Agency and the Attitudes: Responsibility Through Reasoning
-
Upload
khangminh22 -
Category
Documents
-
view
1 -
download
0
Transcript of Agency and the Attitudes: Responsibility Through Reasoning
Agency and the Attitudes: Responsibility Through Reasoning
Matthew Heeney
Submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy under the Executive Committee
of the Graduate School of Arts and Sciences
COLUMBIA UNIVERSITY
2020
Abstract
Agency and the Attitudes: Responsibility Through Reasoning
Matthew Heeney
Are we morally responsible for what we believe and intend? If so, what is the nature of this
responsibility, and how does it differ from our moral responsibility for our outward bodily deeds?
How is our moral responsibility for belief and intention grounded in mental action? I argue that
we do bear a species of moral responsibility for our beliefs and intentions. But our beliefs and
intentions are nonvoluntary—we neither believe nor intend ‘at will.’ This raises a pressing
question about how we can be legitimately held accountable for the attitudes. Given that we do
not choose our attitudes in the same way we choose to perform ordinary intentional actions, how
do we exercise agency in belief and intention? My answer is that responsibility for the attitudes
is grounded in a fully intentional yet nonvoluntary form of mental action. This is a thinker’s
reasoning to a conclusion in thought (or inferring). Drawing on the work of G.E.M. Anscombe, I
argue that reasoning is active because it is constituted by the very species of self-conscious
practical knowledge as intentional bodily action. This practical knowledge positions a thinker to
answer the justificatory demands that mark our responsibility for the attitudes.
i
Table of Contents
Acknowledgements................................................................................................................. iii
Dedication ................................................................................................................................ v
Introduction.............................................................................................................................. 1
Chapter 1 Responsibility and the Attitudes ............................................................................. 6
Resentment and the attitudes ............................................................................................... 7
The form of attitudinal responsibility ................................................................................ 19
Attitudes and the voluntary ................................................................................................ 35
Attitude-formative mental agency ..................................................................................... 45
Conclusion: Three desiderata ............................................................................................. 58
Chapter 2 The Mark of Mental Action ................................................................................... 60
The criterial question .......................................................................................................... 60
Instrumentalism: bodily action .......................................................................................... 64
Instrumentalism: remembering, imagining ........................................................................ 74
Instrumentalism: deciding, judging ................................................................................... 79
Against instrumentalism .................................................................................................... 85
Two objections ................................................................................................................... 95
Judging and deciding as inference ................................................................................... 100
Conclusion: Two new desiderata ..................................................................................... 103
Chapter 3 Inference and Attitudinal Commitment............................................................... 106
Attitudinal agency as inferring ........................................................................................ 106
The Taking Condition....................................................................................................... 107
The appeal to rules ............................................................................................................114
ii
The appeal to dispositions .................................................................................................116
Taking as commitment ..................................................................................................... 123
Worries and refinements .................................................................................................. 141
Practical deliberating ....................................................................................................... 153
Judging and deciding as inferring .................................................................................... 160
Conclusion ....................................................................................................................... 164
Chapter 4 Inferring as Intentional Agency ........................................................................... 167
The Action Thesis ............................................................................................................. 167
The problem of action and the problem of inference ....................................................... 171
Intentional action and practical knowledge ..................................................................... 176
Inferring and practical self-consciousness ....................................................................... 199
Intentional action as inferring to a conclusion ................................................................. 218
Conclusion ....................................................................................................................... 227
Chapter 5 Responsibility Through Reasoning ..................................................................... 229
Recounting the desiderata ................................................................................................ 229
Desideratum 2: Nonvoluntariness.................................................................................... 237
Desideratum 3: Productivity ............................................................................................ 240
Desideratum 4: Non-instrumentalizing............................................................................ 245
Desideratum 5: The Unity of Action ................................................................................ 247
Desideratum 1: Responsibility-grounding ....................................................................... 259
Conclusion ....................................................................................................................... 269
Appendix: Intention, Constitutive Norms, and the Toxin Puzzle ........................................ 270
Bibliography ........................................................................................................................ 278
iii
Acknowledgements
My first debt of gratitude is to Professor Rovane. Early on in this project, Professor Rovane
helped me to set my sights on normative questions that now anchor my philosophical outlook on
intentional agency. In addition to helping to oversee the project as it (slowly) unfolded, she urged
me to broaden my horizons and reorient the philosophical concerns that framed my original
proposal. I would have been a poorer philosopher had I worked with anyone else.
In addition to guiding the project with helpful instructions, Akeel Bilgrami and Katja Vogt
have encouraged me to take a more capacious approach to these matters than I was initially
inclined to.
Some extremely helpful feedback has improved the dissertation. Here I want to
acknowledge audiences (and organizers) at Columbia University, Florida State University
(especially commentary by Gordon Cooper), Université Paris 1 Panthéon-Sorbonne, and the
University of Helsinki. I extend a special note of thanks to the external examiners on my
dissertation committee, Pamela Hieronymi and Alfred Mele, for their commentary, questions,
and objections.
My work on this dissertation was nourished and sustained by many valuable friendships. I
learned a great deal about how to do philosophy from conversations with Thimo Heisenberg. His
gift of sharp philosophical acumen paired with deep intellectual empathy sets a standard to which
I continue to aspire. Discussions with Thimo and with Jake McNulty brought me to appreciate
philosophical work outside of the very narrow band of interests that I brought to Columbia.
Borhane Blili-Hamelin's warmth and openness encouraged me to come to Columbia and helped
me to get my bearings in the first years of the program. I had many very probing and insightful
discussions with Isabel Kaeslin in bars, cafes, the subway, our respective apartments, outings in
iv
the city, and over the course of long bicycling expeditions outside New York. In various ways,
Isabel helped me to find the deeper cares motivating me to do philosophy—and to write about
them. I continue to admire her passionate zeal for the committed philosophical life.
I was lucky to find lasting philosophical role models in Simon Brown, Nemira Gasiunas-
Kopp, and Jorge Morales. Our reading groups in the philosophy of mind and perception taught
me to set my sights on philosophical ideas that matter to me and to articulate these ideas to
others. (They also taught me how to read a room when talking nonsense.) I continue to look to
these philosophers for inspiration. More recently, Kate Pendoley’s generosity and patience has
helped to improve many of the ideas in these pages by subjecting them to job market-grade
scrutiny. I could not have made these ideas communicable without Kate's, Nemira’s and Simon's
guidance. Their friendship has helped push me over the finish line.
I think back fondly to the working groups we set up to keep from having to write alone.
Katie McIntyre, Borhane Blili-Hamelin, Devin Morse, and Alex Rigas kept me accountable. I am
grateful to these friends for their company.
I acknowledge my family and friends back home in California. My mother, father, sister,
and brother gave incalculable moral support and encouraged me to challenge myself by enrolling
at Columbia. I would be rudderless without their support when the chips are down. The chance to
reconnect with my steadfast friends Jeremy, Ben, and Alex has always made me anticipate trips
home. I thank them for their steady friendship, especially when I am not the best at reaching out.
It is hardest to find a way to thank Yvette. New York was not home until she moved here,
and much of this dissertation was written by her side in our studio apartment. My gratitude to her
extends far beyond the bounds of this dissertation; nevertheless, I dedicate it to her. And no, that
does not mean she has to read it.
Deo gratias.
1
Introduction
Are we morally responsible for what we believe and intend? If so, what is the nature of this
responsibility, and how does it differ from our moral responsibility for our outward bodily deeds?
How is our moral responsibility for belief and intention grounded in what we do? These
questions set the agenda for this dissertation.
I argue that we do bear a species of moral responsibility for our beliefs and intentions. This
responsibility is sourced in rational norms that govern the attitudes. But our beliefs and
intentions are nonvoluntary—we neither believe nor intend ‘at will.’ This raises a pressing
question about how we can be legitimately held accountable for the attitudes. Given that we do
not choose our attitudes in the same way we choose to perform ordinary intentional actions, how
do we exercise agency in belief and intention? My answer is that responsibility for the attitudes
is grounded in a fully intentional yet nonvoluntary form of mental action. This is a thinker’s
reasoning to a conclusion in thought (or inferring). Reasoning to a conclusion, in turn, is a fully
intentional but non-voluntary mental action. These mental actions position an agent to reply to
the distinctive justificatory demands that mark our responsibility for the attitudes.
Accordingly, much of this dissertation is concerned with what it is to infer to a conclusion.
More specifically, it is concerned to issue an action-theoretic account of inferring—an account
that explains why inferring is something that a conscious thinker does, rather than something that
merely happens in a thinker’s mental life. A parallel concern is to better understand how inferring
to a conclusion compares to familiar forms of intentional bodily action. So, I have comparatively
little to say about inferring from the point of view of empirical psychology, epistemology, or
decision theory. My focus is on what makes inferring the distinctive kind of mental episode that
it is, and the grounds on which inferring to a conclusion differs intrinsically from non-inferential
2
mere transitions in thinking. Still, I hope that this action-theoretic focus complements, rather than
departs entirely, from these more traditional philosophical perspectives on inference. For
instance, we can better understand how inferring to a conclusion transmits an internalist species
of epistemic warrant from premise to conclusion by understanding why inferring is active, rather
than passive. At least, that is the hope.
The best way to describe what the dissertation is about and what it argues is to summarize
the chapters to follow.
Chapter 1: Responsibility and the Attitudes argues that we are morally responsible for our
beliefs and intentions. My argument takes off from a remark of P.F. Strawson’s in “Freedom and
Resentment” concerning the ethics of interpersonal rational persuasion. Strawson indicates that
rational persuasion takes place only from within the ‘engaged stance,’ wherein reactive attitudes
of blame, resentment, and indignation are intelligibly directed towards another. I argue that this is
because belief and intention are constituted by rational norms that render a thinker wrong to
persist in an attitude that he recognizes as falling short of them. Attitudinal blame is just such a
charge. I vindicate the legitimacy of attitudinal blame against three challenges: (i) it is
inappropriate to inflict bodily punishment on another on account of his beliefs or intentions; (ii)
many of our beliefs and intentions fall outside the range of specifically moral concern; (iii) we do
not possess voluntary control over our beliefs and intentions. In meeting these challenges, it
emerges that we are answerable for belief and intention, in that we are rightly subject to the
question, “Why do you believe or intend as you do?” and open to a range of reactive attitudes
should our reply fail to satisfy this justificatory demand.
If we are responsible for the attitudes, we require an account of the agency that we exercise
over our beliefs and intentions. I close the chapter with three desiderata for an account of this
attitudinal agency. First, the account must show how the agency at issue works to ground our
responsibility. Second, the account must respect the nonvoluntariness of belief and intention.
3
Third, the account must explain how attitudinal agency is productive of beliefs and intentions. In
the face of these desiderata, I propose that reasoning to a conclusion is a fitting candidate for that
in which attitudinal agency consists. But is reasoning to a conclusion an action?
Chapter 2: The Mark of Mental Action addresses this question. We cannot do so without
discerning the criterion that distinguishes mental actions from mental episodes that merely
happen to us. I explicate and evaluate an instrumentalist answer to this criterial question about
mental action. According to the instrumentalist, mental actions are distinguished by the fact that
an agent can perform mental actions without reliance on distinct causal-instrumental means. I
argue that this instrumentalist reply does not appropriately conceive of the relationship between
reasoning and deciding to act. This is because deciding is not something that can be done
‘directly,’ and yet also not something that is brought about by dint of reliance upon causal-
instrumental means. The upshot is that the instrumentalist cannot provide a stable reply to the
criterial question. I suggest that the relationship between deliberating and deciding is best
understood on the model of the relation between premise and conclusion when a thinker draws
an inference in thought. Inferring to a conclusion seems to mark a form of mental agency that is
prior to the sort at issue for instrumentalism.
This chapter’s course of argument returns two more desiderata on an account of attitudinal
agency, in addition to the three set out at the end of Chapter 1. My fourth desideratum is that the
account must be non-instrumentalizing, in that we cannot construe the relationship between
premise and conclusion in causal-instrumental terms. Inferring is neither a ‘basic action’ nor a
non-basic mental episode an agent acts to bring about. But this fourth desideratum brings in a
fifth. The instrumentalist applies to attitudinal agency and ordinary bodily action the very same
model of intentional action. In this, the instrumentalist respects the unity of action across the
mental and bodily domains. If we wish to part with this model, we must explain how it is that
4
attitudinal agency counts as action on the very same grounds as ordinary intentional bodily
action; we must respect the unity of action. This is our fifth desideratum.
Over the course of the rest of the dissertation I argue that inferring to a conclusion,
understood on its own terms, can satisfy all five of these desiderata. I offer an account of
inferring over Chapters 3–4.
Chapter 3: Inference and Attitudinal Commitment takes up the traditional Taking Condition
on inferring, according to which every inferring involves a thinker’s appreciation that the
premises rationally support the conclusion (Boghossian 2014). The challenge is to specify what
this ‘taking’ consists in without generating a regress of premises (again, à la Lewis Carroll). I
consider rule-based and disposition-based ways of making out this Taking Condition but find
both unsatisfying. To solve these problems, I draw on Akeel Bilgrami’s (2006) and Carol
Rovane’s (1998) irreducibly normative conception of the attitudes of belief and intention. On this
view, what it is to believe that p involves a suite of rational obligations to uphold the truth of p in
further dealings (and analogously for intending to φ and obligations to uphold the goodness or
value in φ-ing). Once we understand the attitudes in irreducibly normative terms, though, we
need not appeal to any separate premise in which a thinker’s taking consists. Rather, a thinker’s
taking consists in a suite of attitudinal commitments.
Chapter 4: Inferring as Intentional Agency draws on Anscombe’s Intention (1957) to explain
why inferring is intentional rational agency. For Anscombe, practical knowledge of what I am
doing and why is constitutive of intentional bodily action, marking the difference between a
series of mere bodily movements and an intelligible order of means directed to an end. I
generalize Anscombe’s approach to the case of inferring. A thinker’s self-conscious appreciation
of what follows from what is constitutive of inferring, distinguishing an intelligible premise-
conclusion order from a mere transition in thinking. By answering the ‘problem of action’ and
the ‘problem of inference’ in the very same way, I show that inferring is intentional agency and
5
that intentional bodily action is but an extension of the agency we exercise in inferring to a
conclusion. Intentional agency consists in a distinctive kind of self-consciousness that is present
equally in inferring and intentionally acting.
Chapter 5: Responsibility Through Reasoning explains why inferring to a conclusion
satisfies all five of our desiderata. I first argue that inferring to a conclusion meets desiderata 2–5,
concerning the shape of attitudinal agency. Inferring is nonvoluntary, productive, and non-
instrumentalizing. As for the unity of action, the argument of Chapter 4 shows how the
phenomenon of inferring to a conclusion is the action-making factor of ordinary intentional
action. In Chapter 5, I approach this point from another angle, by way of Pamela Hieronymi’s
(2009) distinction between two kinds of rational agency, managerial control over the body and
evaluative control over the attitudes. As against this, I argue that both of Hieronymi’s kinds of
agency are sourced in a thinker’s inferring to a conclusion. Lastly, inferring to a conclusion
positions a thinker to satisfy the justificatory demand communicated by the ‘Why?’-question that
marks our answerability for the attitudes of belief and intention. Responsibility for belief and
intention is grounded in a form of rational agency that we exercise in reasoning to a conclusion.
I want to enter a caveat about the dissertation’s overall sweep of argument. Throughout, I
am concerned with intentional agency as it is manifested in mature, rational, and rationally
reflective human adults. This restriction also goes to what I have to say about inferring to a
conclusion—I am interested in notions of inferring that capture a notion of epistemic agency. So,
even if young children and some ‘higher’ animals can be said to intentionally act or engage in
reasoning in some attenuated sense, they are not the paradigm case I am concerned about in this
dissertation.1
1 I thank Alfred Mele for bringing me to make this point.
6
Chapter 1
Responsibility and the Attitudes
As rational moral agents, we bear moral responsibility for our overt intentional actions—those
deeds we knowingly and deliberately bring about by moving our bodies. Of course, for the
purposes of ordinary moral appraisal the class of overt intentional actions comprehends far more
than mere bodily movements. It also includes the events and states of affairs that these bodily
movements help to bring about (e.g., the events that I bring about by setting off a causal chain
that eventuates in the realization of my purposes). This more capacious field of overt intentional
action typically occupies the forefront of our moral concern. And understandably so: it is in the
performance of overt intentional action that an agent’s will is made manifest in the world.
Moreover, it is here that an agent’s will matters most to his fellows. The events and states of
affairs that occupy the forefront of our ordinary moral concern—those which are worthy of moral
praise and blame, as opposed to those which are merely welcome or unwelcome—occur or
obtain in the public arena wherein we press our claims against one another. Accordingly, our
ordinary thought and talk about moral responsibility centers on such overt will-manifesting
intentional actions.
Yet we are morally responsible for more than just overt intentional action. We also bear
moral responsibility for features of our own minds. Specifically, we are morally responsible for
our attitudes of belief and intention. It is easy to overlook this responsibility, for three main
reasons. First, we are reluctant to inflict outright bodily punishment on an agent on account of his
beliefs. Second, there are many beliefs and intentions that fall outside the range of specifically
moral concern. What is the specifically moral significance of my belief that San Francisco is a
city in California, or of my intention to walk to the store after my afternoon appointment? Third,
7
and most importantly, we do not possess the kind of control over our attitudes that we exercise
over our own bodies in the performance of overt intentional actions. My purpose in this chapter
is to show that, despite these factors, there is nevertheless a deep sense in which we are
responsible for our attitudes. This is because our beliefs and intentions are constitutively
regulated by rational norms which agents, qua believers and intenders, are obligated to respect.
To bring this out, I begin with a suggestion from P.F. Strawson. In “Freedom and
Resentment,” Strawson indicates that we cannot reason with another unless we do so from a
certain stance or point of view. Understanding why this is so will reveal our responsibility for
belief and intention. Then, I’ll go on to distinguish several forms of responsibility; clarifying the
form of responsibility appropriate to our attitudes will disarm the worries pertaining to bodily
punishment and moral significance. But the third worry concerning control will stay with us
throughout this study. Unlike overt intentional action, beliefs and intentions are nonvoluntary.
The nonvoluntariness of the attitudes raises a challenging question: how can it be legitimate to
blame another for violating the rational norms governing belief and intention? Taking Angela
Smith’s (2005; 2008) Rational Relations View as a point of departure, I suggest a way forward
that animates the rest of this dissertation.
Resentment and the attitudes
P.F. Strawson’s “Freedom and Resentment” (2003 [1962]) focused philosophical attention
on our interpersonal practices surrounding responsibility. Strawson introduced the notion of a
reactive attitude—an attitude such as gratitude, blame, resentment, indignation, or forgiveness—
that we take to others and to ourselves. These reactive attitudes register a “demand for goodwill
or regard” from others as well as from ourselves (2003: 78). Strawson’s concern is to show that
these attitudes are not suitable candidates for rational justification or for modification in the light
8
of a general metaphysical thesis like determinism. I leave these metaphysical matters to one side.
I want to focus on what Strawson noticed about the objective attitudes that contrast with reactive
attitudes such as resentment or gratitude, and the stances from which each set of attitudes is
appropriate.
According to Strawson, we take up reactive attitudes from a stance of engagement. From
this engaged stance, we interpret a person’s bodily behavior as “intelligible in terms of conscious
purposes” (2003: 88). For instance, it is from the engaged stance that I can understand your arm-
raising as an attempt to get my attention so as to pose a question, as opposed to a mere reflex
movement that does not indicate any meaningful will or purpose. Since it is from the engaged
stance that your “behavior is intelligible in terms of conscious purposes,” it can only be from the
engaged stance that I interpret your behavior as manifesting (what Strawson calls) your quality
of will: your due regard (or undue disregard) for my claims to respect and proper treatment.
Reactive attitudes, Strawson says, are reactions to the quality of will manifested in a person’s
intelligibly purposive behavior. That is why they are appropriate only from within the engaged
stance (2003: 76–77).
This means that the reactive attitudes can be modified, suspended, or abandoned in two
distinct ways. First, it may emerge that I was mistaken to interpret your behavior as a moral
trespass manifesting ill will. In this connection Strawson mentions such familiar pleas as ‘He
didn’t mean to,’ ‘He hadn’t realized,’ or ‘He couldn’t help it’ (2003: 77). In the light of such
pleas, the reactive attitudes of resentment and indignation are modulated accordingly. It is
notable that in such cases it remains appropriate to relate to the other from within the engaged
stance. Only the apparent trespass gets reinterpreted, not the agent who acts from conscious
purposes. Such suspensions, Strawson tells us, “do not suggest that the agent is in any way an
inappropriate object of that kind of demand for goodwill or regard which is reflected in our
ordinary reactive attitudes”; they “suggest instead that the fact of injury was not inconsistent with
9
the agent’s attitude and intentions being just what they demand they should be” (2003: 78). In
this first kind of suspension, I abandon the reactive attitude of resentment in the recognition that
there is, in fact, no moral transgression to react to.
But a second sort of suspension is more far reaching: it renders its target an inappropriate
target for the reactive attitudes taken as a whole. When this second sort of suspension comes into
play, one no longer sees the other as an agent whose conscious purposes are manifested in overt
intentional actions that register (or fail to register) due regard for moral demands. One thus
ceases to take the engaged stance altogether to the other. One takes up, instead, what Strawson
labels the objective stance or the objective attitude. “To adopt the objective attitude to another
human being,” writes Strawson, “is to see him, perhaps, as an object of social policy; as a subject
for what, in a wide range of sense, might be called treatment…to be managed or handled or
cured or trained” (2003: 79). It is simply inappropriate to take one of the reactive attitudes to
such a being. In taking up the objective stance, one realizes that there is no thoroughgoing
“participant relation” possible with the other at all. To recognize this is to recognize that no
reactive attitude is rightly taken to the other.
Taking this objective attitude towards another does not preclude all forms of love. Nor does
it preclude attitudes of mercy or charity. So it is too strong to say that in taking up an objective
attitude we conceive of another as a mindless artifact or mere object. Still, the objective attitude
significantly constrains the range of sensible interpersonal dealings. Since it will be important for
my argument, I quote Strawson at length on this point:
The objective attitude [i.e., the attitudes appropriately taken to another from within the
objective stance] may be emotionally toned in many ways, but not in all ways: it may
include repulsion or fear, it may include pity or even love, though not all kinds of love. But
it cannot include the range of reactive feelings and attitudes which belong to involvement
or participation with others in inter-personal human relationships: it cannot include
10
resentment, gratitude, forgiveness, anger, or the sort of love which two adults can
sometimes be said to feel, reciprocally, for each other. If your attitudes towards someone is
wholly objective, then though you may fight him, you cannot quarrel with him, and though
you may talk to him, even negotiate with him, you cannot reason with him. (2003: 79)
Strawson’s last constraint is significant for our inquiry. From the objective stance we might fight,
talk to, or negotiate with another. But reasoning with another can only be done from within the
engaged stance. This is curious. Why is fighting, talking, and negotiating possible with another
from within the objective stance, but not reasoning? What is special about reasoning with another
that makes it possible only from within the engaged stance? Well, what is it to reason with
another?2
We can pursue this question by way of a simple example. Suppose you and I disagree about
what we ought to do tonight. You want to attend an off-Broadway play. I want to see a movie.
Both of us prefer spending the evening together over going to our separate ways. Since we
cannot see the play and see the movie tonight, I want to conform your will to mine and you want
to conform mine to yours. We both seek to influence the will of the other.
There are two ways you can influence my will that do not count as reasoning. The first is by
treating my will like any other instrument. Suppose you slip me a drug that induces a
predominant desire to attend an off-Broadway play. Once this desire arises in me, my will is sure
to follow suit: I will decide that we attend the play. But this way of bringing my will into accord
with yours bypasses my own rational self-governance. While I do end up with the desire to see
the play, I do not do so by way of anything like a decisive practical reason to do so. And where
2 When Strawson says that from the objective stance “you cannot reason with him,” I take it that he means to
exclude the sort of interpersonal persuasion I am about to describe—cases in which I attempt to convince you of a
conclusion to which I already hold. But there are also cases of reasoning with someone in which both parties pose a
question to which neither has the answer, where the aim is to find an answer that is equally compelling to both of u s,
from a shared set of premises. I treat with the first reading of ‘reasoning with someone.’ But the second sense, too,
must be guided by the aim of self-governance in the way I go on to describe.
11
there is no presentation of reasons, there can be no interpersonal activity of reasoning. In treating
my will like an ordinary instrument, you treat me as an object to be managed or controlled;
thereby, you have opted out the sort of participatory relationship that, Strawson indicates, marks
the engaged stance. To influence my will in the indicated fashion just is to relate to me from the
objective stance. From this stance, you cannot reason with me.
Secondly, you might attempt to influence my will by intimidating me with a threat of
violence. Unlike the desire-inducing drug, your communication of this threat does present a
reason for me to see the play—the reason being that seeing the play will spare me whatever pain
you promise to inflict should I refuse to see it. Here too, though, we do not reason with one
another. Rather, your threat is an attempt at coercion that precludes me from considering the
relative merits or demerits involved in seeing the play. In this, you once more treat my will as
something to be treated or handled, rather than engaged with in joint participatory activity.
These scenarios bring out the fact that reasoning with another is an attempt at persuasion.
Persuasion cannot be the coercive attempt to force the will of another to conform to one’s own.
Rather, if your aim is to persuade me, you must provision reasons that I will find compelling.
These will be reasons that I am positioned to appreciate from my own point of view. If your
attempt at persuasion is successful, then I will decide for these very reasons to pursue the plan of
action for which you’ve made a case. But I must be convinced: I must settle on going to the play
out of the normative force of the reasons that you have presented.3 In this way, persuasion
3 I want to note two complications regarding this simple reflection on persuasion. First, if you present to me reasons
to go the play that I do not or cannot yet recognize as considerations that point up the rightness, goodness,
valuableness, or desirability of going to the pay, then you have further work to do: you need to show me why these
considerations really are reasons to attend the play. These are not reasons for pursuing the relevant first -order plan of
action per se, but metalevel reasons to value something I do not already value. This sort of metalevel exchange often
occurs in persuasion and is a frequent part of making two points of views mesh in the mutual adoption of a shared
plan of action. I leave this complication out of the main text only because it is incidental to the example I am
describing. Second, in the course of our joint deliberation you might present to me reasons that do not bear directly
on going to the play. These might be reasons sourced in our relationship, or in the history of our joint ventures (e.g.,
‘You chose what we did last weekend,’ ‘If you go to the play, I’ll pick up the check at dinner’). While these do not
12
exhibits a due regard for another’s capacity for rational self-governance. Or better: interpersonal
persuasion aims at the other’s rational self-governance.4 I propose that this is a constitutive aim
of interpersonal persuasion, in that it makes persuasion what it is and sets a standard for good
cases of persuasion.
I have been trying to clarify why reasoning with another is a distinctive mode of
interpersonal influence, with an eye to discerning why it can only happen within the engaged
stance. At this point it is worth reiterating that reasoning with another is an attempt to influence
the other—to induce in the other some change according to one’s own purposes (cf. Rovane
1998: 76–77). All we have said so far is that any such attempt cannot count as reasoning with
someone unless it is an attempt to persuade the other. In turn, persuasion constitutively aims at
the other’s self-governance. This does not render interpersonal reasoning any less an attempt to
change another’s will. In our scenario, you are attempting to change my will from being settled
on seeing a movie to being settled on seeing a play. All I have said is that reasoning is distinctive
because it is an attempt at interpersonal influence that works by provisioning reasons that
facilitate, rather than hinder, the other’s rational self-governance.
Sometimes persuasion is seamless. You might communicate to me reasons so compelling
that upon taking them in our wills instantly harmonize on a joint plan of action. In such cases one
merely voices a suggestion and the matter, for both, is settled. (In my relationship, “Pizza first?”
bear directly on going to the play—and are hence impure modes of rational influence (cf. Rovane 1998: 81)—they
nevertheless figure frequently in attempts to ‘sweeten the deal’ to provide incentives for another. I do not think they
are always coercive, but of course can constitute coercion in special circumstances. I take it though that we do not
need necessary and sufficient conditions for an incentive’s counting as coercive to make the present point, namely
that interpersonal reasoning aims at (non-coercive) persuasion. 4 I am tempted at this point to employ Rovane’s phrase agency-regarding relations (1998). As Rovane unpacks this
notion, agency-regarding relations are typified by “pure and open rational influence”—the honest presentation of the
merits strictly of doing what one recommends—that is mindful of the other’s rational point of view (cf. Rovane
1998: 75–85). But as I said in note 2, ordinary persuasion can be impure by offering additional incentives and can
attempt to rationally influence another’s point of view by offering metalevel reasons to value things that the other
does not, before the attempted influence, already care (sufficiently) about. This does nothing to take away from
Rovane’s main point that pure and open rational influence is criterial for personhood.
13
works like this.) I suppose such cases count as reasoning with another. But reasoning with
another frequently involves a deeper level of contention between the two parties. These will be
cases where your aim is to change my will from a plan of action that I endorse for reasons I take
to be decisive. Again, not all cases of reasoning with another involve this dynamic. But our
interpersonal endeavor cannot really be called an instance of interpersonal reasoning if it is not
the type of joint activity that at least admits of this dynamic. And in these more contentious
cases, you must transmit to me reasons that are stronger than the ones currently in my
possession. Here, reasoning with me requires that you (a) transmit to me a practical case that
conflicts with the one I already possess, therein introducing dissonance into my own rational
point of view, and (b) trust me to litigate the issue rationally in coming to the conclusion that best
accords with all of the reasons I possess after the transmission.
This last item (b) is closely related to the constitutive aim of persuasion. To persuade me
from the plan I endorse (going to the movies) to the plan that you endorse (going to the play), I
must assess both plans along with the reasons for each. Then I must settle my will in accord with
the strongest case. Suppose that your reasons for going to the play really are stronger than the
reasons for going to the movies. In trusting me to recognize this, you are trusting me to recognize
that I am rationally constrained, indeed rationally compelled, to decide to go to the play.
(Otherwise you will not persuade me, but merely incentivize me in a concessionary spirit to
accede to your demand.) This is because in the envisaged case I will be flat out irrational to
persist in the predominant desire to go to the movies—and the charge of irrationality must be one
that matters to me. Upon presenting your practical case for going to the play, I will not be
persuaded to your way of thinking unless my coming to decide to go to the play is also a
recognition that going to the movies is unwarranted in the light of the very reasons you have
presented. In these circumstances, the recognition that we ought to go to the play must also be a
recognition that I will be wrong to persist in my former preponderant desire to see the film. It is a
14
hallmark of specifically rational interpersonal influence that I shall be at fault to persist in a plan
that goes against the reasons within my possession.
Furthermore, this must also be a charge that you can level against me. You cannot be said to
be reasoning with me if there is not at least the possibility of faulting me for persisting in a plan
of action that I can be trusted to recognize as unwarranted in the light of all you have brought to
bear on our shared deliberative question. So, reasoning involves the possibility of targeting
another with distinctive patterns of interpersonal criticism. These criticisms will rest on the
charge that I am failing to abide the rational norms that govern the interpersonal practice in
which we are engaged. (I am failing to abide ‘the rules of the game.’) These will be norms of
epistemic and practical rationality, with the weighing of evidence, utility, risk, and the
preferences. Most fundamentally, we cannot reason together if it not possible to fault me for
persisting in a plan that I recognize by my own lights to be unwarranted.
Coming back to Strawson, we can see now just why interpersonal reasoning cannot be done
from the objective stance. The engaged stance, Strawson tells us, is one of participatory
interpersonal involvement. The reactive attitudes taken up from within that stance reflect a deeply
rooted “human commitment to participation in ordinary inter-personal relationships” (2003: 81).
We see now that interpersonal reasoning—the good faith communication of reasons with the
mutual aim of persuading one another—is just such a jointly participatory activity. Reasoning
with another requires both parties to acknowledge and abide rational norms that govern the
attitudes of belief and intention. These norms, together with the reasons proffered by each party,
dictate what each party ought to settle upon. In this way, interpersonal reasoning always allows
for the possibility of thoroughgoing normative criticism according to how well each party abides
the rational norms that govern reasoning. It is this sense of ‘shared rational rules,’ and the
possibility of normative criticism that these rules bring into play, which makes reasoning with
another a participatory joint activity. Engaging in such an interpersonal participatory activity is,
15
Strawson says, simply not possible from the objective stance. We can only do it from the
engaged stance.
Fittingly, a range of reactive attitudes might become appropriate in the course of reasoning
with another. In the course of our interpersonal reasoning you might charge me with bad faith,
obstinateness, equivocation, feigned ignorance, wishful thinking, evasion, lack of charity, or
refusal to face the facts. More seriously still is the charge of outright irrationality—a failure to
resolve a rational conflict that is within my ken and which puts me at odds even with myself.
These are all fully-fledged reactive attitudes in Strawson’s sense. They are reactions to one
agent’s failure to exhibit due regard for the rational norms that govern the interpersonal activity
in which both are mutually engaged. Evidently, an agent exhibits such failure by persisting in an
attitude or set of attitudes that fall short of those demanded by the rational norms that govern our
interpersonal activity.
Now, I have envisioned these reactive attitudes (obstinateness, equivocation, evasion, etc.)
as arising in the course of reasoning with another. But once this is pointed out, it is evident that
such attitudes emerge from the failure to abide the rational norms that govern the sets of attitudes
that constitute each individual’s own rational point of view (cf. Rovane 1998: 19–26; Bilgrami
2006: 212ff). And these norms in turn can be traced to the specific attitudes of belief and
intention. Belief, it is commonly held, is constituted by the norm of truth, so that part of what
makes an attitude a belief and not some other kind of attitude is that one recognizes that one
ought to believe that p if and only if p is true. It is less commonly appreciated that intention, too,
takes a constitutive norm, namely that an agent ought to intend to φ if and only if φ-ing ought to
be done (or that it is not the case that the agent ought not to φ—see Shah 2008). These norms
impose deliberative norms of their own: an agent ought to believe that p only out of reasons that
bear on p’s truth; an agent ought to intend to φ only for reasons that point up the rightness,
goodness, value, or desirability of φ-ing. The norms on belief and on intention agglomerate into
16
more general rational norms that govern how an agent’s beliefs and intentions should relate as a
whole. Some of these norms include the following:
Instrumental Requirement: An agent ought not [Intend to φ and believe that ψ-ing is
necessary to φ-ing and fail to intend to ψ].
Intention Consistency: An agent ought not [Intend to φ and believe that φ-ing is
impossible].5
Many further normative dimensions of the attitudes will be explored in coming chapters. For
now, I only want to point out that the rational norms that govern interpersonal reasoning also
govern intrapersonal reasoning, traceable as they are to an individual agent’s attitudes of belief
and intention. If one cannot form, modify, or abolish one’s own attitudes in the light of such
rational norms, it is hard to see how one could engage in reasoning with another.
The upshot is that I may take reactive attitudes towards myself for failing to believe and
intend in accord with the rational norms that constitutively govern my attitudes. When I change
my mind in the course of reasoning to some new conclusion, I must think myself correct to do so
and at fault for persisting in the attitude I recognize to be unwarranted. Failure to abide such
norms can occasion self-directed reactive attitudes. For instance, one might fault oneself for
backsliding on a firm resolution, for procrastinating in the face of serious work to be done, for
fooling oneself into believing a delusion. All of these reactive attitudes lodge distinctively
normative criticisms to oneself on account of one’s own erroneous beliefs and intentions. This
fact is consonant with Strawson’s own insistence on the close links between other- and self-
directed reactive attitudes—to the ties between the standards according to which we are prepared
to appraise others and those according to which we judge ourselves (2003: 84–85).
5 See Bratman (1987, 2009a). In articulating these norms, I opt for a ‘wide scope’ reading in the manner indicated by
John Broome (2013).
17
I will use the label attitudinal blame for such attitude-targeting reactive attitudes, whether
they are directed to oneself or to one another. Like the more familiar reactive attitudes targeting
overt intentional action, our attitude-targeting reactive practices are rich and varied, with many
contrasting strengths and valences. ‘Attitudinal blame’ is a term intended to gather these diverse
attitude-targeting practices together under a single label. What unites them is that they are
reactions to an agent’s failure to abide the normative oughts sourced in the attitudes of belief and
intention, and the wider norms of inquiry, reasoning, and deliberation that stem from such
attitudinal oughts.
There are further parallels with the more familiar reactive attitudes of resentment,
indignation, forgiveness, and the like. In particular, attitudinal blame can be suspended in two
ways. First, there is a ‘local’ suspension. Here, one ceases to regard an agent as attitudinally
blameworthy in respect of one attitude or a small set of attitudes in his repertoire. But the agent
is still conceived as an appropriate target for attitudinal blame. This local suspension of
attitudinal blame corresponds to the suspension of resentment in the light of such sincere excuses
as ‘He didn’t mean to,’ ‘He hadn’t realized,’ ‘or ‘He couldn’t help it’ (2003: 77). In the case of
attitudinal blame, local suspensions come in the train of such excuses as ‘He didn’t understand
that the beliefs conflict,’ ‘He didn’t realize that his plan could not come off without ψ-ing,’
‘Unknown to him, his goal was in fact impossible to achieve,’ or ‘He didn’t grasp that this
conclusion is implied by what he does believe.’ In the wake of such pleas, we may suspend
attitudinal blame with respect to the specific attitude(s) excused; but we will not cease holding
the agent to the demand to abide the rational norms that govern his attitudes as a whole. (It is
telling that all these excuses appeal to the lack of some sort of self-conscious grasp on the agent’s
part, which will figure centrally in Chapter 4.) Second, there is a ‘global’ suspension of
attitudinal blame. This is to cease to hold the agent to any demand to abide rational norms at all.
This global suspension effects an objective and detached stance towards another, from which (as
18
I have been stressing, and as Strawson concurs) one simply cannot engage in reasoning with the
other.
Strawson was quite right, then, that we reason together only from within the engaged stance.
The ultimate explanation for this fact is that we bear moral responsibility for our attitudes just as
we do for our ordinary overt intentional actions. This moral responsibility is evinced by our
responsibility practices, in particular by the suite of attitude-targeting reactive practices I have
been describing. Such practices register an agent’s fidelity or infidelity to the rational norms that
govern our beliefs, intentions, and proper inquiry. Attitudinal blame charges an agent with
blameworthiness for failing to abide these norms.
I conclude that we bear moral responsibility for our attitudes of belief and intention. Or at
least, our inter- and intrapersonal practices surrounding responsibility indicate that this is so.
As I mentioned at the outset of this section, Strawson claims that these practices are not
candidates for revision in the light of a general metaphysical thesis like determinism. Strawson
says in “Freedom and Resentment” that the engaged stance is too firmly rooted to be dislodged
by such a thesis. I do not think we need to follow Strawson’s general metaphysical point to
appreciate that these firmly lodged interpersonal practices present a presumptive case for our
moral responsibility for the attitudes. But nor do I think we depart from Strawson’s point to seek
an explanation of why it is legitimate to subject another to attitudinal blame. After all, Strawson
himself emphasizes at several points that our reactive attitudes are reactions to the rational
agency of others—to the quality of another’s will as it is manifested in behavior that is
appropriately interpretable as intelligible in the light of conscious purposes. I am tempted to say
that the tie between agency and responsibility is an inextricable nexus recoverable from within
the very engaged point of view that, according to Strawson, does not admit of justification from
some stance outside of its confines.
19
If that is right, then the natural next question to ask is: In virtue of what species of agency
over our attitudes is it legitimate to subject another to attitudinal blame? This is a question that I
shall take up towards the end of this chapter. First, I want to say more about the form of
responsibility that we bear for the attitudes of belief and intention.
The form of attitudinal responsibility
At the outset of this chapter I related three apparent obstacles to the claim that we are
morally responsible for belief and intention. The first is that we do not think it fair to subject
another to bodily punishment on account of his beliefs and intentions. The second is that a wide
range of our beliefs and intentions are morally insignificant, so that it seems overblown to speak
of moral responsibility for e.g. my belief that San Francisco is in California.6 And the third is that
we do not exercise the same kind of control over our attitudes that we do over our overt
intentional actions. In this section, I want to address the first two of these obstacles by clarifying
the form of our moral responsibility for the attitudes.
Let us begin by following Angela Smith (2015) in distinguishing two questions about moral
responsibility:
a. What is it for an agent to be responsible for x?
6 One quick reply to this worry is to point to those beliefs and intentions that do constitute morally significant
trespass. Intentions to do harm fall in this category. And racist, sexist, or otherwise prejudicial beliefs may constitute
morally significant epistemic prejudices. But such examples are of no help for the broader and deeper claims I wish
to forward in this thesis. My claims in this dissertation pertain to the normative standards that constitutively govern
our beliefs and intentions, and about the species of agency that we exercise with respect to such norm-constituted
attitudes. So, what I say about the responsibility we bear for the attitudes must cover such humdrum cases as
ordinary failures of practical rationality, believing something in the face of decisive contrary evidence, and the like.
Only by working with such general failures can I hope to establish something about our responsibility for our own
minds in these ordinary respects.
And in any case, focusing on morally prejudicial beliefs and intentions to do harm will not answer a related
worry: we do not speak of legitimate bodily punishment even for such morally significant attitudinal trespasses as
intentions to do harm or prejudicial beliefs. Intolerant as such prejudicial beliefs may be, intolerant also is the
attempt to bodily punish those who erroneously hold them. And, while the law rightly punishes for attempted
crimes, the mere intention to perform a crime is not yet an attempt and so , I reckon, not justly punishable. I should
like to stay on the right side of these intuitions in this thesis.
20
b. In virtue of what is an agent responsible for x?
Question (a) asks about the nature of responsibility. It asks: in what does an agent’s
responsibility for x consist? What is involved in an agent’s being responsible for something?
Question (b) asks something different. It asks: what conditions must an agent satisfy relative to x
in order to bear responsibility for x? As Smith has it, question (b) asks after the agency-
conditions that one must satisfy to be responsible for x (Smith 2015: 104). Our answers to
questions (a) and (b) are independent of one another. Two theorists may agree about what it is for
an agent to be responsible for x and yet disagree about the agency-conditions that must be
satisfied for this status to obtain. For instance, these theorists might agree about that which
responsibility consists in but disagree about whether, for this to obtain, the agent must have
enjoyed alternative possibilities at the point of action. These theorists agree over question (a) but
disagree over question (b). Similarly, two theorists might agree over question (b) and disagree
over question (a). They might agree (say) that responsibility does not require alternative
possibilities. But they might disagree about the nature of responsibility: about what
responsibility ultimately comes down to.
Since our answers to these questions can diverge in these ways, we may treat them
separately. I will focus in this section on question (a): what is it for an agent to be responsible for
his beliefs and intentions? I shall return to question (b) near the close of this chapter. Following
an important paper by Gary Watson (1996), philosophers have lately distinguished three notions
of moral responsibility: responsibility as accountability, as attributability, or as answerability.7
7 Gary Watson distinguished attributability from accountability and defended the former notion as a legitimate form
of “aretaic appraisal” in his (1996). (See discussion below.) David Shoemaker argues that attributability,
accountability, and answerability are three separable subspecies of responsibility that come apart in certain cases
(2011). Smith (2015) holds that answerability is the basic and sole notion of responsibility. I do not take a stand on
this question. My question is: which notion of these notions captures our responsibility for the attitudes?
21
Strat with accountability. Accountability is a matter of holding an agent responsible. We
hold an agent responsible for x by subjecting him to a range of moral or legal responses,
sometimes including bodily punishment. But accountability need not take this form: “Holding
people responsible,” says Gary Watson, “involves a readiness to respond to them in certain
ways” (1996: 235). Such responses are typified by the Strawsonian reactive attitudes that we
have taken as our point of departure: attitudes such as blame, resentment, indignation,
forgiveness, and the like. These responses involve one party being “responsible to the other for
complying with the demands” that have been breached (1996: 235). So far, this much is familiar:
overt intentional actions are held to normative demands to exhibit due regard for another.
(Attitudes of belief and intention, in contrast, are held to their own cannons of normative
demands—demands to observe the rational norms that govern interpersonal reasoning.
Attitudinal blame is thus best understood as a form of responsibility-as-accountability. I’ll return
to this point.)
Blame (attitudinal and otherwise) is a characteristically forceful mode of appraisal. Because
of this, we feel that it is unfair—and not merely mistaken—to blame someone for something for
which he does not in fact deserve to be blamed. (Even praise can be unearned.) This is especially
so when holding another accountable involves imposing sanctions (such as bodily punishment)
that deprive someone of his autonomy or deprive him of some good. In light of this, Watson
notes that accountability is subject to its own demands, this time sourced in concerns having to
do with the fairness of blaming someone who has apparently committed a significant trespass.
What makes it fair to hold an agent accountable for x? For one thing, it seems that x must be,
in some sense, legitimately put down to that agent. This brings us to Watson’s second face of
responsibility—responsibility as attributability. According to Watson, an agent is responsible in
the attributability sense for x (roughly) when x expresses or manifests commitments that are
constitutive of the agent’s practical identity (or ‘deep self’). One of Watson’s concerns is to show
22
that attributability is a robustly normative form of responsibility, and not a mere stand-in for non-
normative judgments of efficient-causal connection.8 In particular, Watson pressed that holding
an agent accountable is not the only ‘deep form’ of normative assessment: attributability, too,
permits a form of blame.
Key to this latter leg of Watson’s argument is the notion of aretaic appraisal, or appraisal
from an aretaic perspective. As I read Watson, aretaic appraisal is an assessment of the agent’s
evaluative commitments, as they find expression in overt action. To evaluate an action as flowing
from such evaluative commitments is, Watson thinks, to interpret that action as having been
authored by the agent in such a way that it (the action) gives voice to the agent’s “adoption of
some ends among others” (1996: 233). Here is Watson on the point:
if what I do flows from my values and ends, there is a stronger sense in which what I do is
my own: I am committed to them. As declarations of my adopted ends, [such actions]
express what I’m about, my identity as an agent. They can be evaluated in distinctive ways
(not just as welcome or unwelcome) because they themselves are exercises of my
evaluative capacities […] Because aretaic appraisals implicate one’s practical identity, they
have ethical depth in an obvious sense. (1996: 233)
As Watson points out, when my intentional action expresses my identity as an agent, my action is
itself “an exercise of my evaluative capacities.” The action can be attributed to me in the same
deep sense in which my commitments about what is good, right, and worthy of pursuit can be
attributed to me: such actions and such commitments constitute my practical identity—my
standpoint as a reflective being who cares about things, takes on projects, endeavors to execute
plans, and so on. This, I take it, is what Watson means when he says that I author the actions that
express my practical identity (1996: 234).
8 He was principally replying to Susan Wolf (1990).
23
Though such aretaic appraisals constitute a form of blame, Watson stresses a distinction
between blame in the sense of accountability. Accountability-blame involves, we said, reactive
attitudes such as resentment and indignation. But they also involve forms of sanction and
punishment that, on any given instance, you or I might not be entitled to mete out. Though we
lack such a “reactive entitlement,” still we can assess an agent from the aretaic perspective
(1996: 230).9 So, on Watson’s picture we can attribute an action to an agent in such a way that
grounds a deep form of normative criticism, without thereby holding the agent accountable for
that action. This will occur whenever we attribute a faulty action to an agent but do not possess
the right sort of social standing to subject the agent to blame or other forms of sanction. For
instance, a student might confide to me that his partner has behaved in an untoward fashion
towards him; I might empathize with my student and render the appropriate aretaic appraisal of
his partner, but it is nevertheless ‘not my place’ to blame the student’s partner for what has
transpired in their relationship. (Cf. the discussion of ‘shoddiness’ in Watson 1996.) Such cases
do not point to a defect in the very idea of responsibility in the attributability sense; it only
speaks to the fact that attributability and accountability are different faces of responsibility. Fairly
holding an agent to account for an action involves more than merely appreciating the authorship
at work in the action that we attribute to the agent’s evaluative commitments and practical
identity. For instance, it may involve being appropriately related or situated with respect to the
agent; not being situated to the agent in the right way, one may find oneself without the standing
to hold him accountable.
Answerability can be introduced by way of G.E.M Anscombe’s opening discussion of the
mark of intentional action in Intention. (We will treat with Anscombe’s work in much more detail
in Chapter 4.) Anscombe wrote that intentional actions are those “to which a certain sense of the
9 For challenge on this point, see Angela Smith (2008), who argues that attributability can indeed license the full
range of reactions that Watson prefers to keep grouped under accountability -blame.
24
question ‘Why?’ is given application; the sense of course is that the answer, if positive, gives a
reason for acting” (1957: 7). For Anscombe, this question seeks normative rather than
explanatory reasons. It does not ask for (say) a neurophysiological explanation of how it was that
an agent’s body came to move. Rather, it asks for something like a practical case for why the
action was a good, right, rational, choiceworthy, or sensible thing to do (from the agent’s point of
view). Or, the ‘Why?’-question asks for the aim toward which the agent hoped to contribute by
acting as he did.
Anscombe writes that the ‘Why?’-question can be “refused application,” and in these cases
the episode in question is not an intentional action at all (1957: §6). For instance, the ‘Why?’-
question is refused application when the respondent professes ignorance about what he was
(apparently) doing. When an agent replies that “I didn’t know that I was doing that,” it simply
makes no sense to hold that agent to an expectation to summon a practical case for the action.
(This is the reason why Anscombe took a certain sort of self-knowledge of what one is doing as
constitutive of intentional action. Again, I will spend a great deal of Chapter 4 on this issue.)
Though Anscombe does not explicitly mention the point, it seems that the ‘Why?’-question
must be refused application when it is asked of an (apparent) agent who is normatively
incompetent and cannot appreciate the normative significance of reasons. It simply does not
make sense to ask such an agent to provide the reasons that justify his conduct. Instead, it seems
that the ‘Why?’-question in such cases can only ask after causal or explanatory reasons that offer
an “objective” explanation of why a certain episode occurred. This points up a key point. The
‘Why?’-question presumes not only that the event in question can be justified, but additionally
that the agent is capable of appreciating this justification and acting from it. Again, this will be a
major feature of Chapters 3–4.
However, Anscombe notes that the ‘Why?’-question is not refused application by the reply
that the agent had no particular reason for acting. She writes, “The question is not refused
25
application because the answer to it says that there is no reason, any more than the question how
much money I have in my pocket is refused application by the answer ‘None’” (1957: 25-26).
Most importantly, the ‘Why?’-question is not refused application when the respondent issues
a poor response to the question. A poor response to the ‘Why?’-question is one that does not
offer a (sufficiently) good practical case for the action in question. This will be the case when the
agent’s reply to the ‘Why?’-question fails to justify the action, elucidate its point, or demonstrate
something valuable or choiceworthy in the prospect of its performance. When this is the case, the
agent is open to a distinctive brand of normative criticism. Depending upon the circumstances,
the agent who fails to satisfy the justificatory demand can be blamed, held accountable, and
subjected to the reactive attitudes. In a broader range of non-moral cases, we may simply fault
the agent for his irrationality, which, as I have stressed in connection with the attitudes, is no less
a robustly normative form of criticism.
We may now grasp how the applicability of Anscombe’s ‘Why?’-question constitutes a form
of responsibility. Here I can do no better than Angela Smith’s recent formulation: when an agent
is answerable for something, she stands as “an appropriate target, in principle, for requests for
justification regarding that thing and […] she is eligible, in principle, for a variety of moral
responses depending upon how well or poorly she meets this justificatory request” (2015: 103).
Smith helpfully brings out the dual character of answerability in this passage. To be answerable
for x is to stand in a relation that permits intelligible justificatory demands with respect to x and
to be open to moral (or otherwise normatively significant) responses depending on how well the
justificatory demand is satisfied.
At this point it is important to ward off two potential misunderstandings. The first bears
directly on issues of moral significance with which I started this section. Smith and Pamela
Hieronymi (2014) both argue that we are answerable for all of our intentional actions, and not
only those that take on a special moral significance. Whether a moral response is called for has to
26
do with the specific action at issue and the circumstances in which it was performed. It might
also have to do with the “reactive entitlements” that we enjoy relative to the agent whose action
we are appraising. To say that we are morally responsible for all of our intentional actions is not
to say that all of our actions take on a specifically moral significance. It is rather to say that we
are answerable for morally significant actions and insignificant actions alike.10
Secondly, we should not we be misled by the interpersonal slant to the foregoing
characterizations of answerability. What matters is that an agent is the intelligible and
appropriate source of reasons for the action, not that there is an interlocutor in the wings who is
prepared to pose the ‘Why?’-question. We may conceive of the interlocutor as a merely
hypothetical device to illustrate the notion of answerability.
Having distinguished three forms of responsibility, how do they relate to one another? My
claim is that answerability is a pervasive form of responsibility: if an agent is responsible in any
sense for x, then he is at least answerable for x. Now, T.M. Scanlon (1998) and Angela Smith
(2015) stake out a stronger claim: that answerability is the sole form of responsibility,
encompassing both attributability and accountability. But this stronger claim is not necessary for
our purposes. It will be sufficient to show how answerability, as the pervasive form of
responsibility, constitutes a fitting link between the sort of aretaic appraisals that mark
attributability and the reactive attitudes we take to another when, in fairly holding him
responsible, we deem him accountable for what he has done.
Take attributability first. An action cannot be attributable to an agent unless it is also one for
which the agent is answerable. To attribute an action to an agent is to interpret it as manifesting
the evaluative commitments that comprise his practical identity. These evaluative commitments
are in turn normally held for reasons: it is only insofar as these evaluative commitments manifest
10 Pace McKenna (2012: 17). I return to this point below.
27
an agent’s normative competence in treating with reasons that they can function to constitute his
practical identity at all. And so, to act out of these evaluative commitments, to express them in
action, is ipso facto to be in a position to justify the action by way of the reasons that bear on the
evaluative commitments from which the action issues. To put the point another way: if we were
to attribute to an agent an action for which he is not answerable, the normative dimension of
attributability simply drops out of the picture. We could say at most that the action was produced
and controlled by psychological states and occurrences within the agent. But the special sort of
normative or “ethical depth” that Watson claims for the notion of attributability would be
downgraded (as it were) to mere efficient-causal transactions. To remain a robustly normative
notion of responsibility, then, attributability for x must entail answerability for x.
Against this line of thought, David Shoemaker argues that cases of irrationality illustrate the
possibility of attributability without answerability (2011: 606–615). According to Shoemaker, the
agent’s irrationality is indeed properly attributable to the agent, as we may render aretaic
appraisals of the agent on account of his irrationality. But, he claims, the irrational agent is not
answerable for his irrationality, since it is implausible to hold this irrationality is an expression of
his evaluative commitments. Shoemaker rightly deems this an implausible result: we do take
agents to be responsible for their irrationality, and our theory of responsibility ought to reflect
this commonsense datum.
In reply, notice that an agent who is overall irrational is not thereby an inappropriate target
of the ‘Why?’-question that marks answerability. Irrationality is not nonrationality. Posing the
‘Why?’-question to a nonrational agent—one who does not possess a minimum degree of
competence in taking in and appreciating the normative force of reasons—is a pointless exercise.
The robust justificatory demand that the ‘Why?’-question communicates simply falls to the
ground when the recipient is nonrational. This is different from the case of one who does possess
this sort of competence, but who in a given instance exercises this competence poorly. It very
28
much is to the point to ask an agent to justify outright irrationality, as such cases constitute
failures by the agent’s own lights in what to do, what to intend, or what to believe. In contrast to
a nonrational agent, an irrational agent is nevertheless an apt target for such demands for
justification.11
Now consider the connection between accountability and answerability: an agent cannot
fairly be held accountable for x unless he is answerable for x. If an agent is not the appropriate
target for demands for justification, then it cannot be fair or appropriate to direct blame,
resentment, or indignation to that agent on account of what he has done. To be sure, perhaps
some form of instrumental sanction might be appropriate—say, to keep the relevant subject from
unduly harming others. But I take it for granted that this is not what we are normally up to in
taking up a reactive attitude towards another. After all, part of the motivation for centering
discussion on the reactive attitudes is precisely that they register and embody ‘deep’ forms of
normative appraisal of an agent’s conduct. Genuine accountability, then, is appropriate only
when the agent in question is answerable for the trespass in question, and furthermore when the
practical case that the agent provides (or could provide) does not justify the action in question.
If this is correct, then answerability is the pervasive form of responsibility. An agent can
neither be held accountable nor attributed an action for which he is not answerable. This will
allow us to understand answerability as a common factor in all appraisals of responsibility, as the
link between deeming an attitude properly attributable to an agent and holding him accountable
for the attitude by subjecting him to attitudinal blame. Let us now apply these lessons to the
attitudes more explicitly.
11 This line of reply differs from Angela Smith’s own reply (Smith 2012: §II). Smith construes responsibility for
irrationality as reducible to responsibility for the separate attitudes that constitute the irrationality. But our reply
better accommodates Shoemaker’s point that we think of agents as responsible for their irrationality per se.
29
If we are responsible in any sense for our attitudes of belief and intention, we must at least
be answerable for them. We are answerable for our beliefs, intentions, and for the attitudinal
complexes resulting from their combination. This only means that, as Smith has it, we are
“appropriate targets, in principle, for requests for justification regarding” our beliefs and
intentions, and “eligible, in principle, for a variety of moral responses depending upon how well
or poorly [we] meet this justificatory request” (2015: 103).12 Of course, Smith’s formulation
speaks of moral responses. But we have already seen that our responsibility practices are
equipped with a range of rational responses in the guise of attitudinal blame. Answerability for
the attitudes, then, means that one is eligible for attitude-targeting reactive practices such as
attitudinal blame according to how well one discharges the justificatory demand embedded in the
‘Why?’-question when it requests grounds for one’s attitudes. In the case of the attitudes, this
demand is discharged when an agent can show that he is on the right side of the rational norms
that govern beliefs and intentions. In this way, answerability for the attitudes—just like
answerability for overt intentional action—is a robustly normative conception of responsibility,
permitting a robustly normative sort of criticism, albeit of a non-moral sort.
To bring this out, let us follow Anscombe in noting the cases where the ‘Why?’-question
finds and fails to find application. First, the ‘Why?’-question finds application even in cases
where the agent’s response is poor. Poor responses are those that fail to justify the belief or
intention at issue, or which fail to obviate some other rational criticism. Suppose you ask me why
I believe that I will pass my dissertation defense, and I reply that having this belief helps me to
sleep at night. If such a response were possible, you might rightly take up the reactive attitude of
attitudinal blame, faulting me for believing (or alleging to believe) something for reasons that do
not at all bear on the truth of what is believed. Or, if you ask me why I intend to φ even in the
12 Smith herself agrees that we bear answerability for the attitudes.
30
belief that φ-ing is not something I can do, you can fault me for an inconsistency in my plan. The
point to notice here is that the justificatory demand embedded within the ‘Why?’-question is not
undermined by my poor response. If this were so—if a poor response meant that the ‘Why?’-
question is not applicable after all—then attitudinal blame would be impossible, and
answerability would amount to a non-normative notion after all. The possibility of a poor reply to
the ‘Why?’-question is what gives to answerability its robust normativity.
For similar reasons, the ‘Why?’-question is given application when the agent replies, “No
reason.” Suppose I reply to the ‘Why?’-question by saying, “I do not believe p for any particular
reason—it’s just something I’ve always held to be true.” This reply will surely strike the
questioner as unsatisfying. But what is important is that this reply, unsatisfying though it is,
shows the ‘Why?’-question to enjoy application. Even though I believe p for no particular
reason, it is nevertheless true that, just because p is my belief, I am related to it in such a way that
renders me an intelligible and appropriate target for a demand for reasons that bear upon p’s
truth. The fact that I cannot summon those reasons does not make me an unintelligible and
inappropriate target; it only shows that I fail to satisfy the justificatory demand involved in
believing. Here we find another point of similarity with respect to intentional actions. We are
related to our own actions in such a way that renders us open to justificatory demands in relation
to their performance. Acting for no reason does not render that demand misplaced or wrongly
directed. At most, it can render the agent’s reply unsatisfying.
It will occur to the reader that there is a significant difference between believing that p for no
reason and acting for no reason: the former, but not the latter, violates a rational requirement.
While we are rationally required not to believe that p in the absence of convincing evidence that
p, no such demand precludes intentional actions undertaken with no reason in mind—so long as
the action itself is not one that is prohibited in the circumstances. But this does not count against
the analogy that I wish to draw. It shows only that “No reason” can, in some cases, be a
31
satisfactory reply to the ‘Why?’-question when it is asked about an agent’s intentional action, but
that “No reason” is always an unsatisfactory reply to the ‘Why?’-question when asked about an
agent’s belief. This is a difference in the rational norms that govern acting and believing, not a
difference in the relation that an agent bears to his actions as opposed to the relation he bears to
his beliefs. That relation, I suggest, is the same in both cases.
To say that we are answerable (and so responsible) for our attitudes, then, is not to say that
an agent bears responsibility only for the beliefs and intentions for which he possesses a
sufficient justification. We are responsible, rather, for all of the attitudes in our repertoire that
admit of or are susceptible to justification.
To be sure, some of our attitudes are not susceptible to justification. Compulsive desires, for
instance, might ‘have a grip’ on one even in the knowledge that the object of such desires are not
to be pursued. In such case, the ‘Why?’-question fails to find application. If I reply to “Why do
you desire that?” by citing a nonrational compulsion (such as a neurological condition), then I
have not offered grounds, but an explanation for why I happen to have the desire that I do.
Recall: Anscombe tells us that intentional actions are those “to which a certain sense of the
question ‘Why?’ is given application; the sense of course is that the answer, if positive, gives a
reason for acting.” Giving a set of merely explanatory reasons for my bodily movement is not a
way of answering the ‘Why?’-question, then, but a way of denying its application (since merely
explanatory reasons are not reasons for the performance of an action). Similarly, in the case of
the attitudes the sense that gives the ‘Why?’-question application is that in which the answer, if
positive, furnishes a normative justification for what is believed or for what is intended. If the
answer to the ‘Why?’-question does not furnish such grounds, then the ‘Why?’-question does not
find application.
When we interpret our responsibility for the attitudes as answerability for what we believe
and intend, the link between answerability and the other forms of responsibility comes into better
32
focus. I have already argued that both accountability and attributability entail answerability. But
we have also discovered grounds for the claim that answerability also entails attributability. I am
not answerable for that which cannot be attributed to me in the robust normative sense that
Watson describes. This is because if there is some item that is not attributable to me—that does
not indicate my ‘practical identity’ in a way that permits aretaic appraisal—then I am simply not
an appropriate target for requests for justification of the item in question. ‘Why?’-question
communicates a demand for justification that presumes attributability. If the ‘Why?’-question is
directed to me in respect of something that is not attributable to me, it seems I am within my
rights to deny the question force (why should I be expected to justify that?). With respect to
attributability and answerability, then, something stronger than my claim of pervasiveness holds:
an agent is answerable for x if and only if x is attributable to her.13
This certainly holds in the case of belief and intention. Answerability for the belief that p or
intention to φ means being the appropriate target for justification for p or for φ-ing, respectively.
And if this ‘Why?’-question finds application, then the attitude can be attributed to me from the
purview of aretaic appraisal: it can be seen as constituting my epistemic or practical identity as a
rationally self-governing agent outfitted with the normative competence to appreciate the
normative force of reasons. In contrast, when the ‘Why?’-question fails to find application—
when my reply is “But I don’t believe that,” or when my attitude is simply impervious or not
susceptible to justification—the attitude cannot be attributed to me in a way that permits deep
forms of normatively significant aretaic appraisal. In such cases, the attitude is just one more fact
about me, alongside my height, hair color, or place of birth.
A similar link holds between answerability for an attitude and accountability for it.
Attitudinal blame is taken up in response to—or in anticipated response to—a poor response to
13 I thank Alfred Mele for requesting more clarity on this point.
33
the ‘Why?’-question. On account of its normative depth, attitudinal blame (qua blame) carries a
characteristic force or ‘sting’ that renders it unfair to direct to undeserving agents. One can also
lack the proper standing to direct attitudinal blame to another. All of this is just to say that
attitudinal blame is a kind of accountability, in Watson’s sense. Subjecting another to attitudinal
blame is a way of holding an agent accountable to the rational norms that constitutively govern
beliefs, intentions, and combinations thereof. The link between answerability for an attitude and
accountability for it is thus clear. To hold an agent accountable for an attitude by subjecting him
to attitudinal blame is to charge the agent with a poor response to the ‘Why?’-question. And a
poor response to the ‘Why?’-question, in turn, just is a failure to evince a proper regard for the
constitutive norms that govern the attitudes. So, while it is possible for an agent to be attributable
and answerable but not accountable for a belief or intention, typically these three forms of
responsibility come together. One place where they come together, as we have seen, is in the
context of reasoning together to a shared conclusion.
Let us return at last to the worries I announced at the beginning of the section. The first
worry is that a wide range of our attitudes fall short of moral significance. My belief that San
Francisco is a Californian city, or my intention to walk to the store this afternoon, hardly rise to
matters of serious moral concern. But this only means that the attitude-targeting reactive
practices by which we hold one another accountable for belief and intention are frequently not
explicitly moral responses. As we have shown with ways that the ‘Why?’-question is applicable
even when the agent has a poor response, attitudinal blame even for ‘merely’ rational failings is
nevertheless a fully normative way of faulting an agent for what he believes or intends. It is a
way of lodging criticism that carries a characteristic ‘force’ or ‘bite.’ This matters for the worry
with which we began this section. It allows us to see why we are responsible for a great many of
our attitudes—namely, all of those that admit of, require, or are susceptible to justification—and
not only those that have an explicitly moral significance.
34
We can also see why we are answerable for the attitudes despite the fact that we do not
bodily punish those who believe or intend wrongly. This refusal to administer bodily punishment
is not due to the fact that we are not responsible for beliefs and intentions. It is rather put down to
the fact that our accountability-practices are keyed to the norms to which one is answerable. The
violation of rational norms on believing and intending certainly place an agent at fault. But they
do not constitute a moral trespass worthy of bodily punishment or sanction. We hold one another
accountable to rational norms by expecting justifications when appropriate and by placing
burdens on others to furnish those justifications. Our refusal to punish those who violate rational
norms says something about the norms to which our justificatory demands are anchored, not to
the presence or absence of responsibility.
Watson, recall, notes something else about responsibility in the sense of accountability. He
writes that, because blame carries a characteristic force, it must be fair or reasonable to blame
another. According to Watson, it is “unfair to impose sanctions upon people unless they have a
reasonable opportunity to avoid incurring them” (1996: 237). Attitudinal blame, even if it is not
a form of sanction, still counts as a form of accountability. As a form of accountability, blame
can be deserved or undeserved, and Watson is lucid on the point that we must exercise some
form of agency over our attitudes in order to be fairly accountable for them. This leads to the
third factor occluding moral responsibility for the attitudes: the fact that we do not control our
beliefs and intentions in the same ways we exercise control over overt intentional action. We
must understand just what this lack of control consists in, and why it constitutes a challenge to
attitudinal blame.
35
Attitudes and the voluntary
Our overt intentional actions are voluntary. The voluntariness of overt intentional action
provides a satisfying legitimation of such reactive attitudes as blame, resentment, and
indignation. Our attitudes of belief and intention, in contrast, are not voluntary. That is to say, we
do not have voluntary control over what we believe and what we intend. This is why we are
reluctant to ascribe moral responsibility for the attitudes: if we do not possess voluntary control
over the attitudes, how can it be legitimate to direct attitudinal blame to another (or even to
oneself) on account of what one believes or intends? A satisfying answer to this question is the
task of the dissertation at large. In this section I say what ‘voluntariness’ amounts to and why it
matters for responsibility for overt intentional action. Then I will say why our attitudes, while not
voluntary, are nevertheless not involuntary.
I have two senses of ‘voluntary’ in mind.14 In the first reason-based sense, an action is
voluntary if it can be done for any reason that recommends its performance. Our overt intentional
actions are certainly voluntary in just this sense. For any overt intentional action-type I am able
to perform, I can token an instance of that type for any reason that points up the rightness,
goodness, desirability, or value of doing just that thing. Ditto for the events and states of affairs
that I am able to act in order to realize. Many of these reasons will be instrumental ‘further’
reasons: they will count in favor of raising my arm in virtue of the right, good, desirable, or
valuable consequences that raising my arm will go on to bring about. But the reasons-for-which I
raise my arm needn’t be further instrumental reasons. For I may raise my arms just for the very
sake of doing it, for the intrinsic joy I take in raising my arm. So, this first sense of ‘voluntary’
links the performance of an action with considerations that the agent takes to recommend its
14 Cf. Hieronymi (2006, 2009). My terms ‘reason-based sense’ and ‘will-based sense’ come from McHugh (2011),
who makes the same distinction I have in mind.
36
performance. Voluntariness in the reason-based sense underscores that an action-type can be
done for any reason that recommends its performance. The reason-based sense of voluntariness
concerns the plurality of reasons for some given overt intentional action, rather than the
connection between reason and action per se.
This link permits a form of appraisal that carries an obvious ethical significance. When an
action is voluntary in this first reason-based sense, we can interpret the performance of the action
as an expression or manifestation of the agent’s evaluative capacities and ability to set ends for
himself (cf. Watson 1975, 1996). Thus, voluntary actions are those that are performed in the light
of, and so express, the agent’s conception of the good (or other ethically relevant appraisals of
value). Because of this connection, appraising an agent’s actions also appraises the agent himself.
This is because, as we have seen in connection with Watson’s discussion of attributability, our
appraisals of value articulate a certain practical standpoint, a ‘place to stand,’ from which we
intervene in the world in order to change it according to our purposes. When action is not
voluntary there is a question about how an overt intentional action can express the kind of quality
of will that, Strawson tells us, our reactive attitudes are primarily reactions to.
I turn to a second sense of ‘voluntary.’ In this second will-based sense, an action is voluntary
if it is brought about in the right way by a mental state or episode that represents the very action
in question. This second sense of ‘voluntary’ is associated with the concept of willing, as when
we say that an agent wills to φ or that φ-ing was, at the point of action, the agent’s will. A
paradigmatic type of voluntariness in this second will-based sense is manifested when an agent
acts from a prior intention. A great many overt intentional actions are like this. In the morning I
might form the intention to walk to the café in the afternoon; come the afternoon, I execute this
intention just by doing the very thing that my intention represents.15
15 This will-based sense of voluntariness roughly corresponds to Hieronymi’s notion of managerial control (cf.
Hieronymi 2006; 2009). I discuss this notion more fully in Chapter 5 in connection with the unity of action.
37
Whereas the reason-based sense of ‘voluntary’ carries an obvious ethical significance, this
will-based sense of ‘voluntary’ carries an obvious action-theoretic significance. It is frequently
because a bodily movement is voluntary in this second sense that it counts as an intentional
action at all. Intentional bodily actions are those bodily performances by which an agent brings
the world into conformity with his will (cf. Hieronymi 2009: 140ff.) When a performance is
voluntary in the will-based sense, its performance is a deliberate action—something that the
agent means to do. It is thus a will-manifesting intentional action, rather than a mere event.
Furthermore, the will-based sense of voluntariness grants us a commonsense notion of control
over an event or state of affairs. On at least one notion of intentional control, it consists in the
degree to which the agent can make the world correspond to the practical representations that he
has of it (cf. Shepherd 2014). I exert control over my body just to the extent that my bodily
movements correspond to my intentions to move my body in the relevant ways.
Correspondingly, I fail to control my body to the extent that they occur in the absence of my
intentions or even in outright defiance of my intentions. This notion of control extends to my
immediate environment: I exercise control over my immediate just to the extent that I can bring
that environment into accord with my intentions for how the environment ought to be. In general,
voluntary control is a matter of bringing the world into accord with my mind.
These two senses of ‘voluntary’ throw moral responsibility for overt intentional action into
sharper relief. When an overt intentional action is voluntary in both the reason- and will-based
senses, it is a deliberate performance that expresses the agent’s evaluative conception of the
good. Thereby, the quality of the agent’s will—the regard or disregard that the agent displays for
his fellows—can be the target of such reactive attitudes as resentment, blame, and indignation.
These twin senses of voluntariness, I am now suggesting, serve as conditions for the possibility
of an agent’s quality of will manifesting in outward behavior. They are fundamental for making
the reactive attitudes intelligible, at least when these attitudes concern “actions which confer
38
benefits or inflict injuries over and above any conferred or inflicted by the mere manifestation of
attitude and intention themselves,” as Strawson has it (2003: 76).
In light of this intimate connection between voluntariness and moral responsibility, it is
significant that our attitudes of belief and intention are voluntary in neither the reason- nor the
will-based sense.
Begin with the reason-based sense. I cannot believe that p or intend to φ for any reason that
recommends having the belief that p or the intention to φ. Consider Kavka’s (1983) well-known
toxin puzzle (cf. Pink 1996: 147ff). Kavka asks us to imagine an eccentric billionaire who on
Monday offers you $1 million to intend to drink a mild toxin on Wednesday. Upon verification
that you possess the intention on Tuesday to (Drink the toxin on Wednesday), the billionaire will
deposit the reward into your account. Of course, the mild toxin will have no lasting ill effects.
But such effects are not to the point, anyway: the billionaire’s offer does not require that you
drink the toxin. You stand to gain $1 million just by intending on Tuesday to drink it on
Wednesday. Earning this reward seems the easiest thing to do in the world: just form the
intention today to drink the toxin on Wednesday. And yet, you cannot instill in yourself the
intention to drink the toxin for the reason that you stand to gain a financial reward by having the
intention. Your intentions register a rational response to reasons that (you take to) recommend
performing the intended action. They do not register a rational response to reasons that
recommend merely having the intention to act, even though in this case there is a compelling
reason to have the intention in question. Notice that the upshot of this puzzle is not that we are
barred from forming the intention to φ for instrumental reasons. I certainly can form the intention
39
to φ out of an appreciation of instrumental reasons; but these will typically be instrumental
reasons to φ, rather than instrumental reasons solely for having the intention to φ.16
Kavka’s toxin puzzle generalizes to the case of belief. Suppose I offer you $1 million to
possess the belief that Michael Dukakis won the presidential election in 1988. You will have
great trouble earning this reward, though. Your belief about who won the election in ’88 is
responsive to evidence that bears on who in fact won the election, and not responsive to the
benefits to be gained by having a given belief about who won the election. Since your belief
about the election is not responsive to Kavka-esque practical, you cannot instill in yourself a
belief that Michael Dukakis won in 1988—even for a million dollars. You can come to believe
that Michael Dukakis won the 1988 election only in the light of reasons that (you take to) bear
positively on the truth of this proposition. The incentive to believe, though it evinces a positive
rationale for having the belief in question, does not count in favor of the truth of what it is you
are incentivized to believe. (See Hieronymi 2006.)
And so, our attitudes are not voluntary in the reason-based sense: we cannot form the
intention to φ for any reason that recommends having the intention to φ, just as we cannot form
the belief that p for any reason that recommends having the belief that p.17 (I will register some
caveats about this point below.)
Nor do we exercise will-based voluntary control over our attitudes. Will-based voluntary
control is epitomized by an agent’s φ-ing out of an intention to φ, where the intention to φ plays
an explanatorily salient role in the agent’s φ-ing. So, will-based voluntary control over the
attitudes of belief and intention will involve an agent acquiring the belief that p or the intention
to φ just out of an intention to acquire the very attitude in question. This would mean exercising
16 An important class of exceptions to this claim stems from Buridan-type situations wherein an agent chooses
between two alternatives that are incompatible and equally worthy of choice. See the Appendix for more discussion.
I thank Alfred Mele for drawing my attention to this issue. 17 I have learned from Thomas Pink’s (1996) discussion of similar matters.
40
the very same kind of intention-based control over the attitudes that we normally exercise over
our overt bodily movements. If we possessed this sort of control over our attitudes they simply
would not be what they are. This is because, as a constitutive matter, beliefs aim at the truth and
intentions aim at what ought to be done.18 These constitutive aims mean that we cannot believe
and intend ‘at will,’ with the same sort of voluntary control that characterizes our overt
intentional actions. To underline this point, think again about reasons to have the belief that p
that do not bear on p’s truth, or reasons to have the intention to φ that do not bear on the
desirability of φ-ing. This occurs when an eccentric billionaire offers an agent $1 million to
believe something that the agent knows to be contrary to fact, or to intend something the agent
knows is not worth doing. If we did possess will-based voluntary control over our beliefs and
intentions, there should be no bar to earning such incentives by acting from intentions to acquire
the belief in question.
Notice here a connection between the two sense of voluntariness. It is because our attitudes
are not voluntary in the reason-based sense that they cannot be voluntary in the will-based sense.
It is because we cannot form the belief that p or intention to φ for any reason that recommends
having that belief or intention that we cannot form attitudes just out of the intention to have the
attitude in question. If our attitudes were voluntary in the reason-based sense, then we would be
able to form the belief that p or intention to φ for reasons that do not bear on the truth of p or
desirability of φ-ing; we would thus be able to form beliefs or intentions for reasons based in an
eccentric billionaire’s $1 million incentive to have beliefs that go contrary to our own knowledge
or intentions to perform pointless or undesirable actions. Having formed such an intention to
acquire the belief that Michael Dukakis won the 1988 presidential election or the intention to
drink the toxin on Wednesday, I would then be free to go on to act from this very intention in
18 For a well-known argument along these lines, see Williams (1973: Ch. 9).
41
acquiring the attitude in question. But since the attitudes are not voluntary in the reason-based
sense, they cannot be voluntary in the will-based sense.19
There is a caveat to this point. To say that our attitudes possess constitutive aims does not
rule out cases in which an agent comes to find that his ‘real’ reasons for believing p or intending
to φ in fact do not bear on the truth of p or the desirability of φ-ing. Someone might come to
understand that he holds an intention to φ or a belief that p only due to social pressure, rather
than out of an appreciation of the reasons to φ or evidence for p, respectively. Such an agent
learns that his attitudes are not sensitive to the right reasons—those that bear on the truth of the
belief’s content or the desirability of the intention’s content, respectively. But these cases do not
speak against the idea that beliefs and intentions have constitutive aims. To say that beliefs and
intentions have constitutive aims is to say that it is part of the nature of belief and intention that
an agent ought not believe that p for reasons other than those that count as evidence for p and
that an agent ought not intend to φ for reasons other than those that count in favor of φ-ing. But
this ‘ought’ can be violated without the attitude in question ceasing to count as a belief or an
intention. Such violations only mean that the relevant attitude is a defective instance of its kind
and must be revised. The agents in our envisaged scenario will experience a correlative rational
tension or dissonance upon realization that the reasons he believes and intends as he does are
reasons of the wrong kind. It is precisely this susceptibility to rational pressure in the light of
reasons of the wrong kind that evinces the constitutive nature of the rational norms in question.
19 Carol Rovane asks whether my point is that the actions I can perform as a means to some further end are al so
actions I can perform ‘at will.’ I believe that my present point about the lack of either reason - or will-based
voluntariness over the attitudes cuts deeper than this. If we cannot form an attitude voluntarily in either of these
senses, there is a real question about just how it is that attitude-formation can be something for which an agent is
responsible. Without voluntariness, there is not only a constraint in the types of reasons-for-which I can hold a given
attitude, but also a total lack of an ordinary notion of will-based control over my own mind. I thank Carol Rovane
for raising this question. Since our attitudes are not voluntary in both senses of ‘voluntary,’ we cannot choose what
to believe or what to intend with the same sort of agency that we employ in choosing to move our bodies.
42
A second caveat: we might imagine scenarios in which one does manage to exercise will-
based voluntary control over the attitudes. Earlier in this chapter I envisioned a scenario in which
you slip me a pill that induces a specific attitude (such as an intention to see an off-Broadway
play). Suppose a company produces a pill that, when ingested, induces the belief that I will pass
my dissertation defense. Ingesting this pill and having this belief may help me to sleep at night,
so I ingest it. But suppose that I also have decisive evidence that I will not pass the dissertation
defense. Taking this pill is a way of exercising will-based intentional voluntary control over my
beliefs. These attitude-inducement scenarios are conceivable.20 However, they are interesting
only because they depict a kind of self-deceptive manipulation of our minds that is difficult to
bring about without such artificial means. To instill in myself a belief that p that is not responsive
to evidence for or against p is to deceive myself with respect to p, for I have made myself believe
that p whether or not I possess reasons that count in favor of p.
True, this belief will sit uncomfortably among the rest of the attitudes within my repertoire.
But the possibility of bringing about a belief that is insensitive to considerations bearing on the
truth of the content of what is believed means that we cannot rule out will-based voluntary
intentional control over the beliefs. It is nevertheless the case that we cannot exercise such will-
based voluntary intentional control over our attitudes in the paradigmatic way that we normally
exercise agency over our attitudes. As the rest of this discussion will bring out, the paradigmatic
ways that we exercise agency over our attitudes respects the constitutive rational norms that
make our beliefs and intentions what they are. So, even in the face of the possibility of belief-
implanting pills and other forms of self-deception, it is still true that the normal and
20 Here I concur with Pamela Hieronymi (2006: 61–64). Hieronymi argues ultimately that taking such a pill does not
mean that one believes that p voluntarily, only that one voluntarily brings it about that one believes that p. While I
find much in common with Hieronymi’s conclusion and her course of argument, it remains that case that after
ingesting the pill I am left with a belief that rationally conflicts with the rest of the attitudes in my repertoire, and if
the pill is lastingly effective it is nevertheless true that the belief in question has been installed.
43
paradigmatic way of influencing our own minds does not involve such self-deception and thus
that non-self-deceiving mental agency cannot be voluntary in the will-based sense. The rest of
what I have to say about the nonvoluntariness of the attitudes should be read as having this
caveat in mind.
The fact that our attitudes are not voluntary—in either the reason- or will-based sense of
‘voluntary’—presents a perplexity. We saw above that the voluntariness of ordinary overt
intentional action permits a close connection between action and moral responsibility, so that our
interpersonal reactive attitudes can register the quality of will manifested in an agent’s outward
bodily deeds. But since the attitudes are not voluntary, it appears that this link cannot be forged.
How then can it be legitimate to take up a reactive attitude towards another on account of his
attitudes of belief and intention? How can an agent bear moral responsibility for an attitude over
which he does not possess voluntary control?
The rest of this dissertation will be given over to answering these questions and resolving the
perplexity that gives them force. Before embarking on that work, we need to distinguish the
involuntary from the nonvoluntary.21 Involuntary occurrences are those over which we possess
no control whatsoever, and in respect of which one is merely patient rather than an agent. A
sneeze—or at least some sneezes—are like this. They simply come over one, seizing one’s body
before one can give a thought to preventing, directing, or modulating its occurrence.
And it also seems as though some token attitudes are like this, too. Take perceptual beliefs,
for instance. The generation of perception-based beliefs—such as my occurrent belief that my
laptop is just before me, or that I am presently sitting in the Rose Main Reading Room—are
induced in me involuntarily. They are taken in automatically when my perceptual-cognitive
system is functioning as it should. It is telling, then, that our responsibility practices do not
21 Here too I follow McHugh (2011).
44
comprehend a robust form of attitudinal blame for an agent who suffers from a mistaken
perceptual belief that results from a malfunction in the perceptual-cognitive cognitive system. An
agent is not attitudinally blameworthy for thinking that a straight stick is bent when looking at it
through water; nor is he blameworthy for thinking the lines of a Muller-Lyer illusion to be equal
in length. In fact, our responsibility practices include excuses that exempt an agent from
attitudinal blame. ‘My perceptual system was undergoing an illusion which landed me with a
faulty belief’ is just one sort of valid excuse, one that lets an agent ‘off the hook’ for having a
mistaken perception-based belief without thereby rendering the agent an unfit target for attitude-
targeting reactive attitudes as a whole. So, even while mistaken perceptual beliefs owing to
perceptual illusions are bad in that they mark a falling-short of the constitutive norm governing
belief, they do not on that account open the agent suffering from the illusion to attitudinal blame.
This I put down to the involuntariness of (at least a wide range of) our perception-based beliefs.
So it seems that the involuntary might indeed undermine any attempt to explain the sort of
agency that we exercise over our attitudes in virtue of which it is legitimate to target an agent
with attitudinal blame.
Not all of our attitudes are involuntary in this way, though—even if all of our attitudes are
not voluntary. For in between the voluntary and the involuntary there is a middle category: the
nonvoluntary (cf. McHugh 2011). If we exercise any agency at all over our attitudes, this agency
must be of a nonvoluntary sort—bearing in mind the caveat I entered above about self-deception.
Let us say that a form of agency is nonvoluntary when it (a) cannot be exercised for any reason
that recommends the target of agential control and (b) does not consist in bringing the world into
accord with some mental state or episode that represents the target of agential control. These two
constraints correspond to the reason- and will-based senses of ‘voluntary,’ respectively. Together
they allow for a form of agency that is restricted with respect to the range of reasons that
recommend its exercise and which does not consist in acting so as to bring the world into accord
45
with some prior intention representing the very state of affairs to be brought about. I underscore
that at this point the possibility even of nonvoluntary control over the attitudes cannot be taken
for granted. For there are those who insist that we can be morally responsible only for that over
which we exercise voluntary control.
Attitude-formative mental agency
Earlier we separated two questions about moral responsibility:
a. What is it for an agent to be responsible for x?
b. In virtue of what is an agent responsible for x?
In reply to question (a), I argued that we are answerable for our beliefs and intentions, and that it
is in virtue of this answerability that we hold one another accountable for the attitudes by taking
up attitude-targeting reactive practices (such as attitudinal blame). I now want to turn to question
(b). To ask question (b) is to ask after the conditions that an agent must meet in respect of x to be
morally responsible for x. In the case of responsibility specifically for the attitudes, this question
(b) takes on added urgency. The attitudes, we have seen, are nonvoluntary. Thus, we cannot
appeal to voluntariness as a responsibility-grounding agency-condition for the attitudes as we can
in the case of overt intentional action.
Volitionists about moral responsibility deny this possibility entirely. They claim that we are
morally responsible only for what is voluntary. On this view, attitudinal blame is never
legitimate. For the volitionist, we might say that an agent’s attitudes are bad—subpar, below a
standard, not as they ought to be—but never blameworthy. To illustrate the difference between
the bad and the blameworthy, consider relative judgments of physical attractiveness. To say that I
am less handsome than my brother is not to blame me for coming out the worse in the
comparison. But insofar as I do come out the worse according to some standard that measures
46
value, the assessment is a kind of normative evaluation. It is not my fault that I am less
handsome than my brother, but that does not mean that the evaluation is not a matter of assigning
value according to some ideal or some standard. But there is a distinction here between the bad
and the blameworthy. Here is Neil Levy on the point:
Volitionists agree that we can assess agents upon the basis of their morally relevant
attitudes […] What they deny is that finding that an agent is morally flawed is necessarily
to hold that agent responsible for her flaws; [volitionists deny] that all negative assessment
is blame. After all, volitionists might point out, even [opponents of volitionism] find fault
with agents without holding them to be at fault: when these faults are the product of
transient mental illness, for instance. It is, therefore, false that there is no conceptual room
for a distinction between the bad and the blameworthy, and [opponents of volitionism] owe
us an argument for closing the gap between the two. (Levy 2005, p. 6)
On this volitionist account of moral responsibility we cannot subject agents to attitudinal blame.
We cannot fault another on account of his beliefs and intentions; the most we might do is to
deem that they are not as they ought to be. Attitudes can be bad, but an agent cannot be
blameworthy for his beliefs and intentions. The challenge from the volitionist is to explain how it
is we satisfy the agency-conditions for responsibility for the attitudes—whatever these may be—
even though our attitudes are nonvoluntary.
Angela Smith has taken up this volitionist challenge (2008). Smith’s strategy is to enter two
agency-conditions on moral responsibility in general, and to argue that both of them are met
when an agent enjoys the right sort of relations among his evaluative judgments. She terms this
the Rational Relations View (2005).
Smith’s first agency-condition on moral responsibility is that moral blame,
47
unlike many other forms of negative assessment, seems to imply something about our
activity as rational agents […] Moral criticism in general […] can only be directed to a
person with regard to things that involve her rational activity in some way” (2008: 380–
381).
This is why it is illegitimate to direct moral blame to me for being less handsome than my
brother. Being less handsome than my brother is not something I have exercised my rational
agency in order to achieve. (Things would be different if I had done something to deliberately
disfigure my appearance.) Call this the Activity Condition on moral blame. It is clear enough how
the Activity Condition gets a hold in blame for overt intentional action. In the case of the
attitudes, though, the Activity Condition seems to call for some species of mental agency that is
exercised in the formation of belief and intention. I return to this point below.
The second condition centers on the justificatory demand at the heart of answerability.
“Moral criticism,” Smith writes, “by its very nature, seems to address a demand to its target. It
calls upon the agent to explain or justify her rational activity in some area, and to acknowledge
fault if such justification cannot be provided” (2008: 281). This is the very demand that renders
an agent answerable for overt intentional action as well as for the attitudes of belief and
intention. Here, Smith is noting the very same link between answerability and moral blame that I
noted earlier. When an agent is answerable for x, then he is the appropriate target for demands for
justification for x and is the proper target of blame should he proffer a poor reply to the ‘Why?’-
question. So, another reason it is inappropriate to blame me for being less handsome than my
brother is that I cannot be called upon to justify by attractiveness. I am simply the inappropriate
target for such a justificatory demand. Call this the Demand Condition on moral blame.
With Smith, I find both conditions highly intuitive. Notice how the Activity and Demand
Conditions are jointly satisfied when an agent performs a voluntary intentional action. An overt
intentional action is an occurrent happening, an occurrent event whose performance is subject to
48
the agent’s executive control. This satisfies the Activity Condition. At least paradigmatically,
overt intentional actions are performed out of a sense for the reasons that recommend its
performance. The agent’s understanding of the practical case in favor of acting works to satisfy
the Demand Condition. The reason overt intentional action can jointly satisfy both the Activity
and the Demand Conditions, though, is that overt intentional action is voluntary. My action
meets both conditions because I can act for any reason that recommends acting as I choose and
exercise control in bringing the world in conformity with that choice. This only serves to
reemphasize the question before us: how can these conditions be satisfied in the case of beliefs
and intentions that are not voluntary?
According to Smith, attitudinally blameworthy beliefs and intentions “are not failures of
[voluntary] choice, but failures of judgment and the ‘demand’ we make of [the agent] in this case
is that she acknowledge the inappropriateness of her attitude and reassess the objectionable
judgments they reflect” (2008: 383). It is telling that Smith’s characterization of the rational
activity that legitimates attitudinal blame elides quickly into that which satisfies the Demand
Condition. For Smith, the rational activity that satisfies the Activity Condition for attitudinal
blame is not voluntary choice per se but, Smith tells us, “rational activity more generally” (2008:
383).
What is this ‘more general’ rational activity? According to Smith, when we criticize another
on account of some belief or intention, we are targeting other attitudes that are implicitly
‘behind’ the attitude singled out for criticism. In another paper, Smith takes these implicit
attitudes to be evaluative judgments which “make up the basic evaluative framework through
which we view the world” (2005: 251). Such judgments “comprise the things we care about or
regard as important or significant” (2005: 251–252). Importantly, Smith claims these evaluative
judgments do not need to be explicitly entertained; nor need they be arrived at by a process of
reasoning or even rational reflection. They are simply the bedrock web of evaluative judgments
49
whose consequences ripple throughout the agent’s other beliefs and intentions. When we blame
an agent on account of an attitude of belief or intention, the target is the evaluative judgment the
that the belief or intention reflects. When I hold an evaluative judgment of the sort that Smith is
interested in, rationality dictates that the other attitudes within my repertoire ought to shift
accordingly. In this way, the surface-level beliefs and intentions singled out for attitudinal blame
are dependent on our evaluative judgments; Smith’s claim is that attitudinal blame targets these
deep-level evaluative judgments (cf. 2005: 260ff). In turn, “we take these attitudes [viz., the
deep-level evaluative judgments], like our explicit choices and decisions, to reflect our own
judgmental activity,” even though “we may never have consciously entertained the judgments in
question” (2005: 264). Indeed, Smith claims that that “we may even be surprised by our own
reactions in certain circumstances, precisely because they reveal rational or evaluative
commitments which we were hitherto unaware of holding” (2005: 264). So, the evaluative
attitudes in question need not even be something of which we possess self-knowledge. The
overall view is that the “rational activity more generally” that Smith has in mind is not any kind
of mental action, but simply the fact that the agent possesses a network of evaluative judgments
which is reflected in some belief or intention that comes in for attitudinal blame.22
At this point, though, it is easy to see the justice in the volitionist’s complaint. For it is hard
to see how a standing network of implicit evaluative judgments can satisfy the Activity
Condition, even when we interpret that condition as calling for “rational activity in general.”
Indeed, it is hard to see what is active my possessing a standing network of evaluative judgments
which may even fall outside of my own ken. In what sense can such a network be put down to
my activity qua rational agent? What is the mark that separates such a network of evaluative
22 True, Smith does say that “the justificatory demand implicit in moral criticism is a demand to reassess, modify,
and in some cases apologize for those [evaluative] judgments” (2008: 383). But it is not clear that she conceives of
these as activities per se.
50
judgments from the things that I cannot be morally blamed for, such as my physical appearance?
Obviously, a person’s network of evaluative judgments, like his physical appearance, can come
in for evaluative assessment. My evaluative judgments can be rational or irrational, warranted or
unwarranted; and my surface-level attitudes might accord or fail to accord with the deeper-level
evaluative judgments that my beliefs and intentions reflect. In the same way, my appearance can
be pleasant or unpleasant, attractive or unattractive. In neither case, though, is it evident that
these evaluative assessments carry the characteristic force of genuine blame. The complaint I am
describing is that there is nothing in the idea of such an interconnected network of evaluative
judgments (and the ‘surface-level’ beliefs and intentions that such evaluative judgments reflect)
that can make good on the Activity Condition. For I might possess just such a network of
evaluative judgments without engaging in any rational activity whatsoever. Though Smith
presents this network of evaluative judgments as constituting an agent’s ‘rational activity in
general,’ it is hard to see why she is entitled to this claim since it is unclear just why a standing
network of evaluative judgments is ‘active’ in the right way.23
We arrive at the following impasse. The volitionist’s account of the agency-conditions on
moral blame respects both the Activity and the Demand Conditions. Precisely because of this, on
the volitionist view we can be morally responsible only for voluntary overt intentional actions.
This means that the attitude-targeting reactive practices we examined in this chapter must either
come in for revision or be downgraded as consisting in non-forceful evaluative assessment.
Against the backdrop of these consequences, Smith’s Rational Relations View seems promising:
on this latter view we can be open to genuine forms of moral blame on account of our beliefs and
intentions. Thus, we can accept that the attitude-targeting reactive practices amount to genuine
forms of attitudinal blame. However, while this option respects our attitude-targeting
23 For a similar criticism of a view that resembles Smith’s Rational Relations View, see Pink (2016: 12 –16).
51
responsibility practices as they are, it cannot respect the Activity Condition on blame. Neither
option leaves our starting assumptions unscathed.
To break through this impasse, we require a notion of attitude-formative mental agency.
Attitude-formative mental agency will consist in the nonvoluntary mental activity that consists in
the productive generation of beliefs and intentions for which an agent is subsequently
answerable. I propose that we look to the mental activity of reasoning, the very form of activity
that Strawson indicated as possible only from within the engaged stance. The mental activity of
reasoning—understood as a nonvoluntary but responsibility-grounding form of agency—is a
promising candidate for a way out of our impasse. This is because, as I will point out shortly,
reasoning is an occurrent activity within a thinker’s dynamic conscious stream of thought. Prima
facie, the mental activity of reasoning differs in just the right way from the standing network of
evaluative states with which Smith seeks to satisfy the Activity Condition. There are three broad
points that indicate why this is so, having to do with (a) the ontology of conscious mental
episodes, (b) rational agency, and (c) the formation of attitudes.
(a) Ontological considerations. Actions are event-like: an action is a singular, non-
repeatable particular.24 That is why the paradigmatic example of an action is a bodily movement,
for every token bodily movement is an isolable occurrence with a determinate spatiotemporal
location and an individuating suite of causal and other relational properties. In order to speak of
intentional mental actions, we should first need to find a proper candidate type in the right
ontological category—mental entities or aspects of mind that possess such an event-like
ontological profile. Are our mental lives the scene of such occurrences?
24 I shall simply presume this Davidsonian (2011: Essay 6) thesis here. The thesis rules out some views about what
actions are at the ontological level. Alvin Goldman (1970) argues that actions are abstract exemplifyings of
properties. More recently, Helen Steward (2012) and Jennifer Hornsby (2013) have argued that actions are not
momentary events, but rather processes. Steward and Hornsby can reconcile themselves to this discussion, though,
insofar as I shall respect the general idea that actions belong within the broad class of episodic occurrences, set apart
from states in ways that I describe presently.
52
Yes—but to see why, we must take care to observe a distinction between mental states and
mental episodes. For our purposes, the most important types of mental states are an agent’s
standing propositional attitudes—his beliefs, desires, intentions, and so on. I follow
philosophical mainstream in accepting that each token attitude is individuated by two features:
the (sentence-like) intentional content that the attitude represents, and the attitudinal modality in
which it is represented. Propositional attitudes are temporally non-dynamic. They do not possess
temporal parts. They do not enter into causal relations with other aspects of mind (though
perhaps they may be influenced by causal events). An agent’s attitudes simply persist. It would
seem, then, that mental states cannot themselves be mental actions. They do not possess the
requisite ontological profile to count as particularistic occurrences.
Mental episodes, in contrast, are temporally dynamic. They include all of the conscious
events that occur within what James (1890) labeled the “stream of consciousness.” Mental
episodes may be momentary or temporally extended. They include the sorts of events that begin
and end “in an instant” (such as a person’s consciously noticing something), or they may be
temporally extended activities that an agent can be up to (“in the present progressive,” so to
speak—such as one’s consciously calculating something, or daydreaming about an upcoming
vacation). 25 They run the gamut from conscious perceptual experiences to “higher level”
conscious cognitive episodes—conscious thoughts, I take it, also figure in the stream of
consciousness.26 Like attitudinal states, token content-bearing mental episodes are partly
individuated by the content that they represent and by the attitudinal modality under which they
occur. But they are also token-individuated by their temporal location in the agent’s stream of
consciousness, and the relations that they enter into with other token episodes. The conscious
25 See Vendler (1957) for an influential classification of episodic verb-types. For application to conscious mental
occurrences, I have been influenced by Helen Steward (1997), Matthew Soteriou (2007, 2013), and Sebastian Watzl
(2017). 26 For a defense of this position, see Tim Crane (2013: Ch. 13).
53
belief that The defendant is guilty tokened at t is a distinct episode from the same conscious
thought tokened at some later time.
So understood, mental episodes possess the event-like temporal profile that also
characterizes ordinary intentional bodily actions. Like ordinary bodily actions, the mental
episode of John’s noticing something occurs at a determinate time (even if we remain silent
about the spatial location of John’s noticing something), and has a suite of causal and other
relational properties that contribute to its individuation. If this much is allowed, then we can
understand the possibility of genuinely intentional mental actions by classing them as mental
episodes in the stream of consciousness. We can understand intentional mental actions as
episodes that play a role in, and help to fix the nature of, other aspects of mind. To stress the
episodic character of mental actions, I will speak (somewhat awkwardly) of an agent’s judgings,
decidings, imaginings, rememberings, (etc.), and, later, his inferrings. These terms refer to
episodes of judging, deciding, and so on; they are distinct from the judgment or the decision
reached, which are naturally construed as the stable thought-contents that persist as mental states
or attitudes on the part of the subject.
Reasoning belongs within the category of occurrent mental episodes. As I will stress in later
chapters, reasoning is something that an agent can be up to or which can go on for an extended
period of time in the conscious stream of thought. On that score, reasoning is at least within the
correct ontological category of the kind of thing that can satisfy the Activity Condition on moral
responsibility.
(b) Reasoning and rational agency. Reasoning is related to our specifically rational agency
in an obvious way. Here we can make a rough distinction between epistemic reasoning and
practical deliberating. Epistemic reasoning aims at belief or knowledge; practical deliberating
aims at an intention for action (or simply action forthwith). I have already remarked that
reasoning and deliberating are occurrent mental activities that ‘go on’ in the stream of thought.
54
But so too are daydreaming, idle mind wandering, counting, attending to something, and trying
to remember something. However, these are not cases of reasoning. What distinguishes
reasoning is that it is mental activity that aims at some conclusion. True, not all bouts of
reasoning conclude in this way. Sometimes I am interrupted in the midst of figuring out what is
the case or what to do. Other times my reasoning is stopped short because I come to understand
that I do not have sufficient information to come to any conclusion at all (which is a kind of
meta-conclusion). In all cases of reasoning, though, I am engaged in an occurrent mental activity
that aims at reaching a conclusion.
To reach a conclusion, though, I must do so via consideration of premises that support that
conclusion. If I do not reach my conclusion by way of conclusion-supporting premises, then I
have not properly terminated my reasoning—I have merely cut it off abruptly, by fiat. I will say a
great deal more about the tie between premise and conclusion in reasoning as we go on. The
thing to notice at this point is that the premises entertained in reasoning are reasons that bear, or
at least seem to bear, on the conclusion that terminates my reasoning successfully. Obviously I
might make mistakes about this, and conclude that p for reasons that do not in fact count at all as
evidence for p. What matters, though, is that the considerations that do figure in my reasoning,
and for which I conclude that p, must at least seem to me to support the conclusion. Otherwise
they simply are not premises, for I cannot conclude from them the conclusion that p. This is due
entirely to the rational norms constitutive of belief that p.
The very same is true of practical deliberating. Suppose I reason to the practical conclusion
that I ought to walk to the café this afternoon. I must do so via the consideration of premises that
bear positively on this conclusion. Often these considerations take the form of instrumental
reasons, reasons that point up an instrumental connection between some goal to which I am
already committed and the conclusion that terminates my practical deliberation. For instance, I
think that I need to stay awake this afternoon and get some exercise, and that walking to the café
55
is a way of doing both things at once. Putting aside whether this latter belief constitutes
knowledge on my part, it cannot constitute a premise in my practical deliberating unless I take it
to be the case. The premises of practical reasoning, then, frequently operate in our reasoning by
specifying more and more proximate means I can take to realize my end. But it remains the fact
that in order to so much as count as a species of reasoning, it must be case that I aim to reach a
conclusion by way of premises that, at the conclusion of my reasoning, will count as reasons for
my conclusion.
When engaged in conscious reasoning an agent treats with reasons and attempts to arrive at
a conclusion via consideration of these reasons. So it appears that reasoning manifests our
rational agency in an obvious way. This is another mark in favor of investigating reasoning as a
candidate mental episode-type for satisfying the Activity Condition together with the Demand
Condition.
(c) Reasoning as attitude-productive mental agency. Lastly, engaging in reasoning is
paradigmatically a way of reaching new beliefs, intentions, and actions. If I am conflicted about
whether p, you will no doubt instruct me to engage in reasoning as to whether p (or engage in
reasoning with me about whether p). If I am unsure of what I ought to do, surely you will instruct
me to enter into practical deliberation with a view to settling on an intention for future action. If I
already possessed the belief that p or an intention about what to do, your advice would be
useless. Your advice presupposes that by engaging in epistemic or practical reasoning I will
generate a novel attitude, one that I did not possess when my reasoning began. It is part of our
commonsense picture of reasoning that reasoning at least aims for the production of novel
attitudes.
Reasoning, then, aims at a premise-guided termination in the generation of a new attitude.
The generation of a novel attitude is itself a mental episode in the conscious stream of thought.
As we have seen, such episodes terminate reasoning only if their occurrence is linked to the
56
agent’s treating with reasons that support the very attitude that is generated. It will be helpful to
have a ready handle to refer to such attitude-formative and reasoning-terminating mental
episodes. In the epistemic case, reasoning terminates in episodes of judging-that-p; whereas in
the practical case deliberating terminates in episodes of deciding-to-φ. As I use these terms, to
undergo a judging-that-p or a deciding-to-φ is to occurrently form the stably persisting belief that
p or intention to φ, respectively. We can thus think of judging-that-p and deciding-to-φ as the
event that is the onset of the persisting belief that p or intention to φ.
I emphasize that judging-that-p and deciding-to-φ are episodes in the conscious stream of
thought on a thinker’s part. So, judging-that-p and deciding-to-φ need to be held distinct from
judgments and decisions, understood as ‘judgmental’ and ‘decisional’ thoughts (cf. Dorsch
2009). Judgments and decisions are persisting states that are not temporally dynamic. They
belong in a totally distinct ontological category in mental life.
A second stipulative point: judging-that-p and deciding-to-φ are mental episodes that consist
in, respectively: the occurrent formation of the belief that p undertaken for reasons that (the agent
conceives as) evidence for p and the occurrent formation of the intention to φ undertaken for
reasons that (the agent conceives as) reasons for φ-ing. So, judging-that-p and deciding-to-φ
need to be distinguished from the mere occurrent thought that p or thought of φ-ing. As I will use
these terms in this dissertation, not even an occurrent thought that p in the awareness of one’s
epistemic acceptance of p, or an occurrent thought of φ-ing in the awareness of one’s practical
commitment to φ-ing, count as judging or deciding, respectively. Judging and deciding are
attitude-formative events linked to the considerations that the agent takes to support the content
of these very episodes. I make this stipulation just to focus our attention on the character of
judging and deciding as functioning to conclude reasoning successfully in the generation of a
novel attitude undertaken in the light of reasons that support the very attitude that is generated.
57
Summing up, reasoning is an activity that engages our rational agency in the production of
novel attitudes of belief and intention. In this, reasoning is similar to overt intentional action.
Intentional actions are episodic in temporal profile: actions are particulars that occur at a specific
place and time. Also like overt intentional actions, reasoning manifests our rational agency. Just
as I perform a bodily movement for reasons that recommend its performance, so too I reach
conclusions in reasoning by way of premises that I acknowledge as reasons for the very
conclusion that terminates my reasoning. Lastly, reasoning to a conclusion involves
(paradigmatically) the production of some change—in this case, a change in my standing
attitudes of belief and intention by terminating reasoning in a judging-that-p or a deciding-to-φ.
This is akin to the changes I bring about by engaging in bodily action, which is likewise
productive in nature. On these counts reasoning appears at least to be a promising candidate for
satisfying the conditions for moral responsibility for the attitudes and the reactive attitudes we
take to others’ attitudes.
Before concluding this chapter, I want to pause over one important feature of reasoning. As
an occurrent mental activity, reasoning is voluntary. I can engage in reasoning for any reason that
recommends doing so; and I can begin reasoning just out of an intention to do so, by directing
my attention to considerations that bear on the matter I hope to settle. But, while reasoning is
voluntary, judging-that-p and deciding-to-φ are not voluntary. If judging-that-p and deciding-to-φ
were voluntary, then we would indeed possess voluntary control over the attitudes. This follows
from our stipulative notions of judging and deciding, according to which judging and deciding
consist in the occurrent formation of novel attitudes of belief and intention, respectively. So,
while I might open a deliberative (epistemic or practical) question by engaging in reasoning, my
reasoning must aim at conclusion in a nonvoluntary type of mental episode, a judging-that-p or a
deciding-to-φ.
58
It looks, then, as though we might break through the impasse described in this section by
paying closer attention to the interplay between reasoning and the attitudes. What we must
understand is how reasoning to a conclusion might count as an intentional but nonvoluntary
species of rational agency, by the exercise of which an agent satisfies both the Activity and the
Demand Conditions on moral blame.
Conclusion: Three desiderata
In this chapter I have argued that we bear a distinctive form of responsibility for our
attitudes of belief and intention. We are answerable for our beliefs and intentions, in the sense
that we stand subject to a justificatory demand to provision reasons why we believe or intend as
we do, and are open to distinctive patterns of forceful normative criticism according to how well
we are prepared to discharge this demand. True, this species of responsibility rarely rises to the
level of serious moral concern, and it would be unjust to bodily punish those who fall short of
the rational norms that govern the attitudes. Despite these factors, I have tried to show that our
responsibility for belief and intention is no less a fixture of the inter- and intrapersonal reactive
practices that constitute our stances of participatory engagement.
But in virtue of what do we bear this responsibility? How can we bear responsibility for the
attitudes when we do not have the same sort of control over them as we enjoy over overt
intentional action? We require an account of the attitude-formative mental agency that
legitimates attitudinal blame. This account must meet the following three desiderata:
Desideratum 1: Responsibility-grounding. The account must explain how it is that exercising
this species of agency works to ground our responsibility for belief and intention. The account
must be responsibility-grounding. As we have seen, the shape of our attitudinal responsibility is
that of answerability. So, the account of mental agency that emerges must explain how it is that
59
we are fit subjects for requests for justification on account of our attitudes, and subject to forceful
normative criticism according to how well a thinker meets this demand.
Desideratum 2: Nonvoluntariness. But our account of responsibility-grounding attitude-
formative mental agency must respect the nonvoluntariness of our beliefs and intentions. Our
attitudes are nonvoluntary in a twofold sense: we cannot believe that p or intend to φ for any
reason that recommends having the attitude in question, and nor can we directly executive
intentions to believe that p or intend to φ.
Desideratum 3: Productivity. In addition to nonvoluntariness, the account must also explain
how it is that manifestations of this sort of agency are productive. How does the exercise of a
nonvoluntary form of mental agency produce or generate novel beliefs and intentions for which
we subsequently bear answerability?
I have suggested that a promising way forward in meeting these desiderata is to concentrate
on reasoning that terminates in a judging-that-p or a deciding-to-φ. But can judging and deciding
count as a type of mental action that meets these desiderata? That is the question I turn to in the
next chapter.
60
Chapter 2
The Mark of Mental Action
The criterial question
To what extent are we agents, rather than patients, in our mental lives? Which mental episodes
count as intentional mental actions on the agent’s part? What is the mark or criterion that
distinguishes genuine mental actions from those conscious mental episodes that merely happen
to us? Call this latter question the criterial question about mental action.1
Aside from its intrinsic interest for the philosophy of mind and action, our answer to the
criterial question bears directly on the legitimacy of attitudinal blame. A primary route to
responsibility is via fully intentional actions—those actions an agent does knowingly and
deliberately. But is conscious mental life the scene of such intentional performances? We cannot
say so without an idea of that which separates the active from the passive in mental life. If we are
to say that we are responsible not only for our outward bodily deeds but also for aspects of our
own minds, we must first answer the criterial question.
This is why I argued in Chapter 1 that our inquiry into responsibility for the attitudes must
proceed by investigating the species of agency we exercise over our own attitudes. In what does
such attitudinal agency consist? In particular, we must determine whether two mental episode-
types in particular—judging-that-p and deciding-to-φ—can rate as fully intentional mental
actions in their own right, and if so on what grounds. As I remarked near the end of the last
chapter, these mental episode-types have an attitude-formative character. It is in judging p to be
1 With this, I mean to ask a set of questions about mental action that Donald Davidson raised about bodily action.
Davidson asked: “What events in the life of a person reveal agency; what are his deeds and his doings in contrast to
mere happenings in his history; what is the mark that distinguishes his actions?” (2011: 43). Note at the outset that
my concern is with mental actions—the mental counterpart to fully intentional bodily performances. There may be a
wider class of ‘agentive’ mental phenomena that don’t quality as fully intentional mental actions in this sense.
61
true that I acquire the belief that p, and it is by deciding to φ that I acquire the intention to φ. Yet
it is highly contentious whether these attitude-formative mental episode-types can rate as
intentional actions. This is because we cannot place judging and deciding either within or outside
the scope of mental agency without knowing what it is in general for a mental episode to count
as active or passive. Hence our focus in this chapter on the criterial question.
My aim is to explicate and evaluate an answer to the criterial question that rests on a
distinction between what an agent does directly—without reliance on distinct causal-instrumental
means—and what is merely brought about as the receptive causal product of what the agent does.
Only the former count as action, according to this approach. I call the view instrumentalism since
its proponents assign great significance to an agent’s use of ‘mediating’ causal-instrumental
processes to bring about some intended mental episode. In this, the instrumentalist carries over to
the mental domain Donald Davidson’s famous conclusion that “our primitive [i.e., direct]
actions—those we do not do by doing something else, mere movements of the body—these are
all the actions there are. We never do more than move our bodies: the rest is up to nature”
(2011b, 59). Keeping to Davidson’s conclusion, the instrumentalist claims that every mental
action is direct; the rest is up to nature.
Proponents of instrumentalism include Strawson (2003), O’Shaughnessy (2008: 543–547;
2000: 382ff), and Dorsch (2009). Dorsch, for instance, describes the view this way:
The contrast between straightforward [i.e., direct] and mediated [i.e., indirect] deliberate
agency […] is best drawn […] in terms of the fact that, while the latter involves the
instrumental reliance on certain epistemic or merely causal processes and their passive
effects, the former does not. The relevant processes are thereby characterized by the fact
that, once they are triggered by us and progress normally, they lead by themselves—that is,
without further help or involvement of agency and other factors external to them—to their
desired or intended outcome. (2009: 41)
62
Here, Dorsch explicitly links passivity with a mental episode’s being brought about by a causal
process that an agent instrumentally relies upon to bring it about. If that is right, then the
instrumentalist can sort out the active from the passive in mental life by looking to the “epistemic
or merely causal processes” that we rely upon to bring about desired or intended mental
episodes.
In treating with instrumentalism, though, I will generalize away from any one philosopher’s
views with an eye towards capturing a compelling model of how we exercise agency over our
minds. As we will see, the instrumentalist takes a familiar model of intention-based control as
central to intentional action (at least, in paradigmatic responsibility-grounding cases). This form
of control is typified by an agent’s forming an intention to φ and φ-ing from this very intention.
One reason instrumentalism is attractive is that it applies this paradigmatic model of control to
bodily and mental action alike. Whether over our bodies or over our own minds, key to the
instrumentalist’s view is the agent’s forming and executing intentions without reliance on distinct
causal means to bring about what she intends. In treating critically with the instrumentalist reply
to the criterial question we will also learn something positive about the limits of this intention-
based model of control for mental agency. In what follows I am more interested in evincing these
broader lessons than I am in targeting the views of any one specific philosopher.
I wish to explore what instrumentalism implies for the mental episode-types that are central
for the possibility of attitudinal agency. My worry will center on what the instrumentalist must
say about the relationship between (i) the temporally extended episodes of deliberation and
reasoning wherein I work to settle for myself the answer to some question, and (ii) the attitude-
formative episodes of deciding-to-φ and judging-that-p wherein I arrive at an answer to the
question at issue. As we will see, the instrumentalist comprehends the relationship between (i)
and (ii) in causal-instrumental terms. For the instrumentalist, deliberating (reasoning) is a causal-
instrumental cognitive process that brings about a deciding (judging). It is this feature of
63
instrumentalism I wish to question. Pace the instrumentalist, deliberating (reasoning) is not a
cognitive process that is undertaken in order to causally bring about some deciding (judging).
Rather, the relation between the reasons that figure in deliberating (reasoning) and deciding
(judging) is more broadly normative, in that one comes to a deciding (judging) for the normative
relations recognized between the reasons the figure in deliberating (reasoning) and the deciding
(judging) that the agent undertakes. I argue on these grounds that the two mental episode-types at
the center of this study, judging and deciding, present significant problems for the
instrumentalist. Judging and deciding are neither ‘direct’ nor ‘indirect,’ neither basic nor non-
basic, in the sense relevant for instrumentalism. Since this is so, instrumentalism cannot provide
a reply to the criterial question.
We must find an alternative way of answering the criterial question. My suggestion will be
that we draw the distinction between mental agency and mental passivity in terms of rule-
following. I can do no more than motivate the rule-following approach in this chapter, since the
rest of the dissertation will work towards an account of that in which the agency of attitude-
formative mental actions, considered as rule-governed occurrences, consists. The purpose of this
chapter is to motivate an approach to mental agency that takes reasoning and deliberating as the
paradigm of intentional mental action.
Over the next few sections, I will explicate instrumentalism first with respect to ordinary
bodily action. Then I will demonstrate the instrumentalist criterion of mental action with respect
to two mental episode-types for which it seems particularly suited: remembering that p and
imagining x. I bring this instrumentalist analysis to bear on the mental episode-types relevant for
attitudinal agency: deliberating, reasoning, deciding, and judging. Then I shall try to set out the
aforementioned worries about instrumentalism. That argument allows us to draw some positive
upshots about a better reply to the criterial question what this might mean for how to draw the
scope of mental agency.
64
Instrumentalism: bodily action
Action is productive. To perform an action is to bring about an event or a change. To
intentionally raise my arm is to bring it about that my arm moves from a stationary to an upright
position. Now, some of the changes that I bring about by acting are done by doing something
else. Suppose I press a button on a remote control attached to my keychain; when pressed, this
button emits a signal that unlocks my car’s door. Here, the event that I bring about is the car
door’s unlocking. But I bring this event about by doing something else: button-pressing on the
remote control. What is involved in doing one thing by doing another? Action theorists
distinguish between basic, non-basic, and direct action to answer this question in a principled
way.
First, we need to separate the results of an action from its consequences.2 The result of an
action is the event or change in which the action (partially) consists. Results are constitutive of
the actions whose results they are. The result of my action of arm-raising is the rising of my arm
from a stationary to an upright position. Without the occurrence of this event of arm-rising, no
action of arm-raising can occur. However, while results are intrinsic and so necessary for the
actions whose results they are, the occurrence of a (token) result that is associated with a given
action-type is never by itself sufficient for the occurrence of an action of that type. This is just to
say that while every time I raise my arm there is an arm-rising, it is not the case that I raise my
arm every time it rises. My arm might rise due to non-actional reflex.
An action’s consequences, in contrast, are the effects that the action causes. In contrast to its
results, consequences are extrinsic to the action whose consequences they are. My arm-raising
might have a plethora of consequences. It might cause a lecturer to grant me attention to pose a
question, or annoyance in the person sitting beside me. Neither of these causal effects are
2 See McCann (1998: Part II), Stoutland (1968), Danto (1970), Smith (1988), and Enç (2003).
65
intrinsic to my arm-raising. Consequences, then, are neither necessary nor sufficient for an
action; they are simply events causally downstream of an action.
The distinction between results and consequences permits a preliminary construal of basic
and non-basic action. It will be helpful to work with a simple example.
Key: In approaching my car from ten feet away, I depress a button on a remote control
attached to my keychain. When the button is pressed, the remote control emits a signal that
unlocks the driver’s side door.
Here, the (putative) action of unlocking my car door has a result that is necessary but not
sufficient for its occurrence: namely, the event of the door’s unlocking. But this result is also a
consequence of the result of another (putative) action that I perform: that of pressing the button
on the remote control. I unlock the car door by pressing the button on the remote. In general:
when an agent S performs an action of ψ-ing by performing some distinct action φ-ing, the result
of S’s ψ-ing is a consequence of the result of S’s φ-ing. When this is the case, we say that S’s φ-
ing is basic relative to his ψ-ing, and S’s ψ-ing is non-basic relative to his φ-ing.3
A most basic action is one in relation to which there is no distinct action that is basic relative
to it. When this is the case, I will say that the agent φ’s directly, or that φ-ing is a direct action. In
Key, my button-pressing is basic relative to the door’s unlocking. But it is not a direct action,
since pressing the button is non-basic relative to another action that I perform: namely, moving
my thumb. This is because the result of my button-pressing is a consequence of the result of the
distinct action of thumb-moving. Since (let us suppose) my thumb-moving is not the
consequence of any distinct action’s result, my thumb-moving is the most-basic—and so,
direct—action.
3 I focus on causal non-basicness. See Goldman (1970: Ch. 2) for different kinds of non-basicness.
66
There is another aspect of doing one thing by doing another that we must capture. This is the
idea of an agent’s doing something in order to do something else. When I unlock the car door by
pressing the remote by moving my thumb, I do so for the sake of a purpose or goal that, in
moving my thumb, I aim to realize. We need to capture this purposive element. Three conditions,
(A)–(C), must be met for an agent to φ in-order-to ψ.
(A) An agent φ’s in-order-to ψ only if the agent intends to bring about the result of ψ-ing. If
the agent does not intend the result of ψ-ing, then its occurrence can be nothing more than
unintended consequence of whatever direct action he does perform. It is this condition (A) which
prevents us from saying that every one of my arm-raisings is performed in order to disturb a
local collection of air molecules. Without this first condition, the things I bring about by dint of
doing something else cannot count as agentive in any meaningful sense.
In correspondence, Alfred Mele asks if the following scenario is a counterexample to this
first condition (A). Imagine an agent, Fred, who knows that a nuclear reactor is set to explode
soon. The only way to prevent this disaster is to enter a 10-digit code into a nearby console. Fred
does not know the code and has no way to learn it before the reactor’s melting down. He begins
typing numbers into the console at random. Luckily, he strikes upon the correct sequence and
prevents the reactor’s melting down. Has Fred typed in the code in order to shut down the
reactor? If so, the case appears to count against (A), since then Fred has entered in the correct
code, and has done so in order to prevent the reactor’s melting down, but cannot be said to intend
to prevent the reactor’s melting down. The difficulty lies in a belief-condition on intending to φ.
Proposals vary as to the specific condition involved, but Mele supplies a modest version: “One
cannot intend to φ while believing that one will not, or probably will not, φ” (1992b: 176). Yet
Fred certainly believes that he probably will not prevent the reactor’s melting down, since he
strikes the numbers on the nearby console at random. Since Fred violates the belief-condition on
intending and so cannot intend to prevent the reactor’s melting; still, Fred seems to enter in the
67
code in-order-to prevent the reactor’s melting. Understood in this way, it does seem that Fred’s
case presents a counterexample to condition (A).
In reply, I suggest we follow Will Small (2012) in making a distinction between intention
and aspiration. The difference between aspiring x and intending to φ comes down to a number of
factors, including know-how, belief that circumstances are appropriate, and other factors that
ground the belief-condition on intending. I might aspire to do something or to bring something
about that I have no inkling how to do or produce. For instance, an adolescent might aspire to
become an astronaut without having any conception of the types of training this career will
require. Most importantly, while there is a belief-condition on intending, no such condition
constrains an agent’s aspiring to bring something about. We should say that Fred aspires, but
cannot intend, to prevent the reactor’s melting down. While Fred acts intentionally in the hope of
realizing his aspiration (in entering random combinations into the console), he cannot be said to
have intentionally prevented the reactor’s shutting down. Fred’s preventing the reactor’s
meltdown is a happy accident, a lucky turn, but not something he does intentionally. Such chance
occurrences bear too accidental a relation to what Fred does do intentionally to fall within the
scope of the in-order-to relation I am attempting to clarify here.
(B) An agent φ’s in-order-to ψ only if he acts from an instrumental belief in the efficacy of
φ-ing. The agent must reasonably believe that the result of ψ-ing can come about as the
consequence of the result of φ-ing.4 Lacking such an instrumental belief, we cannot say that the
agent conceives of his means-action as bearing any significant relation to the end he intends to
bring about. The occurrence of that end will strike the agent as a mere happy coincidence. Such
an instrumental belief needn’t rise to the level of knowledge. Nor must it attribute to one’s
selected means a guarantee of success. Oftentimes we engage some means in the belief of likely
4 Cf. Goldman’s discussion of ‘means-end knowledge’ in (1970: Ch. 2).
68
success. To simplify discussion, though, I will restrict attention to instrumental beliefs that
represent one’s chosen means as causally sufficient for the results of the intended end.
Instrumentalism is most plausible when these simplifications are in place.
(C) An agent φ’s in order to ψ only if he φ’s for the reason that the result of φ-ing is
efficacious of the result of ψ-ing. The basic idea behind (C) is that settling on and engaging a
means is something done for reasons; the reason-for-which an agent φ’s, when he φ’s in-order-to
ψ, is precisely the causal-instrumental relation that the result of φ-ing bears to that of ψ-ing. This
is reflected in, among other places, an agent’s practical deliberation about what to do. Practical
deliberation frequently (if not always) involves settling on an action the result of which bears the
right causal relation to the results of the goals we intend to realize. And this (purported) causal
relation is what constitutes for the agent an instrumental reason to φ.
Following John Gibbons, we can say something stronger here. The notion of a means is
already inflected with normativity. To see φ-ing as a means to ψ-ing when one intends to ψ is to
recognize an instrumental reason to φ to bring about the result of ψ-ing. For there is a sharp
distinction between grasping two events as merely related by a causal relation and understanding
one event as a means to a further end. As Gibbons has it, “To see A as a means to B is to see the
connection between A and B as a reason for A-ing. To see A as a means rather than merely as a
cause or a necessary or sufficient condition is to see it as something you have reason to do”
(2009: 90). So, in employing some means for the reason of its efficacy of a further end, the agent
also acts for a reason—namely, the instrumental reason that φ-ing is a means to ψ-ing. This is our
condition (C).
Since this condition (C) will be important to the argument to follow, let me buttress it by
considering a putative counterexample from Alfred Mele.5 An agent Joe sees that cutting off his
5 I thank Professor Mele for posing this question in correspondence.
69
head is a means to killing himself. But he also recognizes that there is no reason to kill himself
and no reason to cut off his own head. Yet the case of Joe is not a counterexample to condition
(C). For all that has been said, Joe merely recognizes a causal relation between cutting off his
head and committing suicide. To see the former as a means to the latter, Joe has to at least put
himself in the position of conditionally taking up the intention to committing suicide to ask
himself how he would kill himself were he to (attempt to) do so. To look upon cutting off his own
head as a means, Joe must enter into the practical analog of suppositional epistemic reasoning. In
suppositional epistemic reasoning, a thinker supposes that p to discern p’s epistemic entailments.
Similarly, if Joe is to discern the means to ψ-ing (as opposed to causal relations among event-
types) he must at least take up a suppositional intention to ψ, to discern what he would have to do
to bring about a ψ-ing were he to intend to ψ.
In the absence of condition (C), the occurrence of the agent’s end cannot be something
brought about intentionally. Consider our Key case again. Suppose I satisfy conditions (A) and
(B), but not (C): I endeavor to unlock the car door from ten feet away; I believe that pressing the
button on the remote control is efficacious of this result; but when I do press the button on the
remote, I do not do so for the reason of its efficacy of my end. Rather, I press the button only for
the reason that doing so gives off a pleasant tactile sensation. In these circumstances, it seems
proper to say that while I have the intention to unlock the car door, and the right sort of
instrumental belief linking the button-pressing to the door-unlocking, these two events are not
integrated into the sort of instrumental act-complex that allows us to say that I act in order to
bring this end about. Instead, the occurrence of my intended end figures as an unintended, though
happily convenient, consequence of something else that I do intentionally (induce the pleasant
tactile sensation). If this is right, then (C) is a necessary condition on an agent’s doing something
in-order-to bring about some further end.
70
We can now enter a substantive necessary condition on an agent’s doing something in order
to do something else.
In-order-to: S successfully φ’s in-order-to ψ only if: S successfully φ’s; and the result of ψ-
ing is a consequence of the result of φ-ing; and (A) S intends to bring about the result of ψ-
ing; and (B) S φ’s from a reasonable instrumental belief that the result of ψ-ing is a
consequence of the result of φ-ing; and (C) S φ’s for the very reason that the result of ψ-ing
is a consequence of the result of φ-ing.6
We will repeatedly revisit this In-order-to condition in connection with mental action. I will refer
to actions so related as causal-instrumental act-complexes (or ‘act-complexes,’ for short). To
signify that an act-complex meets this condition, I will use the hyphenated expression ‘in-order-
to,’ as in ‘Heeney presses the remote in-order-to unlock the car door.’
As Arthur Danto (1963, 1965) originally pointed out, it seems as though whenever an agent
φ’s in-order-to ψ, there is some action the agent does directly—without having to do anything
else. Otherwise, the agent will never so much as commence acting at all.7 We can use the
foregoing notions to furnish a necessary condition on doing something directly:
Direct: An agent φ’s directly only if the agent forms an intention to φ and φ’s out of that
very intention.8
6 I have said that condition (A)–(C) are individually necessary, but not sufficient. Sufficient conditions might involve
reliability of the relevant causal chains and the lack of ‘causal deviance,’ skill on the part of the acting agent, a nd so
on. The argument to follow rest only on the necessity of these conditions, especially (C). 7 Danto’s argument has recently been criticized by Thompson (2008) and Lavin (2013). I propose here to grant
provisionally the need for direct actions, though, since one upshot of the arguments to come will be that certain
mental episode-types (deliberating and deciding, primarily) do not fit easily within this scheme. 8 One might worry that Direct presupposes the so-called ‘Simple View’ about the relation between intention and
intentional action. On the Simple View, an agent intentionally φ’s only if he has an intention to φ. Readers who deny
the Simple View are free to substitute the following amended version of Direct: An agent φ’s directly only if either
the agent forms an intention to φ and φ’s out of this very intention or φ’s from an attitude that lies within the
motivational potential of some other intention within the agent’s repertoire. Even this amended version of Direct
71
Why does Direct take off from our prior notion In-order-to? Because Direct requires that the
agent φ out of the very intention to do just that. This cannot be done when the agent does one
thing in-order-to do another. I certainly can form the intention to Unlock the car door when I am
ten feet away from the car; but I cannot do so out of that very intention. I must cast about for a
distinct instrumental means and form the intention to employ these means in-order-to execute the
further intention. Direct, then, precludes the use of distinct causal-instrumental means, since such
means require that I form a distinct intention.
We have defined basic action, non-basic action, and direct action all while taking the core
notion of an action for granted. Which events are actions?
According to many action theorists, some event counts as an action just in case there is some
act-description under which the event is intentional.9 One immediate consequence of this view is
that the class of an agent’s actions cannot be specified extensionally, since the very same episode
will be intentional under some act-descriptions and nonintentional under others. ‘Heeney’s
unlocking the car door’ is an act-description under which the event of the door’s unlocking
counts as an intentional action on my part; ‘Heeney’s disturbing the nearby prowler’ may refer to
the very same event (viz., the door’s unlocking), but fail as a description of what I intend to bring
about (I did not even know that the prowler was lurking by my car). Our first step in isolating the
class of events which count as actions, then, is to heed the descriptions under which events can
be described as intentional.
‘Heeney’s pressing the remote’ and ‘Heeney’s unlocking the car door’ are certainly two such
descriptions. But do these two act-descriptions refer to two distinct actions? Or do they rather
precludes the formation of a distinct intention, one to do a distinct action, ψ-ing; the intention to φ and the
‘intention-serving’ attitudes within its motivational potential are the only allowable action -making attitudes. For the
notion of a motivational potential and how it interfaces with the Simple View, see Bratman (1984). 9 See again Davidson (2011: Essay 3), Anscombe (1957), Moya (1990).
72
refer to the very same event under two distinct descriptions? If I φ in-order-to ψ, do I do one
thing or two?
Instrumentalism is predicated on the view that in such cases there is but one thing the agent
does. To claim this is to accept a so-called coarse-grained view of action-individuation. On this
view, the act-description ‘Heeney’s unlocking the car door’ refers to my direct action, the one
whose result is the moving of my thumb; the act-description ‘Heeney’s unlocking the car door’
refers to this action by citing one of its consequences.
For the coarse-grained theorist, it turns out that ‘non-basic action’ is a misnomer. The
distinction between basic and non-basic action is not a distinction among actions, but rather a
distinction among the descriptions we employ to think and talk about actions. Though the act-
descriptions ‘Heeney’s moving his thumb,’ ‘Heeney’s pressing the button on the remote,’ and
‘Heeney’s unlocking the car door’ are intensionally distinct, they are extensionally equivalent,
referring to the very same event—in this case, the movement of my thumb.
The coarse-grained theorist can respect the distinction between an action’s results and its
consequences with the notion of semantic indication (Smith 1988). The act-description ‘Heeney’s
unlocking the car door’ semantically indicates the event of the door’s unlocking, while ‘Heeney’s
pressing the remote’ semantically indicates the depressing of the remote’s button. Adapting a
definition by A.D. Smith (1988: 404), we can say:
Act-description D is non-basic relative to a distinct act-description D′ if and only if the event
semantically indicated by D is an effect of the event that is semantically indicated by D′.
Act-description D is a direct act-description if and only if there is no such distinct act-
description D′.
The coarse-grained theorist claims that when an agent φ’s in-order-to ψ, there is but one thing
that the agent does. The action in question consists in the bringing about of the event that is
73
semantically indicated in the direct act-description. When I unlock the car door by pressing the
button on the remote by moving my thumb, there is but one thing that I do, and the description
‘Heeney’s moving his thumb’ is the direct act-description throwing this action into relief.10
Again, the other act-descriptions here refer to that event by citing the consequences of the event
it semantically indicates.
The upshot for the coarse-grained view is that all actions are direct. Proponents of the
coarse-grained scheme of action-individuation frequently motivate this conclusion on the basis
of the following intuition. There is nothing more that I need to do in-order-to unlock the car door
after my thumb has depressed the button on the remote control. In general: for every causal-
instrumental act-complex, there is nothing more an agent does after the occurrence of the event
that is semantically indicated by the direct act-description of that act-complex.11 After I bring
about a movement of my thumb, I merely need to let the causal chain do its work. To see this,
recall our conditions (A)–(C): I intend to bring about some distal effect, believe that φ-ing will
cause that very effect, and φ for the very reason of its efficacy of this effect. Given that this
causal-instrumental structure is in place, all I need do by way of intentional action is bring about
the event that is at the ‘near side’ of the causal chain; “the rest,” as Davidson puts it, “is up to
nature” (2011: 59).
The instrumentalist reply to the criterial question takes off from this coarse-grained theory of
action individuation.12 Given that distinct act-descriptions of a given act-complex refer to one
10 Actually, there is dispute among coarse-grained theorists whether bodily movements are the results of direct
action-descriptions. Hornsby (1980) argues that direct act-description must semantically indicate events causally
upstream of bodily movements. 11 See Davidson (2011: Essay 3), Hornsby (1980). 12 There are compelling reasons to reject the coarse-grained view. As others have noted, the view faces a serious
difficulty: intensionally distinct act-descriptions appear to refer to events that take place at distinct times and places,
and which bear asymmetrical causal relations to the direct actions that take place at the outset of the relevant act-
complex. The description ‘Heeney’s unlocking the car door’ refers, among other things, to an episode of door-
unlocking, which takes place after and as an effect of the event that the description ‘Heeney’s moving his thumb’
refers to. It seems impossible that these descriptions refer to the very same event. See McCann (1998: Part I) and
74
and the same action, we can separate the active from the passive with reference to the agent’s use
of distinct causal-instrumental means. Whenever we find a causal-instrumental act-complex in
which an agent φ’s in-order-to ψ, we need only discern the direct action that the agent performs.
Since “the rest is left up to nature,” the events downstream of direct action fall outside of the
scope of agency proper.
The instrumentalist thereby furnishes something of a procedure for determining whether
some mental episode x counts as a mental action on the agent’s part: first, determine the
descriptions under which x is intentional; then, determine whether the agent brought x about by
employing distinct causal-instrumental means. If so, then x cannot count as an action; the agent’s
action must lie causally upstream of x’s occurrence. Instrumentalism relies on the intuitive
distinction between what an agent does and what, in contrast, the agent merely brings about by
doing something that eventuates in the event in the question.
Instrumentalism: remembering, imagining
This instrumentalist criterion seems particularly well-suited for remembering and
imagining.13
Mental actions of remembering and imagining certainly possess results.14 When one
deliberately attempts to remember or to imagine something, there are mental episodes the
occurrence of which mark success in what one was attempting to do. When I set about
attempting to remember a student’s name, success in my attempt is marked by the occurrence in
my stream of thought of a conscious mental episode that represents that the student’s name is
Ginet (1990: Ch. 3) for argument along these lines. Still, I propose to grant the coarse-grained view about
individuation for present purposes. Disputing instrumentalism by taking issue with the coarse -grained view on
which it relies would not require us to dwell on specifically mental actions. One of my broader aim here, though, is
to point up some of the special problems that arise when we consider actions in the specifically mental domain. 13 See Dorsch (2009) and Strawson (2003) for characterizations of instrumentalism about mental action. 14 Pace McCann (1979), Steward (2012: 33 fn. 17).
75
(say) Adam. And in setting out to imagine the Eiffel Tower, success means the occurrence of
(what I’ll call) a presentational image of the Eiffel Tower, perhaps as-if-seen from some specific
vantage point (from the north bank of the Seine, say).15 These success-conditions are not trivial.
We all know what it is like to try and fail to recall a person’s name, or to be unable to summon
the desired sort of presentational image. Importantly, the occurrence of these mental episodes is
necessary but not sufficient for the correlative mental action-type (a remembering or an
imagining, respectively) to occur. So, we can designate them as the results of mental actions. (Or
as the mental episode-types that are semantically indicated by mental act-descriptions.)
Is remembering that p a mental action? We must first reckon with the fact that I cannot
remember that p directly. I can φ directly, we said, only if I can form and execute an intention to
φ, and φ out of that very intention. But if I could form an intention to remember that p, then I
would have thereby remembered that p. And if I could thereby remember that p just by forming
an intention to remember it, I would not be in my present predicament of having to remember
that p.
But I can form an intention to bring about a remembering under some distinct description.
Suppose I am speaking with a student after class; I recognize the student’s face but cannot recall
his name. (His name is Adam.) While I cannot form an intention to Remember that this student’s
name is Adam, I can form the intention to Remember this student’s name (where ‘this’ refers
demonstratively to the very student before me). This suggestion gets us closer to satisfying
Direct, but not all the way. While I can form the intention to Remember this student’s name, I
cannot remember the student’s name just by executing this very intention. Unlike raising my arm
out just out of the very intention to raise it, I cannot call the student’s name to mind just by
15 For the view that every imagining of x is an imagining of a possible experience as of x, see Martin (2002).
76
executing the intention to remember this student’s name. If I could do so, I would (as before) not
be in my present predicament.
What I can do—and do directly—is to try to remember the student’s name. Trying to
remember something is a mental action-type is its own right. It has its own characteristic result,
which consists in the activation of a cognitive process that, while hardly transparent to the
conscious subject who engages it, does possesses its own distinctive phenomenology.16
Furthermore, trying to remember cannot be identified with the remembering itself. Otherwise,
there would be no such phenomenon as trying but failing to remember that p. It is possible to
engage in trying to remember something—to ‘chase after’ in conscious thought a salient ‘tip of
the tongue’ sensation—but fail to remember the item in question.
What is the relationship between trying to remember that p and remembering that p, when
the attempt is successful? Notice that the following features all seem to hold true of our scenario,
provided I am successful: the result of remembering the student’s name is a consequence of the
result of my trying to remember it; (A) I intend to bring about a remembering of the student’s
name; (B) I possess a reasonable instrumental belief regarding the causal efficacy of trying to
remember the name; and, most importantly, (C) I try to remember the student’s name for the very
reason of the efficacy of the cognitive process that is activated in trying to remember it. The two
mental episode-types, trying to remember and remembering that p, form a causal-instrumental
act-complex that satisfies In-order-to.
The outcome of this instrumentalist analysis is that remembering that p—the mental episode
that (at least partly) consists in the occurrence in conscious thought of p—is not itself an action.
The action, the instrumentalist tells us, is confined to my trying to remember, to my activating the
16 William James vividly describes this phenomenology. “Suppose we try to recall a forgotten name. The state of our
consciousness is peculiar. There is a gap therein, but no mere gap. It is a gap that is intensely active. A sort of wraith
of the name is in it, beckoning us in a given direction, making us at moments tingle with the sense of our closeness,
and then letting us sink back without the longed-for term” (1890 v. 1: 251).
77
cognitive mental mechanism that causes the student’s name to come to mind. The act-description
‘Heeney’s remembering that the student’s name is Adam’ refers to the mental activity of my
trying to remember the student’s name. This is because trying to remember is the direct mental
action at the nearest end of the causal chain that eventuates in the occurrence of my intended end.
It is because I have resorted to a distinct causal-instrumental means that the instrumentalist
places the event of remembering Adam’s name outside of the scope of mental agency (cf.
O’Shaughnessy 2000: 382ff; Dorsch 2009: 42).
This is how the instrumentalist furnishes us with a criterion of mental action. The
instrumentalist proposes to separate the active from the passive in mental life with reference to
the distinction between what is done directly and what is, in contrast, the receptive causal
product of what is done directly. The In-order-to condition is a useful heuristic in addressing
ourselves to the criterial question because the actions—or act-descriptions—that can be
legitimately substituted within it allow us to separate the active from the passive by specifying
the most direct means, and the end these means are engaged to bring about.
It is important, though, that instrumentalism leaves room for theorists to disagree about what
falls within and what falls outside the scope of mental agency. Instrumentalism just means that
two philosophers who disagree about whether some mental episode-type x falls within the scope
of mental agency disagree about whether x can be brought about without reliance on distinct
causal-instrumental means.
We find just such a disagreement among instrumentalists regarding imagining. Galen
Strawson suggests that (at least in many cases), just as we are receptive in relation to the
occurrence of the remembering that p, so too we are receptive in relation to the occurrence of a
presentational image. Strawson believes that presentational images spring up in one’s awareness
at the slightest suggestion, a suggestion that causes one’s imagination to run off of its own
78
accord. In a portion of his paper that especially highlights the coarse-grained theory of action-
individuation animating the instrumentalist reply, Strawson writes:
One entertains the verbal specification and waits for the mechanisms of imagination—the
(involuntary) spontaneity of imagination—to deliver the [presentational] image […] To
think that the actual content-issuing and content-entertaining that are at the heart of
imagining are themselves a matter of action seems like thinking, when one has thrown a
dart, that the dart’s entering the dartboard is itself an action. (2003: 241–242)
Strawson discerns room for an instrumental analysis of imagining x. He suggests that we
perform mental actions in-order-to bring about a presentational image. So, the only action in
imagining is confined to what is done directly (issuing a “verbal instruction,” according to
Strawson). The occurrence of the presentational image is something in relation to which the
agent, having performed the direct action, is merely receptive. The act-description ‘Heeney’s
imagining of the Eiffel Tower’ just describes my action (the issuance of the verbal instruction) in
terms of one of its salient consequences.
In contrast, Fabian Dorsch assigns to imagining the status of fully intentional mental action.
Unlike Strawson, Dorsch cannot discern in imagining the same kind of reliance on instrumental
means that characterizes successful remembering. For Dorsch, in imagining a sunny forest
we do not thereby think of the link between our employment of this capacity (or similar
capacities, such as the capacity to visualize trees) and the occurrence of the visual
[presentational] image in both causal and instrumental terms […] In particular, we do not
form the intention to use our capacity to visualize sunny forests in rational response to our
intention to visualize a sunny forest and an instrumental belief that making active use of
this capacity is likely to cause the occurrence of the desired visual image. (2009: 43)
79
Here, Dorsch argues that imaging x does not stand in the In-order-to relation to some prior
mental action on the agent’s part. Dorsch classifies imagining within the scope of mental agency
for the very reason that he can discern no distinct mental action that, together with the
occurrence of the intended presentational image, makes up a causal-instrumental act-complex.
But what is important is that while Dorsch and Strawson disagree over whether imagining x
is active, they agree that the instrumentalist reply provides the right criterion of mental agency.
They both hold that episodes which are brought about via instrumental reliance are passive; they
merely disagree over which episodes are brought about this way.
Instrumentalism: deciding, judging
Our purpose all along has been to determine whether judging and deciding can count as
intentional mental actions, and if so on what grounds. We now have ready to hand a developed,
widely espoused, and independently motivated reply to the criterial question. Let us bring this
reply to bear on the attitude-formative episode-types that matter most for attitudinal agency. I
will treat first with deciding (and deliberating), and only afterwards briefly apply the following
instrumentalist analysis to judging (and reasoning).
Is deciding-to-φ a mental action? The instrumentalist answers this question by falling back to
another question: Can an agent decide-to-φ directly, without relying on distinct causal-
instrumental means? Do we decide directly, in this sense?
We do not. An agent can φ directly only if the agent can form an intention to φ and φ out of
that very intention. But we do not form intentions to decide-to-φ; nor can we execute an intention
to decide-to-φ just out of an intention to do just that. Indeed, such an intention—an intention to
decide on some specific course of action—would render the resulting intention superfluous.
80
To bring this out, suppose that I do enjoy just such an intention, say an intention to Decide to
walk to the café. Now, every intention embodies a distinctive species of practical commitment to
the performance of the intended action. In intending to φ, I am rationally accountable for φ-ing
so that failing to φ at the appropriate time opens me to a distinctive pattern of rational criticism,
or even to a species of attitudinal blame in case φ-ing is not an all-things-considered desirable,
valuable, or choiceworthy thing to do. (I shall revisit this specific form of practical irrationality
in later chapters.) So, if I were to enjoy an intention to Decide to walk to the café, this intention
must embody a practical commitment to deciding to walk to the café at noon. But, as several
philosophers have noted, such a commitment obviates the very intention to walk to the café at
noon that is formed in my deciding to do so.17 Why should I enjoy a commitment to adopting the
commitment to walking to the café at noon? Why not just move now to the commitment to
walking to the café at noon by so deciding to act? For most agents, in most circumstances, it is
hard to understand just how one might enjoy the commitment to deciding without therein taking
up the commitment to the action.18 If that is right, then we should not suppose that agents enjoy
intentions to decide to perform some specific future course of action. The upshot is that deciding
to perform a specific course of future action does not satisfy Direct. We do not decide-to-φ
directly.
How then do our decisions come about? Are they entirely passive mental occurrences, which
do not involve our agency in any respect? If this were so, it would be hard to imagine how the
actions that flow from these decisions could be active in any sense whatever. Luckily the
instrumentalist needn’t claim this. While I cannot decide-to-φ out an intention to do just that, I
17 See McCann (1998: Ch. 8), Levy (2013). 18 The ‘most agents, most circumstances’ qualification must be added to accommodate a thought experiment of
Mele’s (2000). Mele’s experiment seems to show that the intention to decide-to-φ is not conceptually incoherent.
This qualification does not weaken the present argument, though, since most decisions are not undertaken in the face
of the demonic constraints that Mele imposes on the decider who figures in his example.
81
can form a distinct intention, say an intention to Make some decision or other.19 I can form such
an intention readily in the face of some practical uncertainty about what I ought to do. In contrast
with an intention to Decide to walk to the café, the intention to Make some decision or other does
not render otiose the eventual decision I do make, the decision to walk to the café.
However, while I may readily form the intention to Make some decision or other, I cannot
execute this intention just out the intention to do so—at least, I cannot do this if I want to come
to a decision on the basis of reasons. To execute the intention to Make some decision or other
and to decide in accordance with reasons I must treat with practical reasons in deliberation. It
appears, then, that I can bring about some decision or other only by engaging in a suite of distinct
mental actions, namely mental actions in which deliberation consists. Such mental actions might
include: posing questions to oneself (e.g., ‘What ought I to do?’); attending to reasons (e.g.,
attending to the potential health benefits of walking to the café); or trying to remember
considerations that will prompt and feed into subsequent deliberation. We find a nice
enumeration of some relevant mental action-types in a recent paper by Lindsay Rettler (2018).
While Rettler is concerned with settling one’s mind on some epistemic matter of fact, the very
same considerations apply to the practical case. According to Rettler,
The mental actions that compose reflection include asking ourselves various questions and
directing our attention in various ways […] One might direct one’s attention toward a
certain class of reasons, such as relevant memories, or away from certain reasons. One
might try to compare or contrast the case at hand with other similar cases. Trying to
remember various reasons, synthesizing them, reconstructing an argument for a
proposition—all these are examples of the various mental actions that are part of critical
reflection. (2018: 2217)
19 See Robert Kane (1996), Alfred Mele (2000), Randolph Clarke (2003).
82
I will call the mental actions that ‘compose reflection’ on some epistemic or practical question
reflective mental actions. In performing a suite of reflective mental actions, one directs one’s
attention to reasons, poses questions oneself to prompt and sustain thought, or tries to remember
considerations that bear on the question one wants to resolve.
It is plausible that reflective mental actions of this sort causally influence the attitudes to
which we arrive in deliberation. If this were not the case, it would be hard to make sense of the
fact that we often instruct others to perform reflective mental actions when they face a practical
uncertainty and need to make up their minds one way or another on the basis of reasons. Faced
with a practical uncertainty, then, I can make up my mind some way or other by performing a
chain of such reflective mental actions—bringing considerations to mind, attending to reasons,
posing questions to myself, and so on.
Reflective mental actions have another feature worth pointing out: they can be done directly.
I can form the intention to perform a reflective mental action and execute that intention
forthwith, without having to form a secondary intention that will bring it about. Indeed, the
voluntary and immediate use of the attention—as when, for instance, I attend to some practical
consideration that recommends one plan of action over others—is a paradigm of direct mental
action.20 There is simply no instrumental ‘distance,’ when it comes to reflective mental actions,
between the agent’s intention to do them and their performance. Similarly, there is no such
distance between reflective mental actions and deliberation. The relation between reflective
mental actions and deliberation is instead constitutive. Deliberation consists in the performance
of a suite of reflective mental actions. Or so it seems consonant with other instrumentalist
commitments to say.
20 William James agreed. In fact, he took all action, bodily as well as mental, to consist in the direction of the
attention. See his Principles of Psychology, especially his chapters on “Attention” and “The Will.”
83
We arrive now at the central matter. Suppose that, faced with some practical uncertainty, I
form (or anyway, acquire) the intention to Make some decision or other. To execute this
intention, I perform a chain of direct reflective mental actions. The causal upshot of such
reflective mental actions is the occurrence of a deciding in my conscious stream of thought. This
will be a conscious deciding to perform some specific course of action, say to Walk to the café.
What is the relationship between the reflective mental actions I perform directly and the
occurrence of my deciding?
It looks as though the relationship is given by the In-order-to condition set out above. The
result of my action of settling on some decision or other, namely the formation of an intention to
Walk to the café, is a consequence of the result of some antecedent reflective mental action on
my part. (A) I intended to come to some decision or other; (B) I believed reflective mental actions
to be a sure route to executing this intention; and (C) I performed reflective mental actions for
the very reason of their efficacy of some decision or other. This means that my reflective mental
actions and the resulting deciding to Walk to the café together compose a causal-instrumental act-
complex.
By this token, the instrumentalist classifies deciding to act not as an action, but as something
brought about by my action. Or better: the act-description ‘Heeney’s making some decision or
other’ refers to my reflective mental actions—but does so in terms of one of their causal
consequences. (Notice that the act-description ‘Heeney’s deciding to Walk to the café’ is not a
description under which any action of mine is intentional; no deciding is intentional under any
description that mentions the specific action decided upon, for the instrumentalist. That is one of
the morals of our argument that deciding-to-φ is not, and cannot occur as, a direct action.) Since I
must rely upon distinct causal-instrumental means (reflective mental actions), the occurrence of
my deciding in my conscious stream of thought is a receptive intentional episode in relation to
which I am passive. It is like the conscious occurrence of the name ‘Adam’ in the midst of my
84
trying to remember This student’s name. Since I deliberate in-order-to decide some way or other,
deciding itself is a passive mental episode, outside of the scope of mental agency.21,22
The instrumentalist provides the same account of reasoning and judging. I cannot judge-that-
p directly. If I could judge-that-p directly, then I could believe that p at will, for any reason that
makes the belief that p good to have. I could believe for reasons of the wrong kind. But I cannot
do this: I cannot believe that I will pass my dissertation defense just because having such a belief
will let me sleep at night. I can, though, form the intention to Judge some way or other for the
sorts of epistemic reasons. I can execute this intention by performing a suite of reflective mental
actions—by, say, attending to the right kinds of evidence. If I am successful, then some belief or
other will be formed by my judging. In that case, I will have successfully performed a suite of
reflective mental actions in-order-to bring about some judging or other. By that token, my
judging-that-p is a passive mental episode that is merely brought about by my performance of
distinct causal-instrumental means.
I mentioned that one noteworthy outcome of this instrumentalist analysis is that deciding-to-
φ and judging-that-p cannot be intentional under any act-description that semantically indicates
the specific action decided upon or specific proposition judged to be true. My deciding to Walk to
the café can be intentional under a more general description, such as ‘Heeney’s deciding what to
do this afternoon.’ Already this raises suspicion. Our entire enterprise has been to legitimate
attitudinal blame. But now we learn from the instrumentalist that no episode of attitude-
formation can count as an action under any description that semantically indicates the specific
21 Someone might wonder whether deciding specifically to φ can be rendered intentional because it is within the
motivational potential of the intention to make up my mind one way or the other (see above, note 8). But this
deciding specifically to φ will be intentional by dint of some other pro-attitude within the motivational potential of
the intention in question. Then the agent must have some species of pro -attitude towards deciding specifically to φ.
This result encounters two problems. First, it is unclear just what such pro-attitudes are, if not intentions (cf.
McCann 1998: Ch. 10). Second, such pro-attitudes to deciding to φ presuppose the very commitment to φ-ing that
deciding to φ inaugurates. This is just to reiterate the previous conclusion that deciding cannot be direct. 22 For defense of this view of the genesis of deciding-to-φ, see O’Shaughnessy (2008: 543–547).
85
attitude therein formed. If we have firmer allegiance to our responsibility-practices surrounding
the attitudes than we do to the instrumentalist reply to the criterial question, we might on that
account look elsewhere for a responsibility-grounding account of attitudinal agency. I caution
against such a quick dismissal. One reason to dwell on the instrumentalist reply to the criterial
question is that doing so will be highly instructive for a positive theory of attitudinal agency. Just
what is wrong with supposing that reasoning and deliberating stand in causal-instrumental
relations with the episodes of judging and deciding (respectively) that conclude these mental
activities? This will be the question we take up in the following section.
Against instrumentalism
I will not, then, attempt to show that the instrumentalist returns a mistaken verdict with
respect to judging or deciding. Instead I will argue that the distinction between causally basic and
non-basic action on which instrumentalism rests is not adequate to the relationship between
deliberating and deciding (reasoning and judging). Deciding-to-φ and judging-that-p are neither
direct nor indirect in the instrumentalist’s sense. Instrumentalism cannot furnish a criterion of
mental action because it cannot return a stable verdict with respect to the attitude-formative
mental episode-types that are important for attitudinal agency. I’ll discuss deliberating and
deciding first, then briefly apply the same lessons to reasoning and judging.
I first want to take issue with the instrumentalist’s conception of deliberating as consisting in
the performance of a suite of reflective mental actions. What is it to deliberate?
As a constitutive matter, deliberating is an answer-seeking activity. One seeks in deliberation
an answer to (what I’ll call) a deliberative question, such as ‘What ought I to do?’ or ‘What shall
I do?’ In pursuit of an answer to one’s deliberative question, one treats with reasons that bear on
live practical alternatives, such as ‘Walking to the café’ or ‘Staying at home.’ These practical
86
alternatives figure as candidate answers to the deliberative question that one has taken up in
deliberation.23 We can model such an ordinary course of deliberation as follows:
1. I ought to: stay awake this afternoon and get some exercise.
2. Walking to the café is the (all-things-considered) best way to stay awake this afternoon
and get some exercise.
3. So, I shall walk to the café.
The first premise modeled here is a normative commitment, or a desiderative state, or an all-
things-considered assessment of what one ought to do. The second premise is a belief about some
means the taking of which would advance the state of affairs depicted in the first premise.
Together these items constitute what Davidson called a primary reason (2001a: 4). Many
construe primary reasons as belief/desire complexes. However, composition by a belief and
desire is not in fact essential to primary reasons. What is essential is that they together provide
(what we can call) an instrumentalizing explanation of why the agent eventually did what he did
(viz., walk to the café). The explanation is ‘instrumentalizing’ in that it consists in two elements:
a goal the agent seeks to achieve or realize and a plan of action that the agent takes to advance
that goal. I shall have much more to say about these matters in the next few chapters. For now I
only want to enter two intuitive starting claims about this model of deliberating, claims that will
provide a suitable starting-place for critical reflection on the instrumentalist conception of the
relationship between deliberating and deciding.
The first of these claims is that the episode of my deciding just is my reaching the conclusion
(3) from premises (1) and (2). There is no other mental event in which my deciding consists. Put
another way, my deciding just is the mental episode that marks the conclusion of my practical
reasoning. If this claim is not granted, it will be difficult to locate my deciding in the conscious
23 For the view that reasons are considerations that bear upon (deliberative) questions, see Hieronymi (2005).
87
stream of thought. In what might a deciding-to-φ consist, if not in the transition in thought that is
the formation of an intention?
Secondly, deciding closes reasoning via the premises that figure in reasoning. When I close
deliberation in a deciding, the reasons-for-which I decide-to-φ are the premises that figured in
my practical deliberating. Like the first claim, this second claim derives its plausibility from
starting intuitions about our ordinary experience of deciding. When I decide on the basis of
reasons, it is no mystery just why I decide as I do; I am always cognizant of deciding for the very
reasons that figured in my prior deliberation about what I ought to do. This is easy to understand
if every deciding for reasons involves recognizing the normative significance of the reasons-for-
which I decide as I do. (This so-called Taking Condition on inference will be defended and
explained more thoroughly in Chapter 3.)
We arrive at the following simple (and no doubt incomplete) account of what is involved in
deliberating-closing deciding. Deliberating consists in treating with premises that together make
up an instrumentalizing explanation of why I decide as I do. The deciding that closes my
deliberating just is the premise-guided transition to the episode that is the formation of a stable
intention to act. To say that this transition is premise-guided is just to say that in deciding I
appreciate the normative force of the practical reasons that are the premises from which I decide.
Observe, though, that the instrumentalist cannot accept even this much as stated. For there is
no evident role here for reflective mental actions—mental actions such as attending to reasons.
To fit reflective mental actions into the picture, the instrumentalist must say something more: that
the intentional episode (3) must come about as the consequence of some distinct direct reflective
mental action. The instrumentalist claims that I treat with (attend to, dwell upon) reasons (1) and
(2) in-order-to induce in myself some deciding or other; the occurrence of (3) satisfies my
intended end of reaching some decision or other. This means that on the instrumentalist’s
account, I attend to reasons (1) and (2) not in the first instance for their normative bearing on my
88
deliberative question. Rather, I do so because it constitutes effective causal-instrumental means
of bringing about a mental episode that I cannot bring about directly. But this seems incorrect.
It seems incorrect because it locates instrumental relations in the wrong place: as inhering
among intentional episodes that make up a causal-instrumental act-complex in my own mind,
rather than among states of affairs in the world and the plans of action I settle upon to realize
these states of affairs. In deciding, we said, I appreciate the reasons that bear on the action I
decide on. Appreciating these reasons involves recognizing that the prospective plan of action at
issue will contribute to the satisfaction of my settled objectives. It is this worldly instrumental
relationship—the relationship between my settled objectives and the plans that will advance
those objectives—that I recognize in deciding on those very plans. This is because it is the
worldly relationship between these items which constitutes an instrumental reason for the plan of
action I decide upon, and it is in recognition of this instrumental reason that I decide as I do.
The instrumentalist, though, illicitly ‘psychologizes’ these relations. For the instrumentalist,
the relevant instrumental relations for the genesis of my deciding hold between mental episodes:
between my attending to the contents (1)–(2) and the deciding, (3), which is the causal upshot of
my attending. The relations are psychologized in that I no longer decide just out of an
appreciation of the normative support that the instrumental reason provides for the plan of action
of walking to the café at noon. On the instrumentalist’s picture, I decide only by doing something
else that causes episode (3) (viz., attending to (1) and (2) as a means of activating some
cognitive-causal process that will bring about (3)). This move locates the instrumental relations
as holding among my employment of a mediating means and its distal effect. For the
instrumentalist, agency in deliberating and deciding is a matter of the causal-instrumental
relations within an act-complex that is composed entirely of intentional episodes.
And so, the instrumentalist’s story of how deciding concludes deliberating is more
complicated than the simple story we set out in describing (3) as a premise-guided transition
89
from (1) and (2). Why should we prefer the simpler account of deliberating and deciding over the
instrumentalist’s psychologized construal? Here there are two interconnected problems.
The first problem is that the instrumentalist’s account in fact presupposes the more basic
rational ability modeled in (1)–(3). Recall condition (C) within the In-order-to relation: an agent
employs a causal-instrumental means φ-ing in order to bring about a distal end ψ-ing only if the
agent φ’s for the very reason of its efficacy of a ψ-ing. In the present context this means that
when I perform some reflective mental action in-order-to bring about some decision or other, I
must attend to reasons for the reason that doing so will bring about some decision or other. But
this is the very same rational ability that is modeled above, in (1)–(3): it is the very same rational
ability at issue in reasoning to a conclusion. The instrumentalist’s psychologized conception of
employing deliberation as a means of bringing about some decision or other adds nothing to the
explanation of how a decision comes about that is not already present in the simple description I
offered above of (1)–(3).
Secondly, and relatedly, the instrumentalist account introduces a vicious regress. Earlier I
remarked that when an agent cannot bring about a ψ-ing directly, he faces a further practical
question. This is a how-to question: What should I do in-order-to bring about a ψ-ing? This is
itself a deliberative question, a question whose answer I seek in subsequent deliberation. The
answer to this question will be some action φ-ing, which is both a direct action and whose result
bears the right sort of causal relation with (the result of) ψ-ing. In settling such a further practical
question, I therein settle how I will directly act, given my further aim of ψ-ing. This latter species
of settling—the sort which results in a fully-formed practical plan of action that is not vulnerable
to a further how-to question regarding means—presents the instrumentalist with a regress.
To illustrate, let us label my decision to walk to the café ‘deciding1.’ The instrumentalist tells
us that deciding1 is not something I do directly, but something I bring about by employing
causal-instrumental means. This raises for me a further practical question—a how-to question
90
about the best possible means of bringing about my end. So, I must settle on deliberating—
performing a suite of reflective mental actions—in-order-to bring about some decision or other.
This requires that I deliberate as follows:
4. I ought to bring about some decision or other.
5. By deliberating (attending to/dwelling upon reasons for action) I can bring about some
decision or other.
6. So, I shall deliberate.
Is mental episode (6) an action? It appears itself to be a decision to act, namely a decision to
perform a mental action that will cause some further mental episode to come about (a decision to
act). Call the episode in (6) ‘deciding2,’ a decision to do something to bring about my deciding to
walk to the café (deciding1). The instrumentalist tells us in turn that deciding2 is not an action,
but in fact the mere receptive causal product of something else that I do by performing some
distinct causal-instrumental means. If so, then I must do something else in order to bring about
deciding2. This will require a further deciding, deciding3, wherein I settle upon the best
instrumental means of bringing about deciding2. But deciding3 is also a deciding to act, and so
the receptive causal product of something else that I do. This will require that I make some
deciding4, wherein I settle on the best efficacious means…Clearly, we are off on a regress.
Given this regress, it cannot be the case that I employ some distinct causal-instrumental
means in order to bring about a decision. In employing that distinct causal-instrumental means, I
will already have settled on something for a reason. And this phenomenon—settling on a plan of
action out of an appreciation of reasons for that action—is the very phenomenon the
instrumentalist attempts to explain by positing some causal-instrumental causal structure in the
genesis of an ordinary decision to act.
91
I want to pause at this point to note a structural similarity between this regress and the one
that Lewis Carroll (1895) famously pointed out for epistemic reasoning. I shall discuss Carroll’s
regress in further detail in Chapter 3. Briefly, Carroll pointed out that if a thinker is not prepared
to reason directly from premises that support an epistemic conclusion, furnishing that thinker
with a rule cannot help him. If I refuse to conclude that ‘Socrates is mortal’ from the premises ‘If
Socrates is a man, then Socrates is mortal’ and ‘Socrates is a man,’ no rule of reasoning can help
me. Suppose you tell me that the indicated pattern of reasoning corresponds to modus ponens,
and so on this basis I ought to accept the conclusion that ‘Socrates is mortal.’ Carroll points out
that I cannot apply the modus ponens rule to the first-order premises without exercising the very
ability to reason to a conclusion that we are attempting to explain. It will be no use to provide me
with yet further rules of reasoning, as this will simply generate a regress of premises from which
no conclusion at all can emerge. Each such further premise will be meta-level rule of reasoning
relative to the premise that precedes it. And so, each meta-level premise takes the thinker further
away from a conclusion guided by reasons that count as evidence for that conclusion.
The regress I am describing now has a similar structure but corresponds to a distinction
between epistemic and practical reasoning. Practical reasoning (paradigmatically) begins with an
end and concludes with a suitable means that an agent can perform forthwith. My practical
deliberating cannot conclude in action I can perform forthwith unless I conclude in an action that
I can perform without reliance on distinct causal instrumental means. But the regress argument I
have just described rules out this possibility of conclusion in direct action. The regress consists in
further and further premises that are instrumental means to something I might be able to do
directly—but with no end at all in sight. Just as in Carroll’s regress I never reach premises from
which I can reach an epistemic conclusion, so too in the regress just described I can never settle
on an action I can perform forthwith from premises that together constitute an instrumental act-
complex connecting a direct action to my distal end.
92
Note that I do not dispute the instrumentalist’s claim that I perform reflective mental actions
directly. I am not claiming that I must perform some intermediate action in order to bring about
deliberation. Deliberation can be entered into just by attending in various sorts of ways upon the
occasion of practical uncertainty. Instead, the argument targets the tacit claim on the part of the
instrumentalist that I might settle on deliberating (reflective mental actions) without thereby
doing the very thing the instrumentalist has sought to explain by positing a causal-instrumental
act-complex.
Before defending this argument against various objections, I want, briefly, to state an
analogous argument for reasoning and judging. Suppose I want to know Blair’s location. I know
that she’ll be at the café if Alice is and I have just learned that Alice is at the café. I am now in a
position to reason as follows:
7. If Alice is at the café, then Blair is at the café.
8. Alice is at the café.
9. So, Blair is at the café.
As before, we can assume that my judging that Blair is at the café just is an inferring, in
which I move from (7) and (8) to episode (9) on the basis of those very same reasons. There is no
separate episode in which my judging consists. Secondly, in inferring that (9), I appreciate the
reasons (7) and (8) that license the inference. Here we do not find that (7) and (9) together make
up an instrumentalizing explanation for the belief that Blair is at the café. The relation is one of
epistemic evidential support. This is a difference we will discuss in subsequent chapters.
The instrumentalist will claim that I attend to, dwell upon, or otherwise treat with reasons (1)
and (2) so as to cause in myself some judging or other; doing so brings about episode (3), which
marks the successful execution of my intention to make up my mind with respect to the
deliberative question I have taken up.
93
Yet, this instrumentalist account presupposes the basic rational ability modeled in moving
from (7)–(8) to episode (9). Before I can practically settle on reasoning—attending to reasons (7)
and (8)—as a way to bring about a judging, I must first settle on an epistemic matter regarding
the effectiveness or the desirability, in my present circumstances, of attending to reasons in order
to bring about some judging. These are indeed epistemic matters: it is a matter of belief whether
adopting some means φ-ing is an effective way to bring about a ψ-ing. This means that the
instrumentalist provides no explanation of how my judging is brought about, since the very same
ability the instrumentalist seeks to explain is at issue in any instrumental-act complex in which
an agent performs a basic action to bring about some intended further event.
This opens the door to regress. The instrumentalist tells us that my judging is not an action,
but something I must bring about with the use distinct causal-instrumental means. As we have
seen above, this raises a how-to question about just which means I ought to employ. I cannot
settle this further practical question without settling on the epistemic question of which means
will in fact be the most effective (reliable, convenient, suitable for present circumstances, etc.). I
must, then, reason as follows:
10. I ought to make up my mind one way or another.
11. By engaging in reasoning, I am most likely to end up with a judging that accords with
my evidence.
12. So, reasoning to a conclusion is the best way to make up my mind.
13. So, I will attend to reasons that will cause some judging or other.
Here, step (12) is an epistemic conclusion, one that consists in acceptance of all-out belief
regarding the best efficacious means to employ in order to bring about some judging or other. But
is step (12) a mental action? It seems itself to be an action, namely an action of judging wherein I
form the belief represented in (12). The instrumentalist must say that such a belief only comes
94
about by doing something else that is more basic. So, I shall need to need engage in some prior
mental activity to bring it about, and so on.
Just as before, it cannot be the case that I employ some distinct causal-instrumental means in
order to bring about a judging. In employing that distinct causal-instrumental means, I will
already have settled on something for a reason. And this phenomenon—settling on something for
a reason—is the very phenomenon we were attempting to explain by positing some causal-
instrumental causal structure in the genesis of an ordinary judging. Once more, we are directed to
the more fundamental phenomenon: inferring for reasons. Although this last run of the argument
involves epistemic reasoning, it the regress nevertheless has the practical character of a regress
of means indicated earlier.
Before moving on to register objections and replies, let me briefly comment on what these
arguments bring out. Again, the point I have been driving is not that the instrumentalist provides
a faulty verdict regarding the action-status of this or that mental episode-type. My point is that
judging and deciding are neither direct nor indirect in the instrumentalist’s sense. Deciding-to-φ
and judging-that-p are not direct mental action because an agent cannot perform a deciding-to-φ
or a judging-that-p just out of an intention to do so. However, we also cannot make sense of the
proposal that deciding-to-φ and judging-that-p are brought about by dint of performing some
more direct mental action. This is because, I have argued, we cannot understand an agent’s using
distinct causal-instrumental means to bring about some deciding or other, or some judging or
other, without presupposing the very ability at issue.
If this is right, then judging and deciding are neither direct nor indirect in the
instrumentalist’s sense. So, there is simply no way for the instrumentalist to issue an answer to
the criterial question with respect of these mental episode-types: regarding them, we simply
know not what to say.
95
Two objections
I want to address two main lines of objection to these arguments. The first objection is an
attempt to stop the regress by refusing to posit some explicit judging or deciding; the second line
takes off from the evident goal-directed nature of deliberating and reasoning.
The regress arguments rest on the following charge: that the instrumentalist’s analysis of
deciding and judging must posit in the explanation the very thing the instrumentalist seeks to
explain—a judging or a deciding. I allege that in performing mental actions in order to bring
about a deciding, I must decide to use those very means; and in performing mental actions in
order to bring about a judging, I must judge those very means to be efficacious of my further end.
Against this, someone might object that reasoning and deliberating do not involve an explicit
judging and deciding in this way.24
The objection can be set out in a number of ways. First, one might suggest that deliberation
begins this way: upon realization that I face a genuine practical uncertainty, I am automatically
thrown into the sorts of reflective mental actions that constitute deliberation. On this suggestion,
I cannot but engage in the sorts of direct mental action-types that cause decidings to come about.
I passively confront the need to make up my mind some way; this sets in play my attending to
various considerations; and this in turn causes some decision or other to come about. Call this the
automaticity reply.
The automaticity reply, though, threatens to erode the very means-end structure in terms of
which the instrumentalist explains the genesis of our decisions. If deliberating is induced in me
automatically, then I cannot really be said to be engaged in deliberating as a means to some
further end. Suppose you ask me for the sum of 5 and 7. My calculating is then immediately set
off and I arrive at the answer, 12. Here, there is no question of my engaging in some mental
24 This objection was put to me by Gordon Cooper in helpful commentary at Florida State University.
96
operation of calculating in order to reach a sum. There is simply a conscious mental process that
is sparked by a stimulus, delivered in the form of your request for the sum. In particular, I am not
performing a calculation for the reason of arriving at the sum, which is our condition (C) on φ-
ing in-order-to ψ. If some mental process, conscious or not, is induced in me automatically, then
I cannot be said to be engaged in that process for a reason. The upshot is that even if this is the
right thing to say about the relationship between deliberating and deciding, it cannot make
instrumentalism more plausible.
Can the instrumentalist claim that deliberation begins automatically and that the agent, in a
supervisory role, intervenes in the process after it has begun? This version of the automaticity
objection certainly preserves the instrumental structure of deliberating in order to decide. But it
does so only at the cost of inviting the regress. On this suggestion, I must make an explicit
deciding, deciding2, to intervene in the deliberative process that has begun automatically and will
culminate in deciding1. Now, deciding2—the deciding to intervene in the deliberative activity
that is now underway—is not something I can do directly (according to the instrumentalist). On
this suggestion, then, I must make yet another deciding, deciding3—which is the decision to
decide2—to intervene in the deliberation that will culminate in deciding1. On this version of the
automaticity objection, there is an explicit need to decide to deliberate.
Moving on from the automaticity objection, consider instead a habituation reply. On this
reply, when I face practical uncertainty I move into deliberation with the kind of deft swiftness or
easy facility that obviates the need to dwell explicitly on alternatives and decide2 to deliberate in-
order-to decide1. On this reply, moving from practical uncertainty to deliberation is agentive and
undertaken in-order-to decide one way or the other; but it is not itself an explicit deciding to use
deliberation as a means to this end. Picture a swift and circumstance-appropriate reflex
movement, such as suddenly ducking to avoid the path of a moving object headed right toward
you. You do not deliberate and decide to avoid the moving object. But doing so is nevertheless an
97
intentional action on your part, indeed an intentional action you perform in-order-to do
something else (avoid being struck).
But we must be careful not to overload what we mean by deciding. Recall the intuitive claim
I entered above: decidings just are premise-guided conclusions that close deliberation. To decide-
to-φ is to transition to an intention to φ for reasons that recommend φ-ing. Decidings can occur
as deftly, swiftly, and fluidly as any other kind of inference, and I would suggest that our mental
lives are absolutely teeming with such inferential activity. The moving from practical uncertainty
to deliberation in order to bring about a deciding is, then, itself a deciding to use deliberation in
order to bring about a deciding to act. This is because the transition from the awareness of some
practical uncertainty to deliberating for-the-reason that doing so is an effective causal-
instrumental means of bringing about some deciding or other counts itself as an inferential
transition. By that token, it counts as a deciding—which is the very thing the instrumentalist was
attempting to explain.
True, such movements in thought may not involve conscious mulling over alternative
options. (Though I think there is something to be said for the idea that all deliberation takes place
against an implicit awareness that one alternative is to do nothing at all, or simply to evade the
deliberative question at issue.) Again, I do not think such alternatives are necessary for deciding.
What is all-important is that decisions are inferential transitions undertaken for reasons (and I
doubt there is anything more basic to which we might hope to reduce this phenomenon).
I turn to a second line of objection. This one takes off from our ordinary image of
deliberating and reasoning. What makes deliberating and reasoning distinctive as the types of
mental activities that they are? Deliberation and reasoning are not idle mind-wandering. Rather,
they are mental activities that are oriented toward a goal. What goal is that? It seems as though
the goal must be deciding and judging, respectively. And so, it seems that an agent engages in
deliberation and reasoning for the sake of some further goal, namely reaching a judging or a
98
deciding. If it is true that (as we say) “inquiring minds want to know,” reasoning and deliberating
are the means we employ in order to get what we want. If we accept the regress arguments,
though, it seems we cannot accommodate this end-directed nature of reasoning and deliberating.
The instrumentalist accommodates the end-directed nature of deliberating and reasoning by
positing reflective mental actions that serve as the efficient causes of the deciding or the judging
that concludes inquiry. As I have argued, this only generates a regress. To accommodate the end-
directed nature of deliberating and reasoning, we cannot posit mental actions undertaken in-
order-to bring about some deciding or judging. We need to make a different kind of distinction
between the episodes the figure in deliberation and reasoning and do away entirely with the
instrumentalist’s reflective mental actions.
To do so, let us heed more closely the relationship that intentional episodes (1)–(3) bear
among themselves in the conscious stream of thought of the deliberating thinker. And that
relation seems not, or at least not originally, efficient causal in nature at all. Rather, the episodes
are related as premises to conclusion. The content of episodes (1) and (2) constitute reasons that
count-in-favor-of the conclusion, (3). The regress for the instrumentalist arises precisely from the
attempt to explain this normative, premise-conclusion relationship by positing reflective mental
actions that are done in-order-to bring about the conclusion. As we have seen, the very notion of
φ-ing in-order-to ψ presupposes the premise-conclusion relationship, since our notion of a means
undertaken in-order-to bring about an end is already a thoroughly normative notion (as we said
above). In shifting away from the instrumentalist’s emphasis on efficient-causal relations
between these intentional episodes, we should seek to understand how premise is related to
conclusion in the stream of conscious thought.
This will be a matter I explore in subsequent chapters. To anticipate that discussion, we
should emphasize what is granted by all parties: that, in some sense, deliberation properly
concludes in deciding and that reasoning properly concludes in judging. The fact that these
99
mental activities are ‘concluded properly’ in deciding or in judging is not something incidental to
them. It is a constitutive feature of deliberation and reasoning that they conclude in the agent’s
being settled upon some course of action or in some belief about what is the case. To say this is
just to say that part of what makes deliberating and reasoning the kind of mental activities that
they are is that they conclude properly in episodes of deciding or judging, respectively.
I suggest that we conceive of deciding and judging as the final causes of the mental
activities of deliberating and reasoning, respectively. Deciding is the telos toward which
deliberating is directed, as judging is the telos toward which reasoning is directed. This
teleological relationship is grounded in the premise-conclusion relations among the intentional
episodes in the agent’s stream of conscious thinking. The key to avoiding the regress arguments
that greet the instrumentalist reply to the criterial question is to take these premise-conclusion
relations as basic, not to be explained in further terms—in particular, not to be explained by
positing additional mental actions that can serve as the efficient-causal spurs to the episode that
closes deliberation or reasoning.
I do not wish to consider, here, just how to understand the teleological relationship between
deliberating (reasoning) and deciding (judging). For all that has been said so far, we can
understand this relationship in causal-dispositional terms (though I will ultimately argue against
this way of understanding the premise-conclusion relationship). On such a view, we enjoy causal
dispositions to move to a conclusion-episode upon the occurrence of the premise-episodes. Here,
the idea of deliberating (reasoning) as a goal-directed activity is the idea that we disposed to
move to conclusions on the basis of reasons. Even this causal-dispositional construal is able to
avoid the regress that confronts the instrumentalist reply to the criterial question.
In sum, we can respect the goal-directed nature of deliberating and reasoning by construing
deciding and judging as the telos of these mental activities, as their final causes. Final causation
is a concept of goal-direction if anything is. Thereby, we avoid the regress that arises from the
100
suggestion that we perform reflective mental actions in-order-to bring about some deciding or
other while preserving the goal-directed nature of these mental activities.
Judging and deciding as inference
According to the instrumentalist, we determine whether a given mental episode-type belongs
within or without the scope of agency by determining whether the episode in question was
brought about by dint of reliance upon distinct causal-instrumental means. I have argued that this
reply to the criterial question is inadequate to the relationship between an agent’s premises and
the conclusion that one reaches by way of these premises. With respect to this relation the
instrumentalist can return no stable verdict.
This conclusion is significant, for it further underscores the nonvoluntariness of attitude-
formative mental agency. If judging and deciding could be comprehended according to
instrumentalist terms, then we might source the agency in judging and deciding not in attitude-
formative mental episodes themselves, but in the voluntary mental activity of reasoning and
deliberating that eventuate in judging and deciding. Then we might claim that while judging-that-
p and deciding-to-φ are not themselves agentive, they are nevertheless the causal-instrumental
upshot of mental action that is voluntary (in both the reason-based and will-based senses of
‘voluntary’ set out in Chapter 1). This route is closed to us. We cannot suppose that judging and
deciding come about as the nonvoluntary causal effect of voluntary mental action. The regress
arguments described here suggest that there must be some prior exercise of agency that explains
the agent’s performance of reflective mental actions for the reason that such actions will
eventuate in some judging or some deciding or other. And it is this prior exercise of mental
agency—the sort that consists in the premise-guided transition to a conclusion in thought—that
we had hoped to explain in the first place.
101
My proposal is simply to take such premise-guided conclusions as our topic in what follows.
What is it to reach a conclusion via premises? Are such transitions mental actions of any sort? If
so, might their performance legitimate attitudinal blame?
We can begin by labeling more clearly these attitude-formative episodes. Over the course of
this chapter I have put down two of their features. First, deciding-to-φ and judging-that-p just are
occurrences that mark the conclusion of a chain of reasoning. Second, these episode-types
involve an appreciation of the normative force of the premises that count in favor of the action
decided upon or proposition accepted as true. This means that whatever else we say about
judging and deciding, to judge-that-p or decide-to-φ is to reach a conclusion via premises. In
this, judging and deciding both possess an inferential character. As I wrote in Chapter 1, judging-
that-p is always to accept that p is true in the acknowledgement of reasons that show p’s truth,
and deciding-to-φ is always to settle on φ-ing in the acknowledgement of reasons that count in
favor of φ-ing. These reasons, in turn, constitute the premises of an inference to conclusion to
which the judging-that-p or deciding-to-φ corresponds. Thus, we can see every judging and every
deciding as an inferring to a conclusion in the agent’s conscious stream of thought.
The upshot of this chapter’s argument is that if we exercise some blame-legitimating species
of attitudinal agency, it is in inferring to a conclusion. Deciding-to-φ and judging-that-p are
inferences that conclude deliberating and reasoning, respectively. The relationship between
deciding and deliberating, and that between judging and reasoning, is the relationship that a
conclusion bears to the premises on the basis of which that conclusion is drawn. This relationship
is, in the first instance, normative and teleological, rather than causal-instrumental.
Indeed, given condition (C) of the In-order-to relation, the concept of a causal-instrumental
act-complex is conceptually and explanatorily posterior to the premise-conclusion relation that
characterizes inference. And once we recognize this fact, our concern in answering the criterial
question should shift accordingly. Our concern will be to center our inquiry on the peculiarities
102
of the premise-conclusion relation as it surfaces in the conscious stream of thought. Instead of
importing to the mental domain a reply to the criterial question that adequately regiments our
intuitions with respect to bodily action, we might better understand the distinction between the
active and the passive in mental life by taking, as our exemplar of mental activity, the drawing of
an inference.
We can bring this out by noting a nearby distinction: that between following a rule and
merely conforming with one. When I make an inference to a stable belief, I seem to follow
something like an epistemic inference-rule that licenses the transition as one that is preservative
of knowledge, truth, or warrant. When I infer to a stable intention, I follow a practical inference-
rule, one that licenses the action decided upon for features having to do with its all-things-
considered goodness, desirability, or value of the action that I decided upon. Furthermore,
following a rule seems to involve the thinker’s appreciation (however tacit) of the fittingness of
the conclusion’s following from the premises in the light of that very rule. All of this is quite
distinct from cases when I enjoy a succession of conscious intentional episodes that happen to
merely conform to a rule. In that case I might happen to light upon a belief or intention after
having had consciousness of reasons that do in fact recommend the truth of the proposition
believed or the action that I now intend. But if the successions of intentional episodes in my
stream of thought merely conforms with the rule, then it seems I do not judge the truth of the
relevant proposition or decide to perform the relevant action. Rather, I simply passively acquire
the belief or intention at issue.
What informs this intuition? Plausibly, it is the sense that when my intentional episodes
merely conform to a rule the intentional episodes that occur before the onset of the belief or
intention at issue do not play a rationalizing role in the production or genesis of the attitude in
question. If there is no rationalizing role then there are no reasons; if there are no reasons then
there cannot be anything like a premise-conclusion relationship among the intentional episodes
103
that make up the putative reasoning. In fact, there is no reasoning or deliberating at all. Since
there is no reasoning and no deliberating when my stream of thought merely accords with a rule,
we have in this case no teleological relationship between a set of intentional episodes and a
concluding episode that figures as their final end. It appears that the distinction between rule-
following and mere conformity with a rule also corresponds with the presence or absence
(respectively) of inferring itself, defined as it is by the premise-conclusion relations among the
intentional episodes that make up a stretch of reasoning.
All of this suggests a different orientation to the criterial question. Instead of investigating
whether attitude-formative mental episodes conform to what is involved in the voluntary exercise
of intentional action we ought to examine the nature of inferring to a conclusion. Then we can
examine whether inferring to a conclusion is something agentive and legitimating of attitudinal
blame or instead passive. In taking this tack, I will be especially concerned with how it is that in
inferring to a conclusion in thought, a thinker follows a rule that licenses the very transition that
culminates in concluding that p or settling upon φ-ing. To understand our agency over the
attitudes, we must understand inferring on its own terms.
Conclusion: Two new desiderata
In Chapter 1, I introduced three guiding desiderata for an account of attitudinal agency.
First, attitudinal agency must be responsibility-grounding. Our account of attitudinal agency
should specify just why, in virtue of this exercise of agency, an agent comes to enjoy
responsibility for his attitudes. Now, we noted that responsibility for an attitude is keyed to the
specific proposition that an agent believes, or the specific plan of action that the agent intends to
execute. So, attitudinal agency must explain how it is that we are responsible for these items.
104
However—and this is our second desideratum—our account must explain this while
respecting the non-voluntariness of attitudinal agency. One straightforward way to explain
responsibility for the attitudes would be to say, flat-footedly, that an agent chooses to believe that
p, or simply decides to intend to φ. Yet can neither believe nor intend at will. If we could, then
we would be in perfectly good rational standing to believe or intend for reasons of the ‘wrong
kind’—reasons that recommend having the belief that p, or the intention to φ, rather than
recommend the proposition believed or action intended. As I argued in Chapter 1, this is quite
unlike the ordinary responsibility that we bear for the attitudes. The upshot is that attitudinal
agency ground responsibility while refusing to impute to ourselves a purely voluntary form of
attitude-formation.
Third, I argued there that attitudinal agency must be productive in form. Our account of
attitudinal agency must make sense of the fact that attitudinal agency is exercised—at specific
times in the stream of thought—to generate novel attitudes. These attitude-formations are events,
or intentional episodes. Should we fail to capture the productive dimension of attitudinal agency,
we will not be able to make sense of core exercises of attitudinal agency, as when an agent
revises his attitudes in the light of new evidence. This third desideratum rules out the
Answerability View, on which attitudinal agency consists simply in holding an attitude for
reasons. What we must understand is how these attitudes come to be held for reasons.
In light of these desiderata, I suggested we understand attitudinal agency as consisting in an
agent’s performance of mental actions, actions that are productive of the attitudes. This raised a
preliminary question, though: which mental episode-types can count as mental actions? The
present chapter has investigated this question, focusing on a criterion of action that is widely
endorsed in the theory of action. This is the instrumentalist reply to the criterial question,
according to which, on a coarse-grained scheme for the individuation of action, we look to
105
whether the agent brings an intentional episode about by using distinct causal-instrumental
means.
We are now in position to appreciate, though, that such an instrumentalist reply to the
criterial question is inadequate to attitudinal agency. This is because we cannot construe the
relationship between deliberating and deciding, or between reasoning in judging, in causal-
instrumental terms. Judging and deciding are neither direct actions nor non-basic mental
episodes an agent acts to bring about. I suggested instead that we investigate rule-following
inferrings as that in which attitude-formative mental episodes consist.
This leads to a new and fourth desideratum: attitude-formative mental agency must be non-
instrumental in character. Deciding-to-φ and judging-that-p do not take their place as causal-
instrumental means or ends. But this fourth desideratum raises a problem of its own.
Instrumentalism rests on a conception of that which makes an episode count as an intentional
action. The instrumentalist takes the very conception of that which makes an episode actional
from the bodily domain and applies it to the mental domain. Even if the instrumentalist is
mistaken with respect to mental action, one sure advantage of the view is that it offers the very
same criterion of action across the mental and bodily domains. The instrumentalist can
accommodate the unity of agency: for the instrumentalist, the mental actions in which attitudinal
agency consists count as actions on the very same grounds as the ordinary bodily actions we
perform in-order-to achieve our long-run goals and projects.
I have argued we must walk away from the instrumentalist criterion. What does this mean,
though, for the unity of agency? This is certainly a feature we ought to respect, if possible. So we
come to a fifth and final desideratum on a theory of attitudinal agency: it ought to make good on
the unity of agency. Our account of that which makes a mental episode count as a mental action
ought also to apply mutatis mutandis, to bodily actions as well.
106
Chapter 3
Inference and Attitudinal Commitment
Attitudinal agency as inferring
The regress arguments set out in Chapter 2 show that we cannot use the model of voluntary
intentional action to understand the agency we exercise over our attitudes in judging-that-p and
deciding-to-φ. Over this and the following chapter, I pursue in earnest the suggestion that
responsibility-grounding attitudinal agency consists in an agent’s inferring to a conclusion. My
purpose in this chapter is to better understand the difference between inferring to a conclusion
and non-inferential transitions in conscious thought. Only in the next chapter do I argue that
inferring to a conclusion, understood along the lines I set out here, qualifies as a fully intentional
(though nonvoluntary) exercise of rational agency.
I will argue that inferring to a conclusion consists in a kind of rule-following. The best
reason to pursue such a conception is that it makes ready sense of the traditional Taking
Condition on inferring. The Taking Condition stipulates that every inferring involves a thinker’s
‘taking it’ that his premises provide rational support for his conclusion. The notion of rule-
following is well-suited to accommodate this Taking Condition, since it is part of the notion of
following a rule that the agent enjoys some appreciation of the rule that guides his performance.
The difficulty—much addressed in recent work on inferring—lies in just how to specify that in
which an agent’s ‘taking’ consists. In particular, it must be shown that the agent’s taking does not
launch a regress of premises from which no inferring at all can take place.
I will attempt to provide an account of inferring that satisfies the Taking Condition while
avoiding such a regress. I propose that the rules that guide a thinker’s inferring are not separate
premises at all. Instead, the rules are located in the very form of the first-order premises from
107
which the thinker infers to a conclusion. This suggestion will seem natural when we mind the
distinctive first-personal perspective or point of view from which a thinker addresses himself to a
live question in reasoning or deliberating. From that position, a thinker’s attitudes figure as
commitments that register rational permissions or obligations. As irreducibly normative states,
such attitudinal commitments cannot be reduced to the manifestation of causal dispositions. But
they do permit a regress-blocking way of making out the Taking Condition; and this, I will go on
to press in the next chapter, gets the agent into the act.
Before setting out, I want to enter a few points of clarification. To save space, I will run
through the entire argument focused solely on epistemic reasoning—reasoning to a belief. Once
the account is on the table, I will quickly carry the same morals over to the case of practical
deliberating to an intention. Also, I am concerned throughout with conscious reasoning, not with
any sort of informational transactions beneath the level of the thinker’s awareness. This is
because such transactions cannot satisfy the Taking Condition; and anyway, it is not plausible
that they count as active in any sense, let alone the sort of robust intentional agency with which
we hope to ground responsibility for the attitudes. Lastly, I will work only with simple examples
of all-out deductive inference, since concentrating on more complicated sorts of reasoning would
not help to illuminate the basic points I wish to elaborate here.
The Taking Condition
I pointed out in Chapter 1 that reasoning possesses an episodic temporal profile. It takes the
form of an occurrent process, something that goes on in a thinker’s life. Though reasoning has
obvious consequences for a thinker’s persisting epistemic states, in reasoning episodic
occurrences play the primary role. Ontologically, what figure in reasoning are conscious
108
intentional episodes: particular datable content-bearing episodes in a thinker’s conscious stream
of thought.
Reasoning is not the mere idle passing by of intentional episodes, though. Reasoning is
distinct from daydreaming and mind wandering. Part of what makes reasoning distinctive is that
it is mental activity that takes on a final or teleological end: a thinker cannot be engaged in
reasoning unless the activity in which he is engaged will properly conclude in a judging-that-p.
For this to be so, the intentional episodes that figure in reasoning relate to one another in specific
ways (or at least admit of specific types of relations). Namely, the episodes’ contents must be
related as premise to conclusion. The premise-conclusion relation is one of normative or rational
support.
Notice, though, that there is a difference between (on the hand) enjoying an intentional
episode at t that represents a normative relation among contents and (on the other) inferring from
one set of contents to another. A thinker might enjoy an experience of the former without thereby
accomplishing the latter. I might enjoy an intentional episode with the content If Alice is at the
café then Blair is at the café. This episode represents a logical relation between two contents, to
the effect that Blair’s presence at the café is conditional upon Alice’s presence there. But there is
no need to assume that in enjoying such an episode I infer anything at all. Inferring involves
something more than merely enjoying an intentional episode that represents a content that
exhibits normative relations among its sub-propositional parts. There is in inference an additional
dynamic element at work.
To infer is not, then, simply to represent. It is to take a step—make a transition, a shift, a
leap, a movement in thought—to a novel content. It’ll help to have an example before us.
Consider a thinker who performs the following bout of reasoning:
1. If Alice is at the café, then Blair is at the café.
109
2. Alice is at the café.
3. So, Blair is at the café.
In line with what we’ve already said, episodes (1), (2), and (3) are distinct intentional episodes:
tokened at distinct and successive times in the stream of thought. Episodes (1) and (2) constitute
the premises of the thinker’s reasoning, while (3) marks the conclusion with which his reasoning
terminates. To respect the dynamical nature of reasoning, we represent the premise-conclusion
relation as holding not within a single intentional episode but across a chain of them. Episodes
(1) and (2) constitute the premise for the thinker’s conclusion (3), so that the premise-conclusion
relationship is one that spans the intentional episodes within the stream of thought. Modifying a
practice of John Broome’s (2014a), I’ll refer to episodes (1) and (2) as premise-episodes and
episode (3) as the conclusion-episode.
Is this all there is to inferring? Can inferring be understood as a succession of intentional
mental episodes whose contents, taken together, exhibit premise-conclusion relations? No. For
an agent may enjoy a close succession of episodes (1)–(3) without inferring anything at all. The
episodes may simply come across the thinker’s mind without his recognizing that their contents
bear premise-conclusion relations. This is a common enough occurrence in conscious life, or at
least in the conscious life of a dissertation-writer. Call any such case an experience of mere
succession. What is the difference between inferring to a conclusion and undergoing a mere
succession of intentional episodes?
It appears causation cannot provide the missing element. Paul Boghossian (2014) supplies
the following illustration. Suppose a severely depressed person enjoys an intentional episode that
represents I am having so much fun followed an episode that represents But there is so much
suffering in the world. There might be a regular, reliable, and non-deviant causal relation linking
110
these episodes. Yet the depressive does not infer the second from the first. This is mere
succession, not inferring. (I will address a more sophisticated appeal to casual dispositions soon.)
Evidently, what is missing is some species of recognition or acknowledgement on the
thinker’s part of the normative relations holding among the contents of the intentional episodes
that make up his reasoning. To infer is not to experience a mere succession of normatively
related intentional episodes, but to move from one set of episodes to another in the light of these
normative relations. For it seems that only if an agent moves to a conclusion in the light of
premises can we speak of the intentional episodes as falling into ‘premises’ and ‘conclusion’ at
all for the thinker. Furthermore, the inference itself must be taken out of this selfsame
recognition of these relations. Put another way, premises (1) and (2) must serve as the reasons-
for-which the thinker moves to episode (3). Otherwise, the content of the premise-episodes can
play no role in explaining why the thinker infers as he does.
Following Paul Boghossian (2014: 5), call this the Taking Condition on inferring:
Taking Condition: Inferring necessarily involves (I) the thinker taking his premises to
support his conclusion and (II) drawing his conclusion because of that fact.1
Notice that the Taking Condition involves two sub-conditions. Not only must the thinker
appreciate the support his premises provide for the conclusion, he must also move to the
conclusion out of this very appreciation. The thinker’s appreciation of the rational support that his
premises grant to his conclusion must play some explanatory role in the fact that the thinker
concludes as he does. I’ll call these the two sub-conditions on the Taking Condition.
Taken together, these two sub-conditions on the Taking Condition render inferring to a
conclusion a thoroughly normative phenomenon. As Ulf Hlobil (forthcoming) argues, it is a
1 The numerals are my own, added to separate out two different questions we can ask about the Taking Condition:
What does it mean to ‘take’ one’s premises to support one’s conclusion? What is it to draw a conclusion because one
‘takes’ the premises to support the conclusion? I deal only with sub-condition (I) in this chapter.
111
consequence of the Taking Condition that every thinker who infers to a conclusion is committed
to the goodness of his inferring: in inferring to a conclusion that a thinker takes to be supported
by his premises, he is thereby under a rational obligation to accept that the very reasoning
undertaken is good reasoning. Of course, this does not mean that the reasoning is in fact good
reasoning. A thinker’s reasoning might be unsound, invalid, or simply confused. But a thinker
cannot infer to a conclusion unless it is at least apparently a good inference. As we shall see
below, this means that inferring is perspective-dependent—essentially something that is done
from within a thinker’s rational point of view.
The Taking Condition seems to make the right sort of difference with respect to the examples
we’ve considered so far. What distinguishes inferring from mere succession is that when I infer I
meet sub-condition (I): I take the content of (1)–(2) episodes to provide rational support for (3).
Sub-condition (II) guarantees that I move to the conclusion out of this very appreciation of the
rational support provided by the content of the prior episodes. Whereas the severe depressive
who enjoys the intentional episodes (t1) I am having so much fun; (t2) But there is so much
suffering in the world does not appreciate any relation of rational support. And if there is no
relation of rational support at all, there can be no question of coming to the ‘conclusion’ out of
such an appreciation. Prima facie, it seems as though it is (I) and (II) which serve as the
inference-making condition.
But in what does such an inference-making ‘taking’ consist? Just what is it to take the
content of one set of intentional episodes to provide rational support for another?
It is natural to suggest that a thinker’s taking consists in some explicit representing that p
provides rational support for q. And the most natural way of explicitly representing such a
support relation is simply to enjoy a belief, or an occurrent thinking, that p supports q. Following
McHugh and Way (2016), call such a belief a taking-belief. To satisfy both (I) and (II) of the
Taking Condition, the taking-belief must figure explicitly in the premises from which the thinker
112
reasons to the conclusion. This means that we must recast the reasoning (1)–(3) as follows (I
underscore intentional contents for convenience):
4. If Alice is at the café, then Blair is at the café.
5. Alice is at the café.
6. If Alice is at the café, then Blair is at the café and Alice is at the café provide rational
support for Blair is at the café.
7. So, Blair is at the café.
Here, premise (6) is a taking-belief. Its content represents a relation of normative support
between the contents (4)–(5) and the conclusion (7). This taking-belief is one of the very
premises from which the thinker infers to the conclusion. Since this is so, one of the reasons-for-
which the thinker infers that (7) is the taking-belief itself.
As Lewis Carroll (1895) observed, this is just where problems arise. For with respect to this
latest construal of the thinker’s reasoning we can raise the very same question that inspired the
Taking Condition. What makes the thinker’s movement from episodes (4)–(6) to episode (7) a
genuine inferring, rather than a mere succession? In particular, what does the thinker who infers
his conclusion (7) possess that is not possessed by our earlier examples, when the thinker infers
(3) from (1)–(2)? For it seems as though the thinker who infers (7) from (4)–(6) must take the
earlier premises to support this conclusion. If this is right, we face the very same question as
before: In what does the thinker’s taking consist?
We cannot answer this question by positing yet another taking-belief. That would only lead
to yet more premises:
8. If Alice is at the café, then Blair is at the café.
9. Alice is at the café.
113
10. If Alice is at the café, then Blair is at the café and Alice is at the café provide rational
support for Blair is at the café.
11. If (8)–(9) and (10), then Blair is at the café.
12. So, Blair is at the café.
Step (11) constitutes a taking-taking-belief: a belief that one’s taking-belief rationally supports
the conclusion (12). Positing such a taking-taking-belief launches us into a regress. This regress
equally counts against even the taking-belief (6). For the regress-launching taking-taking-belief
demonstrates that taking-beliefs (of any order) can add nothing to the thinker’s reasoning that is
not already present in the first-order intentional episodes (1)–(3) with which we began.
So much for taking-beliefs. What if the relevant species of taking consists in something less
than explicit belief, such as an intuition that a thinker’s premises support her conclusion? (Cf.
Boghossian 2014: §7.) Yet it is not clear how such a taking-intuition can succeed where taking-
beliefs fail. Even if we suppose that the Taking Condition is met by taking-intuitions, it still
seems as though to satisfy the Taking Condition the relevant intuitions must figure in every
inference that the agent makes. As McHugh and Way have it, “it’s not clear how a taking
intuition is supposed to get you to draw your conclusion if not by providing an extra premise”
from which to reason (2016: 320). And that was the very regress-launching problem confronting
fully fledged taking-beliefs.
This leaves us in the following position. The Taking Condition secures the inference-making
features that distinguish inferring from experiences of mere succession. But so far, we have
struggled to understand what the ‘taking’ at issue involves without generating a regress of
premises that will prevent the thinker from ever moving from premise to conclusion.
114
The appeal to rules
On a rule-following conception of inferring, taking consists in a thinker’s being guided by
rules. In the sort of cases we are considering, it is plausible to assume that the rules in question
are inference rules that the thinker implements in carrying out his reasoning. While natural and
compelling, this way of rendering a rule-following account view suffers from the very problems
we have diagnosed with taking-beliefs and taking-intuitions.
For illustration, consider the following explicit case of rule-following. Suppose I am
assembling a piece of furniture and do not know what step comes next. I consult the assembly
instructions, which include step-by-step diagrams for assembling the furniture properly. The
instructions prompt me to gather four wooden planks and arrange them in a certain manner.
Doing this allows me to proceed, moving on to the next step in accordance with the instructions.
Here it seems as though we have direct analogues for the elements of the Taking Condition. The
instructions serve as an explicitly articulated rule (or set of rules) for proceeding in the right way
to the next step in my project. There is accordingly no difficulty in understanding how it is that,
in moving to the next step I (I) take it that the instructions provide rational support for the very
step that I take, and (II) take the next step out of this very appreciation. This is because the rules
written into the assembly instructions are rules for correctly moving from an unassembled heap
of planks and screws to a sturdy piece of furniture.
Compelling though it is, we cannot import this model to help us to specify a thinker’s taking.
Doing so involves attributing to the thinker some mental representation that corresponds to the
instruction sheet—some explicit rule for moving from one state to the next. If it is to play the
same role as it does in our furniture story, this explicit state must represent the very rule that the
thinker means to follow in inferring to a conclusion. It seems that such a state must represent an
115
inference-rule: a rule that licenses the thinker to move from a belief (or set of beliefs) to a
conclusion. Boghossian (2008: 472) represents modus ponens this way:
(MP): If you are rationally permitted to believe both that p and (p → q), then, you are prima
facie rationally permitted to believe that q.
The suggestion is that a thinker consults (MP) in the very same way the builder consults the
assembly instructions. Just as the assembly instructions direct the builder to assemble four
wooden planks, (MP) directs the thinker to infer that q when he believes both p and (p → q). The
thinker’s taking consists in his appreciation that the inference-rule (MP) provides rational
support for inferring that q, because the inference rule (MP) is a rule for correctly moving from a
set of premises to a conclusion that follows from those premises.
This analogy can only describe the thinker’s reasoning if we ascribe to the thinker some
explicit rule-belief that represents (MP). What’s more, we must suppose that this rule-belief
figures in the very reasoning that terminates in the thinker’s concluding just what (MP) directs
him to conclude. Without these two assumptions it will be hard to say just what sort of resource
(MP) really provides for the thinker.
As Boghossian (2008) points out, herein lies our trouble. We have fallen into the very same
difficulty as those theorists who posit a taking-belief or a taking-intuition; accordingly, the very
same regress will greet this (simple) appeal to rules. To see how the regress arises, we can model
the thinker’s reasoning with (MP) as follows:
13. If Alice is at the café, then Blair is at the café.
14. Alice is at the café.
15. (MP): If you are rationally permitted to believe both that p and (p → q), then you are
prime facie rationally permitted to believe that q.
16. Steps (13) and (14) instantiate the antecedent of (MP).
116
17. So, Blair is at the café.
We will naturally raise the very same questions regarding (13)–(17) that we raised earlier about
taking-beliefs and taking-intuitions. What does step (15) provide that was not already given to
the thinker in the earlier reasoning (1)–(3)? It seems that we can imagine these intentional
episodes (13)–(17) taking place without the thinker’s inferring anything at all. What the thinker
requires, we will go on to say, is some additional premise that registers his grasp that (17)
follows from what has gone on before. We will then have to insert into the thinker’s reasoning
some further step, to the effect that steps (13)–(16) license the movement to the conclusion (17).
As we have seen, positing such a further step invites the regress.
This simple appeal to rules, then—wherein we model a thinker’s reasoning by close analogy
with what goes on when a builder assembles a piece of furniture from a set of instructions—is no
better off than proposals we have already rejected. But this is not to say that every appeal to rules
is hopeless. It only points up a constraint on how the appeal is made. We mustn’t attribute to the
thinker any explicit rule-belief that figures as a separate step in his inferring to the conclusion.
Guidance by a rule is not to be construed according to the simple analogy we set out with
assembling a piece of furniture: we do not ‘consult’ inference rules in reasoning the way a
builder consults a set of instructions for what to do next.
The appeal to dispositions
John Broome (2013, 2014a, 2014b) proffers a dispositional account of rule-following that is
sensitive to this constraint. As Broome acknowledges, every rule-following account of inferring
worth its name must respect the idea that inference-rules guide a thinker from premise to
conclusion. Broome’s suggestion is that guidance by a rule consists in a complex disposition; the
117
manifestation of this disposition does not involve imputing to the thinker some explicit rule-
belief that figures as a premise in his reasoning.
According to Broome, such rule-dispositions (as I’ll call them) have two distinct
dispositions as components. The first component is a disposition to make transitions in thought
that conform to inference-rules. For instance, a (rational) thinker possesses a disposition to make
the transition from p to q. The transition from p to q just is the manifestation of this disposition.
As we have seen, though, even a reliably well-functioning disposition of this sort cannot be
sufficient for inferring. The manifestation of a reliable disposition can explain patterns of
intentional episodes that conform with a rule, but not get us patterns that evince a thinker’s
following a rule. This is where the second disposition enters the picture. Broome tells us that
upon the manifestation of the disposition to transition to an intentional episode in accordance
with a rule, we enjoy a second disposition to appreciate the rightness of that very transition. So,
in addition to the disposition to transition from p to q, the thinker is also disposed to appreciate
that the transition from p to q seems right.
Broome insists that this disposition to seem right is not itself a phenomenal feeling or state.
It rather consists in (yet another) complex disposition (Broome 2014a: 22ff). Specifically, the
disposition to seem right is an attitude to (a) the manifestation of the first disposition (i.e., the
transition in thought that is the inferring) and (b) the rule that (putatively) guides the
manifestation of the first disposition. On Broome’s view, this latter disposition essentially
involves the recognition that one’s transition in thought is open to correction by the rule that
guides it. This disposition may be a mere counterfactual disposition, though, since a thinker may
never go through the process of checking his inferring. So, the disposition to seem right is an
attitude towards the transition in thought and the rule that guides that very transition; this attitude
consists in a (counterfactual) disposition to check the inferring for rightness relative to that very
rule.
118
It is one thing to seem right relative to a rule, and another to be right relative to a rule.
Brome accounts for this distinction by building the notion of being right out of that of seeming
right. On his view, to be right relative to rule R is to enjoy a “steady disposition” that inferrings
of this type are right relative to R (2013: 239). To correctly follow R—which may itself be an
incorrect rule of reasoning—is to enjoy a steady disposition of this sort. To correctly follow a
correct rule of reasoning is, correspondingly, to enjoy a steady disposition for inferences guided
by such a correct rule to seem right: “[your] reasoning is correct if and only if you correctly
follow a correct rule” (2013: 247).
Broome has provided a thoroughly dispositional approach to rule-following. To follow a rule
in reasoning is to manifest a rule-disposition. A rule-disposition consists in both the transition in
thought as well as a distinct disposition consisting in the agent’s being guided by the rule. And to
be guided by a rule R is to manifest a (suitably robust) disposition to make transitions in thought
that conform to R and for such transitions to seem right—which in turn consists in a (perhaps
counterfactual) disposition to accept criticism for incorrectly conforming with R. It will be
important for our discussion of Broome’s view that these are separate dispositions: the
disposition to conform to a rule is a separate disposition from the one manifested in the
transition’s seeming right to the thinker. Before moving to that discussion, I want to bring one
more aspect of Broome’s view on the table.
According to Broome, inferring to a conclusion is active—something that an agent does,
rather than something that merely happens to him. Tellingly, Broome claims that inferring is
agentive because it is a rule-guided activity. Consider the thinker who infers q from p. On
Broome’s view, the inference to q will be accompanied by this very inference seeming right. This
seeming right counts, Broome thinks, as a form of endorsement on the agent’s part of the very
transition in thought that has taken place. As Broome has it, a “process’s seeming right to you is
a sort of personal endorsement from you. This is part of what makes the process yours:
119
something you do” (2013: 238). One needn’t endorse the specific rule that guides the inferring; it
is enough, Broome thinks, to believe the content to which a thinker transitions in being guided by
the rule (2013: 233). On Broome’s account, rule-following is agentive because it involves a
thinker’s endorsement of the transition in thought from premise to conclusion; accepting the
deliverance of the rule-disposition is the agency-making endorsement.
It appears that Broome has provided just what was needed: an account of inferring that
distinguishes between rule-following and mere conformity with a rule by appealing to the Taking
Condition. On Broome’s account, it is insufficient for inferring just to enjoy a disposition whose
manifestations conform with inference-rules. The transitions in thought cannot count as
inferrings unless they seem right to the thinker, and this seeming right is the part of the account
that satisfies the Taking Condition. What’s more, it seems that Broome can also explain just why
inferring is something active on the agent’s part, rather than something that passively occurs in
thought—precisely by appealing to non-regress generating notion of rule-following.
These advantages notwithstanding, I wish to part ways with Broome’s proposal (and with it,
the very project of providing a rule-following account of reasoning in terms of dispositions
alone). The problems I will set out for Broome’s proposal arise from his separating into distinct
existences (a) the disposition manifested in a thinker’s transition to a conclusion-episode from
(b) the disposition manifested in this very transition’s seeming right to the thinker.
To set the stage for these worries, note the tight cluster of interconnections among the
notions of rule-following, norms, and reasons (cf. Boghossian 2014). To follow a rule is to treat
it as prescriptive, in at least this minimal sense: in following a rule, what occurs next will be
incorrect according to the rule that I mean to follow if it does not accord in with what the rule
demands. This is just to say that the rule and what it demands must have a normative bearing on
what happens next. To this extent at least, to follow the rule I must treat the rule and what it
demands as constituting a reason for me to act in accord with it—to do as the rule demands.
120
When a thinker infers to a conclusion-episode by following a rule, then, the thinker takes the rule
as demanding (or at least licensing) the very inferring that the thinker performs; the thinker takes
it that he does well relative to the rule and would be doing poorly were he to fail to do what it
demands. In so doing, the thinker takes the rule as normative.
It is to capture these tight interconnections that we impose the Taking Condition and seek a
rule-following account of inferring. The promise of rule-following accounts of the Taking
Condition lies in the hope of recovering these matters of normative significance. In fact, it is an
understatement to say that these notions—that of a rule, of a norm, and of a reason—are merely
‘tightly interconnected.’ It is difficult to see how any one of these notions might be reduced to
another in the network, so that we might define one in terms of the others (or even the network as
a whole in terms of something else). This is just the point that I will press against the idea of
reducing rule-following to a complex set of dispositions.
The chief problem with Broome’s notion of a complex rule-disposition is that it severs these
interconnections by cleaving apart the disposition manifested in moving to a conclusion-episode
from the disposition that is manifested in this transition’s seeming right to the thinker. Once these
dispositions are separated into distinct existences, the thinker’s taking (seeming right) merely
‘runs alongside’ the disposition that is manifested in the transition from p to q. The closest
relation that we can ascribe to these two dispositions is to say that they share a common cause.
There is no reason, though, to suppose that two joint effects of a common cause relate causally to
one another. And if these two dispositions cannot relate (causally or otherwise) to one another,
then one cannot guide the other. The upshot is that the rule that the thinker follows cannot
provide normative guidance for the thinker’s inferring. As I’ve just noted, guidance by a rule
involves treating the rule’s prescriptions as a reason for doing what the rule recommends. And it
is this reason—the reason that is sourced in the normativity of the rule I mean to follow—that
must rationalize (serve as the reason-for-which) my transition to the conclusion-episode.
121
With this, we have effectively re-discovered a motivation for sub-condition (II) of the Taking
Condition, namely that that the thinker infers because of the relations of rational support that he
discerns between the premise and conclusion. When sub-condition (II) goes unmet there cannot
be normative guidance in the sense relevant for rule-following. Should we accept Broome’s rule-
dispositions, the thinker’s grasp of the relations of normative support is merely a byproduct of
the manifestation of the disposition to transition to the conclusion. On this view, the most that we
may say is that every inferring is occasioned by a thinker’s taking it that his premises support his
conclusion. We cannot say that a thinker moves to the conclusion because of this very
appreciation of rational support. This result falls short of what we hoped to explain by appealing
to guidance by a rule.
This first problem feeds into a larger issue concerning agency in inferring. Appealing to rule-
dispositions cannot make good on the idea that rule-following is what renders reasoning active.
To make this second worry vivid, consider yet another conception of inferring. As it happens, it
is a candidate conception that Broome himself wants to dismiss. Broome labels this the Jogging
Account (2013: §13.3). According to the Jogging Account, “in reasoning you call to mind some
of the premises, and doing so jogs into operation an automatic process that causes you to acquire
a conclusion-attitude” in a way that is reliably correct (2013: 226).2 On this view, an inferring is
indirectly brought about by attending to the premises. This exercise of attention sets into motion
some ‘merely ballistic’ process that causes the occurrence of the conclusion-episode. The Jogging
Account is so named because this ‘jogging’ activity is modeled according to the way we often
‘jog’ our memories into throwing up a forgotten item into consciousness. (See my discussion of
instrumentalist accounts of remembering and imagining in Chapter 2.)
2 So described, the Jogging Account is reminiscent of Strawson’s (2003) account of mental agency as consisting
almost entirely in merely ‘ballistic’ cognitive processes. It also accords well with the kind of indirect reflective
mental actions that we rejected in the last chapter. Though Broome does not argue for the intuitive claim that such
automatic processes cannot be actional, the argument in Chapter 2 fills this lacuna.
122
Broome rejects the Jogging Account on the following grounds:
The jogging account is a poor account of active reasoning. If, when you call things to mind
during the course of reasoning, all you do is jog into operation an automatic process,
reasoning would scarcely be an act of yours. Most of it would not be done by you; all you
would do is call the premises to mind. The core of reasoning would be a passive process,
which sometimes needs a jog. But intuitively there is more to reasoning than that. (2013:
226)
Broome rejects the Jogging Account because according to it the agency exercised in reasoning
cannot reach into the inferring itself. At the most it consists in the prompting that sets automatic
processes in motion. According to the Jogging Account we merely act indirectly on our minds. I
share Broome’s intuition. The Jogging Account is a poor model of reasoning to a conclusion, and
it is hard to see, on this model, just why inferring to a conclusion—rather than attending to a set
of contents that cause the occurrence of a conclusion-episode—should count as active in any
sense.
However, Broome’s own view is close to the Jogging Account in relevant respects. The
grounds on which we join Broome in rejecting the Jogging Account count equally against an
account in terms of rule-dispositions. To bring this out, let us supplement the Jogging Account
with Broome’s conception of guidance by a rule. Suppose I attend to the content p. This use of
the attention ‘jogs’ the occurrence of the episode q. The cause of this episode q in turn ‘activates’
a disposition for this very occurrence to ‘seem right’ to me. We may grant that this ‘seeming
right’ consists in the very sort of further complex disposition that Broome envisions. Does the
manifestation of this rule-disposition render inferring any more active? It seems not. The agent’s
action is still confined to the exercise of attention in order to bring about some conclusion-
episode; the occurrence of this episode is entirely passive, and so too is the manifestation of the
123
disposition to see this occurrence as ‘seeming right’ to the thinker. So, it seems that even with the
addition of the rule-disposition, the Jogging Account remains insufficiently active. If this is right,
then it is hard to see how agency really can be located in the manifestation of a rule-disposition.
If the manifestation of a rule-disposition introduced agency into the account, then we should
expect its addition to the Jogging Account to seem prima facie agentive. It appears, then, that the
manifestation of a rule-disposition cannot furnish for our account of inferring the missing
agentive element.
Why is this? I propose that it is because of the first problem we have already pointed out for
the theory. The sort of endorsement that figures in Broome’s rule-disposition view can only be
the joint product of the cause that is common to whatever brings about the occurrence of the
conclusion-episode. The thinker’s endorsement cannot—on either Broome’s rule-disposition
view or the Jogging Account—play any role in guiding the thinker to the conclusion. So long as
this problem goes unfixed, any endorsement on the thinker’s part will be after-the-fact, and an
after-the-fact endorsement of some occurrence does not render the occurrence itself something
that the agent did. Such an endorsement only renders the occurrence something that, having
taken place, the agent now approves of. It seems, then, that we cannot provide a rule-following
account of inferring only in terms of complex sets of dispositions.
Taking as commitment
Our problem remains. While the Taking Condition plausibly captures the distinction
between inferring and mere succession, we have yet to specify that in which a thinker’s taking
consists. Appealing to taking-beliefs and taking-intuitions generates a regress, while appealing to
taking-dispositions cannot satisfy both sub-conditions (I) and (II) of the Taking Condition. We
need to capture the notion of a thinker’s being guided by a rule without thereby positing either
124
some separate intentional state that figures in the thinker’s reasoning or rule-dispositions that
reliably manifest in certain transitions seeming right to the thinker.
For the duration of this chapter I will pursue a positive account of sub-condition (I) of the
Taking Condition with an eye towards these constraints. (Sub-condition (II) will be taken up in
the next chapter, when we turn to an explanation of the agentive character of inferring.) My
proposal will locate the relevant rules within the agent’s grasp of the very first-order premise-
episodes from which he infers to the conclusion. To properly introduce this proposal, though, we
must first expand the theoretical resources we bring to the problem of inferring. To understand
inferring, I want to suggest that we first heed the peculiar first-personal perspective, or rational
point of view, of the thinker who undertakes a bout of reasoning.
THE ENGAGED STANCE
As I indicated in Chapters 1 and 2, as a constitutive matter reasoning is an answer-seeking
activity. Paradigmatically, reasoning terminates with an inferring to a conclusion that answers the
question that the thinker has taken up.3 I will call the questions that invite and sustain inquiry
deliberative questions. Deliberative questions may be epistemic or practical. Epistemic
deliberative questions ask whether p; practical deliberative questions ask whether to φ. What is
it, and what is it like, to take up a deliberative question?
One cannot so much as address oneself to a deliberative question without occupying a
distinctive stance on the subject matter into which one seeks to inquire. From this stance, one is
in position to register and to respond to reasons that bear on one’s deliberative question. From
this point of view, ordinary facts and states of affairs take on more than merely descriptive
3 I say this is so paradigmatically because strictly speaking it is not always true that inferring closes inquiry.
Sometimes, I cannot discern whether p unless and until I have settled whether q; and I may need to infer that q in
order to progress properly to the question whether p. So, while reasoning always terminates with an inferring,
inferring does not always close inquiry.
125
significance. They also take on an immediate normative salience for the thinker. It is from this
stance that a fact or a state of affairs is grasped as having some normative force for the thinker—
as being immediately normatively relevant for the revision and formation of the thinker’s own
further thought and action. I will call this stance the engaged stance of inquiry. So conceived, the
engaged stance of inquiry is distinctively first-personal.4 When a thinker occupies the engaged
stance, it is always with the self-conscious understanding that I am inquiring into whether p from
my own rational point of view.
Following Carol Rovane and Akeel Bilgrami, we should distinguish the engaged stance qua
rational point of view from a conscious subject’s phenomenal point of view.5 The phenomenal
point of view is the perspective from which a sentient being registers qualitative sensations, such
as the felt redness of a red patch of carpet. A subject who possesses a phenomenal point of view
is one for whom there is ‘something it is like,’ phenomenally, to be that subject. The phenomenal
point of view is easy to confuse with the engaged stance, since both are first-personal
perspectives that a subject takes on the world. But it is possible to coherently imagine a
phenomenally conscious subject who inhabits the phenomenal point of view but has no grasp of
reasons or other normative relations.6 Such a creature is wholly reliant on its surrounding
environment for its sustenance, but there seems no conceptual bar to the possibility that it might
enjoy a coherent phenomenal experience of the passing show. At the very least, it would take
substantial philosophical argument to show the incoherence of such a creature; we cannot begin
by assuming that the engaged stance perfectly corresponds with the phenomenal point of view of
a conscious subject.
4 See esp. Burge (2013b, 2013c). 5 See Rovane (1998), Bilgrami (2006: Ch. 4–5). Here as ever I owe much to their work on this topic. The view that
attitudes are commitments stems from work by Isaac Levi. 6 Compare Galen Strawson’s ‘Pure Observers’ (2003), or Bilgrami’s Oblomov (2006: Ch. 4 ). But perhaps these
thought experiments already bring in the close connection between agency and the engaged stance.
126
A rational point of view is different from such a phenomenal point of view. As I have
indicated, it is from the rational point of view that one registers the normative force of
considerations that constitute reasons that bear on one’s deliberative question. Two well-known
illustrations bear this out. Consider, first, Moore’s Paradox. The paradox is that the claims p and I
do not believe that p do not have contradictory truth-conditions. And yet, there is an immediate
rational tension whenever a thinker asserts these two claims. With many, we may understand this
tension as evincing the import of a thinker’s (speaker’s) rational point of view: for it is from that
point of view that a thinker’s beliefs are regulated by the constitutive norm governing belief
(viz., believe that p if and only if p is true). It is only from the engaged stance that a thinker is
subject to the immediate rational tension in asserting p and I do not believe that p. Or, to take a
similar point from Bernard Williams, consider the propositions p and ¬p.7 The assertion of both p
and ¬p can mark a disagreement among two speakers; or, it can mark a logical contradiction.
Whether it marks a contradiction will depend upon whether both p and ¬p are held within a
single rational point of view. Burge’s point is that it is only from within such a point of view that
a thinker can be moved by, and implement, the reasons within his possession.
I have taken care to explicate the engaged stance using examples drawn from epistemic,
rather than practical, inquiry. It is important not to identify the engaged stance with what we
might call the point of view of freedom: the point of view that is appropriate to an agent’s
planning of future-directed action. From the point of view of freedom, one’s own future seems
open, as up to one’s own decisions. When one occupies this point of view, one is concerned to
settle practically the question of one’s future conduct. It is from this point of view that one
arrives at intentions as to what one shall do, rather than predictive beliefs about what one will do.
In this, the point of view of freedom is certainly a rational point of view, and indeed one from
7 See Williams (1978), and Burge’s discussion at (2013b: 388, 393).
127
which an agent registers and responds to reasons. When one occupies the point of view of
freedom, one seeks to settle the question of what one shall do with reference to one’s
assessments of what is good, right, desirable, and valuable. This cannot be done unless one is in a
position to register and treat with reasons. This means that an agent cannot so much as occupy
the point of view of freedom without also taking up the engaged stance of inquiry. Yet the two
points of view cannot be identified: it is possible to take up the engaged stance of inquiry without
thereby occupying the point of view of freedom.
To see this, consider those cases in which one enters the engaged stance by taking up an
epistemic deliberative question regarding one’s own future conduct. Sometimes, reaching such a
prediction can inform one’s course of practical deliberation. Suppose I receive a dinner invitation
tonight from an old friend passing through town. But I also have an important meeting tomorrow
for which I must prepare this very evening. I then enter the point of view of freedom, to decide
upon a practical course of action: shall I go to dinner, or instead stay home to prepare for my
meeting tomorrow? I might think: What I really ought to do is both: attend the dinner and simply
work late into the night upon coming home. This would be an ideal plan, but for the fact that I
have a terrible habit of failing to return to work at night after returning home from evening
socializing. My record in this respect is abysmal, and I know this about myself. Given this, it
makes sense for me to ask an epistemic, rather than a practical, question about how I am likely to
behave in the future. If I am unlikely to work after returning home, it behooves me to decline the
dinner invitation. To be sure, in arriving at the prediction that informs this choice I have taken, in
some sense, a passive or merely observational stance with respect to my own future conduct.
Nevertheless, I have arrived at the prediction by taking the engaged stance to worldly facts and
states of affairs that bear upon an epistemic, rather than a practical, deliberative question. I have
concluded as I have in view of strong inductive evidence that constitutes a compelling case for
thinking that I will not in fact work after returning home from dinner. In contrast, the point of
128
view of freedom seems not to allow me to look upon my own future conduct through such a
merely predictive lens. When I occupy the point of view of freedom, my future course of action
seems to be a matter that can be settled (or at least strongly influenced) by my undertaking a
present commitment to resume my work after returning from dinner. Given this disparity, the
engaged stance must be recognized as something more general than the point of view of
freedom, in that one can inhabit the former without thereby stepping foot in the latter. This helps
to clarify the engaged stance as a distinctive perspective that makes reasons available for
reasoning; it is not already a point of view wherein one settles practically what one shall do.
(Though, to be sure, the inverse implication does hold: if one takes up the point of view of
freedom, one must therein occupy the engaged stance of inquiry.)
Now, introduction of the notion of a stance or a rational point of view raises a host of deeply
controversial philosophical matters that cannot be addressed here. Still, I hope that it is clear
enough that a full specification of the Taking Condition must introduce some such notion of a
rational point of view. As Williams’ point and Moore-type paradoxical statements seem to
evince, we need the notion of a point of view in order to make sense of the Taking Condition.
What can a thinker’s taking be, if not something locatable from the thinker’s rational point of
view?
Having introduced this notion of an engaged point of view, I turn now to drawing out its
significance for inferring. My claims will be three: All inferring happens from the engaged
stance; From this stance, one’s intentional attitudes figure as normative commitments on the
thinker’s part; Qua normative commitments, these attitudes take on an imperatival rule-like
character insofar as a thinker’s attitudes enjoin him to discharge attitudinal commitments in
further thought and action. The upshot: when a thinker is in a position to infer, his intentional
episodes are themselves rules for reasoning. By understanding the imperatival dimension of
beliefs and intentions, we locate all the normativity required for sub-condition (I) of the Taking
129
Condition without having to posit any separate intentional state in which the thinker’s taking
consists. I’ll take these three claims in turn.
INFERRING AND THE ENGAGED STANCE
I have already foreshadowed the first of these claims and will quickly dispense with it.
Every inferring is undertaken from within the engaged stance. Inferring is reasoning, and
reasoning is constitutively an answer-seeking activity. It is the quest for an answer to a
deliberative question that distinguishes reasoning from idle daydreaming or aimless mind
wandering. To take up a deliberative question is to orient oneself towards considerations that are
relevant to one’s answer, so that those considerations can be implemented in the course of
inquiry. If one cannot do this, then one can merely contemplate the question rather than earnestly
pursue an answer to it.8
To see the same point from another angle, consider: what is it like to fail to occupy the
engaged stance? Such a stance will not be first-personal at all. Even if from such a stance a
thinker can recognize reasons, the thinker will be unable to implement those reasons in the
formation or revision of attitudes.9 If a thinker is not at all able to implement reasons in inquiry,
then there can be nothing genuinely at stake for the thinker’s progression of thinking. From
within the confines of such a disengaged stance, a thinker can be only a passive observer of his
own thoughts. They pass by, adrift and undirected, to assail the thinker.10 And if the thinker
cannot so much as register and respond to reasons, there is no way in which the thinker can
8 I do not endorse any sort of verificationism, according to which understanding the meaning of a proposition
consists in a grasp of what might count as evidence for it. No claim here is being made about understanding
meaning. An agent, in contemplating a deliberative question for which he knows not what could count as a
normatively significant consideration that bears on that question, can still understand what the question asks. My
claim is just that the agent cannot make real progress inquiring into that question until he has some sense of what
could count as a reason for or against some possible answer, or range of answers, to it. 9 For these points, see Burge (2013a). 10 Compare Bilgrami’s thought-experimental character Oblomov, in (2006: 161–166).
130
satisfy the Taking Condition. So long as we impose the Taking Condition on inferring, then, we
must also recognize that inferring can only happen from within the engaged stance.
ATTITUDINAL COMMITMENT
From the engaged stance, a thinker’s attitudes do not figure as mere descriptive facts that are
true of him. By ‘descriptive facts,’ I mean such ordinary facts as the fact that I am over six feet
tall and born of American parents. These facts are fully graspable from a third-personal point of
view. To be sure, a thinker can conceive of features of his own mind in just this way. I can take a
detached point of view with respect to my own mind by reporting that I become annoyed by
tardiness and frustrated with messiness.
In contrast, from the engaged stance a thinker’s attitudes figure as normative commitments
on one’s part. I call these attitudinal commitments.11 A thinker’s attitudinal commitments register
his rational obligations or permissions. I want to elucidate the notion of an attitudinal
commitment by minding the fashion in which one’s extant attitudes figure, from the engaged
stance. But I caution the reader at the outset that this is a mere expository strategy: as we shall
see, attitudinal commitments are often themselves the product of a thinker’s inferring.
In taking up a deliberative question, a thinker brings a set of beliefs, desires, or intentions to
bear on that question. That is, one always addresses oneself to a deliberative question from some
prior deliberative position: a network of extant attitudes whose contents bear on the question that
one seeks to answer. So, if I have addressed myself to the question whether q, and already
believe that p and that (p → q), then the deliberative position I bring to bear on whether q
includes the beliefs p and (p → q). Often one’s deliberative position is rather more impoverished
11 In addition to Rovane (1998) and Bilgrami (2006: Ch. 4), see also recent work by Pamela Hieronymi, who writes
of belief and intention as ‘commitment-constituted attitudes’: attitudes that embody an agent’s reply to a deliberative
question. Hieronymi (2005, 2006, 2009). See also Moran (2001) for a similar characterization of ho w our attitudes
figure from the first-personal stance of engagement.
131
than this. There is also the possibility that one’s deliberative position itself undergoes refinement
or revision in the course of inquiry. Still, no account of inferring can pass over the significance of
the prior deliberative position. That is because inferring is conceived of as a movement from
some set of content-bearing intentional episodes to a novel intentional episode. The intentional
episodes from which one infers are always (at least partial) elements comprising one’s prior
deliberative position. Put another way: elements of a thinker’s prior deliberative position figure
as the premises from which he infers to a conclusion. The question is how, from the first-personal
engaged stance of inquiry, this prior deliberative position figures for the agent carrying out the
inferring.
The premises of one’s inferring certainly do not figure as causally-efficacious states
impelling the thinker to some conclusion or other. One simply does not ascribe to the items
within one’s prior deliberative position this species of merely ballistic significance. To do so is to
construe one’s inquiry as a matter of hurling oneself along by the causal force of one’s states and
episodes. This is not to say that the states and events that realize transitions in reasoning play no
causal role in the generation of novel contents. Neither is it to foreclose the possibility that, as
causal theorists of action maintain, reasons-explanations are a species of efficient-causal
explanation. It is just to say that from the engaged stance, as one is taken up with a deliberative
question, the attitudes from which one reasons are not conceived of as causal-instrumental means
of bringing about some novel intentional episode. That is precisely the mistake we diagnosed in
the instrumentalist account of mental action back in Chapter 2; it is also what gives rise to the
Jogging Account that we have followed John Broome in rejecting. As Stuart Hampshire puts the
point, a person “who changes his mind, in response to evidence of the truth of a proposition, does
not act upon himself, nor does he bring about an effect” (1975: 100).12 Even if this ‘change in
12 I came to this apt quotation by way of Moran (2001: 132, n. 11).
132
mind’ involves the bringing-about of some causal effect (e.g. at a neuronal or sub-personal level
of description), the first-personal engaged stance of inquiry is not one from which I am
concerned to employ such causal relations to bring about an intended effect in my own mind.
Rather: I grasp the content of the episodes that comprise my prior deliberative position as
having a normative bearing on further thought and action. In fact, it is already a mistake to speak
of attitudes as figuring as such in the engaged stance, at all. In the first place, it is worldly facts
and states of affairs which are taken to bear normative significance for one’s inquiry. I do not
reason from the fact that I believe that Alice is at the café, but from the (apparent) fact of Alice’s
being present at the café.13 It is not psychological states that are ‘reason-giving,’ but the worldly
facts and states of affairs represented in their contents which provide reasons. As Gareth Evans
observed, our attitudes are directed outwards, towards the facts and states of affairs that the
attitudes work to put us in touch with.14
Obviously, this does not mean that my own attitudes are inaccessible to me from the
engaged stance. It only means that reasoning does not and cannot begin by considering what
must be the case given one’s own psychological states. To begin inquiry from a determinate
deliberative position is to begin inquiry with a certain picture of what is the case, of what the
world is like.
My prior deliberative position is given to me, from the engaged stance, in terms of
normatively salient worldly facts and states of affairs. This means that my prior deliberative
position at once provides a normative direction for my inquiry as it progresses. If I already
believe that Alice is at the café, this (apparent) fact—that Alice is present at the café—can be
treated as a reason that bears on a range of further deliberative questions. And in this, the
13 This is a point that recent ‘anti-psychologist’ philosophers of reasons and normativity have pointed out. See e.g.
Jonathan Dancy (2000) and Maria Alvarez (2010). 14 See Evans (1982) and Moran (2001). Broome is also onto this same point in insisting that inferring is an operation
on contents of attitudes, not on attitudes per se (2013: Ch. 13).
133
(apparent) fact that Alice is at the café is a state of affairs that my further inquiry mustn’t
contradict, as such a contradiction amounts to an incoherent picture of the world. It is here that
we arrive at the notion of an attitudinal commitment. If p is represented by an intentional episode
that figures in my reasoning, then I am obligated to ‘live up to’ p in subsequent thought. What
does it mean to ‘live up to’ p in subsequent thinking? At a minimum, it means refusing to
endorse anything that outright contradicts with p. It also means endorsing p’s implications,
material and logical. This is what it means to say that my occurrent intentional episode is more
than merely a relation to a proposition, but a commitment that I am concerned to uphold. Akeel
Bilgrami fittingly describes beliefs as internal oughts: the belief that p incorporates the rational
obligation to accept the consequences of p and to reject what is inconsistent with p (2006: 212).
This is an apt construal of an attitudinal commitment.
In this, attitudinal commitments are akin to interpersonal commitments familiar from
everyday ethical life. Suppose I promise to help you move to a new uptown apartment on
Saturday afternoon. My promise to you inaugurates a range of commitments on my part, which
figure as oughts that I am obligated to observe in further thought and action. To take some of the
most obvious: I ought not to make alternative plans on Saturday afternoon; I ought to stay in
Friday evening to conserve energy, and come Saturday I ought not to lie to you to escape my
obligation to help you move that day. These commitments all arise from the fundamental
obligation to help you move on Saturday. I am obligated to live up to this commitment by
discharging it in the very action of assisting you with your move. (What it means to ‘discharge’ a
commitment is not something we can explain without having on the table a conception of
inferring. For it seems that to discharge a commitment is just to do what the commitment obliges
me to do, out of an appreciation of the normative force of the very commitment that I discharge
in action. My proposal over this and the following chapter is that to infer, and to intentionally
act, is just to discharge a commitment.)
134
Come Saturday, I can do well or worse relative to my promise to help you move. One way
of going worse will be to disregard my promise to you and stay home on Saturday afternoon. My
commitment to assist you brings in train the possibility that I might fail to live up to this very
commitment when the time comes. A commitment to act involves the possibility of falling short
of it. However, falling short of a commitment does nothing to impugn the presence of the
commitment itself (see Bilgrami 2006: Ch. 4, §§3–4). If I leave you in the lurch on Saturday,
there is no conclusive ground to judge that I never really committed to helping you move at all.
There is only conclusive ground to criticize me for falling short of my commitment—for not
doing what, by my own lights, I ought to do. In cases where a person falls short of his
commitments, the mistake is in the performance, as Anscombe had it, and not in the commitment
itself (1957: 57).
This differs sharply with causal dispositions. Unlike normative commitments, a causal
disposition’s failure to manifest does impugn the presence of the disposition in question. Take a
paradigm exemplar of a disposition: the fragility of a plate. The plate’s fragility consists in its
being disposed to shatter in a range of specifiable conditions. But suppose that the plate persists
intact after being thrown, kicked, dropped, and otherwise abused. This will constitute evidence
that the disposition to break was mistakenly attributed to the plate, and that it is in fact not fragile
at all, but rather quite durable. In general, an entity’s reliably failing to manifest a disposition in
relevant circumstances counts as evidence that the entity does not really possess the disposition
in question. We will come to judge that the plate is not fragile after all, but instead rather durable
(fit for camping and not fine dining).15
15 This assessment is complicated by the fact that the manifestation of a disposition might be ‘masked’ by a range of
conditions that suppress or otherwise prevent its manifestation even in the presence of the relevant conditions. For
discussion, see Schlosser (2011). This complication can only bear on the present argument if it is assumed that
failure to discharge a commitment can always be subsumed to such cases of masking. To say this, though, would be
to embark on a reductive project of explaining normative commitments in thoroughl y causal-dispositional terms,
which is the very approach I sought to reject in distancing myself from Broome’s account of rule -dispositions.
135
As I’ve already indicated, though, our responsibility-practices evince quite a different
pattern of evaluation and revision in the case of commitments. Should I fail to discharge my
commitment, the appropriate response is not to conclude that I never enjoyed the commitment at
all, but rather to criticize me for failing to live up to what, by my own lights, I ought to do. This is
not to say that it is never appropriate to conclude that I never really had the commitment at all to
help you move. If you find that I am not at all disposed to accept criticism for my failure, nor to
acknowledge any fault or to re-commit to doing better in the future, then it may be appropriate to
conclude that I never really had the commitment at all.16 Yet this pattern of evaluation and
revision only serves to underscore the difference with causal dispositions: once more, we find
that in the case of normative commitments the error is located in the performance, rather than in
the attribution of the disposition.
With Rovane and Bilgrami, I am suggesting that this very notion of a commitment
characterizes the attitudes that comprise a thinker’s prior deliberative position as he addresses
himself to a deliberative question from the engaged stance. It is just that the commitments at
issue are more broadly rational, rather than ethical—but no less normative for all that. From the
engaged stance, my belief that p is a commitment to p’s truth. My commitment to p’s truth is
something that I can dutifully discharge or something I can fail to live up to. I discharge my
commitment to p by accepting p’s consequences and by rejecting those propositions that conflict
with p. Just like promising, though, the commitment to p’s truth is something of which I can fall
short. This will be the case when I refuse, evade, or delude myself with respect to p’s
consequences; or when I maintain both p and ¬p (or a separate belief that I recognize to conflict
with p). Crucially, though, falling short of my commitment to p does not count as conclusive
evidence that I do not really believe that p. In such cases, the mistake will be in my performance
16 See Bilgrami (2006: Ch. 4, §4).
136
(in the normative sense pointed up by Anscombe). The possibility of falling short—of making a
mistake, of standing in need of correction—is a direct and ineliminable consequence of
conceiving of intentional episodes in these normative terms. This means that my commitment to
p renders me liable to a distinctive pattern of rational appraisal.17 All of this differs sharply with
the case of causal dispositions. From the engaged stance, the items within my prior deliberative
position do not figure as causal dispositions, but as normative commitments that I aspire to live
up to. Attitudinal commitments are commitments of just this kind.
ATTITUDES AS RULES
So far, I have argued that from the engaged stance the beliefs that make up a thinker’s prior
deliberative position take on a normative profile. They figure as a suit of rational permissions and
obligations that a thinker must live up to in subsequent epistemic dealings. We are now in a
position to show how these attitudinal commitments furnish the rules for a thinker to reason by.
Suppose my prior deliberative position includes not only the attitudinal commitment to p,
but also to r. This latter belief involves its own normative profile, given to me from the engaged
stance in terms of a suite of rational obligations and permissions. Just as I am committed to
accept the consequences of p and to reject that which p contradicts, so too I am committed to r’s
consequences and to the rejection of its contraries. This will be the case for all of the attitudes
that figure in my prior deliberative position—all of the attitudes that have a bearing on the
deliberative question that I am addressing in inquiry. So, my commitment to p and my
17 The reader will wonder: is this not the core of the very sort of answerability for belief that I am attempting to
ground? I shall address this circularity worry only after the account of sub-condition (I) of the Taking Condition is
on the table. As I will go on to address more fully: here I am characterizing the nature of intentional attitudes; the
question of that in virtue of which intentional attitudes have this nature is a separate, and more fundamental,
question—one that I seek to answer by way of the irreducibly normative account of inferring issued here.
Circularity worries only get a grip if we assume a foundationalist account according to which normative
commitments are got from some non-normative causal disposition, which I think is not possible given the previous
arguments.
137
commitment to r must agglomerate together into a commitment that incorporates the truth of p
and the truth of r. My subsequent thought and action must live up to the commitment to (p ˄ r),
and not just one of these commitments alone. We find the very same agglomeration in the case of
ordinary ethical commitments. If I commit to helping you move on Saturday afternoon and to
reading a student’s essay draft Saturday morning, then I must live up to both of these
commitments on Saturday. The very same species of constraints and oughts that spring from my
commitment to helping you move will also spring from my commitment to reading the draft
essay in the morning.
But now suppose that I learn that r has a consequence that conflicts with p. I cannot live up
to both of these commitments, because I cannot discharge a commitment to a contradiction or
other type of inconsistency. (I am happy to take this as a brute, basic fact about belief and
epistemic norms.) So, I must resolve this tension, and do so in the direction of fidelity to the facts
as I am best able to discern them. From the engaged stance, then, I am gripped by the demand to
achieve coherence and consistency among my commitments. Rovane calls this a demand to
achieve rational unity. I will achieve rational unity among my attitudinal commitments by either
abolishing the commitment to p, or to the commitment to r, or to inquiring further as to whether r
really does bring in a conflict with p.
From the engaged stance, then, my commitments on the whole agglomerate into overall
demands to achieve rational unity among all of the attitudes from which I am prepared to reason.
This is why, in Rovane’s terms, the engaged stance is one from
which contradictions and conflicts ought to be resolved, implications ought to be accepted,
preferences ought to be ranked, means and consequences ought to be taken into account—
138
and more generally, it is something from which all-things-considered judgments ought to
be reached and implemented. (1998: 21)18
Rational unity is achieved when the thinker’s intentional states and episodes satisfy the
normative prescriptions of epistemic (as well as practical) rationality. An outright contradiction
among my beliefs, or a failure to accept that which (by my own lights) my beliefs entail,
amounts to rational disunity. Rovane’s characterization underscores another face of the
normativity of my prior deliberative position, this time with respect of the global demands for
consistency and coherence among all of a thinker’s attitudes. The demand to achieve rational
unity is a mark of the engaged stance, one that shows up perspicuously when we consider the
fact that rational commitments agglomerate together.
In sum, the engaged stance is normative in two closely related respects. First, it is normative
in that it is from the engaged stance that I am put in touch with worldly facts and states of affairs
qua reasons that take on an immediate force for me. The second normative aspect of the engaged
stance has to do with the prior deliberative position from which I take up a deliberative question.
This prior deliberative question carries its own normative significance: it comprises a set of
normative commitments which agglomerate into an overall demand for rational unity.
Let us observe how the demand for rational unity plays out when a thinker addresses himself
to a deliberative question. I am concerned to reason to Blair’s present location. My prior
deliberative position includes the beliefs that If Alice is at the café, then Blair is at the café and
the belief that Alice is at the café. The second of these beliefs involves a commitment to accept
its consequences; the first belief involves a commitment to accepting Blair is at the café as just
one such consequence. These two commitments agglomerate together into a rational demand to
18 I am unsure whether Rovane would agree that the overall demand to achieve rational unity arises from the specific
normative demands embedded in our attitudinal commitments. I have stated things this way because it coheres
nicely with a ‘reasons-first’ approach to rational norms that I find otherwise persuasive.
139
discharge these commitments by accepting that Blair is at the café. It is by discharging these
commitments with the inference that Blair is at the café that I live up to them in subsequent
epistemic dealings.
The moral I wish to draw from this is that from the engaged stance, a thinker’s first-order
intentional episodes are already rule-like. My belief that p is a commitment to uphold the truth of
p in further thought and action; the same goes for all of the beliefs in my repertoire that together
impose an overall rational demand for unity among my commitments. To occupy the engaged
stance is to be in a position to recover and to receive the normative force of these rational
commitments. I cannot so much as set out on my inquiry without addressing myself to a
question; I address a question from a prior deliberative position; and, from this position, I am
concerned to discharge my attitudinal commitments. This means that a thinker needs nothing in
addition to the first-order intentional episodes that make up the premises of his inferring in order
to grasp the relations of normative support that are relevant for the Taking Condition.
This allows us to say, finally, that in which a thinker’s ‘taking’ consists. It consists in an
appreciation of the obligations rooted in the attitudinal commitments making up his prior
deliberative position. The mistake is to search for some separate premise to include within the
thinker’s reasoning. It is a mistake to do so because this strategy overlooks entirely what it is
like, from the first-personal point of view of the reasoning thinker, to enjoy beliefs from which I
am prepared to reason. The premises that comprise my prior deliberative position are themselves
rules to reason by. To take one’s premises to provide support for the conclusion is to grasp that
the conclusion is a consequence of an attitudinal commitment that one has undertaken. It is in
this sense that we can say that attitudes are rules. Since there is no extra premise attributed to the
thinker about what follows from what, there is no threat of regress.
I’ll call this the commitment-theoretic account of the species of taking that is significant for
the Taking Condition. I emphasize that my discussion in this chapter has been confined to sub-
140
condition (I) of the Taking Condition, which requires of every inferring that the thinker take his
premises to provide rational support for his conclusion. My point has been that this taking is not
a separate premise at all, but simply the committal dimension of the attitudes as they appear to us
from the engaged first-personal stance. Another noteworthy feature of the commitment-theoretic
approach is that it makes out a clear sense in which a thinker is guided by rules when reasoning
to a conclusion. These rules, though, are not explicitly represented as a separate premise that
makes a claim about what follows from what. They are rather encoded in the very first-order
attitudes that constitute a thinker’s reasoning.
Importantly, the commitment-theoretic approach respects the deep interconnections that I
noted earlier between rule-following, norms, and reasons. We said earlier that when an agent
follows a rule, he treats the rule as prescriptive of what he ought to do next, so that the thinker
treats the rule as a reason for doing what it recommends. These interconnections explain why,
when the thinker follows a rule in reasoning to a conclusion, he takes himself to be doing well
relative to the very rule he means to follow. The present account sources these interconnections
in the thinker’s attitudinal commitments. When a thinker discharges an attitudinal commitment,
he also takes himself to be doing well relative to the demand for rational unity by living up to the
very commitments he discharges in reasoning. We needn’t ascribe to the thinker a specific belief
about what follows from what to account for this. We only need to point out that inferring is self-
conscious: that the thinker, in inferring, knows just what he is doing. This is a matter I’ll address
more fully in the next chapter.
141
Worries and refinements
With the main contours of the commit-theoretic approach on the table, I want to address
some possible worries and introduce some refinements into the account before drawing
analogous conclusions about practical deliberating.
THE PLACE OF INFERENCE RULES
One immediate consequence of this commitment-theoretic approach to ‘taking’ is that it
removes from the thinker’s ken any explicit representation of inference rules such as (MP):
(MP): If you are rationally permitted to believe both that p and (p → q), then, you are prima
facie rationally permitted to believe that q.
We have already seen why (MP) causes trouble when it is enlisted as an explicit premise in a
thinker’s reasoning. But can we really dispense with them altogether in an account of what goes
on when a thinker reasons to a conclusion?
I am suggesting that we can. A thinker does well by the lights of such inference rules when
he discharges those of his attitudinal commitments which count as instantiations of such
argument-forms. But they need not be grasped as such by the thinker who infers in accord with
them. To suppose that we learn rules the other way around—by first grasping their form, and
then coming to understand that some particular combination of contents instantiates that form—
will make the inference-rules unlearnable. If explicit articulations of (MP) have a role in our
concrete epistemic dealings, it is by grasping their form in some first-order content. Otherwise,
the attempt to learn an abstract inference rule like (MP) will presuppose the very thing we are
attempting to explain.
142
IMPERATIVAL ASPECTS OF BELIEF
On the commitment-theoretic approach to taking, attitudinal commitments take on an
imperatival aspect. A thinker cannot take up a deliberative question without occupying a prior
deliberative position. And to occupy a position to reason from, one must appreciate that the
attitudes in the prior deliberative position enjoin the thinker to live up to the individual attitudinal
commitments as well as the overall demand to achieve rational unity. From the engaged stance,
my belief that p possesses an imperatival dimension: the belief that p also calls me to live up to it
in further epistemic dealings. This imperatival aspect really does fall directly from the idea that
there are attitudinal commitments.
One might worry about construing a thinker’s beliefs as imperatival commitments on a
thinker’s part. Ordinarily, imperatives are communicated in a distinctive mood that differs
sharply with indicative expressions. Open the door! is a familiar indicative; John opens the door
a familiar indicative. Am I suggesting that all of the beliefs within a thinker’s prior deliberative
position are cast in an imperatival mood? If so, this would threaten an extraordinary revision to
our usual way of ascribing beliefs to thinkers; after all, the ‘that’-clauses we use to ascribe such
beliefs employ an indicative mood, and it would be odd if such ascriptions were to be held to
fundamentally misconstrue the contents of the attitudes ascribed.
Luckily, no such revision is forced on us. Indeed, such a revision of ordinary belief content
is inadvisable on its own terms: by itself such a proposal does not help to inform us of that in
which the imperatival dimension of belief consists. Suppose we do revise the content of every
belief to include an indicative proposition and an imperative committal element. Such a revision
might go this way: the contents of ordinary beliefs such as (a) Alice is at the café can be recast as
having the content (a*) Alice is at the café, and uphold this very belief!. But with respect to the
content of (a*) we can ask: In what does its imperatival dimension consist? We certainly cannot
143
further revise its content to (a**) Alice is at the café, and uphold this very belief!—and uphold
this very belief in turn!. This move courts the very same regress worries with which we are by
now familiar. So, it can be no help to us to add some imperatival ‘tag’ to the ordinary indicative
contents related within ‘that’-clauses.
The imperatival dimension of belief is located not in the content of any particular belief, but
in the very form that makes believing distinctive (cf. Kietzmann 2018). If believing is a way of
relating to a proposition, the imperatival dimension of belief is rooted in what is special about
this, rather than some other, way of relating to a proposition. What it is to believe that p—rather
than hope, suppose, or dread that p—is (at least in part) to come in under the normative
commitments that I have described so far. One cannot properly be said to believe that p without
also appreciating the normative commitment to uphold p’s truth in further epistemic dealings by
e.g. accepting p’s consequences and resolving tensions with p among the other attitudes in one’s
repertoire. From the engaged stance the very modality (or form) of belief possesses an
imperatival aspect. From the engaged stance, one’s belief that p is not the mere representation
that a fact obtains, but also comprehends a normative commitment to upholding the truth of this
fact in further epistemic and practical dealings.
COMMITMENT AND BELIEF: TWO WORRIES
I want to turn now to two worries about what the commitment-theoretic approach to taking
may imply for what it is an agent believes.
The first worry is that the commitment-theoretic approach threatens to overburden belief. If
belief does possess such an imperatival dimension, then it seems as though I must grasp and
pursue these imperatives whenever I am cognizant of any one of my beliefs—or at least, any one
of the beliefs that figures in the prior deliberative position I take with me into the engaged stance.
But any one of these beliefs will have its own vast array of implications and entailments; and for
144
all I know, each one of my beliefs may have implications of its own that generates conflicts with
other beliefs that do not presently occupy my attention. This poses a remarkable burden upon me
as I go to prosecute my inquiry into a deliberative question. It begins to appear as though the
imperatival dimension of the beliefs within my prior deliberative position will frustrate any
attempt to take the truth of those beliefs for granted in reasoning to new conclusions. The worry
is that as a practical matter, the imperatival dimension I am attributing to the form of belief will
undermine any attempt to prosecute a deliberative question in inquiry.
This burden is eased, though, by two important features of attitudinal commitment. For one
thing, there is nothing in this account requiring me to be cognizant of all my epistemic
commitments. Rather, the only imperatives relevant for me are those that lie within my ken—
those which are accessible, reasonably held, appropriately sensitive to evidence, and so on. As I
put it above, I treat my belief that p as a commitment that I must live up to in my further
epistemic dealings. All that this requires is that I am prepared, on pain of criticism, to live up to
this commitment when the right sorts of circumstances arise—when the relevant sort of
epistemic dealings are undertaken.
For another thing, one may also fall short of one’s commitments, and in that way fail to live
up to them. This possibility is part of what it means to have a commitment at all, when we think
of commitments as irreducibly normative attitudes rather than those with a merely causal or
dispositional significance. Clearly, such an agent is irrational in falling short of his attitudinal
commitments. Every view must allow room for such failures in irrationality.
This last consideration might tease the deliberative burden that I am claiming agents do in
fact carry while retaining the burden that agents are rationally obligated to undertake. This latter
burden is eased by the presence of what we might call rational excuses: special extenuating
circumstances that comprehend the inevitable limitations in an agent’s information, perspective,
145
expertise, and other factors.19 Such rational excuses come into play because many of the logical
consequences of p will simply be unknown to the agent. And some of the logical consequences
of p can surface only after lengthy investigation on the agent’s part, or in the light of technical
expertise that the agent lacks and cannot be fairly expected to carry out. When an agent falls
under such a rational excuse, it does not count against his commitment to p that he fails to
endorse one of p’s logical consequences. What matters when the agent falls under such an
excusing condition is that the agent is prepared to endorse the logical consequences of his belief
should he be made aware of them, and that he takes himself to come in for legitimate criticism
should he fail to do so. When this condition is in place, ignorance is a legitimate excuse: in such
cases there is no reason to fault the agent for failing to live up to his attitudinal commitments.20
Such rational excuses remove the burden of inquiring into the logical consequences of one’s
extant attitudes ad infinitum, allowing agents to (rationally) move to the deliberative matter at
hand.
However, taking this line raises another question—a second worry about the relationship
between belief and commitment. The reply I have just entered trades on a distinction between
that which an agent believes and that to which an agent is attitudinally committed. An agent
might believe that p; and it may be the case that, unbeknownst to the agent, p implies q; the agent
is certainly committed to q, though the agent does not know this. This is because an agent is
attitudinally committed to all of the implications of his beliefs, whether he knows them or not.
The worry I want to consider now concerns whether this separation is licit. For it rests upon a
distinction between what an agent believes and what he is committed to; and it may be suggested
19 Here I recapitulate the solution of Bilgrami (2006: 372, n. 8). 20 This is not the place to generate a final accounting of the kinds of rational excuses that figure in our ordinary
judgments, and in any case the question seems to be highly contex tual. There are certainly times when ignorance is
not an excuse: I may be responsible for attending a departmental meeting even if I have lazily failed to read the
notifications in my e-mail inbox. When the agent does not fall under a legitimate rational excuse, the agent can
certainly be at fault for failing to have a belief that is the logical consequence of one of his beliefs.
146
that the normative content of a belief includes all of its consequences (whether the agent
possesses self-knowledge of these consequences is another matter). If this is right, then the
distinction I relied on above—that between the content of an agent’s beliefs and that to which the
agent’s beliefs commit him—collapses. On this view, we speak only of the total stock of an
agent’s commitments. Even if the agent does not realize that p implies q, q belongs within the
normative content of his belief that p, since p really does imply q. Since (as Carol Rovane
informs me) Isaac Levi held to such an identification, I will call it Levi’s View of the relationship
between belief and commitment.
Levi’s View retains the distinction between living up to and falling short of one’s
commitments. On Levi’s View, we will describe the agent we have been considering as someone
who so far has failed to live up to his commitment to p; he will live up to this commitment only
when he accepts that q (and, for that matter, all of the other implications of p and of q). In light
of this, it is tempting to put my differences with Levi’s View down to a verbal dispute.
We might think that a proponent of Levi’s View will simply use the descriptor ‘commitment
to which I have lived up’ as a synonym for what I have called a belief. On this usage, the
distinction between those commitments within and those outside a thinker’s ken is just the
distinction between commitments the thinker has discharged and those of which he has fallen
short. Our thinker who believes that p, but who has no inkling that p implies q, and who does not
believe that q can simply be described, by the proponent of Levi’s View, as someone who has
only partially lived up to his commitment to p. But on reflection, this verbal suggestion cannot be
quite right. Consider an agent who believes that p and that ¬p. Clearly, such an agent is
irrational, and if he understands p and ¬p his irrationality is criticizable. The point, though, is that
the agent does possess both beliefs, even while falling short of both the commitment to p (and its
consequences) and the commitment to ¬p (and its consequences).
147
A better tack, then, is to think of what I have called beliefs as simply commitments that lie
within my ken. My beliefs are those attitudinal commitments that are accessible to me and can be
called up for conscious reflection. This allows us to make ready sense of the idea of an agent who
falls short of the commitments that are constitutive of his beliefs. A proponent of Levi’s View is
free, I think, to conceive of my talk about the attitudes in this way. The key point, though, is to
bear in mind the ethical and action-theoretic aims of this dissertation. We need to capture the
distinction between commitments inside and those outside of a thinker’s ken, since it is only for
falling short of the former that an agent is subject to attitudinal blame. It is for the purpose of
preserving this important ethical distinction that I insist on a distinction between the content of
an agent’s belief and that to which the agent, on account of his beliefs, is attitudinally committed.
On the commitment-theoretic approach to taking, the imperatival dimension of belief is a matter
of the modality through which an agent relates to a content. Attitudinal commitments, from here
on out, are not located within the content the belief itself. Rather, the truth of the content is that
to which the agent is committed to uphold and respect.
BOOTSTRAPPING
We must treat with ‘bootstrapping’ worries regarding belief.21 I have said that in taking my
belief that p into the engaged stance, I am thereby committed to the truth of p and all that p
entails, barring some legitimate rational excuse. But p’s most conspicuous entailment is p itself.
This would seem to make every one of my beliefs problematically self-justifying: in believing
that p, I am also attitudinally committed to p’s truth. The worry is that I will be committed to a
belief of mine that is untrue, held for the wrong kinds of reasons, or is the outcome of some other
rational failing on my part. Surely this is a worrying result of the view that intentional attitudes
possess an implicit imperatival dimension.
21 Bilgrami (2006: 372–3) deals with the same worry and issues a similar reply. See also Moran (2001).
148
To defuse this worry, it will be helpful to recall and slightly expand upon a point that Gareth
Evans made in a related context (Evans 1982: Ch. 7). Evans’ question was: What is involved
in—how do I go about—self-ascribing the belief that p? He proposed that the method whereby I
self-ascribe the belief that p is the very same method whereby I determine an answer to the first-
order deliberative question whether p. If an interlocutor asks me whether I believe that Alice is at
the café, I furnish my reply not by ‘introspecting’ some interior domain of mental facts, but by
taking up the first-order deliberative question: What is Alice’s present location? Commentators
after Evans have used the metaphor of transparency to describe this phenomenon: it is said that
the second-order introspective question of whether I believe that p is transparent to the first-
order deliberative question of whether p. Even if this transparency phenomenon does not hold in
all contexts, I submit that it does hold always when a thinker inhabits the engaged stance of
inquiry. For it is in the engaged stance that an agent is prepared to assess the reasons that bear on
the deliberative question that is the subject matter for inquiry. Part of the broader significance of
Evans’ point, I take it, is to get us on to characterizing the engaged stance as we find ourselves
when we take up a deliberative question. And this brings in train a deep distinction between two
ways of treating with facts that are true of one’s own mind.
For it to be so much as possible to answer the second-order introspective question by
engaging in such a deliberative procedure, it must be the case that the deliberative procedure
itself is such that deliberative questions can be answered from the engaged stance only by
minding the worldly considerations that take on a normative bearing for one’s question. (This is
the slight extension of Evans’ point that I alluded to.)
The upshot is that there are really two ways of considering one’s own present epistemic
profile—the suite of epistemic (and other attitudinal) commitments that at any given time
characterize one’s own mind. One can look upon them as introspective data, as states and facts
that are true of one. In looking upon them this way, one takes one’s belief that p or intention to φ
149
as a brute fact that is true of one. Just as I am over six feet tall and born of American parents, so
too I have the belief that p and the intention to φ. In contrast, one can take up the engaged stance
towards the contents of these attitudes. To do so, though, is not to take up the engaged stance
towards an attitude, but to be engaged in (e.g.) believing or intending something with that very
content. When one takes up the engaged stance towards the (apparent) truth that p, one is
believing that p; and when one takes up the engaged stance towards φ-ing, one therein
undertakes a practical commitment to φ-ing, for it is in intending to φ that one is ‘in touch’ with
the considerations that make φ-ing good, desirable, or valuable.22 From this stance, Evans’ point
is relevant: we are confronted with deliberative questions. And once we are confronted with
deliberative questions, we are thereby concerned with worldly considerations that bear
normatively on these questions. (This is why I have favored the locution of ‘taking up’ a
deliberative question: it is not a matter of merely attending to some fact, but of putting oneself in
a position to determine whether a fact obtains.)
But this means that there really is no bootstrapping worry. From within the engaged stance,
one does not look upon one’s belief that p as a self-justifying fact. From within the engaged
stance, one’s belief that p strikes one as the best response to the worldly considerations qua
reasons that bear on whether p. When I take up a deliberative question, the attitudes that figure in
my prior deliberative position are not self-justifying at all; rather, my very mode of access to my
attitudinal commitments is via the worldly considerations that bear positively upon them. Fears
about bootstrapping—and, more generally, about persisting in some attitude in the full
knowledge that is not the best response to one’s reasons—crop up only when one’s attitudes are
conceived as mere facts that are true of one. It is from that standpoint that one seeks to preserve a
mental fact irrespective of the considerations that bear positively or negatively upon it. Since my
22 I thank Carol Rovane for helping me to emphasize this point.
150
concern has been solely with how items in one’s prior deliberative position show up for an agent
from the engaged stance of inquiry, the worry about bootstrapping does not arise. Rather,
concerns about bootstrapping—and relatedly, questions about self-deception—arise when our
own minds show up for us as objects in the world we seek to arrange according to our purposes.
Having treated two worries about the imperatival dimension of the attitudes, let me now
draw the upshot for inferring and rule-following. From the engaged stance, one’s attitudes are
themselves already rule-like. (Later, the same point will be made about the attitude of intention.)
In taking up a deliberative question on which (the belief that) p has a normative bearing, p itself
is taken as a settled fact that can play a justificatory role with respect to any new proposition that
figures in subsequent inquiry. In addition, p has the character of a commitment: p is something
that the agent ought to live up to by refusing to endorse a proposition with which p conflicts.
(Unless the reasons for this new proposition militate against p, in which case the demand for
rational unity is to be satisfied by abandoning p in favor of this new proposition.) From within
the engaged stance, the deliberating agent’s beliefs—at least, those within the thinker’s prior
deliberative position—assume this rule-like character. Put metaphorically, the attitudes that an
agent takes into the engaged stance are rules to reason by. This normative feature of believing
(and intending) only comes into view once we assume the first-personal engaged stance of a
thinker who is occurrently occupied with a deliberative question.
TAKING: SUFFICIENT FOR INFERRING?
I have inherited from Boghossian the following construal of the Taking Condition:
Taking Condition: Inferring necessarily involves (I) the thinker taking his premises to
support his conclusion and (II) drawing his conclusion because of that fact.
151
I have been concerned in this chapter only with sub-condition (I): with specifying what it is a
thinker’s ‘taking’ amounts to. I have said (to repeat) that the taking at issue consists in a thinker’s
self-conscious recognition of an attitudinal commitment (or set of attitudinal commitments). This
is a necessary, but still not sufficient, account of what is involved in inferring to a conclusion. For
we still need an understanding of condition (II): of how the self-conscious recognition of an
attitudinal commitment can play a ‘becausal’ role in producing the thinker’s transition to the
conclusion that terminates the inferring.
Some philosophers challenge this assessment. They claim that (I) is both necessary and
sufficient for inferring. On this view, all that it is to infer from premise to conclusion just is to
recognize that the conclusion follows from the premise. The suggestion here is that all there is to
infer is to ‘take’ the premises to provide sufficient support for the conclusion.23 This is because
premise and conclusion are not causally related at all; there is rather a constitutive relation
between thinker’s belief in the conclusion and his belief that the conclusion follows from
available premises. If we take this line to heart, we might question whether we have not already
said all that needs to be said about inferring by providing this characterization of how a thinker’s
attitudes figure as rules to reason by.
There are two problems, though, for this approach to inferring. First, it cannot account for
the productive dimension of inferring. By engaging in reasoning, a thinker can expand the beliefs
within his repertoire given other things he knows, and thereby conclude reasoning with beliefs
that he did not have when he took up a deliberative question. It is hard to see how we might
accommodate the productive dimension of inferring view that sub-condition (I) is sufficient for
inferring. This is because the productive dimension of inferring is secured by the generative
nature of inferring: in inferring, a thinker goes from beliefs held to the endorsement of a new
23 See Valaris (2014) and Marcus (forthcoming).
152
belief. For this to be possible, it seems that the premise and the conclusion must be ‘distinct
existences,’ at least insofar as the conclusion-episode existentially depends on the premise-
episodes.24
This leads us to the second problem for the view that taking is sufficient for inferring, one
that I have already foreshadowed in my description of an attitudinal commitment. Earlier, I wrote
that it is part of the nature of commitments that an agent may fall short of them, and that falling
short of a commitment to p does not at all impugn the ascription of this commitment to the
thinker. But this also means that there is always a ‘gap’ between an agent’s recognition that his
commitments require him to φ and the agent’s discharging this selfsame commitment out of the
very commitment recognized. To be sure, this is a conceptual, and not a causal, gap: in a wide
swath of cases, recognizing and discharging a commitment is done in one and the same episode.
Even this conceptual gap, though, frustrates the proposal that taking is sufficient for inferring.
For there is always the possibility that an agent might recognize that his attitudinal commitments
require him to infer that q and nevertheless fail to do so. I might utterly fail to live up to what my
attitudes require of me. True, such a failure introduces rational disunity and opens a thinker to
rational criticism. But the very possibility of such a ‘gap’ between commitment and performance
is important for any theory to recover, and it is hard to see how these latter approaches to taking
can recover this phenomenon. We can understand irrationality as the failure to live up to one’s
attitudinal commitments.25
24 Kietzmann is on to a similar point when he writes that an inferring and the conclusion-episode bear an asymmetric
existential dependence, so that the conclusion-episode relies upon the inferring (2018: 5–6). I am unsure of this way
of putting the point, though, since the inferring need be nothing o ther than the movement form premise to
conclusion; there need be nothing over and above this movement from premise to conclusion that causes the
conclusion-episode. 25 Bilgrami (2006: 240ff) offers this very characterization of irrationality.
153
Practical deliberating
The commitment-theoretic approach also covers practical deliberating. I mentioned that
deliberative questions may be either epistemic or practical. Practical deliberative questions, no
less than epistemic ones, must be taken up from the engaged stance. And just like the epistemic
case, the engaged stance is one from which worldly facts and states of affairs carry an immediate
and implementational normative force for subsequent thought and action. These normatively
salient facts and states of affairs do not, in practical deliberating, figure as evidence for the truth
of a proposition. But they can be considered evidence of another kind: evidence for the value,
rightness, justice, goodness, etc. of performing some course of action.
We often reason from ends to means. In doing so, our intentions figure as attitudinal
commitments just as our beliefs do. Intentions involve a commitment to their own execution in
action. From the engaged stance, one’s extant intentions figure as imperatives to act just as
intended. This distinctive species of commitment to future action brings in train the commitment
to do what is necessary in order to achieve the intended outcome. So, just as the belief that p
involves a commitment to accept p’s entailments, so too the intention to φ involves a
commitment to intend that which (from the agent’s point of view) is necessary for φ-ing. These
commitments to intend to perform the means to φ-ing in turn commitments to execute those very
intentions in action.
I leave it open whether practical deliberating is always concerned with means, or whether
instead can encompass settling on an end as well. All that matters for present purposes is that
practical deliberating takes place from the engaged stance from which the attitudes within one’s
prior deliberative position figure as commitments that the agent must ‘live up to’ in subsequent
154
practical dealings.26 (In any case, I am suspicious that in the rough and tumble of inquiry
practical and epistemic concerns are as neatly separated as the topics have come to be treated in
philosophical treatments of these matters.)
Just as in the epistemic case, then, our practical attitudes figure as rules to reason by. My
intention to φ enjoins me to φ and incorporates a demand also to intend to do that which is
necessary for φ-ing. It is part of the very nature of these attitudes that from the engaged stance
they figure as commitments that I must uphold in subsequent thought and action.
Here it might seem that we arrive at a stark dissimilarity between the epistemic and the
practical. In the epistemic case, it is natural to think of the relevant rules on the model of modus
ponens—as rules for deductive truth-preservation. In the practical case, I might enjoy an
intention to φ and a belief that ψ-ing is a necessary means to φ-ing. Discharging this commitment
in acquiring the intention to ψ does not ‘preserve truth’ in the way that I preserve truth when I
discharge the commitment to (p; p → q) by acquiring the belief that q. How then can we apply
the same style of rule-following account to the case of practical deliberating?
To answer this question, we should first take note that even in these epistemic cases truth is
not all that is preserved. The thinker’s warrant for believing that (p; p → q) is also preserved
across the transition. With Tyler Burge, we might think of epistemic warrant as a basic normative
good that attaches to beliefs and to transitions among beliefs.27 Epistemic warrant consists in
meeting normative standards that are constitutively associated with belief. Satisfying such
standards constitutes for the thinker a right to rely on the belief in further epistemic dealings
(such as reasoning, attitude revision, and so on).28 These standards pertain to representing the
26 Carol Rovane has suggested to me that deliberation as to ends belongs to practical inquiry, which might
seamlessly interact with practical deliberating about means. 27 See Burge (2013). See also Graham (2012, 451). In what follows I borrow from my treatment of diachro nic
agency in Heeney (2020). 28 Burge (2013a, 1–4).
155
truth reliably well, given limitations in a thinker’s perspective, information, and resources.29 To
say that such standards are constitutively associated with belief is to say that part of what it is to
be a belief is to be governed by norms pertaining to the truth reliably well relative to one’s
information and perspective. Epistemic warrant consists in doing as well as possible in the
service of satisfying such constitutive normative standards. I propose to understand practical
warrant analogously. Just as an epistemic warrant is a right to rely on a belief or transition among
beliefs, we may conceive of a practical warrant as a right to rely on a decision, an intention, or an
intentional action. What makes the warrant practical is that the agent relies on the warranted
item in the pursuit and attainment of rational aims.
Now, we have said that epistemic warrant consists in success relative to normative standards
that are constitutively associated with belief. If we carry this conception over to the case of
practical warrant, we should attempt to specify the normative standard which, when fulfilled by a
decision, intention, or intentional action, confers upon the agent a warrant to rely upon the item
in question in the prosecution of his rational aims. In what, then, does practical warrant consist?
Well, what are the normative standards that are constitutively associated with decisions,
intentions, and intentional actions?
As a first pass, consider what I’ll call an executive approach to this question. The executive
approach keys the relevant normative standards to the role that decisions and intentions play in
the future-directed planning and execution of coordinated schemes of action. On the executive
approach, decisions and intentions constitutively aim at their own execution in action. An
intention to (say) Work at the library on Wednesday morning is successful qua intention when it
causes an action that marks the intention’s execution. The executive approach can accommodate
‘internal’ requirements on an agent’s overall set of intentions, since mutually conflicting
29 Burge (2013a, 1).
156
intentions cannot together agglomerate into an overall realizable success-condition. On this
executive approach practical warrant is achieved when the success-conditions of intentions are
satisfied in the actions that constitute successful intention-executions.30
The trouble with the executive approach is that it does not specify a meaningful constraint on
deciding and intending. The only intentions it renders defective are those with unrealizable
success-conditions—and it is questionable whether agents can genuinely intend actions they take
to be unrealizable. On the executive approach practical warrant is secured just by having an
executable intention and by executing that intention at the appropriate time. This is quite unlike
the proposal that belief constitutively aims at truth. The truth-aim for belief amounts to a
substantive constraint on which beliefs count as defective qua beliefs in a way that the executive
approach does not in the case of deciding and intending. The executive approach sets an ‘aim’ for
intention that is trivially easy to satisfy.
The proponent of the executive approach might reply that this result is mitigated by the
presence of other intentions in the agent’s repertoire. Perhaps practical warrant consists not only
in satisfying the success-condition embedded in this or that intention, but instead in the pursuit
and realization of one’s overall system of coordinated intentions.
But I doubt this reply can do the work required of it. Not every intention to act is an element
in a larger coordinated plan of action, and with respect to these intentions the reply faces the very
same worry about triviality. What’s more, the triviality worry also confronts the agent’s overall
system of coordinated intentions. What makes such a coordinated plan one that confers a
practical warrant? It appears that even on this latter reply, the answer is just that the agent intends
to realize that plan (or the goal subserved by that plan). This does nothing to resolve our worry
30 This executive approach appears to be endorsed by Michael Bratman in at least one place. Bratman he writes that
an “intention aims at its own realization in coordination with one’s overall system of intentions” (2009 b: 25). I am
unsure whether Bratman, in later work, remains committed to this executive approach to the aim of intention.
157
about triviality: the larger coordinated plan is warranted (or confers warrant) just in virtue of
being intended. This does not provide a meaningful constraint.
These problems attend the executive approach due to a failure to distinguish an intention’s
success-condition from its correctness-condition. The success-condition for my intention to Work
at the library on Wednesday morning is that I work at the library on Wednesday out of this very
intention. It is a separate matter, though, whether this intention is the correct one for me to have,
given my other rational goals and the reasons I take to count in favor of their pursuit. It is this
latter question we are interested in, since our concern is to specify the normative standards
which, when fulfilled, confer upon the agent a practical warrant.31 So, even if an intention
includes the attitudinal commitment to carry it out by performing the action it represents (as well
as the means required to perform this action), this is a separate question from what renders an
intention a good intention as such.
Let us consider instead a normative approach to our question about the constitutive aim of
deciding and intending. On the normative approach, decisions and intentions aim not at their own
execution, but at what an agent ought to do. The correctness-condition for deciding and intending
is specified accordingly: an agent’s decision or intention to act is correct if and only if the agent
ought to perform the action that is decided upon or intended. Or, as Nishi Shah amends the
condition: an agent’s decision or intention to act is correct if and only if it is not the case that the
agent ought not to perform the action that is decided upon or intended.32 This normative construal
applies readily to intentional actions themselves: an agent’s intentional action is correct if and
only if it is not the case that the agent ought not to perform that very action.
31 On the distinction between an intention’s success-condition and its correctness-condition, see (Shah 2008). 32 See again Shah (2008). This amendment is required in order to handle Buridan cases in which one is faced with
two incompatible courses of action that are both rationally permissible.
158
On this normative approach, the correctness-condition on deciding and intending is specified
with reference to what an agent ought to do. What an agent ought to do, in turn, is determined by
the preponderance of the normative reasons there are to act, given perspectival limitations in the
agent’s information, perspective, and so on. Warrant is conferred on the agent’s decision,
intention, or intentional action when his decision or intention is sourced in a reliable competence
in satisfying the indicated correctness-condition.33
Unlike the executive approach, the normative approach provides a non-trivial constraint on
deciding and intending. For instance, if I reach an all-things-considered evaluative judgment that
the balance of reasons favors working at the library over working at home, but I decide against
this balance to work at home, then the resulting intention to work at home violates our proposed
correctness-condition. Even if the resulting intention represents a stable plan that will lead me in
a reliable way to execute this very intention in action, my intention to work at home (rather than
the library) is nevertheless defective qua intention.
To draw together and summarize the proposal: practical warrant is a positive normative
status that consists in reliably satisfying standards that are constitutively associated with states
and episodes that serve a practical function in the pursuit and attainment of an agent’s rational
aims. The states and episodes in question are primarily decisions, intentions, and intentional
actions themselves. The normative standard at issue is the correctness-condition on deciding and
intending: an agent’s decision or intention is correct if and only if it is not the case that the agent
ought to perform the action decided upon or intended. When an agent’s decision or intention
33 I understand normative reasons as constituted by objective facts and states of affairs. But it does not follow from
this view of reasons that what an agent ought to do—the ought relevant for satisfying the correctness-condition on
deciding and intending—is fixed by all of the objective reasons that there are to act, irrespective of limitations in
one’s perspective. Epistemic as well as practical warrant is keyed to such limitations. That is why there can
epistemically warranted but false belief. Warrant is not a matter of infallibly detecting all of the reasons that there
are. Warrant is less demanding than this. Warrant is obtained by doing as well as one can in the service of reliably
satisfying standards that are constitutively associated with deciding, intending, acting, judging, inferring, and
believing. So, while what reasons there are to act is one question, another question is what the agent ought to do,
given the reasons within his possession . The second question is the relevant one for assessing warrant.
159
stems from a psychological competence in meeting this standard, the agent enjoys a practical
warrant for the item at issue.34 The agent thereby accrues a right to rely on that decision or
intention in further practical dealings. Correspondingly, an intentional action is warranted when
it issues from a decision or intention that is warranted in turn, or when the action’s performance
constitutes satisfaction of the relevant standards—i.e., if the action itself marks a reliable
competence in satisfying the relevant norm. What is preserved in these practical transitions, then,
is practical warrant so conceived. The rules that figure in practical deliberating are rules for
preserving practical warrant. If my intention to φ is practically warranted, and my belief that ψ-
ing is necessary for φ-ing, then my intention to ψ will preserve the very same practical warrant.
Let us return to the question of the apparent asymmetry between truth-preserving deductive
epistemic inference and the sort of transitions between practical states and episodes. I suggest we
see epistemic warrant as the relevant status that is preserved across such epistemic transitions.
Truth-preservative deductive inference is one—admittedly central—type of warrant-preserving
transition. But it is not the only kind; indeed, preservation of truth is not even necessary for
warrant-preservation, since there can be false yet epistemically warranted beliefs that are the
outcome of such transitions. Truth matters for epistemic warrant in that epistemic warrant marks
a competence in representing the truth reliably well. Truth-preserving deductive inference is one
way of satisfying this norm.
If it is the preservation of warrant (instead of truth per se) that is fundamental for inferential
transitions, then there really is no asymmetry between the epistemic and the practical case. Both
epistemic as well as practical inferring is a matter of preserving warrant—when the inferring is a
34 For his part, Burge sketches a quasi-biological construal of practical warrant, writing that “A cognition’s […]
operating well practically is serving the function of forming states that benefit the individual’s survival for mating,
or that serve some other practical—perhaps moral—function or purpose” (2013a: 3, n. 5). I do not take my construal
to part ways significantly from Burge’s. It is, though, geared towards a higher level of cognitive sophistication on the
part of the relevant individuals.
160
good one. Correlatively, a bad inferring, epistemic as well as practical, will be one that fails to
preserve warrant across the transition. Admittedly, there is no direct analogue of deductive truth-
preserving inference in the practical case. But this is put down to a peculiarity about the norms
that constitutively govern belief, not to a conceptual hurdle to the very notion of a warrant-
preserving practical inference.
We can now draw the key parallel with the epistemic case. To make a practical inference is
to discharge an attitudinal commitment. And the ‘taking’ relevant to satisfying the Taking
Condition on practical inferring is the very same as the sort at issue in epistemic inferring: it
consists in recognition of the imperatival dimension of the attitudinal commitments making up
the agent’s prior deliberative position. This does not mean attributing to the agent some separate
state that explicitly represents these commitments; what it is to take up the engaged stance is
already to appreciate the normative purport of these imperatives. And this allows, in turn, a rule-
following conception of practical inferring along the very same lines as I have been elaborating
for the epistemic case.
Judging and deciding as inferring
Attitudinal commitments, we have seen, may be epistemic as well as practical. At this point,
it is worth reiterating a point that I made in the last chapter concerning the relationship of
judging-that-p and deciding-to-φ to inferring. To judge-that-p just is to perform an epistemic
inference, whereas to decide-to-φ just is to perform a practical inference.
Sometimes these equivalences are disputed. It is said that judging is in some sense the
mental analogue of the act of assertion. To judge that p is to give a form of inner ‘assent’ or
‘endorsement’ to p as it is held ‘before one’s mind,’ as it were. On this usage, a judging is just an
161
occurrent belief, one to which a thinker is now attending in conscious thought.35 An occurrent
belief, though, is quite different from an inferring. As I have stressed throughout this chapter,
inferring is constitutively an appreciation of a rational relation between premise and conclusion.
To occurrently entertain the belief that p involves no such rational relation. I may simply
occurrently dwell upon my belief that Blair is at the café, without any thought as to what this
portends for any of the other beliefs within my repertoire. Analogous considerations might be
raised to dispute the parallel equivalence between deciding-to-φ and making a practical inferring
to the intention to φ.36
I do not deem it profitable to engage in lengthy argument about just which conception of
judging and deciding to adopt. Instead, it will serve for the purposes of this study to recall just
why we have come around to the problem of inferring at all. I will simply point out that the
alternative conception of judging and deciding is not relevant for the purposes we have set for
ourselves—the purpose of discerning a species of attitudinal agency that grounds responsibility
for the attitudes.
Recall our interest in judging-that-p and deciding-to-φ. We are seeking to ground
responsibility for the attitudes in a species of intentional mental action that generates the attitudes
of belief and intention. Such attitude-formative mental episodes must be such that by their
performance the agent is in a position to answer Anscombe’s ‘Why?’-question. This is necessary
for securing answerability for the attitudes, the pervasive form of responsibility that w bear for
35 For this usage, see Soteriou (2005). 36 In the case of deciding-to-φ there is a separate worry. I might make a ‘snap’ deciding, without dwelling at all on
reasons that I take to support the course of action I decide upon. It is questionable whether such ‘snap’ decidings
count as inferrings, or indeed whether they have a rational basis whatsoever. As I will go on to explain, though, my
interest is with attitude-formative mental episodes that ground stable attitudes for which an agent bears
responsibility. Snap decisions give rise to no stable intention to act as decided—after making a snap decision, one is
just as likely, and indeed cannot even be rationally faulted for, ignoring the decision and the course of action that it
recommends.
162
belief and intention. This is why, in Chapter 1, I proposed to investigate responsibility-grounding
attitudinal agency by minding the peculiar status of judging and deciding.
Inferring, as we have understood it here, provides just the materials we were looking for.
This is because the ‘Why?’-question seeks the reasons that figure as premises from which the
agent infers to the conclusion. We can bring this out by first noting the inappropriateness of
answering the ‘Why?’-question with a set of reasons that did not figure as premises to the
conclusion targeted by the question. Suppose I respond to the ‘Why?’-question with reasons that
justify having the belief that p, or having the intention to φ, rather than reasons that justify p or
that justify φ-ing. Consider these exchanges:
“Why do you believe that you will pass your dissertation defense?” “The belief that I will
pass my dissertation defense helps me to sleep at night.”
“Why do you intend to go to seminary?” “Oh, I don’t see anything worthwhile about going
to seminary. But the intention to go to seminary is an intention it makes me feel good to
have.”
These replies to the ‘Why?’-question are unsatisfactory. But they are unsatisfactory not in that
they answer the ‘Why?’-question with reasons that fall short of justifying the belief or the
intention at issue. Rather, they are unsatisfactory in that they do not answer the ‘Why?’-question
at all.37 They bypass that question altogether by citing the wrong kind of reasons: reasons that
justify having the belief that p or the intention to φ and which do not (also) justify p or φ-ing. But
in virtue of what are these answers to the ‘Why?’-question unsatisfactory?
The explanation for the unsuitability of these replies, I submit, is precisely that the ‘Why?’-
question asks for the very reasons that rationalize the belief or intention targeted. The ‘Why?’-
37 Their inappropriateness can be traced to the following norms on belief and intention: Believe that p only for
reasons that constitute evidence for p; intend to φ only for reasons that justify φ-ing.
163
question asks, not just why I have the belief or intention that I do, but more specifically what
considerations I took in favor of the content of the attitude in question. Just as when I direct the
‘Why?’-question to elicit the reasons-for-which an agent performs an intentional bodily action,
when I direct the ‘Why?’-question to another, I am not after some epistemic or practical case for
the attitude in question; nor do I seek to elicit some practical case for merely having the belief or
intention in question. Rather, I ask for the agent’s own reasons why he believes or intends as he
does. Thus, the ‘Why?’-question asks: What epistemic or practical considerations does the agent
take to support the proposition in which he believes, or the plan of action that he intends to
execute?
In short: the ‘Why?’-question, when directed to a belief or an intention, seeks the premises
from which the agent inferred to the conclusion that constitutes the content of the belief or
intention that is asked about. And, because every inference satisfies the Taking Condition, to cite
these premises is thereby to cite the reasons-for-which the agent holds the attitude in question—
the reasons that the agent, in inferring to the conclusion, took to support that very conclusion,
and came to that conclusion for the very reason of this appreciation. So, every agent who infers
to a conclusion is positioned to issue an appropriate reply to the ‘Why?’-question regarding the
attitude that the conclusion constitutes. In turn, being in a position to issue an appropriate reply
to the ‘Why?’-question means that the agent is an appropriate target of the ‘Why?’-question
when it targets one of his beliefs or intentions.
Being an appropriate target for the ‘Why?’-question, though, does not ensure that one’s
answer to that question is satisfactory. I might answer the ‘Why?’-question with considerations
that are appropriate to that question, but insufficient for justifying the proposition that I believe
or the action that I intend to execute. I might have reasoned badly (committed some fallacy in
reasoning, say), or I might have inferred to an intention on the basis of practical premises in
which I ought not to have placed stock (an akratic desire, say). In either sort of case, my reply to
164
the ‘Why?’-question is appropriate in that I have provided what was asked for, but nevertheless
insufficient in that the reasons I cite do not amount to a compelling epistemic or practical case for
the conclusion of my reasoning or deliberation (respectively).
The possibility of such appropriate but insufficient replies to the ‘Why?’-question is
important to recover. For it is this possibility which opens an agent to attitudinal blame. I could
not be open to attitudinal blame unless my inferring were open to criticism. For this to be the
case, it must be possible for me to answer the ‘Why?’-question more or less well. Here, ‘more or
less well’ has to do with the strength or weakness of the premises from which I inferred my
conclusion. An insufficient reply to the ‘Why?’-question, understood as communicating having
reasoned poorly from prior premises, does open an agent up to a range of distinctive sorts of
normative criticism. This is why it is important to underscore the interconnections between
inferring to a conclusion and being in position to proffer an appropriate reply to the ‘Why?’-
question.
Conclusion
I have attempted in this chapter to provide the materials for a rule-following conception of
inferring. I have concentrated solely on sub-condition (I) of the Taking Condition. This sub-
condition requires that an agent ‘take it’ that his premises provide support for his conclusion.
However, we cannot make out this sub-condition by positing formal inference rules as a premise
from the thinker infers; nor, I argued, can we decompose a thinker’s taking into a complex set of
causal dispositions. Taking up a compelling view of the attitudes from Levi, Bilgrami, and
Rovane, I have suggested that we cannot characterize inferring to a conclusion without minding
the peculiar first-personal stance of engagement from which a thinker addresses himself to a live
deliberative question. From that stance, the thinker’s attitudes themselves exhibit the requisite
165
normativity. From the engaged stance the beliefs and intentions from which I infer are already
rule-like. They possess an imperatival dimension, figuring as normative oughts that I am
obligated to discharge in further thought and action. Given the normativity of the attitudes
themselves (from the engaged stance), we do not need to appeal to taking-beliefs, taking-
intuitions, rule-beliefs, or taking-dispositions. The problem all along has been one of description:
of failing to take heed of what it is like to occupy the stance of the thinker in the midst of
reasoning.
It is important to note explicitly what I have not shown, and have not attempted to show:
namely, some sort of reductive account of rule-following to something else. And I have
specifically argued against one sort of reduction, a reduction of rule-following to the
manifestation of causal dispositions. Rather, I have followed Boghossian (2008, 2014) in taking
rule-following as a primitive notion, unanalyzable to the non-normative. I have not yet attempted
an account of the second-sub-condition of the Taking Condition. Sub-condition (II) requires that
an agent infer to a conclusion because of the very taking that figures in sub-condition (I). That is
the task for the next chapter.
I conclude by giving an indication of our location in the broader dialectic of this dissertation.
I have argued—well, stipulated—that judging-that-p and deciding-to-φ are simply inferrings to a
conclusion. Inferring to a conclusion, in turn, can be conceived as a rule-guided mental activity,
in that in inferring a thinker takes it that his premises are commitments which he is obligated to
discharge. This points up the close connection between the premises from which the agent infers
and the fact that, having inferred, the agent is therein the appropriate target of the ‘Why?’-
question. So, there is a close connection between inferring and responsibility for the attitude that
is formed in inferring. In the next chapter, I will argue that when we understand what it is for a
thinker to infer to a conclusion because of this very appreciation of his attitudinal commitments,
we see why inferring counts as a fully intentional (yet nonvoluntary) exercise of rational agency.
167
Chapter 4
Inferring as Intentional Agency
The Action Thesis
I have argued that attitudinal agency—the conscious mental episodes of judging and
deciding that legitimate attitudinal blame—consists in a thinker’s inferring to a conclusion.
Inferring to a conclusion, in turn, involves a thinker’s appreciation of the normative support that
his premises provide for his conclusion, in line with Boghossian’s (2014: 5) Taking Condition on
inference:
Taking Condition: Inferring necessarily involves (I) the thinker taking his premises to
support his conclusion and (II) drawing his conclusion because of that fact.
In Chapter 3, I argued that sub-condition (I) of this condition ought to be understood in
commitment-theoretic terms. But I have not yet said how it is that an agent comes to infer to a
conclusion because of the attitudinal commitments from which he infers. It is my purpose in this
chapter to address this second sub-condition of the Taking Condition.
In doing so, I will also attempt to fulfill another promise of this dissertation: to show that
inferring to a conclusion is an exercise of rational agency. More specifically, I will argue that
inferring to a conclusion in thought is an intentional yet nonvoluntary action. I label this the
Action Thesis.
Since the argument of this chapter depends upon the possibility of intentional and
nonvoluntary action, let me reiterate what I mean by these terms.
There are two senses of voluntary. In the first reason-based sense, an action is voluntary if it
can be done for any reason that recommends its performance. Arm-raising is voluntary in just
this sense. I can raise my arm for any reason that makes raising my arm a good, rational,
168
desirable, or appropriate thing to do. An action is voluntary in the second will-based sense if it is
brought about in the right way by a mental state or episode that represents its very occurrence.
The paradigm of voluntariness in this second sense is intention-executing bodily action. Take
again the case of arm-raising. My arm-raising is voluntary in this second sense when my raising
it comes about as the product of an intention to make my arm rise. Voluntariness in this second
sense captures a dimension of control over our actions. I control an unfolding event just to the
extent that it unfolds in accordance with my representation of what I want (most) to happen. It
does not matter for our argument just what we designate as the relevant representational states.
They may be efficacious desires, distal or proximate intentions, volitions, or tryings. Each of
these state- or episode-types are posited to capture voluntariness in this second, will-based sense
of voluntariness. I shall speak of intentions throughout, but the reader may substitute the
preferred candidate to play this role.
Inferring to a conclusion is nonvoluntary in both of these senses. While you can infer to a
conclusion q for any reason that you take to provide sufficient rational support for q, you cannot
infer to q for any reason you take to support making that very inference and possessing that very
conclusion as a stable attitude. If inferring were voluntary in this sense, then there would be
nothing keeping us from inferring to conclusions for reasons of ‘the wrong kind’—those that
have nothing to with the rational support that the premises grant to the conclusion. Neither are
our inferrings prefigured by an explicit intention just to infer as we do. I do not form the
intention to infer from p to q and execute this intention by performing the indicated inference. If
that were so, then we would have no trouble inferring ‘at will,’ again irrespective of the rational
support that the premises grant to the conclusion. This would make inferring to a conclusion
voluntary in the way that trying to remember is a voluntary mental activity. The result would be a
distortion of the ordinary notion of an inference. There would in this case be no close link
169
between the conclusion to which a thinker arrives and the (apparent) reasons that count in favor
of that very conclusion.
Despite being nonvoluntary in these senses, I will argue that inferring is nevertheless fully
intentional. What is it for an action to be intentional? It is a daunting action-theoretical task to
specify necessary and sufficient conditions for an action’s being intentional. I will settle for a
sufficient condition. It is sufficient for an action to count as intentional if it is done for reasons.
Even if there are intentional actions that are not done for reasons, every action that is done for
reasons counts as intentional. But what is it for an action to be ‘done for reasons’? I shall operate
under the plausible assumption that when a thinker infers from p to q because of the support that
p provides for q, his inferring to q is done for the reason that p in the relevant sense. This
wording echoes sub-condition (II) of the Taking Condition. To understand inferring, we must
understand the Taking Condition; and to understand the Taking Condition, we must understand
the sense in which inferring in thought is an intentional nonvoluntary action.
I will also assume that it is not necessary for an action to be intentional that it is undertaken
out of an intention so to act. Imposing this condition on an action’s being intentional simply
identifies intentionality with the will-based sense of voluntariness, thereby foreclosing the very
possibility of intentional nonvoluntary action. I hope that the discussion to follow will motivate
keeping open the theoretical possibility of nonvoluntary action. But there are also independent
grounds to resist identifying intentionality with will-based voluntariness. Many action theorists
recognize the possibility of actions that are intentional but neither preceded nor accompanied by
a specific intention so to act (Bratman 1984). Consider my walking to the department. Walking to
the department is certainly an intentional act of mine. And so is every step that I take along the
way. But I do not from an intention for every single step that constitutes my journey. Putting
170
aside popular philosophical treatments of this possibility, it does not seem forced on us that an
intentional action occurs without being preceded by a prior intention so to act.1
I’ll begin by pointing out a telling symmetry between inferring and intentional action. Just as
Wittgenstein asked for the difference between action and non-actional bodily movement, so too
we face the problem of distinguishing inferring from non-inferential transitions in thought.2
Taking a cue from recent Anscombean treatments of intentional agency, I’ll show that the
arguments of the last chapter point up the need for a non-decompositional answer to both the
problem of action and the problem of inference.3 After exhibiting Anscombe’s non-
decompositional approach to intentional action, I apply it to the case of inferring. Key to this
Anscombean approach is the claim that practical knowledge of what I am doing and why is
constitutive of intentional action, marking the difference between a series of mere bodily
movements and an intelligible order of means directed to an end. In like manner a thinker’s self-
conscious appreciation of what follows from what is constitutive of inferring, distinguishing an
intelligible premise-conclusion order from a mere succession of intentional episodes. It is this
practical self-consciousness, and not any efficient-causal link, which secures sub-condition (II) of
the Taking Condition. In this, overt intentional action as well as attitude-formative mental action
1 Bratman claims that while each step is not preceded by a prior intention, each step does fall within (what he calls)
the motivational potential of my intention to walk to the department. Even though not every intentional φ-ing is
undertaken out of an intention to φ, it is still the case that some other intention explains the intentionality of φ-ing.
See Bratman (1984). One might attempt to account for the intentionality of inferring in an analogous way. On this
suggestion, an agent enjoys an intention to decide what to do or judge what is the case, and the specific deciding -to-
φ or judging-that-p belongs within the motivational potential of this intention. But this would count as an
objectionable form of instrumentalism of the sort I argued against in Chapter 2. How does an agent arrive at such an
intention? Only by exercising a more fundamental species of rational agency —and it is this exercise of agency that I
am attempting, here, to explain. The wider action-theoretic upshot is this. If my argument in this chapter is correct,
then there are nonvoluntary intentional actions whose intentionality is not to be explained with reference to any
intention, thus refuting what Bratman calls the Single Phenomenon View. 2 Also, both contrasts admit of middle ground cases. There are bodily movements that seem agentive even if not fully
intentional, and there are transitions in conscious thinking that are not inferential but nevertheless not cases of idle
mind wandering. My focus in this chapter is on the stark contrasts, leaving the interesting middle cases for another
occasion. I thank Katja Vogt for bringing this to my attention. 3 My understanding of Anscombe’s work has been informed by work by Will Small (2012), Joh n Schwenkler
(2015), and Lucy Campbell (2017). Later in the chapter I draw on Ursula Coope’s (2013) interpretation of Aquinas.
171
likewise consist in a distinctive species of self-consciousness. This line of argument vindicates
the Action Thesis and sub-condition (II) of the Taking Condition at once.
The problem of action and the problem of inference
The problem of action is summed up in Wittgenstein’s question: “what is left over if I
subtract the fact that my arm goes up from the fact that I raise my arm?” (1953: §621). (Put aside
the issue whether Wittgenstein meant this question literally. He probably did not.) To ask for the
thing left over after we ‘subtract’ the fact that my arm goes up from the fact that I raise my arm is
to ask after the isolatable element that, when admixed with an arm-rising, renders the event an
intentional action of raising my arm. What is the extra factor that renders a non-actional bodily
movement a full-blooded intentional action? So understood, the question presumes that
intentional bodily action and non-actional bodily movements share a common factor. The
presumption is that when I raise my arm intentionally, a non-actional bodily movement is
situated ‘inside’ the action as a standalone component, one that is just the same kind of thing that
can happen of itself. Following Lavin (2016), call this the decompositional approach to
intentional action.
The decompositional approach construes the action-making factor as something wholly
extrinsic to the bodily movement that is the action. We see this feature at work in the causalist
conception of intentional action. According to the causalist, a bodily movement is an action when
it is caused ‘in the right way’ by a suitable belief-desire pair (or by an intention, volition, or
trying that is in turn the causal product of such a pair). As Davidson pointed out—and I am not
here placing him within the standard causalist camp—the contents of these belief-desire pairs
lend themselves to a distinctive pattern of explanation of the bodily movement that they cause.
As I described in Chapter 2, the relevant explanations are instrumentalizing. This is because the
172
desire points up something which the agent sees as (prima facie) good, valuable, rational, or
worthy of pursuit, while the belief represents that an action (under some description) will
promote or further the very element picked out by the desire. Such a belief-desire pair permits an
instrumentalizing explanation of action because it represents a bodily movement as an
instrumental means to a further end. For the causalist, a bodily movement is an intentional action
just in case it is caused ‘in the right way’ by such a complex. (When understood in this way, it
makes no difference whether the most proximal cause of the bodily movement is an intention
rather than a belief-desire pair. This is because the intention itself is conceived to be the causal
product of such a complex.) For the causalist, the problem of action is to specify the belief-desire
complex and the causal relation that unites this complex to a mere bodily movement. Since
causes must be distinct existences from their effects, the causalist will say that the action-making
factor, the belief-desire pair, is extrinsic to the bodily movement that it causes. On this view, a
bodily movement is rendered actional in virtue of its extrinsic causal history and not in virtue of
anything intrinsic to the action itself. This decompositional approach really does take
Wittgenstein’s question literally: according to it, that which is ‘left over’ when we subtract the
fact that my arm goes from the fact that I raise my arm is a belief-desire pair that represents the
arm-rising as contributing to the satisfaction of my desires.
Now consider a corresponding problem of inference. The problem of inference is to specify
the distinction between a thinker’s inferring to a conclusion and a thinker’s undergoing a mere
succession of intentional episodes in the conscious stream of thought. Echoing Wittgenstein:
what is left over when we subtract the fact that a thinker experiences such a mere succession
from the fact that the thinker inferred to a conclusion? To drive the question home, consider the
following inferring:
1. If Alice is at the café, then Bill is at the café.
173
2. Alice is at the café.
3. So, Bill is at the café.
As I explained in Chapter 3, each of these items (1)–(3) corresponds with a conscious content-
bearing episode in the thinker’s stream of thought. But what distinguishes a thinker’s inferring
that (3) from merely undergoing a mere succession of intentional episodes, without any inkling at
all that the contents of (1) and (2) together count as conclusive reasons for the content of (3)? We
have here the very same problem that Wittgenstein diagnosed in the case of bodily action. Just as
Wittgenstein’s question asks for the factor that is present in intentional arm-raising but missing
when one’s arm merely rises, here we ask after the factor present in a thinker’s inferring to a
conclusion but missing when he undergoes such a mere succession of intentional episodes in the
stream of thought.
On a decompositional approach to the problem of inference, the inference-making features
are wholly extrinsic to the transition between premise- and conclusion-episodes in the thinker’s
stream of thought. However, a decompositional approach to inference cannot succeed. This
argument will rehearse some considerations we have already encountered in Chapter 3 but will
place them in a new light.
Consider first a causalist conception of inferring. On this suggestion, an inferring occurs just
when the premise-episodes cause the conclusion-episode ‘in the right way.’ (On such a view, the
inferring is the token occurrence of the conclusion-episode, just as the causalist identifies
intentional action with bodily movement.) But we need not appeal to deviant causal chains to
imagine a counterexample to this suggestion. Recall Boghossian’s (2014: 4) depressive, who
experiences a reliable and non-deviant transition from I am having so much fun to the episode
representing But there is so much suffering in the world. The depressive does not infer the latter
from the former; he is simply subject to the operation of a blind psychological association. This
174
psychological association can be as reliable, non-deviant, or counterfactually robust as the
causalist prefers. Such features cannot make the connection an inferential one. There seems no
legitimate sense in which the depressive’s resulting belief that there is so much suffering in the
world is based on his belief that he is having so much fun. It was in the light of such cases that
we followed Boghossian in placing the Taking Condition on inferring. The problem for the
causalist about inferring is that there is nothing in the idea of a reliable efficient-causal link
between the premise- and conclusion-episodes that involves the thinker’s appreciation that his
premises support his conclusion. And yet, without incorporation of this appreciation, it is hard to
see how the relevant episodes can figure in a premise-conclusion relation at all.4
To satisfy sub-condition (I) of the Taking Condition, we might posit an additional cause of
the conclusion-episode. The suggestion is to add another premise as cause:
4. If Alice is at the café, then Bill is at the café.
5. Alice is at the café.
6. If (If Alice is at the café, then Bill is at the café) and (Alice is at the café), then Bill is at
the café.
7. So, Bill is at the café.
With premise (6) we set off a Lewis Carroll-style regress of premises from which no inference at
all can issue. For if the thinker cannot infer from an appreciation of (4) and (5) together, there is
no way that a premise like (6) can make the difference. This way of incorporating a thinker’s
‘taking’ cannot make for a satisfying causalist solution to the problem of inference.
4 Katja Vogt raises the following question regarding Boghossian’s case of the depressive. Perhaps the depressive’s
psychological association is not ‘blind’ per se, but rather the product of some other quasi -rational form of
association linking the suffering in the world as rendering cheerfulness inappropriate. My reply is that insofar as the
depressive faults himself for having fun in recognizing the suffering of the world, then he has performed a kind of
inference, which is just the sort of dynamic mental occurrence I am trying to explain in this chapter. But insofar as
the depressive’s occurrent thought that There is so much suffering in the world strikes him as an unfit reason to not
enjoy himself, he is as it were the passive bystander subject to a blind psychological force.
175
So, we might try another tack. Perhaps the thinker’s taking consists not in any additional
premise about what follows from what, but in the causal play of dispositions. As we have seen,
Broome (2013) suggests that a thinker’s taking consists in a two-track disposition: a disposition
for conclusion-episodes to occur immediately after premise-episodes and a corresponding
disposition for this succession to ‘seem right’ (which consists in turn in yet further dispositions).
As I argued in Chapter 3, this appeal to rule-dispositions renders the thinker’s taking a mere
epiphenomenon. No such account can respect sub-condition (II) of the Taking Condition, which
requires that the thinker move to the conclusion because of an appreciation of the relations of
rational support that the premises provide for the conclusion.
These considerations point up a dilemma for causalism about the problem of inference. To
satisfy the Taking Condition, the causalist can either posit an explicit premise about what follows
from what as an efficient cause of the conclusion-episode; or else account for the thinker’s taking
by positing a network of causal dispositions. However, the first horn gives rise to a regress from
which no inferring at all can issue, whereas the second horn undermines sub-condition (II) of the
Taking Condition. The upshot is that the causalist cannot account for both sub-conditions of the
Taking Condition together.
This dilemma stems from a sole source. That source is the decompositional approach to the
problem of inference. These causalist approaches to the problem of inference certainly are of a
piece with decompositional approaches to the problem of action. Like in the case of intentional
action, the causalist claims that the inference-making features are wholly extrinsic to the
transition between premise- and conclusion-episodes in the stream of thought. But since the
inference-making features are wholly extrinsic to the transition in thought, we cannot capture the
way that a thinker’s ‘taking’ plays a guiding role in the thinker’s arrival to the conclusion-
episode. The only way to secure the right relation between the succession of episodes and the
thinker’s appreciation of relations of rational support is to divorce ourselves from the
176
decompositional approach to inferring. So long as the inference-making factor is something
wholly outside the succession of intentional episodes, we cannot accommodate both sub-
conditions of the Taking Condition together.
To evade the dilemma, we apparently require a non-decompositional approach to inferring.
On such an approach, we do not identity the inference-making factor as an extrinsic factor in a
Wittgensteinian equation (cf. Lavin 2016: 612). Rather, we source the inference-making feature
in something that is intrinsic to every inferring. On this competing approach, we will look to
something intrinsic to inferring—something which makes an inferring what it is. My suggestion
is to draw on a recently recovered non-decompositional approach to intentional action sourced in
Anscombe’s Intention. After laying out the rudiments of this approach to action, we will be
positioned to apply its morals back to the problem of inference.
Intentional action and practical knowledge
My concern in this section is to exhibit Anscombe’s non-decompositional approach to
intentional action, in order to later conceive of inferring to a conclusion in analogous terms. The
hallmark of Anscombe’s approach is an emphasis on the distinctive knowledge that a rational
agent possesses of occurrent intentional action. The approach is non-decompositional in that this
practical knowledge is conceived to be the formal rather than efficient cause of intentional action.
Practical knowledge of what I am doing and why articulates a ‘mere heap’ of bodily movements
into an ordered unity of means undertaken for the sake of an end. In virtue of this constitutive
role, it is only because an agent possesses this knowledge that he is intentionally acting at all. I
wish to trace Anscombe’s route to this conclusion. We will find it an instructive one.
177
THE ‘WHY?’-QUESTION AND INTENTIONAL ACTION
Lavin (2016) argues that Anscombe’s non-decompositional approach takes off from a
starting place that is altogether different from the decompositional theorist’s way of posing the
problem of action. Anscombe does not attempt to locate intentional actions by isolating an
intrinsically non-intentional factor that renders a non-bodily movement an intentional action.5
Instead, intentional actions are distinguished by their susceptibility to a characteristic form of
explanation. As I quoted her in Chapter 1, Anscombe starts with the claim that intentional actions
are “actions to which a certain sense of the question ‘Why?’ is given application; the sense of
course is that the answer, if positive, gives a reason for acting” (1957: 9). Anscombe registers an
immediate circularity worry in this formulation, since it tells us no more about intentional actions
than that they are done for reasons (Stoutland 2011: 24).
Anscombe works her way out of the circle by considering, first, cases where the ‘Why?’-
question fails to apply. The question ‘Why are you doing that?’ is refused application when the
agent answers, ‘I wasn’t aware that I was doing it’ (1957: 11). If an agent (sincerely) answers in
this way, then there cannot be an intentional action going on. Much of what I will elaborate in
this section will go towards explaining why this is so.
An important complication is that since intentional actions are description-sensitive, a
person might know that he is doing something under one description and fail to know it under
another (1957: 11–12). For instance, I might know what I am doing under the description ‘MH’s
pouring himself a gin and tonic’ and yet reply that ‘I wasn’t aware I was doing that’ when asked
why I am mixing gasoline and tonic. In this, my knowledge of what I am doing resembles my
knowledge of what I believe and what I intend. Like our beliefs and our intentions, “to say that a
5 Indeed, Anscombe’s approach is notable for the way she eschews the attempt to isolate some action -making mental
element (see e.g. Intention, §27; Stoutland 2011: 28; Schwenkler 2015: 5). This is another reason to take
Anscombe’s approach seriously in an attempt to lay bare the active character of inferring to a conclusion: as I have
argued, inferring is not rendered intentional on account of any intention to infer just as one does.
178
man knows he is doing X is to give a description of what he is doing under which he knows it”
(1957: 12). Every intentional action, then, will be known by the agent performing that action at
least under one description. Intentional actions are not the kind of thing that can go on ‘behind
the agent’s back,’ as it were.
What’s more, an agent knows his own occurrent action in a distinctive way: namely, non-
observationally. Anscombe has it that we do not know of our own actions by observing, as I
might know by sight that a bus is crossing the street one block away. Now, not all of the things
that an agent knows non-observationally are intentional actions (1957: 13–15). A person also
knows the position of his limbs without observation since “nothing shows him the position of his
limbs” (1957: 13).6 Still, the claim that we know our intentional actions non-observationally
helps to further delimit cases where the ‘Why?’-question fails to apply. It is refused application
when the agent answers either with “I was not aware I was doing that,” or when, if the agent
knows he is doing it, he knows it only by observing himself do it (1957: 24). Intentional actions
are known, and this knowledge is not based on observation.
But what distinguishes intentional action from other things (such as one’s present bodily
position) that are known without observation? Here Anscombe introduces the notion of a mental
cause. Mental causes include such things as feelings or images that figure in an efficient-causal
explanation for why the relevant event took place. “A mental cause,” Anscombe has it, “is what
someone would describe if he were asked the specific question: what produced this action or
thought or feeling on your part: what did you see or hear or feel, what ideas or images cropped
up in your mind and led to it?” (1957: 17–18).
The operative notion of ‘cause’ is opaque in this portion of Intention (cf. Stoutland 2011: 25–
26). But I think we do no injustice to Anscombe’s discussion to conceive of mental causes as
6 Anscombe makes a special point here of justifying the use of the word ‘knowledge.’ An agent knows the position of
his own limbs because he can be wrong about this very fact (§8).
179
events that constitute merely explanatory reasons: reasons that explain why something occurred,
but which do not justify the event in question.7 Suppose you ask me why I stayed up all night last
night. One way of refusing application of this question will be to reply that the constant sirens
just outside my window kept me from falling asleep. The sound of the sirens is here cited as an
explanatory reason for the occurrence of some event (or rather, the non-occurrence of the event
of my falling asleep). This explanatory reason does nothing to justify my staying awake at night,
and indeed my reply is not even offered under the guise of providing such a justification. But it
does offer a satisfying explanation of why I was up all night, and upon receiving this explanation
there is nothing more you need to know in order to understand why it was I was up all night.
Contrast this citation of a merely explanatory reason with the provision of a justifying reason:
one that provides a point or purpose furthered by my staying up at night (“I wanted to see the
referendum results in real-time”). Providing such a point or purpose—what Anscombe, later in
Intention, calls a ‘desirability-characterization’ (1957: 72)—communicates that which is
furthered by my staying up all night. In contrast, citing the merely explanatory reason (“Sirens
kept me up all night”) does nothing to justify my staying up all night, because it is not offered as
even a potential justification of doing so. (Obviously, some justifying reasons might fail to justify
the behavior in question.) The ‘Why?’-question is refused application when the answer cites just
such a mental cause (or merely explanatory reason for why something has occurred).
Anscombe distinguishes such mental causes from an agent’s motives, which do grant the
‘Why?’-question application. Motives divide into three sub-kinds (Stoutland 2011: 25). First,
backward-looking motives include such things as revenge or gratitude. To answer the ‘Why?’-
question by citing a backwards-looking motive is to cite a past occurrence as the “ground of an
action” (1957: 20). For instance, my receipt of a birthday gift can be the ground of my action of
7 In support of this interpretation, notice that Anscombe says that the ‘Why?’-question is refused application when
the agent cites evidence for the occurrence of the event in question (1957: 15–16).
180
sending off a note of thanks. Such backward-looking motives are distinct from mental causes in
that citing a backward-looking cause mentions, perhaps implicitly, something good or evil that
the agent discerns in the past occurrence (1957: 21); this needn’t be the case when a mental cause
(merely explanatory reason) is cited in response to the ‘Why?’-question. Second, what
Anscombe calls motives-in-general provide re-descriptions of an action which point up the
proper light in which to interpret an action (1957: 21). For instance: “Why are you holding your
arms up?” “I am stretching my sides.” This reply to the ‘Why?’-question re-describes the action,
since here the very same event falls under two descriptions, only one of which captures my
purposes. Notice that this reply to the ‘Why?’-question does not refuse its application, even
though the description that figures in the question is substituted for a distinct description. This
leads to the third kind of motive, forward-looking motives. When the ‘Why?’-question is
answered by citing a forward-looking motive, some future event is given in reply: “Why are you
packing up your bag?” “To work at the library.” Anscombe calls such forward-looking motives
intentions. The ‘Why?’-question has application in all three cases. This is because in each case, to
cite the motive or the intention is to communicate a consideration that counts in favor of the
action in question. As I mentioned above, this is not so when an agent cites a mental cause
(merely explanatory reason).
There are two other telling cases where the ‘Why?’-question finds application, both of which
I mentioned earlier in Chapter 1. I might reply to the question by saying “No reason.” For
Anscombe, this reply no more refuses application to the ‘Why?’-question than the reply “None”
refuses application to the question, “How much change is in your pocket?” (1957: 25). Secondly,
the ‘Why?’-question can be legitimately answered with “No particular reason,” so long as,
Anscombe says, we can “make sense” of the person so acting (1957: §18; Stoutland 2011: 26). In
many cases, ‘No particular reason’ can indicate that I am doing something merely for the sake of
doing it (for its own sake), where there is no further objective I achieve by the performance.
181
Drawing all of this together: intentional actions are events for which the question “Why are
you doing that?” has application. This ‘Why?’-question is refused application when: the recipient
replies “I wasn’t doing that,” or if his knowledge of its going on is based on observation, or if his
answer cites a merely explanatory reason. The ‘Why?’-question finds application when the
answer cites some good or evil in a previous occurrence; when the answer re-describes the action
in an intelligible (but not merely explanatory) light; when it mentions something future; or even
if the answer is ‘No reason’ or ‘No particular reason.’ Since these types of legitimate answers
make up a wider class than reasons for action, the ‘Why?’-question “can now be defined as the
question expecting an answer within this range” (1957: 28; Stoutland 2011: 26). Following Lucy
Campbell (2017), we can say that for Anscombe the ‘Why?’-question finds application when it
asks about behaviors that have a point: they are behaviors that contribute towards some good
(avoid some evil) or to the realization of some aim.
There are two reasons I have set out to describe all the cases in which the ‘Why?’-question
applies and fails to apply. First, as we have seen, Anscombe’s remarks clarify what intentional
actions are—namely, behaviors that have a point or purpose. Second, the applicable range of
answers to the ‘Why?’-question evokes Anscombe’s all-important starting-point—the claim that
intentional actions are distinctive in that the agent in the midst of an intentional action has a
distinctive kind of knowledge of its performance. As we have seen, the ‘Why?’-question is
refused, rather than answered, when the agent says that his knowledge of the relevant event is
based on observation, or when the agent was flatly unaware of the event at all.
However, this way of demarcating the class of intentional actions leaves us with a question.
What is it about the nature of intentional action that places an agent in position to answer the
‘Why?’-question? To answer this question, we need to delve further into Intention, wherein
Anscombe indicates the content of an agent’s practical knowledge of what he is doing. This will
enable us to then go on to say why applicability of the ‘Why?’-question is criterial for intentional
182
action, and ultimately why practical knowledge serves as the formal principle that makes
intentional action what it is.
THE CONTENT OF PRACTICAL KNOWLEDGE
We can make progress on these questions by treating with one of Anscombe’s central
examples in Intention. The example concerns a person who operates a water pump in order to
poison the inhabitants of a nearby house (1957: §23ff). Here as ever we have an event that falls
under a range of true descriptions. What is more, there are several distinct descriptions under
which the event is intentional. Call these intentional descriptions. Anscombe tells us that in the
midst of his pumping, the man can be truly described as intentionally:
A. Moving his arm;
B. Operating the water pump;
C. Replenishing the water supply;
D. Poisoning the inhabitants of the house. (1957: 45)
Notice that each of the intentional descriptions in A–D can be incorporated into a corresponding
‘Why?’-question:
A′. Why are you moving your arm?
B′. Why are you operating the water pump?
C′. Why are you replenishing the water supply?
D′. Why are you poisoning the inhabitants in the house?
183
Each of these ‘Why?’-questions A′–D′ finds application—each one falls within the range that
Anscombe demarcates in the opening sections of Intention.8 Since this is so, the agent in this
example must possess practical knowledge that he is: (A) moving his arm; (B) operating the
water pump; (C) replenishing the water supply; and (D) poisoning the inhabitants. This is
because the ‘Why?’-question will simply fall to the ground should the agent reply that “I didn’t
know that I was A-ing/B-ing/C-ing/D-ing.” This secures an important preliminary feature of
Anscombe’s non-decompositional approach to action: as the agent in this example acts, he must
possess practical knowledge of what he is doing under each of the intentional descriptions A–D.
(If not, then the description in question cannot be counted as an intentional description and drops
out of our consideration.)
In fact, Anscombe tells us that the agent must possess practical knowledge of more than
simply act-descriptions A, B, C, and D taken separately (1957: §26). These descriptions A–D are
interrelated and hold together as a unity. If an agent knows any of the descriptions A–D, he must
know all of them and know them in a particular way. This we can bring out by noting two series
of questions we might pose to the agent in the midst of his action.9
The first is our familiar ‘Why?’-question. Suppose we ask the agent, ‘Why are you X-ing?’,
where ‘X’ stands for any of the intentional descriptions A–D. The agent can issue a legitimate
reply to this question by citing any one of the intentional descriptions in the series that comes
8 Anscombe is clear in Intention that each of the intentional descriptions A–D describe one action: the man’s moving
his arm is his operating of the water pump, which is his replenishing of the water supply, and so on (1957: 46). In
this, Anscombe offers, like Davidson, a ‘coarse-grained’ individuation of actions. Crucially, though, Anscombe does
not share Davidson’s view that all the man does is the most proximate event in the series. (At least, this
interpretation is not forced upon us by anything that Anscombe says in Intention.) As I read Anscombe, the view is
that distinct act-descriptions correspond to distinct parts within the same unity. More on this to come in the main
text. We must not read into Anscombe all of the features that are frequently grouped together under the heading of
the so-called ‘Davidson-Anscombe coarse-grained theory of act-individuation.’ Doing so risks obscuring the non-
decompositional approach to intentional action that we are seeking in Intention. 9 In what follows I have learned from Small (2012: §2.2), Schwenkler (2015: §2), and Lavin (2016: 619 –622).
184
after X.10 ‘Why are you moving your arm up and down?’ can be answered with ‘Because I am
operating the water pump’; and the question ‘Why are you operating the water pump?’ can be
answered with ‘Because I am replenishing the water supply,’ and so on. (The replies needn’t cite
the next adjacent intentional description in the series; the agent can answer ‘Why are you moving
your arm like that?’ with ‘Because I am poisoning the inhabitants,’ trusting, thereby, that context
will indicate that moving his arm is poisoning the inhabitants. See Intention, p. 46.). In this way,
as Small (2012) puts it, we can move up the A–D series to arrive at progressively more articulate
answers to the ‘Why?’-question. The answers are progressively more articulate in that with each
round of question and answer one gets a better sense of just what the agent is doing—we arrive
at a clearer picture of the agent’s purpose.
Just as we may move up the A–D series in this way, we may also ‘move down’ the series.
We move down the series not by asking ‘Why?’, but ‘How?’. If we ask the agent, ‘How are you
poisoning the inhabitants?’ the agent may reply by citing any of the prior descriptions C–A: ‘I am
(D) poising the inhabitants by (C) replenishing the water supply/(B) operating the water
pump/(A) moving my arm.’ Whereas moving up the A–D series elicits the purpose for which the
agent acts, moving down the series elicits the means that the agent employs in order to realize
that purpose (1957: 46).
In a moment, I will return to this peculiar feature of the A–D order. Why is it that we can
both ‘move up’ the series by asking ‘Why?’ and ‘move down’ the series by asking ‘How?’ This is
something that we might expect Anscombe to explain, and indeed she gives us the tools to
explain it, as Will Small (2012) points out.
10 If the agent is asked ‘Why are you poisoning the inhabitants of the house?’, then Anscombe tells us the agent can
cite the further distal purpose for which he acts, such as: ‘I want to replace this crop of leaders with another one.’
See the discussion in Intention at §27ff and Stoutland 2011: 27. Anscombe does say that this sort of reply constitutes
a ‘break’—though not a sharp one—with the A–D series.
185
Anscombe writes that “When terms [i.e. intentional descriptions A–D] are related in this
fashion, they constitute a series of means, the last term of which [i.e., D] is, just by being given
as the last, so far treated as end” (1957: 46–47). In this connection, Anscombe speaks of the last
term in the series “swallowing up” the previous ones, in that it is perfectly appropriate the cite D
in a response to the question ‘Why are you A-ing?’ (1957: 46).
This allows us to characterize more fully the content of practical knowledge. An agent does
not know each of the intentional descriptions separately. Practical knowledge must be more
articulate than this. The agent knows that the intentional descriptions of his action hang together
as a unity—an instrumental act-complex such that each description in the complex bears a
teleological relation to the other items in the series. The mark of such a unity is that, given the
series A–D, we can ‘move up’ the series with the ‘Why?’-question as well as ‘move down’ the
series with the ‘How?’-question. Practical knowledge is not just knowing that one is A-ing/B-
ing/C-ing/D-ing, but also that one A’s in order to B, B’s in order to C, C’s in order to D. Only
such a more articulate practical knowledge can provide a principled explanation of why the agent
is positioned to reply to any of the ‘Why?’-questions A′–D′. Taken together, the agent’s replies to
all of the ‘Why?’-questions A′–D′ draw on practical knowledge of how intentional descriptions
A–D hang together as a unity. The content of practical knowledge is a unified order of means
undertaken for an end: when an agent acts, he knows what he is doing, why he is doing it, and
how he is getting it done.
Anscombe tells us that we cannot understand practical knowledge without understanding
practical deliberation (1957: 57). She subsequently devotes §§33–44 of Intention to this topic.
For our purposes, the key point to take away from Anscombe’s discussion of practical reasoning
is the following. When an agent begins with a practical commitment to D-ing, the action A-ing is
186
reachable by a consideration of progressively more specific means to D-ing.11 Practical reasoning
or deliberation is in this sense calculative. I begin with the practical commitment to poisoning
the inhabitants in the nearby house. How can I poison them? By replenishing the water supply
with poison. How can I replenish the water supply? By operating this pump. But how can I
operate this pump? By moving my arm up and down. In this way, as Schwenkler puts it,
“practical reasoning is a form of thought that proceeds from a general action-type to a particular
bodily movement by identifying means to a given end” (2015: 5). (For Anscombe’s talk of
‘calculation,’ see e.g. Intention, p. 79.)
This form of thought is exemplified, Anscombe tells us, by the Aristotelian practical
syllogism (understood in this calculative way). But she also enters a word of caution: “if
Aristotle’s account were supposed to describe actual mental processes, it would in general be
quite absurd” (1957: 80). Her emphatic point is that it is not the case that every intentional action
of which an agent possesses practical knowledge is preceded by an explicit bout of practical
reasoning in which premises which correspond to the A–D order are explicitly entertained or
contemplated. Rather, her view is that the rational order which corresponds to that bout of
reasoning—i.e., the A–D teleological order of end-directed means itself—“is there whenever
actions are done with intentions” (1957: 80). Put another way, the unified teleological order
exemplified by the series of descriptions A–D is sometimes arrived at by dint of practical
reasoning that is completed well before the agent ever approaches the handle of the water pump.
11 The phrase ‘practical commitment’ does not occur in Intention. Still, I think that this is what Anscombe has in
mind with her discussion of wanting over §§36–41. First, practical commitments are all-out rational commitments,
and this appears to correspond to the fact that ‘wanting,’ in Anscombe’s sense, “is neither wishing nor hoping nor the
feeling of desire, and cannot be said to exist in a man who does nothing towards getting what he wants” (1957: 67 –
8). Anscombe goes on to say that wanting is connected to the good as truth is connected to judging (1957: 76).
Throughout her discussion, Anscombe insists that the relevant sense of ‘wanting’ and ‘good’ are not ethical: they
correspond instead to the agent’s own appraisals (hence her talk of heinous action-descriptions figuring as
‘desirability characterizations’ for heinous men). All of this supports the idea that what Anscombe has in mind is a
notion of practical commitment, so that my practical commitment to D-ing serves as a kind of rational ‘baseline’
from which, by consideration of more and more specific means, I can act with an eye towards D-ing.
187
But whenever the agent operates the water pump with an eye to poisoning the inhabitants, the A–
D order is present. I shall return to this issue in a later section.
What we have to understand, now, is how this is so. How is it that intentional action, as
evinced in the A–D series of intentional descriptions, itself has the very same structure—the very
same means-end or teleological unity—as the practical knowledge that an agent enjoys of his
intentional action as it occurs? To answer this latter question, we must understand what it is for
practical knowledge to be action-constitutive—what it means to say that practical knowledge is
‘the cause of what it understands.’
THE ACTION-CONSTITUTIVE ROLE OF PRACTICAL KNOWLEDGE
At this point we come to a worry that occupies a few sections of Intention (§§28–32).
Understanding Anscombe’s response to this worry shows how practical knowledge constitutes
intentional action, and how the resulting picture resists the temptation to ‘decompose’ intentional
action into non-intentional atoms.
Practical knowledge is non-observational: I do not have practical knowledge of my action by
observing its occurrence. In connection with the A–D order, we have seen that practical
knowledge is knowledge of what I am doing, why I am doing it, and how I am getting it done.
But some of the things I do—some of the things that are happening—are at some remove from
my own body. It follows from what Anscombe has told us so far about practical knowledge that
an agent in the midst of action enjoys a non-observational knowledge of events taking place out
in the distal environment. In operating the pump, I know non-observationally that the inhabitants
are being poisoned. Anscombe shares in the reader’s perplexity at this consequence. How can I
know non-observationally that something out there is going on?
It is tempting to ease this perplexity by making practical knowledge a composite. On this
suggestion, practical knowledge is made up out of (on the one hand) the agent’s non-
188
observational knowledge of what he intends, together with (on the other hand) observational
knowledge of what occurs in the agent’s immediate distal environment (cf. Donnellan 1963). On
this mixed view, practical knowledge is only partly non-observational and concerns only the
willing, trying, or intention.
Anscombe pushes against this conception, pointing out that sometimes the only intentional
descriptions available to an agent mention distal occurrences (1957: §30). She concedes, though,
that sometimes my practical knowledge of something occurring in the distal environment
presupposes ordinary knowledge that is “based on observation, inference, hearsay, superstition,
or anything that knowledge or opinion ever are based on” (1957: 50). She writes:
When knowledge or opinion are present concerning what is the case, and what can
happen—say Z—if one does certain things, say ABC, then it is possible to have the
intention of doing Z in doing ABC; and if the case is one of knowledge or if the opinion is
correct, then doing or causing Z is an intentional action, and it is not by observation that
one is doing Z; or in so far as one is observing, inferring etc. that Z is actually taking place,
one’s knowledge is not the knowledge that a man has of his intentional actions. (1957: 50)
The crucial point for Anscombe is that in these circumstances I possess non-observational
practical knowledge that I am Z-ing, even if this presumes some prior observational, inferential,
etc., knowledge that I can Z by ABC-ing. My practical knowledge will fail should these
presumptions be mistaken. But where the presumptions are knowledgeable my practical
knowledge of Z-ing is secured. This is because the practical knowledge of my Z-ing by doing
ABC, while presuming a connection between ABC-ing and Z-ing, is nevertheless neither based
on nor inferred from the observational knowledge that ABC-ing is a way to Z. (Cf. Moran 2004:
48–49.)
189
Consideration of this point leads Anscombe to this formulation: “I do what happens” (1957:
53). How is it that my doing can comprehend events that occur in the distal environment? To
answer this question, we have to draw on two of Anscombe’s most enduring contributions: the
proposal that practical knowledge is distinct from speculative knowledge in the way it is open to
correction; and the proposal that practical knowledge bears a constitutive, rather than causal,
relation to the intentional action that it directs.
It is in connection with the first proposal that Anscombe gives one of the most celebrated
examples in Intention, that of the man walking about town with a shopping list (1957: §32). The
man is observed by detectives who meticulously jot down the items that the man places within
his cart. The man’s shopping list and the detective’s list of the items might agree with respect to
the items that figure on it. But they disagree in the following respect. If the contents of the man’s
shopping bag do not conform to the list, then the man’s error lies in his performance, whereas for
the detectives the error is in their list. For the man doing his shopping, the list plays a guiding
role in the prosecution of his ends. The man’s shopping venture will be counted a success only if
the contents of his bags correspond exactly to the enumerated list with which he set out to the
store. Since this is so, the list sets down a standard or rule for the successful execution of his
aims. On the other hand, the detectives’ list plays a different role: it works not to guide the action
and set down criteria for a successful performance, but merely to accurately represent what is the
case. Anscombe’s point is that practical knowledge functions like the shopper’s list, not the
detectives’ list. Where there is a discrepancy between an agent’s practical knowledge and what is
actually taking place, the mistake lies in the performance, not in the practical knowledge.
This is one respect in which practical knowledge is practical. It is directive in that the
content of practical knowledge sets a standard for success. This is in distinction with what
Anscombe calls speculative knowledge, the aim of which is to represent a state of affairs that is
already the case and which fails by its own measure when it fails to do so. And so, Anscombe
190
writes that it is a mistake to cast about for two distinct kinds of speculative knowledge.
Understanding practical knowledge will be impossible—we will continue to find ourselves in
“utter darkness”—so long as we adhere to the speculative conception of knowledge while
“look[ing] hopelessly for the different mode of contemplative [i.e., speculative] knowledge in
acting, as if there were a very queer and special sort of seeing eye in the middle of the acting”
(1957: 57). The solution is to cease thinking of knowledge on a strictly speculative or
contemplative model. The alternative is to construe practical knowledge as directive in the
aforementioned sense (Campbell 2017).
How does this practical, as opposed to speculative, dimension of knowledge clarify
Anscombe’s formula that “I do what happens”? Because it allows us to classify discrepancies
between what occurs and that which I intend to bring about as errors in performance, rather than
errors in my practical knowledge. Clearly, such discrepancies do not get to count as knowledge
anyway—failure to accord with the facts undermines any sort of knowledge, practical or
speculative (cf. Moran 2004: 61). Nevertheless, it is significant just where the error lies in such
cases. When there is accord between the agent’s practical knowledge and the facts, it is because
the facts are fitted in the right way to the agent’s practical knowledge. What I do—what I know
practically—is what happens, because what happens must be fitted to that knowledge.
This first point—that practical knowledge is directive, and so not at all a species of
speculative knowledge—goes some way toward addressing the worry about the very idea of non-
observational knowledge of what happens ‘out there.’ But it cannot be the entirety of the
explanation of the ‘I do what happens’ formula. This is because we can accept that an agent has
directive practical knowledge of some occurrent episode while also accepting that practical
knowledge is not sufficient for the content of that practical ‘knowledge’ to cohere at all with the
world. We need some way of securing a connection between practical knowledge and the
worldly intentional action it knows.
191
Here, towards the end of Intention, Anscombe clarifies the role of practical knowledge as
constitutive of intentional action.12 Here is one place where Anscombe expresses the main idea:
[W]here (a) the description of an event is of a type to be formally the description of an
executed intention [and] (b) the event is actually the execution of an intention (by our
criteria) then the account given by Aquinas of the nature of practical knowledge holds:
Practical knowledge is ‘the cause of what it understands,’ unlike ‘speculative’ knowledge,
which ‘is derived from the objects known.’ This means more than that practical knowledge
is observed to be a necessary condition of the production of various results…It means that
without [practical knowledge] what happens does not come under the description—
execution of intentions—whose characteristics we have been investigating. (1957: 87;
footnote omitted)13
Condition (a) is not about material bodily movements, but about intentional descriptions.14
Anscombe’s claim is that intentional descriptions take on a distinctive form, one that consists in
the teleological unity evinced in the A–D order. As we have seen, this teleological unity
corresponds to the content of an agent’s practical knowledge. Anscombe goes on to say that this
correspondence is no accident. The reason that intentional descriptions possess the form that they
do is because of the way that the events properly brought under those descriptions—viz.,
intentional actions themselves—are constituted by an agent’s practical knowledge. Practical
knowledge, then, is action-constitutive: if not for practical knowledge, we could not so much as
describe bodily movements as executions of intentions.15 It is the agent’s practical knowledge or
understanding of the teleological order in the content of his practical knowledge which renders
12 On this aspect of Anscombe’s view, see especially Moran (2004), Small (2012), and Schwenkler (2015). 13 An editorial remark: I take myself to be correcting a simple typographical error in adding ‘and’ in this passage,
turning condition (a) and (b) into a single conjunction. This seems the right way to read Anscombe here, since she
prefaces this quotation with, “Putting these remarks together…” (1957: 87). 14 On the importance of this point to the general Anscombean approach, see Lavin (2016: 615 ff). 15 Cf. Pink’s (2016) use of the term ‘action-constitutive.’
192
the movements of his body an intentional action at all. With John Schwenkler, we can
characterize practical knowledge as “the formal principle that unifies an action, or that in virtue
of which certain physical happenings are constituted as parts of a person’s intentional activity”
(2015: 6).
What does it mean to say that practical knowledge is the formal principle that unifies an
action? Recall: Anscombe is concerned with sets of descriptions of events that possess a certain
form. That form is the A–D order we have discerned holding among the descriptions under which
what an agent does counts as intentional. Each such description, recall, can be given as an
admissible reply to the ‘Why?’-question. Relative to one another, the intentional descriptions A–
D possess an internal structure. This structure is teleological, in that an agent can ‘move up’ the
A–D order to say what he is doing and why he’s doing it, and ‘move down’ the A–D order to say
how he is getting it done. The question we are asking is: in virtue of what does a series of bodily
movements come under such a teleologically-ordered structure of intentional description(s)?
To bring this question into relief, consider the way we are inclined to describe other
occurrent episodes as they unfold over time (cf. Lavin 2016: 619–622). What makes an
unfolding occurrent process the very process that it is? For instance, what makes a series of
occurrent events ‘hang together’ as the different innings within a single baseball game? The
question seeks, in Lavin’s words, “some principle in virtue of which the phases of the process
constitute a whole” (2016: 619). If there were no such principle, then there would be little ground
for our sense that the very same token process is suspended at the conclusion of the seventh
inning but resumed at the bottom of the seventh inning. Or take an ongoing process with an
inbuilt terminus that is interrupted before reaching its proper endpoint. What makes it the case
that I am walking to the café, even if I never reach the café? It cannot be my eventual arrival at
the café, for I may never reach it. Still, I am walking there, despite the fact that I never arrive (cf.
193
Lavin 2016: 618).16 And what is it for an ongoing process to progress to another phase of its
activity? These questions ask after the principle that as it were ‘organizes’ these events into the
process that it is. Here again is Lavin:
Not just any collection (A, B) constitutes the progress of wider process (C). But when such
happenings (A, B) constitute the progress of some wider process (C), when they are lesser
phases of some more inclusive going on, the whole is in the parts. (Lavin 2016: 619)
By ‘the whole is in the parts,’ Lavin means that the parts are what they are in virtue of the
principle that unifies them into the whole of which they are parts. This principle is the formal
principle that unifies the collection of episodes into the progression underway. I will return to this
point.
We can raise the very same questions about an intentional action that is occurrently
underway. Suppose we take a single time-slice of an agent’s occurrent bodily movement as he is
in the midst of operating the water pump. This is no mere bodily movement. It is an event that
falls under an intentional description; as such, it bears the aforementioned structural relations to
the events falling under the other intentional descriptions that belong within the A–D order. Why
is this so? What is the principle of unity in virtue of which our chosen time-slice “constitutes the
progress of some wider process”? This is the question that practical knowledge answers, for
Anscombe. For Anscombe, it is the agent’s practical knowledge of what he is doing, why he is
doing it, and how he is getting it done which makes whatever occurs within our selected time-
slice a part of an ongoing process, namely the process of poisoning the inhabitants. For
Anscombe, practical knowledge is the formal principle of an action’s internal unity. Practical
knowledge inserts the whole into each of the parts of the agent’s ongoing intentional activity.
16 As Lavin points out, these and other features are at home in the way ordinary thought and talk employs the present
progressive to describe occurrent events. For these features of the present progressive, see Falvey (2000).
194
I have been describing two features of practical knowledge: that it is directive of intentional
action and that it serves as the formal principle that makes an intentional action just what it is.
These two features explain why we enjoy non-observational practical knowledge of what occurs
at some remove from our own bodies. The first feature ensures that where there is an error, “the
mistake is in the performance,” not the agent’s standing purported practical knowledge of what
he is doing. (The practical knowledge in such cases is merely ‘purported’ because, as Moran
points out, failure of correspondence with the world undermines any species of knowledge,
practical or speculative.) Practical knowledge of what I am doing, why I am doing it, and how it
is getting done can comprehend the occurrence of extra-bodily events in that this very knowledge
sets the standard of success for the extra-bodily events in question. When I am successful, the
success will be put down to the practical knowledge of my occurrent action.17 Secondly, practical
knowledge comprehends extra-bodily events because it is the formal principle in virtue of which
what happens counts as something I am doing at all. Practical knowledge does not play an
efficient-causal role in bringing about an action, but a formal or action-constitutive role with
respect the bodily movements and distal events that constitute it.18
PRACTICAL KNOWLEDGE AND DECOMPOSITION
The final task of this long section is to draw out just why Anscombe’s approach is not
decompositional, and indeed why it is impossible to provide a decompositional approach to
action when we begin with the ‘Why?’-question rather than with Wittgenstein’s ‘arithmetical’
17 See Small (2012: §4) and Lavin (2016: 621–622) for argument that practical knowledge secures a principled basis
on which action underway can be considered a genuine process in development. 18 Schwenkler (2015) argues that practical knowledge cannot be wholly non -observational, but that the observational
knowledge implicated in practical knowledge is nevertheless practical insofar as it works to keep an action “on
course” until completion. Schwenkler’s relaxation of Anscombe’s insistence on non -observational practical
knowledge is consistent with what I have said in this subsection, since we might motivate the very same puzzle with
the weaker question: How can there be any degree of self-knowledge in the knowledge of what happens at some
remove from my own body?
195
question. The key claim in this connection is the one that I have just been at pains to describe—
viz., that an agent’s practical knowledge of what I am doing, why I am doing it, and how it is
getting done is the formal principle of unity of an intentional action, or the “the cause of what it
understands” (1957: 87). Intentional action cannot be decomposed into non-intentional and non-
actional components because these components can only be what they are—the parts of a unified
intentional action—unless they are comprehended in an agent’s practical knowledge.
In taking this line, Anscombe’s account of intentional action is analogous to the so-called
Substance Theory of concrete particulars. On this account, the parts of a living organism are
posterior in explanation to the whole that the parts jointly compose. From this point of view, it is
mistaken to seek the organic units that together make up my living body. On the contrary, it is the
fact that each of my gross organic parts are parts of my body that they are what they are. Here is
Michael J. Loux (summarizing Aristotle) on that point:
[T]he organic parts of living beings are such that to specify what they are one must make a
reference to the whole living being whose parts they are. Thus, to specify what a liver is,
one must identify its biological function; and that, Aristotle wants to say, requires a
reference to the role of the liver in the maintenance of the whole living being to which it
belongs. He wants to claim, then, that so far from being reducible to their organic parts,
living beings as wholes are prior to their organic parts […] Thus, as organic parts of living
beings, the parts of substances have a nature or essence that presupposes a reference to
whole whose parts they are; and while they can exist in separation from those substances,
qua existing apart, they are no longer the sorts of things that can be the parts of living
beings. (1979: 171, footnotes omitted)
With respect to the parts that make up a whole organism, then, we might say with Lavin (quoted
above) that the whole is in the parts: the part of an organic unity is what it is only because of the
196
way it is arranged in the whole that it makes up. There is thus a kind of ‘downwards constitution’
in the case of ordinary organic unities: the whole is what makes a given part just what it is.
Though I find this Substance Theory of concrete particulars plausible, it is not my aim to
argue for it here. Rather, my point is that Anscombe has a similar thesis in mind; and the
argument for this Anscombean view of intentional action has been unfolding over the course of
this entire chapter so far. It is because the parts of an intentional action are constituted by the
whole that we cannot arrive at an account of intentional action by decomposing actions down to
their constituent elements. Wittgenstein’s question presupposes that a non-intentional and non-
actional mere bodily movement is an action when produced by the right kind of extrinsic
efficient cause. But on Anscombe’s approach we cannot solve Wittgenstein’s arithmetical
question because there is on this approach no such thing as a mere bodily movement in the
relevant sense.
Let us attempt to pose Wittgenstein’s question with respect to water-pumper in Anscombe’s
example. Wittgenstein asks, “what is left over if I subtract the fact that my arm goes up from the
fact that I raise my arm?” (1953: §621). In the same way, we will ask “what is left over if we
subtract the fact that the man’s arm goes up and down from the fact that he is operating the water
pump?” Read in the spirit of a decompositional approach to intentional action, this question asks
for the non-intentional and non-actional ingredients that, when added with the action-making
element, renders the man’s arm going up and down an intentional action. But this is like asking
for the ingredients that, when added together, make up a human organism. Really the direction of
explanation goes the other way: just as a liver is what it is only when integrated in a whole
human body, so too the man’s arm’s going up and down is what it is only in reference to the
teleological unity that constitutes the parts of the man’s intentional action as means that the man
engaged towards achieving his settled aim of poisoning the inhabitants in the nearby house. The
man’s arm’s going up and down is a means undertaken for an end, and this means-end structure
197
is precisely what the A–D order captures; and the A–D order itself is the content of the man’s
practical knowledge. So, the agent’s practical knowledge makes the bodily movement what it
is—a part of an intentional action that is occurrently unfolding in time.
The upshot is that we cannot answer Wittgenstein’s question in a decompositional way. We
can only say what makes a bodily movement an intentional action with reference to an agent’s
practical knowledge that both comprehends and constitutes the bodily movement in question as a
means directed towards a further end. Practical knowledge is what articulates bodily movements
into an intelligible means-end or teleological unity. Practical knowledge, as the formal principle
of this unity, makes the parts of an intentional action what they are. So there can be no question
of decomposing these parts into atoms that are rendered intentional actions by something
extrinsic to them.
To anticipate later discussion, it is important that this non-decompositional approach to
action does not go as far as utterly disintegrating the parts into the whole.19 On this latter
suggestion, we cannot speak of the parts of an intentional action at all, but merely the whole
intentional action. On such a view, an agent’s fine-grained practical knowledge of the various
intentional descriptions that constitute his intentional action are as it were flattened into but one
intentional description—the one under which he possesses knowledge of his end. (Anscombe
herself might be interpreted this way, since she speaks of the man’s knowledge of his D-ing
‘swallowing up’ his practical knowledge of his A-ing/B-ing/C-ing. Cf. 1957: 46.)
To take this suggestion, though, is to hold that the agent in Anscombe’s example possesses
practical knowledge only of one thing: poisoning the inhabitants. But surely the agent in
Anscombe’s example knows more than this. He also knows how he is getting this done. This is
why I have stressed that we can ‘move down’ the A–D order by specifying progressively more
19 I was pressed on this point by Akeel Bilgrami in conversation (December 3, 2019).
198
articulate means that the agent takes in order to realize his end of D-ing. The suggestion at issue
runs into the mirror of this problem as well. If the agent possesses practical knowledge only of
D-ing (poisoning the inhabitants), then many of the events falling under apparently intentional
descriptions must be refused this status. We cannot ask the agent why he is operating the handle
of water pump, or why he is moving his arm up and down. If the agent possesses practical
knowledge only of poisoning the inhabitants, then these latter descriptions are not intentional
descriptions at all. Correspondingly we cannot pose the ‘Why?’-questions that correspond to
such descriptions: to the question “Why are you operating the water pump handle,” the agent
should have to refuse this question application altogether by saying, “I’m not operating the water
pump handle, I am poisoning the inhabitants.” Now, while it is certainly true that the action-
constitutive role of practical knowledge allows the agent who is operating the water pump handle
to say that he is D-ing (poisoning the inhabitants), this very practical knowledge allows him to
say more than this—namely, he can say that ‘I am A-ing/B-ing/C-ing.’ Were this not the case,
then the agent’s action under these more proximate descriptions—‘I am moving my arm up and
down’ / ‘I am operating the water pump’ / ‘I am replenishing the water supply’—would not be
intentional descriptions at all and his action, under these descriptions, would be unintelligible.
This consequence is implausible. Practical knowledge constitutes the parts of an intentional
action into a unity without thereby disintegrating the parts into the most distal end that the agent
hopes to realize. This is because practical knowledge is not just of an end, but of actions
undertaken for the sake of the end’s realization. The means-end structure is constituted with the
intentional action whose formal principle is the agent’s practical knowledge.
199
Inferring and practical self-consciousness
I propose to vindicate the Action Thesis by generalizing the non-decompositional approach
to action that figures in Anscombe’s Intention. We will see that inferring to a conclusion counts
as intentional action on the very same grounds, and in the very same respects, as ordinary
intentional bodily action. And since the resulting conception is non-decompositional, it will pass
through between the horns of the dilemma facing causalist and dispositionalist accounts of
inferring.
Return now to the Taking Condition:
Taking Condition: Inferring necessarily involves (I) the thinker taking his premises to
support this conclusion and (II) drawing his conclusion because of that fact.
I will read the ‘because’ in sub-condition (II) as announcing a formal cause, or a formal principle
of unity. A thinker’s appreciation of what follows from what plays a similar role as an agent’s
practical knowledge of what I am doing, why I am doing it, and how I am getting it done: it
works to constitute the premise- and conclusion-episodes as the parts of a wider unity in the
thinker’s conscious stream of thought. These parts are explanatorily posterior to the whole in
which they figure, since apart from this inferential order they cannot figure as premises or
conclusion at all.
In working up to these claims, we must begin with a better understanding of what a thinker’s
taking is. In Chapter 3, I argued that a thinker’s taking consists in attitudinal commitments. This
commitment-theoretic account of sub-condition (I) of the Taking Condition, though, does not by
itself tell us what it is for a thinker to infer from p to q because of an attitudinal commitment. To
better understand sub-condition (II) of the Taking Condition, we must appeal to the notion of
200
self-consciousness. I will argue that the ‘because’ in sub-condition (II) constitutes an inferring by
dint of a distinctive kind of self-consciousness of one’s own attitudinal commitments.
SELF-CONSCIOUSNESS IN INFERRING
What does it mean to say that inferring is self-conscious mental activity? Here I want to set
out a recent account of this question proposed by David Jenkins (2018) in working towards a
positive proposal of my own.
Self-consciousness is an awareness that a thinker has of oneself as oneself (Rödl 2007: Ch.
1). When one is self-conscious of some mental episode one is not merely conscious of something
that happens to be oneself. Rather, in one’s self-conscious grasp of the mental episode, one
consciously grasps that one is grasping some mental episode of oneself (Smith 2017). This is
what it means to say that self-consciousness is consciousness of oneself as oneself.
Self-conscious mental activity is thus a conscious grasp of what I am doing or undergoing.
Self-consciousness in this sense is distinct from reflection on what I am doing or undergoing.
Reflection on what one is doing or undergoing involves some separate higher-level state that
takes a first-order mental episode as its object. The best case for distinguishing self-
consciousness from such reflective apprehension is perhaps that of intentional bodily action
itself. Intentional bodily action is (at least frequently) self-conscious without being reflective: a
person who walks to the café may do so self-consciously without enjoying any explicit second-
order reflective state that represents his walking. Indeed, it is a commonplace in our ordinary
intentional activities that such bouts of reflection can interfere and hinder, rather than make
possible, the first-order intentional activity that it contemplates.
To carry this conception over to the case of inferring: inferring from p to q is self-conscious
in that when one infers, one is aware of oneself as oneself in inferring as one does. My inferring
from p to q is self-conscious in that it involves an awareness that I am so inferring. Again, this is
201
not to introduce any explicit reflective belief about my own inferring as it takes place. In
inferring from p to q I need not introspectively reflect upon my own inferring in this way. Such a
reflective belief would be a belief about my first-order beliefs p and q. Perhaps we sometimes
take such a reflective stance towards our own beliefs as such. But this kind of reflective belief is
not necessary for inferring; it is not the sort of self-consciousness that, I am proposing, is
involved in every inferring.
My claim is that inferring involves self-consciousness. Inferring is not merely sometimes
self-conscious. Rather, inferring is necessarily self-conscious mental activity; a purported
inferring that fails to be self-conscious will be no inferring at all. This means that self-
consciousness of inferring is not a matter of apprehending one’s own inferring in a sensory
mode.20 When apprehending an object in a sensory mode, one’s apprehension of the object of
awareness is but contingently related to the object itself. To model self-consciousness of
inference after such sensory modes of awareness is to suppose, for instance, that my inferring q
from p reliably causes a consciousness of my inferring as my inferring.21 But since efficient
causes possess an existence that is distinct from their effects, any such causal relation can be only
contingent. This cannot secure the necessity connection between inferring and self-consciousness
of that very inferring (Jenkins 2018: §4.3).
Since we cannot understand self-consciousness of inferring as apprehension in a sensory
mode, self-consciousness “must be an indissoluble component of the self-conscious act itself”
(Kitcher 2011: 130; Jenkins 2018: Ch. 4). It is here that Jenkins proposes that self-knowledge of
one’s inferring is a conscious state or episode that is contemporaneous with the inferring itself
(2017: 106). On Jenkins’ proposal,
20 This point is frequently recorded in discussions of self-consciousness. See Jenkins (2018, Ch. 4) and references
therein, including O’Shaughnessy (2000: 106–7) and Kitcher (2011: 130). 21 See Koziolek (2017); for persuasive argument against the suggestion, see Jenkins (2018: 99–105).
202
One knows that one is inferring when and because one infers whenever one infers. This is
not because one’s inference causes or is identical to knowledge of one’s inference, but
because when one infers a contemporaneous state which is one’s knowledge of one’s
inferring constitutively depends on the conscious occurrence which is one’s inference. On
this suggestion, when a self-conscious reasoner makes an inference this suffices for
knowledge of the inference where this is understood as a matter of the obtaining of a
contemporaneous state of awareness which constitutively depends on the conscious
occurrence which is the inference. (2018: 106)
On this proposal, inferring is self-conscious because inferring from p to q guarantees that
one knows that one is inferring q from p. For Jenkins, self-consciousness of inference is a
constitutive component of one’s inferring, so that the very act that is the inferring suffices for a
contemporaneous state or episode of self-consciousness of that very inferring. This proposal
appears to secure the right connection between inferring and self-consciousness of inferring.
Jenkins goes on to clarify that this proposal does not require some further explicit judging, since
“Coming to know does not always require conscious judgment in a mediating role” (2018: 108).
This qualification is important, since as already mentioned we must keep self-consciousness
distinct from reflection.
This allows Jenkins to understand what inferring is and when it occurs:
For one to infer q from p is for one to consciously accept p and then consciously accept q
where a contemporaneous state of knowledge that one is inferring q from p constitutively
depends on the conscious occurrence that is one’s acceptance of q […] An inference is a
kind of conscious event and conscious act. Namely, it is the conscious event and act of
acceptance which is one’s acceptance of that which one infers. (2018: 112)
203
Jenkins here identifies an inferring with the episode that marks acceptance of the conclusion. On
this view, inferring is an episode of mere succession which constitutively involves a
contemporaneous state of self-knowledge of the transition from premise to conclusion. Jenkins
goes on to clarify the sense in which this species of self-conscious makes one’s inferring
intelligible to oneself, so that when one infers from p to q, one is thereby in touch with the
rational support that p provides for q.
There is a great deal to recommend Jenkins’ proposal (and I have learned much from his
treatment). But the view is open to objection. The objection can be introduced by pursuing
Jenkins’ answer to the question: When does an inferring take place in the stream of thought?
Jenkins asks: “What determines that one is inferring when one infers? The question is to be
understood as asking what, metaphysically speaking, settles that one is inferring when one
infers” (2018: 121). In contrast to causalist accounts of inferring, Jenkins writes that we
determine when an inferring occurs by way of the distinctive self-conscious knowledge we have
of its occurrence. As I have related, for Jenkins this self-conscious knowledge is constitutively
dependent on the occurrence of the inferring itself. “One’s inferences thus depend for their status
as inferences,” Jenkins has it, “upon one’s knowledge of them” (2018: 121). This proposal does
not stray into the inner sense model of self-consciousness that he rejects. Neither does Jenkins
accept that knowledge is conceptually or explanatorily posterior that which is known; if his
proposal is along the right lines, then just the opposite is true in the case of inference.
The trouble, though, is that this species of self-conscious self-knowledge, like all speculative
knowledge, is factive. I can have speculative knowledge of p only if p takes place or is the case.
My self-conscious self-knowledge that I infer q from p is no different in this respect. Like all
speculative knowledge, self-conscious self-knowledge of my occurrent inferring must be
“derived from the object known,” as Anscombe has it (1957: 87). My self-conscious self-
knowledge that I infer from p to q stands or falls as knowledge according to whether it accurately
204
reflects the facts as they are independently of my self-conscious grasp of the object known. This
is something that Jenkins, too, seems to accept: he repeatedly claims that self-consciousness of
inferring constitutively depends upon the inferring itself.
But if this so, we cannot answer the question when inference occurs, metaphysically
speaking, by citing self-conscious knowledge of inference. This is because self-conscious self-
knowledge of the inferring, on Jenkins’ proposal, constitutively depends on the inferring’s having
taken place. Thus, to cite self-consciousness of my inferring q from p as grounds for asserting
that I have inferred q from p presumes that an inferring has indeed taken place. We can appeal to
self-consciousness of inferring as grounds for asserting that inferring has taken place only if self-
consciousness of inferring determines that the inferring takes place. In importing a speculative
model of self-conscious self-knowledge, though, Jenkins has opted for a direction of constitutive
dependence that runs in the opposite direction.22 Evidently, this problem only arises because
Jenkins is working with what, adapting from our prior discussion of Anscombe, we can term a
speculative conception of self-conscious self-knowledge. On this conception, inferring
constitutively involves a self-consciousness that must accurately reflect the obtaining of some
fact or occurrence of some episode.
Apparently, we require a practical conception of self-consciousness. A practical self-
consciousness will not reflect an inferring that it knows, but rather constitute an inferring by
serving as the formal cause of what it is conscious of. On this alternative conception, rather than
self-consciousness constitutively depending on inferring, the inferring will constitutively depend
on whether an agent possesses the requisite self-consciousness. Call this practical self-
consciousness, in distinction with speculative self-consciousness.
22 In addition, it is difficult to see, on this conception of self-conscious inference, how to vindicate the Action Thesis.
(It should be noted, though, that Jenkins is not out to vindicate any such thesis.) This is because there seems nothing
to distinguish self-consciousness of inferring per se from self-consciousness of purely receptive thought.
205
PRACTICAL SELF-CONSCIOUSNESS
Practical self-consciousness does not constitutively depend upon the occurrence of an
inferring. Rather, inferring constitutively depends on the practical self-conscious grasp that a
thinker enjoys of the normative relations among intentional episodes in the stream of thought.
Practical self-consciousness is to inferring what practical knowledge is to intentionally acting.
As a species of self-consciousness, this practical grasp of one’s inferential activity will be a
grasp of one’s inferring, as one’s own inferring. What makes this species of self-consciousness
properly practical is that it is ‘the cause of what it understands.’ This is just what Ursula Coope
finds in the work of Thomas Aquinas. According to Coope’s interpretation of Aquinas, the power
to make rational judgments (to infer)
can be exercised in a way that is at once, and inseparably, a kind of self-knowing and a
kind of self-causing. More precisely, his [viz., Aquinas’] view is that because intellect is
self-reflective…it is capable of a certain kind of exercise, φ-ing, which is such that one
makes it the case that one is φ-ing by being aware of oneself as φ-ing. (Coope 2013: 7)
In what does this practical self-consciousness consist? How does practical self-
consciousness make inferring possible? I propose that we answer these questions just by drawing
on the Taking Condition, especially sub-condition (II). Just as Anscombe suggests that practical
knowledge of what I am doing and why constitutes an intentional action, practical self-
consciousness of what follows from constitutes a premise-conclusion order from the intentional
episodes within the conscious stream of thought.
An immediate problem arises for this proposal. Above, I followed Jenkins in the view that
self-consciousness of inferring consists in a contemporaneous state of self-knowledge that one is
inferring from p to q, where this knowledge constitutively depends on that very inferring. But
what is the content of this self-knowledgeable state? It cannot be a self-standing state that
206
represents the very inference undertaken, namely the one from p to q. If this were the case, then
the idea of a practical self-consciousness would come down to this: that a state which is
contemporaneous with my inference from p to q, and which represents the very inference from p
to q, makes it the case that I infer from p to q. This leaves us with the need, though, to explain
how it is I come to have self-knowledge of the inference from p to q: how do I come to the self-
conscious grasp of the inference which explains the very inference it is knowledge of? This
question threatens a circular and uninformative account of inferring as constituted by practical
self-consciousness.
In this respect, practical self-consciousness appears to differ from Anscombe’s treatment of
practical knowledge of intentional bodily action. The present threat of circularity can be brought
out by considering this (apparent) difference. As we have seen, Anscombe holds that practical
knowledge of occurrent intentional action is knowledge that constitutes a certain form of
description—namely, the teleological unity evinced by the A–D order. It is not necessary that this
practical knowledge is sourced in some prior stretch of practical reasoning. All the same, it
reflects a possible exercise of practical deliberating, wherein the agent progressively calculates
more determinate means to realize the end from which deliberation begins. It is this practical
knowledge, this self-conscious grasp of the A–D order, which constitutes ‘mere bodily
movements’ as the parts of an intentional action. In this, the form of an intentional bodily action
is given by the practical knowledge wherein an agent grasps its occurrence. (Later, I will offer a
more specific proposal about how to understand how this knowledge is manifest in an intentional
agent.)
However, this cannot be transported to the case of inferring. My inferring from p to q cannot
be constituted by a practical representation that q follows from p. If that were so, then we should
need to appeal to a still further species of practical self-consciousness that represents, at a still
‘higher’ level, the inferential relation between p and q. This is the very supposition that launches
207
a regress. To be sure, it not the same Carroll-style regress we have met with in the previous
chapter, as there is no suggestion here that the consciousness of q’s following from p is a premise
from which I infer that q follows from p. Rather, the regress arises from positing within that
which explains my inferring—a practical self-conscious grasp that q follows from p—the very
thing we want to explain, namely an inferring q from p that includes my taking it that p supports
q.
Luckily, no such ‘meta-inference’ needs to be ascribed to the thinker. Practical self-
consciousness is not a grasp that p follows from q, but rather a grasp of the rightness of the very
inferring that it constitutes. When I infer from p to q, I am self-conscious that inferring from p to
q is correct. Inferring is constituted by this practical self-consciousness of the rightness of the
very inferring that is underway. This self-consciousness of the rightness of the transition between
p and q is constituted by two elements working in tandem.23
The first element of practical self-consciousness is the grasp that the mental activity in which
I am engaged falls under a certain standard of success.24 The standard in question is not merely
truth, but more broadly truth in the service of achieving overall rational unity among my
intentional episodes (Rovane 1998). Practical self-consciousness articulates this demand by
imposing on an inferring a standard of success that must be met by the very inferring of which it
is self-conscious. In undertaking an epistemic inference, this demand for rational unity is itself
governed by the aim of truth: the inference is defective qua inference just in case the truth of the
conclusion is not guaranteed by the truth of the premise. At least, this is so for deductive
inference. Practical self-consciousness of inductive inferring is weaker, demanding only
likelihood of the truth of the conclusion given the (perhaps likely in turn) truth of the premise.)
23 Coope (2013: 10) also attributes this twofold structure to Aquinas, with some differences concerning the role of
rational unity in the overall account. I have learned from Coope’s interpretation of Aquinas. 24 Cf. my earlier discussion of Anscombe’s shopping list example.
208
Practical self-consciousness of practical inferring likewise imposes a normative demand for
rational unity. But here the demand is itself governed with the aim of goodness or promotion of
the agent’s rational ends. In practical inferring, the transition from premise to conclusion is
oriented towards the formation of the sort of teleological unity that characterizes Anscombe’s A–
D order.
Second, practical self-consciousness involves a grasp that the very inferring underway
marks success in meeting the normative demand for rational unity, either by preserving truth or
preserving goodness. My transition in thought from p to q involves not only a grasp that this
transition falls under a normative demand, but also that the transition from p to q marks success
in this very demand. This second element of practical self-consciousness can be mistaken. I
might be mistaken in grasping that the transition from p to q satisfies the normative demand for
rational unity. This is an important possibility to accommodate: in some particular instance, the
inference from (p → q; ¬p) to ¬q might fail to meet the demand for rational unity because, in
that instance, the transition is not truth-preserving. But the thinker may fail to recognize this. Bad
reasoning still counts as reasoning: what distinguishes it from good reasoning is something
which may in some particular instance lie outside the thinker’s ken.
Practical self-consciousness of inferring is consciousness of myself in the midst of the
transition from p to q, of the demand for rational unity that normatively governs the transition,
and of the transition as a successful making-good on that demand. This practical self-
consciousness is not itself an inference; it is, instead, an inference-constituting grasp of a
transition in thought. I shall go on to explain how this practical self-consciousness constitutes the
inference that it grasps. Before doing so, it is important to ward off a potential misunderstanding.
In Chapter 2, I argued that we cannot understand inferring in instrumentalizing terms. That
is, we cannot understand a possible course of inferring going as follows:
209
8. I want to aim at truth.
9. P
10. P → Q
11. The inference from P to Q is truth-preserving.
12. So, I can accept Q as a way to satisfy my aim for truth.
13. So, Q.
This model of inferring is instrumentalizing, I said, because it represents even every epistemic
inferring as something done in order to promote an explicit aim: namely, the aim to produce true
and knowledgeable beliefs. On this model every epistemic inference involves a practical
syllogism. Yet this model takes for granted the very thing we need to explain: if the thinker
cannot infer (13) from premises (9) and (10) together, the additional practical syllogistic
premises (8), (11), and (12) cannot help.
At this point it might be asked whether I have tacitly employed the very same
instrumentalizing conception of inferring. Aren’t the two elements making up practical self-
consciousness—grasp of the normative demand and grasp that the transition in thought satisfies
this demand—analogous to the major and minor premise of a practical syllogism? And if so,
haven’t we snuck in a kind of hidden practical syllogism tucked away inside every epistemic
inference, just as the above model (8)–(13) represents?
No. The grasp that constitutes a thinker’s practical self-consciousness is not itself a premise
that figures in the thinker’s reasoning. A thinker does not infer from his grasp of the intentional
episodes that figure as premises in his reasoning. Rather, it is practical self-consciousness which
constitutes the very relation that his premise-episodes are understood to bear to the conclusion-
episode. So it is not as though the thinker begins with a set of premise- and conclusion-episodes
which are brought into an inferential relation due to the addition of more premises. That really
would lead to the instrumentalizing conception of inferring that I rejected in Chapter 2. Practical
210
self-consciousness does not supply additional premises at all; it rather constitutes an intelligible
premise-conclusion order holding among the contents of conscious intentional episodes.
This point is borne out, I think, by common sense reflection on our responsibility-practices
surrounding answerability for the attitudes. Suppose I do infer from p to q. If I am asked ‘Why
do you believe that q?’, I need cite only p—the premise that, in inferring, I take to support q. The
‘Why?’-question here does not ask for some additional practical syllogism that captures the
desirability of inferring to q in order to satisfy my desire for knowledgeable belief. The ‘Why?’-
question, in asking for the grounds from which I inferred that q, takes this structure for granted.
The elements comprising practical self-consciousness explain how the ‘Why?’-question is able to
presume these elements. Namely, it is because the ‘Why?’-question asks for the premises from
which one inferred to the conclusion cited in the question, and, on the view I am attempting to
set out, it is practical self-consciousness which makes the inference what it is at all.
In sum, along with Soteriou (2013) and Jenkins (2018) I hold that inferring involves
occurrent self-conscious self-knowledge. But whereas they model occurrent self-conscious self-
knowledge as constitutively dependent upon what it is knowledge of, on the present account the
direction of constitutive dependence moves in just the other direction. Practical self-
consciousness is an occurrent conscious episode that makes an inferring what it is. To see just
how this is so, the notion of practical self-consciousness needs to be integrated with Chapter 3’s
account of attitudinal commitment.
WHAT INFERENCE IS
We can now account for both sub-conditions of the Taking Condition by integrating the
foregoing account of practical self-consciousness with what was said in Chapter 3 regarding
attitudinal commitment.
211
Taking Condition: Inferring necessarily involves (I) the thinker taking his premises to
support his conclusion and (II) drawing his conclusion because of that fact.
To briefly recapitulate my treatment of sub-condition (I): a thinker’s ‘taking it’ that his
premises support his conclusion does not consist in some separate premise or a disposition to
move from premise-episode to conclusion-episode. Rather, all of the normativity relevant for a
thinker’s taking lies within the very first-order attitudes from which he infers. Attitudinal
commitments represent the content of the thinker’s premises in a distinctive imperatival fashion:
as propositions or prospective actions that an agent ought to respect in subsequent thought and
action. From the engaged stance of reasoning and deliberating, the premises from which I infer
take on the character of rational obligations. This is because the attitudinal commitment to p, and
the attitudinal commitment to (p → q), agglomerate together in a normative demand to accept q.
From the engaged stance, the premises from which a thinker is prepared to infer are already rules
to reason by.
Inferring involves an occurrent transition between contents in the thinker’s conscious mental
life. Inferring to a conclusion is a dynamic conscious mental episode that involves taking a leap
or step in thought. Inferring to a conclusion, then, involves a thinker’s appreciation of normative
relations holding among contents that span over several conscious intentional episodes in the
stream of thought.
Practical self-consciousness is itself an occurrent episode of self-conscious self-knowledge.
Specifically, it is self-knowledge of the first-order occurrent intentional episodes that figure in the
inferring that is known. Suppose I infer from p to q. Because I infer from the engaged stance, the
intentional episode representing p figures as an attitudinal commitment to uphold the rational
obligations that are grounded in the (apparent) truth of p. These obligations concern e.g.
accepting p’s consequences and rejecting what is incompatible with p. But an inferring involves
212
a dynamic transition from premise to conclusion—here, from p to q in the conscious stream of
thought. Practical self-consciousness is a comprehension that to make this transition and accept
that q is a making-good on the very attitudinal commitment to p that I am, from the engaged
stance, concerned to uphold. What is self-consciously self-known in an inferring is that the
attitudinal commitments that the thinker incurs in a premise are discharged in the conclusion-
episode. This is because when I enjoy practical self-consciousness of a transition between
intentional episodes, I grasp that the episodes in question are (a) assessable according to a
normative standard of correctness and (b) mark success according to that very standard. Practical
self-consciousness of an inferring is a self-conscious grasp that a thinker’s attitudinal
commitment to the premise is discharged by acceptance of the conclusion.
Without practical self-consciousness, one can enjoy only a mere succession of intentional
episodes. This is so even if these intentional episodes are taken in from the engaged stance that
reveals their imperatival dimension. Such a mere succession of intentional episodes would
consist in acknowledgement of one’s rational obligations to p and then in an occurrent conscious
episode that represents q. But there could be in this scenario no grasp that adopting a rational
obligation to q discharges my standing commitments to p. To infer q from p, I must have self-
conscious self-knowledge that accepting q is to make good on my standing commitment to p.
Sub-condition (II) of the Taking Condition requires that an agent infer from p to q because
of his taking it that p supports q. The ‘because’ in sub-condition (II) does not announce an
extrinsic efficient cause, but rather a formal cause, or a formal principle of unity, that makes an
inferring what it is. This formal principle of unity is the thinker’s practical self-conscious grasp
of his attitudinal commitments being discharged in the acceptance of a novel content. Therein,
the thinker grasps that a set of intentional episodes both falls under and succeeds according to a
rational obligation intrinsic to his attitudinal commitment. In this, practical self-consciousness
works in much the same way as practical knowledge of intentional action. Practical knowledge
213
constitutes intentional action by bringing events under a certain form of description, one that
exhibits teleological unity. In the same way, a thinker’s practical self-consciousness of what
follows from what constitutes a series of conscious intentional episodes as a structured unity in
the stream of thought. This structured unity is not a teleological unity of means-end relations, of
course, since an inferring does not exhibit instrumentalizing relations. It is a more fundamental
rational unity, a premise-conclusion order. The parts of this integrated premise-conclusion order
are the occurrent intentional episodes in the stream of consciousness that make up the inferring,
such as the intentional episode that p and the one representing q. When these episodes figure in a
premise-conclusion order, q is self-consciously self-known as a discharging of the attitudinal
commitment to p. Practical self-consciousness of an inferring, like practical knowledge of an
intentional action, is thus ‘the cause of what it understands.’ This means not only that practical
self-consciousness always attends a thinker’s inferring to a conclusion; in addition it means that
without practical self-consciousness what occurs in a thinker’s conscious stream of thought does
not come under the description inferring to a conclusion at all (cf. Anscombe 1957: 87–88).
The thinker’s self-conscious appreciation that accepting q on the basis of p is a way of
unifying the episode representing p with the episode representing q. The premise-conclusion
order that is constituted by a thinker’s practical self-consciousness is a relation between p and q
such that in appreciating that p provides adequate support for q, one also thereby appreciates that
q is rationally reachable from p. Just as in the case of teleological unity among intentional act-
descriptions, then, a thinker can ‘move up’ the premise-conclusion order to discover conclusions
that are supported by grounds, or ‘move down’ from the conclusion to an appreciation of the
grounds for that conclusion. A thinker’s appreciation of this premise-conclusion order allows us
to understand the transition from p to q as a step or movement from a premise-episode to a
conclusion-episode as such. It is certainly appropriate to say, then, that a thinker infers to a
conclusion because of the fact that he takes his premises to support his conclusion.
214
It might be worried that this emphasis on a thinker’s own self-conscious grasp of the
premise-conclusion order ushers in an overly psychologistic conception of inference. It may
appear that a good inference is just what a thinker presumes to be a good inference, however
erroneous the presumption may be. But recall that we are attempting to characterize how an
inferring occurs within a thinker’s conscious stream of thought, from within the thinker’s own
rational point of view. So, we need to distinguish between objectively good or bad inferences and
those that are merely apparently good or bad from the thinker’s point of view. In the good case,
apparently good inferrings involve a practical self-consciousness that constitutes a premise-
conclusion order among intentional episodes that, as an objective matter, involve a premise that
really does support the conclusion. In the bad case, one infers from p to q when in fact p does not
provide (adequate) support for q. This is to ‘appreciate’ a premise-conclusion order where in fact
none exists. But bad reasoning is still reasoning; and reasoning, I am arguing, requires that a
thinker appreciate at least an apparent premise-conclusion order among intentional episodes. By
the same token, there is more to good reasoning than merely seeming right: appreciating a
premise-conclusion order that really does exist among facts is to get on to an objective relation in
the world. So, I have not offered a psychologistic construal of good inference since non-
psychologistic relations of rational support exist in the objective world. (Or at least, such a
conclusion is compatible with everything I have said here.) A good inference is one that is made
up of contents that accurately reflect the facts.
PRACTICAL SELF-CONSCIOUSNESS AND DECOMPOSITION
The foregoing formal causal reading of ‘because’ in the Taking Condition yields a non-
decompositional understanding of inferring. When it comes to practical self-consciousness of an
inferring, we can say with Lavin that “the whole is in the parts.” The parts of an inferring—
namely, the premise-episode(s) and the conclusion-episode—are what they are only in virtue of
215
their being united in a whole. That whole is the premise-conclusion order, an order that is
constituted by a thinker’s self-conscious grasp that accepting the conclusion discharges the
rational obligations in the premise(s). Thus, the premise- and conclusion-episodes that are the
parts of a thinker’s inferring cannot be what they are apart from the thinker’s practical self-
consciousness. A mere succession is not rendered an inferring in virtue of some extrinsic factor,
but in virtue of the way the premise- and conclusion-episodes are apprehended together in the
stream of thought. All of this is strictly analogous to what Anscombe writes ab out the way
practical knowledge is constitutive of action. What makes a bodily movement an action is the
way it is apprehended as a means to an end; in the same way, what makes a succession of
intentional episodes an inferring is the way the thinker apprehends one of these episodes as a
conclusion reachable by the premise.
This non-decompositional approach to inferring evades the dilemma facing causalist
accounts and dispositional accounts, a dilemma that I described in a prior section of this chapter.
Here again is the dilemma. To satisfy the Taking Condition, the causalist can either (i) posit an
explicit premise as a cause of the conclusion-episode, a premise about what follows from what;
or (ii) account for the thinker’s taking by positing a network of causal dispositions. However,
horn (i) gives rise to a regress from which no inferring at all can issue, whereas horn (ii)
undermines sub-condition (II) of the Taking Condition since the play of dispositions renders the
thinker’s taking a mere epiphenomenon. The upshot is that the causalist cannot account for both
sub-conditions of the Taking Condition together.
Yet the non-decompositional approach canvassed here evades both horns. We needn’t
ascribe any additional premise whose vehicle plays an efficient causal role in producing the
inferring. The only premises in the inferring are those pertaining to the first-order matters which
bear on the thinker’s live deliberative question. We thereby sidestep Carroll-style regresses. With
respect to the second horn of this dilemma, the thinker’s ‘taking’ really does play a causal role in
216
the inferring. The difference with extant theories of inferring, though, comes in when we specify
the kind of causality at issue. I contend that the relevant causal role is formal: a thinker’s taking
is the formal principle of unity that unites premise and conclusion in an inferring.
As I examined in the case of overt intentional action, it is important not to take this point
about non-decomposition too far. One might suggest that in taking a non-decompositional
approach to inferring, the separateness of the constituent premise- and conclusion-episodes is (as
it were) dissolved into a whole. According to this suggestion, the thinker enjoys only an
occurrent intentional episode that represents the contents of both the premise- and the
conclusion-episodes together. This suggestion is analogous to one I considered in connection
with Anscombe’s non-decompositional approach to intentional action. There, the suggestion was
that a non-decompositional approach to intentional action subsumes all of the teleological
structure holding among the parts of an intentional action into the ultimate end for which the
agent acts. On this suggestion, it is incorrect (or at least, inappropriate) to say that the man in
Anscombe’s example moves his arm in order to operate the pump in order to replenish the water
supply in order to poison the inhabitants; rather, he simply replenishes the water supply. Here,
the suggestion is that a non-decompositional approach to inferring subsumes all of the premise-
conclusion structure among the parts of an inferring into the conclusion that terminates the
thinker’s inquiry. On this suggestion, it is incorrect (or at least, inappropriate) to say that a
thinker’s inferring consists in two separate intentional episodes representing p and q respectively;
rather, inferring consists in a single intentional episode that represents the rational relation
between p and q together.
As before, the present suggestion threatens to dissolve the articulate structure evinced in the
premise-conclusion order. As I mentioned above, just as we can move up the A–D order to
determine why an agent acts, so too we can move down the A–D order to determine how he
prosecutes his aim. I pointed out above that the same is true when a thinker infers to a
217
conclusion: one can move up the premise-conclusion order to determine the conclusion reached,
and down the order to determine the grounds in the light of which the thinker accepts the
conclusion. This articulate structure is important to preserve in an account of inferring. An
inferring is not merely the occurrent conscious contemplation of a proposition; rather, inferring
to a conclusion involves the grounds on the basis of which one infers.
It might be replied that the premise-conclusion structure is preserved in a single intentional
episode that represents the rational relation between p and q. For instance, I might enjoy an
occurrent conscious thought that If Alice is in the library, then Bill is at the library. This episode
represents that Alice’s presence at the library is a sufficient condition for Bill’s presence there. A
thinker who enjoys such an episode (and understands its content) can thus take Alice’s presence
at the library as a premise for the conclusion that Bill is at the library. While this may be so, the
resulting conception runs counter to the notion of inferring that we isolated back in Chapter 3.
There, I pointed out a distinction between a thinker’s conscious contemplation of a logically
structured content and the thinker’s inferring one content from another. Inferring to a conclusion
involves a thinker’s appreciation of normative relations that span the occurrence of two separate
intentional episodes, whereas the former does not. I might enjoy a conscious intentional episode
representing that If Alice is at the library, then Bill is at the library without inferring the
consequent of this conditional from its antecedent. I grant that conscious mental life frequently
involves the occurrence of intentional episodes that represent a logically structured content. But
such episodes are not my target in this dissertation. My target is the phenomenon of normative
relations that span distinct intentional episodes so that a thinker infers a conclusion from an
appreciation of the normative force of premises.
The upshot is that we cannot solve the problem of inference by seeking an inference-making
element that, when added to a case of mere succession, results in a thinker’s inferring to a
conclusion. It is the thinker’s self-conscious appreciation that the content represented by an
218
intentional episode discharges an attitudinal commitment which makes that episode count as a
conclusion at all.
Intentional action as inferring to a conclusion
Over the course of this chapter I have prosecuted two parallel investigations. I have
suggested that the problem of action and the problem of inference are frequently misconceived in
the same way: as asking for the decompositional ingredients that, when added together, make an
action or an inferring what it is. Both problems can be advanced by denying this assumption. The
chief motivation for taking this non-decompositional line to inferring is that one thereby
sidesteps the dilemma facing decompositional accounts. And there is another upshot in the
offing: if we take this non-decompositional approach to inferring, we can see why inferring is an
action, indeed a type of fully intentional action.
Let’s begin on the side of overt intentional action. How does Anscombe’s approach
illuminate the nature of action? How does her distinctive non-decompositional approach explain
what an action is? Certainly not by proffering a type of efficient cause that precedes a bodily
movement that is by itself considered non-intentional. Rather, we have seen that Anscombe’s
approach begins by (hypothetically) eliciting from an agent the reasons for which he acts. When
the agent responds (or is disposed to respond) in the right way to such a ‘Why?’-question, the
description that figures in the ‘Why?’-question is an intentional one, as is the description by
which the agent picks out the further end for the sake of which he is acting. For Anscombe—as
for Davidson (cf. above, note 8)—an event is an action when it falls under an intentional
description in just this way. Typically, the very same intentional action will fall under several
intentional descriptions, so that e.g. a man’s arm movement is describable as an operating of the
water pump, as a replenishing of the water supply, and as a poisoning of the inhabitants of a
219
nearby house. The teleological unity among these descriptions makes the events so described
count as intentional action. In turn, this teleological unity as a whole is constituted by the agent’s
practical knowledge of what I am doing, why I am doing, and how I am getting it done. This is
what it means to say, with Anscombe, that an event is an action in virtue of its intelligibility in a
certain light: this intelligibility comes down to understanding the means-end relations holding
among the intentional descriptions that figure in the agent’s practical knowledge. For Anscombe,
practical knowledge is action-constitutive.
Here we must register an important aspect of the relation between means and ends,
something I remarked upon back in Chapter 2. The notion of a means is a normative notion, one
that we understand in terms of a reason for action. As John Gibbons puts the point, “To see A as a
means to B is to see the connection between A and B as a reason for A-ing. To see A as a means
rather than merely as a cause or a necessary or sufficient condition is to see it as something you
have reason to do” (2009: 90). This is manifest in ordinary deliberation, wherein an agent begins
with a practical commitment to some end and concludes with a commitment to pursue some
specific means to its realization. Take, for instance, the practical syllogism we considered back in
Chapter 2:
14. I ought to: stay awake this afternoon and get some exercise.
15. Walking to the café is the (all-things-considered) best way to stay awake this afternoon
and get some exercise.
16. So, I shall walk to the café.
The major premise in this practical syllogism specifies my end; the minor premise consists in
some instrumental knowledge about how to realize my end; and the syllogism concludes in my
settling on the means that the minor premise links to the realization of my end. The means
specified in (15) figures as a premise that supports the conclusion that I settle upon. To figure as a
220
premise in this way, step (15) must present to me an instrumental reason to pursue the means, a
reason whose normative force is sourced in the instrumental connection these means bear to the
end I am committed to advancing. This is how the notion of a means is already normative.
I reiterate that Anscombe explicitly denies that every intentional action is preceded by
explicit reasoning of this sort. Her claim, though, is that the teleological unity among the
intentional descriptions of an action is a deliberative order, in that the practical syllogism
“describes an order which is there whenever actions are done with intentions” (1957: 80). This
allows us to better understand what is involved in saying that the intentional descriptions of my
action hang together as a teleological unity. To say this means that when I walk to the café in
order to get some exercise with the further aim of staying awake this afternoon, the teleological
unity among these intentional descriptions is underwritten by a deeper and more fundamental
premise-conclusion order. Hence, in acting intentionally something like the Taking Condition
holds: in performing a means undertaken in the service of a further end, (I) I take these means to
be rationally supported by the premises that are jointly constituted by my end and by my
instrumental know-how; and (II) I perform the means because of this very relation of rational
support. Though Anscombe does not explicitly draw this moral in quite these terms, I submit that
it is a natural way of understanding her claim that the practical syllogism describes an order that
is present in every intentional action. Of course, the premise-conclusion order that underwrites
the teleological unity of an intentional action is not specifically epistemic. But this only means
that the premise-conclusion order at issue is practical, describable in terms of a practical
syllogism rather than epistemic inference. The premise-conclusion order admits of both practical
and epistemic varieties.
On this basis we might be inclined to understand Anscombe’s claim in terms of aptness. On
this conception, my practical knowledge is action-constitutive in virtue of my grasp that the
intentional description of my means is apt to figure as a conclusion that is reachable from certain
221
premises; these premises are in turn given by the intentional descriptions under which I enjoy
practical knowledge of my end and the instrumental know-how that I exploit in furtherance of
this end. However, I might enjoy such a grasp without performing any intentional action at all. I
might idly contemplate the practical syllogism (7)–(9) the evening before the afternoon in
question, and only later, the next day, put this knowledge to use in furtherance of my end. Such a
scenario is quite distinct from what Anscombe seems to have in mind. The line of argument in
Intention characterizes what an action is when considered as a unified process occurrently
underway (cf. Lavin 2016). The premise-conclusion order in its teleological species is present in
the agent’s self-conscious activity as he acts; and it is this self-consciousness which gives an
intelligible structure to what is happening as the agent engages in teleological activity. Practical
knowledge is action-constitutive in virtue of the way it renders a structure and form to an
occurrent bout of bodily and extra-bodily activity as it progresses. (Recall Anscombe’s slogan: I
do what happens.)
If this is the right way to read Anscombe, then practical knowledge is not best conceived as
a static state that represents intentional descriptions that are merely apt to figure in a practical
syllogism. Rather, practical knowledge must be conceived as an activity. My walking to the café
cannot be considered an intentional action unless in walking to the café I self-consciously grasp
this very activity as something I am doing for the sake of a further end. In effect, to self-
consciously engage a means as such is to draw the conclusion of the practical syllogism whose
distinctive teleological order is present whenever an action is done with an intention. Notice that
this interpretation of Anscombe’s claim preserves her emphatic denial that every intentional
action is preceded by some explicit chain of deliberating in thought. It is to say, though, that
every intentional action is constituted by reasoning—it is just that the reasoning in question is
sometimes done with the body, as it were.
222
On this interpretation, action-constitutive practical knowledge is a species of inferential
activity. To act for the sake of an end just is to infer to a conclusion. Inferring to a conclusion, in
turn, consists in a practically self-conscious grasp that my attitudinal commitments are rightly
discharged in the acceptance or endorsement of a conclusion. Anscombe is rightly skeptical that
inferring must take place before the commencement of a bodily movement. Sometimes, the
acceptance or endorsement of the relevant conclusion-episode is the bodily movement that is the
means undertaken for the sake of my further end. These further ends, in turn, are the attitudinal
commitments that constitute the premises from which I infer in acting as I do (and for the sake of
which I act). This practical self-consciousness constitutes a span of normative relations between
intentional episodes—namely, the occurrent intentional descriptions under which I possess
practical knowledge of what I am doing as I am getting it done. In sum: when I act intentionally,
the teleological unity among the intentional descriptions under which I possess practical
knowledge of my intentional action is itself sourced in the practical self-consciousness that
constitutes an inferring. And so, intentional action itself is to be understood as a kind of
inferential activity.
To be sure, this sort of inferential activity is manifested in diverse ways across the three
different cases we have been considering. In epistemic inference, practical self-consciousness
constitutes a premise-conclusion order of evidentiary support; in practical inference, one of
instrumental connection with a further end; and in intentional action, the bodily movement is a
conclusion drawn to instrumentally further such an end. But these variations are species of a
single genus: an agent’s self-conscious grasp that constitutes an intelligible premise-conclusion
order among occurrent intentional episodes.
This result vindicates the Action Thesis—the thesis that inferring to a conclusion in thought
is an intentional yet nonvoluntary action. In the first section of this chapter, I indicated that it is
sufficient for an action to count as intentional if it is done for reasons. Inferring to a conclusion is
223
an episode that is done for reasons in the sense that a thinker’s practical self-conscious grasp of
the rational support that his premises grant for his conclusion is what makes an inferring the very
sort of episode that it is. This is the very ‘formal causal’ sense of ‘because’ that, I argued, allows
us to make sense of sub-condition (II) of the Taking Condition on inferring.
Importantly, this also allows us to make sense of what it is to act for reasons even in the case
of overt intentional action. An agent acts for reasons when he enjoys practical knowledge of his
bodily movement under an intentional description that figures as a conclusion drawn from
attitudinal commitments that are the reasons from which he acts. Since the ‘for a reason’
formulation is understood univocally across epistemic inferring, practical inferring, and
occurrent intentional bodily action, we vindicate the Action Thesis by pointing to the very same
action-constitutive factor across all three cases. Importantly, this way of vindicating the Action
Thesis does not claim that inferring possesses the action-making quality that renders a mere
bodily movement an intentional action. Rather, inferring must be an intentional action because
ordinary bodily intentional actions themselves count as such in virtue of practical knowledge;
and this practical knowledge, I suggest, is itself a way of inferring to a conclusion (‘with the
body,’ as it were). Action is constituted by a kind of self-consciousness that is equally manifested
in certain bodily performances as well as in conscious thought. In this, inferring to a conclusion
counts as intentional action on the very same grounds as overt intentional actions.
I want to close this chapter by considering two worries about this manner of vindicating the
Action Thesis. The first has to do with voluntariness; the second, with intentional control.
The Action Thesis says that inferring to a conclusion is an intentional yet nonvoluntary
intentional action. But if intentional bodily actions are just ways of inferring to a conclusion,
then it would appear that inferring is at least sometimes voluntary. Intentional bodily action is
voluntary in both of our senses of ‘voluntary’: I can φ for any reason that recommends φ-ing, and
224
I can φ just out of an intention to do so. Are we to count voluntary intentional bodily actions as
voluntary inferences?
As to the reason-based sense of voluntariness, notice that even in intentional bodily action
the reasons-for-which I φ are constrained by the range of considerations I take to grant count in
favor of φ-ing. So long as I am rational, I cannot φ for reason R unless I really do take R to
provide sufficient support for φ-ing, and this is not something subject to my whims. So there is
still a sense in which even intentional bodily action is nonvoluntary. Still, there remains a
significant contrast between intentional bodily action and inferring to a conclusion in thought.
Inferring to a conclusion in thought is nonvoluntary in that I cannot form a given belief or
intention for reasons that recommend having that belief or intention, but only for reasons that
recommend the content of the belief or the intention that will be formed thereby. In light of this,
the Action Thesis should be understood as claiming only that voluntariness is not an essential
feature of intentional action—it is a feature that can be present in intentional bodily action but go
missing in the case of inferring to a conclusion in thought. On this reading of the Action Thesis,
inferring to a conclusion in thought is a nonvoluntary intentional action; whereas intentional
bodily action is a form of voluntary inferring (with the proviso that one cannot φ for R unless one
really does take R to support φ-ing, which is not a voluntary phenomenon). This reading of the
Action Thesis allows for voluntary as well as involuntary intentional actions, which is supported
by anyone who thinks that judging and deciding are actions (see e.g. McHugh 2011).
I turn now to the will-based sense of voluntariness. An action is voluntary in this sense in
that I can φ just out of an intention to do so, whereas I cannot infer to the belief that p or to the
intention to ψ just out of an intention to do so. Here it is important to understand how future-
directed intentions, so important for the will-based sense of voluntariness, function in the
diachronic life of a practically rational agent. These intentions are themselves typically the
outcome of reasoning, and where they are rational they are at least apt to come in for revision in
225
the light of reasoning. Future-directed intentions work to spare a cognitively limited agent the
considerable costs of always having to deliberate, here and now, on what to do (cf. Bratman
1987). By settling now on my conduct at a later time, I spare myself the burden of having to
decide then what I shall do. The outcome of this future-directed deliberating is a stable intention
to act, one that I carry forward up until the point of action. What matters for my diachronic life as
a practically rational agent is that at the point of action I should be reliably ‘in touch with’ an all-
things-considered case for performing the best action among the alternatives open to me. And
this is just what such a forward-looking intention enables me to do: intentions work to preserve
and to transmit a correct decision that was undertaken from an earlier time. At the point of
action, I needn’t reenter deliberation to figure out what to do; I can simply execute the intention
that is the persisting outcome of the deliberation I engaged in earlier.25
In turn, the relationship between an intention and an intention-executing voluntary action is
nevertheless an inferential one. Inferring to an intention is nonvoluntary, in that the intentions I
form via reasoning are constrained by my take on what reasons there are to act. The resulting
intention can thus function to preserve (apparently) good reasoning—reasoning that embodies an
appropriate regard for the practical reasons that figure in my deliberating. Having concluded my
practical reasoning in thought, the intention remains as something like a standing
recommendation for how to act, given my earlier nonvoluntary inferring in thought. In ways I
have already explained in connection with Anscombe’s shopping list example, at the point of
action my intention takes on the status of a rule or normative standard that measures the
correctness of the performance I engage in to realize it. To execute an intention, I must enjoy a
practical self-consciousness which constitutes a premise-conclusion order holding among my
standing intention and my performance. In this way, the intention figures as an attitudinal
25 There are many accounts available of how long-range intentions manage to keep us appropriately in touch with the
reasons that figured in earlier deliberation. See e.g. Rovane (1998: 146–159), Ferraro (2010), Heeney (2020).
226
commitment that I discharge in my performance. As I have already argued, this just is to infer to
a conclusion.
Voluntariness in the will-based sense thus plays a key role in the diachronic life of a
practically rational agent. It is because I can φ just from an intention to φ—an intention that is
sourced in deliberation about what I ought to do—that intentions can work to spare us the costs
of having always to redeliberate at the point of action. Seen in this light, voluntariness in the
will-based sense is what allows the execution of a long-term intention to count as discharging the
very attitudinal commitments in the light of which intentions are formed in the first place. If
inferring to a conclusion is fundamentally a matter of discharging attitudinal commitments, then
this second will-based sense of voluntariness in intentional bodily action does not take away
from my claim that inferring is fundamental to all intentional agency. On the contrary, will-based
voluntariness allows us to discharge attitudinal commitments that are not presently in view.
This brings me to the second worry, this one having to do with control. It might be thought
that intentional action is fundamentally a matter of control. In turn, control is plausibly construed
as an agent’s ability to exercise a will-bases sense of voluntariness. On this conception, I control
some event or process just to the extent that I can bring that event or process into accord with an
antecedent intention that I possess for how it ought to occur or unfold. But since inferring to a
conclusion in thought is not voluntary in this will-based sense, it appears we do not exercise
control over what we infer. On these grounds it might be claimed that inferring to a conclusion in
thought cannot be intentional action.
My reply is to resist this identification of intentional action and will-based voluntary control.
We are inclined to make this identification, I suggest, only out of a (reasonable enough)
preoccupation with intention-executing bodily action. To be sure, this sort of control is highly
significant for our diachronic lives. But when we follow Anscombe in better understanding what
is involved in making an intentional action what it is, we ought to resist the identification. While
227
will-based voluntary control is an important feature of some intentional actions, it is not an
essential feature of intentional action as such. Inferring to a conclusion in thought, while an
intentional action, is nonvoluntary and thus not subject to the kind of control at issue in will-
based voluntariness.
We can best understand the Action Thesis as the claim that inferring in thought is a
nonvoluntary intentional action. Inferrings ‘with the body’—intentional bodily actions
constituted by practical knowledge—are indeed voluntary. But even such bodily performances
stand as conclusions relative to the attitudinal commitments in the light of which they are
undertaken.
Conclusion
I began this chapter by drawing a close analogy between the problem of action and the
problem of inference. Since decompositional approaches to the problem of inference face a
dilemma that I described over Chapter 3, I sought to exhibit Anscombe’s non-decompositional
approach to intentional action. I developed a parallel non-decompositional approach to inferring
to a conclusion in thought. Once this non-decompositional approach to inferring to a conclusion
(in thought) is made explicit, however, we see that intentional bodily action is itself an inferential
phenomenon. This vindication of the Action Thesis does not subsume inferring to a conclusion to
intentional bodily action; on the contrary, the argument in this chapter vindicates the Action
Thesis by subsuming intentional bodily action to the phenomenon of inferring to a conclusion in
thought. This vindication of the Action Thesis also allows us to make good on both sub-
conditions of the Taking Condition, thereby completing the action-theoretic account of inference
I began in Chapter 3.
228
In the next chapter, I want to conclude this dissertation’s argument by showing how inferring
to a conclusion (in thought), conceived as an intentional action along the lines described over this
and the previous chapter, grounds our moral responsibility for the attitudes of belief and
intention.
229
Chapter 5
Responsibility Through Reasoning
Recounting the desiderata
I began in Chapter 1 by arguing that we bear moral responsibility for the attitudes of belief
and intention. I also argued there that we seek an account of the attitudinal agency that
legitimates attitudinal blame. To this end, over Chapters 3–4 I set out an action-theoretic account
of inferring. I have argued that inferring to a conclusion in thought is an intentional yet
nonvoluntary mental action. In this concluding chapter my task is to bring this notion of inferring
to bear on the problems with which we began in Chapter 1. These problems concern the peculiar
species of responsibility that we bear for the attitudes of belief and intention. There, I suggested
that responsibility for the attitudes is grounded in a thinker’s performance of mental actions.
Now, my task is to show how inferring to a conclusion in thought—conceived along the lines I
have sketched over Chapters 3–4—grounds the legitimacy of attitudinal blame.
I will begin by recounting each desideratum along with its rationale. This will serve to
summarize the main line of argument of the dissertation up until this point. Then I take the
desiderata individually but out of order, starting with Desiderata 2–5 and concluding with
Desideratum 1 concerning the grounding for our responsibility for the attitudes.
Desideratum 1: Responsibility-grounding. An account of attitudinal agency must explain
why are morally responsible for our beliefs and intentions when we are. It must also work to
legitimate the forceful rational criticism involved in attitudinal blame—blaming another on
account of what he believes or intends. Correlatively, our account of attitudinal agency ought to
230
guide our intuitions about why we are not morally responsible for beliefs and intentions in
circumstances where attitudinal blame is inappropriate.
To satisfy this desideratum, our account must be sensitive to our responsibility-practices
surrounding the attitudes. In Chapter 1, I argued that these practices evince our answerability for
belief and intention. We are answerable for our beliefs and intentions in that we are appropriate
targets for requests for justification for the relevant items. This justificatory demand can be
communicated via Anscombe’s ‘Why?’-question: Why do you believe that p? Why do you intend
to φ? Standing as a proper target for these justificatory demands means that the agent is
susceptible to distinctive patterns of normative appraisal according to how well or poorly the
recipient meets the demand. Suppose I am asked to satisfy the demand to justify my belief that I
shall pass my dissertation defense. I might reply that I believe I shall pass because having this
belief helps me to sleep at night. This is a (practical) reason to enjoy the belief that p, but it is not
a reason why p, as it does not at all bear on p’s truth. Or again, I might be asked to satisfy the
demand to justify my intention to drink a mild toxin. I reply, “I will be rewarded handsomely for
having this intention.” Here again we find a reason to enjoy the intention to drink the toxin but
not a reason to drink it. The reason in question does not all bear on whether to drink the toxin.
These responses to the ‘Why?’-question open me to distinctive patterns of normative appraisal.
These appraisals go beyond ‘mere grading’ (as when I judge that I am shorter and less handsome
than my brother). They can reach genuine forms of blame. There is a sense in which a thinker is
at fault for persisting in the belief that p for reasons that fail to offer the requisite support for p,
or the intention to φ for reasons that fail to render φ-ing good, desirable, valuable, and so on.
At the same time, there are attitudes for which these normative appraisals seem to be ruled
out by our responsibility-practices. Consider my perceptually-based belief that there is a blue
coffee mug before me on the desk. Though these beliefs can certainly be appraised according to a
normative standard (say, accuracy within a specified range), our responsibility-practices do not
231
license attitudinal praise and blame for them. A thinker believes falsely but is not open to
attitudinal blame for undergoing an illusion, for instance. Here, it seems that our responsibility-
practices embody a tacit policy of excusal or exemption.
To satisfy the first desideratum, our account must explain why we are answerable for our
beliefs and intentions, when we are; why we are not, when we are not; and how it is that we
come in for genuine forms of praise and blame, rather than mere grading, on account of what we
believe and what we intend.
In Chapter 1, I registered an Activity Condition on responsibility. According to this
condition, responsibility for x is grounded in agency over x, or with respect to x, or with respect
to an event or condition implicated in the obtaining of x. If an agent is directly responsible for
some episode or state of affairs, it seems reasonable to suppose that the agent did or does
something that explains the occurrence or obtaining of x. In at least a wide range of paradigmatic
cases, responsibility implies agency. Where this is not the case—where the ‘Why?’-question is
put to someone whose agency is not implicated in the state or episode that the question asks
about—it seems that the ‘Why?’-question merely falls to the ground. The ‘Why?’-question is
simply refused application by the reply, “I had nothing to do with that.”
Another reason to impose the Activity Condition centers on an ethical concern regarding the
fairness of blaming someone. As Watson recognizes, holding another accountable for an action
typically involves the imposition of sanctions, penalties, or even bodily punishments. It is unfair
to subject another to such penalties unless the (putative) wrongdoer enjoyed, at the point of
action, the opportunity to avoid incurring them (Watson 1996: 237). Of course, we do not subject
one another to punishment or sanction on account of what they believe or intend. Still, a version
of Watson’s point is applicable to the attitudes. It seems we ought not direct attitudinal blame to
another unless there is something the agent can do to rectify the blameworthy state of affairs. In
addition to the Activity Condition, an account of our agency over the attitudes should make good
232
on this ethical dimension of interpersonal attitudinal blame. We can understand both of these
facets of attitudinal blame by making out a sense in which an agent acts to bring about his beliefs
and intentions.
Desideratum 2: Nonvoluntariness. It would be quick work to explain our responsibility for
the attitudes by carrying over to the case of our mental lives the very same agency-conditions
that ground our responsibility for ordinary bodily action. It is reasonable to suppose that my
answerability for walking to the café is grounded in whatever goes into making my walking an
intentional action on my part. And part of what makes the action intentional, we might think, is
that it is voluntary in the reason- and will-based senses. Recall: an action is voluntary in the
reason-based sense if it can be performed for any reason that recommends its performance; and it
is voluntary in the will-based sense if it can be brought about by an intention to bring it about.
Overt intentional actions are voluntary in both senses. I can walk to the café for any reason that
recommends doing so. The reasons-for-which I settle on walking to the café will then be reflected
in the intention that is formed in my deciding to walk there. This intention goes on to play a
controlling role in my subsequent behavior. Accordingly, I will be able to provision these reasons
in response to a ‘Why?’-question directed at me in respect of my action of walking to the café.1
It is at least a reasonable starting point in an account of my responsibility for walking to the café
that doing so was a voluntary action of mine, for on that count my walking to the café is a
manifestation of my intentional agency.
Yet, this starting point is not available to us in an explanation of our answerability for belief
and intention. This is because our attitudes are nonvoluntary. As I have mentioned in previous
chapters, our attitudes are nonvoluntary in both senses of ‘voluntary.’ In the reason-based sense,
1 Here I am not making the ‘classical compatibilist’ point that contemplation of a different set of reasons would lead
me to perform the very same action I perform. My point is that the twin senses of ‘voluntariness’ that I have set out
in this dissertation form a very plausible starting point for theorizing about why that my performance can be
considered an action for which I am answerable.
233
you cannot believe that p for any reason that makes the belief that p a good belief to have. The
same goes for intentions to φ. Nor can you believe that p or intend to φ in the will-based sense of
voluntariness, by dint of enjoying some attitude-controlling intention to instill in yourself the
belief that p or the intention to φ. For if our beliefs and intentions were voluntary in both of these
senses, there would be nothing to keep us from believing and intending ‘at will.’ In turn, this
would mean that I might believe that p irrespective of the evidence I take to bear on whether p or
intend to φ irrespective of the desirability of φ-ing. And this it seems we cannot do. We cannot
believe or intend ‘at will.’ We must ground our responsibility for the attitudes in a form of
agency that we exercise over belief and intention. But this form of agency must be nonvoluntary.
These first two desiderata thus combine to place an important constraint on attitudinal
agency. We must understand what it is to exercise a nonvoluntary form of agency over our own
minds. In what might such a form of agency consist? In addition to the first two desiderata, then,
we should also hope to learn something about this exercise of agency and why we should count
the generation of (some of our) beliefs and intentions as something that an agent does, rather
than something that an agent merely undergoes in conscious mental life.
Desideratum 3: Productivity. One thing we should hope to learn about this non-voluntary
exercise of agency is the sense in which belief- and intention-generation, qua expression of
agency, counts as a doing or an action on one’s part. Ordinary bodily actions are productive: they
involve the bringing about of events or episodes that, absent the action, otherwise would not
have occurred. My raising my arm from a stationary to an upright position involves my arm’s
taking on that trajectory through space over a given interval of time; Jones’ killing of Smith
involves the occurrence of Smith’s death. As against those who locate mental agency in
something like an epistemic basing relation, I argued in Chapter 1 that our attitudinal agency is
likewise productive in that through its exercise an agent generates novel beliefs and intentions.
In this way, we can understand the generation of belief and intention as playing an analogous
234
role in attitudinal agency as the rising of my arm, or Smith’s death, plays in intentional bodily
action: as constituting the result of the intentional exercise of attitudinal agency.
I closed Chapter 1 with the suggestion that we perform non-voluntary mental actions that
generate beliefs and intentions. I nominated the candidate mental episode-types of judging-that-p
and deciding-to-φ which consist in the generation of beliefs and intentions, respectively.
Desideratum 4: Non-instrumentalizing. One promising way to provide an account in line
with the aforementioned constraint is to construe judging-that-p and deciding-to-φ just like any
other intentional bodily action: as mental episodes that come about as the causal upshot of an
intention. This is the suggestion that I pursued, and ultimately jettisoned, in Chapter 2. According
to the instrumentalist about mental action, judging-that-p and deciding-to-φ are not themselves
mental action-types; instead, they are passive mental episodes that we bring about by performing
proximate means that go on to cause a judging or a deciding. An agent may open reasoning or
inquiry with the intention to come to some judging or deciding or other; but the specific judging-
that-p or deciding-to-φ is never itself an intentional mental action.
I argued in Chapter 2 that we cannot apply this model of attitude-formative mental action. In
the first place, we do not form intentions to judge and decide just as we do. If our attitudes were
formed by such intentions, then the actual judging or deciding at issue would be rendered
redundant, for the agent, in merely possessing the intention to judge-that-p or decide-to-φ, would
already be committed to the truth of p or to φ-ing. And in any case, if I could judge-that-p or
decide-to-φ just out of an intention to judge or decide, then believing and intending really would
be voluntary in just the sense that Desideratum 2 rules out.
This is why the instrumentalist proposes that judging-that-p and deciding-to-φ are the
outcome of distal intentions to do something that eventuates in a belief or an intention. This is an
instrumentalizing conception of judging and deciding, since it holds that judging and deciding
are mental actions that are causally mediated by the agent’s performance of more immediate
235
mental means. On this instrumentalizing conception, I am responsible for the belief that p and
the intention to φ in the same way that I am responsible for the distal causal consequences of the
bodily actions I perform ‘immediately,’ without having to do something else in order to bring
them about. In Chapter 2, I centered discussion on the proposal that we perform reflective mental
actions, such as turning one’s attention to considerations within one’s repertoire or posing
questions to oneself. Some have suggested that these reflective mental actions initiate ‘ballistic’
causal processes that eventuate in one’s coming to some attitude or other.
I argued in Chapter 2 that this model of belief- and intention-formation falls prey to a
vicious regress from which no attitude at all can issue. I will briefly reiterate this argument with
respect to the generation of intention. Suppose an agent performs a reflective mental action in
order to induce some decision or other. But if that is so, then the agent has already decided—
namely, the agent has decided to perform the reflective mental action in question in order to
bring about some decision or other to act. So, we cannot appeal to reflective mental actions that
are at some instrumental ‘remove’ from the mental episode that the agent, in endeavoring to
deliberate, performs the reflective mental action in order to bring about.
The fourth desideratum on our account is that it must be non-instrumentalizing. One way to
frame a non-instrumentalizing account of our answerability for the attitudes is to look for the
locus of mental action not in instrumental mediation, but in the very inferential transitions to a
novel attitude that close reasoning and deliberating. Judging-that-p and deciding-to-φ are not
episodes we act in order to bring about. Instead, to judge and to decide is just to infer to a
conclusion.
Desideratum 5: The unity of action. The picture of attitudinal agency that emerges from these
desiderata seems to differ starkly from ordinary intentional bodily action. Notice, first, that
intentional bodily action is instrumentalizing. This is evinced by the reasons-explanations to
which actions are susceptible. My action of walking to the café can be explained by citing (a) my
236
desire to obtain a cup of coffee and (b) my belief that by walking to the café I will be in position
to realize this desire. This is an instrumentalizing act-explanation: it identifies my action of
walking to the café as a means to some further end that I hope to bring about by dint of its
performance. Because intentional bodily action is instrumentalizing, there is a ready sense for
why the voluntariness of action points up a respect in which an agent controls the course of his
bodily movements. The elements (a) and (b) that make up an instrumentalizing action-
explanation designate reasons that the agent takes to count in favor of the very action underway
(cf. Hieronymi 2011). Clearly, these instrumentalizing action-explanations rest on the
presupposition that the belief/desire complex (or its contents) plays a guiding or executive role in
the obtaining of the action that it explains. And this connection between instrumentalizing act-
explanations, voluntariness, and control plausibly secures a familiar form of responsibility for the
action the agent performs.
This overriding explanation of responsibility is not available to us, though, if we hold to
Desiderata 1–4. This is because these desiderata call for an account of attitudinal agency that is
nonvoluntary, non-instrumentalizing, and yet nevertheless productive and responsibility-
grounding. This leaves us with a potential worry. Even if we ground responsibility for the
attitudes in the indicated sort of attitudinal agency, will it not effectively tear asunder the concept
of ‘action’? Will we not be forced to admit that there are ‘two kinds of agency,’ as Pamela
Hieronymi (2009) argues? It will be all to the good if, instead, we can preserve the unity of
action by explaining why the mental actions that ground responsibility for the attitudes are
nonvoluntary, non-instrumentalizing, productive, and of the very same kind as ordinary
intentional bodily actions. This is our fifth and last desideratum—to provide an account of
attitudinal agency that respects the unity of action.
My burden for the rest of the chapter is to show how the account of reasoning—inferring to
a conclusion in thought—satisfies all five of these desiderata. In prosecuting this account, I will
237
take the desiderata somewhat out of order. As I have indicated, the general shape of the account I
want to offer grounds responsibility for the attitudes in a type of intentional mental action that
normally generates the attitudes for which one is directly responsible. But this must be a type of
intentional mental action that satisfies Desiderata 2–5. In keeping with this, I will begin with
these latter desiderata, showing how the account of inferring offered so far is non-voluntary,
productive, non-instrumentalizing, and of a kind with ordinary intentional bodily action. This
will position us to understand how inferring to a conclusion grounds responsibility for the
attitudes.
Desideratum 2: Nonvoluntariness
Starting with nonvoluntariness: what is it that makes inferring nonvoluntary? Crucial in this
connection is the Taking Condition on inferring (Boghossian 2014: 4):
Taking Condition: Inferring necessarily involves (I) the thinker taking his premises to
support his conclusion and (II) drawing his conclusion because of that fact.
When a thinker infers from p to q, he must take p to support that q. Analogously, when he infers
to the intention to φ, he must take the practical reasons from which he infers to support φ-ing.
The Taking Condition renders inferring nonvoluntary. I cannot infer from p to q for any
reason I take in favor of making this inference. An eccentric billionaire might offer me an
extravagant reward for inferring from p to q. This incentive certainly generates a compelling
reason for me to infer from p to q. But my attempts to net this reward will come to nothing
unless, from the engaged stance of inquiry, I reckon that p provides sufficient support for q. And
the same is true when I infer to an intention. By extension, I cannot enjoy the belief that p or the
intention to φ for any reason I take to count in favor of having that very belief or intention. The
238
reasons-for-which I can believe that p or intend to φ are constrained by what I take, from my
own point of view, to count in favor of the truth that p or for doing φ.
On these points I agree with Pamela Hieronymi (2005, 2006, 2008, 2009). With Hieronymi,
this way of preserving nonvoluntariness leaves room for ways of manipulating our own minds in
such a way as to instill the belief that p or intention to φ in ways that are not sensitive
considerations that support p and φ-ing, respectively. For instance, I might take a belief- or
intention-inducing pill. Upon taking this pill, I will be left with some belief or intention which I
most want to have, irrespective of considerations that normatively support the content of the
belief in question, or which count in favor of performing the action I am left intending. As
Hieronymi (2006) points out, taking such an attitude-inducing pill is fully voluntary: I can take
such a pill for any reason I take to recommend having the belief that p or the intention to φ. For
instance, I might induce in myself the belief that I will pass my dissertation defense for the
reason that having his belief helps me to sleep at night. And I might induce the intention to Drink
the mild toxin for the reason that having this intention will net me a reward from an eccentric
billionaire. But these ways of influencing my own mind do not abide the Taking Condition.
While they do register my appreciation of the reasons for having the belief that p or intention to
φ, their genesis does not involve my appreciation of considerations that bear on p or on φ-ing. In
this, attitude inducement does not abide the Taking Condition, and cannot count as inferring.
They thus present no challenge to the claim that inferring is nonvoluntary—since the indicated
methods are not processes of reasoning at all. They are, rather, ways of fixing one’s attitudes in
ways that circumvent reasoning, which is constrained by the Taking Condition.
To be sure, there is certainly a process of practical reasoning at play in these examples.
Specifically, the examples involve practical reasoning that terminates in the intention to take a
pill that will induce some belief that p or intention to φ. But these stretches of practical reasoning
consist in inferrings which do abide the Taking Condition, where the conclusion is an intention to
239
induce the relevant attitude. These intentions in turn are hostage to one’s taking it that there is
sufficient practical reason to induce the attitude in question. (In the above case, the eccentric
billionaire’s extravagant offer might be worth the ‘cost’ of a false belief.) And these conclusions
are reached by considerations that count in favor of these very intentions to induce the attitude.
I turn to the second sense of ‘voluntariness.’ In this second will-based sense, an action is
voluntary when I can perform that action just out of an intention to do it. Inferring is
nonvoluntary in this sense as well. I cannot infer from p to q just out of an intention to do so, for
the very same reason: my inferring from p to q is constrained by my take on what is the case. If I
could infer from p to q just out of an intention to do so, and then inferring would also be
voluntary in the reason-based sense, since my intentions in turn are constrained by the reasons I
take to support the performance of the intended action. Since inferring to a conclusion in thought
is nonvoluntary in the first (reason-based) sense, it must also be nonvoluntary in this second
(will-based) sense.
At this point, we can use the notion of voluntariness to clarify the relationship between
inferring and reasoning. In contrast with inferring to a conclusion, reasoning is voluntary. I can
set myself the deliberative questions ‘Whether p?’ and ‘Whether to φ?’ for any reason that
motivates the search for an answer. And I can turn my attention to these questions just out of an
intention to do so. Reasoning, then, is a voluntary conscious mental activity. But it is also one
that takes on a concrete formal aim or telos: the resolution of a deliberative question (such as
‘Whether p’ or ‘Whether to φ’) in such a way as to mark overall rational unity among a thinker’s
intentional episodes. It is this formal aim which distinguishes genuine reasoning (practical as
well as epistemic) from idle mind-wandering, brainstorming, or randomizing plays of
association. Reasoning is thus oriented towards conclusion in an inferring. So, while there may
be some stretches of reasoning that do not terminate successfully in the resolution of a thinker’s
deliberative question, reasoning cannot be what it is if it does not aim towards inferring. On the
240
other hand, inferring to a conclusion does suffice for a stretch of conscious mental activity to
count as reasoning.
Desideratum 3: Productivity
Inferring is also a productive type of intentional episode. This is so for two reasons. In the
first place, an inferring is an occurrent episode in the stream of thought. Inferring is something
that happens. Qua occurrent episode, inferring shares an ontological category with intentional
bodily actions and other non-actional events. As in Chapter 1, we can distinguish here between
occurrent episodes and states. The difference is marked by how entities in either category take up
time (Steward 1997; Soteriou 2013). States, while they might have a time of onset, do not take
up time in the way that episodes do. Standing propositional attitudes such as beliefs and
intentions are surely mental states, whereas conscious mental happenings in the stream of
thought—e.g. a judging-that-p or a deciding-to-φ—are occurrent intentional episodes. Evidently,
inferring to a conclusion belongs in this latter class. An inferring is a conscious episode that
occurs at some determinate time in the thinker’s stream of thought. Inferrings involve a
conscious transition, step, or leap to a conclusion.
Secondly, inferring to a conclusion is paradigmatically an attitude-generative transition.
When a thinker infers from p to q, the episode that represents q persists in the form of a
corresponding stably persisting attitude with the same content. This is most familiar in the case of
belief. Suppose I infer from All men are mortal and Socrates is a man to Socrates is mortal. The
conclusion of this inferring is a judging-that-Socrates is mortal. The occurrence of this judging
marks the onslaught of a corresponding belief that Socrates is mortal. Paradigmatically, this
belief takes the form of an enduring attitudinal commitment. This is the sense in which inferring
to a conclusion is attitude-generative: the conclusion of an inferring marks the formation of a
241
corresponding attitude on the thinker’s part. If this were not so, then it would be hard to make
sense of the idea that inferring is a route to novel knowledge. At least paradigmatically, we grow
our store of standing attitudes by engaging in reasoning and inferring to new conclusions. This
could not be done unless inferring were attitude-generative in the sense at issue, at least
paradigmatically. (Later, I shall consider deviations from the paradigmatic cases, in which one’s
inferring to q does not result in the stable acquisition of a novel attitude.)
Given that inferring is paradigmatically attitude-generative, we can designate the formation
of an attitude as the characteristic result of an inferring. In Chapter 2, I described results as the
necessary components of an action that are by themselves insufficient for the occurrence of an
action of the associated type. For instance, the result of my arm-raising is the rising of my arm,
and the result of my killing Smith is Smith’s death. In the same way, the generation of a belief or
intention is the result of an inferring.2 The result of an action is normally that which the action
produces or brings about. This is why actions cannot be identified with their results: my arm-
raising is not the rising of my arm, but rather the episode that consists in bringing it about that
my arm rises (McCann 1998: Ch. 4). An action produces a result but does not consist merely in
the result’s occurrence. Inferrings have the same structure. An inferring is not the occurrence of
the conclusion-episode, as such an episode can perfectly well occur without an inferring having
taken place. An inferring is rather the dynamic episode that partly consists in the occurrence of
the conclusion-episode. So, we can designate attitude-formation as the result of an inferring, just
as arm-rising is the result of arm-raising and Smith’s death is the result of killing Smith. In virtue
of their ontological-temporal category and their status as attitude-generative episodes, inferrings
are productive mental episodes. They are the right kinds of things to count as a thinker’s doing.
2 What about cases in which a thinker infers from p to q but the generated belief that q is not retained? I do not think
that these are cases of an inferring that fails to generate an attitude. After all, the judging -that-q which is the
conclusion of the inference marks the onset of the attitude. In the indicated cases, the attitude is merely exceedingly
short-lived.
242
The natural suggestion is that judging-that-p brings about the belief that p, whereas deciding-
to-φ brings about the intention to φ. Now, it need not be claimed that judging is the only way to
bring about the belief that p; there may be ways of coming to believe that p without the agent’s
judging-that-p. The same goes for deciding-to-φ.
But Matthew Boyle (2009; 2011) has presented a dilemma for this natural suggestion.
Boyle asks: At the time when an agent judges-that-p, does the agent at that time believe that p? If
so, then judging-that-p does not produce a corresponding belief at all, but instead merely
expresses a belief that the agent already holds. On the other hand: if the agent in judging-that-p
does not believe that p at the time of the judging, it is hard to see what judging could possibly
consist in. An agent who judges-that-p must therein take p to be true; and if the agent takes p to
be true, how can we deny that the agent believes that p? This seems to throw us back onto the
first horn of the dilemma.
The way out of the dilemma is to claim that judging-that-p is a dynamical event-type, one
that consists in the formation of a belief that p. Judging and believing are not distinct existences.
Rather, judging-that-p is the onset of the belief that p, so that judging-that-p brings about the
belief that p by being a constitutive earliest temporal part of the belief that p. On this view, the
episode of judging and the episode of coming-to-believe that p are token identical. This reply
blunts the first horn of Boyle’s dilemma. We can maintain that at the time when the agent judges-
that-p, he believes that p in that very moment. But it does not follow that the agent believed p
before the judging took place since the time of judging is the onset of the belief.
Such a dynamical understanding of judging and deciding fits well with my proposal that to
judge-that-p or decide-to-φ is simply to make an epistemic or practical inference (respectively)
that accords with the Taking Condition. What it is to perform an inference involves an agent’s
moving from a (set of) premise-episode(s) to a conclusion-episode. The inference is something
more than the mere occurrence of the conclusion-episode; rather, the inference consists in the
243
dynamical change within an agent’s stream of thought from one episode to another. On this view,
reasoning is a productive species of agency in that it involves coming to new attitudes on the
rational basis of those intentional episodes that form one’s prior deliberative position.
This is a promising strategy, but it faces an immediate question. There seem to be cases in
which an occurrent judging that p fails to give rise to a stable belief that p. Christopher Peacocke
provides this widely-cited case:
Someone may judge that undergraduate degrees from countries other than their own are of
an equal standard to her own, and excellent reasons may be operative in her assertions to
that effect. All the same, it may be quite clear, in decisions she makes on hiring, or in
making recommendations, that she does not really have this belief at all. (Peacocke 1998:
90)
Such cases suggest that an agent can judge-that-p without thereby forming the stable belief that
p. If so, then such a judging will be extrinsic from the stable belief that is typically or even
paradigmatically associated with the judging. This possibility directs us back to the second horn
of Boyle’s dilemma. Addressing this challenge will help point up a distinctive advantage offered
by the commitment-theoretic approach I canvassed in Chapter 3.
Notice that it is part of Peacocke’s example that ‘excellent reasons are operative’ in the
administrator’s judging that undergraduate degrees from countries other than her own (say, the
UK) are of an equal standard to her own (say, those issued in the USA). On my proposal that
inferring is attitude-formative, to say that such reasons are ‘operative’ is just to say that the
administrator infers from these reasons that UK degrees are equal to USA degrees (all else
equal). This means that the administrator, in forming this belief, appreciates that certain of her
own normative commitments are discharged in accepting that UK degrees are equal to those
issued in the USA. On this basis, the administrator falls short of her attitudinal commitments in
244
failing to uphold that UK degrees are of an equal standard to degrees issued in the USA.
Furthermore, this latter belief about the equality of degrees issued in the UK and the USA is one
that ought to surface in her behavior, especially in ‘the decisions she makes on hiring, or in
making recommendations,’ and like matters. Crucially, this is something of which the
administrator possesses self-conscious self-knowledge. On the commitment- and action-theoretic
account of inferring I have sketched over Chapters 3–4, we can readily diagnose this
administrator as an agent who falls well short of her commitments.
This helps to clarify the real upshot of Peacocke’s example. The example purports to evince a
case of judging-that-p without the corresponding belief that p. And we are encouraged to say that
the agent lacks the belief that p because the agent’s actions fail to accord with the judging. This
verdict is too hasty, though. It overlooks an important possibility: that the agent, in making
decisions about hiring or recommendations, falls short of her attitudinal commitments. Here we
have an agent whose attitudinal commitments stand in stark tension. To be sure, the reasons that
the administrator possesses certainly recommend resolving the conflict in the direction of
fairness in hiring and recommendations. But that is just to say that the administrator is
criticizably irrational, in that her behavior fails to live up to her epistemic commitments. (Or, to
say the same thing differently, the administrator fails to discharge her epistemic commitments in
action.) The superior pedigree that the administrator attaches to undergraduate degrees issued in
her own country may be upheld by dint of akrasia, self-deception, bad faith, or any number of
rational faults on the administrator’s part. But casting attitudes as commitments allows us an
important latitude in ascribing them to agents, since it belongs to the nature of a commitment that
an agent can be at fault by failing to live up to it. Failing to live up to one’s commitment does not
disparage the attribution of that commitment. Rather, what disparages the very attribution of a
commitment is the failure to take oneself to be at fault for failing to live up to it, as Bilgrami
argues (2006: 241–242). Our administrator—supposing with Peacocke that her judging is based
245
on excellent ‘operative’ reasons—surely would take herself to be open to rational criticism on the
basis of her failure to live up to her epistemic commitments in action. Otherwise, we should
question whether we really have been given a case in which the agent’s judging accords with the
Taking Condition.
This theoretical alternative is afforded to us only by recognizing that our attitudes are
commitments, possessive of an imperatival dimension that one can fail to discharge in
subsequent behavior. Once we recognize the normative dimension of the attitudes, though, we
can diagnose Peacocke’s administrator as just such an agent. The administrator does have the
belief—the attitudinal commitment—that corresponds with her judging. It is just that the
administrator fails to live up to the commitment that is partly constitutive of that belief. To
challenge our claim that judging-that-p consists in the formation of the corresponding belief that
p, then, we require an agent who judges-that-p, and so satisfies the Taking Condition in doing so,
while nevertheless subsequently refusing to accept criticism (even self-criticism, issued ‘in
private’) for failing to discharge the commitment to p in further epistemic and practical dealings.
I claim that any such case will only dampen our conviction that the agent’s inferring really does
satisfy the Taking Condition.
By extension, since there is no worry about judging-that-p without the corresponding belief
that p, we may accept the dynamical solution to Boyle’s dilemma: judging-that-p consists in the
dynamic formation of the belief that p, and analogously deciding-to-φ and the intention to φ.
Desideratum 4: Non-instrumentalizing
Inferring is non-instrumentalizing. A thinker infers from p to q not for reasons that
recommend this inference, but because the thinker, in reasoning, takes p to bear positively on
whether q. As I discussed before, a thinker might induce the belief that q in some other way (by
246
taking a drug, say). Along these lines, the thinker might induce the occurrence of an intentional
episode that represents that q every time the thinker enjoys an occurrent intentional episode
representing that p. But neither form of inducement counts as an inferring. This is because in
neither form of inducement is the ultimate explanation for the conclusion-episode sourced in the
thinker’s taking it that the episode is supported by a premise-episode. And so, the episodes in
question do not really make for a premise-conclusion unity, at all.
It might be thought that this line of thought breaks down for the case of practical reasoning,
reasoning that terminates in an intention to φ. Practical reasoning is terminated in an intention to
φ undertaken for reasons that recommend φ-ing by pointing up φ-ing as good, desirable,
valuable, and so on. Unless one is deliberating as to a final end, though, practical reasons that
count in favor of φ-ing will do so by showing ways that φ-ing will contribute to some further
aim. If I begin with the aim to ingest some caffeine, then the fact that I can attain caffeine at the
café will count as a consideration in favor of walking to the café. This consideration constitutes
for me an instrumental reason: a reason that counts in favor of walking to the café by pointing up
its positive instrumental contribution to some further aim, that of ingesting caffeine. Doesn’t this
render inferring to an intention instrumentalizing in the forbidden sense?
No. Inferring to the intention to φ for instrumental reasons for φ-ing is in perfect keeping
with what makes for a good inference. What Desideratum 4 rules out is inferring to the intention
to φ for reasons that instrumentally rationalize having the intention to φ, rather than φ-ing itself.
For as we have seen many times, there are instrumental reasons for having the intention to φ that
do not at all bear on whether to φ. But if I could infer to the intention to φ for these reasons of the
‘wrong kind,’ then inferring would be voluntary. Since inferring is non-voluntary, these are not
reasons-for-which I can infer to the intention to φ. Suitably understood, then, inferring to a
conclusion—even to a practical conclusion—is non-instrumentalizing.
247
Desideratum 5: The Unity of Action
While inferring to a conclusion is productive, the fact that it is nonvoluntary and non-
instrumentalizing seems to mark it off from more familiar exercises of rational agency. This is
Pamela Hieronymi’s contention in “Two Kinds of Agency” (2009). There, Hieronymi argues that
the agency we exercise over our attitudes is a complex of two more basic kinds of agency (2009:
138). In this section, I want to set out Hieronymi’s arguments for this view, noting as I do many
deep points of agreement. Ultimately, I will argue that when we better understand each of these
putatively distinct kinds of agency, there are few grounds for distinguishing them after all. There
is but one kind of intentional rational agency, and its paradigmatic exercise is fully on display in
both inferring to a conclusion and executing an intention in bodily action.
Hieronymi endorses a normative conception of the attitudes of belief and intention. She
writes that beliefs and intentions are constituted by a thinker’s commitments, such as a
commitment to (the truth of) p or to (the goodness, desirability, value in) φ-ing. In “Two Kinds of
Agency,” Hieronymi works from an initial assumption about these attitudes: namely, that beliefs
and intentions “embody their subject’s answer to some question or set of questions” (2009: 138).
For instance, a thinker’s belief that p embodies his answer to the question whether p, whereas his
intention to φ embodies his answer to the question whether to φ. “Insofar as one intends to φ,”
Hieronymi goes on, “one is vulnerable to certain sorts of criticisms and open to certain sorts of
questions—in particular, one is open to questions and criticisms that would be satisfied by
reasons that (one takes to) bear positively on whether to φ”—and ditto for belief (2009: 138). For
Hieronymi, a thinker believes that p/intends to φ if and only if the thinker is committed to a
positive answer to the question that the attitude embodies (2009: 139).
One way to exercise control over what you believe or what you intend, then, is to change
your answer to the relevant questions. As Hieronymi puts it, “we control these aspects of our
248
minds because, as we change our mind, our mind changes—as we form or revise our take on
things, we form or revise our attitudes” (2009: 140). Hieronymi calls this evaluative control. You
exercise evaluative control over your attitudes when you answer for yourself positively the
question whether p and therein believe that p, or when you answer for yourself positively the
question whether to φ and therein intend to φ. Evaluative control consists in settling questions for
oneself, which therein form attitudes, and which place a thinker ‘on the hook’ to answer
corresponding ‘Why?’-questions with respect to these very attitudes.
So understood, inferring to a conclusion qualifies as a kind of evaluative control.3 A thinker
infers to a conclusion by recognizing the normative force of the reasons that bear positively on
the deliberative question that he has set himself to answer. This self-conscious recognition of the
premises is what constitutes the conclusion as what it is—an answer to one’s deliberative
question. On the account of inferring I have sketched over Chapters 3–4, it appears that inferring
to a conclusion is a paradigmatic type of exercise of evaluative control.
Hieronymi describes another form of control, managerial control, which is more readily
associated with intentional rational agency. In exercising managerial control, an agent does
things to bring the world into accord with the representations that he has of it. In exercises of
managerial control, we “take actions designed to affect [ordinary objects] according to our
purposes” (2009: 140). I exercise managerial control over the items on my desk, for instance, by
arranging them to accord with my intentions for how they ought to be arranged on the desk’s
surface. Managerial control has two features that evaluative control does not. First, managerial
control is voluntary in the reason-based sense. I can (say) take a sip from the coffee mug on my
desk for any reason that recommends doing so, including instrumental reasons that connect
3 Hieronymi thinks that there are forms of evaluative control that do not consist in inferring to a conclusion (2009:
143, note 5). It is difficult, though, to see what this could amount to, given that evaluative control involves one’s
self-conscious recognition of his take with respect to the question whether p or whether to φ. I argue for this below.
249
coffee sipping to further aims I act in order to realize. Secondly, a kind of “reflective distance”
separates an agent from the objects of his exercise of managerial control (2009: 144). When I act
on an intention to drink from my coffee mug, the mug is the object of my awareness by dint of
figuring in the content of my intention to lift the mug to lips. Since, Hieronymi says, “we form
our intentions by settling for ourselves a question that represents our action under some
description, then it seems that our action is, in some sense, an object of our thought” (2009: 145).
Evaluative control lacks these two features. Evaluative control is nonvoluntary. And when I
settle for myself the question whether p or whether to φ, I do not have the (prospective) belief
that p or intention to φ ‘in mind’ as the object of my thought. When I settle for myself the
question whether p or whether to φ, my awareness is directed toward the (apparent) fact that p or
to the prospect of φ-ing, and not the generation of an attitude of belief or intention with respect to
these items. This contrasts with how the coffee mug before me figures in my awareness as I act
upon it. In light of this contrast, Hieronymi says that evaluative control is without the
“paradigmatic features” that mark intentional bodily actions.
Why, then, count evaluative control as a kind of agency at all? Hieronymi’s argument on this
score appeals to the formation of intention. According to her, “it seems that the forming of an
intention must be an exercise of agency, if anything is” (2009: 150). But forming an intention is
nonvoluntary and not itself the object of reflective awareness. Rather, forming an intention is
done paradigmatically by exercising evaluative control—by settling positively for oneself
whether to act. So, Hieronymi concludes, it turns out agency need not be voluntary and need not
be marked by a characteristic sort of reflective distance.
But why must intention-formation constitute an exercise of agency if anything is?
Presumably because intentions play a role in rendering a great many actions intentional. A great
many of our actions count as intentional because in φ-ing I am doing just what I have the
intention to do. This thought is natural enough, but it cannot be a sure ground for our conviction
250
that the formation of intention is itself an exercise of intentional agency. Hobbes, for instance,
dismisses this very thought. Hobbes holds that active bodily performances are indeed rendered
active in virtue of their relation to prior intentions; but he also holds that these very active-
making intentions are formed passively. Hobbes writes, “I acknowledge this liberty, that I can do
if I will, but to say, I can will if I will, I take to be an absurd speech.”4 Prima facie, Hobbes’
position seems consistent: so far, we have no positive argument for the claim that the formation
of intention must itself be an exercise of rational agency. Call this Hobbes’ Demand: it is the
demand for an explanation why the formation of intention must itself be an exercise of agency.
To rectify this lacuna, we must better understand what is involved in the exercise of
evaluative control. Doing so will show that evaluative control is agentive when it consists in
inferring to a conclusion; inferring to a conclusion, in turn, is a manifestation of intentional
agency (as per the argument over Chapters 3–4). On the other hand, managerial control will be
explicated in the very same terms. The upshot will be that evaluative as well as managerial
control are sourced in the very same sort of inferential agency that is exercised most prominently
in a thinker’s inferring to a conclusion. In this way, the account presented here will preserve the
unity of action not by subsuming inferring to a conclusion to intention-executing bodily action,
but by subsuming intention-executing bodily action to the inferring to a conclusion. (This will
reinforce from another angle a conclusion from Chapter 4.)
To commence the first leg of this argument, let us return to evaluative control. As we have
seen, Hieronymi tells us that evaluative control is exercised in settling positively the question
whether p or the question whether to φ. (In what follows, to save space I will treat with the case
of intending, since the very same considerations apply also to forming the belief that p by
settling positively whether p.) But what is involved in settling positively the question whether to
4 See e.g. Chappell (1999: 16); quoted in Pink (2004: 61). Hobbes identifies action with voluntariness, thus ruling
out the possibility of intentional nonvoluntary action. See Pink (2007) for lucid discussion.
251
φ? In the first case, settling positively the question whether to φ cannot consist merely in the
occurrent thought that one shall φ. One might enjoy an occurrent thought that one shall φ without
ever being committed to φ-ing (as when one merely supposes that one shall φ solely to reason
out the probable consequences supposing one were to φ). But the positive affirmation that that
one shall φ is not sufficient for settling for oneself positively the question whether to φ. This is
because one might affirm that I shall φ in recognition of the commitment to φ-ing while already
being settled on φ-ing. This occurs when one re-affirms one’s commitment to oneself (as I might
do in running through my plans for the day). This is the practical analog of enjoying an occurrent
thought that p when one already enjoys a standing commitment to a positive answer to the
question whether p (and therein believes that p). Settling positively the question embodied in an
attitude cannot simply consist in rehearsing to oneself a positive answer to that question.
Similarly, settling positively the question whether to φ cannot occur if one is already
committed to a positive answer to that question. Settling positively the question whether to φ is
to form the intention to φ. But I cannot form the intention to φ if I already intend to φ. This is
because one cannot possess two tokens of the very same attitude (cf. McCann 1998: 144–145). I
cannot enjoy two intentions to φ: if I already have an intention to φ, then I cannot form this
intention over again. The most I might do is reaffirm an intention that I already enjoy (say, in the
face of anticipated weakness of resolve to follow through with my plan to φ). If settling the
question whether to φ is a way of forming the intention to φ, then, there is no way to settle again
the question whether to φ if I have already positively settled this question.5
5 Of course, we sometimes reopen a question we have already settled. For instance, I might reopen the question
whether to φ even though I already possess a settled long-term intention to φ. It is rational to reopen the question
whether to φ when one encounters novel reasons counting against one’s settled plans. But if one closes such a
reopened inquiry by settling on the very same intention to φ, then I am inclined to say that the intention is abolished
and a new one with the same content as before reformed. If the intention to φ is never abolished, though, then one
never really abandons the intention; rather, one engages in practical reasoning under the supposition that one’s mind
is not really settled when in fact it is. Thanks to Carol Rovane for bringing this to my attention.
252
If settling positively one’s answer to a question is to count at all as a way of controlling
one’s attitudes, then, it must be the case that settling positively an answer to a question works to
generate a novel attitude that was not already present at the time of settling.
How can settling the question whether to φ work to generate a novel intention to φ? And
how can an intention to φ “embody” this positive appraisal of the case for φ-ing? We find a clue
in Hieronymi’s description of what this settling involves. She tells us that settling the question
positively whether to φ therein renders one “vulnerable to certain sorts of criticisms and open to
certain sorts of questions—in particular, one is open to questions and criticisms that would be
satisfied by reasons that (one takes to) bear positively on whether to φ” (2009: 138). Here,
Hieronymi indicates that settling on φ-ing—rather than being already settled on φ-ing—is a
transition that renders one answerable for φ-ing. Being answerable for φ-ing means that one is in
position to offer reasons that support φ-ing, reasons that bear positively on the course of action
one (now) intends to execute. For one to settle positively the question whether to φ, then, one
must do so in the acknowledgement of reasons that recommend φ-ing. At least, this must be so if
we are to preserve the close connection between becoming settled on φ-ing and being in position
to answer the ‘Why?’-question with respect to the intention that is therein formed. Call this the
answerability constraint on settling positively the question whether to φ: whatever else we want
to say about this transition, this transition must position the agent to answer the ‘Why?’-question
whose applicability marks a thinker’s answerability for the intention that results from this very
transition.
Notice, now, that the answerability constraint rules out the prospect of an agent’s settling
positively the question whether to φ in the absence of grasping grounds that count in favor of φ-
ing. In the absence of any such grasp of positive grounds for φ-ing, it is difficult to see just why
settling suffices to place one in position to answer the range of questions that Hieronymi
identifies. Surely part of the interest in Hieronymi’s notion of evaluative control is that settling
253
positively the question whether to φ differs starkly, in its normative profile, from the onset of a
sneeze or the coming-to-mind of random thoughts that pass away just as soon one’s awareness
lights upon them. In contrast, we can readily satisfy the answerability constraint by making it
constitutive of settling positively the question whether to φ that an agent settles this question
positively in appreciating some range of positive grounds for that answer, and answering in
recognition of these very grounds. When I settle positively the question whether to walk to the
café, then, I do so in recognizing the normative force of reasons why I ought to walk there, given
my other aims (such as ingesting some caffeine and stretching my legs). But if the intention
comes over me without such a recognition—if I simply find myself with the intention to φ that
embodies a positive answer to the question whether to φ—it remains mysterious just why I am
rightly subject to the justificatory demands that give force to the answerability constraint. The
best way to meet this constraint, I submit, is to accept the present suggestion: an agent settles
positively the question whether to φ in the recognition of grounds that positively recommend φ-
ing.
Furthermore, it will not suffice for meeting the answerability constraint to say that an agent
settles positively the question whether to φ in recognition of grounds that positively recommend
φ-ing unless the agent’s intention to φ is formed because of that very recognition. If the
recognition of grounds is not the operative explanatory factor in the agent’s coming to settle the
question whether to φ, then we are no better off satisfying the answerability constraint than we
were before we took up the suggestion regarding an agent’s recognition of grounds.
At this point we have simply imposed the Taking Condition on settling positively the
question whether to φ. An agent cannot settle positively the question whether to φ unless (I) in
the recognition of grounds that positively recommend φ-ing and (II) the agent settles positively
the question whether to φ because of this very recognition. And this means that when an agent
settles positively the question whether to φ, the thinker transitions to the commitment to φ-ing
254
because the intention to φ stands as the enduring conclusion to which the agent inferred from
premises that the agent took to support φ-ing. If this is right, then to settle just is to infer to a
conclusion And if to settle is to infer, then it must be that the exercise of evaluative control is to
infer to a conclusion in thought.
I have been arguing that in order for Hieronymi’s notion of evaluative control to do the work
that she enlists for it—in order for evaluative control to be a way of forming novel attitudes for
which one is subsequently answerable—evaluative control must consist in a thinker’s inferring to
a conclusion. Hieronymi explicitly rejects this result (cf. 2009: 143, n. 5). Nevertheless, it has the
advantage of answering Hobbes’ Demand. Recall, Hobbes’ Demand is a request for an
explanation why the formation of an intention must be agentive. Why can’t it be that our
intentions are both causes of bodily movements that render the latter actions and mere passive
effects? The answer to Hobbes’ Demand is that intentions are formed in a thinker’s exercise of
evaluative control, which in turn consists in the thinker’s inferring to a conclusion, which in turn
is an exercise of intentional rational agency (as I have argued over the last two chapters).6
But this still leaves us with Hieronymi’s view that there are two kinds of intentional agency,
which conflicts with our fifth desideratum. I now want to argue that even managerial control
consists essentially in the very same phenomenon of inferring to a conclusion. This is the second
leg of the argument-sketch I described above.
Managerial control, recall, is exercised when we “take actions designed to affect [ordinary
objects] according to our purposes” (2009: 140). Hieronymi construes managerial control as the
more familiar kind of agency, since it exhibits the two paradigmatic features that mark our
intentional doings. These are the features of voluntariness and reflective distance. I want to focus
6 Hieronymi reserves the term ‘action’ for exercises of managerial control. She allows that there are mental actions,
but these are voluntary actions which are undertaken in order to bring one’s own mind into accord with one’s
intentions, and so exercises of managerial control.
255
on this second paradigmatic feature. In what respect does managerial control exhibit reflective
distance? Presumably, it is because managerial control is full-blooded intentional action, wherein
an agent φ’s out of an intention to φ. (Or the agent may φ out of a belief-desire complex that
represents φ-ing as conducive to some (further) end that the agent aims to achieve. Again, to save
space I will leave these belief-desire complexes to one side and speak more simply of intentions
to act.) The reflective distance that distinguishes managerial control can be put down to the fact
that when I take actions designed to affect objects according to my purposes, I act from an
intention; the content of this intention represents the purpose or state of affairs that I act on
objects in order to realize. Such intentions do not influence our behavior ‘behind our backs.’
Rather, it is by dint of acting from an intention that an agent knowingly and deliberately bringing
about some state of affairs in the world. In acting from intention, I act self-consciously from the
intention that represents the very action I endeavor to perform. (In addition, the intention from
which I self-consciously act represents this action under a distinctive practical modality, as
something that ought to be done.) This seems a fitting explanation of why managerial control is
marked by reflective distance. It is because I act self-consciously from an intention so to act that
I attain the reflective distance in virtue of which my φ-ing is an object of my awareness as I φ.
Managerial control, then, consists in the self-conscious execution of intention. But this only
brings us to the question of what it is to self-consciously execute an intention. Put another way:
What is the relationship between intention and intention-executing action? I want to motivate my
own reply to this question by examining cases in which an agent fails to execute an intention.7
Consider an agent who intends to (φ at t in circumstances C) but that at t the agent fails to φ even
7 The view I will develop over the next portion of the argument stands in stark contrast to the standard causalist
theory that is regnant in the theory of action. On this view, the intention to φ causes an occurrent φ-ing in the same
way as any other efficient cause gives rise to its effects. I will not argue at length against this causalist reply here,
since any such argument would amount to a reiteration of arguments that I have canvassed over Chapters 3 –4
against causalist and dispositionalist accounts of inferring.
256
in the belief that C obtains. The agent has not forgotten the intention to φ, has not changed his
mind about whether to φ, and has not, at the moment of action, somehow lost the ability to φ.
Rather, the agent simply fails to φ. Following Samuel Asarnow (2019), call any such case one of
execution failure. Execution failures are commonplace. Such failures are vividly illustrated, for
instance, in the failure to get up and out of bed in the morning as one’s alarm clock drones on
and on (the example is Asarnow’s). The important thing to notice about execution failures is that
they do not show an agent to be broken or defective (cf. Finkelstein 2007). An execution failure
is not a malfunction in a causal mechanism. Execution failures are rational failings. By dint of
his execution failure, an agent is at odds with himself as to how he ought, by his own lights, to
act.
In fact, this is just what Hieronymi’s commitment-constituted conception of the attitudes
helps us to understand. As Hieronymi has it, an intention embodies a positive answer whether to
φ; the failure to execute one’s intention, then, is a failure to do what one takes as what one ought
to do. This the very same sort of rational dissonance at issue in holding two beliefs in tension, or
failing to infer from p to q even while one takes it that p entails q, therein falling short of the
commitments that constitute what it is to believe that p. But as we saw in Chapter 3, falling short
of these commitments does nothing to undermine that one really does intend to φ or believe that
p. This is because the possibility of falling short of a commitment is worked into the very notion
of a normative commitment—and it is this possibility that marks normative commitments off
from efficient causes.8 We can understand execution failures as cases in which an agent falls
short of a normative commitment—in this case, a commitment to discharge a plan of action.
Intentions, no less than beliefs, are commitments—commitments to act that answer to an agent’s
8 It is this deep distinction between normative commitments and efficient causes that leads some to deny the
possibility of execution failures. But I am more confident in the existence of execution failures than I am in the
causalist/dispositionalist analysis of intention.
257
sense of what he ought to do based on the reasons within his possession. So far, this seems very
much in line with what Hieronymi has to say about attitudes as commitment-constituted (cf.
Hieronymi 2006).
This way of understanding execution failures illuminates the relationship between intention
and intention-executing action in successful cases. We have already said that managerial control
is intention-executing action that possesses reflective distance in virtue of the fact that intentions
are executed self-consciously. We see now that intentions are commitments to act. Recalling an
important conclusion from Chapter 3, commitments possess an imperatival dimension:
commitments are directive and not merely descriptive. An intention to act, then, is not merely a
description of some possible state of affairs but in part an imperative or instruction directing an
agent to act so as to bring about that state of affairs.9 In sum, managerial control, since it
possesses reflective distance, involves an agent’s self-consciously acting from an instruction
embedded in his intention.
I have been tracing out the downstream consequences of accepting an irreducibly normative
commitment-theoretic conception of intention. Notice, now, that this conception raises anew the
very same rule-following perplexities that greeted us when we turned, in Chapter 3, to the nature
of inference. For it turns out that an intention, qua irreducibly normative instruction, constitutes a
rule for the agent to follow. To execute an intention is to follow a rule—namely, the rule
directing the agent to realize the state of affairs that his intention represents. Like any inference
rule, intentions lay down an evaluative standard for assessing the agent’s subsequent
performance. (We saw this same feature in Chapter 4 in connection with Anscombe’s shopping
list example.) For instance, my intention to walk to the café sets the standard for correct conduct
9 This imperatival dimension of intention does not mark intention off as distinct from belief. To reiterate a conclusion
from Chapter 3, beliefs also possess such an imperatival dimension, since the belief that p is a commitment to the
truth of p and so an imperative to accept p’s consequences and, in general, to do one’s epistemic duty in holding to p.
258
on my part, so that my later performance will be mistaken should I fail to execute this intention
in action.
We can thus raise the very same considerations with respect to intention-executing action
that originally motivated the Taking Condition on inference. Without running through the
argument of the first part of Chapter 3 all over again, we summarize its main points: An agent
who self-consciously executes an intention to act must both (I) take it that his present action is
correct relative to the rule that is his intention and (II) perform his present action because of this
very appreciation. It turns out that the relationship between intention and intention-executing
action is, at bottom, an inferential one. That is why an agent is irrational, and not merely
defective, when he undergoes an execution failure. The failure to execute one’s intention in
action is akin to other rational failings, such as when a thinker falls short of his attitudinal
commitments to infer the conclusion mandated by the premises in his possession.
The upshot is that managerial control is sourced in the very same species of rational agency
we have been studying all along. Rational agency—both over our own attitudes and over our
own bodies and the objects in the surrounding environment—is constituted by a practical self-
consciousness. This self-consciousness is practical and agentive in that it works to constitute the
relevant performances—either transitions in thinking, or movements of the body, or even
changes in the extra-bodily environment—as what they are, namely intentional actions. So, there
are not two kinds of agency. Or at least, there are not two kinds of agency, one which
corresponds to our agency over our attitudes and one which is exercised over our bodies. In both
cases, the relevant responsibility-conferring performances are made what they are by an agent’s
self-conscious appreciation of the normative commitments within his repertoire. In the case of
attitude-formative mental agency, that awareness takes the form of an appreciation of what
follows from what; in the case of agency over one’s own body, that appreciation is given by a
sense of what I am doing and why. Put this way, it rather appears that intention-executing bodily
259
agency is but an extension of the inferential agency that we exercise in the formation of our
attitudes of belief and intention.
True, intention-executing bodily action—Hieronymi’s ‘managerial control’—is marked by
two features that do not characterize our agency over the attitudes. Inferring to a conclusion in
thought to form a new belief or intention is neither voluntary nor possessed of the reflective
distance that renders intention-executing bodily action so familiar to us. What Hieronymi terms
evaluative and managerial control, though, are simply two species of the very same genus. In
fact, it is the inferential agency that we exercise over the attitudes which is primary, and the
agency we exercise over our bodies simply an extension of our agency over belief and intention.
Subsuming both evaluative control and managerial control to the sort of rule-following self-
governance I have described over Chapters 3–4 unifies our notion of intentional agency. There
are not two kinds of agency, but two species of the same kind.
Desideratum 1: Responsibility-grounding
The title of this chapter indicates that we attain moral responsibility for our attitudes of belief
and intention by dint of reasoning. (The title harkens to a recent work of John Broome’s,
Rationality Through Reasoning [2013]). I want to begin by recalling to the reader what I mean by
the term ‘reasoning’ and how inferring figures in reasoning. Reasoning is a conscious mental
activity that aims at the resolution of a deliberative question (such as ‘Whether p’ or ‘Whether to
φ’) in such a way as to mark overall rational unity among a thinker’s intentional episodes. This
aim distinguishes genuine reasoning (practical as well as epistemic) from idle mind-wandering,
brainstorming, or randomizing plays of association. Reasoning is thus oriented towards
conclusion in an inferring. So, while there may be some stretches of reasoning that do not
terminate successfully in the resolution of a thinker’s deliberative question, reasoning cannot be
260
what it is if it does not aim towards inferring. On the other hand, inferring to a conclusion does
suffice for a stretch of conscious mental activity to count as reasoning.
I caution the reader against interpreting these provisional remarks in accord with a default
causalist theory of motivation and action. I do not claim that reasoning is an activity by dint of
the fact that it progresses in accord with and as the efficient causal effect of some intention on the
part of the agent to resolve the deliberative question at issue. My present point is a constitutive
one. It is about what makes some stretch of mental activity the kind of activity that it is.
Reasoning that is not oriented toward inferring cannot be reasoning at all.
So, to say that we attain responsibility through reasoning is to claim that we attain
responsibility for those attitudes—and only those attitudes—which emerge from reasoning. In
turn, this means we are responsible for all and only those attitudes that correspond to a
conclusion in thought.
INFERRING GROUNDS RESPONSIBILITY
The first task is to explain how inferring to a conclusion grounds answerability for that
conclusion. Here, Boghossian’s (2014: 4) Taking Condition bears much of the explanatory load.
Taking Condition: Inferring necessarily involves (I) the thinker taking his premises to
support his conclusion and (II) drawing his conclusion because of that fact.
Over Chapters 3–4, I presented an account of inferring that respects this condition. On this
view, the premises from which one infers figure, from the engaged stance of inquiry, as
normative commitments on one’s part. This is what the ‘taking’ that Boghossian alludes to in
sub-condition (I) amounts to. (Contra to many, this does not require positing ‘separate’ taking
beliefs about what follows from what; this is how the account proffered in Chapter 3 avoids the
Carroll-style regress argument that has preoccupied many recent writers on this topic.) And I
261
argued that sub-condition (II) calls for a formal causal reading. A thinker’s self-conscious grasp
of his attitudinal commitments is the formal principle that renders the transition in thought an
inferring to a conclusion (rather than a non-inferential transition in thought).
Reasoning, I said earlier, begins when a thinker addresses himself to a live deliberative
question, such as Whether p or Whether to φ. When reasoning terminates in a thinker’s inferring
to a conclusion, the thinker answers one of these (or a related) deliberative questions.10
Terminating reasoning by inferring to a conclusion positions a thinker to reply in the right way to
the ‘Why?’-question that marks our answerability for the attitudes (cf. Hieronymi 2008). This is
due to the Taking Condition. To bring this out, consider a thinker who infers from p to q. On the
account of inferring I have offered, the thinker’s inferring from p to q entails that he regards his
premise, p, as an attitudinal commitment with a normative rather than merely descriptive profile.
From the engaged stance, to (occurrently) believe that p is to take oneself to be committed to
accepting p as well as p’s implications. In believing that p, I also come under the rational
obligation to accept that which follows from p. But I infer from p to q only if I self-consciously
appreciate that accepting q will constitute my discharging that very attitudinal commitment to p’s
consequences. This self-conscious appreciation of the relation between p and q is the formal
principle that unifies two intentional episodes, one representing p and the other q. It is this
unification that allows us to speak of a premise-conclusion relation in the stream of thought at
all. Therein, I adopt the attitudinal commitment to the truth of q and to its implications.
Suppose now that I am subjected to the justificatory demand of the ‘Why?’-question. Having
inferred q from p, I am now positioned to reply in way that grants the question application: “Why
do you believe that q?” “Because p.” Here, the ‘Why?’-question asks for the premise from which
10 Sometimes, reasoning concludes not in an inferring that answers the deliberative question, but in a meta -level
conclusion about the paucity of considerations currently available to one. This is an inferring that terminates
reasoning in a responsibility-grounding way. But for sake of clarity I will leave it to one side.
262
I inferred that q; my reply merely recites this very premise. Because the relevant transition in
thought constitutively involves a self-conscious understanding of the premise-conclusion relation
between p and q, I am legitimately subject to justificatory demands with respect to conclusions
that terminate my reasoning. Reasoning guarantees responsibility.
Of course, it might be the case that p provides no support at all for q. In that case, my
reasoning is bad reasoning. But this does not render the justificatory demand inapplicable. Quite
to the contrary, it means that my response to the ‘Why?’-question is a poor one, which does
nothing to justify the conclusion I have purported to defend. Rather than rendering the ‘Why?’-
question inapplicable, then, a poor reply opens a thinker to attitudinal blame and the expectations
and practices that are of a piece with blame. This might encompass a range of blame-responses,
such as communicating to me a better case for p, or a counter argument that shows ¬p. It is no
part of the present account that an agent is morally responsible for his attitudes only if he can
answer the ‘Why?’-question in such a way as to provide a rationally compelling case for the
attitude in question. This consequence would simply rule out the appropriateness of attitudinal
blame at all. It is not good reasoning which makes an agent answerable for his attitudes,but
reasoning good or bad.
Why is this the case? Why is an agent responsible for good and bad reasoning alike? It is
because in inferring, an agent thereby demonstrates a capacity to recognize and respond to
reasons. An agent cannot infer to the conclusion that p without recognizing at least apparent
evidence for p; nor can he infer to the conclusion that is an intention to φ without recognizing at
least an apparent practical case for φ-ing. In responding to (apparent) reasons in the right way
the agent also demonstrates an ability to abide the constitutive norms that govern beliefs,
intentions, and complexes made up out of these attitude-types. If an agent can appreciate these
constitutive norms and act in accord with them, he is thereby ‘on the hook’—rationally liable—
to respect them in epistemic and practical dealings. If an agent reasons badly, then, the agent can
263
be criticized according to rules that the agent recognizes as carrying prescriptive force. (Such
criticism might take the form of simple correction, or more forceful kinds of attitudinal blame.)
This explains why inferring to a conclusion satisfies the Demand Condition.
Inferring to a conclusion also satisfies the Activity Condition. As I have stressed, inferring to
a conclusion is a nonvoluntary type of occurrent intentional mental action. So, to infer to a
conclusion is to do something, therein placing an agent at fault when inferring incorrectly. And
since inferring is an action, there is indeed something the agent can do to avoid incurring
attitudinal blame. The agent can voluntarily commence a bout of reasoning that will conclude in
his reaching a better conclusion, one that better accords with the facts as he can discern them. So,
we can make good on Watson’s point that holding a person accountable for something can only
be fair when the person had the opportunity to avoid incurring the penalties involved in being
accountability for the trespass at issue.
RESPONSIBILITY WITHOUT INFERRING?
Being answerable for a belief that p or intention to φ also presupposes that the belief that p
or intention to φ is an attitude that corresponds to a conclusion to which the agent inferred. If an
agent has not inferred to the belief that p or intention to φ, then the agent has not arrived at the
belief or intention in a way that involves an appreciation of reasons that recommend the belief or
intention in question. Such an agent will not be positioned to give the ‘Why?’-question
application by citing a reason, or at least an apparently supportive consideration, for p or for φ-
ing. In turn, where the ‘Why?’-question is refused application, the agent is not answerable at all
for the belief or intention.
One might object to this line of thought by pointing to other ways in which an agent might
be able to answer the ‘Why?’-question without having inferred to a conclusion. Perhaps one
enjoys a more general basing-relation among the beliefs p and q, so that one believes q because
264
p and can answer p in reply to the ‘Why?’-question. But, the objection continues, no conscious
occurrence in the thinker’s stream of thought needs to occur in order for this basing relation to
hold among the thinker’s beliefs that p and that q.11 This line may appear to grant everything
needed to explain an agent’s answerability for an attitude, without even attempting to satisfy the
Activity Condition that we have followed Angela Smith (2008) in placing on moral
responsibility.
In reply, we must question what is actually involved in the proposal. The lynchpin of the
suggestion is that a thinker’s belief that q might be based on his belief that p, without his ever
having inferred q from p. Suppose this agent is the target of the ‘Why?’-question: he is asked to
justify his belief that q in such a way that failure to do so adequately will open him to forceful
rational criticism. (After all, if this condition is not met then he is simply not responsible for his
belief in the relevant sense.) There are three relevant possibilities. At the time when the ‘Why?’-
question is communicated, either
a. the agent has considered the case in favor of q, appreciated that p sufficiently supports q,
but has so far failed to infer to q; or
b. the agent has considered the case in favor of q and concluded that q out of an
appreciation of his commitment to p; or
c. he believes both p and q but is, at the time of the question, ignorant as to the evidential
relation that p bears to q.
The agent in case (a) is clearly irrational and responsibly so. It is legitimate to hold the agent
accountable for this irrationality by targeting him with forceful rational criticism. An agent who
understands that p implies q, who fails to undertake the commitment to the truth of q, and who is
impervious to criticism on this score cannot be said really to have the commitment to the truth of
11 One view in the literature that appears to make much of such static basing relations is that of Angela Smith
(2005), which I discussed back in Chapter 1.
265
p at all (cf. Bilgrami 2006: 241–242). Here the agent is answerable, if not for the belief that q
then for falling short of other attitudinal commitments within his repertoire.
Moving to case (b): if the agent has acknowledged prior to the ‘Why?’-question the rational
relation between p and q, and answers the ‘Why?’-question by citing p as a reason to believe that
q, then the agent has already inferred to q from p. Recall: the thinker’s taking it that p establishes
q is the formal principle that constitutes a premise-conclusion relation between p and q,
respectively. If the agent acknowledges this relation, then he already has inferred that q; therein,
the agent is answerable for q along the lines I have just explained.
This brings us to case (c). We may envision the agent in this case giving one of two replies to
the question, “Why do you believe that p?” First, the agent might reply, “No reason.” What we
say about this case depends upon how we interpret this reply. Heard in one register, ‘No reason’
indicates that the agent does not at all appreciate that his beliefs are subject to constitutive norms,
such as the norm that one ought not to believe that p in the absence of sufficient evidence. If the
agent’s reply of ‘No reason’ indicates that he does not accept such a constitutive norm, it is hard
to fathom that the agent is the right sort of target at all. Such an agent cannot meet the Demand
Condition on moral responsibility.
Things are different with intentional action. Recall: Anscombe tells us that the ‘Why?’-
question is granted application even when an agent replies ‘No reason,’ as the question ‘How
much change do you have in your pocket?’ is granted, rather than refused, application when the
reply is ‘None.’ But this is not a difference in the kind of responsibility at issue—we are
answerable for our beliefs, our intentions, and our actions in the very same way. Rather, the
difference is traceable to the constitutive rational norms that govern beliefs and actions,
respectively. With the exception of special classes that do not admit of further justification—such
as perceptually-based beliefs or foundational beliefs at the ‘bedrock’ of one’s repertoire—it is
epistemically impermissible to believe that p for ‘No reason.’ The norms that constitutively
266
govern action are less demanding: one is justified in φ-ing so long as it is not the case that one
ought not to φ (cf. Shah 2008). ‘No reason’ grants application to the questions ‘Why are you φ-
ing?’ and ‘Why do you intend to φ?’ since an agent needn’t possess a convincing case for φ-ing
to be justified in φ-ing; his warrant for φ-ing lapses only in the recognition of a compelling case
why he ought not to φ.
There are other ways to interpret an agent’s reply of ‘No reason’ to the question ‘Why do
you believe that p?’ If his persistence in believing p is put down to compulsion or other brute
psychological forces, then once more the agent fails to come under the Demand Condition. Our
responsibility-practices accommodate such cases by exempting or excusing the agent from the
burden to provision a justification for p.
Alternatively, ‘No reason’ might communicate a kind of epistemic confession (as it were).
Heard in this register, ‘No reason’ communicates that the agent recognizes that one ought not to
believe that p in the absence of sufficient evidence but confesses that, in his belief that p, he falls
short of that very norm. In this, the agent knowledgeably falls short of his own attitudinal
commitments. I do not think we can really make sense of this scenario unless we also suppose
that the agent experiences significant rational tension from within his own point of view, a
tension spurring him to achieving rational unity in the direction of fidelity to the facts (as far as
they can be discerned). For if the agent does not come under such a rational tension, then it is
questionable whether he really does take himself to fall short of the norm enjoining belief only in
the light of sufficient evidence. Such an agent is clearly answerable for his belief that p. Such an
agent is rationally liable to provision a justification for p, and in his failure to do so opens
himself to attitudinal blame.
Are such cases counterexamples to the view that we attain responsibility for the attitudes by
reasoning? No. To see why, consider analogous cases with respect to overt intentional action.
Suppose I have been designated to bring wine to the party tonight. I understand this commitment
267
and I am concerned to discharge it by performing the relevant actions. But suppose further that I
fall short of these commitments by failing to come to the party with the wine. Just like the agent
in our example about belief, I have simply failed to discharge my commitments in action and, in
a confessional register, own up to that fact when I arrive to the party wineless. Such examples do
nothing to count against the idea that my responsibility for bringing the wine is grounded in
voluntary wine-acquisitive bodily actions. We only have to say that in the case in question I have
utterly omitted to employ this agency properly towards doing what I know I ought to have done.
In this example, my omission to discharge my commitment is blameworthy: not because of what
I have done, but because of what I have failed to do.
We may understand believers and intenders in strictly analogous terms—as agents who have
failed to discharge standing commitments by engaging their agency in the right ways. Such
omissions are failures to seek grounds for one’s attitudinal commitments to p or to φ-ing. They
are failures to subject one’s attitudinal commitments to the kind of inquiry that terminates in an
inferring that corresponds to the ungrounded commitment. We bear responsibility for these
omissions for the very same reasons we bear responsibility for omissions to perform other duties.
Understood in this way of reading case (b), attitudinal blame remains legitimate, since the agent
in our example falls under rational norms and attitudinal commitments, understands these norms
and commitments, and is able to discharge them in action. (This line regarding omissions may
also bear on what we say about an earlier case, case (a).)
There is a final way of reading case (c), wherein an agent believes both p and q but is
ignorant of the evidential relation that p bears to q. Upon being asked the ‘Why?’-question, the
agent may be prompted to reason (however briefly) and infer that q from p. Having inferred q
from p, the agent is properly responsible for q. In such a case, the occasion of the ‘Why?’-
question prompts the agent to engage in a bout of reasoning (however brief). Here, the ‘Why?’-
question works to bring the agent to question the grounds of some belief that he already holds,
268
and in performing this reasoning and terminating it with a judging-that-q out of recognition of
the normative force of his commitment to p, he answers the ‘Why?’-question satisfactorily in a
way that licenses attitudinal praise. In this way, asking the ‘Why?’-question can occasion the
agent to discharge a commitment that he has, before being asked, omitted to discharge properly.
To this it will be asked what we are to say about the agent’s responsibility for the belief that
q at the moment the ‘Why?’-question is asked but before the agent has reasoned to the
conclusion that q from the premise that p. At this moment, the agent is responsible in the sense
indicated a paragraph back, in the sense that he attitudinally blameworthy on account of an
omission. But something else can be said: the ‘Why?’-question prompts the agent to take
responsibility for the belief that q by reasoning to it from grounds that constitute evidence for q.
In this way, posing the ‘Why?’-question to another—or even to oneself—can function in a
forward-looking manner, as a way of leading one to express one’s rational agency in assuming
ownership over an attitude. I take it that the same forward-looking dynamic operates in the
domain of overt intentional action, as when I am asked pointedly just what I plan to do with my
day. I may have no plans at all when the question is asked; but the question might be
communicated in such a way as to communicate that I really ought to have something productive
planned. (This occurs frequently in the domestic life of a dissertation writer.)
Upon consideration of these cases (a)–(c), we see that reasoning—or the agent’s illicit
omission from engaging in reasoning—undergirds responsibility for the attitudes. If the agent
cannot engage in reasoning, then the most we can do is to merely evaluate the overall network of
his attitudes as rationally suboptimal in holding to p and q but failing to grasp that p is grounds
for q. But the agent’s ability to reason from one attitudinal commitment to the generation of
novel commitments licenses deeper, forceful forms of appraisal: attitudinal praise and blame. For
it is only by engaging in reasoning that the agent can exercise his agency in discharging the
rational commitments constitutive of being a believer or intender at all. In this light, ‘mere
269
basing’ in a network of evaluative commitments is not sufficient to legitimate attitudinal blame.
The agent must recognize these relations in such a way that, in this very recognition, the agent
exercises intentional agency in the formation of novel attitudes—or at least have the standing
ability to do so.
Conclusion
I began this dissertation by arguing that we bear responsibility for the attitudes of belief and
intention. But this raised a question about the legitimacy of attitudinal blame. Given that we do
not hold our attitudes voluntarily, what makes it legitimate to hold one accountable to the rational
norms constitutive of belief and intention?
In effect, the argument of this dissertation is that this question carries only an apparent force.
Attitudinal blame is legitimated by a species of intentional action that agents exercise over their
own attitudes. This very sort of intentional action is also fundamental to the agency we exercise
over our bodies in executing long-term intentions. Responsibility for the attitudes is grounded in
a thinker’s inferring to a conclusion. Inferring is a fully intentional yet nonvoluntary exercise of
rational agency. We are responsible for inferring, ultimately, in virtue of its normative grounding
in our commitments and values—and by the way that, in inferring, an agent attains self-
knowledge of discharging these commitments and values in the attitudes and deeds that they
require. By working out our responsibility with regard to inferential transitions in the mind, we
come to understand a prior dimension of our responsibility for our intentional bodily actions.
270
Appendix: Intention, Constitutive Norms, and the Toxin Puzzle
I have been arguing that Kavka’s puzzle lends support for the claim that beliefs and
intentions are involuntary. And I have been pressing this claim by putting forward a specific
interpretation of Kavka’s puzzle—an explanation for why a rational agent cannot win the
eccentric billionaire’s prize. With others, I have claimed that this explanation comes down to a
close relationship between believing that p and reasons that bear on p’s truth, and between
intending to φ and reasons that bear on the goodness, value, or desirability of φ-ing. This close
link, I claim, frustrates a rational agent’s attempt to win the prize. The reason why you cannot
win the prize—and the reason why the attitudes are involuntary—is because you cannot come to
believe that Dukakis won the presidency in 1988 absent evidence that this is so, and nor can you
intend now to (drink a toxin two days from now) that you will not have reason, then, to drink.
Alfred Mele (1992b) disputes this putative explanation of the involuntariness of the
attitudes.1 Mele argues that it is possible to win the eccentric billionaire’s prize money, and that
this possibility counts against the purported close link between reasons and intentions. To show
this, Mele asks us to imagine an agent named Ted. An evil genius has conditioned Ted to drink
any nearby toxin. Ted’s conditioning runs so deep that he is sure to drink any nearby toxin
whether he intends to do so or not. That is, the evil genius has not conditioned Ted to like
drinking toxin. Ted has not been conditioned to see something good or valuable in drinking
toxin. Rather, Ted is reliably conditioned to drink toxin irrespective of his own will. What is
more, Ted knows this about himself. He knows that he will drink any nearby toxin whether he
wants to drink it or not, and whether he intends to drink it or not. For Ted, the prediction ‘I am
sure to drink any nearby toxin’ is as good as certain.
1 I thank Professor Mele for bringing this article to my attention in private correspondence.
271
Now imagine that the eccentric billionaire offers Ted the chance to win the prize. To win $1
million, Ted only needs to intend now (Monday) to drink the toxin on Wednesday. Ted
deliberates roughly as follows (cf. Mele 1992b: 177–179). Ted knows that he will drink the toxin
on Wednesday. The only thing that is up to him is whether he drinks it intentionally. But on this
question Ted is indifferent. He has no preference at all that might lead him to choose to drink the
toxin intentionally over drinking it unintentionally, and vice versa. So, while Ted recognizes that
he has no reason to drink the toxin intentionally, he also believes that he has no reason not to
drink the toxin intentionally (that is, no reason to choose what will ultimately amount to his
drinking it unintentionally). This means that Ted is in a ‘Buridan’s ass’-type situation in respect
of his prospective intentional toxin-drinking. He must choose between two incompatible
alternative courses of action that are equally worthy of choice. He must choose, but he cannot
choose for a reason that favors one course of action over the other.
This makes Ted quite different from you or me. You haven’t been conditioned to drink any
nearby toxin. So, you are not confident that you will drink the toxin unintentionally even if you
decide not to drink it. Unlike the case with Ted, whether you drink the toxin is up to you. So, you
are not in a Buridan-type scenario. Your choice is not between two equally preferred alternatives.
To win the prize, you must decide to do something that, after your deciding, you will both have
no reason to do and have the power not to do. That is what makes the toxin puzzle puzzling for
you and for me.
These peculiarities position Ted to reap the reward. Ted is in a Buridan-type scenario. In
such scenarios we can settle on an alternative even in the absence of grounds that favor that one
over another. Indeed, we can typically settle on one such alternative in the light of incentives that
attach to making up one’s mind, rather than those pertaining just to the course of action we
decide upon. Suppose you are indifferent about whether to work this afternoon in the library or
the café. However, your friend would like to meet you wherever you choose to go and is also
272
indifferent as to whether to meet at the library or the café. Though there is no reason to settle on
working at the library over working at the café (and vice versa), there is a reason to settle one
way or the other. The reason to settle on one alternative over the other is to realize another goal
of yours: having a settled mind will allow you to coordinate with your friend about where to
meet later today. Ted’s situation is similar in relevant respects. While he has no reason to decide
to drink the toxin intentionally and no reason to drink the toxin unintentionally, he does have a
reason to decide to drink the toxin intentionally, since making this decision will net him the
reward.
But if this is right, then Ted wins the reward—even in the recognition that he has no reason
to drink the toxin intentionally, no reason to decide to drink the toxin intentionally, and no reason
to persist in the intention to drink the toxin intentionally after deciding to do so. This appears to
count against the explanation that I have offered in terms of a close connection between reasons
for intending to φ and reasons for φ-ing. Mele explains the toxin puzzle not in terms of any such
link between reasons and intentions, but between intentions and associated beliefs about whether
one will act as intended. There is controversy in the literature about the strength of this condition.
Mele operates with a relatively weak version according to which “an agent who intends to φ does
not believe that he (probably) will not φ” (1992b: 177).2 This condition only rules out my
intending to φ in the belief that I am more likely not to φ than I am to φ. Crucially, Mele
continues, this condition “leaves it open whether the agent believes that he will not intentionally
φ” (1992b: 177). Of course, you and I typically assume that if we intend to φ then we will φ
intentionally, because we typically deliberate only about courses of action that will come about
only if we bring them about intentionally. But Ted’s situation is different. He knows that he will
drink the toxin whether or not he does so intentionally. Ted is like us in deliberating only
2 I substitute ‘φ’ for Mele’s ‘A” in quotations.
273
between the options that are genuinely open to him, but he is unlike us in that it is not open to
him whether he drinks the toxin. Since this is so, the belief condition on intending that prevents
us from winning the reward does not apply to Ted. I cannot win the reward because I have no
reason to drink the toxin. When I have no reason to drink the toxin, I believe I will not attempt to
drink it. Since Ted knows he will drink the toxin whether or not he attempts to do so—and
because he registers no reason that favors intentionally drinking it over drinking it
unintentionally—the belief condition that prevents you and me from winning the reward does not
prevent Ted from doing so. Mele concludes:
Thus, even if our convictions about what we will not even try to do are typically products
of our convictions about our reasons, we need not suppose that intending to φ is
conceptually more tightly tied to having reasons for φ-ing than to the absence of a
conviction that one will not even attempt to φ. (1992b: 184–185)
The upshot is that we can account for the inability to win the toxin by way of the relationship
between intentions and (the absence of) beliefs about what therein intends to do—not by positing
any sort of deep connection between reasons to φ and intentions to φ.
Another one of Mele’s aims is to challenge the following traditional action-theoretic thesis:
PA. An agent intentionally φ’s only if he φ’s for a reason.3
Ted’s case also casts (PA) into doubt, since Ted decides to drink the toxin intentionally
without a reason for doing so (over drinking the toxin unintentionally), therein intends to
3 Mele (1992: 193, n. 2) lists the following proponents of (PA): Goldman (1970: 76); Davidson (2011: 264), and
Anscombe (1957: 9). However, including Anscombe as a proponent of (PA) might be too quick. Anscombe holds
that intentional actions are events for which a certain sense of the question ‘Why?’ finds application; later in
Intention, Anscombe grants that the answer “No reason” grants the ‘Why?’-question application (2011: 25). See
Chapters 1 and 4 of this dissertation for additional discussion.
274
intentionally drink the toxin without a reason, and, eventually, even drinks the toxin intentionally
for no reason.
If Mele is right, then we cannot appeal to any such connection between reasons, intentions,
and intentional actions in our explanation of the involuntariness of the attitudes. It is the
connection between intentions to act and beliefs about one’s own future actions which explains
why we cannot reap the eccentric billionaire’s rewards.
I grant to Mele that (PA) is too strong. Not every intentional action is done for a reason for
that action. But this need not count against the traditional view linking reasons with intentions.
We only need to heed better just what this connection does and does not rule out.
To sort these matters, recall that Ted can only win the reward because Ted (unlike us) faces a
Buridan-type situation when deliberating about whether to drink the toxin intentionally. Mele is
correct to say that we can settle these scenarios by ‘plunking’ for one alternative over another
and do so in the direct of reaping certain extrinsic incentives attaching to having one’s mind
made up. But it was precisely in recognition of these scenarios that I follow Shah (2008) in
placing the following constitutive norm (CN) on deciding and intending:
CN: An agent’s belief or intention to φ is correct if and only if it is not the case that the agent
ought not to φ.4
(CN) places a standard of correctness on deciding-to-φ and intending to φ. A deciding-to-φ
or intention to φ is incorrect qua deciding or intention if it fails to abide this standard. I will be
irrational should I persist in an intention to do something that I know I ought not do.
Notice that there is a much stronger thesis in the vicinity. This would be the claim that the
constitutive norm regulating deciding and intending is as follows:
4 Given that Shah also takes (CN) as the constitutive aim of intending, I take it his reply to Mele’s case of Ted would
run along similar lines. But since he does not discuss Mele’s argument, it is worth elaborating the reply.
275
R: An agent’s belief or intention to φ is correct if and only if the agent ought to φ.
(R) is stronger than (CN) because the deliberative ‘ought’ is fixed by the preponderance of
reasons within an agent’s possession. If (R) were the constitutive norm regulating our decisions
and intentions, then (PA) really would be a condition on intentional action (at least, on rational
action). (CN) is weaker than this, though. (CN) allows agents to decide between two
incompatible courses of action that are equally worthy of choice. Indeed, (CN) even allows an
agent to decide-to-φ for no reason. All that (CN) rules out is the prospect of an agent’s deciding-
to-φ in the knowledge that the preponderance of reasons counts against φ-ing.
Notice, also, that (CN) will prevent most of us from winning the prize. When you consider
the eccentric billionaire’s offer, you understand (however tacitly) that (CN): any intention to φ
that emerges from deliberation will be correct if and only if it is not the case that you ought not
to φ. But you also know that you ought not to drink the toxin since the preponderance of reasons
counts against doing so. You cannot form an intention that you know will be defective qua
intention. (CN) guarantees the following Preservation Principle:
Preservation Principle: If an agent S is warranted at t1 to decide to (φ at t2) then S is
warranted to φ at t2 out of this very decision, provided there are no relevant changes in S’s
deliberative circumstances in the period between t1–t2.5
This Preservation Principle insures that no agent who is warranted to decide to φ will end up
being less warranted in φ-ing out of that very decision at a later time (so long as relevant
deliberative circumstances remain fixed). The Preservation Principle is itself guaranteed by the
fact that deliberation, deciding, and intending are all constitutively regulated by the standard of
5 I defend the Preservation Principle in the course of my treatment of diachronic autonomy in Heeney (2020: §4).
My thinking on these matters owes much to Pink’s (1996: 153) Reason-Apply principle and Bratman’s (1987: 55)
‘intention-action’ principle.
276
correctness articulated in (CN). This is because (CN) ties the correctness of deciding-to-φ and
intending to φ to the absence of preponderant reasons against φ-ing.
However, (CN) does permit ‘brute selection’ in a Buridan-type situation. Faced with two
incompatible alternatives that are equally worth pursuing, (CN) allows that an agent can correctly
decide on one alternative over the other. Such an agent needn’t do so for some specific reason
that favors his chosen alternative. The agent may elect to choose either alternative so long as
neither one runs against the preponderance of reasons within his possession.
This means that Ted’s prizewinning decision to intentionally drink the toxin satisfies (CN).
Since there is no reason not to drink the toxin intentionally and no reason not to drink the toxin
unintentionally, Ted can correctly decide to drink the toxin intentionally or correctly decide to
drink it unintentionally. We might wonder whether Ted has a reason not to drink the toxin. But
remember that this is not something that is up to Ted at all. Because of his peculiar situation, he
is tasked with deciding whether to drink it intentionally—not whether to drink it. It is Ted’s
deciding to drink the toxin intentionally that accords with (CN). So, too, does Ted’s enduring
intention to intentionally drink the toxin. Even though Ted has won the reward upon deciding to
drink the toxin intentionally, it remains the case after his deciding to act that he has just as much
reason to drink the toxin intentionally as he has to abandon his intention post-decision. This is the
case because, recall, Ted registers no preference at all between intentional and unintentional
toxin drinking. (See Mele 1992b: 185–189 for supporting argument.) The sequence taking Ted
from deliberating to deciding and eventually acting, because it is constitutively regulated by the
norm (CN), also accords with the Preservation Principle.
However, as I have already intimated, (CN) and the Preservation Principle cannot be
respected in a toxin scenario that features agents like you or me—agents who have not been
conditioned to drink any nearby toxin whether intentionally or not. Should you or I decide to
277
drink the toxin, we will be left unwarranted to drink the toxin after having decided. This
circumstance violates (CN) and the Preservation Principle.
The upshot is that we can preserve a version of our original view, the one that links
intentions to reasons perform the intended action. Of course, in the light of Mele’s case we
cannot hold (PA). Nor can we hold that an agent intends to φ only if the agent has preponderant
reason to φ. But we can repair to weaker versions of these claims by recognizing (CN) as the
constitutive standard of correctness that normatively regulates decision and intention. This
constitutive standard legislates a permission to intend actions that one is not rationally barred
from performing. This weaker claim can explain why you and I cannot win the reward but Ted
can do so. This is because Ted faces a Buridan-type scenario that licenses his intending to drink
the toxin, whereas you and I do not.
278
Bibliography
Allison, Henry, Kant’s Theory of Freedom (Cambridge: Cambridge University Press, 1990).
Alvarez, Maria, Kinds of Reasons: An Essay in the Philosophy of Action (Oxford: Oxford
University Press, 2010).
Anscombe, G.E.M. Intention (Cambridge: Harvard University Press, 1957).
Asarnow, Samuel, “On Not Getting Out of Bed,” Philosophical Studies 176, no. 6 (2019): 1639–
1666.
Bilgrami, Akeel. Self-Knowledge and Resentment. (Cambridge: Harvard University Press, 2006).
Boghossian, Paul, “What is Inference?” Philosophical Studies 169, no. 1 (2014): 1–18.
Boghossian, Paul. “Epistemic Rules,” Journal of Philosophy 105, no. 9 (2008): 472–500.
Boyle, Matthew, “‘Making up Your Mind’ and the Activity of Reason,” Philosopher’s Imprint 11,
no. 17 (2011): 1–24.
Boyle, Matthew, “Active Belief,” Canadian Journal of Philosophy 39 (2009): 119–147.
Bratman, Michael, “Intention, Practical Rationality, and Self-Governance,” Ethics 119, no. 3
(2009a): 411–443.
Bratman, Michael, “Intention, Belief, Practical, Theoretical,” in S. Roberson (ed.), Spheres of
Reason: New Essays in the Philosophy of Normativity (Oxford: Oxford University Press,
2009b): pp. 29–61.
Bratman, Michael, Intention, Plans, and Practical Reason (Cambridge: Harvard University
Press, 1987).
Bratman, Michael, “Two Faces of Intention,” Philosophical Review 93, no. 3 (1984): 375–405.
Broome, John. “Comments on Boghossian,” Philosophical Studies 169, no. 1 (2014a): 19–25.
279
Broome, John. “Normativity in Reasoning,” Pacific Philosophical Quarterly 95, no. 4 (2014b):
622–633.
Broome, John. Rationality Through Reasoning (London: Wiley-Blackwell, 2013).
Burge, Tyler. Cognition Through Understanding (Oxford: Oxford University Press, 2013).
Burge, Tyler. (2013a) “Introduction,” in Burge (2013): pp. 1–52.
Burge, Tyler. (2013b) “Our Entitlement to Self-Knowledge,” in Burge (2013): pp. 68–87.
Burge, Tyler. (2013c) “Reason and the First Person,” in Burge (2013): pp. 383–406.
Buss, Sarah, “Autonomous Action: Self-Determination in a Passive Mode,” Ethics 122, no. 4
(2012): 647–691.
Campbell, Lucy, “Two Notions of Intentional Action? Solving a Puzzle in Anscombe’s
Intention,” British Journal for the History of Philosophy 26, no. 3 (2018), 578–602.
Carroll, Lewis, “What the Tortoise Said to Achilles,” Mind 4, no. 14 (1895): 278–280.
Chappell, Vere (ed.), Hobbes and Bramhall on Liberty and Necessity (Cambridge: Cambridge
University Press, 1999).
Clarke, Randolph, Libertarian Accounts of Free Will (Oxford: Oxford University Press, 2003).
Coope, Ursula, “Aquinas on Judgment and the Active Power of Reason,” Philosopher’s Imprint
13, no. 20 (2013): 1–19.
Crane, Tim, Aspects of Psychologism (Cambridge: Harvard University Press, 2013).
Dancy, Jonathan, Practical Reality (Oxford: Oxford University Press, 2000).
Danto, Arthur, “Causation and Basic Actions,” Inquiry 13, no. 1–4 (1970): 108–125.
Danto, Arthur, “Basic Actions,” American Philosophical Quarterly 2, no. 2 (1965): 141–148.
Danto, Arthur, “What We Can Do,” Journal of Philosophy 60, no. 15 (1963): 435–444.
Davidson, Donald, Essays on Actions and Events (Oxford: Oxford University Press, 2011).
Davidson, Donald, “Actions, Reasons, and Causes,” in Essays on Action and Events (Oxford:
Oxford University Press, 2011a): pp. 1–19.
280
Davidson, Donald, “Agency,” in Essays on Action and Events (Oxford: Oxford University Press,
2011b): pp. 43–62.
Donnellan, Keith, “Knowing What I Am Doing,” Journal of Philosophy 60, no. 14 (1963): 401–
409.
Dorsch, Fabian, “Judging and the Scope of Mental Agency,” in L. O’Brien and M. Soteriou
(eds.), Mental Actions (Oxford: Oxford University Press, 2009): pp. 38–71.
Enç, Berent, How We Act: Causes, Reasons, and Intentions (Oxford: Oxford University Press,
2003).
Evans, Gareth, The Varieties of Reference (Oxford: Oxford University Press, 1982).
Falvey, Kevin, “Knowledge in Intention,” Philosophical Studies 99, no. 1 (2000): 21–44.
Finkelstein, Claire, “Acting on Intentions,” in B. Verbeek (ed.), Reasons and Intentions
(Ashgate, 2007): pp. 67–84.
Frege, Gottlob. Posthumous Writings, ed. H. Hermes, F. Kambartel, and F. Kaulbach (Oxford:
Oxford University Press, 1979).
Gibbons, John, “Reason in Action,” in Mental Actions, ed. Lucy O’Brien & Matthew Soteriou
(Oxford: Oxford University Press, 2009): pp. 72–94.
Ginet, Carl, On Action (Cambridge: Cambridge University Press, 1990).
Goldman, Alvin, A Theory of Human Action (Princeton: Princeton University Press, 1970).
Graham, Peter, “Epistemic Entitlement,” Noûs 16 no. 3 (2012): 449–482.
Hampshire, Stuart, Freedom of the Individual (Princeton: Princeton University Press, 1965).
Heeney, Matthew, “Diachronic Agency and Practical Entitlement,” European Journal of
Philosophy 28, no. 1 (2020): 177–198.
Hieronymi, Pamela, “Reasons For Action,” Proceedings of the Aristotelian Society 111 (2011):
407–427.
281
Hieronymi, Pamela, “Two Kinds of Agency,” in Mental Actions, ed. Lucy O’Brien & Matthew
Soteriou (Oxford: Oxford University Press, 2009): pp. 138–162.
Hieronymi, Pamela, “Responsibility for Believing,” Synthese 161, no. 3 (2008): 357–373.
Hieronymi, Pamela, “Controlling Attitudes,” Pacific Philosophical Quarterly 87, no. 1 (2006):
45–74.
Hieronymi, Pamela, “The Wrong Kind of Reason,” The Journal of Philosophy 102, no. 9 (2005):
437–457.
Hlobil, Ulf, “Inferring by Attaching Force,” Australasian Journal of Philosophy, forthcoming.
Hornsby, Jennifer, “Basic Activity,” Aristotelian Society Supplementary Volume 87, no. 1 (2013):
1–18.
Hornsby, Jennifer, Actions (London: Routledge, 1980).
James, William, The Principles of Psychology (Dover Publications, 1890).
Jenkins, David Russell, How Inference Isn’t Blind: Self-Conscious Inference and Its Role in
Doxastic Agency, Ph.D. dissertation, King’s College, London, 2018.
Kane, Robert, The Significance of Free Will (Oxford: Oxford University Press, 1996).
Kavka, Gregory, “The Toxin Puzzle,” Analysis 42, no. 1 (1983): 33–36.
Kietzmann, Christian, “Inference and the Taking Condition,” Ratio 31, no. 3 (2018): 294–302.
Kitcher, Patricia, Kant’s Thinker (Oxford: Oxford University Press, 2011).
Koziolek, Nicholas, “Infering as a Way of Knowing,” Synthese (2017). https://doi-
org.ezproxy.cul.columbia.edu/10.1007/s11229-017-1632-4.
Kripke, Saul, Wittgenstein on Rules and Private Language (Cambridge: Harvard University
Press, 1982).
Lavin, Douglass, “Action as a Form of Temporal Unity: On Anscombe’s Intention,” Canadian
Journal of Philosophy 45, no. 5 (2016): 609–629.
Lavin, Douglass, “Must There Be Basic Action?” Noûs 47, no. 2 (2013): 273–301.
282
Levy, Neil, “The Good, The Bad, and the Blameworthy,” Journal of Ethics and Social Philosophy
1, no. 2 (2005): 1–16.
Levy, Yair, “Action Unified,” Philosophical Quarterly 66, no. 262 (2013): 65–83.
Loux, Michael, Substance and Attribute: A Study in Ontology (Indiana: University of Notre
Dame Press, 1979).
Martin, MGF, “The Transparency of Experience,” Mind & Language 17, no. 4 (2002): 376–425.
McCann, Hugh, The Works of Agency: On Human Action, Will, and Freedom (Ithaca: Cornell
University Press, 1998).
McCann, Hugh, “On Mental Activity and Passivity: A Reply to Thalberg,” Mind 88, no. 352
(1979), 592–596.
McHugh, Conor and Way, Jonathan, “Against the Taking Condition,” Philosophical Issues 26,
no. 1 (2016): 314–331.
McHugh, Conor, “Judging as a Non-Voluntary Action,” Philosophical Studies 152, no. 2 (2011):
245–269.
McKenna, Michael, Conversation and Responsibility (Oxford: Oxford University Press, 2012).
Mele, Alfred, “Mental Action: A Case Study,” in L. O’Brien and M. Soteriou (eds.), Mental
Actions (Oxford: Oxford University Press, 2009): pp. 17–37.
Mele, Alfred, “Deciding to Act,” Philosophical Studies 100, no. 1 (2000): 81–108.
Mele, Alfred, The Springs of Action: Understanding Intentional Behavior (Oxford: Oxford
University Press, 1992a).
Mele, Alfred, “Intentions, Reasons, and Beliefs: Morals of the Toxin Puzzle,” Philosophical
Studies 68 (1992b): 171–194.
Moran, Richard, “Anscombe on ‘Practical Knowledge,’” Royal Institute of Philosophy
Supplement 55 (2004): 43–68.
Moran, Richard, Authority and Estrangement (Princeton: Princeton University Press, 2001).
283
Moya, Carlos, The Philosophy of Action: An Introduction (Cambridge: Polity Press, 1990).
O’Brien, Lucy, Self-Knowing Agents (Oxford: Oxford University Press, 2007).
O’Shaughnessy, Brian, Consciousness and the World (Oxford: Oxford University Press, 2000).
O’Shaughnessy, Brian, The Will, v. 2 (Cambridge: Cambridge University Press, 1980).
Peacocke, Christopher, Truly Understood (Oxford: Oxford University Press, 2008).
Peacocke, Christopher, “Conscious Attitudes, Attention, and Self-Knowledge,” in B. Smith and
C. MacDonald (eds.), Knowing Our Own Minds (Oxford: Oxford University Press, 1998):
pp. 63–98.
Pink, Thomas, Self-Determination (The Ethics of Action, v. 1) (Oxford: Oxford University Press,
2016).
Pink, Thomas, “Intentions and Two Models of Human Action,” in B. Verbeek (ed.), Reasons and
Intentions (Ashgate, 2007): pp. 153–179.
Pink, Thomas, Free Will: A Very Short Introduction (Oxford: Oxford University Press, 2004).
Pink, Thomas, The Psychology of Freedom (Cambridge: Cambridge University Press, 1996).
Proust, Joëlle, The Philosophy of Metacognition: Mental Agency and Self-Awareness (Oxford:
Oxford University Press, 2013).
Proust, Joëlle, “Is There a Sense of Agency for Thought?” in L. O’Brien and M. Soteriou (eds.),
Mental Actions (Oxford: Oxford University Press, 2009): pp. 253–279.
Rettler, Lindsay, “In Defense of Doxastic Blame,” Synthese 195 (2018): 2205–2226.
Rödl, Sebastian, Self-Consciousness (Cambridge: Harvard University Press, 2007).
Rovane, Carol, The Bounds of Agency (Princeton: Princeton University Press, 1998).
Scanlon, T.M., What We Owe to Each Other (Cambridge: Harvard University Press, 1998).
Schlosser, Mark, “The Metaphysics of Rule-Following,” Philosophical Studies 155, no. 3 (2011):
345–369.
284
Schwenkler, John, “Understanding ‘Practical Knowledge,’” Philosopher’s Imprint 15, no. 15
(2015): 1–32.
Sehon, Scott, Teleological Realism: Mind, Agency, and Explanation (Cambridge: MIT Press,
2005).
Shah, Nishi, “How Action Governs Intention,” Philosopher’s Imprint 8, no. 5 (2008): 1–19.
Shepherd, Joshua, “Deciding as Intentional Action: Control Over Decisions,” Australasian
Journal of Philosophy 93, no. 2 (2015): 335–351.
Shepherd, Joshua, “The Contours of Control,” Philosophical Studies 170, no. 3 (2014): 395–411.
Shoemaker, David, “Attributability, Answerability, and Accountability: Toward a Wider Theory
of Moral Responsibility,” Ethics 121, no. 3 (2011): 602–632.
Small, Will, “Practical Knowledge and the Structure of Action,” in G. Abel and J. Conant (eds.),
Rethinking Epistemology: Volume 2 (Berlin: De Gruyter, 2012), pp. 133–227.
Smith, A.D., “Agency and the Essence of Actions,” Philosophical Quarterly 153 (1988): 401–
421.
Smith, Angela, “Responsibility as Answerability,” Inquiry 58, no. 2 (2015): 99–126.
Smith, Angela, “Attributability, Answerability, and Accountability: In Defense of a Unified
Account,” Ethics 122 (2012): 575–589.
Smith, Angela, “Control, Responsibility, and Moral Assessment,” Philosophical Studies 138
(2008): 367–392.
Smith, Angela, “On Being and Holding Responsible,” The Journal of Ethics 11 (2007): 465–484.
Smith, Angela, “Responsibility for Attitudes: Activity and Passivity in Mental Life,” Ethics 115
(2005): 236–271.
Smith, Joel, "Self-Consciousness", The Stanford Encyclopedia of Philosophy (Fall 2017 Edition),
Edward N. Zalta (ed.), URL = <https://plato.stanford.edu/archives/fall2017/entries/self-
consciousness/>.
285
Soteriou, Matthew, The Mind’s Construction: The Ontology of Mind and Mental Action (Oxford:
Oxford University Press, 2013).
Soteriou, Matthew, “Content and the Stream of Consciousness,” Philosophical Perspectives 21,
no. 1 (2007): 543–568.
Soteriou, Matthew, “Mental Action and the Epistemology of Mind,” Noûs 39, no. 1 (2005): 83–
105.
Steward, Helen, “Action as Processes,” Philosophical Perspectives 26, no. 1 (2013): 373–388.
Steward, Helen, A Metaphysics for Freedom (Oxford: Oxford University Press, 2012).
Steward, Helen, The Ontology of Mind: Events, Processes, States (Oxford: Oxford University
Press, 1997).
Stoutland, Frederick, “Summary of Anscombe’s Intention,” in A. Ford, J. Hornsby, and F.
Stoutland (eds.), Essays on Anscombe’s Intention (Cambridge: Harvard University Press,
2011).
Stoutland, Frederick, “Basic Actions and Causality,” Journal of Philosophy 65, no. no. 16
(1968): 467–475.
Strawson, Galen, “Mental Ballistics, or: The Involuntariness of Spontaneity,” Proceedings of the
Aristotelian Society 103, no. 3 (2003): 227–257.
Strawson, Galen, “The Impossibility of Moral Responsibility,” Philosophical Studies 74, no. 1
(1994): 5–24.
Strawson, P.F., “Freedom and Resentment,” in G. Watson (ed.), Free Will 2nd ed. (Oxford:
Oxford University Press, 2003 [1962]): 72–93.
Thompson, Michael, Life and Action: Elementary Structures of Practice and Practical Thought
(Cambridge: Harvard University Press, 2008).
Valaris, Marcus, “Reasoning and Regress,” Mind 123, no. 489 (2014): 101–127.
Vendler, Zeno, “Verbs and Times,” Philosophical Review 66, no. 2 (1957): 143–160.
286
Watson, Gary, “Two Faces of Responsibility,” Philosophical Topics 24, no. 2 (1996): 227–248.
Watzl, Sebastian, Structuring Mind: The Nature of Attention and How it Shapes Consciousness
(Oxford: Oxford University Press, 2017).
Wedgwood, Ralph, “The Normative Force of Reasoning,” Noûs 40, no. 4 (2006): 660–686.
Williams, Bernard, Descartes: The Project of Pure Enquiry (Oxford: Penguin, 1978).
Williams, Bernard, Problems of the Self (Cambridge: Cambridge University Press, 1973).
Wittgenstein, Ludwig, Philosophical Investigations, trans. E. Anscombe (Oxford: Oxford
University Press, 1953).
Wolf, Susan, Freedom Within Reason (Oxford: Oxford University Press, 1990).
Wu, Wayne, “Mental Action and the Threat of Automaticity,” in A. Clark, J. Kiverstein, and T.
Vierkant (eds.), Decomposing the Will (Oxford: Oxford University Press, 2013): pp. 244–
261.