Andrew Chrucky, Critique of Wilfrid Sellars' Materialism, 1990

CHAPTER 9

PERCEPTION

      My main interest in this chapter is to argue that an analysis of perceptual reports about physical objects yields phenomenal (sense data) reports which are foundational for empirical knowledge. This may be called the epistemological problem of perception. But I am also interested in examining the ontological problem of perception, which is the problem of giving an explanation of both veridical and unveridical perceptions. The crucial distinction here, for Sellars, is between giving an analysis and giving an explanation. It is Sellars' claim that this distinction has often been conflated -- creating a confusion. Epistemology is concerned with the analysis of perception; ontology with its explanation.

      The key question is whether a sense data analysis of perception is possible. If it is, then sense data reports can provide foundations for empirical knowledge -- a thesis conceded by Sellars. On the other hand if the explanation of perception requires the postulation of something like sense data, then these must find a place in a Materialistic ontology. Sellars is interested in solving both the epistemological and the ontological problem. His considered judgment on the epistemological problem is that analysis of perception does not yield sense data; while an explanation of perception does require the positing of something like sense data, namely sense impressions (or sensings) construed adverbially as states of people. And in the Scientific Image these sense impressions are recategorized as sensa, which are absolute processes.

      I find no difficulty in accepting Sellars' explanation of perceptual reports as causally mediated by sensa; but I do find difficulty in accepting his reasoning for rejecting sense data through an analysis of perceptual reports. Counter to Sellars I accept the availability of sense data through phenomenological reduction (if that can be considered an analysis), and, furthermore, I accept the validity of the so called Sense-Datum Inference. Since Chisholm has formulated an explicit argument against this inference, I will defend it from his criticism.

I. PERCEPTUAL REPORTS

      The first task is to understand the nature of perceptual or observational reports as such. In the Manifest Image we say that we perceive physical things having proper and common sensibles (Aristotle's way), or having primary and secondary qualities (Locke's way). If we add that physical objects really have the properties which we perceive them to have, we have what is commonly called Naive or Direct Realism. The Manifest Image includes Naive (Direct) Realism as its epistemological and ontological bedrock. However, Naive or Direct Realism has a difficulty in accounting for the phenomena cited in the so-called arguments from illusion, relativity, time-lag, the causal chain; and such mental images as hallucinations, after-images, eidetic images, and phosphenes. The argument from illusion uses such cases as, for example, when we see something that looks like A, or that we mistakenly believe to be A, but which on closer inspection it turns out that what we were looking at was something else, B, or there was nothing there at all to be seen.

      In view of the possibility of unveridical perceptions, the classical approach to the problem of perception is to reject Naive (Direct) Realism. This is accomplished by accepting an ontology that makes qualitative properties subjective. This is offered, first, as either an analysis or as an explanation of unveridical perceptions, and then the analysis or explanation is extended to account for what happens in veridical perceptions as well.

      Were these arguments from 'illusion' successful? Frank Jackson in his admirable book Perception{1} points out that these classical arguments beg the question. His reason is that in all such arguments we can at most conclude that there is a need to distinguish appearance talk from reality talk, and this no one disputes. What is, however, in dispute, he points out, concerns the possible ontological and epistemological inferences that can be made from particular kinds of appearance talk.

A. DISAMBIGUATION OF 'PERCEPTION'

      To state the problems of perception more clearly, and to avoid arguing at cross purposes, we first have to disambiguate perception terms because they can be used in contexts other than the ones philosophers are interested in.

      To get oriented in perception talk, we have to make a number of classifications and distinctions. I will do this from a formal perspective. There is no necessity to examine particular cases in ordinary language, since all such cases can be claimed to fall under one or another of the forms to be distinguished. And for my purpose, which is to determine which forms allow which inferences, it is sufficient to concentrate on the forms.

      First there are uses of 'seeing' which are not perceptual. In the example 'I see what you mean', 'seeing' is being used almost synonymously with 'understanding'. In the example 'I see that there is still time' may be paraphrased as 'I conclude that . . .' The uses of 'see' and other perceptual verbs which philosophers are interested in require a relevant type of sensorial stimulation. Exactly how to formulate this relevancy is a bit of a problem. However, in the literature on perception, we find two sets of distinctions constantly being used. The first is between propositional and non-propositional forms; the other between opaque and transparent forms. I would like to distinguish these so that I can locate myself and Sellars in this classification.

      We may begin by distinguishing on the basis of surface grammar between propositional and non-propositional types of perception. Propositional perception can be characterized linguistically as containing a reference to a perceiver, a psychological verb of perception (e.g.,'seeing', 'hearing', etc.), and having as its accusative a propositional clause: thus, 'I see that this ice cube is pink'. While non-propositional perception takes as its accusative a name or a descriptive phrase, as, for example, 'I see this pink ice cube'.

      To understand perception talk reflectively, we must be able to state the kinds of inference patterns perceptual reports enter into. The first inference pattern common to all normal uses of perception talk (both propositional and non-propositional) is that they entail that what is perceived is true and that the object of perception exists; thus, 'S sees that D is F' entails 'D is F' (which in turn entails 'D exists') and 'S sees D' entails 'D exists'. This is an expression of what Gilbert Ryle, and Sellars in agreement, call the 'achievement' sense of perception.

B. SURFACE GRAMMAR: PROPOSITIONAL PERCEPTION

      I will first distinguish various forms of propositional perception without trying to reduce or subsume one form under another (which may be possible).

  • (A) S sees that this is a(n) D.
  • (B) S sees that this is F.
  • (C) S sees that this D is F.
  • (D) S sees that this is the D'.
  • (E) S sees that this D is the D'.
  • (F) S sees that this is N.
  • (G) S sees that this is he.
  • (H) S sees that this D is he himself.
  • (I) S sees that this is something.
  • The letters are variables that take the following types of substituends: 'S' for persons, 'D' for common names, 'D' for definite descriptions, 'F' for properties (monadic and polyadic), 'N' for proper names.

          All forms of perception have demonstrative references. The reference to a place is expressed by the demonstrative 'this', while the reference to a time is expressed by the tense of the verb. We could make this explicit by rewriting (A), for example, as 'S sees that this here, now is a(n) D'. Let me quickly add that although I have distinguished these various forms, I cannot do them justice here. That would require a separate treatise. All I can do is focus on a particular form that is essential for a critical appraisal of Sellars. This form is (C). As I have already cautioned the form itself can be confused with what I would call a non-perceptual form. For example, it could be confused with something like

    S sees that 2 + 3 = 5.

          In this example the word 'see' has a non-perceptual use. Here the word is used synonymously with something like 'understand' or 'realize'. What is wanted for perceptual cases is the acquisition of information through the senses. But again we must be cautious because of cases like

    The doctor, by looking at John, sees that John is sick.

    Here the difficulty is that the information that John is sick is apparently inferred. We want to isolate a primary sense of propositional non-inferential seeing, or, which may be called primary propositional seeing. With these precautions we can distinguish within propositional perception between a transparent and an opaque form.

    1. TRANSPARENT FORM

          The transparent form allows substitution of co-referential terms in the subject of the accusative. To exhibit the underlying meaning of such transparent propositional perception claims, I will rewrite

    S sees that this D is F.

    as,

    S sees of this D that it is F.

    or

    S sees this D to be F.

    From this form we can infer

    S sees that something is F.

    or

    (Ex) S sees that Fx.

    For 'x' we can substitute freely any co-referential term, description, or indexical without affecting the truth value of the statement. A person may indicate the x in various ways, as by pointing or describing without committing himself to the sort he is referring to.

          And since we are working with the achievement sense of perception it follows from

    S sees that this D is F.

    that

    (Ex) Fx

    2. OPAQUE FORM

          The opaque form, by contrast, does not allow substitution of co-referential terms in the subject of the accusative proposition. Quantified,

    S sees that this D is F.

    becomes,

    S sees that (Ex) (x=D) & F(x)

    Substituting different co-referential terms for the x will result in an altered truth value for the whole statement.

          Again, because we are assuming the achievement sense of perception, from this form we can infer

    (Ex)[(x=D) & F(x)]

    and

    F(D)

    C. SURFACE GRAMMAR: NON-PROPOSITIONAL PERCEPTION

          By contrast, in the non-propositional form of perception the verb takes as its accusative the name or description of some object or event. Let us focus on the form

    S sees this D.

    As with propositional perception, we may distinguish a transparent (non-epistemic) and an opaque (epistemic) form. Non-epistemic perception was distinguished by J.F. Soltis, G. Warnock, and F. Dretske,{2} and is now usually referred to as 'simple seeing', and will be designated by 'seeing(n)'. Simple seeing can be distinguished from epistemic (opaque) seeing (designated 'seeing(e)') by the following traits:

          (1) Simple seeing allows substitution of perceptually possible co-referential terms for the objects of perception; while opaque (epistemic) perception does not.

          It should be noted that the types of co-referential terms are restricted to those which can be used perceptually, as qualified by F. Jackson in the following passage:

    I suppose 'see' to be such that 'S sees D and D = D'' entails that S sees D'. That is, 'D' is here subject to substitution (of co-referring terms) and so 'S sees ___' is a transparent construction. {3}

    Dretske expresses the requirement this way: "if S sees(n) D, and Di is an associated description of D such that S can seen Di, then S sees(n) Di."{4} The restriction is meant to exclude such cases as cited by Cornman in the following passage:

    if any volume of water is identical with some conglomeration of H(2)O molecules, then seeing, tasting, and feeling water, we are seeing, tasting and feeling a conglomeration of H(2)O molecules. {5}

          If we want to include this as a separate type of seeing, we may label it 'seeing(u)', where the subscript 'u' is to remind us that it is an unrestricted type of seeing. Sellars, however, rejects this approach as trivial.{6}

          (2) Opaque (epistemic) perception requires an inference from 'S sees D' to 'S believes that he sees D'; while simple seeing does not.

          (3) Dretske's criterion for simple seeing is that D be visually differentiable. Apparently Chisholm also allows for this kind of seeing when he writes:

    there is no paradox involved in saying that a man sees a dog without taking what he sees to be a dog. It may be, however, that we should hesitate to say that he sees a dog if he didn't take it to be anything at all.{7}

          (4) Simple seeing also allows for a person seeing(n) something and yet not being aware of seeing(n) anything. This may happen in what is called subliminal seeing. For example, suppose the a number is flashed on a television screen at a speed below the threshold of conscious discrimination. We then ask the person to tell us a number that comes to his head. If he comes up with the number flashed, there is reason to say he saw it subliminally or subconsciously.

          (5) I take it that Dretske accepts the following analysis of seeing(n):

    'S sees(n) D' entails 'D appears(n) f to S' (in which 'appears(n)' is a non-epistemic appearing).

          Because cases can be imagined where a person can visually differentiate something, as in an after-image, yet not be causally stimulated by stray radiation hitting his eyes, which in this case the cause may be internal neuronal activity, it follows that the normal conditions of seeing are only contingently related to seeing(n).

          On the basis of such cases, Dretske explicitly remarks that the substitution instances of 'D' are not to be restricted to physical objects, but may include such things as after-images, mirages, phosphenes, and hallucinations.{8} Consequently he also rejects all causal analyses of seeing(n); in particular, he rejects Chisholm's analysis in Perceiving,{9} saying that Chisholm's analysis does not reflect common sense use.{10}

          Dretske's critics have taken issue with this characterization of this non-epistemic or simple seeing. One issue is whether no beliefs whatever, or no particular beliefs are entailed by 'S sees(n) D'. As I understand this, Dretske requires that the subject of non-epistemic perception have the capacity to have conceptual responses to his experiences, including the capacity to have beliefs. But no particular belief is entailed or relevant.

          In general, then, although I have divided perceptual reports by considering their forms into propositional and non-propositional perception, the form itself may be misleading as to the logical status of such reports. Logically speaking,

    S sees D.

    is either an elliptic propositional report, which can be expressed as 'S sees that this is D', or it is a case of simple seeing.

    II. SELLARS' POSITION

          I will now examine Sellars' position in terms of the distinctions that I have made. Does Sellars' position fall within these rubrics? The answer is that he is, for the most part, conscious of these distinctions and has a place for them.

          He remarks that the problem of perception can be viewed as an epistemological problem or as an explanatory (constitutive, causal, ontological) problem.{11} The epistemological problem concerns the analysis and justification of perceptual reports, and here the relevant forms are those of propositional perception. By contrast, the constitutive problem involves explaining the ontological conditions for the possibility of both veridical and unveridical propositional perceptual claims, and here the non-propositional non-epistemic forms are relevant. In fact Sellars' concerns about whether the sensa (sense data) or sensing (adverbial) theory is correct is a problem about the status of non-epistemic perception. And although the constitutive problem of perception plays a major role in Sellars' philosophy, he does not call it the "problem of perception."{12} The traditional problem of perception, he tells us, is an epistemic problem.

          In view of this bifurcation, Dretske's attempts to give an 'analysis' of primary epistemic seeing in terms of a non-epistemic seeing in the following way:

    S sees that D is f =
    (1) D is f.
    (2) S sees(n) D.
    (3) The conditions under which S sees(n) D are such that D would not look, L, the way it now looks to S unless it was f ('L' represents the totality of appearances of D to S).
    (4) S, believing the conditions are as described in (3), takes D to be f.
    {13}

          This would be rejected by Sellars as an analysis -- on the ground that epistemic facts cannot be analyzed by non-epistemic facts. However, Sellars could (with qualification) accept this as an explanation.

          Classical approaches to perception have tended to appeal to appearances, without clearly distinguishing analyses from explanations. But exactly how to construe this appeal, and to determine the entailments is tricky. Sellars, in a rare expression of confidence, writes that the pendular swing of philosophical controversy about the problem of perception can be made to stop because "the tools are at hand for a decisive clarification of traditional puzzles about perceptual knowledge."{14} I am not so confident.

    III. EPISTEMOLOGICAL PROBLEM

          The epistemological problem is to analyze both veridical and unveridical cases of perception. And to determine whether analysis provides us with what can be used as foundations for empirical knowledge. In view of the possibility of unveridical perceptions of physical objects, if there are to be foundational perceptions, they must surely be, in some sense, minimal. So the issue of perceptual analysis is, as I see it, a question of perceptual reduction or minimalization.

          Let us start with a veridical opaque propositional perceptual reports that S sees that this D is F, e.g., S sees that this apple is red. In several essays, Sellars distinguishes within the propositional content a component that is presupposed or believed in -- namely, the subject of the thought, D; and the component which is believed that -- namely, the predicate, F, of the thought.{15} His reasoning is that in a case such as seeing that this ice cube is pink, I presuppose that I am looking at an ice cube. And I may have learned this not on the basis of seeing, but because you told me it was ice. So, presupposing it is an ice cube I am looking at, I see that it is pink.

          How can such reports be minimalized? I will summarize Sellars' position as a series of steps, and offer commentary. The first step in his analysis is to say that, at least for opaque objects, we can distinguish the object and its parts. Sellars says that what we see of the object D is actually part of D -- the facing surface, D' -- and we may see that D' is F.{16}

          The above formulation is a bit coarse because it conceals distinctions. The question comes to whether the following inference is valid:

    S sees that D is F. (opaque form)
    D' is a part of D.
    D' is F.
    Therefore, S sees that D' is F.

          This obviously won't do because there are hidden parts -- like the inside of an apple which is not seen even if we tried noticing it. So the way to overcome this objection is to add a premise, namely,

    D' can be seen by S.

          Now the problem is that although S could see D', yet he may not be paying attention to any part of or any specific part of D; so that as a matter of fact he does not notice D'. So we have to add the premise,

    S notices D'.

    But D' itself may have parts which are not F; so even if S notices D', it does not follow that S notices that D' is F. So what we actually need is a new perception, namely,

    S sees that D' is F.

    But this line of reasoning shows that it is not the case that 'S sees that D is F' (in the opaque sense) implies 'S sees that D' is F'. What it shows is that 'S sees that D' is F' is a new perception.

          Sellars seems to be mixing epistemic and non-epistemic uses of perception. His remarks makes sense with the transparent mode of propositional perception. But even here some of the above qualifications apply

    S sees that D is F. (transparent form)
    D' is a part of D.
    D' is F.
    S can be seen by S.
    Therefore, S sees that D' is F. (transparent form)

    The transparent mode of perception requires for its explanation the inclusion of non-epistemic perception, such that 'S sees that D is F' implies 'S sees(n) D'.

          Yet in some sense it is true that I see that D is F because I see that D' is F. But to say this is to depart from giving an analysis to giving an explanation. Jackson handles this problem nicely by introducing the 'in virtue of' relation.{17} It is meant to express, at least for part-whole relations, that the whole is a whole because (in virtue of) the parts standing in certain relations to each other. So the above line of reasoning would be transposed to the claim that S sees that D is F in virtue of seeing that D' is F.

          But we can maintain the epistemic relation between the opaque forms of perception of wholes to the perception of parts by noting the possibility of switching perceptions from wholes to parts. Call this perceptual reduction.

          This whole-part problem presents additional difficulties. How much and what parts of an opaque object have to be seen before we can say that we see the object? This is a difficulty because we never see the whole object with its inside and back; but only a part and facing surface.{18} Descartes, for example, tackled this problem when he wondered whether he saw people below on the street as he watched from an upper floor window, because what he saw were moving cloaks. {19} Sellars does not discuss the above problems, but his explanation of perceptual reports is consistent with these considerations.

          However, in his endeavor to provide an analysis, he goes on to distinguish parts from properties of physical objects. And within properties he distinguishes between the dispositional or causal properties, on the one hand, and the occurrent properties (Aristotelian proper and common sensibles), on the other.{20}

          The second step in his analysis is to claim that we do not see the causal or dispositional properties; we only see the occurrent ones.

          As a sympathetic reconstruction of Sellars' intentions, his combined claim would be that when we look at D, we can through perceptual reduction see D' of D, and we can see that D' is F (where 'F' is an occurrent property). So, for example, when we look at a red apple, we can see of it it's facing surface and we can see that it is red.

    A. APPEARANCE TALK

          The next problem in analyzing perceptual claims is to find a place for appearance talk. Without rehashing the arguments from illusion, considerations of the causal chain in perception, the time-lag argument, and so on, I simply assume that there are cases of unveridical perception. When we talk about unveridical perception, it must be to indicate a failure of one of the modes of epistemic perception we have distinguished, namely an opaque or transparent forms of propositional perception; or the epistemic form of non-propositional perception. But it seems to make no sense to talk about an unveridical mode of non-epistemic perception.

          Unveridical perception or doubt of the veracity of perception is normally signalled by the use of 'appear' words. These are such words as 'seems', 'looks', 'feels', 'smells', 'sounds', 'tastes', and their cognates. The generic form of these words are 'appear' and 'seem'. But we should distinguish between perceptual and non-perceptual uses of appear words.

          Chisholm has distinguished three uses of appear words,{21} and Sellars has expressed agreement at least with the distinctions -- if not with their analyses.{22} The first use of appear words is epistemic and distinguished by sanctioning the inference from 'D appears to be F to S' to 'S believes that D is F'. The role of 'appears' in this use is synonymous with 'seems'.{23} It is used primarily as an expression of judgment e.g., 'It seems that D is F' is equivalent to 'I judge that D is F'. C. Mundle, in a similar vein, talks about as 'estimative use' of 'seems', as, for example, 'It seems that this is larger than that'.{24} By contrast, the use of 'appear', though synonymous with the use of 'seem' in some contexts, has generally different uses.

          The second use of appear words distinguished by Chisholm is comparative. Using the comparative sense, from sentences such as 'D appears F to S' we can infer 'D appears like F-things usually do (in standard or optimal contexts)'.

          The third use of appear words is noncomparative. This use does not allow the types of inferences sanctioned by the epistemic or comparative uses. For example, if 'D appears F to S' and S suspects, let's say, some abnormality in the lighting, then S may not believe that D is F. As to the relation of phenomenal uses of appear words to the comparative uses the situation seems to be this. Comparative uses of appear words presuppose noncomparative uses.

          I suspect that Chisholm's noncomparative use of appear words has two species, but which Chisholm does not distinguish at the appropriate place of his discussion. One subspecies can be identified with what Sellars generically refers to as 'ostensible perception', which has the species 'ostensible seeing', 'ostensible hearing', etc., to correspond to various sensory modes. To say that there is an ostensible perception is to indicate a suspension of judgment as to whether a perceptual report is true or not. Sellars put it this way:

    P ostensibly sees a cube of pink facing him/her edgewise as a cube of pink facing him/her edgewise is, for our purposes, identical with the idea that
    There appears to P to be a cube of pink facing him/her edgewise.
    {25}

          We could also characterize it as the use of a perceptual report with the achievement sense suspended. Firth's use of the term 'percept' in his "Sense-Data and the Percept Theory", I take it, is synonymous with Sellars' 'ostensible perception'.

          The other subspecies of noncomparative uses of appear words is what Chisholm in his Theory of Knowledge calls the "descriptive, phenomenological use."{26} The objects of phenomenological appearances are the proper and common sensibles of Aristotle.{27} Frank Jackson calls this the phenomenological or, for brevity, the phenomenal use, which is characterized "by being tied to terms for color, shape, and/or distance."{28}

    B. CRITICISMS OF THE DISTINCTIONS

          Those who are opposed to the introduction of sense data as phenomenologically given (and hence as essential to a foundationalist view of empirical knowledge) have realized that the first step in the defense of sense data is to recognize the distinct character of phenomenal appearance talk, and have tried either to eliminate phenomenal talk altogether or reduce it either to the epistemic or the comparative use, or have claimed that phenomenal talk is inscrutable. Jackson mentions the following philosophers who take these approaches: J. Bennett, D. Armstrong, J. W. Roxbee-Cox, and G. Pitcher. According to Jackson, Bennett offers a radical reduction to the comparative form. Bennett, according to Jackson, claims that 'x appears y to S' implies 'x produces in S the kind of sensory state S usually is in when he perceives x as y'.{29} Pitcher and Roxbee-Cox offer an epistemic reduction, as follows: 'x appears y to S' implies 'S acquires belief z' or 'S causally receives belief z' or 'S believes z'; 'x appears y to S' implies 'S believes x is y' or 'S is inclined to believe that x is y' or 'S has a suppressed inclination to believe that x is y'.

          According to Jackson, Armstrong offers a counterfactual epistemic analysis. He says that 'x appears y to S' is equivalent to 'But for the fact that S had other, independent, beliefs S would have acquired the belief that x is y'.{30}

          Richard Rorty too offers an epistemic reduction. He writes:

    For the appearance-reality distinction is not based on a distinction between subjective representations and objective states of affairs; it is merely a matter of getting something wrong, having a false belief.{31}

          Another attack on the phenomenal use of appear words -- a move, I may point out, not sanctioned by Sellars -- is to take a skeptical stance about the possibility of demarcating them. We find Dretske, who is probably thinking of Goodman's grue paradox, expressing the following skepticism: "I know of no non-circular way to specify what is to count as a visual characteristic."{32} Sellars, by contrast, cites Aristotle's proper and common sensibles as the appropriate occurrent sensorial properties. But he insists that they always have a physical sense in perceptual contexts.

          As far as I can tell, Sellars' efforts against the introduction of phenomenal sense data, and, hence, against foundations of empirical knowledge are crystallized in his denial of a phenomenological or phenomenal use of appear words. In his polemics against Chisholm, Firth, and Cornman, Sellars claims that it is either synonymous to 'ostensible perception', or that such a use of appear words does not exist, or that it is a contrived use to describe non-conceptual states.

          Sellars' approach to this issue of analyzing perceptual reports can be developed on the basis of analyzing the following three statements from Sellars (with some minor modifications on my part):

    (a) S sees that D, over there, is F.
    (b) It looks to S that D, over there, is F.
    (c) It looks to S as if there were an F(D) over there.{33}

    Sellars considers the entailments of these three types of reports. Among these entailments he distinguishes a 'propositional content' and a 'descriptive content'.

    C. PROPOSITIONAL CONTENT

          The propositional content entailed by all three reports is 'S has the thought that D, over there, is F'. The three reports differ in their implied judgments about this propositional content: (a) endorses it, (b) partially endorses it, and (c) does not endorse it.

          Sellars' approach seems to suggest -- at least for the appearance reports -- that they are properly made only if they imply the D-or-D condition ('doubt or denial condition') as examined by H. P. Grice in his "The Causal Theory of Perception". According to this approach:

    When someone makes such a remark as "It looks red to me" a certain implication is carried, an implication which is disjunctive in form. It is implied either that the object referred to is known or believed by the speaker no to be red, or that it has been denied by someone else to be red, or that the speaker is doubtful whether it is red, or that someone else has expressed doubt that it is red, or that the situation is such that though no doubt has actually been expressed and no denial has actually been made, some person or other might feel inclined towards denial or doubt if he were to address himself to the question whether the object is actually red.{34}

          The question Grice asks is whether there is a use of such expressions as 'It looks red to me' where the D-or-D conditions are missing. In other words is there a distinct phenomenal use of appear words? To settle this question, Grice distinguishes four types of implications, and considers four different possible vehicles of implications:

    (a') what the speaker said (or asserted), or (b') the speaker ("did he imply that . . ."), or (c') the words the speaker used, or (d') his saying that (or again his saying that in that way); or possibly some plurality of these items.{35}

          Instead of following Grice in his exposition, we can invoke a distinction that Sellars uses between what he calls semantical or S-assertibility{36} and dialectical or D-assertibility.{37} S-assertibility consists in those statement utterances which are sanctioned by formal and material rules of implication. D-assertibility, on the other hand, consists of those utterances "grounded in the requirements of unambiguous communication."{38}

          Grice is, in effect, classifying (a') as an S-assertible utterance, and the rest as D-assertible utterances. D-assertibility is determined by background rules of what may be called normality, sincerity, and appropriateness. It probably involves standards of etiquette and ethics. Within such a background or context, it makes a difference who says what to whom, when, where, etc., etc. In any case, normally one does not say 'It looks red to me' in an informative manner unless the D-or-D conditions are satisfied. But I think this applies to any remark which would communicate something that is obviously seen or known. For example, suppose everything is normal and it is obvious that we are both looking at a red ball, what would be the point of my saying to the other person 'This is a red ball', if I know that he knows that it is a red ball, and I know that he knows that I know that it is a red ball? Saying it would express a truth, but it would be contextually or D-assertibly odd. Now it seems that the same sort of oddity would occur in the same context if I said 'This ball looks red'. It would be true, but pointless. There would be no S-assertible oddness in saying 'This ball looks red and I believe that the D-or-D conditions do not exist'. And if this is an intelligible thing to say, then there seems to be a phenomenal use of appear words.

          Sellars disagrees. If I understand him correctly, he would say that if the D-or-D condition as an implication of an appearance statement is dropped, then you are left with a straightforward report that x or something, anyway, is F. In other words, the 'looks' use in this context becomes a contrived way of speaking for something being F.

          We can glean Sellars' answer from how he analyses the following, which he takes to be a necessary truth:

    x is red. iff .x would look red to standard observers in standard conditions.

          This seems like an analysis of 'x is red' in terms of' x looking red'. After all, the right hand side is composed of three claims which can be separated, namely 'x looks red to S in conditions C', 'S is a standard observer', and 'C are standard conditions'. And it also seems that the first conjunct could be further analyzed into 'x looks red to S' and 'the conditions are C'. And it also seems to be the case that the former conjunct can be analyzed as a triadic relation between x, red, and S.

          Sellars disagrees with the above that 'looks' statements are analyzable, and this for the simple reason that 'looks' is not a relation word. It's logic is tied up with the presence or absence of standard observers and conditions. In that case, it would seem the use of 'looks' is tied to the seeing of something. And the focus of attention has to shift to primitive modes of seeing.

    1. DESCRIPTIVE CONTENT

          Sellars would agree that appear statements as well as perceptual reports imply, as he puts it, a common descriptive content, which would be that S sees that D is F on the condition that the propositional content is true and S knew that the circumstances were normal.{39} But Sellars admits that such a counterfactual analysis does not specify the occurrent descriptive content of the given reports.

          He also notes that the common descriptive content can be called an 'experience'. But he immediately cautions us about the possible ambiguity between an experiencing and that which is experienced. Let us refer to these as 'act' and 'content' respectively. He concentrates on a description of the experienced content, and suggests first an externalist, comparative analysis, but rejects it as inadequate. He is troubled by how to characterize directly or intrinsically this content.{40}

          To understand why Sellars is troubled we must be conscious of the alternative which is he trying to avoid. This alternative is the classical account that something -- a sense datum -- is F. Keeping this as the terminus he is avoiding, Sellars -- as a moth to fire -- keeps returning to this issue.{41} The closest that Sellars comes to describing the descriptive content is when he writes: "If understood in a neutral way, Something, somehow a cube of pink in physical space is present in the perception other than as merely believed in.";{42} and: "Strictly speaking, in its character as perceptual object it consists of a volume of pink -- where 'pink' is, in Quine's sense, a mass term, a word for a kind of stuff."{43} Note that Sellars thinks that this reductive analysis still leaves us with physical matter in physical space. At best, according to Sellars, we have 'S sees a bulgy pink expanse'.

          Reflecting on this, Sellars writes that this is a minimal account:

    So minimal, indeed, that it might be thought that my account hardly differs from the traditional view that what we perceive of objects are sensa which somehow 'belong' to them.{44}

          But he goes on to say that this account constitutes an explanation and not an analysis. I disagree. This account can do double service.

          All Sellars is claiming is that the minimal account is conceptually tied to concepts of physical objects. But conceptual dependence is compatible with epistemic independence. The obvious example from Sellars' writings is the conceptual dependence of the Manifest Image on the Original Image. This is the claim that our concept of a physical object is a truncated concept of a person. But we can agree with that, and yet say that now we can use our truncated concept of a person to identify physical objects. In other words, for epistemic purposes the concept of a physical object is independent of the concept of a person.

          By analogous reasoning, we can say that though minimal perception, which truncates causal properties and parts unseen, is conceptually dependent on the concept of a physical object, it is epistemically independent of seeing physical objects.

          What warrants Sellars calling the objects of minimal perception 'stuff' as a physical category? Recall that a physical object was characterized by (1) being in time, (2) being in space, and (3) having causal properties. But in the above account of perceptual analysis, step two required an abstraction from causal properties; so that a minimal perception was of an occurrent property located in space and time. But in calling something 'stuff' suggests that it has causal properties. So it seems Sellars' account is either inconsistent, or he is using the word 'stuff' in the sense of parts; but again the sense of 'parts' cannot be in the sense of causally integrated parts, but merely the sense of spatially and temporally distinguished parts. But what is the point of calling the objects of minimal perception a physical entity? The only physical properties that we are aware of are location in space and time. But that sort of entity is no different from the sense datum of classical perception theory. Moreover, calling the object of such minimal perception either subjective or objective would also be improper since that would presuppose knowledge of the causal connections into which these entities enter. So neutralized, the minimal perception that Sellars is talking about becomes identical to Lewis' 'given' for which "a distinction between the subjective and the objective . . . is irrelevant to givenness as such."{45} I would also point out that the given would be neutral between being physical(1) and physical(2) since that distinction is a function of causal connections. Concretely, if Sellars means to claim that the predicate 'pink' in analyses cannot be reduced to anything more basic than physical pink stuff, as long as the word 'physical' does not carry causal implications, we can say, with Lewis, that the pink is given.

    2. RE-EXAMINING THIS CONCLUSION

          If my interpretation is unacceptable to Sellars, it must surely be that I have overstepped the boundaries of analysis and have entered the realm of theory and explanation. But, in that case, I am puzzled where the demarcation point is. If there is to be a reductive analysis for Sellars, it must be based on a model of something public in the Manifest Image. Such a model is a physical object or the facing surface of a physical object. So why won't this do an a model?

          It is clear that sense impressions as posited unapperceived entities are theoretical entities, but I am not sure whether Sellars would call the result of a phenomenological reduction a theory as well. In any case, it is clear that Sellars believes there is a limit to phenomenological reduction. I would raise the question, then, whether if that limit is overstepped we get into the realm of theoretical positing? Is there a limit to acceptable abstraction or truncation? Is this resistance consistent with Sellars' principles? It could be objected that Sellars is dealing only with the perception of physical objects and not with such subjective phenomena as after-images, hallucinations, and such. So that his analyses of the former do not apply to his analyses of the latter. But I think it would be wrong to take this tack. For Sellars concepts of subjective states are learned by analogy with concepts of physical objects. So it would seem that even subjective states have an irreducible physicality to them. After all our subjective states are felt as states of our bodies. Pain, for example, is located in parts of our bodies. Even phantom limb pains seem to be located in physical space.

          Sellars conclusion is acceptable in part as a phenomenological description of both our perceptions of physical objects, and of our subjective experiences as appearing to be located in physical space. This much is conceded by Firth in his examination of the percept and sense data theories.{46} Firth's 'percept' is what Sellars calls an 'ostensible physical object'. However, both Sellars and Firth ignore the availability of such subjectless events as flashes of light or the claps of thunder as models for sense data, or as subject of perceptual reduction. Given such a model we do not need to abstract the properties of physical objects from the causal dispositions of the physical substance. Subjectless events are ideal for this. Sellars recognizes this, but uses these models for theoretical construction. My point is that these models can be used for phenomenological purposes.

    3. RESTRICTIONS

          Let us see how Sellars places restrictions on phenomenological reduction. In the essay "Phenomenalism" he is considering three models by which to form a concept of sensing a sense datum. The first model he uses is that of seeing the facing surface of a physical object. Since I believe, contrary to Sellars, that this model will do, I will examine his reasoning for rejecting it.

    Such perception-theoretical expressions as 'directly see', 'directly hear', etc., are given a logic which parallels, in significant respects, the logic of the verbs 'to see', 'to hear', etc., as they occur in everyday perceptual discourse. Thus, to such statements as

    Jones saw a book and saw that it was blue.

    there correspond such statements as

    Jones directly saw a bulgy red expanse and directly saw that it was bulgy and red.

          And just as seeing that is a specific form of knowing that, a variety of observational knowledge, of observing or perceiving that, so directly seeing that is construed as a variety of directly observing or perceiving that, and, hence, as a specific for of directly knowing that. Again, just as seeing x is a form of perceiving x, so directly seeing is introduced as a specific form of directly perceiving x, or, as the term is introduced, sensing x.{47}

          So far, I have no objections. However his following claims are dubious.

    (AT) The fact that 'sensing x' is introduced on the model of 'perceiving x' as ordinarily used brings with it a number of implicit commitments not all of which can be dodged without cutting the theory off from the roots of its meaning.{48}

          Let me remind the reader that Sellars, writing on another occasion about models and theories, stressed that a theory is constructed by using positive, negative, and neutral analogies.{49} Moreover he made a distinction between material (or first-order) and formal (or second-order) analogies. Given these analogical resources, there is much leeway in the construction of theories on the basis of models. A theory is, then, introduced on the basis of a model and what Sellars calls a 'commentary'. A commentary is simply the statement indicating which features are similar, different, neutral, and the order-level which is operative. In view of this, how are we to deal with the last cited passage? Sellars seems to be suggesting that a number of similarities have to be included if the model is to be used at all. To make sense of this, I will interpret Sellars to be making the following presuppositions:

    (1) The model has an essence.
    (2) All or some of the essential properties must be included in the positive analogy.

          My concern is with (2). Let me cite an example of categorial refinement which Sellars uses and which seems to violate this prescription. He speculates that the Manifest Image is the result of introducing the concept of a physical object by truncating the concept of a person. Specifically a person's actions are spontaneous or habitual. A physical object is derived from the concept of a person by truncation the spontaneous actions -- leaving only habitual actions. Now if the truncated property of spontaneous activity is essential to the concept of the model (i.e., a person), then isn't the introduction of physical objects a result of "cutting the theory off from the roots of its meaning"? If it is, I find nothing objectionable in this. So Sellars must have something else in mind.

          Frank Jackson has a related worry. He considers truncating the concept of red from the concept of being colored:

    Suppose someone writing on secondary qualities who observes that 'X is red' entails that X is colored, and decides to introduce the term 'red*' to mean precisely what 'red' means except that 'X is red*' does not entail that X is colored.{50}

          If this is the sort of truncation that Sellars has in mind which would be a "cutting the theory off from the roots of its meaning," then I agree that these kinds of implications cannot be removed. But I don't see that this applies to the model of a physical object.

          The problem seems to be to decide how to preserve the intelligibility of a theory when using negative analogies with a model. I suppose the problem is analogous to the problem of how much of the Cheshire cat can we remove before it is absurd to say that there is a smile on (of) a cat. This would suggest that we must preserve at least a subject-predicate form.

          Let's reflect on another use of analogy is Sellars that caused difficulty. Bruce Aune{51} found difficulty in accepting Sellars' facing surfaces of physical objects as models for sense impressions. Facing surfaces are, after all, things; yet Sellars conceived sense impressions as states of persons. Such a trans-category analogy seemed absurd to Aune. The reason Aune had for saying this is that the logical form of the surface and the physical object is a subject-predicate form, 'Fx'; while the logical form of a person having a sense impression seemed to Aune to have a relational form, 'xRy'. Sellars correction was to give an adverbial analysis of a person having a sense impression, as, for example, 'person S sensing redly', so that this formulation too had a subject-predicate form.

          The upshot of this discussion seems to be this. In using analogies for theoretical purposes, formal analogies must be preserved.

          With these precautions in mind, lets examine what implications Sellars has in mind for physical objects as models for sense data which cannot be rejected without "cutting the theory off from the roots of its meaning."

          One such commitment rests on the fact that in ordinary perceptual discourse the consequence from saying that

    Jones saw a book.

    to

    There was a book (i.e., the one that Jones saw).

    is valid. The theory, thus introduced, brings with it, therefore, a commitment to the consequence from

    Jones sensed a red and triangular expanse

    to

    There was a red and triangular expanse (i.e. the one that Jones sensed).{52}

          I find nothing wrong with this. It is a variation of the Sense-Datum Inference, which I will defend against its alleged refutation by Chisholm -- a refutation apparently accepted by Sellars.

    Another commitment rests on the fact that in ordinary perceptual discourse the objects of perception typically exist before they are noticed and after we have turned away; in short they can and do exist unperceived. The theory, introduced on this model, brings with it the implication that the red and triangular item which Jones sensed is capable of existing unsensed.{53}

          Again I find nothing objectionable here except for an ambiguous use of the term 'sensed'. In one use, it is synonymous with 'noticed'; in another, it could be construed as a state of the person. Sellars is trading on this ambiguity. If the analogy with the seeing of the surface of physical objects is to apply, 'sensing' must be interpreted as 'noticing'.

    Other implications are that items which are sensed can appear to be other than they are, and that the fact that a sensed item is red and triangular can no more depend on the fact that someone senses that it is red and triangular, than the fact that a table is round and brown depends on the fact that someone perceives that it is round and brown. {54}

          I cannot allow such an implication. Allowing it would lead to an infinite regress. One of the reasons for introducing sense data is to explain how it is possible for things to appear other than they are. And the explanation is that there is something which actually has the apparent property. Now if sense data could appear other than they are, this would require positing a second sense datum which actually has the apparent property. And the process could be repeated for this sense datum as well, and so on to infinity. So it is necessary to introduce a negative analogy here. Sense data are not entirely like physical object in this respect. And I don't see how denying this implication leads to the kind of absurdity illustrated by Jackson in proposing to remove the concept of color from the concept of red, or to a trans-categorial analogy against which Aune objected.

    D. SENSE-DATUM INFERENCE

          The next issue I wish to examine concerns Chisholm's sense-datum inference. What superficially seems to be one strand of thought, turns out to be three. The first is Chisholm's denial that the locution

    (I) 'D appears f to S' entails 'There is something which is f'

          What is the nature of this entailment? Is it to explain, as Sellars wants, the existence of perceptual propositional attitudes? Anyway, something like this inference is called the "sense datum inference" in Chisholm's "The Theory of Appearing."{55}

          To illustrate the invalidity of (I) he cites examples which have, save one, the form

    (II) 'D appears f to S' entails 'There is an appearance which is f'

          Notice that the conclusion of (I), unlike that of (II), is non-committal on the category of the subject. The term 'something' is a transcendental term (in the scholastic sense) applicable across categories. Whereas the term 'appearance' may or may not carry some categorial baggage. So schema (II) is or is not equivalent to schema (I).

          Whether schema (I) is considered valid by Sellars is puzzling. From such cases as 'D appears f' he is willing to infer that

    something, in some way red is in some way present to the perceiver other than as thought of.

    And adds, "This does not mean that I accept the "sense datum inference." {56} Again in Science and Metaphysics, he repudiates this inference for a contrasting 'sense impression inference':

          Unlike the latter [sense datum inference], the 'sense impression inference' is not a matter of explaining the fact that an object looks red (or that there looks to be a red object) in terms of the idea that there is something, a sense-datum, which is red; for a sense impression of a red rectangle is neither red nor rectangular.{57}

    So it is clear that Sellars rejects the sense-datum inference. But it is not entirely clear for what reason.

          Since Chisholm provides reasons for rejecting the sense-datum inference (and Sellars is aware of this), perhaps he accepts Chisholm's refutation.

          Chisholm's refutation consists of claiming that an argument of the schema-form (I) is not valid because it has the following counterexamples.

    (a) 'That animal looks centaurian' entails 'There is something centaurian'
    (b) 'The pail feels empty' entails 'There is an appearance which is empty'
    (c) 'The woods sound inhabited' entails 'There is an appearance which is inhabited'
    (d) 'The curtain appears green' entails 'There is appearance which is green'.{58}

    Elsewhere, he gives the following counterexamples:

    (e) 'The man appears tubercular' entails 'The man presents an appearance which is tubercular'
    (f) 'The books appear worn and dusty and more than two hundred years old' entails 'The books present appearances which are worn and dusty and more than two hundred years old' entails'.{59}

          What Chisholm accomplishes with these counterexamples, except for (d), is to show that not all uses of appear words can be used in this inference form. All of these appearance words, except for (d), are cases either of Chisholm's epistemic or comparative uses of appear words. Only example (d) exhibits a noncomparative use of an appearance word. And example (d), smuggled in with the other examples, is made to appear illegitimate by association. But we can resist this attempt at contamination. I find nothing wrong with example (d).

          To secure the sense-datum inference against such counterexamples, we only need to specify that the valid inference form is to be restricted to what Chisholm called 'phenomenological' or 'phenomenal' uses of appearance words. And then only example (d) qualifies.

          So modified, the only objection Chisholm has to the sense datum inference is that the conclusion contains the reifying words 'something' and 'appearance'. So his proposal is, in effect, to modify the formulation of the conclusion so as to exclude substantive terms. He opts for an adverbial formulation. So example (d) would be modified to 'The curtain appears green' entails 'I am appeared greenly to'.{60} Sellars position here is that although 'The curtain appears green' can be phenomenologically reduced to something, in some way green is in some way present to the perceiver other than as thought of. By this he means that some physical stuff is seen. He intends to emphasize that there is no perceptual category available other than that of physical things. He also means to hold that appearance statements are not analyzable, because they do not express a relation between a person and something else.

          Schemata (I) and (II) should be contrasted with what is also called the 'sense-datum inference' in Perceiving. The schema of this inference is illustrated by Chisholm in the following steps:

    (III)
    (1) 'He sees a boat' entails (2) 'A boat appears in some way to him.'
    (2) can be transformed to
    (3) 'A boat presented him with an appearance,' which entails
    (4) 'He sees an appearance.' {61}

          Chisholm rightly objects to the inference from step (3) to the conclusion (4), especially if one has a tendency to make the further conclusion: (5) 'He does not see a boat'. But I think the reason for objecting to this inference is for equivocating on the word 'see'. Those who have been interested in making such an inference have used two well known techniques. The first technique is to distinguish the uses of 'see' in (1) and (4). The distinction introduced is between what can be 'seen(d)' directly (which would apply to (4)) and what can be 'seen(m)' mediately (which would apply to (1)). According to this view, medium sized specimens of dry goods are always mediately perceived; while appearances are directly perceived. The second technique -- actually a variation of the first -- is to substitute for 'sees' in (4) the term 'senses'. Once these equivocations are taken care of the inference pattern is acceptable, provided that phenomenal appearances are involved.

          However, there is another equivocation that must be guarded against in the sense datum inference. Consider example (d). Suppose (i) the curtain is green, (ii) the curtain looks green, and (iii) an appearance is green. As Chisholm himself points out we have to distinguish a dispositional use of color words when applied to physical objects, from a phenomenal occurrent use. However, a seeing(m) of physical green entails seeing(d) phenomenal green

    E. INTENTIONALITY OF SENSATIONS

          One of Sellars' major criticisms of Western philosophy is, as he puts it, the assimilation of sensations to the conceptual order. This, he says, is a fundamental mistake. A root cause of this mistake is due to being misled by a grammatical analogy between the phrases

    a sensation of a white triangular thing

    and

    a thought of a white triangular thing.

          This, according to Sellars, is a doubly misleading analogy. First, the grammatical similarity suggests that sensations are species of thinking, which they are not; second, the grammatical similarity suggests that having a sensation is a relational fact between a person and a sensation, which it is not. These two tempting suggestions, says Sellars, must be resisted. (1) Sensations themselves, according to Sellars, are non-cognitive; although they can be the objects of cognitions, as, for example, when I know that I have a pain. (2) And having a sensations is not a relation between a person and a sensation; rather, sensations are adverbial states of a person.

          Both of these contentions have been denied by Sellars' critics. Sellars' denial that sensations are cognitive has been challenged by Romane Clark, while Sellars' adverbial analysis has been challenged by, among others, Frank Jackson. Here I will examine claim (1). I examine claim (2) under the heading Ontology of Perception.

          Romane Clark has now written several essays devoted to this topic. He specifies his difference with Sellars by way of how they handle the following triad of claims:

    (1) Perceptions are judgments (thoughts).
    (2) Sense impressions are constituents of perceptions.
    (3) Sense impressions are not conceptual or cognitive items.

          He goes on to claim that Sellars would accept (1) and (3) but deny (2). Clark, on the other hand, accepts (1) and (2) but denies (3).

          There is, however, a sense in which Sellars accepts (2). Sellars says that sense impressions are the necessary causal conditions of perceptions.

    the direct perception of physical objects is mediated by the occurrence of sense impressions which latter are, in themselves, thoroughly noncognitive. Furthermore, this mediation is causal rather than epistemic. Sense impressions do not mediate by virtue of being known.{62}

          For Clark the causal role of sense impressions cannot be exhaustive of their role:

    It is not as though a judgment exhausts the content of a perception with some correlated sense impressions serving as mere causal mechanisms, the material triggers for the occurrence of the judgment. For this would fail to distinguish perceptions from other judgments which may be materially induced by sensory stimulation, but which in no sense are perceptions, and which in no sense count as seeing what is judged.{63}

          Clark points to a disanalogy between perceptions and thoughts: sensuous experiences outrun our capacity for verbalization. I may see the hue of a color but have no adequate word for its particular hue.

          Sellars is conscious of these difficulties when he writes "Even if we take into account the fact that perception involves a causal dimension . . . we feel that the distinctive character of seeing has not yet been captured."{64} In addition, Sellars' answer is strangely at odds with what he said while discussing the thoughts of musicians and painters. There he said that these artists, in some sense, think in a sensorial medium:

    visual perception itself is not just a conceptualizing of colored objects in a certain context -- but, in a sense most difficult to analyze, a thinking in color about colored objects.{65}

          In this essay Sellars provides no analysis of this claim. However, Clark, in insisting on the representational character of sense impressions, I believe, is providing such an analysis. Clark's claim amount to this. When Sellars says that people can think in color about colored objects, this is to be interpreted that sense impressions are acting analogously to the way predicates act in propositions. And this amount to interpreting claim (2) of the triad as saying that sense impressions are predicate components of perceptual judgments. As has been discussed in the chapter "Language and Thought", Sellars discussion of representational systems in animals is compatible with Clark's interpretation.

          Clark's contribution is to dwell on Sellars' view that thoughts are analogous to speech, and to extend this analogy of speech to sense impressions as well. Speech, thoughts, and perceptions are analogous in three respects: all are intentional, all have truth values, and all are subject to reflexive, non-inferential knowledge.

    The model of sense impressions . . . is simply this: as predicates are to declarative sentences, as ascriptions are to assertions, as concepts are to judgments, as ascribing is to stating, so, too, sense impressions are to perceptions. . . . Sense impressions are the predicate "words" of perceptual, mental "assertions." . . . Thus, the occurrence of a sense impression is the predication of a sense quality to an object, and so it is in this way that sense impressions are cognitive or conceptual elements of sensuous beliefs.{66}

    Clark explicitly states that his analysis applies only to 'basic perceptions'.

    These are basic in two senses: First, they are ascriptions of simple sensuous qualities. Second, they have no internal, logical complexity. Each is simply a (single) qualitative ascription to a (single) object of reference.{67}

    and, "Basic perceptions we understand to be those sensuous judgments which ascribe the proper and common sensibles to what is demonstratively before one."{68} Clark notes that Everett Hall had made a similar claim, except that he wrote that "perceptions are descriptive or predicative throughout." To which Sellars responded:

    how [can] a pure perception. . . be a sentence, and yet be 'predicative throughout. . . . Must not pure perceptions contain expressions referring to an object in order to be able to characterize an object?{69}

          To this objection Clark replies that Sellars is correct on insisting on a referring component, and provides the referential component with the following qualifications: First, basic perceptions are postulated idealizations "to account for the empirical basis of the acquisition of the more complex, but standard, forms of sensuous belief."{70} Second, the reference is carried demonstratively: "It is the material occurrence of the impression in the given context which provides the demonstrative reference of that experience,"{71} and

    Sense impressions are rather special functions for they are defined only over a special and limited domain. They are functions taking the demonstrative references embodied in their own occurrences as arguments.{72}

          Clark adds that Sellars' perspicuous language, Jumblese, provides the ideal model for basic perceptions. In this language there are no predicate expressions; the names are simply written in different styles. However, Clark concedes that the cognitive role of sense impressions is not given phenomenologically:

    It certainly does not follow from the fact that sense impressions are not themselves objects of epistemic acts in standard cases of perception, nor cognitions, that they play no conceptual role in sensuous judgments.{73}

          Notice that Clark is using the term 'conceptual'. In view of the way I have demarcated non-conceptual representations and conceptual representations (in ch. 5), we must be clear that Clark means to make a case for sense impressions as playing a non-conceptual representational role. This is clear from the fact that he excludes logical connectors in sensuous judgments. And Clark is apparently conceding to Sellars the claim that sense impressions are not phenomenologically given. This leaves at least two possible ways in which they can play a representational role. The first way is that sense impressions are the objects or vehicles of unconscious cognitions; the second that they are potential objects or vehicles of conscious cognition.

          Richard Aquila, in commenting on Clark, seems to opt for the second interpretation: "All we need grant is that sense impressions are not themselves items which become 'mental words' in perception: rather they are merely potential tokeners of mental words."{74}

          Firth's observations in his classical essay "Sense Data and Percepts" becomes pertinent here. He too said that from a phenomenological point of view only 'percepts' ('ostensible physical objects') are given; but called attention to the possibility of introducing an awareness of sense data through a 'perceptual reduction'. A perceptual reduction would be a learned mode of perception. When we reflect on the prescriptions given by Moore or Price for "directly perceiving" sense data, we can understand their recommendations in two ways: either they were prescribing how to pick out -- perhaps by abstraction -- sense data in ordinary perception; or they were prescribing an imaginative perception according to a recipe. In either case, they were not providing a description of what was phenomenologically given, but specifying a possible mode of perception satisfying their criteria.

          In his famous essay "Is There a Problem About Sense-Data?", G.A. Paul{75} failed to discern anything like Moore's sense data in phenomenological perception, and concluded that sense data were mythical. His mistake, in hindsight, was to believe that sense data were part of the phenomenologically given. Evidently he did not consider the possibility of alternative modes of perception.

    IV. ONTOLOGY OF PERCEPTION

    Sellars works with the principle that all theoretical concepts (including introspective reports) must be introduced by analogy with concepts available in the Manifest Image. We can call this the Principle of Theory Introduction, or PTI. I will assume that PTI is correct.

          **The subjects for ontological speculation, for Sellars, are the minimal perceptual propositional reports as 'S ostensibly sees a volume of pink stuff', or, as he puts it "Something, somehow a cube of pink in physical space is present in the perception other than as merely believed in."{76}

    According to the above account, perceptual experience is a unique blend of a conceptual and a non-conceptual state. The latter is required by what is not inappropriately characterized as the common sense conviction that even when there merely seems to be a physical object which is red and rectangular on the facing side in front of one, something which is somehow a physical object which is red and rectangular on the facing side is present in the experience other than as a mere object of belief.{77}

          He then tries to explain that the act-object and the adverbial theories of perception differ according to what they make of the above formulation:

    It is this conviction that, to put it differently, an actuality which is somehow a physical object which is red and rectangular on the facing side is present in the experience which is, as I see it, the common sense core which the various forms of both act-object and adverbial theories of sensation attempt to build into a full-fledged theory. In other words they are, from this point of view, attempts to pin down the place-holders 'something' and 'somehow' in specific matter-of-factual terms.{78}

    In a footnote appended to the occurrence of 'something' in the first paragraph, he wrote, "Notice that 'something' is a transcendental so that 'somethings' are not limited to objects."{79}

          The explanation consists, first, in theoretically positing the existence of sense impressions -- in our case -- of a pink volume. Using Leibniz' distinction between perception and apperception, Sellars writes that sense impressions as theoretical entities are never (consciously) apperceived, but that they are (unconsciously) perceived.{80} And since sense impressions are theoretical entities they must comply with the Principle of Theory Introduction -- this means they must be introduced on some model.

    The consequence of this line of reasoning which assumes that there is no simpler category than the physical by which we have experiences -- even though we know by theory that we are in fact reacting to sensa -- is the conclusion that we are living in an illusory world. In "Sensa or Sensings" Sellars put it this way:

    If one thing is clear, it is that in perception we do not take what in point of fact . . . are sensa (or sensing) to be such. . . . Might we not, so to speak, mis-categorize them as items in the physical environment. Of course, such a taking would be a mis-taking.{81}

    A. THE EQUIVALENCE OF SENSA AND SENSINGS

          Sellars holds an adverbial sensing theory in the Manifest Image and a sensa theory in the Scientific Image. The relation between these theories in Sellars' vintage writings has been to talk about a categorial transposition from sensing to sensa. However, I suspect, because of the recent concentration of attention to the logic of the adverbial sensing theory which has unearthed difficulties, Sellars has, in response, shifted his stand to what appears to me to be the claim that the sensing and sensa theories are (or could be made to be) logically equivalent.

          I am not going to rehearse the logic of the adverbial sensing theory, but simply mention the course of criticisms which brought Sellars to his new position. The testing of the viability (possibility) of the adverbial theory was started by Frank Jackson{82} who introduced the so called Many-Property Problem. Sellars{83} responded to this criticism, and in the course of developing a logic of adverbial states uncovered another problem, the so called Identity Problem (or Co-incidence Problem). Thomas Vinci{84} reviewed the situation and introduced another difficulty by challenging Sellars' thesis that all talk about events is reducible to talk about propositions. He wrote: "Sellars is mistaken in thinking that ordinary event talk refers to propositional entities."{85} I don't know whether Sellars was aware of Vinci's criticism because his reaction to the state of affairs was almost simultaneous. Sellars response to these difficulties, as I understand it, is to push for a reconciliation between the sense-datum and the sensing (adverbial) theories -- a thesis I am going to call the Equivalence Thesis. Meantime, others -- oblivious to this move by Sellars -- have either tried to defend a reformulated version of the adverbial theory (Tye);{86} or after having noted the weakening of the adverbial theory, have taken up the defense of the sense-datum theory (Casullo).{87}

          Sellars move towards the Equivalency Thesis was first made in the Carus Lectures and followed up in "Sensa or Sensings: Reflections on the Ontology of Perception". The reconciliation is effected by changing the model for the concept of a sense impression. In his essay "Phenomenalism", he had argued that a theory of sense impressions has to be introduced by a model, and he opted for an adverbial theory. The model for sense data, as Sellars understands it, is a physical object and its properties. The model for the adverbial sensings is a person doing or suffering something: for example, a person waltzing. And these two models seem to give us two incompatible theories.

          To make his case for the Equivalency Thesis in "Sensa or Sensings", Sellars uses a dialectical approach. He introduces two characters, Jones and Smith, who respectively hold the sensa (sense-data) and sensing (adverbial) views. What Sellars does is transform each position through a number of steps to a common position.

          First, both Jones and Smith agree that the analysandum for analysis are cases of opaque propositional perceptions, having the form: 'S sees that this D is f.' For example, 'Jones sees that this ice cube is pink.' And they agree that from a phenomenological or conceptual analysis point of view it would be correct to say of a veridical perception, such as, 'S sees that this is ice' that it requires a knowledge of causal properties, and these are not, in a strict sense, seen. So that it will be improper to say from a phenomenological perspective that physical things are seen. Rather what are seen are the properties of physical objects. And Jones and Smith try to formulate what they can agree to as a case of minimal seeing. They can agree that

    "Something, somehow a cube of pink in physical space is present in the perception other than as merely believed in."{88}

          Second, they agree that the object of perception 'D' is believed in (rather than, strictly speaking, seen) or, put otherwise, it is taken that 'something is D'. The model for taking would be presupposition in Straw- son's sense.{89} In other words, the subject of perception is assumed or presupposed to exist and to be correctly referred to. While it is that which is expressed by the predicate word or phrase that is believed about the subject of perception.

          Third, they agree to an analysis of the common content of both veridical and unveridical perceptual reports. In order to be able to talk about the content of both veridical and non-veridical cases, Sellars introduces the category of ostensible perceptions, in which the achievement implication is neutralized.{90} Thus, a neutral formulation would be: "P ostensibly sees a cube of pink facing him/her edgewise."{91} This is intended to be equivalent to: " There appears to P to be a cube of pink facing him/her edgewise."{92}

          The next step marks a radical shift. Sellars moves from an analysis of perception to its explanation. I say this is a radical shift precisely because classical sense-data theorists intended their theories to be analyses of perceptual experiences. But I have already expressed my disagreement with Sellars about that. But having disagreed with Sellars about the limits of analysis, does not preclude me from agreeing with Sellars about explanations.

          What Sellars is assuming, then, is that the sensa and sensing theorists are not in disagreement over the analysis of perception, but over the correct categorial expression of the ontological explanation.

          Fourth, Jones and Smith

    agree that sensations are constituents of situations in which there appears to P to be such and such an object in front of him, and that the introduction of these items as having this status is a theory to be justified by its explanatory power.{93}

    Fifth, they agree that "such phrases, involving verbal nouns, as 'a sensation', 'a thought', 'a feeling' are in some way derivative from the corresponding verbs, thus 'senses', 'thinks', and 'feels'."{94}

    1. DISAGREEMENT

    The divergence between Jones and Smith comes over the explanation proposed to account for ostensible seeings, such as:

    (1) P senses a red rectangle

    Both agree that 'a red rectangle' does not refer to either a surface or anything in physical space. Sellars assumes that the concepts of physical objects and physical spaces entail concepts of causal properties.{95} And we have assumed that causal properties are not seen. At this point I have to break in to ask: if the causal properties are being neutralized why continue calling an 'ostensible seeing' of something 'physical'? Why not call the neutralized object of 'ostensible seeing' a sense datum? This is all that H.H. Price, for example, wanted for a sense-datum:

    What the red patch is, whether a substance, or a state of a substance, or an event, whether it is physical of psychical or neither, are questions that we may doubt about. But that something is red and round then and there I cannot doubt.{96}

          I believe that Sellars hasn't resolved this difficulty.

          In any case, Jones' and Smith's disagreement is about the correct categorial form of the theory of sense impressions. Jones claims that (1) has the form

    'xVy'

    ['x' represents some person; 'V' is the transitive verb 'senses'; 'y' represents 'a red rectangle', which is construed as a referring expression]. This logical form is to be represented by the following convention:

    (1A) P senses an s(red rectangle).

    Jones now stipulates (by definition) that (1A) entails:

    (1A-1) (Ex) s(red rectangle)x . senses (P,s).

    Smith, by contrast, claims that (1) has the form

    'P Adv V'

    ['P' represents some person; 'Adv' represents 'a red rectangle', which is construed as an adverbial phrase; 'V' is the intransitive verb 'senses'].

          This logical form is to be represented by:

    (1B) P (a red rectangle) senses.

    which by stipulation entails:

    (1B-1) (Ex) x (red rectangle) senses.

    2. MOVES TO RECONCILIATION

          Smith, in an irenic mood, noting the possibility of a move from

    (2) P dances a waltz.

    to

    (3) A waltzing is taking place.

    On this model he accepts a move from (1) to

    (4) A (red rectangle)ing is taking place.

    and hence to

    (1A-2) There is an s(red rectangle).

    Jones responds that on this logic Smith should also accept the entailment from (1A) to

    (5) P(s) (red rectangle)s . And Smith agrees. That is to say, for Smith ' (s)(red rectangle)' is a verb modifier.

    Sellars then presses Jones with the model of a sense datum being the seeing (in the achievement sense) of a physical object such that

    P sees D, sees of it D', and sees D' to be F.

    Sellars is introducing the distinction between seeing as which is achievement neutral, and the achievement sense of seeing to be. If Jones and Smith have agreed on a phenomenological reduction of perception to the level of 'ostensible perception', then it means they have neutralized the achievement sense of perception. And the form expressed by (1) is achievement neutral, and analogous to a seeing as. But, on the other hand, as a theoretical explanation the sensing is to be understood by analogy to the achievement sense as a non-epistemic seeing.

          Sellars points out that Jones does not sense his sensum as a sensum. To say that this must take place would entail accepting the propria persona principle (the myth of the given), which he has rejected. Should we then say the same thing about the 'phenomenal properties', namely that we do not necessarily sense them as phenomenal?{97}

    But if, even within the model, we can distinguish items which, though naturally taken to be facing surfaces of physical objects are, with respect to what is seen of them, simply rectangles of red and cubes of pink, then phenomenological analysis has made available a neutral sense of 'red rectangle' which is available for use in his theory.{98}

          This is a significant move. Ontological commitment resides in the subjects of (opaque) propositional perceptions. These are presupposed to be things with causal properties. Sellars is, however, conceding a neutrality to such things as a volume of pink, or a pink expanse. This is so because phenomenological reduction has turned such subjects into variables, so that

    the resources of the first type [concepts of occurrent sensible properties] can be said to be constant, those of the second type [kinds of objects and causal properties] change.{99}

    and,

    [The Manifest Image] combines a restricted and fixed domain of concepts pertaining to occurrent perceptible properties with a changeable domain of thing kinds and causal properties.{100}

          To reconcile Jones and Smith, both have to agree to neutralize the subject terms of their respective theories. This is, I take it, the turning point in Sellars' analysis. He is accepting a neutralized form of sensing, such that it is improper for Jones to say that he is sensing a phenomenon, and for Smith to say that the sensing is modifying a person. They agree to neutralize the nature of the subject of the phenomenal properties. They agree that "it is this 'something somehow' which they have been attempting to flesh out by a theory of visual sensations."{101}

    3. THE RESOLUTION

          The irreconciliation between Jones and Smith rests on their models. Jones is committed to the model of physical objects as the objects of sensing, while Smith is committed to the model of a human being as the subject of sensing.

    it is difficult to see how the gap between Jones' sensa which are tied to perceivers by causal ties and Smith's objectless sensings, which are logically tied to perceivers, can be bridged.{102}
    To get around the gap, Sellars suggest that both Jones and Smith could adopt a third model of sensing, and try to reformulate their respective claims in this third model. Instead of using the model of subject bound verbs which both Jones and Smith have been invoking, the new model is to use subjectless verbs instead:

          Many philosophers have attempted to use another category of verbs as their model, namely those verbs which have no subjects -- or, better, take such strange subjects as 'it'. Consider the verb 'to thunder'. Except in a metaphorical sense nothing thunders . . . There are sounds in physical space which, produced by physical objects, are not themselves 'physical objects' as specified by philosophical paradigms.{103}

          Using the model of 'it thunders', Smith can substitute for the subject in

    P (s)(middle C#s

    the subject of the model 'it', getting

    It (s)(middleC#)s.{104}

    Once this is accepted by both Smith and Jones, their appeals to physical objects and persons as models is neutralized, their positions becomes equivalent save for a terminological distinction. They can both agree to the following:

    when there appears to P to be a physical object with a red and rectangular facing surface:

    (a) There is a sensum or objectless sensing which is an (s)(red rectangle) or (s)(red rectangle)ing;

    (b) P is aware of (conceptually refers to) this sensum or sensing;

    (c) P mistakes this sensum or sensing to be the red and rectangular facing surface of a physical object.{105}

          This equivalence is made possible because the concept of a physical object makes uses of causal properties, while the concept of a property does not.{106} This is a crucial point because the concept of a physical object in phenomenological reduction becomes a variable, while the concept of an occurrent property remains invariable, as is clearly expressed in following passages:

    if we look at these two these two types of conceptual resource belonging to the framework of perception, we notice that whereas the resources of the first type [i.e., concepts of occurrent sensible properties] can be said to be constant, those of the second type [i.e., concepts of physical objects] change.{107}

          Both the adverbial and the sensa theories are, so to say, theory laden with an appeal to persons and physical objects respectively. And these appeals can be neutralized by using the model of a subjectless report such as 'it thunders'. The result is that "Jones' sensa which are tied to perceivers by causal ties and Smith's objectless sensings, which are logically tied to perceivers, can be bridged."{108}

          To justify this move, Sellars says that this is part of a "revisionary metaphysics, a revision within the manifest image -- that it is a constituent of P as seen by a revisionary eye."{109} This seem odd from my perspective since I allow processes (events) as described by 'it thunders' as part of the common sense framework.{110} However, Sellars does not have that option. He has defined the Manifest Image as based on substances, so that the acceptance of events is either an addition or a revision of the Manifest Image.

          What do I make of Sellars' Equivalence Thesis? I have already given my reasons for thinking that such neutral sensa are available to conscious awareness through perceptual (phenomenological) reduction. What Sellars' reasoning shows is that sensa (sense data) are also sensibilia, which, according to Russell, are "those objects which have the same metaphysical and physical status as sense data, without necessarily being data to any mind."{111}


    [Go to Chapter 10]