O SET THE STAGE for the discussion[1], I will rehearse and clarify a
well-known dispute between A. J. Ayer and J. L. Austin concerning whether
perceptual judgments are inferences. Both in his Sense and Sensibilia[2] and in his "Other Minds,"[3] Austin carefully distinguishes
recognizing that p from inferring that p. For the purpose of
comparing his position to Ayer's, we might put his basic claim in this way:
given the way words such as "recognize" and "infer" are used outside
philosophical discussions, one clearly distinguishes instances of recognizing
from instances of inferring. Yet Ayer does not dispute that, but replies that
while non-philosophers do make a sharp distinction between the two, it is
arbitrary for philosophical purposes.[4] Claims based upon one's having
recognized something are sufficiently like claims based upon one's having
inferred, Ayer supposes, that it is useful to treat them as instances of a
common category. So the issue is not whether the distinction is recognized
outside philosophical circles, but whether it is a defensible and useful one to
make. Clearly, Austin insists upon the distinction because he supposes that
failing to make it will promote philosophical confusion; indeed, he argues that
one traditional problem of skepticism is largely due to this confusion.[5] In his "Other Minds," Austin tries
to suggest how recognizing differs from inferring by showing how the sorts of
questions or challenges brought to bear differ between the two sorts of
claim:[6] for inferences, one wants
a rehearsal of the pieces of evidence and an account of their connections to
the judgment; for perceptual claims of recognition, one explores whether the
observer had the opportunity to see what he claimed to have seen, whether he
had acquired the expertise to recognize the sort of thing he claimed to have
seen, and whether the circumstances were free of evident distraction and
defect. But his readers' appreciation of these things depends upon their having
granted already that the distinction is not arbitrary: since Austin expects us
to recognize that the distinction is justifiable, he does not defend it
explicitly. After we have compared the two sorts of claim, we will examine a
defense of the distinction which accords with Austin's discussion.When we turn to an examination of claims based upon inference and ones made as perceptual judgments of recognition, we discover a number of important similarities which must be addressed by any account which insists nevertheless upon the importance of distinguishing between them:
1. When we explain our perceptual mistakes, we often do so in a way which makes what we did look like an inference. For we like to articulate characters in common between what we (mistakenly) thought we saw and what we did (in fact) see which led us to mistake one for the other. And that makes it appear that we reached our (mistaken) judgment through a consideration of those common features.
2. When we explain how we, unlike others present, have been able to see what we did see, we again talk about our judgment as though it were a species of inference. For we point to the one or several characteristics which enabled us to see what we saw.
3. In general, for any perceptual judgment which one makes, there is an inference which might have been made in those circumstances and which "mimics" or "emulates" the perceptual judgment. Because of this, if I recognize something which you don't size up, I may point out characteristics which will enable you to infer that what I claim is true. This fact about perceptual judgments is what enables a seasoned veteran to teach a beginner to recognize what he can recognize. Imagine teaching a child to recognize oak trees: one will first point out the features characteristic of oaks--the patterns of their leaves, the way they branch, features of their bark, their usual habitat--in short, the sorts of thing from which one's pupil may infer that he is seeing an oak. Eventually, from such training, the pupil may learn to recognize certain trees as oaks. So a perceptual judgment looks like a very rapid, unexpressed inference. In addition, gaining new information can expand one's ability to recognize or see things, just as it can similarly expand the number of inferences one can make.
4. One's claims about what he sees often incorporate non-perceptual beliefs or things he takes for granted, just as do his inferences. As Dretske has shown,[7] if you ask me whether the water you put on the stove is boiling, I may be able to see that it is even though I cannot see that it is water which is boiling: I take it for granted that the stuff in the pot is water. And if I believe that it is my car which I was seeing through the window, I can properly claim to have seen a red-haired man steal my car. Because it is clear that at least part of such a perceptual judgment requires an inference (an act of theft presenting no unique visual patterns), it is tempting to suppose that the entire judgment must consist of inferences.
5. As an empirical matter of (causal) fact, whenever one recognizes X as a Y or this as an X, he must be enabled to do so in virtue of certain visible or otherwise evident features of the thing he sees or perceives. Even when the observer cannot identify those features which enabled him to perceive what he has perceived, nevertheless there must have been evident features which were causally necessary for his perceptual judgment. If I really do recognize my father as he comes down the street, then had certain of his features been obscured by makeup or obstacles, I would not have recognized him. And so it looks as though whenever one makes a perceptual judgment, there are visual (or sensory) "cues" which will serve as the "data" for a reconstruction of the perceptual judgment as an inference. For clearly, in some sense one must have been aware of those "cues," even if one is unable to articulate them later.
Compared to these impressive sorts of similarity between inferences and perceptual judgments, the differences between them look at first to be fairly subtle. In the first place, one can be enabled by virtue of certain characteristics visibly displayed by X to recognize it as a Y without deliberately or consciously having taken notice of those characteristics. While the presence or availability of those "cues" may be causally necessary to one's seeing what he sees, the observer may nevertheless be wholly unable to list them in any helpful way; whereas, in the case of an inference, he must be able to list or advert explicitly to the data he has used (or, if he has already forgotten them, he must earlier have gone through such a list). One might, of course, try to patch up the comparison by invoking unconsciously used data or implicit reasons, after the manner of Harman,[8] Pitcher,[9] and perhaps Pollock,[10] but that risks begging the question at issue. (In his Foundations of Empirical Knowledge, (Macmillan & Co Ltd: 1958), A. J. Ayer says that "I simply take it for granted that this is a pen," since "our belief in the existence of such things are inductive, in the sense that they imply more than would be implied by a mere description of the experiences on which they are based," adding, "they often do not involve any conscious process of inference." (page 60) And later, Ayer speaks of an "unreflecting assumption of the existence of a material thing." (page 237) ) Secondly, not only may a person fail to mention the cues which arguably he used in his perceptual recognition, often--when questioned about them specifically--one misdescribes those characteristics. Thus, as is perfectly familiar to us, one's claim to have recognized a friend at a party may be far more reliable than one's response to questions about how the friend looked or what he was wearing. Indeed, it is a crucial feature of one's ability to recognize someone or to see that something is the case, that the reliability of his perceptual judgment outstrips his ability to say accurately how he was enabled to see what he saw.[11] Sometimes, the perceptual judgment in question seems too simple for its relevant "cues" to be specified: how indeed is one enabled to see that the wall is yellow, if seeing is really inferring? Sometimes, the difficulty is that we do not really have a classification scheme of the "cues" which is as finely graded as the things we are enabled to perceive: one can often tell, from the way the kitchen smells, what is cooking, though one would be unable to describe the odor except as "the odor of burnt toast" or "the odor of roast beef." And sometimes, the difficulty seems to be the enormous complexity behind the judgment. Frequently a person acquires the ability to tell that his spouse or best friend is angry without being able to specify what, in those circumstances, enabled him to tell; and even others who observe him in action may be hard pressed to specify the operative "cues." Often a parent can recognize his child at great distances, or from unusual perspectives, or from a largely obscured view, in a way which defies specific explanation. And that is exactly the feature of an exercise of a perceptual ability which makes it distinct from an inference: its reliability outstrips any set of inferences which hopes to "emulate" or "replicate" the whole set of perceptual judgments which are exercises of that perceptual ability.[12]
No doubt when one, say, recognizes his child from a partial view, there are causal "cues" which serve as necessary conditions; the problem in one's or someone else's specifying these cues arises largely from their complexity and redundancy. Imagine that any of numerous sets of facial features will enable a parent to recognize his child; and suppose that several sets of these features are available to the parent in the current circumstances; then he may be able correctly and reliably to recognize this as his child without being able to identify that set of facial characteristics which enabled him to do so.[13] In fact, because of such redundancy, the parent may falsely suppose that he was able to recognize his child by one characteristic whereas in fact, it was some set of characteristics not including that feature which was the "but-for" causal condition of his recognition. In this sort of instance, the parent's perceptual judgment ("that is my child") is far more reliable than any attempt by him to infer the conclusion from explicitly noted data: for an inference is only so good as its data, and in this instance, the data which the parent is able to articulate do not conclusively support the judgment that he is seeing his child. Moreover, there is no guarantee that an independent investigator could do better: he might confirm the reliability of the parent's repeated recognitions of his child over time and varying circumstances, but be unable to offer sets of causally sufficient "cues" which would with equal reliability have supported inferences "replicating" the parent's judgments.
This should not be surprising. When certain "cues" are the causally sufficient conditions for a person's being able to recognize someone, then for the observer to bring off his recognition, it is not necessary that he be able to articulate or advert to the "cues;" it is necessary only that they be in fact displayed to him.[14] Unlike an inference, the credibility of a perceptual judgment of recognition does not depend upon one's beliefs about how he was able to arrive at his judgment; all that is required is that he reliably make the judgment upon the presentation of causally necessary conditions--conditions which may go unidentified. Similarly, there are additional conditions which are causally necessary for one to see that this is his child: one's eyes, optic nerves, and brain centers must be in certain states within certain tolerances; and beyond such conditions of the observer, the environment must either possess certain characteristics or anyway lack certain other characteristics. But once again, in order to observe that something is the case, a person need not advert to or use beliefs about these conditions: it is sufficient that the causal conditions be satisfied; the person need not know that they are.
Thus, although some perceptual judgments may be "reconstructed" fairly accurately as inferences, not all can be so "reconstructed." For one thing, the use of explicit inferences presupposes the ability to make other judgments non-inferentially. This "ultimate" dependency of rule-following judgments upon "intuitive" judgments is at the bottom of the common sense view that inferences rest upon observation: if one cannot, e.g., simply see what the data are, then he will not be able to infer things from them. And for another, we may be far more confident about a perceptual judgment's reliability than about any identification of its "implicit data." Just as inferences are liable to go wrong, so one's exercise of perceptual abilities is fallible. To be sure, sometimes what makes a perceptual judgment "go wrong" may be analogous to the corresponding flaw in an inference: a background belief may have been false, or some of the crucial "cues" may have been obscured, or sometimes one may have been incautious or hasty. But sometimes, unlike instances of inferences, no one can say why, on that particular occasion, a person's ability to recognize things such as this failed him.
We are now in a position to understand why, as Austin points out in his "Other Minds," challenges and questions pertaining to a perceptual claim of recognition follow a different pattern from those pertaining to an inference. If one wants to challenge or gauge the credibility of an inference, he needs to look at the data used and at the grounds which related them to the sort of conclusions drawn. But if one wants to challenge or scrutinize a perceptual judgment of recognition, one likely has no explicit record to consider. And since the reliability of the perceptual judgment may outstrip that of any inference suggested as its "replication," examining a "replication" may not be an accurate way of evaluating the credibility of the perceptual judgment. Instead, one must look to the past performances of the perceiver--to see whether it is likely that he has developed the ability which he claims to have used on this occasion, and to see whether this occasion was the sort to have permitted its reliable exercise. (In a closely analogous way, when we want to determine whether Jones' tennis playing on this occasion was skillful or merely lucky, we look to whether he has had the opportunity to develop such skill and whether the present circumstances were propitious for its exercise; we do not demand that Jones cite a set of explicit rules for playing tennis, since that will not in general be a reliable indicator of what we wish to assess.) After all, when we question a person who claims perceptual knowledge, we want to find out whether the person probably did see what he supposed he recognized, not whether he would have been able to infer what he saw from some explicit set of data.
Thus we find Austin's insistence on the distinction between recognizing and inferring to be fully justified. As a result, Austin is supported in his further claim that certain sorts of fact about the perceiver and about the thing perceived need not be substantiated by the perceiver in order correctly to be said to have seen what he supposed himself to have seen. Although these items would be part of the data of an inference which "emulates" the perceptual judgment, the perceiver's grounds for believing them are not relevant to whether he saw what he said he saw: he did not claim to have inferred, but to have recognized. Thus, the distinction between recognizing and inferring forestalls one traditional entrance of the problem of skepticism. For if seeing really were the making of an inference, then the observer would certainly have to include in his grounds for his perceptual judgment some details concerning these "cues;" and he would also need to have subsidiary evidence tending to substantiate such details. And since one scarcely ever has relevant and conclusive evidence about such things, traditional skeptical problems would consequently assert themselves and force the denial that one ever really does see such a thing as that X was at the party. But since recognizing is not the making of an inference, these details are not data which the observer needs to have adverted to and made sure of; instead, they are causal conditions which simply must have been satisfied--independently of what the observer knew or reasonably believed. Indeed, the observer can make an inference which goes quite in the opposite direction: realizing that he has seen X at the party, and realizing that doing so required a host of causal conditions to have been satisfied, he is able to infer that these conditions must have been satisfied.