My interest in the content of experience and in the issue of whether experiences as such can justify beliefs has many sources. At present, it stems from the fact that in a number of writings I have been defending what I call the “moderate” conception of the architecture of empirical warrants. According to such a view, in order to possess a warrant for a given empirical belief such as “Here is my hand” it is not enough merely to have a hand-like experience. It is also necessary to make some collateral assumptions such as “There is an external world”, “My sense organs are generally reliable”, “I am not the victim of a lucid and sustained dream”, and possibly some more. My views thus engage directly both with those of dogmatists, such as Jim Pryor, and of conservatives, like Crispin Wright (Wright Reference Wright1985, Reference Wright2004; Pryor Reference Pryor2000).Footnote 1
While, within that kind of debate, there is much discussion about the role of collateral assumptions within the structure of perceptually-based warrants – whether they are needed or not and whether they need be independently justified or not – very little is usually said about perception itself and what conditions it must fulfill in order to play a justificatory role. Still, an account of that is owed too, within the project of explaining how perceptual warrants are constituted. The present paper will deal with that issue. Let me point out from the start, however, that the moderate architecture of empirical warrant could be retained even if one disagreed with the way in which I will account for the constraints perceptions have to meet in order to play a justificatory role, once the relevant assumptions are propositionally in play.Footnote 2
1. AT THE ORIGINS OF THE DEBATE BETWEEN CONCEPTUAL AND NONCONCEPTUAL THEORISTS
A good entry point to start tackling the issue is to briefly consider John McDowell's rendition, in his Mind and World (McDowell Reference McDowell1994), of Wilfrid Sellars' attack against the “Myth of the Given” (Sellars Reference Sellars1956). Very roughly, the idea is that if we think of experience as devoid of any structure and therefore of a representational content with correctness conditions, like sense-data theorists would do, or even those who think that experiences, as such, have only a phenomenal content, then our perceptual experiences could not justify those empirical beliefs which are based on them. For, how could an experience devoid of any representational structure support the truth of a specific empirical proposition, such as “Here is a red table”, which is a structured entity that represents the world correctly or incorrectly? Moreover, such an experience would be compatible with a multiplicity of possible conceptually structured propositions, not all of which would be relevant for the specific belief which we would like to justify by means of it. For instance, it would be compatible with “There is a red rectangular-shaped object here”; but also with “There is a white rectangular-shaped object bathed in red light”; or with “There is a hologram of a red rectangular-shaped object”, or even with “There are two juxtaposed triangular red-objects here” and so on and so forth.
If “the Given” is an unstructured datum, it cannot play a justificatory role with respect to our beliefs about objects in our surroundings. However, if one endorses the idea that only beliefs can justify other beliefs, in a coherentist fashion,Footnote 3 because they are the only entities, which possess the right kind of structure to play such a role, we incur other problems. First, if beliefs are never justified by experience, although they are caused by it, then none of our beliefs will ever be justified by how things are in the world; contrary to our deep-seated intuition that at least when perceptually based beliefs are at stake this would be the case. Rather, beliefs will always be justified thanks to other beliefs, as long as they comply with some coherentist principles. Yet there can be maximally coherent systems, which, however, are very different from each other. How should we choose among them? Which one would give us the (largely) correct account of reality?
The second problem raised by a coherentist response to the Myth of the Given is the one that perhaps is dearest to McDowell. Namely, the Kantian issue of how our concepts can have empirical content if they are merely caused by experience. If the experience has just a causal role in the formation and application of our concepts, then nothing precludes the following inverted-spectrum possibility from obtaining. There could be people that, in front of red objects, have a green experience that they conceptualize as red, while others, in front of the very same objects, have a red experience, which they conceptualize as red. If we grant that both of them have the concept red,Footnote 4 it follows that that concept is not individuated, at least in part, by how things appear to one. For the different experience in the two cases would not lead to two different concepts. Thus, concepts would not have an empirical content after all. Indeed, they would turn out to be just abstract rules that a cognitive system could implement even if it did not have any color experience, as long as it were able to categorize all and only red objects on the basis of whatever kind of information, which may well have nothing to do with how that color looks to us.
In order to solve both these problems, as is well known, McDowell puts forward the view that the content of perception is conceptual. This means that in experiencing the world by means of our senses we passively deploy our conceptual abilities. Hence, on the one hand, our concepts are endowed – one would like to say replete – with empirical content. On the other, perceptual content is structured; indeed it has propositional structure, and can thus serve as a justification for the beliefs formed by taking it at face value. Moreover, according to McDowell, in perception we take in facts – the world itself – yet in an already conceptualized way. Accordingly, we are in touch directly with the world, yet we take it in in such a way that it is immediately suitable to justify our empirical beliefs, because it is already structured and conceptualized just as those propositions we go on to believe if there are no reasons to think that our sense organs might have somehow deceived us.Footnote 5
Here I will not take issue with the idea that in perception we take in facts and with McDowell's ensuing disjunctivism, to the effect that whenever our experience is not veridical, it is not an instance of the psychological type “perception”. I think this view is extremely problematical. Yet a discussion of it would take us too far afield.Footnote 6 Let me just say that, in order for experiences to be appropriately structured so as to constitute a warrant for specific empirical beliefs (by courtesy of some very general assumptions, according to moderatism, whether or not a subject is able to entertain them), there is no need to think of them as ways of taking in facts. It is enough that they be structured to match the propositions believed on their basis. The epistemological question to be addressed is: can they have such a structure only if they involve the passive exercise of concepts? Before considering it in more detail, however, let me say a few words about another issue raised by McDowell's position. Namely, the metaphysical problem of whether only creatures endowed with the relevant conceptual repertoire can have a perceptual representation of the world around them.
2. THE METAPHYSICAL PROBLEM
As is well known, McDowell has been fiercely criticized for holding a view that, on the face of it, seems to imply that only human adults would perceive their surroundings. Indeed this seems to run contrary to both common sense and scientific inquiry, which concur in granting perceptions to infants and even animals who, arguably, lack the conceptual repertoire needed, in McDowell's account of the content of perception, to have those very perceptual representations.Footnote 7
Furthermore, he has been criticized on the score that his theory would provide circular possession conditions for observational concepts, because one would need to have the concept red already in order to perceive something as red. This would pose insurmountable problems for those who, like McDowell, think that our concepts are acquired and are not innate. For how could one offer a psychologically plausible account of how humans can acquire the concept red, say, if subjects must already possess it, in order to be able to perceive red objects as red and thus have the kind of perceptual experience on which their acquisition of that concept should, allegedly, be based?Footnote 8
Another canonical objection is the one relative to the finesse of grain of perception that exceeds our color vocabulary, for instance. As I have maintained elsewhere (Coliva Reference Coliva2003; cf. also Brewer Reference Brewer1999; Kelly Reference Kelly2001), although it is true that color perception is finer grained than our color lexicon, we can form demonstrative concepts, such as that shade of red, which, as short-lived as they might be, since we may be unable to recognize that shade on future occasion, may well be operative while perceiving the shade in question and while discriminating it from other ones in one's perceptual scene. Yet, in my view, this is no comfort to the metaphysical thesis McDowell is committed to, regarding perception. Namely, that only creatures with the relevant conceptual repertoire can have the corresponding kinds of perceptions. For all the argument from the finesse of grain shows, when supported by the considerations just advanced, is that we can have fine-grained concepts, which match the content of our color experience. Yet, it does not establish that only creatures with those concepts can have those experiences. Hence, I think it is useful to distinguish two theses:
Dependence Thesis: a subject can have a perception with the representational content that P if and only if he has the concepts necessary to characterize it in a canonical way.
Immediate Conceptualization Thesis: for those who possess the relevant concepts, an experience with content that P is immediately conceptualized – i.e. independently of judgment – by means of the concepts necessary to characterize its content in a canonical way.
The argument from the finer-grained content of experience is silent on the first thesis. It is therefore powerless to adjudicate the issue between conceptual and non-conceptual theorists, which hinges on precisely that point. Similarly, the argument that conceptualists have often invoked in defense of their position, namely the one from the justification of empirical beliefs, according to which justificatory relations hold only between conceptual contents, and therefore perceptual experiences can justify empirical beliefs, as such, only if they have a conceptual content, is equally silent on the first thesis. For the second thesis would be enough to account for the conceptualist intuition that only conceptualized experiences could justify the corresponding beliefs.
We can therefore conclude that there are no arguments in favor of the first thesis, while there are actually two against it. Namely, the argument from the perception of infants and animals and the one from circularity, especially in its “acquisition” version. Thus, I think the metaphysical question whether only creatures endowed with a conceptual repertoire can perceive the world around them must be answered in the negative.
3. THE EPISTEMOLOGICAL PROBLEM
We can now return to the epistemological problem, encapsulated in the argument from the justification of empirical beliefs, of whether only conceptualized experiences can play a justificatory role with respect to the corresponding beliefs.
3.1. Peacocke and Burge
Especially two theorists – i.e. Christopher Peacocke and Tyler Burge – have put forward the view that although perceptions are not conceptual, they have representational contents and therefore correctness conditions. Moreover, they have both maintained that perceptions can, as such – i.e. unaided by judgment –, provide a warrant for the corresponding empirical beliefs.
Their characterization of the very content of experience is partially different. They both think it is a kind of map of the environment around the perceiver. However, Peacocke introduces the notion of “scenario” – a sort of relatively coarse-grained way of filling out the space around the perceiver, which can be correct or incorrect – and that of “protoproposition” to account for kinds of cases where the scenario by itself would not discriminate between different ways of filling in the perceptual scene.Footnote 9 Burge, in contrast, does not make use of these notions. He thinks that perceptual content is a kind of “map” that fills in the space around the subject and has singular and general elements. That is to say, it singles out specific objects (or shapes, or colors, etc.) and attributes properties to them (being to the right of, to the left of, above or below something else, etc.). Now, the number of perceptual attributives, for Burge, is limited and consists of spatial, temporal, and numerical ones, as well as of attributives that indicate food and, possibly, conspecifics and mates.
However, the problem remains of how a perception, which, as such, is not conceptualized and has no propositional structure, could justify an empirical belief. Obviously, it cannot do it if by “warrant” one means a justification, which is constitutively articulable by a subject. For articulability requires conceptualization in judgment. Hence, the proposal would end up embracing a sophisticated form of coherentism, where experiences as such have non-conceptual and non-propositional structure, yet are endowed with that kind of structure in judgment and the latter enters the justification of further beliefs and judgments.Footnote 10
It is perhaps for this reason that Peacocke and Burge have proposed to countenance a different notion of warrant, called “entitlement”,Footnote 11 which would allow experiences as such – that is, independently of being conceptualized and of having propositional structure – to play a genuinely justificatory role (once certain further conditions are met).
Now, according to Burge, entitlements do not require subjects to have access to them – so they need not even be conscious experiences –, even less that subjects be able to justify why they are warrants. Rather, it is enough that the perceptual system works properly and therefore satisfies the natural norm that characterizes it. Roughly, that norm requires that the system produce correct representations of the environment, at least in the kind of surroundings in which the system has evolved, in such a way as to be capable of representing them correctly. Hence, when placed in a different kind of environment, the system would still produce an entitlement, although it would not yield a correct representation. Burge's entitlements are therefore clearly defeasible. Yet, since they are not enough to produce a warrant for general beliefs such as “There is an external world”, they do not give rise to the unpleasant consequence of providing an entitlement for such a belief when the system is placed in a kind of skeptical environment. Finally, in order to have an entitlement for a given belief, the transition from the perception to the belief based on it must preserve the former's content, while there are no defeaters.
This last requirement is actually a little problematical. For if defeaters are conceived of internistically, as reasons to think that one's perception and further cognitive operations may be unreliable, then subjects should be able to conceptualize them and even to conduct inquiries to make sure whether they obtain. This, however, would prevent young children and the unsophisticated from having perceptual entitlements for their beliefs. Yet, this would spoil the very notion of entitlement, which, in this context, is meant precisely to overcome the unpalatable result that ordinary notions of justification would have with respect to the epistemic status of young children's beliefs. Burge is aware of this problem and suggests thinking of defeaters “externistically”, as unfavorable conditions, which need not be within a subject's ken (Burge Reference Burge2003: 544). However, it now seems that by being placed in uncongenial environments subjects would, after all, lose their entitlements for their perceptual beliefs, because these unfavorable conditions would indeed obtain. Burge's notion of entitlement, therefore, risks being very close to traditional reliabilist accounts of justification. Furthermore, the very idea that there be natural norms characteristic of perception which should hold a priori, is problematic and, if one does without it, it is even clearer that entitlements do not represent a real improvement over reliabilism.
It seems dubious that the notion of representation carries with it and a priori, as Burge has it, a commitment to its veridicality. In particular, the notion of representation seems simply to involve the idea of providing a potential layout of the world around a perceiver, which can be correct or incorrect, depending on whether it corresponds, or fails to correspond to such a portion of environment. Yet, were it incorrect, the representation would be a representation nonetheless. Nor does the notion of representation seem to entail that one must be or have been, at least in the past, in causal relations with the objects it allegedly represents. There can be representations of unicorns even if they never existed. Nor does it seem to be a natural norm that perceptual representations are factive and therefore entail the correctness of their content. To hold that much seems a conventional aspect, perhaps fostered by the use of English, which does not hold in other languages (but, clearly, such a use could be stipulated in either language). Now, without recourse to natural norms characteristic of perception, Burge's position does not seem to be substantially different from forms of reliabilism, which have been extensively criticized at least as theories of justification, and from which he wishes to steer away. Furthermore, one should be suspicious of a notion of warrant that is compatible with the possibility that subjects affected by blind-sight could have an entitlement for their perceptually based beliefs. For, in that case, what would differentiate a warranted belief from a mere shot in the dark, at least from a subjective point of view?Footnote 12
Let us now turn to Peacocke's position. Some general tenets are similar to Burge's. In particular, that a subject need not know that he has a perceptual entitlement in order to have it and even less how it manages to give him a warrant for his beliefs. The differences concern both some details and some general background assumptions. Let us start with the former. First, for Peacocke the transitions from perceptual contents to the corresponding beliefs are “relatively a priori”, in particular when beliefs do not go beyond the concepts whose possession conditions are exhausted by the capacity to exercise them when one has the corresponding experiences. An example of such a transition would be the one from a red perceptual content to the belief “This is red”. Secondly, Peacocke explains why such transitions would lead to mostly true beliefs in a sort of evolutionistic way. For his idea is that perceptual systems have evolved to yield correct representations of subjects' surroundings. Therefore, if we take our experiences at face value we will form mostly true beliefs, as long as the latter require the exercise of merely observational concepts.Footnote 13
Turning now to the general background assumptions, it has to be noticed that Peacocke's position is crucially different from Burge's insofar as, for him, rational transitions are those that are such from a subject's own point of view. Hence, in making the transition from a red perceptual content to the corresponding belief, a subject must know what it means to be red, be aware that he has sufficient reasons to move from the experience to the belief and do so because he is aware of these reasons (Peacocke Reference Peacocke2004: 176). A further element that helps shed light on the differences between the two proposals is that for Peacocke an experience must be conscious in order to give a subject an entitlement for the corresponding belief. As we saw before, in contrast, Burge's position seems to be compatible with the fact that even a subject affected by blind-sight could have an entitlement for the belief based on it. Hence, Peacocke is much more of an epistemic internalist than Burge. Indeed, one might even wonder whether he is actually making use of a notion of warrant that is crucially different from the usual, internalist notion of justification. Sure, there is no requirement that the entitlement be articulable by the subject himself, yet a subject should know, at least implicitly, a great number of things with respect to the conditions in which warrants are supposed to obtain. Now, are we really sure young children and the unsophisticated do? The standard move is to make all this knowledge implicit. This strategy,Footnote 14 however, is problematical, at least when the conditions a subject should meet in order to have it are not clearly specified. More importantly, supposing the conditions were clearly specified, wouldn't the ensuing view come down to a form of sensible, yet by all means traditional internalism?
Be that as it may, let us now consider an objection that I think can be raised against both accounts and that will allow us hopefully to put forward a proposal about the content of perceptual experience in order for it to fulfill a justificatory role. We could put the point by asking: how many genuinely perceptual entitlements do we have? For, according to Burge, perceptual attributives are very limited and, for Peacocke, purely observational concepts are just a few. If so, perceptual entitlements are either too many – many more than those we would, intuitively, wish to allow for – or too few.Footnote 15 Let us consider a subject who has a perception with the content of a small spherical red shape. This content is compatible with a great number of beliefs, e.g. “Here is a red sphere”, “Here is a red apple”, “Here is a red ball”, “Here is a red spherical candle”, etc. Hence, if that perception, as such, can warrant each of these beliefs, we would end up having perceptual entitlements not only for the belief we have actually formed on that basis, e.g. that there is a red apple, but also for many more, some of which would be incompatible with the one we have formed. There are two problems with this proposal, however. First, if perceptual experiences should be a guide to the appropriate beliefs, how could they fulfill their role if they licensed any number of different and potentially incompatible beliefs? Second, if entitlements are conceived of as guides to the truth, then how could they play such a role when perceptions as such could justify any number of potentially incompatible beliefs?
If, in contrast, and as Burge's and Peacocke's writings seem to suggest, that perception gives one an entitlement to believe only “Here is a red spherical shape”, then perception as such would provide us with entitlements for very few beliefs. Even the humdrum belief that there is a red apple in front of us would have to be justified by the interplay between our perceptual and our conceptual systems. As before, there are only two ways in which this can happen, either by mobilizing beliefs, or else by letting concepts shape our very perceptions, just as McDowell used to maintain. The former option would actually incur in a sophisticated form of coherentism, as we saw. The latter would not, but it would represent an important concession to McDowell's position.
I would actually recommend opting for this second horn of the dilemma. Interestingly, however, this would show that the debate between conceptual and nonconceptual theorists regarding the content of perception is much less sharp and clear-cut than is usually thought of. For the two positions need to interact in fruitful ways if we wish to answer both the metaphysical problem of perception – and hence allow that many more creatures beside human adults could actually enjoy fully representational perceptual states – and the epistemological problem of how experiences, as such, can justify our beliefs based on them. The kind of solution I am proposing is a kind of divide et impera. For I think nonconceptualists are right with respect to the metaphysical problem, while conceptualists are right regarding the epistemological one.
3.2. Siegel
Before analyzing this position in more detail, let me review yet another proposal, recently made by Susanna Siegel (Reference Siegel2010). According to Siegel, in perception itself, without the aid of concepts, we are presented with “rich” contents. She writes: “When you see a bowl of fruit, you can (usually) recognize the kinds of fruit in the bowl” (p. 78) and, according to her, such a recognitional ability is purely based on the content of one's perceptions unaided by concepts. Notice that for Siegel such recognitional abilities may well be based on the representation in visual experience of natural kind properties (cf. p. 114). Now, of course I think she is right to hold that we can learn to recognize kinds of fruit, of tree, and so on. I also agree with Siegel that these recognitional abilities may alter the phenomenology of one's visual experiences (see her nice discussion of how knowledge of Cyrillic can change the phenomenological aspects of one's visual experience while looking at that kind of alphabetical characters). Finally, I agree with her that some recognitional abilities regarding, say, moms, dads, types of food and one's immediate and familiar surroundings (see Millikan Reference Millikan1998), may take place without possession of the relevant concepts. What, however, is more difficult to accept is that recognitional abilities about natural kind properties do not involve the exercise – however passive it may be – of a conceptual repertoire. That is to say, it may be that without the aid of the concept apple a child can recognize the fruit in the bowl as the same one he saw the day before and it may obviously be the case that he is able visually to distinguish it from pears and oranges. This ability, however, has nothing to do with the ability to visually recognize the property of being an apple taken as a natural kind one. To be able to do that seems clearly to require the concept of natural kind. Hence, one should be sensitive to the fact that two identical-looking objects may not be instances of the same natural kind (e.g. water and twater, or jade) as well as to the fact that two different-looking objects can be instances of the same natural kind (e.g. water and ice). Precisely because natural kind properties are independent of the way things look to us, I take it that it is not possible to perceive natural kinds properties as such. Hence, I agree that perceptions have a representational content that is not exhausted merely by shapes and colors and that represents objects (and people) as having specific, or at least characteristic contours, colors, odors, textures, tastes, sounds and voices. However, I disagree that in experience, unaided by concepts, we represent more than that and, in particular, I do not agree that we represent natural kind properties.
One may then wonder whether the kind of recognitional abilities here under discussion would not suffice for concepts' possession, as Ruth Millikan claims. In that case, it would be clear that concepts would already figure in perception. Yet, even granting that these recognitional abilities are not based on the exercise of concepts – or do not themselves constitute the possession of the relevant concepts – it is quite clear that the proposal would not take us too far with respect to the solution of the underdetermination problem. Take an environment in which there are both apples of the delicious variety and candles with the same shape and color, as well as decorations in the house resembling apples. The visual experiences would all be relevantly similar. Hence, once more, experiences unaided by concepts would not be sufficiently fine-grained to warrant “Here is an apple” any more than “Here is an apple-looking candle”, etc. contrary to what the entitlement theory maintains. Thus, if we want experiences to warrant – no matter how defeasibly that might be – the relevant belief, concepts have somehow to be in play. That way, the experience could be taken at face value to form the corresponding belief, absent defeaters. By contrast, without the exercise of concepts already in experience, what belief would that very experience be a guide to? They would all be on a par. To stress, the point of requiring concepts to enter experiences is not that of making them indefeasible warrants for the corresponding beliefs. For, after all, it may be the case that the apple-looking candle has been put in the bowl, unbeknownst to the subject who is accustomed to his environment and knows that usually apples are in the kitchen and apple-looking candles are not. Rather, the point is to allow experiences to be guides to the relevant beliefs, which may then be true or false. Without the exercise of concepts, they are not sufficiently fine-grained to guide a subject, even when raised in that kind of environment, to any of the possible beliefs licensed by those very experiences.
4. THE PROPOSAL
Thus, I submit that we have to grant that also creatures without concepts can have genuine perceptual representations of their surroundings (and their bodies). Yet, in order for these perceptual experiences to enter the structure of empirical warrants, they must involve the passive exercise of concepts. Furthermore, the underdetermination problem reviewed before, teaches us that these concepts had better be as fine grained as the ones that enter the corresponding beliefs and hence exceed merely observational ones.Footnote 16 Otherwise, we would end up having either too many or too few perceptual justifications. Hence, the perceptual content of creatures endowed with concepts is partially different from the one of those who do not possess them. While the latter is simply a kind of map of their environment with some specific perceptual attributives, the former gets refined and restructured in such a way that those elements fill in a conceptual, propositional structure. That is to say, while nonconceptual creatures merely enjoy a sort of “pictorial” representation of – say – a red sphere in front of them, creatures endowed with the relevant concepts see that there is a red apple in front of them, where this propositional and conceptual structure is immediately presented to them in the very perception itself, and has all the empirical content perception would provide and which its mere entertainment in thought may well lack. Finally, the passive exercise of concepts in conscious experience complies with the internalist requirement that perceptual warrants be, at least partly, internal states of a subject, which are given to him, although this does not require the articulability on his part of his own warrants. For there is no need that a subject be able to conceptualize his perceptual experience as a seeing that things are thus and so. It is enough that he actually sees that there is a red sphere in front of him. Yet, such a warrant is indeed articulable, at least in principle, by those who happen to have the relevant conceptual repertoire. Hence, it has all the characteristic features of a respectable internalist justification.
Finally, let me briefly address the issue of the epistemic status of the beliefs of young children and of the unsophisticated. It is part of the picture I have been advocating that non-conceptual creatures cannot have perceptual justifications, while they can have perceptions. Yet, it has to be noticed that they could not have beliefs either. For beliefs require concepts, which, ex hypothesi, they lack. Thus, if we are dealing with early-age infants or non-conceptual animals, this would not be a problem. On reflection, however, the view is not problematical when taken in connection with older children either. For they will have as many perceptual warrants as they are capable to form, given their conceptual resources, which, crucially, on the present proposal, match the number of beliefs they are actually able to entertain.Footnote 17
Thus, to sum up and conclude: perceptions as such, i.e. unaided by judgments, yet by courtesy of some general background assumptions, which need not be conceptualized by a subject, can justify empirical beliefs based on them. Yet, this can be the case only for those creatures who have a conceptual apparatus and are therefore in a position passively to exercise it in perception.Footnote 18
