AN ALLEGED LEGEND

CHARLES G. ECHELBARGER

State University of New York, Oswego

Published in Philosophical Studies 39 (1981) 227-245. (Received, in revised form, 2 May, 1980)

      In this paper, I will attempt an exposition of some recent criticisms of Wilfrid Sellars on the notion of The Given', and a proof that Sellars is right about this notion, the proof being one which takes its light from the inadequacy of the criticisms which will be revealed. In [1], William S. Robinson argues that Sellars has not shown that 'The Given' is a 'myth'. Robinson mostly confines his scrutiny of Sellarsian text to 'Empiricism and the Philosophy of Mind' (EPM). Robinson's claim is that none of the arguments in EPM logically require that the Given is anything less than 'legendary' because

To say of an object that it is a myth is to imply that that object does not exist. To say that it is a legend, however, implies neither its existence nor its nonexistence. Thus, if I say that I will show that the given is a legend, it must not be anticipated that I will attempt to prove that there is a given. My aim will only be to argue that it has not been shown to be a myth.
Robinson notes that EPM contains explicit arguments against "several views which are based on the acceptance of a given", and that "an argument against the given as such" has to be derived from Sellars' "more particular arguments", such arguments being derivable by 'easily justifiable generalizations" ([1], p. 83).

      Robinson's orderly procedure is to construct some formal definitions concerning the notion of givenness, having culled the materials for the definitions from ideas which are either explicitly in or directly implied by the text of EPM. The remarks in EPM which he finds 'particularly salient' are

...the point of the epistemological category of the given is, presumably, to explicate the idea that empirical knowledge rests on a foundation of non-inferential knowledge of matter of fact. ([2], p.128)

...they all, i.e. forms of the myth of the given have in common the idea that the awareness of certain sorts – and by ‘sorts’ I have in mind, in the first instance, determinate sense repeatables – is a primordial, non-problematic feature of ‘immediate experience’ ([2], p.157)

Robinson's formal definitions of the key concepts are

Def. I. Something is given (is a single given item) if and only if
  1. it is a primordial awareness of sorts, and
  2. it is required to explicate a foundation view.
Def. II. An awareness is primordial if and only if there are no necessary conditions for its occurrence except
  1. conditions which follow analytically from the concept of an awareness of sorts;
  2. having a sensation or image; and
  3. conditions which are also necessary conditions for having sensations or images.

Jointly, Def. I and Def. II comprise Robinson's attempt to capture Sellars' portrayal of the notion of givenness. Next, Robinson sets forth a specific argument by Sellars against the givenness of sense-contents. Then he attempts to construct from Sellarsian materials an argument against the Given as such by generalizing the specific argument. Though Robinson does not seem to be aware of the fact that Sellars' specific argument is a reductio aimed at showing the internal inconsistency of classical sense-datum theories, Robinson does realize the importance of pursuing the strategy of rejecting as unessential to some type of Givenness epistemology at least one of the generalized counterparts of the premises of the specific argument. If successful in such an attempt, Robinson would thus have made his case for the reasonableness of not rejecting (as mythical) the giveness of sense-contents. This point would block a generalization to the mythicality of the Given as such. One of the key components of Sellars' specific argument is

If (e) sensing is analyzable in nonepistemic terms, then (g) (Some) epistemic facts can be analyzed without remainder into non-epistemic facts.
Sellars' reductio works by first obtaining (e) from earlier steps, negating (g) as an independent premise, and obtaining the negation of (e). Robinson attempts to counter the argument by refuting 'not-(g)'. He first shows that (continuing his numbering) a certain proposition

(11) "X makes a primordial use of a concept" entails "X immediately subsumes something under a concept"
is constructable from assertions and arguments explicitly made by Sellars, and that (11) is a necessary condition of 'not-(g)'. Second,he focuses on the obviously necessary condition of (11), viz.

(11a) "X makes primordial use of a concept" materially implies "X immediately subsumes something under a concept".
The crucial step in Robinson's argument is not to refute (11a) but to show that we need not accept (11a) since there is at least one conceivable circumstance in which its antecedent is true and its consequent is false. Sellars' argument thus would not logically force one to relegate the Given to mythology, even if in fact the Given is in the same position as the gorgons and the harpies. Indirectly, Robinson is defending (e) by means of a defense of its generalized counterpart.

(e') Givenness is analyzable in non epistemic terms.

Robinson chooses to reformulate (e') in terms of entailments between propositions rather than in terms of the analyzability of concepts. As a result, it is eventually made clear that Sellars must defend what Ronbison formulates as

(7) "X has a primordial awareness of sorts" entails "X has non inferential knowledge"

Robinson shows that the only arguments which could be constructed from EPM are ones which ultimately depend on (11) as a 'rock bottom principle'. By means of some (earlier) insightful textual exegesis, Robinson makes us realize that Sellars' notions of primordial awareness and making primordial use of a concept are connected, the latter being necessary for the former (in the view of classical theorists). It also emerges that primordial use of a concept is defined by reference to a specific type of concept: 'ur-concepts', and that ur concepts are by definition those which are not dependent on the class of physical concepts. The class of physical concepts are said to be those which cannot be acquired simply in virtue of having sensations and images. Ur-concepts, then, are those which can be so acquired. Hence, if it is possible to make primordial use of a concept without immediately subsuming something under a concept, then (11a), (11), etc. need not be accepted by one who respects reason. Clearly, everything for Robinson depends on his establishing the possibility that (11a) is false. He tries to present such a possibility to the mind's eye by giving an example of what it would be like to make nonsubsumptive use of a concept. He presents his example as follows.

If I count the books on my shelf, I may say, as I point, 'One, two, three,...'. These words reflect a state of mind corresponding to the properties 'first book counted by me on this occasion', 'second book counted by me on this occasion', etc. But I have not subsumed these books under concepts corresponding to these properties. Nor, indeed, could I have correctly done so until I had counted them. Even if I had said 'This is the first book, this is the second book...'. This would only look like subsumption, but it would actually be counting. Of course, my counting does provide a ground for subsumption, by myself or by another, of these books under such concepts as 'first book counted by me on this occasion'. But this is not to say that the original counting is itself a case of subsuming. ([1], p. 103)

      What sort of sense-datum would essentially depend upon a non-subsumptive use of concepts? Robinson admits that, if sense data were supposed (as they have been supposed by many) to be entities which can exist and have the properties they have whether anyone is aware of them or not, then

I do not know of any way of thinking of such an entity being given, except by way of thinking of it as subsumed under a concept. One must begin with an entity having a property. Then one has to add the awareness of it. Finally, one has to get these connected, and the only way it appears that one can do this is to make the awareness a use of a concept which subsumes the entity under a concept. ([1], p . 103)
Robinson suggests a conception of sense data alternative to this one. According to this alternative, a sense datum is not an entity that can exist and have the properties it has without anyone being aware of it. He says
...on this view there is nothing to be brought under a concept until a use of a concept has already occured. A red sense datum, for example, would just be a primordial awareness of the sort red, involving a use of the (ur-)concept 'red', but not a subsumptive use. Such a use is analogous to the use of the concept 'first book counted by me on this occasion' in counting. ([1], p. 103)
As in the case of counting, such nonsubsumptive uses of (ur-)concepts are also supposed to provide a ground for a 'later subsumptive use of that concept'. ([1], p. 104). Robinson says that this is "enough to make the notion of a nonsubsumptive use of a concept intelligible" and so proceeds to infer the possibility of (11a)'s falsehood, etc. I am still in the dark about this notion of a primordial awareness involving nonsubsumptive uses of ur concepts. I will argue that the obscurity of my understanding is the fault of Robinson's notion.

      First, I will argue for the following claims:

      I. Even if the physicality of Robinson's concept-example is ignored, the 'analogy~ between the allegedly non-subsumptive use of 'ur-red' and the non subsumptive use of 'first book counted by me on this occasion' is no analogy at all.

      II. There can be nonsubsumptive uses of concepts only if the concepts so used are physical concepts, not ur-concepts, because nonsubsumptive uses of ur-concepts could provide no ground whatever for a later subsumptive use of any concept.

Ad I. Any analogy which might be claimed to exist between the nonsubsumptive use of 'first book counted by me on this occasion' and the non subsumptive use of the ur-concept 'red' is at best extremely weak. Anyone who could on a given occasion use the concept 'First book counted by me on this occasion' nonsubsumptively would very probably have been able to use the concept 'book' subsumptively on an earlier occasion. If it were known that a certain person was unable to use the concept 'book' in a subsuming or classifying way, one would be justified in supposing that no spontaneous utterance of 'first book counted by me on this occasion' by this person is the expression of a nonsubsumptive concept-use. A reason for totally rejecting the claim of analogy is that the concept 'first book counted by me on this occasion' can have no correct application unless the book which I in fact count now exists in such a way that no one need have counted it. The same cannot be said of Robinson's kind of sense-datum. Moreover, the book exists in such a way that it could have been mis-counted by me as the second book counted by me on this (noninstantaneous) occasion without thus bringing into being two numerically nonidentical books. The latter mode of existence, receptive to such counterfactual descriptions, cannot be attributed to Robinson's kind of sense-datum. His kind of sense-datum is individuated by being sensed. It can't be sensed incorrectly because, by nature, it has no existence or identity independently of being sensed. This kind of sensing essentially involves (is?) the nonsubsumptive use of ur-concepts. The fact that Robinson's concept example ('first book counted by me on this occasion') is a physical concept means that he simply begs the question when he proceeds to infer that, since the latter concept can be used non-subsumptively, ur-concepts (by definition non-physical concepts) can also be used non subsumptively. No analogy has actually been made. Robinson has only made a claim that relevant analogy exists, and has offered no substantiation for this claim at all. A number of important philosophers, Strawson for example, have argued in great detail that the reason why a logically individuating description (like Robinson's concept-example) can be used nonsubsumptively is precisely because it is logically tied to the framework of physical things in Space and Time. To proceed as if such arguments were known to be unsound or as if they did not exist is begging the question at hand.

Ad II. Robinson's whole case depends on the viability of a conception of sense-data alternative to the varieties of sense-data discussed by classical sense-datum theorists. That is, in order to find a counterexample to (11a), he must explain what kind of sense-datum would avoid the kinds of problems posed by Sellars for classical sense-datum theorists. If it should turn out that primordial use of concepts which can be acquired simply in virtue of having sensations and images (nonphysical concepts, i.e. ur-concepts) logically fails to contribute to explicating the possibility of having noninferential knowledge of matter of fact, then all of Robinson's reasons for considering some form of givenness to be at least 'legendary' would collapse. He would have presented no viable alternative to the view (endorsed by Sellars) that non inferential knowledge of (contingent singular) matters of fact is possible only from within a conceptual framework of physical objects in Space and Time. I will argue that his alternative is not viable because logically incoherent. But I must first explain the relevance of the argument against the general approach of Robinson's paper. At this point, the alert reader will say

Either the argument you are about to give is roughly Sellars' or it isn't. If it is (roughly) his, then neither your paper nor Robinson's is very interesting because all that you do is to point out something in the Sellarsian text which Robinson just overlooked. If it is not Sellars' argument but yours, then Robinson is at least partly right in claiming that the arguments of EPM are insufficient by themselves to show that the Given is a myth.
My response to this challenge is as follows: (i) Robinson has called attention to a serious flaw in EPM. The flaw is that the only arguments against the Given as such which seem constructable from EPM are arguments which perhaps lend a considerable amount of support for the conclusion that appeal to the Given does not in fact account for noninferential knowledge, and so is dispensable on the ground of parsimony. It is not evident that EPM contains arguments which show that appeal to the Given logically could not account for noninferential knowledge. There does not seem to be argument in EPM which demonstrates the notion of the Given as such to be logically incoherent. One wants arguments of this kind in philosophy before classifying philosophical views as myths. Robinson's paper is interesting because of this apparent lacuna in EPM. It is therefore appropriate for Robinson to attempt to find an empirically plausible example which seems to illustrate a possibility which EPM does not appear to exclude on purely conceptual grounds. (ii) I intend in what follows to show that, despite its initial plausibility, the notion of nonsubsumptive use of an ur-concept leads to logically absurd results and thus that Robinson has not provided a pausible example countering Sellars' views.

      But will these arguments be mine or Sellars'? They will be extensions by me of arguments which are implicitly presupposed in EPM. Robinson shows considerable ingenuity in constructing from EPM premises which do not explicitly appear therein. I shall do something similar though I make no claim of ingenuity.

      Robinson cannot be faulted for focusing primarily on the passage in EPM which seems to be the most explicit characterization of the Given. ("All forms of the myth of the Given have in common the idea that awareness of certain sorts... is a primordial... feature of 'immediate experience'"). It is natural to pick the one which seems clearest and best connected with the general purposes of the essay. In fact, however, the Given is characterized in several other ways in EPM. So, if some of the other characterizations are more helpful in clearing up EPM's apparent defects, they should be considered carefully.

      It is obvious that Sellars wishes EPM to be viewed in the context of historically relevant epistemological themes. He never says in EPM that his own could be described as generally Kantian, but it is reasonable to infer that conclusion based solely on the overall structure of EPM. Two points are particularly noteworthy in this connection. (i) The main thrust of the long section directed against sense-datum theories is to show that these theories attempt to bridge the gap between obviously cognitive acts of noninferential knowing, the 'object' of which is a fact, and noncognitive acts of sensing, the objects of which are particulars. Sellars argues at length that the gap cannot be bridged. (ii) In section VI ("Impressions and Ideas: A Logical Point") Sellars reveals what he considers to be the deepest mistakes about immediate experience and knowledge in Locke's and Hume's views and then describes a different view which constitutes the rejection of these specific mistakes. He calls the latter view "Psychological Nominalism" and declares himself basically friendly to it. He says

...Locke, Berkeley, and Hume all take it for granted that the human mind has an innate ability to be aware of certain determinate sorts -- indeed that we are aware of them simply by having sensations and images.

If (association between words and classes of resembling particulars) is not mediated by facts either of the form x resembles y or of the form x is f then we have a view of the general type which I will call psychological nominalism according to which all awareness of sorts, resemblances, facts, etc., in short all awareness of abstract entities -- indeed even all awareness of particulars is a linguistic affair. ([2], p. 160)

Clearly what Sellars is rejecting here is the idea of the nonlinguistic immediate awareness of abstract entities as a ground or explication of non inferential knowledge (Psychological Realism?). It is also clear from the rest of EPM that he does not object at all to the idea of noninferential knowledge (of our environments or of our own sensory or conceptual states), nor to every kind of explanation of how noninferential knowledge is possible. He does not even object to a metaphysical explanation of how non inferential knowledge is possible. Nominalism, after all, is a metaphysical position, so Psychological Nominalism depends on a metaphysical backdrop. It is precisely the abstractness of such abstract entities as one is alleged to be aware of nonlinguistically as a condition of having noninferential knowledge that Sellars considers to be a deep logical source of trouble.

      What I have just said may seem to be no different from what Robinson takes Sellars to mean by the Given. In fact, though, Robinson does not seem to see that abstractness is the culprit and his failure to realize this point is recorded in the way that he formulates one of his main definitions, viz.

Def. I. Something is given (is a single given item), if and only if
(i) it is a primordial awareness of sorts, and
(ii) it is required to explicate a foundation view. ([1], p. 85. My emphasis, CE)

Surely an awareness is a particular event, process, act, state, or concrete thing of some kind. So, if the analysis of what Sellars means by the Given that I am constructing now is moving in the right direction, Robinson's analysis is seriously imperfect. One may reply that this point would not necessarily mean total collapse of Robinson's case because it could still be maintained that Robinson has produced a promising example of a form of givenness alternative to the one that Sellars has in mind. I will show on the basis of general principles implicit in EPM that such an alternative form of givenness is also logically unacceptable.

      Robinson's concept-example involves what he calls the nonsubsumptive use of a concept. Consider an ordinary physical sortal concept, e.g. the one expressed by 'table'. Subsumptive use of such a concept would have a form illustrated by the open sentence

x is a table.

Robinson's example of a nonsubsumptively usable concept is not a straight sortal. It involves a physical sortal ('book'), but it is what might be called a logically individuating description, an essential part of its individuating power being made possible by the occurrence in it of the expressions 'me' and 'this'. Consider next a structural point about nonsubsumptiveness. What is the smallest modification that would have to be made to the subsumptive context 'x is a table' in order to transform it into a nonsubsumptive context? At the vert least, one would have to eliminate any suggestion of predication. But the solitary sortal 'table' does not accomplish an illustration of a nonsubsumptive use of a concept, a use which is experientially engaged as opposed to idle conceptual reverie. Such a use would be illustrated by a locution tantamount to

this-table,

which is neither a sentence, nor an open sentence, but could be the subject of a sentence. What Robinson argues can be nonsubsumptively used, however, is obviously not physical sortal concepts. His sense-datum notion is based on an example of use of an attributive concept ('red'). He argues that ur-concepts parallel to attributive concepts applicable to physical objects (e.g. 'red') can be nonsubsumptively used; thus, 'ur-red'. If we were to try for an illustration of the nonsubsumptive use of the concept 'red' which applies ordinarily to physical objects, it would seem to be

this-red.

But can this expression serve as the subject of a sentence? It is not clear that it can do so because in the case of a concept-use that is within the framework of physical object concepts, 'red' is at least an attributive concept, and is capable of intelligibly occurring in predicative contexts; thus, 'x is a table and x is red'. It may seem logically harmless to propose a sentence like

this-red is pale

as having a subject illustrating nonsubsumptive use of a concept. Recall, however, that what one knows inferentially or noninferentially is expressible by means of sentences. Noninferential knowledge of ordinary (physical) matters of fact is, according to Robinson's proposal, made possible by the nonsubsumptive use of an ur-concept. We are now exploring the logical structure of such nonsubsumptive uses. If these uses provide us with subjects of sentences, do they also provide us with predicates? If they make non inferential knowledge of the form 'x is red' possible, it seems that they must provide predicates. In the case of an expression like 'this-table' we can employ 'table' nonsubsumptively and then in the very same sentences in which it can appear as grammatical subject also employ 'table' subsumptively. Thus, ‘This-table is a table' is a sentence in which 'table' occurs intelligibly and univocally in both the subject and the predicate. The implication seems to be that only a sortal term can properly occur as the expression of a non subsumptive genuine subject-expression. Something is undeniably odd about 'this-red is red' as compared with ‘This-table is a table' or compared with ‘This-table is red'. This implication would be important if shown correct because Robinson says of nonsubsumptive uses of concepts

...My counting provides a ground for later subsumption under such a concept as 'first book counted by me on this occasion'. In the case of a primordial awareness of red, the initial nonsubsumptive use of ur-red provides a ground for a later subsumptive use of that concept. ([1], p. 103-4)
The oddness of the occurrence of 'red' in 'this-red' will soon emerge as a symptom of deep trouble in Robinson's view. My final arguments will take as their point of departure the question "In what sense could nonsubsumptive uses of ur-concepts provide a ground for later subsumptive uses of concepts?"

      Robinson's counterexample depends on a certain definition of what it is to have and use a concept. Robinson formulates this definition as part of his attempt to demonstrate that Sellars has not shown the Given to be a myth. In other words, part of what Sellars does not do according to Robinson is to argue against a general definition of having and using a concept which would support those not ready to consign the Given to Mythology. Robinson says

... a use of a concept is the actualization of an ability to be in a state corresponding to a property, while a subsumptive use of a concept is the actualization of an ability to be in a state corresponding to an assertion. ([1], p. 102)
Evidently what Robinson means here is that nonsubsumptive use of a concept is the actualization of an ability to be in a state corresponding to a property. Let us continue with our assumption that the form of this kind of state is illustrated by locutions like 'this-red'. Robinson's primordial awarenesses involve being in such states and it is thus that they are awarenesses of sorts. Apparently then they are 'of sorts' in that they correspond to properties. So, it seems initially right to say that, for Robinson,

A primordial awareness of red

is just

A 'this-red' awareness corresponding to (the property) red.

      Sellars thinks that one of the most common sources of the inference to sense-data has been the problem of analyzing as a group the following three kind of situations:

  1. Seeing that x, over there, is red

  2. Its looking to one that x, over there, is red

  3. Its looking to one as though there were a red object over there.
One of the aspects of these three situations, something that they share in common according to Sellars and the classical theorists, is the aspect of descriptive content. Descriptive content is one aspect to which we are committed by characterizing situations as being of these three kinds. The other aspects are propositional content ('that x, over there, is red'), and (the lack of) endorsement of propositional content. The crux of the classical problem Sellars formulates as "What is the intrinsic character of descriptive contents?" Sellars has his own answer to this question, an answer which requires that descriptive contents can only be characterized in the terms of a highly developed system of physical concepts, including physically-red.

      Robinson's aim is to argue that descriptive contents (Primordial awarenesses) can be intrinsically characterized by ur-concepts. He does not clearly explain the nature of such intrinsic characterization. It is clear that involving a use of a concept which corresponds to the property red intrinsically characterizes primordial awarenesses of red. It is not clear what it is in virtue of which such awarenesses can be thus intrinsically characterized. Does the awareness itself correspond to a property as I have ventured to suggest? Does the correspondence in this case hold because of similarity between the awareness and the property? This suggestion is useless because similarity in this case would have to be similarity-in-a-respect. If so, both the awareness and the property would exemplify the property red. It would be disaster, as logicians know, to allow 'Red is red' as a significant sentence in a well-formed language Does the correspondence between the awareness and the property hold by the property being somehow included in the awareness? What then would be the status of those other properties which are predicated of the awareness (e g. the property of being mental)? If, as seems likely, inclusion is just another name for predication, then red is again being predicated of the awareness. As Sellars has pointed out, such an option would require the intelligibility of "prying the term 'red' loose from its prima facie tie with the category of physical objects". Robinson's introduction of his notion of a primordial awareness of a sort is for the purpose of showing that ur-concepts can be used nonsubsumptively and that such a use provides a ground for a later subsumptive use of the same concept. But the very notion of a primordial awareness of a sort would then take for granted what Sellars insists is highly controversial. On the exemplification-construal of 'correspondence to a property' entertained immediately above, we would be required to assume the intelligibility of predicating 'red' of a nonphysical particular in order to understand the essence of a primordial awareness of red. Such an exemplification-construal of 'inclusion of a property' is thus extremely unpromising. How 'inclusion of a property' might be nonpredicatively construed, I cannot say.

      What ur-concept is to intrinsically characterize primordial awarenesses? The property of involving a state corresponding to red does intrinsically characterize the awareness. But is the concept of that property an ur-concept? Robinson does not indicate that he thinks the latter is an ur-concept. The only ur-concept of which he speaks in this context is the ur-concept of red. Nothing in the argument I have just presented rules out the possibility that primordial awarenesses could exemplify red. It only shows that we are not entitled to assume a concept of a primordial awareness which logically necessitates that they be (e.g.) red.

      On balance, Robinson's idea seems to amount to the following: The 'right' intrinsic characterization of a primordial awareness can be made in virtue of the nonsubsumptive use of the ur-concept of red which is involved in the awareness. The awareness is a complex state, one sub-state of which is a concept-use. The right relation between the awareness and the property is entail ed by whatever the relation is between the concept-use and the property. Continuing with the apparatus developed so far, we may say that Robinson's primordial awareness is intrinsically of red just in case it involves a concept use-state illustrated by the phrase 'This-red' and that it is this concept-use which fundamentally determines this intrinsic characterization by virtue of the fact that it is the concept-use-state which corresponds to the property red.

      The next question is then obvious. What is it in virtue of which such a conceptual-use-state corresponds to just that property to which it does correspond? Similarity seems to be just as bad a suggestion in this case as before. Iikewise for the notion of including the property red. The question is an urgent one because what is required of any Given is that it explicate a foundation view of noninferential knowledge. For this purpose, there can be no mismatch or misfit between a noninferentially known fact and the Given which explicates the noninferential knowing. A primordial awareness of red involving (per impossible, one assumes) a 'this-red' concept use which corresponded to the property blue would not provide an explication for noninferentially knowing that there now seems to be a red object over there. If there were a dyadic extensional relation between a nonsubsumptive (ur-)conceptual state and a certain property (a relation in the same family as '... is on top of...') then the kind of misfit just alluded to is possible because all such relational truths are contingent. Such a Given would be useless because not totally reliable. So construed, a this-red ur-conceptual state could fail to determine the right intrinsic characterization of the awareness.

      Such possibility of failure could be guarded against by explaining that the kind of relation involved is not an extensional relation at all. If, for example, one said that properties simply are concepts (a la Ockham), then no misfit would be possible. But the price for such security is too steep. To say that a this-red conceptual state corresponded to the property red would then only mean that it is the ur-concept of red which is nonsubsumptively used in a nonsubsumptive use of the ur-concept of red. No doubt, everything is what it is and not another thing.

      Consider then the possibility that the correspondence is a matter of the conceptual state being 'of' the property in a still different sense. According to this sense, the state is of the property by virtue of being directed upon it in something like the manner in which all states of consciousness are 'directed upon' according to the early Brentano. The advantage of this construal of Robinson's 'correspondence' is that it is like extensional relations in seeming to have distinct terms, but it would be nonextensional in that the nature of its 'subject term' guarantees the nature of its 'object term' and thus grounds noncontingent truths, which are not logically trivial truths of the kind made possible by purely logical relations. We would be construing 'this-red' conceptual states as having intentionality. The awareness which involved such a state could then be said to borrow intentionality from it. Another advantage of this construal is that it involves nothing like the classical sense-datum postulated for the purpose of exemplifying those properties which things other than sense data seem to exemplify, but perhaps don't exemplify. The fallacy of this common inference could not be attributed to Robinson. It is clear that Robinson is trying to avoid this kind of fallacious postulation. On this score, he seems to be successful.

      Nothing is gained against Sellars by this success. If this conception of sense-data which I have just (correctly, I believe) attributed to Robinson is the one that he means to present as a serious candidate for a Given alternative to the one he thinks Sellars has effectively criticized, then Robinson has avoided the frying pan for the fire. For this proposed conception of sense data entails a crucial part of the very meaning of the Given which was actually paramount in Sellars' mind in EPM and which (as I have already charged) Robinson seems to have overlooked: Givenness as involving immediate awareness of abstract entities. To make Robinson's Def. I more exact,

Def. IR. Something is given (is a single given item) if and only if
(i) For some x, some y and some z, x is a primordial awareness, y is a This-F' conceptual state, x involves y, z is the property F, y is directed upon z, and
(ii) x is required to explicate a foundation view.

What Def. IR does is to make fully explicit the meaning of 'of a sort' in 'A primordial awareness of a sort' according to the construal of Robinson's primordial awarenesses of sorts which I am now suggesting is virtually forced upon his reader. The notion of the Given which I am arguing is actually the one Sellars had primarily in mind would be:

Def. IS. Something is given (is a single given item) if and only if
(i) For some x, some y and some z, z is the property F, y is a 'This-F conceptual state, x is a primordial awareness, x involves y, y is directed upon z and
(ii) z is required to explicate a foundation view.

What Def. IR and Def. IS have in common is the notion of a conceptual state involved in a primordial awareness such that the conceptual state is directed upon a certain property. In IS, Sellars says givenness is a property of proper ties; a property which properties have derivatively by standing in a 'relation' to a primordial awareness. Robinson thought that Sellars was saying that givenness is a property of those primordial awarenesses which indirectly stand in the same 'relation' to certain properties. Robinson's notion of the given, as I will soon show, does not escape general points in Sellars' critique of the Given because it includes the very notion that was the main target of Sellars' attack: The so-called 'relation' of being (intentionally) directed upon.

      Assume what we have just attributed to Robinson: that sorts are intention al objects of primordial awarenesses and that the occurrence of such awarenesses is what is meant by the occurrence of sense data. How does the occurrence of such sense-data serve to explicate the possibility of noninferential knowledge of contingent singular truths? How does the occurrence of a primordial awareness 'of red' explicate the possibility of noninferentially knowing that (e.g.) there now seems to be something red over there? In this case, the question seems to answer itself because the only property involved in the propositional content of what is noninferentially known is the very property that is the intentional object of the primordial awareness, because only one object is taken to have that property, and because the property in question is logically monadic. Consider noninferential knowledge having a more logically complicated propositional content. For example, noninferential, knowledge that

There now seem to be something red and square adjoining some thing blue and triangular over there.
How many primordial awarenesses of sorts would be required to explicate noninferential knowledge having the latter propositional content? If one says "As many as there are words in the previously indented sentence stand ing for sorts", then the result would be that this set of primordial awarenesses would provide the same explication for the sum of five different noninferential knowings. The fne different knowings would not add up to the one knowing just described. One could try to deal with this problem by allowing for primordial awarenesses of complex sorts, e.g. redsquare. The question then arises: How would primordial awarenesses of relation-sorts contribute to the explication? Relations may not seem to present a problem if one confines his attention to symmetrical relations like adjoining. As will soon become apparent, asymmetrical relations do present a serious problem in this context. Suppose that one allowed only three awarenesses: one of the complex sort A-B, one of the complex sort C-D, and one of relation R. How do these explicate the noninferential knowledge that there seems to be something which ex emplifies A and B, something else which exemplifies C and D, and that the frst has R to the second? Where R is asymmetrical, it is obvious that the propositional content of the noninferential knowledge in this case is not adequately grounded or explicated by the occurrence of the three primordial awarenesses. If one changed the last conjunct of the propositional content to read 'The second has R to the first', the three awarenesses would continue to provide for noninferential knowledge of this fact whatever explication it provides for noninferential knowledge of the original (incompatible) fact.

      One might then try to stipulate that in such cases there is one awareness of a single complex sort. Consider the complex sort

redsquareadjoiningbluetriangular.
Such a move would be symptomatic of desperation, as may be realized from the following consideration. If adjoining is in some sense a component of this complex sort, then either it occurs predicatively or it does not occur predicatively. If it does not occur predicatively, then the same should hold for other relations which are components of complex sorts, e.g. Iargerthan in redsquarelargerthanbluetriangular. Indeed, if larger-than does not occur predicatively in that sort, then no good reason can be given to show that red or any of the others occur predicatively in it or in any other complex sort. The result is chaos. Awareness of complex sorts in which component sorts, especially the asymmetrical relations, do not occur predicatively would provide the same explication for two cases of noninferential 'knowledge' where the facts thus 'known' would be logically incompatible with each other. One would 'know' both facts. But this is impossible.

      Making the components occur predicatively also leads to unacceptable consequences. Where red, blue, larger-than, etc., are supposed to confer the right predicative structure on the complex sorts, it is obvious that the com plex concept which is involved in the awareness of this complex sort would have to be used subsumptively. The awareness would thus no longer be primordial. Indeed, it would be a case of noninferential knowledge, and so require a Given to explicate it. Because of the factor of immediate subsumption under a concept, therefore, this option would support principle (11), which Robinson regards as the principle to be refuted.

      Suppose one responded that primordial awarenesses should not be considered as explicating noninferential knowledge of such 'subjectless' facts as that "There now seems to be something such that...", but to explicate non inferential knowledge conceming a person's perceptual states; the fact, e.g. that

There now seems to S to be two objects such that the one is larger than the other, the facing surface of the one is red and square and the facing surface of the other is blue and triangular.
This suggestion fares no better than any of the others. What promordial awareness(es) of SORTS could there be which served to explicate having noninferential knowledge of the fact expressed by the indented passage directly above which does not equally explicate having noninferential knowledge that
There now seems to S to be two objects such that one is larger than the other, the facing surface of the one is blue and square and the facing surface of the other is red and triangular?
There are those defenders of sensory givenness who might attempt to avoid the difficulties raised here by presenting 'adverbial' theories of sensing. But a moment's reflection should satisfy one that, if the relation of sensing-in certain-manners (or being primordially aware in certain ways) to noninferential knowledge is some kind of epistemic relation (grounding, explicating, found ing), the same problems will break out again. Notice that if this relation were considered to be strictly causal and nonepistemic then it is by no means obvious that these problems will beset an adverbial view of sensing as such. Such a view might be fruitful as part of a causal explanation of how non inferential knowledge of one's own perceptual state can be obtained. I shall not develop this gambit since Sellars has done so repeatedly and in detail. To make the connection causal and nonepistemic, of course, eliminates sensing as a source of the Given. It is the impossibility of any principle of unity holding among the manifold of diverse sorts that is the source of the trouble I have just uncovered. The echo of Kantian themes is loud and clear.

But are the arguments in this vein to be found in EPM? Not explicitly, it must be admitted. Yet, if one reflects on certain parts of EPM with Sellars' remarks on Psychological Nominalism in mind, such arguments can reason ably be taken to be presupposed by Sellars. The parts of EPM just alluded to are in the section "Impressions and Ideas: A Logical Point". Therein, Sellars attempts to "... clear the way of traditional obstacles to understanding the status of such things as sensations of red triangles". ([2], p. 155)

Thus, suppose I were to say that, while the experience 1 am examining is not a red ex perience, it is an experience of red. Does not the existence of a sensation of a red triangle entail the existence of a red and triangular item? ([2], p. 155)
The point just made is that the latter kind of reasoning is one way of attempt ing to pry terms like 'red' loose from the framework of physical concepts. He continues
... there is no doubt but that historically the contexts ...sensation of... and '...impression of... were assimilated to such mentalistic contexts as '..believes...' ...in short to contexts which are either themselves propositional attitudes or involve propositional attitudes in their analysis. This assimilation took the form of classifying sensations with ideas or thoughts. ([2], p. 155)
Robinson might well agree that the attempt to assimilate sensations to thoughts is a mistake but insist that he has not tried to assimilate 'sensation of'-contexts to nonextensional contexts. He might say that he has merely claimed that the idea of primordial awareness of sorts providing a ground for a later subsumptive use of a concept is at least an intelligible idea and is not ruled out by Sellars' arguments. What Robinson has done, however, is to indirectly assimilate primordial awareness-contexts to nonextensional contexts, if the correspondence of primordial awarenesses to properties is determined by an intentional quasirelation 'holding' between nonsubsumpt ive concept-use-states and properties is the key ingredient in his kind of sense-datum.

      The spirit of Sellars' antipathy to the assimilation of ...'sensation of...' contexts to nonextensional contexts like '...believes...' moves one easily in the direction of similar criticisms of Robinson's suggestion. It is precisely because of the fact that Robinson's kind of primordial awarenesses of sorts are defined so that they may keep one foot in logically extensional reality (by seeming to have a relation-like form) and another foot in the logically nonextensional order (by being logically noncontingent) that they necessarily do not contribute to an explication of noninferential knowledge that involves subsumptive uses of concepts embedded in propositional attitude-contexts. Robinson's attempt to provide an intelligible alternative to those ideas which he admits Sellars has effectively criticized thus fails because of essential features of his kind of sense-datum. His attempt to argue that Sellars has not shown the Given to be a myth (anything worse than a legend) therefore fails also. So, it is fair for one sympathetic to Sellars to claim that, in view of the above results, the presumption of innocence is now in Sellars' favor and the burden of proof of nonguilt has fallen back on the partisans of the Given. Robinson has provided no clear evidence that the Given is anything more substantial than a myth. The prospect that this support could be provided now seems more remote than ever.


NOTE

This paper was in part written with the support of a Summer Research Fellowship (JAC/UAC) from the State University of New York. The author is grateful to Robert Carnes for comments on an earlier version.


BIBLIOGRAPHY

[1] Castaneda, Hector-Neri (ed.): 1975, Action, Knowledge, and Reality (Indianapolis, Bobbs-Merill).

[2] Robinson, William S.: 1975, 'The legend of the given', in [1].

[3] Sellars, Wilfrid: 1966, Science, Perception, and Reality (New York, Humanities).