Our beliefs are a map of our world. Philosophers have written widely and informatively on major aspects of belief, but an important dimension of the topic still needs analysis: the conditions under which a cognitive state – say a perception or memory – may be taken to entail belief. I begin by clarifying the problem.
1. COGNITIVE SENSITIVITY AND DOXASTIC LIMITATIONS
Consider an ordinary perception like seeing a tree-lined country road. At a glance one sees trees, pavement, roadposts, and much more. Clearly one receives information – that the trees are in full foliage, that the pavement is gray, that the roadposts have lights, and so forth for many things seen. Must one come to believe each proposition expressing an item of the information perceptually presented to the senses? Many philosophers have thought so,Footnote 1 but the answer, in my judgment, is negative. Much information we receive does not enter our system of beliefs even if we stand ready for it to enter there if needed. This has been argued in detail and often acknowledged.Footnote 2 But the distinction between beliefs themselves and information that disposes us to form beliefs has not yet been adequately assimilated into philosophical literature.
The same point holds for the parallel distinction regarding intentions and other propositional attitudes. Consider intending, which is in a way like believing. Much as we are disposed to believe a proposition (p) clearly entailed by another proposition we believe, we are disposed to intend to do things we take to be necessary for achieving something we already intend. Examples abound. Let me describe two influential cases that have not been generally recognized as attributing belief and intention where the evidence needed to justify that is at best insufficient. One concerns linguistic meaning, the other intention.
For H. P. Grice, in saying something, x, to another (an “audience”), normally “A must intend to induce by x a belief in an audience, and he must also intend his utterance to be recognized as so intended.”Footnote 3 For instance, if I say to you ‘The meeting is at 3’, I must (normally) intend that you come to believe this by recognizing my intention to get you to believe it. Why, however, does it not suffice, in using English normally, simply to intend to say this, perhaps presupposingFootnote 4 rather than believing that you will believe it? And, for you to understand me, why does it not suffice for you simply to comprehend my statement? Why need you have a higher-order belief about my intention to induce a belief in you? A natural answer (implicit in some of what Grice says), is that if, for instance, I had misspoken and said ‘Jim is coming’ when I meant ‘Tim is coming’, I would (if I caught the mistake) believe that I intended to say ‘Tim’; and if you understood my (speaker) meaning as concerning Tim, you would, if asked what I meant, grant that you believed that I intended to say ‘Tim’. Moreover, if asked why you believe Tim is coming, you might well say it is because that is what I intended to tell you. These points are true but (as we shall see) do not require holding Grice's view.
As to the action-theoretic case, consider this passage from Searle:
[O]ne identifies the Intentionality [of an action] by its conditions of satisfaction. Using this test the content of the intention in action is (My arm goes up as a result of this intention in action.)
… in any real-life situation the intention in action … will include not only that my arm goes up but that it goes up in a certain way, and at a certain speed, etc.Footnote 5
The implicit reasoning is apparently in part this: I wouldn't really be raising my arm if it rises other than because I intend to raise it in certain way, a way that includes, as a cause of the arm's rising, my intending to raise it; hence, the content of my intention includes this crucial condition of its own satisfaction. This reasoning can be supported much as Grice's apparent reasoning can be. Suppose I find my arm going up precisely when I intend to raise it but, because a machine caused its rising on being triggered by reading my intention from a brain scan, I lack the sense of raising it. Here I may say (or readily agree if someone says) that I didn't intend to raise it in that way, or perhaps that I intended to raise it in the usual way.
Should we ascribe beliefs and intentions in the way these cases illustrate? Granted, if a normal adult who made the naming error I did were asked, ‘Did you intend to get me to believe that Tim (not Jim) is coming?’, the natural answer would be ‘yes’. But plainly, asking a question presents the hearer with the propositionally formulated prospect represented in the question. Where this kind of presentation of a prospect fits information already in our cognitive system, we naturally form beliefs of propositions that seem true given that information. But the fact that we often assent to such propositions without the sense of forming new beliefs does not show that the assent reflects beliefs already held rather than beliefs newly formed. Consider the proposition that you have just seen more than five English letters. Did you believe this simply by virtue of (in some way) receiving the information as you read that sentence? We need not accept this. Note that from the fact that you need not be aware of any special event of forming the belief upon being asked the question, it does not follow that you already had the belief. Belief-formation might be a natural response to certain information that the believer has not previously encountered in propositional form but is, for the believer, so obviously correct as to yield belief immediately on comprehending consideration. This seems highly plausible, and it is philosophically important (and doubtless important for understanding human psychology in ways best left to empirical science).
Note, too, that there is no need in such cases to appeal to the (supposed) belief to explain anything nor, indeed, any apparent way to explain why the belief would be formed. This is not to imply the apparently mistaken view that belief-formation requires considering the proposition in question. If, e.g., one were interested in the approximate number of English letters in a message written in Chinese and English mixed, one might form the belief that there are many simply on quickly viewing the message while flipping through a diary. Testimony-based beliefs are also commonly formed from simply “hearing” the propositions thereby believed without their being considered (as they may be if, e.g., complex or puzzling). But in testimony cases there is a plausible explanation of why the beliefs in question are formed.
The letter-counting example is one of many that illustrate profligate predication of beliefs. A further reason to avoid ascribing (de dicto) beliefs where some philosophers have tended to do so (as in cases like those cited from Grice and Searle) concerns what we might call doxastic requirements. These include possession of all the concepts that essentially figure in the propositional object of a (de dicto) belief, one whose object is indicated by a truth-apt that-clause.Footnote 6 Call this the conceptual capacity requirement. Imagine a child viewing an electrical box atop a telephone pole. A three-year-old might see the same colors and shapes which, when adults see them, may yield their believing that there is an electrical box there; but the typical three-year-old cannot have exactly this belief. The child can believe the electrical box to be gray; but this de re (objectual) belief – roughly, believing a thing (perhaps under some primitive description) to have a property – does not imply ability to believe that the electrical box is gray. Similarly, a tiny child's saying to Mama that Timmy is coming to play does not require an intention with (e.g.) the content: to get Mama to believe that Timmy is coming, by means of recognizing my intention to say this.
These points suggest a doxastic qualification principle – one having plausible counterparts for other propositional attitudes:
The conceptual capacity constraint: If a person, S, believes that p, S has at least the (relevant) conceptual capacity of the least conceptually developed being capable of having that belief.
Granted, everyday usage enables us to say that, e.g., the child believes that the electrical box is for birds to enter in the rain. But this de dicto locution is best taken to mean something better expressed as ‘The child believes the electrical box to be for birds to enter in the rain’. This ascribes a de re belief. If the conceptual capacity requirement is sound, then until the child learns enough about electricity, the propositional belief just specified is not possible. This is indicated in a number of ways. One is that the child would (typically) have no sense of the box's electrical functions. Related indications of insufficient capacity to hold the belief should be evident, and there is no need to argue the point here.
2. INTENTIONALISTIC EXPANSIBILITY
There are many kinds of cases in which we form new propositional attitudes. Asking people questions whose answers are obvious to them but not yet believed will tend to evoke the corresponding new beliefs; describing something one would immediately see as desirable may evoke some new desire; and similar points hold for other propositional attitudes. Still, intentionalistic expansibility is limited by our conceptual capacities. Conceptual capacity is not just a matter of understanding language or propositional content. It also concerns the kind of recognition that goes with seeing what is said if a belief is linguistically expressed to one and with discerning some of its implications. I understand the sequence, 2 is larger than 1, 3 larger than 2, and so forth ad infinitum. But assuming finite mental capacity, could I have an infinite set of dispositional beliefs whose contents are (distributively) all the propositions in question? Surely not, and this applies to dispositions to believe as well. At some point or other, the comparative proposition I would have to understand (p), would have such complexity – whether in comparisons of numbers in rows of the forms just indicated or in superscripted numbers or in some other form – that, if I tried to get p in mind, I could not do so owing to my conceptual limitations. I could neither have the thought that p nor, if I could read a formulation of it, remember its first part by the time I read or considered its last part.
The indicated limitation is broadly conceptual, but the case poses a problem regarding the nature of understanding: even if one can conceive all the raw materials in a proposition, it may be so complex, if only in an iterative way, that one cannot have it in mind or otherwise recognize it as distinct from a similar proposition with different content. There are at least three plausible comprehensional standards that I suspect are violated by positing infinite sets of beliefs (in beings like me). The first is the thinkability constraint: if a person, S, cannot have the thought that p, then S cannot believe that p.Footnote 7
This is intuitively plausible given the naturalness of taking objects of belief to be things we can have in mind, think of, or even focus on.Footnote 8 The idea is roughly that what one can believe at all one can believe occurrently.
But consider something closely related:
The comprehensibility constraint: If, even given a complete mastery of a language, L, powerful enough for formulating p, S cannot understand any formulation of p in L, then S cannot believe p.
This is weaker than the thinkability constraint, at least assuming that one could understand a proposition presented to one in writing or speech even if one could not have it in mind “all at once” in the way that constraint requires. That constraint requires conscious accessibility “at a glance”; the comprehensibility constraint, by contrast, takes as sufficient memorial accessibility, which might allow comprehension by partial (perhaps sequential) recall or, perhaps, a kind of reconstruction by remembered segments. Unlike the thinkability constraint, this one does not imply that beliefs must be capable of being occurrent.
Both these formulations are closely related to a third:
The discriminability constraint: If S cannot, at least by drawing inferences and attending to examples, discriminate p from certain appropriately contrasting propositions, then S cannot believe that p.
Discriminability is not obviously implied by the other formulations, and it does not imply thinkability, though it apparently implies comprehensibility and presumably recognizability (still another candidate for a conceptual capacity constraint). One element in discriminability is the ability to see certain entailments and non-entailments. Another is ability to explain what the proposition says, at least by using examples.
A great deal must be said to explicate these constraints. My purposes do not require doing so. One reason is that my main points about belief and belief-formation require only taking at least one of the constraints to hold (for at least de dicto beliefs). Another is that, like similar plausible constraints, they all preclude possession of infinitely many beliefs in finite minds. When Hamlet said, “What a piece of work is a man, how noble in reason, how/ infinite in faculties … in action how like an angel, in apprehension how like/ a god!” (Hamlet 2, iii), he was mistaken if he implied that, even dispositionally, we can believe infinitely many propositions. There is no need to deny that one can have a de re belief, of a sequence, that it has an infinite number of elements. But this does not imply comprehending access even to any arbitrarily chosen finite member of that sequence.Footnote 9
Given these considerations, a number of initially plausible epistemic principles must be replaced by more complicated formulations. Consider first the idea of the conjunctivity of justification: If one is justified in believing p and in believing q, then one is justified in believing p & q. Suppose p is at the limit of one's comprehension. Then the principle will not hold, nor is there an easy repair, at least on the assumption that we can comprehend what we are justified in believing. The principle allows that, on the psychological side, q could be as complex as p and that, on the normative side, p and q could be independent and such that one's justification for their conjunction is lower than for either alone. On at least the first count, what has been called the “multi-premise closure principle” fails: the principle that if one is in a position to know that p and to know that q, then one is in a position to know their conjunction.Footnote 10
The kind of problem I am raising for belief besets other propositional attitudes. Searle is again illustrative: “intentions, like desires, are closed under conjunction.”Footnote 11 Thus, if you intend to A and intend to B, you intend to A & B. The crucial point driving such closure views is apparently that for attitudes having a propositional object or, as with infinitives, objects individuated with comparable fineness, many people are disposed, ordinary cases, to realize conjunctivity. This point is supported by the fact that – as brought out already – commonly dispositions to form propositional attitudes are realizable as imperceptibly as the manifestation of the relevant attitude already possessed, and are realizable under similar conditions. But human understanding imposes limits. A parody of Wittgenstein's Tractatus encapsulates part of my point as applied to any propositional attitude: Whereof one cannot understand, thereof one cannot believe.
3. THE MULTIPLE PATHS TO BELIEF-FORMATION
We can resist overascribing beliefs without underestimating the myriad beliefs people normally form. The examples of belief-formation so far given suggest that many experiences produce beliefs. The sources of belief is a topic having dimensions that require empirical inquiry of a kind that goes beyond philosophical reflection, but philosophical reflection concerning the concept of believing and related concepts does indicate a number of ways in which beliefs arises. I begin with perception.
3.1 Perception and belief
Perception is a source of many beliefs. A historically important instance of this view is Reid's claim that “the perception of an object implies both a conception of its form and a belief of its present existence.”Footnote 12 One might perhaps stretch his language and interpret him as affirming only a disposition to form a de dicto belief and an unwillingness to doubt the propositions in question. But much that we cannot (given our psychological make-up) doubt need not be believed. A moment ago I could not have doubted that the temperature of the surrounding air is below 1,000 degrees centigrade, but I had no such belief. Indubitability does not entail belief.
One might here speak of “implicit belief.” But if this term implies actual belief as opposed to a certain disposition to form it given consideration of the relevant proposition, it seems inappropriate here. It is plausibly applied where belief-ascription explains something in the context. In a cool room, my putting on a sweater, even in the midst of avidly telling a story, might be explained by my believing the room is cool, whether this occurs to me or not. By contrast, in the case of temperature below 1,000 degrees, we have no idea what would be explained by my believing the room is below that temperature or what would explain my forming this belief.
Reid speaks of conviction as if it might result from perception in the way de dicto belief often does, and it seems best to conclude that he is not observing the distinction between dispositional beliefs and dispositions to believe. Still, even if he must be read as thinking that at least some de dicto beliefs are formed with every perception, we need not take him to have considered perceptually formed beliefs to be occurrent in a sense implying that they manifest themselves in consciousness in the way my belief that p does when I have the thought that p. In any case, most of what I say about de dicto belief will hold also for de re belief, but my space does not permit addressing that case here.Footnote 13
If Reid did not intend to take perception of external objects to entail belief, some philosophers clearly have. Penelope Maddy (Reference Maddy1980: 171) has held that
P perceives an x at 1 if and only if
i) There is an x at 1;
ii) P acquires perceptual beliefs about x, in particular, that there is an x at 1;
iii) the x at 1 is involved in the generation of his perceptual belief state in an appropriate way …Footnote 14
She says that this account is intended to capture
a strict sense of perception according to which, for example, a hunter does not perceive a pheasant hen sitting motionless in her nest if she blends too perfectly with her surroundings for him to distinguish her from them … he does not acquire any [beliefs] which are about her in a strict sense – for example, the perceptual belief that there is a pheasant hen in such-and-such a location. (Maddy Reference Maddy1980: 171)
She recognizes a “loose sense” in which the hunter may have beliefs about the hen. Maddy may be thinking of believing (de re) of the hen that it is brownish while having no sense of what it is that has this color (other than, say, that spot); but she clearly takes perception to entail having at least some de dicto beliefs.
The points made so far might bring to mind the adage, “Seeing is believing.” If this means that if we see that something is so, we believe that it is, it seems true. But it is not unnaturally taken to mean that if we perceptually see some object or event, we then believe that it exists, normally under a description that indicates some property salient in our sensory field (Reid surely thought something at least close to this). Consider again the print you now see. Do you believe that it is not in red typeface? Don't you “believe your eyes”? Suppose you do believe it is not red. We still have no good reason to hold that for every property clearly present in your visual field (and which you can conceptualize), you believe that the object in question has the property or even to have that property: being red, being on white paper, being ahead. This idea implausibly inflates the doxastic content of the mind and (in my view) misrepresents the most basic way in which perception conveys information.
In cases like this, it may be useful to speak of virtual belief of a proposition. This is not belief – and Section 5 will indicate important differences between such “beliefs” and actual beliefs – but virtual beliefs represent part of what might be called one's information base: the body of information that is, if not at one's fingertips, then ready to occupy that place doxastically. Here is a (rough) first approximation:
S virtually believes p at t when (and perhaps only when) (1) S does not believe p at t; (2) S has a readily realizable disposition to believe p at t; and (3) the disposition is based at least partly on S's being, at t, experientially presented, in thought or in perception, or being in readiness to be so presented (e.g. by recalling some incident easily recovered from memory), with some property or relation such that (a) p ascribes it to something, (b) having the thought that p would (given the circumstances S is in at t) normally lead S to believe p, and (c) S would normally form the belief spontaneously without reflecting on p or seeking some premise for it.Footnote 15
There are propositions one is about to come to believe which this formulation does not include, as where one is listening to ongoing testimony or viewing a horse approaching the finish line. In these cases one is on the verge of believing, but need not – at t – virtually believe, the to-be-discovered proposition(s) in question. Virtual believing is relative to information one in some way has at t.
What is virtually believed may, if it becomes actually believed, be believed inferentially provided (in my usage) the inference underlying it is in some way “implicit” in S's information base, as where, asked whether one believes that 300.444 is larger than 300, one immediately comes to believe this. Here the premise is presumably something like the self-evident arithmetic proposition that an integer plus a decimal is larger than that integer. Compare a non-inferential case: believing that you are now reading black print. If you are, and you then consider whether you are, can you resist believing this? I cannot.
If seeing something normally yields certain virtual beliefs, this should help to explain why it has been taken to produce (propositional) believing automatically. If, in a well-lit room, we think about what we see (at least about what it is that one sees), we will normally quite spontaneously, and often irresistibly, form certain beliefs about one or another thing we see, and may form them so quickly and naturally that we have no sense of doing so. This is partly because they are non-inferential, but also in part because, unlike information contained in certain pictures, they are based on property discernments that are (for the perceiver in question) readily propositionalizable: roughly, capturable by a proposition that is comprehensible to S and attributes the relevant properties. Still, it is apparently only or mainly where what is seen, or an aspect of it, has some significance for the perceiver that seeing a property of the thing produces the corresponding property-ascriptive belief without any thought of the proposition's coming before the mind. Seeing need not produce believing even when one has the relevant thing in view.Footnote 16
It also seems – and this is important for understanding the action-theoretic implications of this paper – that responding to a property of something, such as the height of an obstacle in our path which we automatically step over, does not entail believing it to have that property. Discriminating the property, such as the height of the obstacle, may suffice to guide our steps so that we do not trip. I doubt that perception can guide behavior only through producing highly specific guiding beliefs, such as those specifying actions the agent believes to suffice for achieving the agent's aim in the situation.Footnote 17 This is at any rate not self-evident. It seems possible that a general belief relevant to guiding action, say that one must avoid mossy pavement while running, may open a non-doxastic pathway – one not mediated by any specific belief – from perception of such dangers to action that “automatically” avoids them. In at least many such cases, the guiding propositions would be virtually believed, but I leave open the empirical question whether perceptually supplied information can guide action only if the agent has at least a virtual belief of the relevant content.
Given the rough terms in which I have described virtual belief, one may wonder whether the concept is needed. One reason is that, as the term suggests, philosophers and others – including, importantly, survey researchers – often ascribe beliefs where virtual belief is what best answers to both their relevant data and the cognitive state they are justified in ascribing. Another is that virtual beliefs, as dispositions to form beliefs but not actual beliefs, play a different role in thought and action. The content of our beliefs is, in a direct way, presuppositionally available to us to yield premises in inference and bases for action. Playing this role is indeed partly constitutive of belief. By contrast, dispositions to believe must first yield belief-formation for the propositions in question to occupy this role. More will be said in what follows to clarify these points. I would add here only that recognizing the distinction between actual and virtual beliefs gives due weight to important differences in ways we possess and respond to information and to different ways in which our having it guides our thinking and doing.
I leave open whether being disposed to believe p implies virtually believing p. For one thing, S might be disposed to believe p only given reflection, as where S believes q and r, which clearly entail p, but has never thought of them together. Suppose I believe Anne left class an hour early and also believe Chicago is two hours away. These may be mutually irrelevant for me until I am asked shortly after class whether she will be late for a lecture I also believe is in Chicago in an hour. I might say no, but only after inferring this. Here my disposition to believe is apparently a case of what some philosophers regard as implicit (actual) belief.
3.2 Intellectual apprehension, reflection, and imagination
Some of my points may be generalized not only to other perceptual modes but to intellectual apprehension, which is significantly analogous to perception. Apprehending is like seeing in that (1) when an obviously true proposition is before the mind and we comprehendingly consider whether it is true, we tend to believe it, and (2) when we apprehend that p, we believe p. However, just as we can see an object with its color and shape in view without forming the belief that it has that color and shape, we can comprehendingly consider a true proposition with its plausibility quite evident to us, yet not believe it.Footnote 18
Two cautions apply to both perception of the concrete and apprehension of the abstract. The first concerns the naturalness of doxastic overascription. The second concerns justification and knowledge in relation to propositional beliefs.
On the first point, we cannot properly understand perception or apprehension if we overintellectualize them in the way that, perhaps partly because of misunderstanding of the idea that perception is belief-entailing (e.g. that “seeing is believing”) is natural for many philosophers and others studying belief. But there is another reason for overascription. In reflecting on the range of our beliefs, it is natural to do such things as (1) look around with the question of what we believe in mind, which encourages forming beliefs expressing perceptually obvious environmental facts; (2) think of important tenets (or simply views we hold), which encourages forming beliefs of propositions obviously entailed or suggested by them or indeed new beliefs corresponding to new formulations of the tenets; and (3) consider various propositions we think might be candidates for belief, which easily yields coming to believe propositions one was initially only disposed to believe. This is a defective method for discovering what one already believes. If, for example, in wanting to discover or even simply list what we believe about our environment, we look around, we tend to form beliefs we would not otherwise have: for each clearly obtaining prospect that we either propositionally consider – as when asking whether the floor is real wood or woody-looking vinyl – or find salient, as where we glimpse an ominous crack in a windowpane, we tend to come to believe some evident propositions.
A second point important here is that it is crucial for understanding both belief and its justification – whether the belief is perceptual or intuitive – that we understand how experience, especially perceptual experience, can justify belief or render it knowledge. Seeing and apprehending can provide (propositional) justification and produce knowledge whether or not S forms the beliefs they may justify, or acquires the knowledge they may ground. I now realize that I have been hearing the hum of a lawn mower in the distance. Hearing that gave me justification for believing the proposition before I did believe it, and the hum now grounds perceptual knowledge that I do hear it. Phenomenally indistinguishable sensory experience, as with hallucination, may have a similar doxastic effect. Imagining may also produce beliefs. So may simply imaging or even daydreaming. Much of what has been said indicates how belief arises in these cases, and there is no need for further discussion here.
A third, related point is that inference as a route to belief-formation differs from all of these others (though not all inference is belief-forming, as many logic exercises illustrate). But inference is like the other routes in this: its occurrence can explain why we sincerely and spontaneously assent to a proposition when it is presented to us even without our having previously believed it. Such assent is indeed a primary and, conceptually, a partially constitutive manifestation of belief. But, as noted in the case of seeing “directly” that 300.444 is larger than 300, immediate assent need not indicate antecedent belief, as opposed to its formation as a realization of a disposition to believe, nor does it preclude the newly formed belief's being inferentially based on some other belief.
These points are made with parsimony in mind as a philosophical ideal. Where non-doxastic elements can explain why a person sincerely assents to p, we should consider whether the person already believed p or formed the belief at the time. This question is often difficult to answer; but in belief-ascription, as in positing other psychological properties, we should avoid going beyond what is needed for understanding persons and should certainly avoid positing beliefs where the relevant data are adequately explained by positing, in the ways illustrated, only dispositions to believe together with belief-forming tendencies that proceed along one of the paths described here. To posit belief where the behavior or fact in question is better explained by ascribing a disposition to believe is an instance of what I call doxasticism.
4. COGNITIVE SHORT-CUTS: THE ECONOMY OF NATURE
It seems a blessing of nature that the indefinitely large body of information we receive from perception and can receive from thought provides us with non-inferential routes to belief and, commonly, to knowledge. This information need be neither received nor stored propositionally. Another general lesson illustrated by many of the points emerging so far is that in determining what a (human) person believes, we should not use either a regressive method or an interrogative one.
By the regressive method (illustrated by Searle's attempt to characterize intentional action), we back into belief-attribution by asking under what conditions a readily understandable (and usually obvious) fact about human behavior – linguistic or other – may be plausibly thought to hold. We then suppose that S succeeds in the behavior only if S believes that these conditions hold. This method assimilates dispositions to believe to actual beliefs and virtual knowledge and belief to their actual counterparts. Recalling my standing on a friend's balcony, if I had believed the floor might collapse beneath me, I would not have entered; but to take this to imply that I actually believed it would not collapse (as I might after a fire in the building) wrongly suggests that I formed such a belief despite the normality of the building. My cognitive system readies me to form such beliefs; but that readiness does not imply believing the myriad propositions in question.
Complementing the regressive method for ascertaining what beliefs a person has is the interrogative method. We ask ourselves, in some context of inquiry, what we believe. I ask myself, for instance (as I type in English), whether I intend to write in English. With this question in mind now, I believe that I intend to write in English. But my having the intention to write in English does not entail that I actually believe, as opposed to virtually believing, that I intend to write in English. (I cannot disbelieve this, of course.) Note too that a tiny child can intentionally raise its arm before even having the concept of intention, hence before satisfying the conceptual capacity requirement for believing that it intends to raise it.
If perception constantly yields both dispositions to believe and actual beliefs, memory is a storehouse of both. Retrospection, moreover, can bring to mind beliefs we already hold and also produce new beliefs we are disposed to hold on the basis of the recalled situation. Thinking about a canceled flight, I may picture a helpful airline representative. I believed she was efficient; I did not believe, though I do now in recalling her and asking myself questions about her, that she was a native speaker of English. A first approximation of a general belief-formation tendency here is
The experiential recognition tendency: when (a) S is perceiving or apprehending – for instance seeing – a property F, or (b) has an experience as of perceiving or apprehending F, and (c) considers the proposition that (or a proposition to the effect that) something is F, including propositions to the effect that S sees (or otherwise perceives) that something is F, S tends to believe (non-inferentially) that something is F.Footnote 19
Reid insisted on the non-inferential character of the “conviction” he took perception to entail. He could surely endorse this principle, though it is not clear how far it takes us toward understanding his view of the relation between perception and belief.Footnote 20
A similar principle apparently holds for recalling and for introspection. If I recall (or have an experience as of recalling) the scent of lilacs in the breeze, I tend to believe that there was such a smell on the occasion in question. A parallel principle holds for my introspecting (or having an experience as of introspecting) being apprehensive. (The ‘as of’ clauses are needed for the same reason that we need to allow for beliefs to arise from sufficiently vivid hallucinations.)
Both propositions self-evidently entailed by some proposition a person believes and propositions perceptually indicated to a person belong to a category we might call belief-eliciting. Roughly, a proposition p is belief-eliciting for a person, S, at t, if and only if S does not believe p at t but, upon merely considering it, would tend to believe it. The tendency might or might not be inferential as opposed to perceptual, and various factors can inhibit it. But, for each of us, there are numerous belief-eliciting propositions of many kinds, and a subset of these constitutes an important part of our information base. Loosely, some of these propositions may be said to be “known” – and, since they are virtually believed, it is often clear how they may be conceived as virtual knowledge. Their number and variety testifies to the bounty of our constitution. Nature is at once informationally generous and doxastically economical.
We should bear in mind a way in which epistemologists’ deep concern with skepticism can abet doxasticism. If a skeptic challenges my claim to know that p – say, that there is print here – then even though p is of a belief-eliciting kind for me, I may now form only the cautious belief that I have a steadfast experience as of print before me and the related belief that the only good explanation of such an experience entails that p. I may now believe p inferentially. The example well illustrates at least two things: (1) that many propositions believable only inferentially in one situation may, in other situations, be believed non-inferentially by the short-cut from information conveyed non-propositionally – say by acquaintance with properties of objects – to a doxastic representation thereof; and (2) how the regressive method of arriving at belief-attributions may aid doxasticism.
Might one posit “unconscious inferences” where beliefs that seem non-inferential are supported by a ground, such as a sense-experience? Some philosophers are inclined to do so.Footnote 21 One can insist that there is such a thing, but this view is not warranted by the apparent fact that the brain may process information in a way that can be represented as a transition from one or more propositions to another they support (or appear to S to support). If inference is simply information processing of the kind apparently occurring in the brain when a carpenter's seeing, say, the bend in a board instantly yields the belief that it is warped, then there is no need to dissent. But from the concept of such perceptually responsive belief-formation, nothing follows about what the brain does, nor is determining that a task of philosophy.
Suppose, however, that, as is usual in epistemology, we understand inference as something that, in many cases, we may at will draw or resist drawing; that is valid or invalid; has one or more premises and a conclusion; is equivalent to a kind of reasoning understood as something we do; is an event or process with (at least normally) some manifestation in consciousness; and is at least sometimes under direct voluntary control. Then it is plausible to take inference to be roughly a certain kind of passage of thought from one or more propositions to another, guided by a sense of some relation of support between the former – call it the premise set – and the latter, the conclusion.Footnote 22 The support relation may be entailment or probabilification (and may or may not be present, since an inference can be logically or evidentially defective). On this view, inferences (at least normally) are phenomenally real elements that figure in consciousness, even if incompletely and very briefly present therein.
To posit inferences underlying the formation of perceptual beliefs is an instance of intellectualism. In epistemology, as illustrated by doxasticism, intellectualism is mainly unjustified overascription of psychological states, events, or processes. Such overascription includes inflating higher-order cognition. If, however, we take inference to require a kind of appearance in consciousness, we must be careful not to require, for every inference, either consciousness that one is drawing it or any particular length of time for its appearance in consciousness. The novice in carpentry may form the belief that a board is bent only after laying it on a flat surface and then inferring that it is warped. The experienced carpenter may see the warpage straightaway. This illustrates a common kind of cognitive short-cut: what in learning we can believe only inferentially, with experience we may directly see to be so.
Granted, the carpenter sees the warpage, and sees that the board is warped, by seeing its curvature; but seeing that p on the basis of seeing a property F – including patterns – does not entail that we have inferred p from the proposition that the thing in question is F. To be sure, where this visual basis relation – seeing that p by seeing F – obtains, we both believe p and are disposed to cite F if asked why p. People differ both in how readily they form non-inferential beliefs from perceptual and other experiences, in how readily they draw inferences on such occasions, and in the forms their inferences take. A multitude of empirical questions lurk here. Philosophical clarity on the concepts we use in formulating such questions should benefit both empirical inquiry and philosophical theory.
5. BELIEF AND ACTION
We have seen various kinds of doxasticism as an overascription of beliefs, and I have indicated how the same tendency may yield overascribing intentions and other propositional attitudes, as where we might be considered to have “implicit” intentions or desires when there is only a disposition to form them. In relation to both cognitive and conative cases, action theory must be pursued in a more nuanced way than has been generally recognized. Suppose we take the standard view that action requires at least one guiding belief. Are we guilty of doxasticism in the practical realm?
5.1 The teleological paradigm
Paradigmatically intentional actions are those performed in order to bring something about. Suppose I apply my brakes (in order) to stop for a red light. Must I not believe doing that will achieve this – or bears some instrumental connection to achieving it? Such instrumental belief might be de re, but our purposes do not require dealing with that case separately. It is plausible to conceive action, at least when intentional, as belief-guided, and few have questioned whether (as I have argued elsewhereFootnote 23), when S A-s in order to bring something about, S has an instrumental belief connecting the envisaged act and goal. This paper brings to the fore the question whether intentional action can instead be guided by dispositions to believe. These are, after all, cognitive elements. Might they have much the same action-guiding power as the beliefs they are a disposition to form?
A short answer would be that we can always explain S's intentionally A-ing by citing a belief to the effect that A-ing will realize the relevant goal of the action, but we cannot (or cannot in general) do so by citing a disposition to form such a belief. Such an instrumental belief might be that one must brake in order to stop. But the fact that such belief-ascriptions would explain the action does not entail that their applicability is necessary for explaining it. Perhaps, however, some instrumental belief, even if de re, is necessary. Certainly the case is one in which the agent does something that responds to salient perceptible facts of a belief-eliciting kind. Perhaps the appearance of a red light or, especially, an obstacle, tends to evoke the realization (hence belief) that one must brake. Another spontaneous purposive action is reaching for grapes on a plate in order to eat them. This seems possible only given a belief to the effect that eating them requires reaching for them. Might the action be merely done from habit? There may be at least two cases of habit relevant here: in one, there is a doing that is not intentional – non-intentional, as opposed to unintentional. Biting one's fingernails can be such a case. In the other, an intentional action is triggered without forethought but is still belief-guided, if only by de re belief. Consider shaking hands on meeting someone. One might believe this to be the thing to do. With this case we enter the domain of conventional and also rehearsed behavior. These and other cases require a broad conception of the content of intention.
5.2 The indefinite scope of intention
The role of habit in relation to intentional action raises a question concerning the scope of instrumental belief and, correspondingly, of intention. Consider learning a piano piece. At first one may consciously aim at a certain fingering. Here it is plausible to posit instrumental beliefs regarding ways of moving one's fingers. Once the piece is memorized, one may be able to play it upon deciding to do so while having the quasi-indexical belief that this is how it goes, where a single cognitive picture or “script” carries the content of the instrumental belief. In cases like this, the whole performance is intentional and its actional parts are each either automatic in a sense that does not entail being intentional (or unintentional, of course) or are intentional in something like the following way: as de re believed to be necessary, or (consistent with that) as non-intentional doings believed, de re, to belong to the intended overall performance. Either way, we can account both for surprise at mishaps and for the presence of an appropriate instrumental belief, yet we need not posit individual instrumental beliefs governing each finger movement anticipated as part of the performance. We need not; but my view leaves open that we may, as where a pianist might be self-consciously trying to render every note expressively. This might lead to forming many action-guiding beliefs during the performance. Here, by contrast with much complex behavior, positing the usually unneeded beliefs would explain aspects of the performance and is explained by the pianist's particular focal aims.
The first case – that of the automaticity of scripted behavior – might be understood in terms of something like a program. The details of such a program require empirical investigation that would take us outside philosophy, but the philosophical point is that once a pattern of behavior is sufficiently familiar and, especially when it is remembered clearly, it may be governed by an intention whose content is a script, in the sense that (1) if S is unimpeded and fully successful, what triggers the intention to realize the scripted content – a kind of global intention, we might say – yields the entire sequence of doings without the formation of intentions having focal contents corresponding to each constituent doing that is a basic act-type for the agent; (2) S is disposed to believe, and may virtually know, of each element in a suitable subset of these doings, that it is part of the intended behavior; (3) S is disposed, on the basis of the governing intentional script, both to correct errors in the behavior, such as playing a wrong note, and to consider the erroneous doing unintentional and the corrective action intentional; and (4) the entire behavioral sequence is guided by the scripted intention and one or more related instrumental beliefs: by those beliefs’ creating non-doxastic pathways from perception to action, including both scripted and corrective action. Beliefs, then, have an indirect as well as a direct guiding role: they can empower non-propositional information to guide action.
The global intention view I am sketching refines the belief-desire model of intentional action so as to explain complex behavior without yielding to doxasticism or other kinds of intellectualism. A further advantage of the proposal is that it avoids attributing a kind of causal power to dispositions to believe which they apparently do not have: the power to produce actions. Something should now be said to indicate why they apparently lack it. Part of the answer concerns orders of psychological properties.
5.3 Two orders of cognitive disposition
Dispositions to believe are higher-order properties, being dispositions to instantiate the dispositional property of believing. This does not entail lacking causal power, but at least the typical cases of dispositional properties serving as causes of events (such as actions) require both an event elicitor, such as the thought of a proposition, p, that is the content of the belief S is disposed to form, and an event manifestation, such as affirming that p. For dispositions to believe, the constitutive event manifestation is forming the relevant belief. This event seems the best candidate for that by which dispositions to believe might be thought to explain actions, say changing behavior in the way one's newly formed belief calls for. But we must not assume that the formation of a belief can explain what a belief itself can.
Granted, forming a belief entails the presence of that belief and so may seem to explain what the latter explains. The action-explaining role of belief, however, commonly requires its persisting as a sustainer of actions over their duration; but belief-formation is momentary and cannot play a temporally continuing role, hence cannot sustain temporally extended actions. Take an example. Believing I ought to phone an aging aunt may, upon my noting a phone, cause (or figure essentially in what causes) my phoning her. This deed is temporally extended and a constitutive manifestation of the belief. By contrast, being disposed to believe I ought to phone her has no action as a constitutive manifestation and, in part for this reason, is causally further from action than the dispositional belief it may produce. Dispositional beliefs, then, can be seen as potential causes or partial causes of action; dispositions to believe are at best potential causes of the genesis of such “direct” causes and are best seen as affecting action mainly through those causes.
There is a wider action-theoretic perspective for determining whether belief is required for intentional action. Consider the plausible view that intentional action is action performed in order to achieve something and is explainable by appeal to the agent's wanting to achieve that and having an instrumental belief to the effect that the action conduces to achieving it. In this light, two kinds of variables seem needed to explain intentional action: a motivational variable to explain the energy underlying it, and a cognitive one to explain its “direction” (where the direction is not necessarily physical but a matter of where the instrumental belief leads the agent, which may be in thought alone). In broad terms, desire without belief would be directionless; instrumental belief without desire would be powerless.
Desires, then, are to be understood partly in terms of yielding action given a belief indicating how their object is realizable; beliefs – most notably those with instrumental content – are to be understood partly in terms of their tendency to yield action given appropriate desires.Footnote 24 The very conception of desire (wanting) requires a connection to belief. This is a systematic consideration. One cannot account for the role of desire in action (or intention) without positing instrumental beliefs.
A supporting epistemological consideration is this. It is part of the concept of both beliefs and desires that actions of a certain kind (including at least those that seem goal-directed) are evidences of their presence. It is also part of the concept of intentional action that it is explainable by appeal to certain desires and beliefs. This does not by itself entail that dispositions to believe cannot also play the instrumental role of cognitively guiding behavior. But it is difficult to see how they can do this given that they are not propositional attitudes. We might at least say that the concept of intentional action is implicitly tied to cognitive propositional attitudes that have a kind of instrumental content. Figuratively, whereas beliefs represent lines drawn on our map of the world, dispositions to believe represent our inclinations to draw them; but they provide no path to follow until the inclination is realized in drawing one, and there may be obstacles preventing that realization, whereas even erasable lines are actual.
Although we identify dispositions to believe by the propositional content of the beliefs whose formation manifests them, they do not themselves have instrumental or other propositional content. This is one reason why they are unfit for the guiding role that instrumental beliefs play in action. Given this point together with the other considerations noted in this section, it is reasonable to conclude that intentional action requires a doxastic basis – though far fewer beliefs than would be posited if dispositions to form beliefs are assimilated to actual beliefs. Intentional action (including activities) may be scripted. That indicates the potential complexity of the content of the underlying propositional attitudes, but it does not support the view that mere dispositions to believe may play the causal and explanatory role played by those attitudes.
6. SOME NORMATIVE IMPLICATIONS: TWO DIMENSIONS OF COMMITMENT
This final section presents yet another kind of reason why it is important to ascertain whether dispositions to believe can play the kind of action-explaining role that beliefs do. It will help to compare belief with intention and show how this paper clarifies responsibilities of self-regulation in the behavioral as well as the cognitive domain.
Intention may be broadly conceived as a commitment of the will, and we tend to realize our intentions when we can – though reversal and irresolution are not unusual.Footnote 25 Belief may be conceived as a commitment of the intellect; we tend to presuppose their propositional content in thought and inference and to be guided by them in action. But dispositions to believe are not commitments of the intellect, nor do they imply a tendency toward action comparable to that of intentions. Unrealized intentions are usually a kind of agential failure and are not common in those who optimally combine resoluteness and prudence. Unrealized dispositions to believe are normal, multitudinous, and indispensable posits for explaining cognitive properties while avoiding doxastic inflation of the mind.
A similar point holds for dispositions to intend: unlike intentions themselves, they do not represent predominant motivation, the executive kind that is characteristic of intentions and governs intentional action. If they are motives at all, as where they are constituted by certain desires, they do not constitute executive motivation; and if too many yielded intentions, agency would be burdened by an overabundance of plans and by conflicting action tendencies. Beliefs not only guide action but also provide premises that may guide inference. Dispositions to believe are one step away from both roles, though the step may be, as we have seen, quick and easy. Critical capacities we can bring to bear on monitoring our actions and belief should also be applied to those dispositions. Many of them do not yield beliefs in the normal course of life, as should be evident from our examples. But some do. Where that is likely, a self-critical person can often block belief-formation if that is needed.
Our critical faculties, then, should operate on more than one level. If we are morally and intellectually virtuous, these faculties operate both at the level of action and, on the mental side, the level of beliefs, desires, and intentions. Dispositions to believe, like other dispositions to form propositional attitudes, are higher level. But our critical capacities should operate at that level as well. One might think, given the distance of dispositions to believe from belief itself and from action, that monitoring them is of minor importance. But sometimes that distance is short or likely to be traversed. Suppose, however, that it is not likely to be. Why should a disposition to believe a false view or to form a hostile desire matter if it is outweighed by perhaps admirable opposing dispositions? It might matter. There are certain precipices we should avoid approaching even if we have generally reliable safety procedures meant to keep us from falling.
Quite apart from how readily we can prevent our dispositions to believe from yielding beliefs that we rationally should not hold, the extent and stability of our intellectual virtue is still affected by our dispositions to believe. If, for instance, one has certain prejudices regarding alcoholic consumption and realizes, or should realize, that one is disposed to believe someone to be an alcoholic upon seeing the person take a third glass of wine at dinner, one should bring to bear background knowledge about such things as operative conventions, alcohol tolerance, and the individual in question. This disposition to believe could be an instance of what is now widely called implicit bias.Footnote 26 There are of course more reprehensible prejudices than this, as well as more seriously faulty tendencies toward poor reasoning. Intellectually virtuous people are self-critical concerning such matters as their views of people, their moral standards, their uses of evidence, and more. The point here is crucial for understanding epistemic virtue: being self-critical requires more than appraising what one actually believes. It matters greatly what significant dispositions to believe we have, especially when they are strong or likely to yield belief-formation.
Some of our dispositions to believe are rather like our inclinations to intend; the latter, often grounded in long-standing desires, are important for moral virtue. Good will may predominate over immoral tendencies, but the most virtuous among us do not depend – regularly, at least – on this predominance. They have rooted out such tendencies if they have indeed harbored them. A similar point holds for dispositions to believe. We can become sensitive to the difference between good and bad information, between adequate and inadequate grounds for belief, and between good and bad inferences. If a measure of self-trust is needed for innovation, a measure of self-criticism is needed to avoid error. Dispositions to believe propositions one should resist believing are among the things one should try to eliminate. They are like hazards in one's home: we can be obligated to eliminate them even if caution will prevent their doing harm.
Our experience of the world provides us with a vast amount of information. Testimony is particularly important since it not only propositionally focuses the information attested to but also commonly provides much non-semantic information, say about intonation and facial expression, to which belief-formation may also be an appropriate response, depending on the perceptual circumstances of the relation between attester and recipient. As the case of testimony indicates, we naturally provide one another information in propositional form. But the prominence and naturalness of testimonially supplied information – particularly if we include, as Reid did, the testimony of the senses – must not be taken to support the idea that we receive information only propositionally or indeed retain it in memory only propositionally. If the overall position of this paper is correct, belief, action, and the information-responsiveness of mind can be conceived less intellectualistically than has been common among many philosophers. Perception is richly informative regarding the properties of its objects; imagination and introspection may be so as well regarding their objects. What I have argued for is that much information we receive from these and other sources and much information that we retain in memory yields dispositions to form beliefs, but does not – certainly need not – automatically produce the beliefs it disposes us to form. The conditions under which those dispositions are realized in belief-formation are various. Extensive empirical inquiry is needed for a full accounting. But the data accessible to philosophical inquiry clearly show that there is good reason not to ascribe beliefs in all the cases where we naturally form them without new information and by merely considering propositions concerning or expressing the information that, in some non-doxastic form, we already have. Agency in the intellectual realm, like everyday overt action, may be explicitly guided by beliefs, but it also benefits from a vast store of information that is not doxastically represented but, when the occasion calls for it, commonly yields belief-formation and is in any case available for formulation, learning, inference, and critical scrutiny. Given the importance of belief for action, its constitutive role in knowledge, and its centrality in intellectual virtue, we should form beliefs only under good conditions and ascribe them only when they, and not just inclinations to form them, are present.Footnote 27
