Iyyun, The Jerusalem Philosophical Quarterly, 44, 1995, 243-72. Naming and Necessity: A Second Look

By Joseph Agassi Tel-Aviv University and York University, Toronto

Abstract Saul Kripke’s study of the truth conditions for modal and for subjunctive conditional statements is meant to validate essentialism. To this end he develops his rigid theory of rigid designators, and he enlists the service of normal intuition. Even were his argument valid, his conclusion is unacceptable since language and intuition are flexible and open-ended, and since essentialist expressions ordinarily serve as mere metaphors. 1. Introduction A synoptic picture of the analytic school of philosophy is hard to acquire: being analytic it cannot incorporate any unifying principle. Initially its self-appointed task was the mere clarification of confusions. A general theory of confusion was never presented; all that this school has offered was the assertion that all philosophical pronouncements are confused. Efforts to present the diverse contributions to analytic philosophy in the older categories of traditional problems of philosophy are useless, since that school aimed at the dissolution (not solution) of all of them, thus implicitly rejecting any classification based on them. (“The puzzle does not exist”, said Ludwig Wittgenstein in his Tractatus Logico-Philosophicus. “To ask a philosophical question is sick”, he added in his Philosophical Investigations.)1

This was early analytic philosophy. Hard as it is to view it synoptically, it is even harder to view synoptically its current variants, especially since current views of its early phase are puzzling. The case of Saul Kripke is particularly puzzling. In his Naming and Necessity he says (93),2 “philosophical theories are in danger of being false”, contrary to Wittgenstein’s teaching. He has presented Wittgenstein as a philosopher who attempted to solve a philosophical problem, in a puzzling oversight of Wittgenstein’s repeated and insistent disclaimers.3

1 Double quotes are used to present quotations and single quotes to present the locutions under discussion. 2 Unless otherwise specified, numbers refer to pages of Saul Kripke, Naming and Necessity, Princeton: Princeton University Press, 1980. 3 Saul Kripke, “Wittgenstein on Rules and Private Language: an elementary Exposition” (Cambridge MA: Harvard UP, 1982). See also my “Wittgenstein and Physicalism, “ Grazer Philosophische Studien, 41 (1991): 67-97, and my “Whatever Happened to the Positivist Theory of Meaning?” Zeitschrift für allgemeine Wissenschaftstheorie 28 (1987): 22-29.

Analysis is not the idea that there are no clear philosophical assertions: it is a set of techniques, of tools and standard uses for them. The challenge of presenting the analytic school’s output synoptically may be met by assorting the analytic literature according to the tools it uses. Moreover, since the very idea of clarification as well as the tools of its execution has undergone some change, the history of this change may be used as a unifying principle for its study. First the analytic school claimed that all hypostatization or reification (the treatment of abstract objects as if they were concrete) is confusion; essentialism was then taken to be the paradigm of this kind of confusion. Kripke changed all this. Having forged some new analytic tools, he has tried to use them to revive essentialism. He is in error: even were his analysis correct, his defense of essentialism still is superfluous hypostatization. 2. Exposition Kripke’s celebrated Naming and Necessity is a fairly recent book that is sufficiently prominent and seminal for sufficiently many years to count as a hardy perennial. Contrary to the analytic tradition4 he advocates essentialism there, using his theory of possible worlds that is reminiscent of the ontology of Alexis Meinong and/or the phenomenology of Edmund Husserl: prima facie it is a hypostatization.5 Whether this impression is correct remains to be seen, of course; its prima facie plausibility is intriguing. The book investigates the meaning of modal and subjunctive conditional statements and the conditions under which they are (objectively) true.6

By definition a subjunctive conditional is a declarative conditional put in the subjunctive mood, namely, roughly having the form ‘if x were the case, then y would be the case,’ or ‘were x then y’, etc. I will take as my rubber-stamp example here ‘if I were king I would lay the stars at your feet’. This statement is heavily metaphorical; the following is not. Were it taken literally, Kripke would reject it; taken as the metaphor that it is, he might endorse it (as holding for its author, Françios Villon).

4 Anti-metaphysicians are neutral to metaphysical disputes, except concerning meaning: they are (or try to be) ultra-nominalists. 5 The role of the qualifier here is to avoid the anachronistic application of Kripke’s “transworld identifications” and/or Lewis’s realism of possible worlds to Meinong and Husserl. What is the sense of “real” here? See Joseph Agassi and Paul T. Sagal, “The Problem of Universals”, Philosophical Studies 28 (1975): 289-94. 6 Much of the book is devoted to a detailed study of designation, for reasons explained below.

To avoid misunderstanding, it should be kept in mind that the task is not to stipulate meanings and truth conditions ― for subjunctive conditionals or any other locutions ― but to analyze their ordinary use in order to discover their usually-intended meanings and thus their current truth conditions.7

My rubber-stamp example as it stands, ‘if I were king I would lay the stars at your feet,’ being metaphorical, is possibly unacceptable. This depends on Kripke’s decision concerning metaphorical sense: is it ordinary or extraordinary? Let us take it literally, then. It is false. Consider any assertion of the form ‘were x the case then y,’ where x is (presumably) false (but consistent).8

It is possible to imagine a world or worlds (presumably) different from ours, in which x is true, and the question then is, roughly, how different does ordinary verbal practice allow one to imagine that difference to be? This concerns the truth condition of a statement of the form ‘possibly x is the case’ or ‘possibly x’ or ‘maybe x’ or ‘perhaps x. ‘ This clearly depends on the question, how far does ordinary verbal practice limit the imagining of possibilities? It is important to notice that the locution ‘were x the case,’ and its rough equivalent, ‘possibly x’, are ambiguous in this very respect. (Otherwise there would be neither need nor room for Kripke’s study.) (The antecedent of my rubber-stamp example, ‘if I were king,’ is unclear: to what kind of a king it alludes is undecipherable.) There is a standard, though perhaps not too ordinary, approach or technique that should be excluded a priori. It offers the choice of the minimal deviation from the given that solves the problem analytically: it is the choice of the slightest deviation from reality which suffices to transform any given subjunctive conditional into a true declarative statement. On this approach a possible world should be chosen such that, considering the sentence ‘possibly x’ it deviates from the real world just sufficiently to render ‘x’ true and to eliminate any subsequent inconsistency. This solves the problem analytically, if at all, as it renders true all assertions of possibilities. This may be found insufficient, not just because it trivializes the problem, but because it does so, if at all, at a very high cost, namely, while leaving (modal) meanings incurably vague, especially since it is not clear how a model

7 Possible worlds are “not discovered by powerful telescopes “ but by introspection and intuition of meanings, though possibly these were initially determined by stipulations (42, 44, 49). 8 Declarative statements that are (bona fide) logically true or false are unproblematic here and they will be ignored.

should be altered minimally so as to prevent a contradiction as a result of changing the truth value of a single sentence (57).9

It is preferable then to admit this as a desideratum for adequate (possibly true, putative) solutions. Let us limit our study to those solutions, for which some but not all subjunctive conditionals are true. Question: under what conditions is this desideratum possible to abide by? Answer: the meaning of all subjunctive conditionals should be fixed a priori (i.e., prior to the decision as to their truth values). Kripke’s task is here chiefly to meet this condition in unique decisions. This is impossible: the ambiguity and openness of the modal and the subjunctive are built into language, and so the general, decisive rule which he is after ordinarily does not obtain; it is impossible to detect a priori any unique meaning of any modal or subjunctive conditional statement. What exactly, then, is the deviation from reality to the realm of the possible is intended in ordinary expressions of modal or subjunctive conditional statements? The answer is highly context-dependent. If and when its context is problematic, then a decision must be made: interlocutors negotiate the matter between themselves if and when they find it important enough. The context-dependence of the meaning of such statements, and hence their truth conditions, is empirically observable in the very prevalence of such negotiations. Noam Chomsky resolves ambiguities syntactically, by offering alternative options for the meaning of ambiguous statements; each option is stated as some unambiguous sentence possibly equivalent to it. He silently assumes that all chosen alternatives clarify all ambiguity and exhibit deep (real, essential) structures, thus rendering all sentences essentially (syntactically) clear. Kripke agrees (62n) and resolves ambiguities semantically this way. As some ambiguities may be resolved semantically or syntactically, followers of Chomsky and Kripke like Asa Kasher must choose.10

When in daily life a statement of a possibility is asserted, how extravagantly do interlocutors permit their imagination to go? In examining the meaning of such

9 This is a variant of an intriguing problem: the decision to alter one’s views minimally under the force of valid criticism, how should it be executed? Popular philosophers of science usually take as unproblematic this, hardly soluble problem. 10 Specific case-by-case examinations are required to distinguish between cases of semantic verbal ambiguities, of syntactic ones and of informative vagueness. See my “The Secret of Carnap”, Philosophia, 10 (1981): 57-88, reprinted in my The Gentle Art of Philosophical Polemics (LaSalle IL: Open Court, 1988). “Logicians have not developed the logic of vagueness”, says Kripke (51n), ignoring Arne Naess.

a statement, how much are interlocutors allowed to vary their admission of what is possible? If no constraint is placed on such exercises, then the conclusion is compelling that they are all true under some readings and false under other readings. Example: translate my rubber-stamp example by unpacking its metaphors11 to, say, ‘if it were within my powers, I would make you feel wonderful’. One then may read the assertion to mean, in a declarative mood, either (a) that I am nice specifically to you, or (b) that I am generally nice. Suppose now that we have to choose between one of these two options as the truth but are not allowed to assert both (for reasons that can be easily built into some context). There is no adjudication between them, and so there can be no rule to satisfy Kripke’s search for the unique truth conditions for the reading of this assertion. When the narrow context of this assertion allows for either of the two readings but the broad one does not, then, the way to find out which of them is intended, is to ask for an elucidation. There is no way around this.

We may wish to have even more leeway; we may take the antecedent of my rubber-stamp example literally and its consequent metaphorically, to be translated to the declarative mode as above. A world can be imagined, then, in which kings are nice, so that in that world, as a king I am nice, so that in that world the rubber-stamp example is true, though, of course, in a world in which only nasty people are kings the opposite obtains. Again, Kripke’s program is blocked.

Kripke is aware of this objection, and he legislates away the kind of license it is based on, and even quite ad hoc (justly or not). To prevent the sort of license employed in the previous paragraph, Kripke suggests that in common practice, while varying our image of the world, we do suppress some kind of variation ― of the first person singular (‘if I were king’), or more generally, of some proper names and even of some common nouns (‘if I were king’). The designating terms, the variation of which he says we suppress, he calls “rigid designators”. (Even the first person singular is rigid, he says, 10n, 49n.) The fixing of common general

11 Metaphors are inherently ambiguous. Ambiguity, says William Fowler, demarcates (live) metaphors (from dead ones). Borderline cases scintillate in poems which revive dead metaphors and conceive new ones; a poem’s context appears swiftly as an (often unusual) association network. For association networks see Max Black, Models and Metaphors (Ithaca NY: Cornell University Press, 1962). This invites the study of interactions between the subjunctive and the metaphorical, especially when combined, since background knowledge and its background pertain to intellectual growth. (See Black, op. cit., 242.) The subjunctive and the metaphorical interact through the association networks that function as the intellectual frameworks that function as contexts. In his hostility to metaphysics Black overlooked this: his celebrated essay is an attack on the Sapir-Whorf hypothesis that all language is imbued with metaphysics. Kripke treats this territory like a minefield.

terms (like ‘king’) he views as the first step towards establishing natural kinds, so that it heralds his essentialism.

The rigidity of some designators is hard to specify, Kripke admits (3): ‘if two objects are identical, then they are necessarily identical’ is obviously true, he says. “What pair (x, y) could be counter-examples?” he rhetorically asks. “Not pairs of distinct objects, for then the antecedent is false; nor any pair of an object and itself, for then the consequent is true.” He concludes: “If ‘a’ and ‘b’ are rigid designators, it follows that ‘a = b,’ if true, is a necessary truth.” Since this notoriously fails for ‘the morning star’ and ‘the evening star,’ since the identity of the two stars with each other was empirically discovered, then their designators seem to be not rigid. They are (4-5): whether designators are rigid or not, some features contingently identify the objects they designate; yet these objects, being identical, are necessarily identical ― “even though one may not know that a priori” (108). Here empirical research ends up in a tautology! Kripke rationalizes this oddity: names behave oddly and naming is a complex operation (109n).

Definite descriptions, then, need not be rigid, and then they are not names (5). Hence, if ‘Scot’ is an abbreviation for ‘the author of Waverley’, then, since for all we know anyone could have written Waverley, ‘Scot’ is no more a name than ‘someone’ (though their ranges usually differ). Hence, often a name we regularly use is ambiguous, since it may be rigid but it need not be. This point, Kripke nicely illustrates, is significant for myths.

Nevertheless, Kripke insists that in ordinary discussions of possibilities some items are ordinarily meant to remain rigid and are so understood by ordinary speakers. Thus, in our rubber-stamp example, the moral nature of kings is as rigid as the noun ‘king’, so that one is not allowed to make them benign or nasty at one’s pleasure. The above objection is thus precluded.

What is the rule for rigid designation? We may make it very strict, since we may easily speak of designators in a subjunctive mood, as in the immortal expression ‘a rose by any other name would etc.’ This, however, is a confusion: we may vary possible world to alter designators, but, Kripke demands, when we discuss the meaning and truth conditions of a subjunctive conditional statement, we should not vary designators; thus, it is not that there is no possible world in which a rose is called by some other name, but that saying that a rose by any other name would smell so sweet, while deviating from a description of the real world to a description of a possible world (where a rose is called by some other name), in our

discourse the word meant to designate a rose still is ‘rose’: That which we call a rose etc. in a possible world which is just like ours except that there a rose is called by some other name, etc. Still, we have only excluded one move in the study of the possible rigidity of a rigid designator; more is needed if we are to be clear about the matter: how are we to decide the meanings of subjunctive conditional and modal statements in ordinary parlance?

My presentation is thus far unsatisfactory, and so perhaps it should be clarified in a manner that involves some preliminary technical moves.

DEFINITION: Let us call a couple of statements “associates” if and only if the two are identical except that only one of them employs modal or subjunctive words.

FIRST CLARIFICATION: A modal statement asserting a possibility/necessity is true if and only if its declarative associate is possibly/necessarily true.

SECOND CLARIFICATION: a subjunctive conditional statement is true if and only if its declarative associate is possibly true.

THIRD CLARIFICATION: the possible truth of a declarative statement is the existence of a permissible substitution that renders it true.

This requires a FOURTH and final CLARIFICATION, to determine the answer to the following question.

QUERY: which substitution is permissible? (50-1) Kripke’s program is to answer this question adequately. The ranges of

substitution, then, will be a natural classification and thus portray essences. Were the constraint on permissible substitution lax, then all permissible statements of the possibility of a state of affair, and all such subjunctive conditionals would be true as a matter of course, since there will always be a substitute that renders their (consistent) declarative associates true. Consider this last subjunctive conditional statement, ‘were the constraint on permissible substitutions lax, then all permissible statements of possibility and all permissible subjunctive conditionals would be true.’ It is trivially true, yet Kripke does not permit it, as will soon transpire.12 To Kripke’s question, then. How wide is the range of permissible variation from the real to the possible? What are the rules of permissible

12 The limitation on Kripke’s system is rooted in the rigidity of his view of language. Essentially, his essentialism is his rigidity. The same hold for Husserl. See K.R. Popper, The Open Society and Its Enemies (London: Routledge, 1945) Index, Art Husserl.

substitution? How lax are they permitted to be? (How much permission is required/permitted/recommended by custom/logic/essentialism?)

The first-person singular can be replaced in any sentence that can be uttered by another interlocutor. When in a discourse in the subjunctive mood about Tom is made, could interlocutors vary the name or the object thus named, since possibly not Tom but Dick was christened ‘Tom’? Obviously, sometimes but not always, such variations are permitted. What can be learned from this? Consider discussions about cats (122-3, 125-6, 135). When a statement in the subjunctive mood is permitted in which they are spoken of as if they were dogs, might one, in the same way, also speak of them as if they were not animals but, say, robots? If not, what of it?

Kripke’s desideratum for an adequate or possibly true or putative answer to this question is revealing: if we are not allowed to substitute for ‘cat’ any other class name except that of some other animal, then we thereby admit that we understand cats to be animals ― per se or intrinsically or inherently or essentially so. This is an error: at times we are allowed to substitute for cats only mammals, or only pets; and then we would thereby admit that we understand cats to be mammals or pets ― again, per se or intrinsically or inherently or essentially so. This is absurd.13

Suppose Kripke is (were) right. What is the lesson from this? It is obviously this: ordinary language is essentialist; hence, either ordinary language misleads, or essentialism is true. If ordinary language misleads, then it misleads the way it misleads us to read any metaphor literally. Hence, we should not, Heaven forbid, hypostatize the essentialism of ordinary language any more than may be (misleadingly) suggested by the magic or religion that is inherent in some of ordinary, commonly used locutions (especially poetic ones).

The confinement of most of the examples from ordinary language to prosaic expressions is a cop out. Wittgenstein made some brief forays into new territories, and he was repeatedly baffled. Only in this context he conceded that some metaphysics is inherent in ordinary parlance, and, he added, it is not binding. Black

13 The line of argument here is essentially Quine’s: “to be is to be the value of a variable. “ Kripke adds, and to be essentially the value of “that” variable, is to dwell within “its” range. That range, then, preexists (10-13)! Essence, then, precedes existence. Does The non-existent object designate rigidly? Significantly, though Kripke considers this question a mere technicality, he avoids discussing existence (21, 21n., 31-3, 58-9, 65-7, 67n, 110. Also, a forthcoming work on the topic is promised, 158.)

and Chomsky say the same.14 Indeed, religious expressions embedded in language are literal for the religious and metaphorical for others: no one wants religion dictated by grammar. Einstein’s ‘God does not play dice’ is meant utterly metaphorically (17).

Does analytic philosophy condone magic or religion, or does it expose them as mere confusions? This is possibly a controverted issue. Still, thus far, thank Heavens for small mercies, no one has suggested that ordinary magical or religious locutions impose on their users magical or religious belief. Otherwise many people would be unable to say even ‘bless you!’ Wittgenstein is read by most of his followers as a defender of ordinary language but hostile to essentialism; he is possibly in error, in viewing ordinary language as essentially anti-essentialist, yet he is right in allowing ordinary language its ordinarily advantageous position while viewing the essentialist (and other) metaphysics inherent in it as not binding.

Does Kripke’s view of ordinary language as essentialist vindicates essentialism or warns against being misled by it? I think he favors the former option, contrary to the repeated observation that modal and subjunctive conditional statements about cats and dogs are meant to include at times all pets, at times all mammals, at times all animals, and at times even all (material) things (‘maybe it rains cats and dogs’.)15 Hence, there is no absolute answer to the question, which substitution limitation on a modal or a subjunctive conditional statement is adequate? It is context-dependent, so that the restrictions Kripke depicts as essential do not represent essences. This invites a textual analysis. To the extent that his text emits the impression that his essentialism is of the traditional kind (rather than that it bears only faint family resemblance to it) that text misleads.

Kripke’s discussion diverges from Quine’s, as he rejects the observation that the scope of a noun, not to say its very meaning, may vary with the context of its use. He discusses the scope of ‘gold’ (123-5) and raises the question of its possibly not being a chemical element. What is this to him he does not say. It seems obvious, however: unlike the classical essentialists, he allows for conceptual change; yet he considers himself essentialist all the same, as he maintains a hard core of unchangeable meaning: as essences are universal, essential meanings should be independent of contexts (39ff.). For example, he is willing to allow that

14 See note 11 above; see my Towards a Rational Philosophical Anthropology (Dordrecht: Kluwer, 1977), Ch. 2. 15 This illustrates the close link between modality and metaphors ― in accord with Black’s hypothesis.

possibly gold is not an element, but not to stray too much from our ordinary inexpert concept of gold. He neglects the possible dependence of the very meaning of ‘gold’ on the acceptance or rejection of alchemy, even though he should permit it, as he agrees (139) that he should allow for a historical continuity of the transmission of the term. Yet ordinary twentieth-century English allows the use of alchemical locutions -- and then they are presumed metaphorical (‘turn dirt into gold’). Should we grant him that every significant concept is essentially unalterable? 3. Development Inexplicably, Naming and Necessity follows the custom of replacing ‘subjunctive’ with ‘counterfactual’ and ‘contrary-to-fact’. These last two labels are indeed synonymous: they were introduced in the classical text Symbolic Logic of Lewis and Langford to designate conditional statements whose antecedents are false, and which are thus obviously true; Kripke’s problem concerns the meanings and the truth-conditions of subjunctive conditionals. Its problem is rooted in the fact that some subjunctive conditionals are false, and so obviously their truth value differs from that of their declarative associates: it is not determined by the (presumed) falsehood of their antecedents; hence, the meanings of subjunctive conditionals obviously differ from that of their associated counterfactual or contrary-to-fact conditionals. Calling subjunctive conditionals counterfactual or contrary-to-fact is therefore most peculiar, particularly in a book on the truth conditions of subjunctive conditionals. (The term ‘subjunctive’ is quite ordinary, unlike the terms ‘counterfactual’ and ‘contrary-to-fact’ which are definitely extraordinary, as many authors of the ordinary-language persuasion have amply and redundantly illustrated.) Since the question of the truth conditions of subjunctive conditionals depends on the question of the meaning and truth conditions of modal statements, and since this is tricky, we may wish to know, why this important?

A supposition was made above, incidentally, that the declarative associates of subjunctive conditionals are contrary-to-fact. This supposition is dangerous, as meaning is meant to precede truth value. (One must understand before one agrees.) The antecedent of the conditional associated with a subjunctive conditional may be true, if only because of human error: a nineteenth-century physicist might say ‘were it possible to split atoms …’ presuming that the laws of nature forbid it. For the purpose of the present discussion, however, suffice it to confine our discourse to subjunctive conditionals associated with contrary-to-fact (synthetic) conditionals

and modal statements of possibility associated with false (synthetic) declarative statements.

Modality is problematic, and its problematic character is known from the scholastic literature, which distinguishes at least possibility by the rules of logic from possibility by the laws of nature, what is permissible by logic (consistency) from what is permissible by the laws of nature (regardless of what these are). These, of course, are named, respectively, “modality de dicto” and “modality de re”. Further, the identification of subjunctive conditionals with statements of laws of nature is normal, for example, in legal parlances; see below. This usage is taken for granted by many a modern writer on the topic.16

Kripke’s view regarding modality de re is different. He postulates that language is (almost) rigidly given and the variations permissible for the determination of the truth conditions of modal (and of subjunctive-conditional) statements are narrower than those permissible by formal logic and wider than those permissible by the (putative) laws of nature. He postulates a ‘physical’ modality that unlike modality de re depicts not natural laws but essences (110). Before discussing whether this move is admissible, it should be noted that it may be suspected of circularity: it plainly heralds his conclusion. Now the traditional essentialists forbid circularity; it is permitted by most modern logicians, certainly by Quine, who views all dictionaries as circular. Kripke himself finds circularity a defect (88n and 160.)

Kripke takes meanings to be not fully fixed by essences, and so they only partly decide truths. More can be done by adding laws of nature. Is this possible? Aristotle took his theory of essences (of natural classification) as his system of natural laws. Whether his system is consistent or not is hard to judge, but certainly with its erosion that began already with Euclid and Archimedes (see below), it became increasingly unfeasible. Essentialism is unfeasible, as noted by many writers since Robert Boyle, in that it is anthropocentric and even animistic. Assuming that laws of nature are not Aristotelian essences, one is thus tempted to do without them. As the use of animism in ordinary parlance without commitment to it is generally condoned, this should cover the ordinary use of essentialism. In brief, there seems to be neither need nor possibility for Kripke’s new essences:

16 See, for example, Karl Popper, “A Note on Natural laws and So-called ‘Contrary-to-fact Conditionals’”, Mind, 58 (1949): 62-6. See also his The Logic of Scientific Discovery (London: Hutchinson, 1959), Appendix *X, and “On Subjunctive Conditionals with Impossible Antecedents”, Mind. 68 (1959): 518-20.

since ordinary usage does not distinguish between the literal and the metaphorical (at least it does not draw a sharp line between live and dead metaphors), his detailed discussion of meanings is redundant.

Can essences be taken literally yet as a framework for some metaphysical system? The affirmative answer to this question will elevate a rather primitive idea to the status of a framework for frameworks! Assuming this answer, can that framework be adequate for modern science? Kripke suggests (138) that there is no problem here, but his defense of his suggestion is so thin that one may wonder how such a careful author permits himself to be so breezy. He says there, “scientific discoveries [even when they are] of species essences do not constitute a ‘change of meaning …’ We need not even assume that the biologist’s denial that whales are fish shows his ‘concept of fishhood’ to be different …. he simply corrects …” The concept of fishhood is not changed but corrected, says Kripke, by striking the whale off of the list of fish. Can we conclude, then, that no genuine (essential!) conceptual change occurs in science, only correction? Did Einstein merely “correct” Newton’s concept of gravity? The affirmative answer, one might note, perhaps, is dangerously close to the rejection of fallibilism, contrary to Kripke’s intent; also, it (the affirmative answer) makes his theory of meaning too trivial to require his elaborate discussion: a free and bold use if it (of the idea that there are no scientific changes, only corrections) will make much of his discourse redundant. It is the denial (à la Pierre Duhem) of the very existence of scientific revolutions: it is a methodology of science masked as a theory of ordinary language

The alternative to this view, the admission of some genuine conceptual revolutions, however, makes Kripke’s discussion of changes that are mere corrections quite insufficient, and even beside the point. It is the real changes in science, that entail profound conceptual changes that he should cope with, and yet he cannot: the changes in question lead to the rejection proper, not to the mere correction, of some concepts (that Kripke takes to signify essences). These rejections are not to be taken as the mere change of meaning. The way the concept of fishhood was altered matters little, as the concept is still alive and well. It is the rejection of some concepts in science in a very radical manner that is inconsistent with Kripke’s essentialism. No matter what these revolutions are and where the line between them and mere alterations should be drawn, it is the revolutions that Kripke should address. He does not, and I suggest he cannot.

Kripke seems to notice that this is not all plain sailing: he admits that there is a limit to the laxity he permits (and to the resultant rigidity that he imposes). For example, he suggests (139), the concept of light as used by a blind person may be different from that ordinary. Yet he dismisses this trouble by adding at once, “‘Concept’ here is used non-technically”. Hence, the technical use of the concept ‘concept’ invites a different discourse. Which? And will it allow more laxity or less? And is the discussion of fishhood technical or ordinary? And is it essentially (!) different in ordinary contexts and in technical contexts? I do not know. All in all it is amazing how many ad hoc adjustments his discussion includes. Nor do the adjustments suffice. Elaborating on shifts of reference Kripke admits (163) the matter “requires more apparatus than I have developed here”, without even mentioning the question, how big should a scientific correction of a classification be before it is deemed a shift of meaning? He does discuss the question, though, how much of an object ― such as a table ― is to be replaced before it is deemed altered (51n, 113); he suggests (53) that the object is changed only if its essential features are changed. This is a case of circularity masked by ‘passing the buck’ along a long chain (160).

What exactly is Kripke’s use of the term “essence”? The scholastic usage is not binding. By his usage, to repeat, modality may be confined to some laws or rules endorsed in ordinary parlance, in usage of language by lay people as opposed to technical usage. For my part, I confess I do not know by what rule some parlance is declared ordinary, and I deem futile the search for a universal rule, as in many (ordinary and extraordinary) circumstances verbal usages are not uniquely determined, as is evidenced from the fact that ordinarily people who want to be clear about what they hear may request elaborations or clarifications without complaining about any breech of any rule. I therefore suggest that we do confine modality to some laws or rules, but that their endorsement depends ad hoc on context and on rules that often evolve “ad hoc”. In addition, clearly, in ordinary parlance neither the traditional “modality de dicto” nor the traditional “modality de re” is systematically excluded, yet Kripke overlooks this. It is obvious that the antecedents of some ordinary subjunctive conditionals express “modality de dicto”, as for example, in ‘were cats able to smile, then all children would love them. ‘ Others express modality de re, modalities well within the laws of nature, such as the antecedent in the following statement which is fairly well-known among historians of science: ‘had Galileo performed the experiment of dropping

weights from the Tower of Pisa, then he would have reported the result of that experiment’.

The situation regarding this subjunctive conditional is very complex. There is the admitted historical assertion that we do not possess a report by Galileo to the effect that he had performed the experiment in question. There is the hypothesis that he did make the report nonetheless; it is usually contested but is generally recognized as admissible, namely possible de re ― and even possible for all we know, which is an idea introduced by William Kneale and totally ignored by Kripke.17 There is likewise the generally admitted possibility de re that Galileo has performed the experiment, and there is a disagreement as to whether he actually did. There are thus two contested (possibly true) statements combined into a subjunctive conditional and there is no agreement as to its truth conditions. Thus, considering this subjunctive conditional false is taken by some, but not by all, to be the denial of the declarative conditional associated with it.

It is harder to say whether considering a subjunctive conditional true does or does not ordinarily preclude asserting its associated declarative conditional true as well. This is an open matter not discussed by Kripke. Moreover, at times we express statements of (putative) natural laws by the use of the subjunctive mood; in legal parlance subjunctive conditionals are called “hypotheticals” (following tradition) and they are the prerogative of expert witnesses: hypotheticals in the legal sense are modal de re in the traditional, medieval sense. Example: testifying to the hypothetical statement ‘had the witness seen the victim drink this poison, he would also have seen the victim drop dead on the spot’ amounts to the expert testimony that drinking this poison kills instantly. It is thus very ordinary custom to call expert witnesses to provide hypothetical answers; customarily, eye witnesses are not asked hypothetical questions unless some specific reason is endorsed by the courts ad hoc.

All this shows the great variety of subjunctive conditionals. It stands to reason that even if we confine ourselves to one kind of subjunctive conditionals, then, ordinarily, the truth conditions by which we normally attempt to judge their truth values may (!) depend on context. This is particularly true for expert witnesses, as they are supposed to assert only those subjunctive conditionals which are (a) both highly corroborated and fairly generally endorsed, and (b) relevant to

17 See William Kneale, “Induction and Probability”, (Oxford: Oxford University Press, 1949), 80. See also Popper’s comments on this in his Logic of Scientific Discovery, (London, Hutchinson, 1959): 428-9.

the case under discussion. It is well-known that there can be no general truth conditions for corroboration, much less for their legal recognition -- at least for want of legal uniformity. Moreover, the ordinary restrictions to which expert witnesses are subject do not hold for research, where the greatest latitude of assertion is invited, as researchers are presumed to debate the question of the truth value of any questionable statement (no matter how well corroborated and how generally received). The greater the latitude, the more falls under the rubric of the questionable, as Karl Popper has tirelessly declared and as Niels Bohr noticed when he sought for “crazy” ideas. What context interests Kripke? This translates to the following question. What is the possible truth which he studies? The clue to the answer to this question is the answer to the question, why, if at all, does he think that the subject-matter of his study is important? More specifically, what is the importance of the modal and/or of the subjunctive mood?

Modal logic has a long history, yet its modern beginning is in an error shared by Frege and Russell, no less. The simplest form of the deduction theorem relates valid inferences or entailments with a single premise each to conditionals (or implications as they were called early in the day). In its original version it looked like this.

(A) a entails b, if and only if, if a then b. This is outlawed by Tarski’s rule forbidding the mixture of locution of the meta-language ( “ … entails …”) and locution of the object-language ( “if …, then …”). This can be corrected18 with ease to read as follows:

(B) a entails b, if and only if, (if a then b) is true. This is grammatical. Indeed, it is false. It is refuted by many a conditional statement whose antecedent is false, a counterfactual conditional, as the false antecedent of a conditional seldom entails its consequent. To make (B) above true we may wish to have a stricter sort of conditional, one which is not invariably true when its antecedent is false; it was called “strict implication”. This led to the theory of strict implication, which is a conditional with a modal consequent associated with a declarative conditional: when ‘if x then y’ is replaced by ‘of necessity if x then y’ or ‘if x then of necessity y’ then the result is a strict implication. What is this necessity? This was deemed problematic, and Lewis and Langford presented a series of axiom systems to possibly (!) overcome the trouble,

18 It is well-known that Kripke has offered an alternative to Tarski’s demands; this is not relevant here, as either alternative should clear the confusion at hand.

all of them messy. It turned out that there is no need for these messy axiom systems, as (B) can easily be rescued thus: (Deduction Theorem) a entails b, if and only if, (if a then b) is a tautology. This is why Quine insists that the unhappy conception of modal logic is unredeemable: the initial question is gone leaving no replacement.

The definition of strict implication offered by Lewis and Langford is this:19 (*) a strictly implies b, if and only if, a entails b. In other words, by the deduction theorem, (*) a strictly implies b, if and only if, (if a then b) is a tautology.

This does not close the discussion, however: a tautology is a statement true in all permissible substitution, and only in formal systems is the range of each variable fully determined so that only in formal systems is the list of permissible substitutions definite and unproblematic. The problem remains, how is the question of the validity of inferences in natural languages to be determined? This includes not only ordinary parlance but, as Imre Lakatos has argued, even most of mathematics,20 be it non-formal, semi-formal or quasi-formal. This is a vexing limitation. It was addressed by Karl Popper, who suggests that parts of a discourse be formalized (or even merely quasi-formalized) ad hoc if and when it is found advisable to do so, i.e., only when the limitation raises difficulties or controversy.21 The majority of students of the problem, however, famous among them are Richard Montague, Yehoshua Bar-Hillel, Saul Kripke, and others, prefer a radical approach, and recommend a search for the formal structure of ordinary parlance (on the supposition that it does exist).

The matter hinges on natural languages being open-ended. It is obviously possible to slice a chunk out of a natural language and formalize it. This is how formal mathematics comes into being. The idea developed that mathematics can evolve entirely within a formal language, especially when powerful computers can be employed so as to scan vast areas. Of course, this means that a human will have to sift the few interesting theorems from vast amount of theorems in the machine’s

19 Clarence Irving Lewis and Cooper Harold Langford, Symbolic Logic (New York: Dover Publications, 1932, 2nd ed., 1959): 236 and 241; see also 175. 20 See Imre Lakatos, “Proofs and refutations” (Cambridge: Cambridge UP, 1976). See my review of it in R. S. Cohen, M. W. Wartofsky and P. K. Feyerabend, eds., Essays in Memory of Imre Lakatos, Boston Studies, 39, 1976, reissued in my The Gentle Art of Philosophical Polemics. 21 See K.R. Popper, “Logic Without Assumptions”, Proceedings of the Aristotelian Society, 47 (1947): 251-92.

output. Obviously, this is quite unfeasible.22 So what is usually done is different: mathematical research continues to evolve in the grey area between the formal and the natural, but with an eye on the formal, namely, with the aim of having a fully formal end-result. But a fully formal text is hardly readable. So in the normal run of things mathematics moves to and fro, except that mathematicians today are much more formally attuned than their predecessors, so that even very informal studies tend to be more formal than they used to be. Yet the question is, what is the role of intuition in the growth of mathematics? There is the view that there is no room for intuition in mathematics as it is fully formal. This is not true: the effort put into formalizing it comes after the intuitions are employed and the view is endorsed that the system deserves the effort of putting it formally. In other words, we want our mathematics both informal and formal, which is attained thus far only extremely partially. Moreover, as Lakatos has observed, both intuition and formal standards improve.23 Still, this is no answer to the question, which is still open. How is this applied to the study of natural languages?

Clearly, the formal study of natural languages cannot capture their openness. Clearly, the study cannot be fully formal, so that its formal character is more of a promissory note than a reality. Yet what was said of formal mathematics in the previous paragraph applies equally to the formal study of natural languages, except that unlike mathematics this study is meant to be empirical, at least in part.

Ordinary parlance is open-ended. Assume nonetheless that its formal structure prevails, though it still is under cover. It may then be consistently viewed as both open-ended and formal. (Alternatively, for the sake of this discussion let us deny that it is significantly open-ended.) This already is license to treat it as if it were formal, as if the ranges of its variables were clearly fixed and as if the distinction between formal and non-formal words in the language were possible (regardless of words like ‘bigger’ that naturally are neither, and of Quine-style quasi-definitions that are quasi-tautologies). Under these suppositions sentence forms are easily definable. Tautology and valid inference are then definable in natural languages in the way accepted in the logic of formal systems.

22 See K.R. Popper, “Indeterminism in Quantum Physics and in Classical Physics”, B.J.P.S., 1 (1950): 117-33, 173-95. His example is the arithmetical inequalities that a computer may endlessly generate. This may be an unintended part of the execution of an interesting program. 23 See Lakatos, op cit.

Still, as the distinction between formal and descriptive words is not sharp, the situation is not so simple, and so we do not quite know what is possible de dicto, since it is defined only for fully formal systems. Nonetheless, the deduction theorem can be endorsed at least in part: the inference from ‘a’ to ‘b’ is decidedly invalid if and only if there is a permissible substitution in which ‘a’ is false and ‘b’ is true and it is known to be invalid if such a counter-example is constructed; likewise, the statement ‘if a then b’ is not a tautology if and only if there is a substitution in which ‘a’ is false and ‘b’ is true and it is known to be not a tautology if such a case is constructed. Hence the deduction theorem is putatively true unless questioned ― in line with Popper’s proposal. For another example, in the modus ponens the substitutable items are statements (propositions) and by the definition of the conditional it is clear that no substitution will give rise to a counter-example no matter what exactly the ranges of the variables in the language in question are, at least prior to the introduction of the modal and the subjunctive conditional operators. This supports Popper’s suggestion to take things ad hoc whenever possible, yet here is a question that cannot be treated so ad hoc: can we satisfactorily introduce modality and subjunctive character into the formal grammar of natural languages?

Can this be achieved under the same (contested) assumptions as before? Suppose it can be achieved. What is the advantage involved? It is that strict implication will thereby be quite redundant. The meanings of subjunctive conditionals will also thereby be quite clear if they will be understood to refer to some (existing or virtual) list or ranges of variables as lists of possible substitutes (possible worlds): if and when they are given, then the truth condition of a subjunctive conditional will be that it is true at least in one substitution within the permitted range. Otherwise, (in line with Popper’s proposal) if and when matters are contested, lists may be drawn or negotiated. So much of Kripke’s proposal is acceptable, and all of its merits are salvaged while its essentialism is taken to be context-dependent and so it becomes innocuous (if not even merely a distant relative).

This is still not the end of the story. Some think that strict implications are more attractive then material implications (declarative conditionals): the idea that any conditional is true that has a false antecedent, they say, is counter-intuitive and therefore unsatisfactory. They wish to admit implications only if some connection hold intuitively between their antecedents and consequents. This is doubtful, as the

word “implication” is a technical term shrouded in confusion since its very inception. This is why Quine has put a taboo on the word “implication” and suggested to use instead only “conditional” (for a compound statement) and “entailment” (for the relation between statements). Now this does not end the story for those for whom what they see as the counter-intuitive character of the conditional makes it unredeemable. Of course, some conditional statements are quite odd sounding, but possibly this holds for all logical forms. (Think of an odd conjunction or disjunction, where the two components are total strangers.) Nevertheless, some conditionals are odd sounding because it is not intuitive to admit all conditionals with false antecedents as true. Be it so; this does not mean the rejection of all conditionals whose antecedents and consequents are “unrelated”: “I will eat my hat”, regularly appears as the consequent of conditionals as ways of denying their antecedents; these are all, clearly, “material implications” proper, used in ordinary discourses. Moreover, even in the strictest case we have the logically true yet counter-intuitive conditional whose antecedent is contradictory, quite regardless of the choice of its consequent. By the deduction theorem this amounts to saying that a contradiction entails any assertion whatsoever. Some call this “the paradox of entailment”, but calling it a paradox only expresses discomfort, not the ability to cope with the fact that intuition does not square with logic. Moreover, the validity of any inference from a contradiction means that when contradictions are allowed logic breaks down. The counter-intuitive character of the validity of an inference from a contradiction is akin to the counter-intuitive character of the statement that after the king is removed from the chess board any move is permitted, even though this statement is equivalent to the intuitively unproblematic statement that after check-mate no rules apply any more, namely, that the game is over by then.

Kripke has revived interest in modal logic, Lewis and Langford style, by attempting to define modality by reference to truth in all of the possible sets of statements resulting from all the permissible substitutions in their associated declarative statements. It is not clear what is the advantage of this, and it would be friendly of those who engage the public with the truth-conditions for subjunctive conditionals to tell the public to what end they do so.

Consider truth conditions in any formal language. Narrowly, a statement is true in a system (say, Euclidean) if and only if it is asserted or may be inferred there. Broadly, the idea is to replace the system by the Book of Nature, except that

we do not know if a formal language is at all possible to which it can be fully and adequately translated. (Russell said in his “Mysticism and Logic” that trivially this language is impossible. The final truth, then, becomes an ideal in Kant’s sense.)

Consider the idea of deducibility, then. It is more taxing than truth conditions, as the conditions for the validity of an inference form are meant to exclude the possibility of a counter-example to it. In formal languages this possibility may be replaced by reference to a (real or virtual) comprehensive list of permissible statements, where permissibility is limited by terminology and syntax, where this includes clear delimitation of the range of each variable. If the formal language is taken to be the best and most perfect possible, if all attempts to transcend it are deemed failures in advance, then one might view that language as the limit of our comprehension, as the limits of “our world” (“my world”, said Wittgenstein, following the British empiricist tradition too closely).

The quest for such an ideal language (the essence of language) may have been behind Rudolf Carnap’s celebrated effort to map the whole of the meta-language into the object-language. The inability to succeed fully in this effort was proved by Gödel and by Tarski (as Carnap conceded at once). This is why Popper deemed these the end of the once so popular (Wittgenstein-Neurath-style) philosophical program to erect one sufficient universal language, let alone the question whether it permits metaphysics or not.24 Carnap gave up his initial Wittgensteinian claim that logic has already saddled this language on us. He then tried to erect a language sufficiently perfect to be adequate for science and metaphysics-free, but he conceded that the project was not within his reach, yet he still found constructing such a language attractive. If phenomenalism is admitted, then the program for a metaphysics-free language for science may still have a hope. Except that phenomenalism itself is metaphysics par excellence. This was the thesis of Quine’s celebrated essay on Carnap’s two dogmas, politely misnamed “The Two Dogmas of Empiricism”, one of which was the claim that the only possible view of meaning as context-free is phenomenalism, itself a metaphysics par excellence.25

24 Kripke succeeded where Carnap failed, as he developed a peculiar system that will not be discussed here. As to Neurath and Wittgenstein, see my essays mentioned in previous notes. 25 See W.V. Quine, “A Comment on Agassi’s Remarks”, Zeitschrift für allgemeine Wissenschaftstheorie, 19 (1988): 117-18.

It still is important to notice that the idea of a formal language was retained, except that now one has the freedom to design formal languages as one wishes. The desire to fall back on a natural language as if it were inherently the best formal one, is alluring to many (essentialist) philosophers, such as Chomsky and such as Montague, and is an error: as Bar-Hillel has argued, language is open-ended in its content as well as structure, namely, in its dictionary as well as ranges of variables, quite apart from the Gödel and Tarski limitations, and quite apart from the fact that (parts of) naive set theory enter logic and enrich natural language. (To this one has to add considerations of the paradoxes that bedevil natural languages, he also observed.)26

The exploration of ranges of variables in a language is the exploration of the rules for compiling (virtual) lists of statements and their possible truth values, or their truth-conditions. Here the straight-forward subjunctive conditional comes in handy (and others are ignored): the truth value of the declarative conditional associated with a given straight-forward subjunctive conditional is possibly true, in the sense that by some permissible substitution it becomes true. The theory of possible worlds should comprehend each consistent set of substitutions as a possible world; the theory concerns indirectly the formal structure of a language. * For formal languages this looks as easy as pie. Not so for natural ones: what substitutions in a statement retain its formal identity? At the very least it should be found out (not decided, if essentialism is true) what words in that language are formal. Agree with Tarski and Quine that there is no sharp division in natural languages between formal and descriptive terms, and the program is doomed to limitation. (Similarly, agree that there is no sharp division between naming and describing and the program is likewise doomed. The inability to demarcate sharply between proper names and definite descriptions is clinched by the ambiguity inherent in cases of synonymy (Sabbath = Saturday, Sabbath = Lord ’s Day); one is rigid, the other is not (its value is Friday, Saturday or Sunday). The same holds for all generic names (“Quixote”). Partial synonymy refutes the ascription of formal features to natural languages across the board. Consequently Quine gave up synonymy altogether, but this move is too radical.

26 See Yehoshua Bar-Hillel, “Language and Information” (Reading MA: Addison-Wesley, 1966), Index, Art Machine translation. John R. Wettersten develops wide-ranging implications of the Bar-Hillel view of language as open-ended. See his “The Place of Mario Bunge”, in J. Agassi and R. S. Cohen, eds., Scientific Philosophy Today: Essays in Honor of Mario Bunge, Boston studies, 76, 1982.

This explains much of Kripke’s concern with the question, which word is formal? And is “identity” formal? Also, how can naming demarcated sharply from description?

Possibly my explanation is an error. But then Kripke owes us another. His many disclaimers concerning the meaning of the term “possible worlds” are needed only because he does not announce his program. I conjecture that it is to explore natural language as if it were formal so as to find out the ranges of the various variables in it and delineate formally their permissible substitutions. The question that concerns him is, given any two sentences, can we decide a priori if they are linked by a permissible substitution?

The announcement of a program by a set of disclaimers naturally raises the questions of its completeness and uniqueness. It seems that Kripke takes for granted the existence of a unique and complete execution of his program. It is clear from what is said here that under the conditions that render any part of the program executable, its execution is not unique. Kripke’s discourse does not shed light on this matter: the question is precluded, whether there may be one set or more of rules of substitution required by the program, since it is to explore a language as given and not as it might become. Some sort of essentialism is thus essentially embedded in Kripke’s very rules of rigid designators. Given a language and exploring its structure and vocabulary, while taking care not to transcend it, and while not legislating rules where there are none, the question is, does the language yield to some rules that may capture its essence?

This question is tricky, and rests on the precise meaning of the word “essence”. It was Kripke himself who said,27 given a finite system, more than one set of rules may capture it fully. It would be impossible then to decide which alternative set is its essence. This shows that language is so open-ended that making its designators rigid will not suffice for its closure. Even if a unique essence of language exists, and even if its rules were rendered explicit, even then, the very success of this venture will change the language its users will learn the explicit rule. Proof: grammar books do alter natural languages for better or worse. It is clear that here Kripke was misled by the idea of Wittgenstein that we should clarify language, keep all of meta-linguistic expressions within it, and nevertheless leave it as it was. This does not hold even for formal languages, ever more so for

27 Saul Kripke, “Wittgenstein on Rules and Private Language: an elementary Exposition”, op. cit.

natural ones. That it does not hold for formal languages can be illustrated by Abraham Robinson’s rejection of actual infinity as strictly meaningless, forces one to view all infinity as inherently open-ended, quite contrary to Cantor’s initial intent, of course, and even to that of Hilbert.28 Kripke says nothing about all this.

Take Kripke’s discussion of Kant’s famous dual dichotomy ― of all knowledge to a priori and a posteriori and of all propositions to analytic and synthetic. By allowing oneself to transcend language one can easily question Kant’s claim that there is no a posteriori ground for the view of statements as analytic and the popular view, pace Kant, that there is no a priori proof of any synthetic statement. The question that bothers all students of this matter, at least since Arthur Pap presented it, is the status of statements placing names in their proper ranges, like ‘red is a color’ and ‘cats are animals’.29 Consider a formal language as a system, as a whole, within which Genzen’s theorem holds: all and only tautologies in it are also theorems in it. What, then, is the status of these statements? Our formal systems do improve as a result of the growth of empirical scientific knowledge; crudely put, the growth of knowledge of the range of the possible enlarges and advances the knowledge of the Book of Nature, of the prohibitions in it ( = the putative laws of nature).

To sketch an example or two is not difficult, though the details require much historical research. Plato and Aristotle disagreed as to whether levity is an essential quality different from gravity or a variant of it. Archimedes proved Plato right.30 The same holds for the idea that the pressure that reduces the volume of a given quantity of air is the same as the dilation that increases it.31 The older theories of heat presented two ranges, degrees of heat and of cold; these were collapsed to one, and it was done slowly, and demanded the collapse of a number of other variables, such as that of the compression of air and its dilation under isothermal transfer of heat energy and temperature change in adiabatic expansions. 32

28 See Abraham Robinson, “Formalism 1967”m in Imre Lakatos and Alan Musgrave, eds., Problems in the Philosophy of Mathematics”, (Amsterdam: North Holland, 1967). 29 In Carnap’s final formal system, in Minnesota Studies, 1 (1956), even the fundamental velocity of light may be rendered analytic. 30 See my Towards an Historiography of Science, History and Theory, Beiheft 2 (1963) and facsimile reprint (Middletown CT: Wesleyan University Press, 1967): 55, 110. 31 See my “Who Discovered Boyle’s Law? “, Studies in the History and Philosophy of Science, 8 (1977): 189-250. 32 See Herbert Dingle, The Scientific Adventure, London, 1953.

This lies outside the debates in logic and in science. Kripke approaches the problem not apropos of any formal language but apropos of English as spoken by his intended audiences. Is that legitimate? Possibly not. Definitely not if the domain of his questions is where language is open-ended and if he deems his answers unique.

The exercise which Kripke is engaged in violates the view of language as open-ended. The discussion is of the truth conditions for subjunctive conditionals. The tool he uses is the translation of a given subjunctive conditional to its associated declarative conditional. (His examples are usually contrary-to-fact conditionals.) And he discusses permissible substitutions, where a permissible substitution are meant to retain the identity of the associated declarative conditional through diverse possible worlds. In formal systems, formal words and the ranges of substitutability of descriptive words are not alterable; they are rigid. Also, in formal systems more constraints can be put on permissible substitutability, such as allowing the change of initial conditions while retaining putative natural laws and making them axioms of the system (or implicit definitions, if one insists).

Kripke appeals to the high court of intuition Of course, some philosophers think that something’s having intuitive content is very inconclusive evidence in favor of it. I think it is very heavy evidence in favor of anything, myself. I really don’t know, in a way, what more conclusive evidence one can have about anything [philosophical],33 ultimately speaking. What should be done when intuitions clash? He seems to say, this happens

rarely and temporarily. For, he continues thus: “But, in any event, people who think the notion of accidental property unintuitive have intuitions reversed, I think” (42). I do not know if I read this passage correctly, and I do not know how Kripke’s appeal to intuition squares with his opinion, quoted above, “philosophical theories are in danger of being false”, but we can let it be: suffice it to notice that in this book he employs intuitions as fundamental. He argues, then, that (usually?) proper names are rigid (despite the possible counter-example given there from lists of names, as in the telephone directory; 8n). He argues intuitively

33 My addition of the word “philosophical “ to the text comes to exclude the evidence of the senses, especially highly corroborative (counter-expected) evidence. Kripke may, of course, say that even this is, in the last resort, a matter of judgment and so of intuition. This renders intuition Pickwickian.

that the word ‘cat’ may be replaced by names of other animals and by no other terms.

Kripke defines “essentialism” as the admission of some modality de re, namely, the admission that the Book of Nature contains some prohibitions, all of which may adequately be stated in current language. Traditionally (or as it was baptized by Popper), the word meant, intuiting what makes a thing what it is, e.g., what makes a cat a cat: it may be an essentialism in the sense that animality is essential to felinity, that some of what makes a thing a cat is its being an animal. To make this respectable one has to put the range of “cat” within the range of “animal” or to introduce an adequate second-order axiom to that effect or to rely on intuition (as essentialists do). Once one does this across the board, one violates familiar open-endedness of natural languages: an unintended corollary to essentialism is the statement that natural languages are essentially closed (and the laws of nature are essential definitions within it, as envisaged by Aristotle).

Taken literally, all this suggests that Kripke tacitly claims to be in possession of the true laws of nature. If, on the contrary, it is intuitively admissible that a language and the knowledge of the laws of nature are open-ended, then it is advisable not to rely on intuition here: intuition may admit open-endedness of all sorts. Any doubt about this may be laid to rest by reading Kripke and seeking intuitive exceptions to what he declares an intuitive generality. For, though it is clear that his policy is to avoid general statements, he admits that he makes some general statements. And intuition about them is greatly shaken by modulating subjunctive conditionals far enough from common experience.

Examples. The statement, ‘were I a Wittgenstein fan I would be an analytic philosopher’ is intuitively plausible, whereas ‘if this were Tuesday I would be a king,’ is not ― or at least less so; it is rendered intuitively plausible by conjuring plausibly enough a country with rotating monarchy and a situation in which the person asserting the declarative conditional associated with the subjunctive conditional in question officiates as a monarch every Tuesday: in such a situation, clearly, that conditional is true. Once we allow that, we can declare that mechanical/spooky feline entities are cats proper by conjuring a world in which every animal becomes mechanical/spooky on odd days of the week yet retains its identity through even days of the week by (the intuitively required) space-time continuity. Those who question this should notice that this is how spooks do retain the identity of their former, living selves. Will this exercise convince them? Do all

speakers of a given language share intuition on it and on its limits? Were the answer in the affirmative, then the current controversy amongst taxonomists as to natural kinds would be ruled out a priori. Hence, if the problems raised by Kripke are soluble, they are not uniquely soluble ― unless intuition is improved (and rendered unanimous) and the rules of language are made fully explicit and generally endorsed and rapidly reach closure. 4. Recapitulation The matter at hand is subtle, as it hinges on the license, even requirement, to idealize. What idealization is permissible and what idealization is question-begging and so not very illuminating? We do not know, of course, and there is practically no discussion of the matter. But we can suggest some very broad rules: Idealizations that trivially contradict views of opponents in a critical debate should not be endorsed without some discussion.

Idealizations that concede to opponents in a critical debate should not be declared invalid by these opponents without some discussion. Idealizations that solve problems in trivial manners are to be critically examined: often they solve other problems in unacceptable manners (which is why they are shunned).

To apply these maxims to Kripke we should know who are his opponents. He cites many colleagues, but, in the absence of a general critical outline of the background to his discourse, I for one do not know what are the leading parties in the field. The merit of Kripke’s work is that by entering a marginal area of language, that of the truth conditions of subjunctive conditionals, he manages to raise quite a number of traditional questions, and he also succeeds to throw new light on them. He also has developed tools for the formal and for the semi-formal study of language. He certainly has achieved some results that should be acknowledged and valued even by opponents to the traditions and ideas of the analytic persuasion. For one thing, he has shown the value of the very study of the question, what can we learn about language assuming it is not open-ended without assuming its detailed closures? His study shows what we may think is important for the retention of natural languages as open-ended, and then we may ask, is the benefit higher than the cost?

This raises many questions about the nature (essence?) of natural languages that are not possible to present within Kripke’s system for obvious logical reasons (mainly that he takes language as rigidly given). It also shows yet again that

finding and describing the limits of an exercise (a language game?) renders it easier to follow. Kripke finds it necessary to prevent misunderstandings of his demand for rendering designators rigid, yet the idea is easy to explain once we show that the rigidity of designation is a limitation we impose on our discourse, not on nature. For, this can be done best by showing that this way we limit the scope of the exercise of imagining possible worlds: our exercise is not a part of the possible worlds. (In some possible worlds our language evolves, yet the rigidity of the language within which we study possible worlds precludes study of this option.)

It is clear from Kripke’s discourse that the meaning and truth conditions of a modal/subjunctive-conditional statement depends on the range of the variation of its descriptive terms. This makes the natural ambiguity of this kind of statements clear, and opens the road to improving their meanings by further stipulation, of the kind that is often exercised when the limit of a language is reached. It also shows that the ambiguities of a language can vary, as, for example, the ambiguities in question built into the vagueness of synonymy in general, and, more particularly, of quasi-definitions and of definite descriptions in ordinary parlance. It brings new light to the search for the formalization of ordinary parlance, for associated fields, such as linguistics and artificial intelligence, and it offers a new impetus to Popper’s proposal to raise the formality of language ad hoc.

The nagging question remains: what is the advantage of the exercise? The answer is, first and foremost, in the challenge that it brings to consider ranges of variables less simply than the ways the classical logicians and mathematicians used to do. The difficulties involved in the classical approaches are not new. The first and central question all doctrines of possible worlds (as well as category theory) face is, are the difficulties thereby eliminated, and with what profit? I am not competent to discuss all that. Secondly, the exercise is important as it has proven to be a new and powerful (too powerful, I contend) instrument for the analysis of ordinary parlance and the question of the limit to which ordinary live discourse can be viewed as a part of a formal exercise. It should also help attempts to see the limits of the extent to which essentialism is embedded in ordinary parlance and do the same for other metaphysical systems (atomism and mechanism are the most obvious candidates, but also quasi-Cartesian mind-body dualism: it was Freud who noticed in his The Ego and the Id (1923) that in ordinary parlance at times I possess a body and at times I am my body). Finally, the failure of efforts to push ordinary parlance to the limit displays its open-endedness. We are indebted to all

students of possible-world semantics for that, chiefly to Saul Kripke, but also to the others whose interesting ideas I had no occasion to examine here. 5. Coda. Possibility as such has baffled many philosophers, at least since Parmenides denied it. And he was right in connecting possibility with error. Hence, if he is right, not only the usually observed facts of the perceptible world are parts of a dream, but also all errors! The idea that error is impossible is, indeed, a recurrent theme in the history of philosophy.

A modern writer, remote from any sympathy with Parmenides, struggled with this. Carl G. Hempel has suggested first as a desideratum for an adequate explanation that it be true. He then took it back but this move forced him into instrumentalism. He had no viable suggestion, I think, as he had no alternative to the two methodologies, inductivism and instrumentalism, both of which defy human fallibility. Explanation has to be possibly true for all we know, to use Kneale’s fallibilist suggestion. Putative truth is a better option. Putative truth is unavoidable anyhow, as it enables us to handle idealizations. Objecting to putative truth is thus objecting to all idealizations, including the application of mathematics to the real world. This will help us see the positive and the negative in idealization in general, when viewed as putative truths. The same goes for such seemingly fictitious scientific entities as virtual velocities.

Strictly, the matter hinges much on the question, what is mathematics? This is a huge question; it much pertains to Kripke’s project. He has wisely avoided it for now, but sooner or later he will find it inescapable. This is a clue to how huge his project is.

Lakatos’ philosophy of mathematics was highly innovative. His idea of mathematics as open-ended goes better than logicism with Popper’s own view of logic. Yet he was in error. The possibility of presenting a mathematical system in a gambit that opens a full-fledged axiom system was anathema to Lakatos: “premature formalization” he called it. Yet he was refuted: such gambits are common today in mathematics. Kripke’s study shows that we may view these axiom systems as open-ended, so that Lakatos was in error in thinking that axiomatization is of necessity closure. Hence there are new ways of viewing mathematics as open-ended, and new ways of viewing as rather open-ended

axiomatic systems, formal or semi-formal. This is exciting, and since Kripke has contributed to it most, his work is exciting. 34

34 Early versions were read by I.C. Jarvie of York University, Toronto, and Richard Reiner of Pittsburgh University. Two anonymous referees wrote reports that were demanding but also helpful.