1. Introductory Remarks

Wilfrid Sellars defends a dual-component account of experience, according to which experience consists of two separable components: a conceptual component, which involves an exercise of conceptual capacities, and a non-conceptual phenomenal component. The most important characteristic of this account is that the non-conceptual phenomenal component in experience does not play any role in justifying beliefs. Against this theory, Anil Gupta raises several powerful objections in his 2012 paper “A Critical Examination of Sellars’s Theory of Perception” and in Chapter 2 of his 2019 book, Conscious Experience: A Logical Inquiry. My purpose in this paper is to defend Sellars’s theory against these objections. In particular, I will address his five most important objections.

First, by way of offering a transcendental argument, Sellars defends an epistemic principle that our ordinary perceptual judgements are likely to be true. On this argument, we are effective agents in the world, and this fact requires that such an epistemic principle be an element in our conceptual framework. However, we could still be thinking and acting agents, even if such an epistemic principle were false. Second, Sellars’s theory of perception has no resources to explain the impropriety of a demand for proof of perceptual judgements. Third, his theory fails to explain the phenomenon that one can retreat safely to an appearance judgement in face of challenges. Fourth, Sellars’s account of perceptual judgements is deeply unsatisfactory insofar as it fails to explain the epistemic role of experience in cognition. Finally, Sellars is correct in taking the propositional given as mythical. Nonetheless, he is mistaken in taking the non-propositional given as mythical.

As we will see in due course, I admit that some aspects of Sellars’s theory are problematic: for example, there are good reasons to reject his view that the ultimate referent of a perceptual taking is a sense impression. Nonetheless, I will argue that his overall theory of perception is defensible, despite these problematic aspects.Footnote ¹

2. The Reasonableness of Fundamental Epistemic Principles

Let us begin by considering the following epistemic principle.

Sellars defends such an epistemic principle by offering a transcendental argument.Footnote ² His argument is very roughly as follows. We are effective agents in the world. To be effective agents, we need reliable cognitive maps of our surrounding world. If epistemic principles such as PJ did not hold, we could not have a reliable cognitive map of our world at all. Therefore, the fact that we are effective agents in the world requires that epistemic principles such as PJ be elements in our conceptual framework. Along this line of reasoning, Sellars argues that fundamental epistemic principles such as PJ are justified, not by virtue of empirical evidence, but as necessary conditions for a hardly disputable assumption, namely that we are effective agents in the world.

Gupta argues, however, that our status as thinking and acting beings does not depend on the truth of PJ.Footnote ³ The most important reason is that perceptual judgements could contain concepts that embody fundamental misconceptions. For example, consider a language community in which the expression ‘up’ is governed by the following two criteria: the perceptual criterion and the conceptual criterion. On the perceptual criterion, one’s judgement ‘a is up above b’ is warranted when the direction of the ray from b to a is the same as the direction of the ray from the centre of the earth to b. In addition, on the conceptual criterion, one’s judgement ‘a is up above b’ is warranted when the direction of the ray from b to a is the same as the direction of the Standard Up, which is defined in the following way. There are two perceptually salient objects in the community, namely, a mountain peak p and a natural satellite s located in geostationary orbit above p. And the direction of the Standard Up is the same as the direction of the ray from p to s. These two criteria do not provide different verdicts about the use of ‘up’ when they are used in places near the mountain; but they come into conflict with each other when they are used in places far away from the mountain because the earth is round. Here let us assume that people in the community do not know that these two criteria governing the use of ‘up’ are inconsistent. In such a case, despite the fact that the rules for the use of ‘up’ defined by these two criteria are, strictly speaking, inconsistent, this expression can nonetheless be effectively used within a certain boundary. As this example shows, it is possible that people’s perceptual beliefs involve a fundamental misconception to the effect that they are, strictly speaking, false, and yet these false beliefs reliably guide their behaviour to a considerable extent.Footnote ⁴ Partly for this reason, Gupta thinks that empirical rationality does not require the truth of PJ.

I agree with Gupta that we could still be thinking and acting agents, even if PJ were false. But I think we can defend PJ in a coherentist way, without assuming that we are effective agents. I have argued for this view elsewhere.Footnote ⁵ Accordingly, let me here confine myself to sketching this view.

First, all living animals use sense organs to obtain information about the world necessary for their survival and wellbeing. We are no exception. Unlike mere animals, however, we are rational beings. In particular, on Kant’s view, it is our conception of ourselves that we are rational beings who can engage in theoretical and/or practical reasoning in order to determine what to believe and/or what to do.Footnote ⁶ What then is our distinctive way of obtaining information about the world as rational believers? We acquire information about the world in a way that is bound by the norms of rationality. In other words, unlike other animals, we are by nature such rational beings whose beliefs are bound by the norms of rationality. Due to this distinctive rational nature of ours, we can engage in reasoning in order to determine what to believe. And this epistemic practice assumes that it is, at least in principle, possible for us to justify something. If we deny this possibility, we cannot defend or criticize any claim, and so the end of rational discourse. Along these lines, we may argue that our epistemic discourse does not begin without any presumptions, and that it is a minimum presumption for our epistemic discourse that it is, at least in principle, possible for us to justify something. Here observe that to deny this presumption is tantamount to denying our nature as rational beings. In this regard, it is noteworthy that even a sceptic can hardly deny this minimum presumption for our epistemic discourse. If a sceptic denies this minimum presumption, she cannot defend her sceptical claim. For this reason, the sceptic who denies this minimum presumption is self-refuting. Hence, we can hold this minimum presumption unless we are given some positive reasons to think otherwise.

Second, epistemic justification (henceforth simply ‘justification’) is a normative concept. We ought to accept justified beliefs, whereas we ought not to accept unjustified beliefs. In addition, justification is a goal-dependent concept. To say that a belief is justified is tantamount to saying that it has a favourable status vis-à-vis our epistemic goal. By contrast, to say that a belief is unjustified is to say that it has an unfavourable status vis-à-vis our epistemic goal. Thus, if our epistemic goal were unreasonable, we would not have to accept any alleged justified beliefs. Accordingly, the normative status of a belief depends on the reasonableness of our epistemic goal. Hence, it is a minimum presumption for our epistemic discourse that our epistemic goal is reasonable, just as it is a minimum presumption for our epistemic discourse that it is, at least in principle, possible for us to justify something.

What then is our epistemic goal? Our epistemic goal is usually understood as having true beliefs and avoiding false beliefs. But we cannot step outside our conceptual framework to judge whether a belief is true (or justified). On this coherentist insight, we have no other way but to evaluate whether a belief is true (or justified) on the basis of our epistemic norms within our conceptual framework. In addition, as Kant insists, it is our conceptual framework that provides the norms, criteria, or rules for defending (or criticizing) any claim. Therefore, it is inevitable to address any justification question on the basis of our conceptual framework. Along these lines, Sellars argues that the best we can do for our epistemic goal is to gradually improve our conceptual framework from within. In particular, he upholds an explanatory coherence theory, according to which our epistemic goal is to gradually improve our conceptual framework, so as to achieve a world picture with a maximum of explanatory coherence.

Third, I accept a Lehrerean conception of coherence. Coherence is usually understood as a matter of how the components in a system of beliefs fit together or dovetail with each other. Except for this sort of metaphorical characterization, however, there is no generally accepted definition of ‘coherence.’ In this frustrating situation, Keith Lehrer offers a very illuminating account of this elusive concept.Footnote ⁷ On his proposal, we can understand coherentist justification in terms of ‘beating all competitors’ or ‘answering all objections.’ On this view, roughly, one is justified in accepting that p just in case one can answer all objections raised against it (or beat all competing hypotheses) on the basis of one’s background system. But I accept a Sellarsian social practice theory of justification. On this view, our concept of being justified in holding a belief has been developed on the basis of our social practice of demanding justification and responding to such demands. Thus, we should understand our notion of justification in accordance with this social practice model of justification. This is an intersubjective model of justification, rather than a subjective model. As a consequence, we must modify Lehrer’s conception of coherence in accordance with this intersubjective model. On this social practice theory of justification, roughly, we are justified in accepting that p just in case we can answer all objections raised against it in our social practice of justification.

Fourth, as Robert Brandom argues, our social practice of demanding justification and responding to such demands requires the default-and-challenge structure of justification.Footnote ⁸ As previously pointed out, it is a minimum presumption for our epistemic discourse that it is, at least in principle, possible for us to justify something. In addition, as also pointed out, our concept of justification has been developed on the basis of our social practice of justification. Thus, it is a minimum presumption for our epistemic discourse that it is, at least in principle, possible for us to justify something in our social practice of justification. Furthermore, the default-and-challenge structure of justification is required for us to justify something in our social practice of justification. The reasons are roughly as follows. In the first place, the infinite regress of justification is impossible. Suppose that a claimer defends his claim by offering a ground, A. A challenger can call this ground into question by saying, ‘Why A?’ To meet this challenge, the claimer might provide another ground, B. The challenger can, in turn, call this ground into question by saying, ‘Why B?’ Here it should be noted that, if the challenger were entitled to keep raising a question ‘Why is that?’ to any reply of the claimer, there would be no claim that the claimer can ultimately justify. In the second place, genuine doubt can take place only against a background of beliefs that are not doubted at the same time; to put the point another way, there must be a conceptual framework within which doubts and settlement of doubt take place. Thus, the possibility of one’s defending something requires that some claims be treated as having default justification unless some challenger provides positive reasons to doubt them; that is, there must be some claims for which the burden of proof shifts to the challenger.

With the above points in mind, let us now consider how to justify PJ. As noted, we have no other way but to address any justification demand on the basis of our social practice of justification. PJ is no exception. And we can defend this principle in the following way. First, we have relied on PJ in order to justify perceptual beliefs about the world. Second, we have no positive reason to think that this principle is false. Third, we have no alternative but to accept PJ in order to pursue our epistemic goal. If these three conditions hold, then we can take PJ as having a default positive justificatory status in our social practice of justification. Let me elaborate on the last point.

As mentioned, it is a minimum presumption for our epistemic discourse that it is reasonable for us to pursue our epistemic goal. Recall that our epistemic goal is to gradually improve our conceptual framework, so as to achieve a world picture with a maximum of explanatory coherence. In addition, we ultimately obtain information about the world through the senses. In other words, our senses are the ultimate source of information about the world. Thus, if we could not rely on our perceptual abilities in order to obtain information about the world, then we could not realize our epistemic goal of achieving a world picture with a maximum of explanatory coherence. For denying the general reliability of our perceptual abilities is tantamount to closing our ultimate epistemic door to the world. Therefore, if we reject the general reliability of our perceptual abilities, we cannot get our epistemic pursuit off the ground. Along these lines, we may argue that we have no alternative but to accept PJ in order to pursue our epistemic goal.

As already pointed out, if the aforementioned three conditions hold, then we can take PJ as having a default positive justificatory status in our social practice of justification. Therefore, unless some sceptic somehow provides us with positive reasons to believe that our epistemic goal is unreasonable, or unless some genuine alternative to PJ is provided, it is reasonable for us to accept PJ. Here notice that my argument so far does not depend on Sellars’s assumption that we are effective agents in the world. Unlike his proposal, my proposal employs the strategy of shifting the burden of proof about PJ to any challenger. Hence, my Sellarsian coherence theory can defend PJ, while admitting Gupta’s objection that we could still be thinking and acting agents, even if PJ were false.

3. The Impropriety of a Demand for Proof of Perceptual Judgements

Gupta’s second objection is that Sellars’s theory has no resources to explain the impropriety of a demand for proof of perceptual judgements. He writes:

But the above objection does not pose a serious problem for my Sellarsian coherence theory either. As discussed in the previous section, the demand for justification should be explained in terms internal to our social practice of demanding justification and responding to such demands; and we can take PJ as having a default positive justificatory status in our social practice of justification. As a consequence, we are justified in holding our ordinary perceptual judgements (or beliefs) unless we are given some positive reasons to doubt them: for example, we are having a hallucination under the influence of a drug. In other words, the burden of proof for our ordinary perceptual judgements shifts to those who challenge them. This is why a demand for justification is improper for our ordinary perceptual judgements, unless one has some positive reasons to doubt them. By contrast, judgements about mathematical facts such as the one that the sum of the angles in a triangle is 180° are not such that they have default justification in our social practice of justification. Here, it is noteworthy that, as Michael Williams points out, “the idea that epistemic justification conforms to a default-and-challenge structure does not conflict with (and so is not an alternative to) the position that Sellars defends in [his 1979 paper].”Footnote ¹⁰ Along these lines, a Sellarsian can explain why it is improper to make a demand for proof of ordinary perceptual judgements.

4. Shifts in Perceptual Judgements

Gupta’s third objection is that Sellars’s theory fails to explain an epistemic phenomenon that one can always retreat safely to an appearance judgement in face of challenges. He writes:

Let me address the above objection by considering three situations in which a subject, S, makes each of the following judgements, respectively:

1. (i) The object in front of me is red.

2. (ii) The object in front of me looks red to me.

3. (iii) It looks as if there is a red object in front of me.

According to Sellars, these cases share two common components. On the one hand, there is a common propositional content that can be expressed by the sentence ‘There is a red object in front of me.’ On the other hand, there is a common non-propositional component as well. Notice that, from S’s subjective point of view, her experiences can be qualitatively indistinguishable in these cases. Nonetheless, there are still important differences. In the case of (i), S ascribes to her experience the propositional content that there is a red object in front of her, and endorses this content as well. If this endorsement is correct, (i) is a case of veridical experience. In the case of (ii), S ascribes the same propositional content to her experience, but she only partially endorses this content. More specifically, although S endorses that there is a particular object that she refers to as ‘the object in front of me,’ she does not endorse that this object is red. Thus, this case is compatible with the possibility that the particular object to which she refers as ‘the object in front of me’ is not really red, but only looks red perhaps due to red lighting above her head. In the case of (iii), S ascribes the same propositional content to her experience, but she does not endorse even the claim that there exists a physical object which she purports to denote as the ‘object in front of me.’ Thus, this case is compatible with the possibility that she is having a hallucination.

Now it is important to observe that the weakest judgement, (iii), is hardly refutable. In this case, as noted, S ascribes to her experience the propositional content that can be expressed by the sentence ‘There is a red object in front of me.’ But except for this, she does not undertake any substantial commitment. Recall that she is not even committed to the existence of what she refers to as the ‘object in front of me.’ Thus, making this kind of appearance judgement is tantamount to making the most guarded judgement, that is, the most cautious, least committal judgement.Footnote ¹² And such a guarded judgement is not vulnerable to correction. This is because in making such a guarded judgement one does not undertake any substantial commitment about how things are: no substantial commitment, no correction. Along these lines, Sellars can explain why one can always retreat safely to appearance judgements in face of challenges. Let me emphasize the reason again: in so sticking to appearance judgements, one does not make any substantial claim about how things are, so that these appearance judgements are hardly refutable.

5. Explanation of Perceptual Judgements

Gupta’s fourth objection is that Sellars’s account of perceptual judgements is deeply unsatisfactory insofar as it fails to explain the epistemic (or rational) role of experience in cognition. In this section, however, I argue that Sellars’s overall theory of perception is defensible, even though we may have to reject his ancillary claim that the ultimate referent of a perceptual taking is always a sense impression.

First, it is worth considering why Sellars introduces the notion of sense impression in the first place. Consider again the three situations in which S makes each of the following judgements, respectively:

1. (i) The object in front of me is red.

2. (ii) The object in front of me looks red to me.

3. (iii) It looks as if there is a red object in front of me.

Suppose that (i) is a case of veridical perception, whereas (ii) and (iii) are its non-veridical counterparts. As noted in the previous section, a veridical perception and its non-veridical counterparts share a common propositional content. The question then is: why do we have similar conceptual episodes in a veridical perception and its non-veridical counterparts? According to Sellars, an experience of seeing (or seeming to see) that p contains a distinctive phenomenal aspect that is absent in the case of merely thinking that p, and we need to introduce sense impressions in order to explain this difference between experience and mere thought. In addition, we also need sense impressions in order to explain why we have similar conceptual episodes in a veridical perception and its non-veridical counterparts. This is because these conceptual responses are triggered by suitably similar proximate causes, which are sense impressions.Footnote ¹³

Now it is worth noticing that Sellars draws a sharp line between the logical space of reasons and the realm of law, so that the above-the-line elements that belong to the logical space of reasons are not conceptually reducible to the below-the-line elements that belong to the realm of law. As noted, sense impressions are the causal antecedents of perceptual judgements. In addition, there is no difference in conceptual content between seeing that p and merely thinking that p, and so the aforementioned distinctive phenomenal aspect that is absent in mere thought is non-conceptual. Therefore, insofar as sense impressions are understood only as non-conceptual episodes in the below-the-line, they cannot play any epistemic (or rational) role. According to Gupta, however, this view of Sellars’s fails to explain the epistemic role of experience in cognition. I will say more on this in due course.

Another problem with Sellars’s theory is related to his view that the ultimate referent of a perceptual taking is a sense impression. According to Sellars, we can characterize perceptual takings as having the functional role of hybrid demonstrative phrases of the form ‘this-such.’ Suppose that S forms a visual perceptual belief whose propositional content can be expressed by ‘This red object is rectangular on the facing surface.’ Suppose also that there is nothing in the external world for S to be referring to as ‘this red object.’ In such a case, according to Sellars, the perceptual demonstrative ‘this red object’ ultimately refers to S’s sense impression. Sellars’s reason is roughly as follows. Our perceptual experiences are always about something. For they are not cases of imagination.Footnote ¹⁴ Thus, even when S is hallucinating, her perceptual taking must have its referent. But there is nothing in the external world for S to be referring to as ‘this red object,’ and so this referent should be a sensation, rather than a physical object. For this reason, Sellars takes sense impressions (or sensations) as the last-resort referents of our perceptual takings.Footnote ¹⁵ In other words, he argues that the ultimate referent of a perceptual taking is always a sense impression. But this view has a problematic implication.

Our perceptual thoughts are normally directed to physical objects in our surroundings, rather than to inner phenomenal states that prompt them. Accordingly, as Sellars admits, at least in standard conditions, S’s perceptual taking represented by a demonstrative phrase such as ‘this red object’ is a taking of a physical object. And S’s illusory perceptual experience can be phenomenally indistinguishable from the corresponding non-illusory perceptual experience. Thus, it seems that S’s perceptual taking, even in a non-veridical case, is a taking of a physical object. Therefore, on Sellars’s view, phenomenology is misleading, because the subject’s perceptual taking is a taking of a physical object, but the ultimate referent of a perceptual taking is instead a sense impression, which is a state of the subject. But this view is vulnerable to the following objection: “Logic and phenomenology provide no reason to suppose that the sensing itself is present in experience. … From the logico-phenomenological viewpoint, it is odd in the extreme to suggest that the demonstrative denotes a state of the subject.”Footnote ¹⁶ I agree with Gupta on this point. Thus, in what follows, I argue that the referents of one’s perceptual takings are not one’s sense impressions, even when one directs one’s attention to one’s perceptual experiences.

First, as mentioned, the difference between seeing that p and merely thinking that p lies in the fact that the former contains an additional component, a distinctive phenomenal state that is absent in the latter. But as Paul Coates argues, we can see that there is a distinctive phenomenal component in experience, not by some kind of direct apprehension of the phenomenal component, but by a second-order reflection on the difference between two different kinds of conceptualized states: seeing that p and merely thinking that p.Footnote ¹⁷ This means that the phenomenal component in experience is not the kind of thing that one can directly apprehend by some kind of inner sense. In this regard, it is worth recalling that ‘sense impression’ is a theoretical term that is introduced to explain why we have similar conceptual episodes in a veridical perception and its non-veridical counterparts. Thus, one can grasp the character of one’s experience only as a conceptualized state, and this conceptualized state is common to a veridical experience and its corresponding non-veridical experiences. Hence, insofar as the subject purports to refer to a physical object in the case of a veridical experience, S will also purport to refer to a physical object in the case of a non-veridical experience, since a veridical experience and its non-veridical counterparts are inner states that are qualitatively indistinguishable from the subject’s point of view.

Second, as Coates also argues, when one directs one’s attention to one’s perceptual experience, one can characterize the experience only in general terms.Footnote ¹⁸ The reason is roughly as follows. When one directs one’s attention to one’s perceptual experience, it ceases to play its normal role, that is, the role of enabling one’s attention to be directed towards the ostensibly seen environment. As noted, a veridical perception and its non-veridical counterparts share a common phenomenal component. Thus, the qualitative aspect of one’s perceptual experience is logically independent of any specific object one is seeing in a veridical case. For this reason, when one characterizes one’s perceptual experience in a way that is logically independent of any specific external object, one cannot help but characterize it only in general terms, by reference to the general links between experience types and the kinds of objects and properties in the subject’s surroundings. For example, S can describe her perceptual experience in the following way: ‘I have a visual experience of the sort I would have when I see a red book in front of me.’ Therefore, what is taken in the case of the most cautious perceptual taking, which is common between a veridical experience and its corresponding non-veridical ones, can be characterized only in general terms.

Third, given the distinction between mere referential purport and referential success, to say that a perceptual demonstrative, such as ‘this red object in front of me,’ actually succeeds in referring to its purported referent is to make a substantial claim, which is fallible. But, as pointed out above, insofar as S’s perceptual experience is the object of her attention, she can characterize the experience only in general terms. And describing one’s experience in this general way is tantamount to taking one’s experience in the most cautious way. Moreover, as Sellars points out, we can get at the most cautious perceptual taking by thinning out our perceptual commitment.Footnote ¹⁹ In so doing, however, we do not intend to change the referent of a perceptual taking. Rather, we avoid making a substantial commitment, which is vulnerable to rational criticism. Therefore, to say that I have an experience of the sort I would have when I see a red book in front of me does not imply that I am committed to the claim that my perceptual taking succeeds in making a demonstrative identification of a particular object (or state). What I am saying is rather that, from my subjective point of view, my experiential state is suitably similar to the experiential state I would have when I see a red book in front of me. To put the point another way, a perceptual taking represented by a demonstrative of the form ‘this-such’ is consistent with the discovery that it lacks its referent. Hence, when a person characterizes her experience only in general terms, she does not have to commit herself to the claim that her perceptual taking succeeds in referring to a particular object (or state).

Fourth, one cannot be directly aware of a sense impression as a sense impression. Suppose that S sees a red book in front of her. In this case, she must exercise a physical object concept in her perceptual experience in order to take herself to be seeing a red book. Here, following Coates, it might be useful to distinguish two senses of ‘awareness’: a conceptual sense of awareness and a non-conceptual sense of awareness.Footnote ²⁰ In the conceptual sense of awareness, we may say that in the above case S is (conceptually) aware of an object as a book (that is, as belonging to a certain kind). And what S is (conceptually) aware of in her visual experience would be nothing other than a book. But, as already pointed out, a perceptual experience contains not only a conceptual component, but also a sense impression (as a non-conceptual phenomenal state). Thus, we may say that S directly experiences a sense impression. In this sense, we may say that, insofar as S is consciously aware of her perceptual experience, she is directly aware of her sense impression (as a component of her perceptual experience). But this kind of awareness does not require that she be (conceptually) aware of her sense impression as belonging to a certain kind. For one’s non-conceptual awareness of an experience does not involve any conceptual classification. Therefore, although one can be directly aware of a sense impression in the non-conceptual sense, one cannot be directly aware of a sense impression as a sense impression.Footnote ²¹ Let me elaborate on this point a bit further.

As noted, we can sometimes direct our attention to our phenomenal experiences by themselves, rather than to external objects in our surroundings. In other words, by adopting a reflective stance, one can form thoughts about one’s experiences as such. For example, S can think of her experience as ‘an experience of seeing (or seeming to see) a red book,’ and what she is conceptually aware of in this case is characterized in terms of concepts that classify the experience as belonging to a certain kind. At this point, it is important to notice that the form of one’s conceptual awareness of a phenomenal experience, such as ‘it looks to me as if there is a red and rectangular object over there,’ is not deictic. This is instead what can be characterized in an adverbial way: I am sensing in an of-a-red-rectangle manner; I am sensing a-red-rectangle-ly.Footnote ²² Notice that S is sensing in an of-a-red-rectangle manner does not imply that there is a red rectangular object which she has sensed. Thus, by going adverbial in this way, we can argue that one’s conceptual awareness of a phenomenal experience is compatible with the fact that one does not succeed in referring to something. To put the point another way, we can argue that sense impressions are not objects of conceptual awareness.

Fifth, Gareth Evans imposes a tracking constraint on demonstrative identification.Footnote ²³ On this constraint, one can identify an object insofar as one can keep track of the object over a period of time (albeit a very short time). For this reason, S can succeed in making a demonstrative identification of a certain object by the use of a hybrid demonstrative of the form ‘this-such’ only under the condition that she can, at least in principle, keep track of the same object over a period of time. And when a hybrid demonstrative such as ‘this red object’ picks out a physical object in the subject’s surroundings, the subject can keep track of the same object over a period of time. But the case for a sense impression is quite different for the following reason.

Suppose that at time t₁ S has a visual experience E₁ of the sort she would have when she sees a red book in front of her, and at time t₂ she has a visual experience E₂ of the sort she would have when she sees a red book in front of her. The question then is whether S can keep track of a certain object by the use of a hybrid demonstrative like ‘this something red’ over a period of time (for example, from t₁ to t₂). As I have emphasized, insofar as one directs one’s attention to one’s perceptual experience, one can characterize the experience only in general terms, that is, in a way that is logically independent of any specific object one is seeing in a veridical case. As also previously pointed out, insofar as one characterizes a perceptual experience only in general terms, one does not have to commit oneself to the claim that there exists something to which a hybrid demonstrative like ‘this something red’ refers. In this regard, it is noteworthy that, insofar as S characterizes a perceptual experience only in general terms, there are no objective criteria by which S can determine whether what she purports to refer to by the demonstrative ‘this something red’ at t₁ is the very same thing as what she purports to refer to by the demonstrative ‘this something red’ at t₂. Moreover, even if E₁ and E₂ are qualitatively indistinguishable, strictly speaking, they are different mental states. Notice that to say that two things are qualitatively indistinguishable does not imply that these two are one and the same thing. For example, two twins can look exactly alike, but these two are nonetheless different persons. Notice also that E₂ can happen independently of E₁, so that they are ontologically independent of each other. Therefore, there are no good grounds for thinking that what S purports to refer to by the demonstrative ‘this something red’ at t₁ is the very same thing as what she purports to refer to by the demonstrative ‘this something red’ at t₂. Along these lines, we may argue that a sense impression (as a mental state) fails to meet the aforementioned tracking constraint on demonstrative identification.

Sixth, and finally, we may admit that having a non-veridical experience is different from imagining something, for the former contains a distinctive phenomenal state that is absent in the latter. But this difference does not rule out the possibility that, in the case of a non-veridical experience, the subject’s perceptual taking represented by a demonstrative of the form ‘this-such’ is consistent with the discovery that it lacks its referent. Therefore, we do not have to accept Sellars’s view that our perceptual experiences are always about something. Instead, we can argue that, in the case of the most cautious perceptual taking, one is not committed to the claim that one’s perceptual taking actually succeeds in making a demonstrative identification of a particular object (or state).Footnote ²⁴

If the above considerations are correct, we may reject Sellars’s ancillary view that the ultimate referent of a perceptual taking is always a sense impression. Despite this problematic aspect of his theory, however, I argue in the remainder of this section that Sellars’s overall theory of perception is still defensible. Among others, we can provide good answers to the following questions.

To begin with, what are perceptual beliefs (or judgements)? As already pointed out, according to Sellars, perceptual takings are primarily the complex demonstrative constituents of perceptual beliefs. Thus, it might be appropriate to restrict perceptual takings to takings as. Nonetheless, he did not give up the notion of a taking that (as distinct from a taking as). What then is the relation between a perceptual belief and a perceptual taking (understood as a taking that)? For example, consider a case in which S visually takes it that a is F. To say that S visually takes it that a is F does not, by itself, imply that she believes that a is F. For example, if S thinks that she is seeing a piece of paper under red light, she can withhold accepting that this piece of paper is red, even though the content of her perceptual taking is that it is red. Nevertheless, in standard conditions, S would not have any positive reasons to doubt such a perceptual taking, and so S would think and act in accordance with such a taking. In this regard, it is important to note that perceptual beliefs are non-inferential beliefs that we come to have through sense perception. Therefore, in standard conditions, to say that S visually takes it that a is F is tantamount to saying that she perceptually believes that a is F. Along these lines, we may argue that a perceptual belief (or judgement) is nothing other than a perceptual taking (understood as a taking that) under a normal, default condition.

What then is the relation between a perceptual belief and a perceptual taking (understood as a taking as)? As mentioned, perceptual takings in this sense are the demonstrative constituents of perceptual beliefs; in other words, such a perceptual taking provides the perceiver with the subject component of a perceptual belief. Therefore, just as the subject component of a proposition is an indispensable part of the proposition, such a perceptual taking is an indispensable part of a perceptual belief.

The next question is: how are perceptual beliefs related to sense impressions? As already pointed out, sense impressions are the causal antecedents of perceptual beliefs. Here it is worth emphasizing that only conceptual responses, which have correctness conditions, can play a rational role, so that sense impressions (understood as non-conceptual episodes) cannot play a rational role. In addition, as already argued, a perceptual belief is a perceptual taking (understood as a taking that) under a normal, default condition. Thus, our ordinary perceptual takings have default justification in our social practice of justification. Thus, unless we are given some positive reasons to doubt them, we will think and act in accordance with our ordinary perceptual takings. If, on the other hand, one has some positive reasons to think that a given perceptual taking is not veridical, one should stop being guided by the perceptual taking. Along these lines, Sellars (or a Sellarsian) can explain the epistemic features of perceptual beliefs in terms of our perceptual takings (as takings that), without recourse to the alleged given in conscious experience.

Finally, we can also deal with the following objection of Gupta’s: “The resources of ‘cause’ and ‘sensing’ are simply insufficient even to demarcate perceptual judgments from other judgments, let alone to capture the rational significance of the demarcation.”Footnote ²⁵ As noted, a perceptual judgement is a non-inferential judgement acquired directly through the senses. For example, we can make a perceptual judgement that p directly by seeing that p. In addition, seeing that p contains a distinctive phenomenal aspect that is absent in the case of merely thinking that p. Thus, a perceptual judgement is accompanied by a distinctive sensory experience. By contrast, a non-perceptual judgement is not a perceptual judgement because it is not acquired directly through the senses. One typical example of non-perceptual judgement is a judgement based on another person’s testimony. And such a non-perceptual judgement does not involve a distinctive phenomenal aspect that is present in the case of seeing that p. Therefore, we can distinguish perceptual judgements from non-perceptual ones by comparing these two different kinds of conceptualized states. For, unlike the case of a non-perceptual judgement, a perceptual judgement is acquired directly through the senses, and so it is accompanied by a distinctive sensory experience.

6. The Myth of the Non-Propositional Given

Finally, according to Gupta, although Sellars is correct in taking the propositional given as mythical, the non-propositional given is not a myth. He writes:

On Gupta’s view, one important reason the non-propositional given is not a myth is that no pure coherence theory can be satisfactory. As I discussed in Section 2, however, a Sellarsian coherence theory of justification is defensible. In addition, as I noted at the beginning of this paper, the purpose of this paper is confined to showing that a Sellarsian account of perception is still a viable approach to perception, despite Gupta’s powerful objections. Thus, let me here confine myself to briefly explaining why I reject Gupta’s key step for the non-propositional given, namely that something extra-conceptual in experience can make a rational contribution to the rationality of one’s perceptual judgement.

According to Gupta, S’s visual experience can make a rational contribution to the rationality of her perceptual judgement; and two factors are at play in her making a perceptual judgement: experience and view. On this account, S accepts a view; she undergoes a visual experience; and she thereby makes a particular perceptual judgement. Here the role of experience is to render rational transitions from views to judgements. Gupta explains the given in experience as follows:

1. (1) (Accept: v) → (Accept: J).

1. (2) Γ_e: (Accept: v) → (Accept: J). Footnote ²⁷

Furthermore, according to Gupta, the transition from a view to a perceptual judgement is analogous to the transition from the premises of a modus ponens argument to its conclusion. Observe that the transition from the premises of a modus ponens argument to its conclusion is rational, because it is a truth-preserving transition. In a similar vein, the transition from a view to a perceptual judgement such as (1) is rational because it is truth-preserving in the sense that if v is correct then J is true. He writes:

However, the above claims are problematic for the following reasons.

The first problem with Gupta’s view is concerned with the alleged rational transitions from views to judgements. As noted, according to Gupta, the role of experience is to render rational transitions from views to judgements. But to say that an experience renders a transition from a view to a judgement rational is to make a normative evaluation of the transition; and such a normative evaluation depends on some norms in terms of which we can determine whether such a transition is rational. But then, what are those norms? It is very doubtful that there are such norms. The reasons are as follows.

In the first place, as Sellars points out, “the concept of a reason seems so clearly tied to that of an inference or argument that the concept of non-inferential reasonableness seems to be a contradictio in adjecto.”Footnote ³⁰ Thus, the transition from a view to a perceptual judgement is rational only when it is analogous to an inferential process. If, however, it were an inferential process, the resulting perceptual judgement would be an inferential judgement. But, as already pointed out, perceptual judgements are not inferential judgements. Thus, the transition from a view to a perceptual judgement cannot be a reasoning process. In this important respect, this transition is not analogous to the transition from the premises of a modus ponens argument to its conclusion.

In the second place, as pointed out in Section 2, we are by nature rational beings whose beliefs are bound by the norms of rationality. And we may say that to be rational is to be bound by the norms of rationality (or the rules of reasoning). In other words, any of our activities would not be rational if they were not bound by the rules of reasoning. Moreover, the rules of reasoning are intersubjective. But the alleged transitions from views to judgements are not bound by any (intersubjective) rule of reasoning. On Gupta’s view, the role of experience is to render rational transitions from views to judgements; and the given depends solely on the subjective character of experience.Footnote ³¹ But there is no intersubjective rule that allows the transition from the subjective character of an experience to a perceptual judgement. Let me elaborate. Nowadays, it is widely believed that human beings are not born as (fully) rational beings. Accordingly, human beings must be trained and educated to become (fully) rational beings. And to say that we have become rational beings implies that it is our second nature to believe or act in accordance with the rules of reasoning, at least to some reasonable extent. For example, it is our second nature to follow the rule of modus ponens; accordingly, insofar as we have successfully been trained to follow this rule, just by accepting the premises of a modus ponens argument in the way required by this rule, we thereby believe its conclusion. But one’s subjective character of experience is not the kind of thing that can be evaluated intersubjectively.Footnote ³² As a consequence, our normal education and upbringing processes do not include a training process by which we can make a rational transition from views to perceptual judgements on the basis of the subjective character of experience. Instead, we are trained to make intersubjectively justified perceptual judgements in such a way that intersubjectively unjustified perceptual judgements can be subject to rational criticism. For example, one is trained to make a perceptual judgement like ‘there is a red object in front of me’ in just those situations where other people are allowed to make the same kind of perceptual judgement in our social practice of justification. Along these lines, we may argue that there is no intersubjective rule that allows the transition from the subjective character of an experience to a perceptual judgement.

In addition, our intersubjective rules of reasoning are what we adopt, and so they are in principle fallible. For example, consider Gottlob Frege’s Basic Law V, according to which the extension of the concept F is identical with the extension of the concept G if and only if the very same objects fall under the concepts F and G. He thought it as a truth of logic, and so, in his Basic Laws of Arithmetic, he tried to derive the fundamental axioms and theorems of number theory partly on the basis of this law.Footnote ³³ But, as is well known, it was subject to Russell’s Paradox. As this example shows, we could be mistaken even about what we take to be a truth of logic (or a rule of reasoning). By contrast, Gupta endorses what he calls ‘the reliability constraint,’ according to which the given in experience is not erroneous. This implies that the alleged transitions from views to judgements are immune to rejection. This, in turn, means that these transitions are not bound by any intersubjective rule of reasoning. If these considerations are correct, there are no good grounds for thinking that an alleged transition from a view to a judgement is rational.

The second problem with Gupta’s view is concerned with how something that has no correctness conditions can make a rational contribution to the rationality of a belief. According to Gupta, an experience is neither correct nor incorrect; nonetheless, the synthetic additions brought about by experiences can enrich a view. He writes:

Having a sensory experience is a causal response to some external stimuli, rather than a normative response. In this regard, it is worth recalling that there are no intersubjective rules of reasoning that govern our experiences so that our experiences are not subject to rational criticism. Thus, as Gupta points out, a sensory experience by itself does not have correctness conditions. But as José Bermúdez points out: “The notion of content stands or falls with the idea of correctness conditions.”Footnote ³⁶ And Gupta also does not attribute any content to an experience. Thus, we may say that an experience has no content. A question then arises: given that a sensory experience is a causal response to some external stimuli, how can such a causal response that has no correctness conditions make a rational contribution to the rationality of a perceptual belief? To put the same point another way, how can such a causal response that has no content bring about a synthetic addition to the contents of a view? How is it possible that a rabbit (some additional content) is pulled out of a hat (something that has no content at all)?Footnote ³⁷

The third problem with Gupta’s view concerns whether the justificatory status of one’s perceptual judgement depends in part on something extra-conceptual in experience. The rationality of a perceptual judgement must be intimately related to the justificatory status of the judgement in our social practice of justification. For example, consider the following case. The object in front of S looks red to her. Let us call this object ‘a.’ But all other normal observers sincerely report that a looks white to them. In such a case, all of us including S can decide to examine a’s colour in the standard condition for examination of colour. If a looks white to standard observers in standard conditions, S is not entitled to claim that a is red. As this example shows, whether one’s perceptual judgement is justified is evaluated in terms of what can be intersubjectively ascertained rather than one’s subjective experience. Note that only what can be intersubjectively ascertained can be presented as a reason for a belief in our social practice of justification. Therefore, only what can be intersubjectively ascertained is relevant for determining whether a perceptual judgement is justified. This implies that the subjective dimension of experience plays no justificatory role in our social practice of justification. Along these lines, Sellars (or Sellarsians) can deny that the justificatory status of one’s perceptual judgement depends in part on something extra-conceptual in experience.

To sum up, until and unless the questions raised in this section are adequately answered, Sellars (or Sellarsians) can reject Gupta’s claim that something extra-conceptual in experience can make a rational contribution to the rationality of one’s perceptual judgement.

7. Concluding Remarks

Admittedly, some aspects of Sellars’s theory of perception are problematic. For one thing, our epistemic rationality does not depend on the fact that we are effective agents in the world. For another thing, there are good reasons to reject Sellars’s view that the ultimate referent of a perceptual taking is always a sense impression. Nonetheless, if my arguments in this paper are on the right track, a Sellarsian account of perception is still a viable approach to perception. In particular, Sellars (or Sellarsians) can provide defensible answers to the following questions. What are perceptual judgements? What is the relation between a perceptual judgement and a perceptual taking (understood as a taking as)? How are perceptual judgements related to sense impressions? How can we distinguish perceptual judgements from the corresponding non-perceptual judgements? In addition, Sellars (or Sellarsians) can reject Gupta’s claim that something extra-conceptual in experience can make a rational contribution to the rationality of one’s perceptual judgement.Footnote ³⁸