nothing is rationally required but coherence, whereas beliefs are true or false, and so objective. But in interpretation, we in facttake some preferences to be irrational, such as the simple preference for painover lack of pain. Some preferences can only be assigned to an agent on the basis of very strong evidence. “Pain is worse than no pain, other things beingequal,” is true. So, prima facie, some “better than” sentences are objectivelytrue.

The conception of preferences suggested by Davidson’s remarks that we maximize agreement about the Good as well as about the True is that a preference for A rather than B is, or always corresponds to, a belief that A is (actually, objectively) better than B. If utterances about the Good have truth-values, as Davidson argues in the essays in his (2004), then preferencesare opinions about those truth-values.

To every preference there corresponds a belief that A is better than B. Pleasures and pains, as well as other “perceptions” of good and bad derived from a person’s particular history, are then akin to the causes that give riseto perception-reports. If that is true, “good for me” is akin to “true for me,” that is, another way of talking about beliefs. On this conception, physical pain and pleasure would, in motivating action, be causes of judgmentsthat “This is bad” and “This is good.” Such judgments are incorporated into



the web of belief and desire. Sometimes the desires and aversions associated with pains and pleasures are over-ridden by other considerations, just as illusory perceptions are. So, we desire to get our separated shoulder replacedin its socket by a rather painful procedure.244

No sensible person should rely just on his own “sensory” judgments aboutthe Good and the Bad, just as no sensible person would investigate scientific questions by using only his own perceptions and measurements. It would be unreasonable to reach opinions about what is better than what using only one’sown experientially-delivered preferences. Just as one learns the difference between one’s beliefs and the truth through social interaction in triangulation, so one learns the difference between one’s preferences and the truths about what is better than what through the social triangulation of “moral education,” which starts with Mommy saying “No, no,” and continues through the rest of the socialization of a person.

While some of this triangulation may be mistaken, just as some instruction about what cows are (“they moo”) is, most of it must be right. To the extent that there is indeterminability in the interpretation of other persons, we should expect an indeterminability in ethics that is compatible with objectivity.

The “ethical’ point of view is one that takes individual, personal

preferences as not decisive, but as data. If we could establish that

everyone’s desires reasonably ought to count and are comparable,245 then we

would have an ethical theory that treated ethics as objective, even though

there were unresolvable ethical questions. The unresolvability would be a

consequence of the indeterminability of interpretation, not the result of

subjectivity. The next section will argue that we have every reason to count

the desires of others in deciding what we ought to do.

d) “Better for me than” and “Better than”

244 As in many other cases, this Davidsonian view is essentially Aristotle’s. Pleasure is the “appearance” of the good—generally reliable, but not always. The intellect can overcome sense experience.245 The problem will be the interpersonal comparison of utilities. Thereare some clear cases. A child’s desire not to drown outweighs my desire to continue fishing, for instance. Given that there are some clear cases, the general thesis that desires can be weighed across persons is plausible, even if there is no routine for determining which of two desires counts more objectively. See Davidson’s (2004a).



What is the relation between what is objectively better and the

preferences of particular agents? I argue that the relation is analogous to

the relation between the truths and the beliefs of agents. While the beliefs

have to be largely true, very many of them are localized, so truth is not

reducible to consensus or to any simple collection of local beliefs. That is,

indexical beliefs about what is true here, now, and for me are largely true,

but the relation between those truths and non-indexical ones may not be

summation.

A Davidsonian can say the same thing about preferences. They are locally

true, by and large, but the relation between local truths and non-local ones

is not simple. My preferences, i.e. my beliefs that A is better than B, often

overemphasize immediate indexical judgments, akin to perception reports. How I

immediately feel about things figure very large in my judgments of value. That

is, indexical value judgments shape preferences to a very large degree. I

prefer what is good for me to what is not good for me.

What is the relation between A is better than B” and “A is better for me

than B”? My Davidsonian “rational” foundation for ethical reasoning, as

opposed to prudential reasoning, argues that “better than” is to “better for

me than” as “true” is to “true for me,” where “true for me” means something

like “true given my evidence” or “true in my location.”

How can indexical preferences distort value judgments, if those indexical judgments are true? Preferences are beliefs about what is good and not good that are largely true. But it can be good that A without it being good that A and B. When that A is good is an indexical belief, that is, from aparticular agent’s perspective, that strongly inclines the agent to believe that A and B is good. That people tend to favor themselves amounts to people giving unreasonable weight to their own judgments of good and bad. That is, value judgments tend to value complex states of affairs by giving undue weightto “indexical” value judgments.

Some would argue that this is not irrational, but that preferences are essentially indexical and so do not express objective values. If there is a difference between “That is good” and “I prefer that,” though, so that preferences really are judgments about value, then the predilection for moving



from “I prefer A” to “A is good” is a natural shortcoming of people, just as our predilection to interpret “I’m cold” as “It’s cold” is.

On reflection, it is quite obvious that a normal agent’s concerns go farbeyond anything identifiable as particular states of the particular organism that is the agent. Usually the identification of one’s interests with the interests of others is limited to a subset of the others, and so is “partiality,” a discrimination. Ascribing such partialities is part of interpreting action and speech, and so part of the rationality that is part ofinterpretation. “Reasonable” attachments beyond “local” judgments are maximized in interpretation. But, if interests it is rational to ascribe extend beyond indexical value judgments to interests in things beyond the self, then reasonable valuation is not based just on what things are better for me.

Such interests are a deconstruction of the primacy and clarity of the subjective preferences that have bedeviled ethical theories since the beginning of philosophy. People routinely care about a lot more than their ownpleasures and pains, their indexical deliverances about what is better than what. We also rationally evaluate degrees of attachment. It may be irrational to sacrifice one’s coat to save a woman’s feet from the mud. Partialities can be evaluated as rational or irrational, and we do so routinely in interpretation.246 So interests in the welfare of others can be rational or irrational.

The maximal scope of concern is concern for the interests of every entity that has interests. If Nagel(1970) and Kant(1785) are right, this maximal scope of concern is rational. If we are motivated by partialities thatit is reasonable to have, then this one, “The Moral Point of View,” (Kurt Baier, 1965) is rational, and reasonably ought to motivate any agent with interests. That is, if Kant and Nagel are right that it is rational to be concerned with the good for every entity that has interests, then it is reasonable to be concerned with the interests of everyone, to have a conception of the Good that abstracts from one’s local valuations.247 246 A partiality can be excessive, as some allegiances to states and sports teams often are. The basis for such judgments is that there is no reason for the partiality. Partiality can be inadequate. A lack of favoritism towards one’s children, for example, calls for explanation by pointing out that the mother is den mother of the group of Cub Scouts of whom her child is a member. 247 A very large issue, to which I do not have a satisfactory answer, ishow to join the rationality of taking account of everyone with the rationality of taking special account of one’s spouse, parents, friends, children, and lovers. Taking special account of one’s relationships requires essentially indexical attitudes. But some partialities seem rational and reasonable. How does that fit with



Thus we can conceive of preferences and therefore interests as perspectives on what is better than what, strongly influenced by indexical desires. The perspective of the particular interests of the agent, what is better “for her,” is reasonably conceived as a conception of “what is better.”So we can coherently take preferences to be beliefs about what is objectively better, and reasonably take “narrow” views on what is better, that is, selfishness and unreasonable partiality, to be due to failures of reason or mistaken beliefs. f) Conclusion

Given that “rationality” is in effect, a projection of what some interpreter takes to be rational, and given the substantial variation in people’s assessments of rationality besed on their own idiolects, we can expect ethics to have a core of agreement, but also a wide range of disagreement among agents whom we have to acknowledge are agents.

The ethics that would follow Davidsonian observations would be objective, in the sense that every ethical question has an objective answer, but this objective answer would in many cases be indeterminable. It would alsobe an ethics such that the most saintly person possible could do the wrong thing. No finite being can know everything there is to know about the current situation. It would also be an ethics which would agree with Aristotle (1980 1094b11-28) that we should not expect precision and answers to every question in such subject-matters.

This chapter has not claimed to dissolve philosophical questions in ethics. There are many genuine issues. Davidson’s insights into semantics and the theory of interpretation provide a better foundation for thinking about ethical questions. Ethical questions can be addressed with the hope of making progress once ethicists grasp that “good” could not be the name of a property,that “ought” and “obligation” have very distinct logics, and that the questionwhether desires are constrained by rationality is “yes.” The first two questions are answered by an adequate semantics. The third question is answered by an adequate understanding of interpretation. So, we can proceed.

Works Referred to:

Aristotle. (1966). Metaphysics. Translated by Apostle, H.G. IndianaUniversity Press, Bloomington, IN.Aristotle. (1981a). On The Soul. Translated by Apostle, H.G. Peripatetic Press, Grinnell, Iowa.Aristotle. (1963) De Interpretatione. In Categories and De Interpretatione, edited and translated by J. L.Ackrill. Oxford: Clarendon Press.

there being an objective good that is entirely non-indexical?



Aristotle. (1981b). Posterior Analytics. Translated by Apostle, H.G. Peripatetic Press, Grinnell, Iowa.Aristotle. (1980). Nichomachean Ethics. Translated by Apostle, H. G.D. Reidel.Armstrong, David. (1997). A World of States of Affairs. Cambridge University Press.Baier, Kurt (1965). The Moral Point of View. Random House.Bale, Alan. (2008). A Universal Scale of Comparison. Linguistics and Philosophy 31, 1-55.Barwise and Cooper. (1981).Generalized Quantifiers and Natural Language. Linguistics and Philosophy 4. 159-219.Bennett, Jonathan. (1988). Events and their Names. Indianapolis: Hackett.Bennett, Jonathan. (2003). A Philosophical Guide to Conditionals. Oxford University Press.Bhowmick, Nilanjan. (2012). That F is G: Defending Quantification. PhD dissertation. University of Connecticut.Blair, Daniel. (2003). Parataxis and Logical Form. PhD dissertation. University of Connecticut.Braun, David and Sider, Theodore. (2007). Vague, so Untrue. Noûs

41 (2):133 - 156.

Brower, Jeffrey. (2010). Aristotelian Endurantism: A New Solutionto the Problem of Temporary Intrinsics. Mind, Vol. 119 . 476. 883-905Burge, Tyler.(1973). Reference and Proper Names. Journal of PhilosophyLXX Number 14.425-439.Burgess, John P. (2009). Philosophical Logic. Princeton University Press.Cappelen, Herman and Lepore, Ernie. (2005). Insensitive Semantics. Oxford: Blackwell Publishing.Carnap, Rudolph. (1937). The Logical Syntax of Language. Routledge and Kegan Paul.Carnap, Rudolph. (1947). Meaning and Necessity. University of Chicago Press.Church, Alonzo. (1950). On Carnap's analysis of statements of assertion and belief. Analysis 10, 97-99.



Davidson, Donald. (1966). Emeroses by Other Names. The Journal of PhilosophyVol. 63, No. 24. 778-780, and Davidson (1980) 225-227.Davidson, Donald (1967a) Truth and Meaning. In Davidson (1984). 17-36.Davidson, Donald. (1967b). The Logical Form of Action Sentences. In Davidson (1980). 105-122.Davidson, Donald. (1968). On Saying That. Synthese 19, 130–46.Davidson, Donald. (1970a). Mental Events. In Davidson (1980). 207-227.Davidson, Donald. (1970b). How is Weakness of the Will Possible. In Davidson (1980). 21-42.Davidson, Donald. (1973). Radical Interpretation. In Davidson (1984a). 125-140.Davidson, Donald,. (1974). On the Very Idea of a Conceptual Scheme. In Davidson (1984a). 183-198.Davidson, Donald. (1977). Reality without Reference. In Davidson 1984a. 215-226.Davidson, Donald. (1978). Intending. In Davidson (1980). 83-102.Davidson, Donald. (1979a). Moods and Performances. In Davidson (1984a). 109-122.Davidson, Donald. (1979b). The Inscrutability of Reference. In Davidson (1984a). 227-242.Davidson, Donald. (1980). Essays on Actions and Events. OxfordUniversity Press.

Davidson, Donald. (1984a). Inquiries into Truth and Interpretation. OxfordUniversity Press.

Davidson, Donald.(1984b). Expressing Evaluations. In Davidson,Donald, (2004b). 19-38.

Davidson, Donald. (1986a). “A Nice Derangement of Epitaphs” in Truth and Interpretation, ed. Ernie LePore. Blackwell. 433-446.Davidson, Donald. (1986b). The Interpersonal Comparison of Values. In Davidson (2004b). 59-74.



Davidson, Donald. (1992). “The Second Person.” Midwest Studies in Philosophy 17:1. 255-267.Davidson, Donald. (1993). Locating Literary Language. In Reed Way Dasenbrock, editor, Literary Theory After Davidson. University Park, PA.:Pennsylvania State University Press, 295-308.Davidson, Donald. (1995). The Objectivity of Values. In 2004b 39-51.Davidson, Donald. (1996). The Folly of Trying to Define Truth.

The Journal of Philosophy

Vol. 93, No. 6 (Jun., 1996), pp. 263-278.

Davidson, Donald. (2004a). Problems of Objectivity, Oxford University

Press.

Davidson, Donald. (2005). Truth and Predication. (Cambridge, Mass:

Harvard University Press.

Dehghani, M.R., R. Iliev, and S. Kaufmann. (2012). Causal

Explanation and fact mutability in counterfactual reasoning. Mind

and Language. 55-85.

Dennett, Daniel, Content and Consciousness (Routledge & Kegan Paul

Books Ltd, 1968.

Derrida, Jacques. (1977). “Signature, Event, Context,” Glyph 1,

172-197.

Deacartes, Rene. (1644). Principles of Philosophy.Elsevir.

Doetjes, J., C. Constantinescu & K. Součková. (2010). A neo-Kleinian approach to comparatives. In S. Ito & E. Cormany, Proceedings of SALT XIX

Doetjes, J. (2009). Cross polar (a)nomalies without degrees. Bucharest Working Papers in Linguistics, Vol. XI, 5-20.



Edgington, Dorothy. (1995). On Conditionals. Mind 104. 235-329.

Elder, Crawford. (2005). Real Natures and Familiar Objects. MIT Press.

Frege, Gottlob. (1892). Uber Sinn und Bedeutung. Reprinted in English translation as “On Sense and Meaning” in Martinich, A. The Philosophy of Language. Oxford UP. 1985. 200-212. Geach, Peter.(1962). Reference and Generality. Cornell University

Press.

Gibbard, Alan. (1981). Two Recent Theories of Conditionals”. In

W.L. Harper, R. Stalnaker, and G. Pearce (eds.), Ifs: Conditionals,

Beliefs, Decision, Chance, and Time. Dordrecht, Holland: D. Reidel. 211–

247.

Goodman, Nelson. (1955). Fact, Fiction and Forecast. Harvard University Press.Goodman, Nelson. (1966), The Structure of Appearance, 2nd edition,

1966, Bobbs-Merrill, p.134.

Grice, H.P.(1967) Logic and Conversation. In Grice (1989).22-41.Grice, H.P. (1989). Studies in the Way of Words. Harvard University Press.Hegel, G.W.F. (1807). The Phenomenology of Spirit. Translated byA.V.Miller. Oxford University Press.Heidegger, Martin. (2000). Introduction to Metaphysics.Yale UniversityPress.Heim, Irene and Kratzer, Angelika. (1998). Semantics in GenerativeGrammar. Oxford: Blackwell Publishing.Hempel, Carl. (1960). Inductive Inconsistencies. Synthese Volume 12 Number 4. 439-469.Hempel, Carl. (1965). Studies in the Logic of Confirmation. In Aspects of Scientific Explanation. The Free Press: New York.Horwich, Paul. (1998). Truth. Oxford University Press.



Iatridou, S. (2000). The Grammatical Ingredients of

Counterfactuality. Linguistic Inquiry 31(2). 231-270.

Jackson, Frank. (1979). On Assertion and Indicative Conditionals.

Philosophical Review 88. 565-589.

Jockusch,Elizabeth L, and Wake, David B.(2002).Falling apart and

merging: diversification of slender salamanders (Plethodontidae:

Batrachoseps) in the American West. Biological Journal of the Linnean Society

76, 361–391.

Kant, Immanuel. (1781). Critique of Pure Reason.

Kant, Immanuel. (1785). Groundwork of the metaphysics of Morals. In Kant (1983) Ethical Philosophy, James Ellington, translator. Hackett, and Kant (1911) Akademie Ausgabe Kant’s Gesammelte Schriften. Volume Four. Konigliche Preussiche Akademie der Wissenshaften, Berlin.Kaplan, David. (1968). Quantifying In. Synthese 19 (1-2):178-214.

Kaplan, David. (1990). “Words,” Proceedings of the Aristotelian Society.

Keefe, Rosanna and Smith, Peter. (1997). Vagueness: A Reader. MIT Press.Keefe, Rosanna. (1998). Vagueness by Numbers. Mind, vol. 107.565–

579.

Kennedy, Christopher. (1997). Projecting the Adjective. PhD dissertation, UC Santa Cruz.Kenny, Anthony. (1963). Action, Emotion and Will. London: Routledge.

King, Jeffrey C.(2001). Complex Demonstratives, MIT Press, Cambridge.Klein, Ewan. (1980). “A semantics for positive and comparative adjectives.” Linguistics and Philosophy 4:1–45Klein, Barry. (1983). Plurals and Events. (Cambridge: MIT Press)

Kratzer, A. (Forthcoming). The Event Argument and the Semantics of Verbs. http://semanticsarchive.net/Archive/GU1NWM4Z/



Kratzer, Angelica. (2012). Modals and Conditionals. Oxford UniversityPress.Kripke, Saul. (1980) Naming and Necessity, Harvard UP.Langendoen, D.T. and Postal, P.M. (1986). The Vastness of NaturalLanguages. Oxford, Basil Blackwell.Larson, Richard. (1995). Olga is a Beautiful Dancer. LSA NewOrleans, January, 5-8.Larson, Richard and Segal, Gabriel. (1995). Knowledge of Meaning.Cambridge, MA: MIT Press.Latour, Bruno and Woolgar, Stephen. (1979). Laboratory Life: the Social Construction of Scientific Facts. Los Angeles: Sage Publications.Lepore, Ernie and Ludwig, Kirk. (2005). Donald Davidson Meaning, Truth,Language, and Reality. Oxford University Press.Lepore, Ernie and Ludwig, Kirk. (2007). Davidson’s Truth-Theoretic Semantics. Oxford University Press.Lewis, David. (1976). Probabilities of Conditionals and Conditional Probabilities. Philosophical Review 85 (3). 297-315.Lewis, David. (1970). General Semantics. In D. Davidson and G. Harman, (eds.). Semantics of Natural Language. (pp.169-218). Dordrecht: D. Reidel Publishing Co.Lewis, David. (1979). Score-Keeping in a Language Game. Journal of

Philosophical Logic 8, 339-59.

Lewis, David. (1986). On the Plurality of Worlds, OUP 1986, page 213,

Lewis, David. (1991). Parts of Classes. Blackwell.Ludwig, Kirk and Ray, Greg. (2002). Vagueness and the Sorites Paradox. Philosophical Perspectives 16. 419-461.Lycan, William. (1984). Logical Form. MIT Press.Lycan, William. (2001). Real Conditionals. Oxford University Press.Mackie, John (1977) Ethics: Inventing Right and Wrong. Viking press.Mackie, John (1980). The Cement of the Universe: A Study of Causation. OxfordUniversity Press.McCall, Storrs. (1976). Objective Time Flow. Philosophy of Science 43:3. 337-362.McDaniels, Kris and Bradley, Ben. (2008). Desires. Mind Vol 117. 267-302.

http://www.jstor.org/sici?sici=0031-8108(197607)85:3%3C297:POCACP%3E2.0.CO;2-T

http://www.jstor.org/sici?sici=0031-8108(197607)85:3%3C297:POCACP%3E2.0.CO;2-T



McGee, Vann. (1985). A Counterexample to Modus Ponens. Journal of Philosophy, 82: 462-471.McTaggart, John M. E. (1908). The Unreality of Time, Mind 17. 457–474.Millikan, Ruth. (1984). Language, Thought, and Other Biological Categories. MIT Press.Moline, Jon. (1969). Aristotle, Eubulides, and the Sorites. Mind 78. 393-405.

Moltmann, Frederike. (2009). Degree structure as trope structure:a trope-based analysis of positive and comparative adjectives.Linguistics and Philosophy 32, 51-94.Montague (1973). The Proper Treatment of Quantification in English. In Hintikka, Moravcsik, and Suppes (eds.) Approaches to Natural Language. Dordrecht. 221–242.

Mostowski, Andrej. (1957). On a generalization of quantifiers. Fundamenta Mathematicae 44.12–36.

Nagel, Thomas. (1970). The Possibility of Altruism. Clarendon Press: Oxford.Neale, Stephen. (2001). Facing Facts. Oxford University Press.Pietrosky, Paul. (2000). Causing Actions; (Oxford: Oxford UP).

Plato. (1939). Parmenides. In Plato and Parmenides. Translated by F.M. Cornford. Routledge & Kegan Paul.Plato. (1952). Phaedrus. Translated by Hackforth. CambridgeUniversity Press.Plato. (1977). Phaedo. Translated by G.M.A. Grube. Hackett Publishing.Plato. (1990). Theaetetus. Translated by Levett. Hackett Publishing.Plato. (1993). Sophist. Translated by Nicholas White. HackettPublishing.Prior, Arthur. (1968). Papers on Time and Tense. Oxford: Clarendon

Press.



Putnam, Hilary. (1973). Meaning and Reference. Journal of Philosophy 70. pp. 699-711.Putnam, Hilary. (1975). “The Meaning of "Meaning” in: Keith

Gunderson (ed.), Minnesota Studies in the Philosophy of Science, Vol. VII

Language, Mind and Knowledge. Univ. of Minnesota Press, Minneapolis.

131-193.

Quine, W. V. (1948). On What There Is. In Quine, W. V. 1953a. 1-19.Quine, W. V. (1951). Two Dogmas of Empiricism. The Philosophical Review 60, 20-43. Quine, W. V. (1953). Reference and Modality. In From a Logical Pointof View, Harvard University Press. Quine, W.V. (1953a). From a Logical Point of View. Harvard UniversityPress.Quine, W. V. (1953b). Three Grades of Modal Involvement. In The

Ways of Paradox and other Essays. Harvard University Press, 1976. 158-

176.

Quine, W. V. (1960). Word and Object. Harvard University Press.

Quine, W.V. (1969). Epistemology Naturalized. In Quine (1969a).

69-90.Quine, W.V. (1969a). Ontological Relativity and Other Essays, ColumbiaUniversity Press.Quine, W.V. (1970). Philosophy of Logic. Cambridge, MA.: Harvard

University Press.

Ramsey, Frank P. (1929). General Propositions and Causality. In

Braithwaite, ed. The Foundations of Mathematics. Littlefield, Adams &

Co: Patterson, NJ. (1960).

Rayo, Augustin. (forthcoming). “A Plea for Semantic Localism,”

Nous.



Rescher, Nicholas. (1964). Plurality Quantification. The Journal of Symbolic Logic, vol. 27 373-74.

Roberts, Craige. (1996). Information Structure: Towards an integrated formal theory of pragmatics. In Jae Hak Yoon and Andreas Kathol (eds.) OSUWPL Volume 49: Papers in Semantics. The Ohio State University Department of Linguistics.

Ross, W. D. (1930). The Right and the Good. Oxford University Press.

Rumfitt, Ian. (1993). Content and Context: the Paratactic Theory

Revisited and Revised. Mind, Vol 102, July 1993, pages 429-454.

Saussure, Ferdinand. (1977). Course in General Linguistics, trans. W.

Baskin Glasgow: Fontana/Collins.

Scanlon, Timothy. (1998). What We Owe to Each Other. Harvard

University Press.

Schaffer, Jonathan (2003). Is there a fundamental level? Nous 37

(2003) pp. 498-517.

Schein, Barry. (1993). Plurals and events. MIT Press.

Schwartzchild, Roger, and Wilkinson, Karina.(2002). Quantifiers in Comparatives:A Semantics of Degree Based on Intervals. Natural Language Semantics

10. 1-41.

Seuren, Peter. (2005). Eubulides as a 20th-century semanticist.

Language Sciences, 27(1), 75-95.

Smith, Michael. (2012) "Agents and Patients, Or: What We Learn about Reasons for Action byReflecting on Process-of-Thought Cases" forthcoming in Proceedings of the Aristotelian Society.Smith, Michael. (2012) The Rational Foundations Of Morality



Smith, Michael. (2012) "A Constitutivist Theory of Reasons" forthcoming in Law, Ethics, and Philosophy.Stalnaker, Robert.(1978). Assertion. In Stalnaker, Robert (1999) Context and Content. Oxford University Press. 78-95.Stalnaker, Robert. (2002). Common Ground. Linguistics and Philosophy 25:

701–721.

Stalnaker, Robert. (2003). Ways a World Might Be. Oxford University Press. Strawson, Peter. (1959). Individuals. Methuen.Tarski, Afred. (1933). “The concept of truth in the languages of the deductive sciences” (Polish), Prace Towarzystwa Naukowego Warszawskiego, Wydzial III Nauk Matematyczno-Fizycznych 34, Warsaw; reprinted in Zygmunt 1995, pp. 13–172; expanded English translation in Tarski 1983, pp. 152–278.Taylor, Richard (1957). The Problem of Future Contingencies. The Philosophical Review, Vol. 66, 1. 1-28.Tolhurst, William and Wheeler, Samuel Wheeler. (1979). "On Textual Individuation” Philosophical Studies 35, 187-197.Tooley, Michael. (1997). Time, Tense, and Causation. Oxford University Press.

van Inwagen, Peter. (1990). Material Beings. Cornell University Press.

Wheeler, Samuel. (1972). Attributives and their Modifiers. Nous, 6:4. 310-334.

Wheeler, Samuel. (1974). Inference and the Logical “Ought.” Nous,8:3. 233-258.

Wheeler, Samuel. (1975). Reference and Vagueness. Synthese, Volume

30. 367-379.

Wheeler, Samuel. (1978). Quantification in English. Philosophia, Volume 8, No. 1. 31-42.

http://www.jstor.org.ezproxy.lib.uconn.edu/stable/2182851?&Search=yes&searchText=future&searchText=taylor&searchText=richard&searchText=problem&searchText=contingencies&list=hide&searchUri=%2Faction%2FdoBasicSearch%3FQuery%3Drichard%2Btaylor%2Bthe%2Bproblem%2Bof%2Bfuture%2Bcontingencies%26acc%3Don%26wc%3Don&prevSearch=&item=4&ttl=1872&returnArticleService=showFullText



Wheeler, Samuel. (1979). On That Which is Not. Synthese 41.155-173.

Wheeler, Samuel. (1980). Natural Property Rights as Body Rights. Nous, Volume XIV, No 2. 171-194.

Wheeler, Samuel. (1983). Megarian Paradoxes as Eleatic

Arguments", American Philosophical Quarterly, Volume 20:3. 287-295.

Wheeler, Samuel. (1986). Persons and their Micro-Particles. Nous,Vol XX:3. 333-349.

Wheeler, Samuel. (1991). True Figures. In Hiley (ed.)The InterpretiveTurn. Cornell UP. 197-217.

Wheeler, Samuel. (2007) “Wahrheit, Metapher und Unbestimmheit,”

in Zur Metapher, ed. Czernin and Eder, Wilhelm Fink Verlag, 2007.

Wheeler, Samuel. (2010). Pure Realism. Annales Philosophici, University Of Oradea’s Annals Of Philosophy, Volume 1, Issue 1. 83-100.

Wheeler, Samuel. (2011). Davidson, Derrida and Difference. In

Malpas, ed. Dialogues with Davidson, MIT Press. 29-42.

Wheeler, Samuel. (2012). Remembering Donald Davidson: His 1967

Undergraduate Philosophy of Language Course. In Baghramian, ed.

Donald Davidson: Life and Words. Routledge. 65-70.

Weiss, Paul. (1938). Reality. Carbondale: Southern Illinois

University Press.

Weiss, Paul. (1958). Modes of Being. Carbondale: Southern Illinois

University Press.

Williamson, Timothy.(1994). Vagueness. Routledge.



Wilson, Neil L (1959). Substances without Substrata. The Review of

Metaphysics, XII/4(48): 521-539.

Wittgenstein, Ludwig. (1953). Philosophical Investigations. Macmillan.

Neo-Davidsonian Metaphysics: From the True to the Good

Documents

Transcript of Neo-Davidsonian Metaphysics: From the True to the Good