UNDERSTANDING, JUSTIFICATION AND THE A PRIORI

DAVID HUNTER UNDERSTANDING, JUSTIFICATION AND THE A PRIORI (Received in revised form 28 November 1995) What I wish to consider here is how understanding something is related to the justification of beliefs about what it means. Suppose, for instance, that S understands the name Clinton and has a justified belief that it names Clinton. How is S s understanding related to that belief s justification? Or suppose that S understands the sentence Clinton is President, or Jones assertive utterance of it, and has a justified belief that that sentence expresses the proposition that Clinton is President, or that Jones said that Clinton is President. How is S s understanding related to the justifications of these beliefs? My aim is to explore the following claim. (T) Understanding is an a priori source of immediate prima facie justification for beliefs about what things mean. If knowledge is justified true belief, then, according to (T), true beliefs so justified constitute a priori knowledge. I believe this claim promises to throw interesting new light on the epistemic character and potential of Mind. In order to assess whether (T) is even remotely plausible, several preliminary clarifications are in order. I will begin with some observations about understanding, and then proceed to a discussion of justification. I will conclude with a discussion of a priori justification. 1. The objects of understanding are things that have or express meaning or content, such things as sentences, words, and speech acts. 1 One can, for instance, understand the sentence Clinton is President, or the words is President, or Jones assertive utterance of that Philosophical Studies 87: 119 141, 1997. c 1997 Kluwer Academic Publishers. Printed in the Netherlands.

120 DAVID HUNTER sentence. To correctly understand something is, roughly speaking, to know what it means, or to grasp its meaning. But meanings, contents and propositions are not themselves understood. One reason for thinking this is that it is not clear what it might be to misunderstand, say, the proposition that Clinton is President, whereas it is clear what it is to misunderstand the sentence Clinton is President, or some one s assertive utterance of it. 2 States of understanding are to be distinguished from events of coming to understand. It is one kind of thing for S to understand a sentence, word or speech act, and quite another kind of thing for S to come to understand a sentence, word or speech act. For instance, for S to understand the name Clinton is for S to be in a state of a certain kind, that of understanding. But for S to come to understand the sentence Clinton is President is for something to happen to S; it is for S to enter a state of understanding. Likewise, believing that Clinton is President is a state, while to come to believe that Clinton is President is an event. Although coming to understanding something is an event, it is not an action. Coming to understand a sentence is not something one does, since one s understanding of a sentence is not, in the relevant sense, under voluntary control. One cannot, for instance, simply choose to understand the sentence Clinton is President as expressing the proposition that it is cold outside, if one already understands it as meaning that Clinton is President. And one cannot choose or decide how to understand Jones utterance of that sentence: one simply understands it, if at all, as expressing one thing or another. Nor can one simply decide to understand a word that one knows has no established meaning as, say, a name for Clinton. Of course, one can stipulate or decide that it is to be understood as a name for Clinton, that is, that it is to name Clinton. But once one decides this, then one is no longer free to understand it as one likes. The process of coming to understand something is, thus, independent of willful control. States of understanding can be evaluated as veridical or not. S might, for instance, understand the sentence Clinton is President as expressing the proposition that Nixon is President. In that case, S s understanding of that sentence is mistaken; it is a misunderstanding. Or again, if S understand Jones to have said that Clinton is President,

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 121 and if that is what Jones said, then S s state of understanding is correct or true. It is also possible for two subjects to understand something in the same way, or to have the same understanding of a thing. S and T might, for instance, both understand the name Clinton as a name for Bill Clinton. In such a case, although S s state of understanding is distinct from T s, what these states have in common is that each is true or veridical just in case the name Clinton names Clinton. States of understanding are not states of belief, since states of understanding and states of belief differ in several fundamental respects. 3 First, whereas states of belief admit of justification, states of understanding do not. One can, for instance, ask whether S s belief that Clinton is President is justified, or whether it rests on adequate grounds. And whether S s beliefs are adequately justified is relevant to evaluating S s rationality, since it is a sign of folly to have a large number of unjustified beliefs. But states of understanding do not admit of justification. One cannot, for instance, ask whether S s understanding of the sentence Clinton is President rests on adequate grounds, or whether S is justified in having understood Jones to have said that Clinton is President. And S s states of understanding are not relevant to evaluating whether S is rational. Similarly, states of perception do not admit of justification. One cannot ask with respect to one of S s perceptual states, whether it is justified or whether it rests on adequate grounds. And states of perception are not relevant in evaluating a subject s rationality. In this respect, states of understanding are like states of perception. States of understanding differ from states of belief in a second, and related, respect. Whereas states of belief are cognitive commitments, states of understanding are but grounds for forming, or for taking on, such commitments. For a subject can fail to believe that things are as she understands them to be, just as she can fail to believe that things are as she perceives them to be. Suppose, for instance, that S has some reason to think that her faculty of understanding has recently been malfunctioning. Perhaps she has been told by several friends that she has misunderstood them, even when they spoke clearly and slowly. And suppose that she hears Jones utterance, and correctly understands him to have said that Clinton is President. Suspecting that she has once again misunderstood, S might well not come to believe that that is what he said. Or suppose that T

122 DAVID HUNTER understands livid as a name for a shade of red, but believes that her understanding of colour words is unreliable. Then she might well reserve judgement, and not believe that livid names a shade of red. Believing that something has some meaning commits one in a way merely understanding it to have that meaning does not. States of understanding are grounds for belief, but are not states of belief: states of understanding and states of belief are in different epistemic categories. 2. Though I do not want to wed acceptance of (T) to a particular account of justification, having one in mind will provide a framework for further articulating (T). Consider, then, the following not unreasonable account of what suffices for a belief to be prima facie justified. (J) S s believing that P is prima facie justified if S s believing that P is based on adequate grounds. 4 What I propose to do is to discuss how (J) applies in the case of beliefs based on states of understanding. (J) concerns the state of being justified and not the activity of providing a justification. 5 That is, (J) concerns what it is for a subject s believing to be justified, and not what it is for a subject to provide a justification for her belief. (J) allows that a subject s believing may be justified even if she lacks beliefs about that justification, and even if she is unable to formulate or recognise that justification. This allowance is reasonable since subjects can have justified perceptual beliefs in the absence of any belief, not to mention in the absence of any justified belief, about what grounds her perceptual beliefs. So (J) allows that a subject s beliefs about what a word means or a speaker says may be justified even though she lacks the resources to formulate or recognise that justification. According to (J), a belief is justified only if it is based on grounds that are adequate. Plausibly, a belief s ground is justificatorily adequate if such grounds tends to produce a high ratio of true to false beliefs. Part of the attraction of this account of justificatory adequate is that it accords with ordinary judgements about which

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 123 grounds for belief are adequate. In particular, it accords with, and supports, the judgement that perceptual experience is a justificatorily adequate ground for belief. Two features of perception account for this. First, it is uncontroversial that perception is reliably veridical: the ratio of veridical to false perceptual states is high. And it is uncontroversial that reliance on perception in forming beliefs is also reliable: the ratio of true to false perceptual beliefs is high. So there is little question that reliance on perception is reliance on a process that is reliably linked to truth. Second, the reason for perception s reliability is becoming known. It derives from casual/explanatory links between perception and the subject matter of perception and perceptual belief. This feature of perception is important since it indicates how the processes involved in perception and in the formation of perceptual belief can malfunction, thereby indicating how, as I will discuss below, a perceptual belief s prima facie justification can be undermined and over-ridden. Understanding compares favourably with perception in these respects. First, it is plausible that understanding is reliably veridical: the ratio of correct to incorrect states of understanding is high. And forming beliefs on the basis of such states is also reliable. Furthermore, it is relatively well known how the processes involved in understanding can malfunction, and how to test for such a malfunction. Moreover, research by cognitive scientists, developmental psychologists and linguists promises to explain the reliability of the process of forming beliefs on the basis of understanding by linking the development of the cognitive capacities involved in producing states of understanding to the subject matter of these beliefs. The point is not that there is promise of a physicalistic or naturalistic theory of linguistic understanding. Even in the case of visual perception there is debate about whether current scientific research project are, or ought to be, naturalistic or physicalistic. 6 Rather, the point is that unlike, say, intuition and clairvoyance, perception and understanding are generally held to be non-mysterious phenomena whose cognitive natures are proving to be scientifically tractable. This difference might well remain even if the best scientific theories of perception and understanding appeal to reference and truth. (J) formulates a sufficient condition for a belief to be prima facie justified. A justification is prima facie if it may be defeated. In

124 DAVID HUNTER the absence of defeaters, a prima facie justification is also ultima facie; in particular, in the absence of defeaters, a belief based on understanding is justified. There are two ways a belief s justification may be defeated. First, a prima facie justification may be over-ridden. S s justification for believing that p is over-ridden just in case S s belief that not-p is justified. That is, a subject s belief will be unjustified even if it is based on prima facie adequate grounds if she has, or there is available, evidence that the belief is false. 7 So the prima facie justification of a belief based on understanding may be over-ridden if there is evidence that the belief is false. Suppose, for instance, that S understands Jones to have said that Clinton is President, and believes on that basis that that is what Jones said. So, according to (J), S s belief is prima facie justified, since it is based on adequate grounds. But suppose further that S has been told by a better located and reliable observer that Jones actually said that Nixon is President. Since this is evidence that S s belief is false, S s belief that Jones said that Clinton is President is not justified even though it is based on understanding. For the prima facie justification conferred on it by understanding is over-ridden by other evidence. Or suppose that S believes, on the basis of her understanding, that livid names a shade of red, but is told by a painter that livid actually names a shade of white. In this case, S s belief that livid names a shade of red is not justified even though it is based on understanding. Second, a prima facie justification may be undermined.s s justification for believing that p is undermined just in case S has a justified belief that the alleged justification for S s belief that p is not in fact justificatorily adequate. Suppose, for instance, that S believes on the basis of understanding that Jones said that Clinton is President. Her belief is thus prima facie justified. But suppose further that S has solid evidence that her auditory faculties have been malfunctioning, interfering with her ability to correctly understand utterances. This is not evidence that her belief is false, but it is evidence that the grounds for her belief are not reliable, and hence do not justify her belief that Jones said that Clinton is President. S s belief that Jones said that Clinton is President is thus not justified, even though it is based on understanding.

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 125 Finally, according to (T) understanding is a source of immediate prima facie justification. The distinction between immediate and mediate justification concerns the ground or basis of a belief. A belief is mediately prima facie justified if it is based on one or more beliefs. So, for instance, if S s justified belief that Clinton is in the Rose Garden is based on beliefs about his travel schedule, then that belief is mediately justified. A justified belief is immediately prima facie justified, on the other hand, if it is not based on any beliefs. So if S s belief that Jones said that Clinton is President is justified, and is based on S s understanding of Jones utterance, then that belief is immediately justified, since states of understanding are not states of belief. Some have argued, however, that a belief can only be mediately justified. 8 In particular, some have argued that a belief can not be justified by a state of perception alone. To my knowledge, no one has explicitly argued that beliefs cannot be justified by states of understanding alone. But,f with only small alterations, the considerations advanced in the case of perception can be extended to cover the case of understanding. Donald Davidson, for instance, offers the following considerations to show that beliefs can not be justified by states of experience alone. We have been trying to see [justification] this way: a person has all his beliefs about the world that is, all his beliefs. How can he tell if they are true, or apt to be true? Only, we have been assuming, by connecting his beliefs to the world, confronting certain of his beliefs with the deliverances of the senses one by one, or perhaps confronting the totality of his beliefs with the tribunal of experience. No such confrontation makes sense, for of course we can t get outside our skins to find out what is causing the internal happenings of which we are aware. Introducing intermediate steps or entities into the causal chain, like sensations or observations, serves only to make the epistemological problem more obvious. For if the intermediaries are merely causes, they don t justify the beliefs they cause, while if they deliver information they may be lying (Davidson 1986, 312). If Davidson s conclusion is correct and if states of understanding are not states of belief, then, contra (T), states of understanding do not justify beliefs about what things mean. But what precisely is Davidson s reason for denying the possibility of immediate justification? There appear to be two objections. One objection turns on the claim that states of perception are mere causes. It follows that states of perception cannot justify beliefs only if what causes a belief cannot also justify it. But, according to the account of justificatory

126 DAVID HUNTER adequacy sketched above, what causes a belief does confer at least a prima facie justification on it, if that cause is reliably linked to truth. So if states of perception and of understanding cause beliefs and are reliably linked to truth, then, according to that account, states of perception and of understanding confer prima facie justification on the beliefs they cause. And Davidson agrees that states of perception both cause belief and are reliably linked to truth. So why does he think that perceptions do not confer immediate prima facie justification on the beliefs they cause? His reason, I think, is expressed in the following passage: the problem is to see how the sensation justifies the belief. Of course if someone has the sensation of seeing a green light flashing, it is likely, under certain circumstances, that a green light is flashing. We can say this, since we know of his sensation, but he can t say it, since we are supposing he is justified without having to depend on believing he has the sensation (Davidson 1986, 311). Davidson s point, it seems, is that what causes a belief can justify it only if the believer has a belief about that cause. In particular, a sensation can justify a belief only if the believes he has that sensation. So, according to Davidson, a state of perception can play a justificatory role only in conjunction with a belief, and so only in a case of mediate justification. But unless some support is given for the claim that a sensation can play a justificatory role only in conjunction with a belief, this line of thought simply begs the question against the proponent of immediate justification. For to hold that a sensation can immediately justify a belief is simply to deny that claim. So, it seems, Davidson s first objection to the possibility of immediate justification simply begs the question. What about Davidson s second objection? It is that states of perception cannot immediately justify beliefs because they may be false. On its face, this objection assumes that perception must be infallible if it is to immediately justify belief. But Davidson does not explain why, and it is not obvious that a proponent of immediate justification need assume this. According to the account sketched above, for instance, there need only be a high ratio of true to false perceptions, for perception to immediately justify belief. So why does Davidson think otherwise? He sometimes formulates this second objection in a different way. In one place, he says that a proponent of immediate justification must explain why we should believe our sensations are reliable, that is,

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 127 why should we trust our senses? (Davidson 1986, 310). And in another he says that the difficulty of transmuting a cause into a reason plagues the anti-coherentist again if he tries to answer our second question: What justifies the belief that our senses do not systematically deceive us? (Davidson 1986, 311). These passages suggest that Davidson s second objection assumes that something justifies a belief only if the believer is justified in believing that that thing is justificatorily adequate. So, in particular, a state of understanding can justify a belief only if the believer has some reason to believe that his understanding is reliable or true. And, presumably, Davidson thinks that the fallibility of perception (understanding) makes this need even more pressing. But, again, this assumption simply begs the question against a proponent of immediate justification. For by assuming that a believer must know whether his sensations are reliable in order for his belief based on them to be justified, Davidson s second objection simply assumes that a sensation can justify a belief only in conjunction with another belief, and hence only in a case of mediate justification. Davidson s considerations thus provide no reason to believe that immediate justification is impossible. Davidson s opposition to the possibility of immediate justification rests, I suggest, on what Alston has called a level-confusion about justification (Alston 1980). More explicitly, it rests on confusing the question whether S belief that p is justified with the question whether S has a justified belief that his belief that p is justified. Evidence of a level-confusion is found throughout Davidson s discussion of immediate justification. Consider, for instance, how he formulates what he considers to be the central epistemological question: How can [a subject] tell if his beliefs are true, or apt to be true? This is the question, what justifies S s belief that his beliefs are true or apt to be true, and concerns whether a subject can know whether she knows anything. Though this is an important question, it is very different from the question, what justifies S s beliefs, which concerns whether S knows anything. And it is this latter question that is relevant to the issue whether it is possible for a belief to be immediately justified; that is, whether immediate knowledge is possible. Davidson s level confusion is abetted, I think, by conflating the state of being justified with the activity of giving a justification.

128 DAVID HUNTER Evidence of this is found in the first passage quoted above, when Davidson moves from the question whether sensation is reliably linkedtotruthtothequestionwhethera believercan say that sensation is reliably linked to truth. Only if one thinks that being justified requires being able to say what that justification is would this transition seem relevant. But it is one thing for a belief to be justified, and quite another for a believer to say or be able to say what that justification is. And only the former is relevant to the question whether a belief can be immediately justified. For a belief might be immediately justified, even if the believer must inevitably appeal to further beliefs in order to justify the claim that it is. I have been discussing how beliefs about what things mean can be justified on the basis of understanding. It is worth noting that beliefs about what things mean may be justified on other grounds. A belief about what something means might be acquired through the testimony of another. One might, for instance, ask a friend what the word livid names, and form a belief about its meaning on the basis of the friend s reply. Or a belief about what some word means might be acquired by studying how it is used by those who understand it. 9 And it is conceivable that the meaning of a word or speech act should be discovered by investigating the factors that make it mean what it does. A belief formed in any of these ways would not be based on understanding (alone), but might nevertheless be justified and even constitute knowledge. But these ways of acquiring beliefs about what things mean are not fundamental or primary. For the acquisition of testimonial evidence itself depends on understanding. And few subjects are familiar with the correct methodology for field linguistics. And nobody knows how a thing comes by its meaning. Since subjects have knowledge of meaning, such justifications for beliefs about what things mean must not be primary or fundamental. Rather, what is fundamental to the justification of beliefs about what words mean and speakers say is that understanding is reliably linked to truth. 3. I have tried to show how understanding is, like perception, a source of immediate prima facie justification. I now want to suggest that,

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 129 unlike perception, the justifications conferred by understanding are apriori. 10 I will argue that although perception is necessary to the proper functioning of understanding, it does not play a justificatory role. Its role, rather, is to make understanding possible. According to (T), understanding is an a priori source of prima facie immediate justifications. There is, unfortunately, no generally agreed upon account of a priori justification. But according to one tradition, whether there are a priori justifications depends on whether there are non-perceptual sources of justification. According to this tradition, a faculty is a source of priori justification if it is a nonperceptual source of justification. This suggests the following thesis. (AP) S is justified a priori in believing that p just in case S is justified in believing that p and the sources of S s justification are non-perceptual. 11 Two points about (AP) should be noted. First, (AP) concerns the source of a belief s justification, and not the source of the belief. That is, (AP) concerns the justification, not the acquisition, of belief. More explicitly, (AP) is consistent with the view that the acquisition of a belief that is justified a priori may depend in various ways on sense experience. I will discuss several ways this may be below. Second, (AP) concerns what it is for a prima facie justification to be a priori, and not what it is for a proposition or truth to be a priori. Consequently, (AP) allows that a belief may be justified both a priori and a posteriori. I will also return to this below. What is it for a source of justification to be non-perceptual? The following suggests itself. (NE) A source of justification is non-perceptual if it is a source of justificatorily adequate grounds that are neither states of perceptual belief nor states of perception. According to (NE) and (AP), if some faculty produces grounds for belief that are justificatorily adequate and that are neither states of belief nor states of perception, then it is an a priori source of justification. Given (NE) and (AP), then, understanding is an a priori source of prima facie justification. I have already argued that states of under-

130 DAVID HUNTER standing are justificatorily adequate grounds for belief. And states of understanding are not states of perception, since one does not see, hear or otherwise perceive what a thing means. The gap between hearing a speech act with understanding and hearing it without understanding is striking, and cannot be bridged merely by listening more intently: failure to understand what someone speaking an unfamiliar language is saying is not invariably a failure of perception. And, I contend, states of understanding are not states of belief. 12 Hence, it knowledge is justified true belief, then understanding is a source of a priori knowledge about what things mean. My aim in this section is to try to make this claim about the character of justifications conferred by understanding seem plausible by exploring it and by responding to objections. Some of the objections rely on a conception of a priori justification whose faults have recently been the subject of considerable investigation. Because this work is still relatively unknown, it is worth examining these faults in some detail. One objection to the view that beliefs about what things mean can be justified a priori turns on the thesis that only necessary beliefs can be justified a priori. But beliefs about what things mean are contingent. So, if the thesis is true, such beliefs cannot be justified a priori. Colin McGinn offers the following argument for the thesis. 13 (1) If a belief is justified a priori, then that justification is available in every possible world. (2) If a belief is justified a priori, then it is true. Hence, (N) If a belief is justified a priori, then it is necessarily true. Even if we suppose that the argument is valid, neither premise need be accepted. McGinn defends (1) with the following discussion. Suppose it knowable a priori that P. Now since the knowledge that P is grounded upon reasons one has independently of observation of (causal contact with) the world, one s evidential state is not contingent upon the vicissitudes of the world. And if one s evidential state does not thus depend upon the ways of the actual world, then one could be :::in qualitatively the same epistemic situation in any world: a priori evidence is constant across worlds, because available without observation of the specific properties of each world (McGinn 1975/1976, 205). McGinn s support for (1) relies on the claim that a priori evidence is available independently of experience. This is the point of his

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 131 remark that a priori evidence is constant across worlds. But the claim is ambiguous as between the following. (a) (b) The justificatory force of a priori evidence is independent of experience. The availability of a priori evidence is independent of experience. Plausibly, an advocate of priori justification is committed to (a). This commitment is, I think, respected in (AP). But (a) does not entail (1), since (1), but not (a), concerns the availability of a priori evidence. (a) is consistent with the view that a belief that is justified a priori in one world may lack that justification in some other world. For this same reason, (a) also does not entail (b). So McGinn s defense of (a) must turn on (b). McGinn s objection is important since linguistic understanding would be unavailable in the absence of perception. It is uncontroversial that perceptual experience is necessary for the development of the faculty of linguistic understanding. A child s linguistic faculty develops (in part) in response to perceptual linguistic stimulus. And it is uncontroversial that actual exercises of this faculty also depend on perception. The objects of understanding, words and speech acts, are themselves objects of visual and auditory perception. So the justification provided by understanding would be unavailable in the absence of perceptual experience. But it is not clear why McGinn assumes that an advocate of a priori justification must accept (b). For (b) concerns a justification s availability, not its character. And it is possible that perceptual experience should be necessary for the availability of certain kinds of non-perceptual justification. 14 For instance, it is possible that perceptual experience should be necessary for the acquisition or development, and even for the exercise, of a non-perceptual cognitive faculty. But such faculties might nonetheless confer a priori justification on the beliefs they produce if perception serves merely to make possible the faculty s operation. Perception need not serve as a source of evidence. So it is not clear that a proponent of a priori justification must accept (b). Moreover, this distinction between the roles experience may play is traditionally accepted by avocates of a priori knowledge. First,

132 DAVID HUNTER it is accepted by Kantians who hold that the proper functioning of the faculty of intuition depends on perceptual experience. For, on their view, intuitive evidence for geometrical beliefs is acquired only through percetpaul study of geometrical constructions. Such justification would be unavailable in the absence of perceptual experience. But, it is alleged, perception s role is merely that of making possible the operation of the non-perceptual faculty of intuition; perception is not operating as a source of evidence. Second, this distinction is presupposed by those who hold that arithmetical beliefs formed on the basis of calculation are justified a priori. 15 The abilities required for arithmetical calculation are acquired and honed only through perceptual interaction with linguistic symbols. And various perceptual capacities are employed in the exercise of these abilities. So such justifications would be unavailable in the absence of perceptual experience. But it is nonetheless commonly held that such justifications are not perceptual. Perception serves merely in the production of arithemetical and logical beliefs, and not as a source of justification. So (b) is not traditionally considered part of the view that there are a priori justification. The initial plausibility of (1) derives, I think, from the idea that beliefs that are justified a priori are, in McGinn s words, available without observation of the specific properties of each world. 16 The idea is that beliefs that are justified a priori are not observational: they are not acquired by observing how things contingently are, and for this reason must be available in every possible world. But this idea supports (1) only if perception is the only form of observation of contingent features of the world. Unless this assumption merely begs the question against a defender of the view that contingent beliefs can be justified a priori, it must be an open question whether it is possible that there should be a non-perceptual forms of such observation. In particular, it must be an open question whether understanding is a non-perceptual form of observation. Indeed, that understanding is a source of a priori observational prima facie evidence for contingent beliefs about what things mean strikes me as a powerful and compelling view. 17 What about premise (2) of McGinn s argument for the thesis that only necessary beliefs can be justified a priori? (2) If a belief is justified a priori, then it is true.

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 133 McGinn does not defend (2), but claims that it seems plausible (McGinn 1975/1976, 205). But, as Casullo has noted, (2) is puzzling since, though it is uncontroversial that knowledge entails truth, McGinn s concern is with justification not knowledge. 18 And, in general, justification does not entail truth. For false empirical beliefs can be justified. And it is not clear that a proponent of a priori justification need accept (2). For it is plausible that an arithmetical belief might be both false and yet justified a priori by calculation. So (2) should also be rejected. But (2) is closely related to a second thesis supporting the view that beliefs about what things mean cannot be justified a priori. It is that a belief is justified a priori only if it is unrevisable. But, as was discussed in section 1, beliefs about what things mean are revisable. So if the thesis is true, semantical beliefs cannot be justified a priori. It is worth noting that the account offered in section 2 of the justification conferred by understanding entails that semantical beliefs are revisable. For according to that account, the prima facie justification of a belief based on understanding can be undermined by evidence of malfunction in the faculty of linguistic understanding. Such evidence, even when misleading, might prompt the belief s revision. But such evidence is, in principle, always possible. All beliefs so justified are thus open to revision. So if a belief is justified a priori only if it unreviseable, then that account of justification is inconsistent with the view that understanding is a source of priori justification. But the thesis that a belief is justified a priori only if it is unreviseable should not be accepted. 19 For it has unacceptable consequences. In particular, as Casullo has noted, it entails that no false belief formed by a process that is self-correcting is justified a priori. 20 A belief forming process is self-correcting just in case it can justify abandoning false beliefs previously formed by it. That is, false beliefs formed by a selfcorrecting process are open to revision by subsequent operations of that very process. Logical deduction is a self-correcting process, since deduction can justify abandoning a false belief formed through deduction. So is arithmetical calculation, since repeating a complex calculation might reveal a mistake in the original calculation, prompting a revision in beliefs. But then, if beliefs justified a priori are unreviseable, such beliefs are not justified a priori. But, surely,

134 DAVID HUNTER false logical and arithmetical beliefs based on calculation and deduction are sometimes justified a priori. So the thesis that a belief is justified a priori only if it unrevisable should be rejected. More importantly, though, that thesis is orthogonal to the primary interest of the distinction between a priori and a posteriori justification. Interest in the distinction has traditionally concerned differences between perceptual faculties, on the one hand, and intellectual faculties, on the other, in hopes of revealing part of what is distinctive of the mental. Debate whether beliefs can be justified a priori has traditionally turned on the question whether there are non-perceptual sources of evidence or belief. But the thesis at issue concerns whether there are incorrigible sources of belief: sources whose products are not open to revision. And this concern surely cuts across the distinction between perceptual and non-perceptual sources of belief since it is possible that there should be an incorrigible, is important and interesting, it is not what has traditionally motivated proponents of the a priori. A third objection to the claim that beliefs about what things mean can be justified a priori turns on the thesis that a belief is justified a priori only if it cannot be defeated by empirical evidence. But, as was discussion in section 2, S s belief that Jones said that Clinton is President can be defeated by empirical evidence. So, if the thesis is true, semantical beliefs cannot be justified a priori. In order to assess this thesis, it is important to distinguish two versions of the thesis that empirical evidence cannot defeat a belief that is justified apriori. (i) (ii) If a belief is justified a priori, then no empirical evidence can over-ride its justification. If a belief is justified a priori, then no empirical evidence can undermine its justification. Neither version of the thesis should be accepted. Consider, first, (i). To begin with, it is not clear that a proponent of a priori justification need accept it. The view that beliefs can be justified a priori is a view about the available kinds of justification. As was noted above, there is no inconsistency in holding that a given belief can admit of both a posteriori and a priori justification. But if S s belief that p can be justified a posteriori, then so can S s belief

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 135 that not-p. So there is no inconsistency in holding that a given belief can be justified a priori even though its revision can be justified a posteriori. Consequently, a defender of the a priori need not accept (i). Moreover, as Casullo has noted, (i) has implausible consequences. 21 Consider the case of a subject s introspective beliefs about her own psychological states and events. Many have held that such beliefs are justified a priori. But it is also widely accepted that such beliefs can be justified empirically by behavioural evidence. And it is conceivable that some such beliefs can be justified by evidence about the subject s brain states. So evidence about a subject s behaviour or brain states might justify her in revising her beliefs about her own psychological states. But it is unreasonable to conclude from this alone that beliefs based on introspection are a posteriori. Whether introspection is an a priori source of knowledge about a subject s psychological life does not depend on whether that knowledge might also be acquired in another way. We have already considered reason for rejecting (ii). Many hold that calculation and deduction confer prima facie a priori justification on arithmetical and logical beliefs. But deduction and calculation depend on the proper functioning of perceptual faculties. Evidence that these faculties have malfunctioned in a given case would undermine the prima facie justification conferred in that case. Indeed, such prima facie justification may be undermined even by misleading evidence of a malfunction. But such evidence might well be empirical. So, according to (ii), calculation and deduction do not confer prima facie a priori justification. But this is, I take it, unacceptable. So (ii) should also be rejected. A fourth objection to the view that beliefs based on understanding are justified a priori turns on the fact that the objects of these beliefs are objects of perception. Words and speech acts are objects of visual and auditory perception, and it is plausible that perception is our only access to these objects. What is more, the objector might continue, this is an important difference between beliefs about what things mean and arithmetical beliefs. For even if the acquisition of arithmetical beliefs depends on perception in the ways discussed above, these beliefs are not about objects of perception. Rather, they

136 DAVID HUNTER are about abstract objects. But, surely, a belief about an object of perception can not be justified a priori. Though compelling, this line of objection confuses the role perceptual experience plays in the acquisition of a belief with the role it plays in that belief s justification. 22 I agree that without perceptual interaction with word tokens and with the speech acts of others, subjects would not acquire beliefs about these words and speech acts. One must see the words on the page, or rely on the testimony of someone who has seen them, in order to form beliefs about them. This is so, not only for beliefs about what they mean, but also for beliefs about their physical properties. So, in particular, one could not acquire the belief that Clinton names Clinton or that Jones said that Bill Clinton is President in the absence of perceptual interaction with Jones. But our concern is with justification, not acquisition. Beliefs about a word s physical properties are plainly justified perceptually. One can, for instance, see that a given word is printed in black, or hear that Jones spoke loudly. Such beliefs are not justified a priori, since their justification is perceptual: it derives from the subject matter s visual or auditory impression. However, one cannot tell merely by looking whether a word as used by Jones names Bill Clinton, or merely through hearing whether a speaker has referred to Bill Clinton. A word s or speech act s semantic properties are not intrinsic to it, and do not supervene on its observable physical properties. These properties are not open to perceptual observation in the way some of its physical properties are. The meaning properties of words and speech acts are not seen or heard, they are understood: understanding provides access to features of the world not (ordinarily) available through perception. But one might note that empirical evidence can over-ride the justification for a belief about a word s or speech act s meaning by revealing that the object of understanding has been mis-identified. Suppose that S believes that Jones said that Bill Clinton is President but that it was Smith, and not Jones, who said that Bill Clinton is President. In such a case, the mistake derives, not, one might claim, from mis-understanding, but from empirical mis-identification. And one might conclude from this that the justification for S s belief that

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 137 Jones said that Bill Clinton is President must also be partly empirical. Two points should be noted in response. First, this objection relies on the principle that a belief is justified a priori only if its justification cannot be over-ridden empirically. And, as was discussed above, this principle is unacceptable. The point is not that beliefs about what someone said cannot be justified empirically, by reference to facts about the speaker s identity. They can. The point is that they need not be. Second, that empirical evidence can reveal that the object of understanding has been misidentified shows at most that beliefs about what is said are partly justified perceptually. It does not follow that such beliefs are not also partly justified a priori. This point can be accommodated in either of two ways. One might represent S s belief as consisting of two component beliefs: a belief that Jones said something, and a belief that what was said is that Clinton is President. Alternately, one might represent S as believing, of Jones, that he said that Clinton is President. In both cases, the aim is to isolate a belief that can be justified wholly on the basis of understanding. Isolating such a belief faces familiar difficulties. But it seems to me that achieving this aim is less important than recognizing that understanding is a non-perceptual source of evidence for beliefs about what things mean. 23 Finally, one might suspect that the use of demonstratives shows that in some cases understanding what is said does depend on perceptual evidence. 24 Tyler Burge, for instance, correctly notes that understanding what is said in using a demonstrative may require perceiving the demonstrative s referent, and concludes that the justification of a subject s beliefs based on such an understanding is partly perceptual (Burge 1993, 480 n.19). This issue is very complex, and a complete treatment is out of the question here. But it strikes me that perception s role in understanding domonstratives is less clear than it might initially seem. Suppose that Jones utters the sentence That man is President, pointing to Bill Clinton, and that he thereby says that Bill Clinton is President. And suppose that understanding Jones utterance requires both seeing his act of pointing, and seeing Bill Clinton. 25 But two points should be noted. First, as things stand perception s role in understanding Jones utterance is not appreciably different from its role in understanding

138 DAVID HUNTER words and non-demonstrative speech acts. Compare, for instance, perceiving the name Bill Clinton in understanding the sentence Bill Clinton is President and seeing Bill Clinton in understanding Jones utterance. In the former case, although understanding the sentence depends on perceiving the word, this does not entail that perception plays an evidentiary role. In normal cases, perception s role is understanding is, I content, analogous to its role in arithmetical calculation. Visual and auditory perception make understanding possible, but do not serve as a source of evidence or justification. So what reason is there to think that seeing Bill Clinton and seeing Jones pointing do play a justificatory role in the production of understanding? The second point concerns what is not perceived in seeing the demonstration and its referent. One does not see that the speaker has referred to Bill Clinton or that Bill Clinton is the referent of the demonstration, anymore than one sees that a use of the name Bill Clinton refers to Bill Clinton. When the name is used, the reference to Bill Clinton is understood, not perceived. As Wittgenstein pointed out, demonstrations can also be misunderstood (Wittgenstein 1953, x28). One can misunderstand someone by misunderstanding their use of a demonstrative, or the accompanying demonstration. Understanding Jones utterance requires understanding his pointing as a reference to Bill Clinton, just as understanding a use of the name Bill Clinton requires understanding it as a name for Bill Clinton. Thus, while it is uncontroversial that perception plays a role in understanding speech acts involving demonstratives, it does not obviously follow that this role is justificatory. Let me conclude by summarizing my discussion. I have tried to make it plausible that understanding is an a priori source of immediate prima facie justification for beliefs about what things mean. In learning to understand, subjects gain access to features of things not normally accessible through perception. And, I have argued, the justifications conferred by understanding are a priori. I believe that this conception of understanding promises to reveal part of what is special and distinctive about the nature of Mind. 26

UNDERSTANDING, JUSTIFICATION AND THE A PRIORI 139 NOTES 1 The following claims are defended in my paper Understanding and Belief, ms. 2 That propositions are not objects of understanding does not entail that propositions are not contents of states of understanding. For states of perception may have propositional content even if propositions are not objects of perception. 3 I discuss this in more detail in my Understanding and Belief, ms. 4 Strictly, (J) should include a time restriction, since a subject s believingd that P might be justified at one time but not at another. For simplicity, I will leave this out. My discussion of (J) is indebted to the discussion in Alston (1985). 5 This point is insightfully discussed in Alston (1980). 6 See Burge (1986). 7 What counts as available evidence is plainly vague. But this mirrors vagueness concerning when a subject s ignorance precludes her belief s being justified. Some amount of ignorance is typically allowed, but there is no obvious line to draw. 8 For a fuller discussion of this, see Alston (1983). 9 This view is proposed in Quine (1960), and Davidson (1973). 10 Burge claims that beliefs based on understanding are justified a priori, but does not elaborate or defend this claim (Burge 1993, 458). 11 My discussion of the a priori is indebted to the excellent writings of Albert Casullo; see, especially Casullo (1988, 1995a, b). 12 For more on this, see my Understanding and Belief, ms. 13 McGinn (1975, 1976, 205); see also Kitcher (1983, 29 30). 14 This criticism is articulated by Albert Casullo, in Casullo (1988, 202 3). 15 This was stressed by Frege; see Frege (1884). 16 See also Kitcher (1983, 29 30) and Kripke (1980, 38). 17 Perhaps this view about observation is what prompts McDowell to remark that understanding of a language makes available to ours senses :::facts about what people are saying (McDowell 1981, 241). 18 Casullo (1988, 208). 19 That thesis is explicitly rejected by Pollock (1974, 320); and by Burge (1993, 461). But it is most insightfully discussed in Casullo (1988, 1995a). The notion of a self-correcting belief forming process is developed by Casullo (1988). 20 Casullo (1988, 291). 21 Casullo (1988). 22 One might respond to this objection by denying that the objects of understanding are objects of perception. Since I am not sure how to develop this response, I will set it aside here. 23 I am indebted to the reference for helpful comments on this and other points. 24 The use of indexicals and non-literal uses of language constitute further complexities for an account of understanding. Normally, uses of indexicals are understood without reflection. On my view, which I cannot defend here, this indicates that contextual factors that determine what is said or expressed normally also determine the contents of understanding. To the extent that non-literal uses are parasitic on literal uses they can be treated as exceptional. It is uncontroversial that reflection on speakers intentions and on contextual factors can assist in coming to know what is said. But this assistance should not be exaggerated or given a