In reorganizing his papers, Willem DeVries came across the following unpublished paper by Wilfrid Sellars, which he sent to me. I have scanned and edited it for the Internet. Andrew Chrucky, June 18, 2004.

BEING AND SOME CATEGORIES

Wilfrid Sellars .

1. Kant is clearly not an "empiricist," yet the concept of experience is central to his philosophy. This dissociation of the terms empiricism1 and 'experience' is an interesting feature of how they have come to be used, not only in the Kantian tradition, but also, for example, in the pragmatic tradition according to John Dewey. The latter, as is well known, equated 'empiricism' with the 'atomistic' sensationalism of Hume and Mill, but used the term 'experience' in the spirit of German idealism and made it the central concept of his naturalistic pragmatism.

2. Philosophical 'isms' are as difficult to define as their political counterparts, and 'empiricism' is no exception, yet paradigm cases are presented by Hume, Mill and, more recently, by logical positivism. The Wittgenstein of the Tractatus is a particularly interesting case, because although he is clearly, in some respects, in the Humean tradition, he nevertheless conceives of the task of philosophy as that of giving an a priori account of what it is to be an object of empirical knowledge. in this respect he belongs in the tradition of Kant, for whom, as far as theoretical reason is concerned, the task of philosophy is exactly that of explicating the concept of an object of experience. Kant emphasizes that this task does not belong to empirical psychology. Like Wittgenstein he conceives of it as a. non-empirical enterprise. in effect he is convinced that it is possible to delineate the essential features of anything that could count as an object of empirical knowledge in any possible world i.e. for any finite mind, however different the world it inhabits might be from ours in its generic traits. Notice; that these differences between possible worlds would concern not just their histories, but the very qualities, relations and nomological connections which characterize the objects which make it up. in other words, Kant aimed at delineating the conceptual structure of the most generic features of the concept of an object of experience. Needless to say, however, he was also concerned to understand how these most generic features take specific form to constitute the concept of an object of human experience.

3. If we can say, as I think we can, that the pure categories are essential moments in the definition of an object of experience in general, then schemata are to be construed as the differentia which specify these generic moments into the specific categories of a variety of finite mind in its possible world; and the 'schematized categories' described by Kant become the categories involved in the explication of the concept of human experience, or, roughly, the experience of those finite centers of experience which share Space and particularly, Time as forms of intuition. For, as Kant sees it, the distinctive feature of human experience is that it is experience of a world of spatio-temporal objects. in evaluating this conception, it must be remembered that Kant equates Space and Time with Newtonian Space and Time, and that he would grant that a world of experience might have a structure which, though not in this sense 'spatial' or 'temporal,' has properties which are analogous to the latter in ways which make possible a schematizing of the pure categories and hence which satisfy the abstract requirements of a concept of a world of experience which has been purified of all contingent features.

4. The points I have been making so far can be summed up in the following statement. Both Kant and Wittgenstein think it possible to give an a priori account of what it is to be an object of empirical knowledge. Obviously the accounts they give differ in interesting ways -- ways which reflect the different conceptual resources on which they could draw. For the intervening century saw two intellectual revolutions which have already wrought irreversible changes in the philosopher's environment. Of these revolutions the most important was the impact of evolutionary theory on what are now called "The Life Sciences."

5. Less important, but by no means insignificant -- though its significance has been exaggerated -- was the revolution in logical theory which triumphed with the Principia Mathematica of Whitehead and Russell; for both Kant and Wittgenstein took as their point of departure in explicating the concept of an object of empirical knowledge, the forms and operations in terms of which the logical theory of their time interpreted the structure of statements and the validity of inferences.

II

6. I have pointed out that Kant believed it possible to explicate the concept of an object of empirical knowledge in a way which abstracts from the specifics of human experience. I shall now proceed to argue that the same is true of the Wittgenstein of the Tractatus. To begin with, it is a familiar and tantalizing fact that he gives an account of objects, not only without giving any examples, but without even indicating what sort of examples it would be appropriate to give. Thus, whereas Kant does tie his abstract account to distinctive features of human experience, Wittgenstein makes no such concession and does indeed give an account of what it is to be an object ueberhaupt.

7. Many contemporary philosophers think of themselves as belonging to the 'empiricist tradition,' and yet are increasingly attracted to those features of Kant's thought which are not peripheral to but, indeed, central to his clash with 'empiricism.' It is, I would argue, no mere coincidence that this phenomenon has been accompanied by a revival of interest in the Tractatus, and a growing awareness of the extent to which it diverges from classical empiricism and is profoundly Kantian in character.

8. Are we to interpret this trend as an absorption of Kantian themes into empiricism or as a reinterpretation of Kantian themes along empiricists lines? These questions call for decisions rather than answers. Which gives us more insight? a contrast between empiricised Kant and historical Kant? or a contrast between Kantanised empiricism and historical empiricism? One thing is clear. The traditions are merging and neither will ever again be the same.

III

9. I pointed out that for Kant, the explication of the concept of an object of empirical knowledge requires a theory of 'categories.' The same is no less true of Wittgenstein, and, indeed, I have been implying that their treatment of 'categories' is importantly similar. Yet on what principle do philosophers collect certain concepts together and label them 'categories'? For offering a theory of categories presupposes that one has such a principle of collection in mind. Historically, the principle has evolved with the theory. The initial clumsiness with which it was formulated reflected the inadequacies of early theory, indeed, the sophistication of the principle of collection is just the sophistication of the theory. The initial move toward a collection of and a theory of categories is to be found in Plato's Sophist. The approach is a subtle one. Indeed as in the case of so many other aspects of Plato's thought, it achieved a degree of insight which was not soon to be equalled. On the other hand, its explication would require an elaborate scholarly apparatus. And in a paper to fit the alotted time, two is company and three a crowd. On the other hand, a "reference to the Aristotelian tradition is indispensable, though when Aristotle characterizes the categories as 'highest kinds' (summa genera) of entity he is building, as usual, on platonic ground. It is this conception of the categories which is one of the abiding themes, if not the abiding theme, of traditional kategorienlehre.

10. To approach the conception of categories as summa genera in terms of what have been traditionally regarded as paradigmatic examples -- thus, substance, quality and relation -- is to embark on a sea of perplexity. What would it be to construe substance as a summum genus? The natural temptation is to think of such a series of classificatory statements as

Fido is a dachshund
Fido is a dog
Fido is a brute
Fido is an animal
Fido is a corporeal substance
Fido is a substance
11. Parity of reasoning would lead us, in the case of quality, to some such sequence as
x is a red
x is a color
x is a perceptual quality
x is a quality
for if we tried
x is red
x is colored
.
.
.
.
it would be necessary at some stage to make a radical change in syntax in order to end up
with x is a quality.
12. On the other hand, if we stick with a.
x is a red
what sort of item could x be? One answer trips readily off the tongue: x is an abstract entity, a universal. On the other hand, we can appeal to the familiar, but heterodox idea of what have been called 'particularized qualities', or 'qualitative particulars.' Actually we have no choice, as we soon discover if we attempt to apply the second strategy to the case of relations. Only a series of even more desperate moves can keep us from falling immediately into the obvious absurdities exploited by Bradley.1 We are, therefore, committed to the former alternative, i.e. that what belongs in the place of 'x' is an expression which refers to an abstract entity. Our search for summa genera leads us to the sequences
Red(ness)2 is a color
Red(ness) is a perceptible quality
Red(ness) is a quality
and
Juxtaposition is a spatial relation
Juxtaposition is a dyadic relation
Juxtaposition is a relation.

13. But if the entities which belong to the categories of quality and relation are universals, then to assert that there qualities would seem to be to make an 'ontological commitment' to abstract entities.

14 . Now I take it as obvious that there are, in some sense of 'are', qualities and relations. So I take it as obvious that there are, in some sense of 'are', universals, and, in general, abstract entities. Yet is it so clear that the statements

There are qualities
There are relations
make an 'ontological commitment'? Or is it possible for a philosopher consistently to assert
There are qualities
but add, in a different philosophical tone of voice,
There really are no such things as qualities and relations?
I shall pick up this theme shortly.

IV

14. The empiricist tradition had little light to throw on categorial concepts. Much of what they had to say can be regarded as a reductio ad absurdum of the principle: nihil in intellectu quod non fuit prius in sensu. Roughly this principle tells us that, in the case of simple and most specific concepts, expressions which belong in the context
concept of x
must also belong in the context
impression of x.
Obviously, the more limited the scope of the latter context, the more limited the scope of the former. At the hands of Hume the principle led to a rejection of the traditional interpretation of substance as a simple idea. Yet even in his reinterpretation of substance it remains a summum genus; and, in general, the Aristotelean conception of categories as summa genera and, hence of categorial concepts as most generic concepts, lingered on, though particular categories were pruned or reconstrued. Thus Hume's intriguing account of generic ideas was intended to apply not only, for example, to the generic idea of triangularity, but also to such philosophically interesting ideas as those of substance, quality, relation, unity and even, it would seem, existence.

15. It is not my purpose in this essay to criticize radical concept empiricism. It is so vulnerable, indeed, that it is no easy task to relocate its insights in a correct account of the conceptual order. One would expect that, properly understood, like the theory of categories, it would turn out to be an a priori truth. Historically, of course, empiricists have never been clear as to the status of their fundamental principles. Are they, perhaps, appeals to Ockham's razor? or are they sweeping generalizations?

16. Another feature of the empiricist tradition is its 'logical atomism,' according to which every basic piece of empirical knowledge is logically independent of every other. Notice that this independence concerns not only what is known, but the knowing of it. The second dimension of this 'atomism' is of particular importance for understanding Kant's rejection of empiricism, although its relation to his theory of categories has not always been clearly understood. That the Wittgenstein of the Tractatus agrees with Kant in rejecting this dimension of the 'logical atomism' of the empiricist tradition -- not without raising unanswered questions about what is to count as a logical connection3 -- is equally relevant to his theory of categories. But this, also, is a topic for which preparation must be made.

V

16. The first major breakthrough in the theory of categories came, as one might expect, in the late middle ages, when logic, like knighthood, was in flower. A new strategy was developed for coping with certain puzzling concepts which were the common concern of logicians and metaphysicians. This strategy is illustrated by Ockham's explication of such statements as
(A) Man is a species
Roughly, he construes it to have the sense of
(B) ·Man· is a sortal mental term
where, since mental terms are to be conceived as analogous to linguistic expressions in overt speech, the quotation marks are designed to make it clear that in statement (A) we are mentioning a concept rather than using it as we would be if we were to judge that Tom is a man.

17. To this we must add, I believe, that whereas in (B) the expression '·Man·' presents itself as a name, it need not, and should not, be so construed. Its 'depth grammar' places it in quite a different box, a parallel will, perhaps be helpful. Consider the statement

(C) 'And' is a logical expression
It is clear that although the grammatical subject of this statement is a singular term,, it need not.be construed as a name. For, clearly, to talk about 'and' is to talk about occurrences of 'and,' (i.e. 'and's). We might, therefore, be tempted to say that in this context "and" is functioning as a general term. Taken seriously, however, this suggestion would require us to reformulate (C) as
(C1) 'And's are logical expressions
But, then, we remember the 'institutional' use of 'the' and see that we can have our cake and eat it too. For we can interpret our original statement (C) to have the form
(C2)(The) 'and' is a logical expression
which is the equivalent of
(C3) 'And's are logical expressions
in the sense in which 'The lion is tawny' is equivalent to 'lions are tawny.'

18. If, now, we take seriously the concept of thinking as 'inner speech' (Plato's 'dialogue within the soul'), then Ockham's analysis of

Man is a species
when developed along the lines just indicated, would construe (B) as
(B1) The ·man· is a sortal mental term
and hence, as equivalent to
(B2) ·Man·s are sortal terms.

19. Notice that although, according to this account, the original statement (A) is to be construed as referring to conceptual items, so that, in a sense, the expression 'man' in this context is equivalent to the expression 'the ·man·', it would be a mistake to rewrite the original statement to read

(A1)The ·man· is a species
For the context
---- is a species
already does the quoting, so that (A1) is, according to the analysis, to be construed as
(A2) The ·the· man·· is a mental sortal term
i.e.
(A3) ·The ·man··s are mental sortal terms
which is, however, false, for ·the ·man··s are mental singular terms. Thus if we introduce the phrase 'distributive singular term' for expressions formed by prefixing the institutional 'the' to a sortal expression, e.g. 'the lion,' then ·the ·man··s would be mental distributive singular terms.

20. Exactly the same point must be made about the statement

(D) Socrates is a substance, i.e. a primary individual.
By parity of reasoning it would have, using the above strategy, the form
(D1) ·Socrates· is a basic mental singular term
in other words
(D2) The ·Socrates· is a basic mental singular term
or, again,
(D3) ·Socrates·s are basic mental singular terms4'

21. Here, again, it would be a mistake to assume that if (D) is, in a sense, about conceptual items, i.e. ·Socrates·a, then it would be legitimate to rewrite it as

(D4) The ·Socrates· is a substance i.e. a primary individual.
For according to the analysis, this would be equivalent in sense to
·The ·Socrates·· is a basic mental singular term
i.e.
·The ·Socrates··s are basic mental singular terms
But the latter are false for although ·the ·Socrates·· are singular terms, they are not basic singular terms, but rather distributive singular terms, and, as such, to be classified with defined expressions. Thus, according to the above analysis, mental assertions or judgments with
·The ·Socrates··s
as their subjects are dispensable in favor of statements with
·Socrates··s
as their subject, just as, in overt speech, statements with
'The pawn's
as their subject, e.g.
The pawn is a chess piece
are dispensable in favor of statements with
'pawns'
as their subject, e.g.
Pawns are chess pieces.

22. What all this amounts to is that to apply Ockham's strategy to the theory of categories is to construe categories as classifications of conceptual items. This becomes, in Kant's hands, the idea that categories are the most generic functional classifications of the elements of judgments.

23. One might put this by saying that instead of being summa genera of entities which are objects 'in the world,' a notion which, as we saw, would force us to construe qualities, relations etc. as empirical objects, categories are summa genera of conceptual items. But while this is, I believe, the correct move to make, it raises the further question -- what is the sense of 'in the world' which applies to 'empirical objects' but not to conceptual items? Indeed, in the world seems to be another category which, if we are to be consistent, must itself be construed as applying to conceptual items.

24. Assuming, however, that this apparent difficulty can be met, let us watch the theory grow. In the first place, once we take this general line, we see that we might be able to distinguish between 'formal' and 'material' categories. For just as it is plausible to say that

Quality is a summum genus of entity
so it seems proper to say that
Color is a summum genus of perceptual quality
This suggests that an adequate theory of categories would involve a theory of distinction between 'determinables' and 'determinates.'

25. in the second place the theory suggests that in addition to such standard examples as 'substance,' 'quality,' 'relation,' etc., this list of categories should be expanded to include not only the 'modalities' but also 'state of affairs' as a 'formal' category, with 'event' and, perhaps, 'action' as 'material' categories subsumed under it.5

26. in the third place it is to be noted that implicit in the above account of categories is a theory of abstract entities. For if

Man is a species
is tantamount to
The ·man· is a sortal conceptual item
then we are committed to the idea that statements about the abstract entity man are dispensable in favor of statements about conceptual items, i.e. those acts and dispositions which involve 'predicates' to which the general term
·man·
would apply.

27. Of course, there is always the temptation to say that to be a mental term of the sort to which the general term '·man·' applies is to be a mental term which stands for the abstract entity man, the character being human. But the above approach can be generalized into the idea that every use of abstract singular terms is essentially classificatory, a matter of classifying conceptual items. Thus, to return to the linguistic level,

'Gelb' (in G) means yellow
and its more regimented counterpart
'Gelb' (in G) stands for yellowness
can be construed as classifying 'gelb's (in G) as ·yellow·s, i.e. as doing in German the job done in our language by the predicate 'yellow.'

28. If so, then

Yellow (yellowness) is a quality
would have the sense of
The ·yellow· is a (one-place) predicate
and 'reduce' to
·Yellow·s are predicates
where to be a ·yellow· is to be an item having a certain conceptual job, which would ultimately be explained in terms of the word-word and word-world uniformities by virtue of which 'yellow's in one language and 'gelb's in German function as they do in basic matter-of-factual statements.6

29. In the fourth place, while the theory permits us to say

Yellow is an entity
ascribing to this, roughly, the sense of
The ·yellow· is a meaningful Mentalese term
it must deny
Yellow is an individual
for the latter would have the sense of
The ·yellow· is a Mentalese singular term
which is false. In other words, 'Yellow is a primary entity' might be true, but not 'Yellow is a primary individual.' It would, therefore be correct to say that there are entities which are not individuals, an ostensibly paradoxical statement which, nevertheless, many perceptive philosophers have been led, by intuitive, if not always perspicuous, considerations to make.

30. Finally, notice that the theory would enable us to explain how a philosopher could be justified in acknowledging that

There are qualities e.g. triangularity
while denying that there really are qualities. For such statements about qualities as have the form
The ·triangular· is . . . .
can be paraphrased without the use of a singular term. Might one not acknowledge that there are chess pieces, e.g. the pawn, while denying that they really are chess pieces, e.g. the pawn. Thus a respectable and philosophically important sense could be given to the claim that although there really are particular conceptual episodes of thinking that something is triangular, there really is no such entity as the quality of being triangular.

VI

31. To appreciate the distinctive features of Kant's 'metaphysics of experience,' it is helpful to approach it via the early Wittgenstein, although this might seem an attempt to illuminate the obscure through the more obscure. For, contrary to the usual conception, Wittgenstein's views are actually as clear and straightforward as he thought them to be.7

32. Wittgenstein conceives of basic empirical truths as consisting of expressions referring to simple objects and predicates which stand for simple matter-of-factual qualities and relations. Expressions for complex individuals and complex characters can in principle be eliminated in favor of these 'elementary' expressions. Wittgenstein does not deny, as is often thought, that expressions which do not occur in basic statements (or are not definable in terms of them) can be meaningful. He simply denies that they refer to or describe objects in the world. Their, meaningfulness may consist in the fact that they enable us to formulate truths about our thoughts about objects in the world. One must also carefully distinguish between the way in which logical connectives are meaningful, although they do not refer to or characterize objects in the world, from that in which such meta-conceptual terms as 'object', 'quality' 'fact' etc. are meaningful.

33. This philosophical denial of existence to complex individuals and characters underlies Wittgenstein's claim that it is an essential feature of the objects, of empirical knowledge that they be simple and have simple matter-of-factual characters. Another essential feature is that these objects be the referents of referring expressions in the living language (and of the corresponding mental terms) of those for whom they are objects. Yet just as simplicity and individuality are not attributes of objects, although objects are simple individuals, so being the referent of referring expression is not a relation between an object and the expression.

34. This feature of Wittgenstein's thought has seemed to many to be a most intolerable paradox, for the following two statements are obviously true:

  1. The objects of empirical knowledge include referring expressions as linguistic or mentalese tokens (tokenings).
  2. Referring expressions- wouldn't refer to objects unless they stood in matter-of-factual relations to objects.
It is only too tempting to conclude from the truth of (A) and (B) that reference is a matter-of-factual relation, and hence that the character of referring to an object is a matter-of-factual relational property. Surely, if
(C) Expression E denotes something
entails
(C') Expression E stands in matter-of-factual relations to something
and
(D) Expression E denotes something
entails
(D') There is something, e.g. Socrates, to which E refers
then
E refers to Socrates
must have the form
R(E, Socrates).

35. But this would be a mistake, as can be seen from certain parallels. Consider

'Oder' (in G) stands for something
i.e.
There is something for which 'oder' (in G) stands, i.e. disjunction.
Clearly this entails that certain empirical facts hold of 'oder' in German. Yet it would be implausible to suggest that standing for is an empirical relation between the German word 'oder' and disjunction, or alternation or, to simplify the point, or-ness.

36. For, according to our strategy, to say what 'oder' specifically means or stands for is to classify 'oder's (in German) as ·or·s. And it is by virtue of being ·or·s, and not by virtue of standing in a supposed empirical relation of standing for, that certain empirical facts must be true of 'oder's.

37. Consider another parallel

(In our games of chess) tall cones play (serve as) bishops.
Clearly in order for tall cones to play the bishop, certain empirical truths must hold of tall cones (in the relevant contexts). But it is the criteria for being a bishop which carry these empirical requirements, not the supposed relation of playing or serving as. Indeed
Tall cones play (serve as) the bishop
can be paraphrased as
Tall cones are bishops
and, obviously, 'are' is not a word for an empirical relation.8

38. One who is half convinced might say that this is all very well and good with respect to the context

---- stands for ....
but does it throw light on 'denotes'? And, indeed, 'denotes,' or 'refers to' unlike 'stands for,' or its Fregean counterpart 'expresses,' is no mere specialized version of the copula. But one can grant this without granting that reference or denotation is an empirical relation.

39. Consider the contrast between

'Centaur' (in E) denotes (or refers to) nothing
and
'Centaur' (in E) stands for the property of being a Centaur.
The former differs from the latter not by saying that 'centaur' fails to stand in a certain empirical relation to anything, but by telling us that the kind -- left open -- for which 'centaur' stands is empty, i.e. is not exemplified by or true of anything. Again,
'Man' (in E) denotes featherless biped
differs from
'Man' (in E) stands for rational animal
Not by doing the radically different job of telling us that certain items stand in an empirical relation, but by telling us that the kind -- left open -- for which 'man' stands is exemplified by or true of all and only those items which also exemplify the character of being a featherless biped.

40. The same general account can be given of the denotation of singular terms, though the details must be left for another occasion.9 Thus

'Parigi' (in Italian) denotes the capital of France
would be construed as
'Parigi'(in Italian) stands for a concept which is materially equivalent to the concept ·the capital of France·.
Accordingly,
'Parigi' (in Italian) denotes something
although it would entail that the word 'Parigi' (in Italian) stands in empirical relations, these relations would not constitute a relation of denoting, but would be the relations which serve as criteria for the classification which is left unspecified by the statement. The cash for such denotation statements would be of the form
'Parigi' (in Italian) stands for the ·Paris· and the ·Paris· is materially equivalent to the ·the capital of France·
which imparts the information that 'Parigi's (in Italian) are ·Paris·s, and hence satisfy the criteria associated with this classification.

4l. I conclude that Wittgenstein was right in claiming that reference is not a matter-of-factual relation, although the fact that a term refers entails that it stands in certain matter-of-factual relations. That it stand, at least contingently, in such-and-such matter-of-factual relations to objects in the world is known by knowing its reference. Those in which it necessarily stands can only be determined by tracing out the implications of its sense, i.e. what it stands for.

42. Parallel points can be made with respect to predicates. Thus, for 'dreieckig's (in G) to stand for triangularity it must be true that they are caught up in certain word-word and word-thing uniformities. No more than in the other cases, however, do the latter constitute a supposed empirical relation of standing for.

VII

43. We can now turn our attention to the fact that for something to be an object of empirical knowledge, statements about it must be verifiable. There must, that is, be a method or strategy for deciding whether it is true or false. Here, again, the correct account has an initial appearance of paradox. Surely, to determine that a statement is true, is to determine that the statement corresponds to a fact. Yet facts are not objects, and on Tractarian principles only objects can have matter-of-factual attributes and stand in matter-of-factual relations.

44 . On the other hand, it is presumably a contingent or matter-of-factual fact that a given statement is true. How can we render consistent the ideas that

  1. facts are not objects;
  2. only objects can stand in matter-of-factual relations;
  3. the truth of a statement is a matter-of-factual relation between the statement and a fact?
Are not true statements those which correspond to facts?

45. The above treatment of categories, standing for, and denotation suggests the way out. According to it

'a is φ' (in L) corresponds to (the fact) that a_is_φ
no more asserts that 'a is φ' (in L) stands in a matter-of-factual relation to something in the world than
E denotes O
asserts that E and something in the world stand in a matter-of-factual relation of denotation. 'That a is φ' is to be construed as a distributive singular term ('the ·a is φ·') which, like 'the ·triangular·,' applies to conceptual items. Again, 'that a is φ is a fact' is to be construed, in first approximation, as having the sense of 'the ·a is φ· is correctly10 assertable.' Thus the original statement does two things: (a) it classifies 'a is φ's (in L) as ·*a is φ·s; (b) it tells us that the latter, which according to (a) include 'a is φ's (in L) are correctly assertable.

46. On the other hand, just as

E denotes O
Though it does not assert a matter-of-factual relation to obtain between E and O, nevertheless entails that E stands in a complex empirical relationship to O, so
'a is φ' (in L) corresponds to the fact that a is φ
while it does not assert that a relation obtains between 'a is φ' and the fact, nevertheless entails that certain matter-of-factual relations hold between 'a is φ's as belonging to L, and the object a -- not the fact that a is φ. It entails, that is, that 'a is φ's as conceptual objects stand in certain complex empirical relationships to other objects. Which objects is determined by the empirical relationships which must be satisfied if 'φ' is to stand (in L) for a certain characteristic and 'a' is to denote (in L) a certain object.

47. Thus the matter-of-factual characteristics of 'a is φ's (in L) as conceptual objects which are entailed by their truth do not constitute a supposed relation of "correspondence to a fact." They are, rather, matter-of-factual relations to genuine objects. In the case of basic empirical truths, on which the above account has been focused, these relations are different in each case, although they can all be subsumed under the general formula

Tokens of 'x is f' (and 'x R y') as expressions (in L) (or in Mentalese) have matter-of-factual characteristics by virtue of which they are linguistic projections -- in accordance with certain semantical uniformities -- of the objects to which these expressions refer.

48. We can sum this up as follows: In the case of basic empirical statements, unless the linguistic episodes which are their tokens stand in a projective relation to other objects in the world, they wouldn't be true. On the other hand, to characterize these statements as true (i.e. as corresponding to facts) is not to specify the projective relation in question.

49. The above line of thought is reinforced by considerations pertaining to logical and mathematical truth. Thus it is a fact that 2+2=4. And

'II+II = IV' (in L) corresponds to the fact that 2+2=4
trips readily off the tongue. Yet it is implausible in the extreme to suppose that in this context 'corresponds' stands for a matter-of-factual relation between 'II+II = IV' (in L) and an extra-conceptual entity. As before, 'corresponds to fact' dissolves into classification,
'II+II = IV's (in L) are ·2+2 = 4·s.
and the ascription of correct assertability to ·2+2 = 4·s. This time, however, the relevant criteria of correctness are intra-linguistic (or syntactical) and do not concern word-thing connections.

VIII

50. As the final step in this exploration of the concept of an object of empirical knowledge, I turn to considerations which were at the very center of Kant's thought,11 but at best implicit in the Tractatus, lurking, if at all, in his obscure dicta concerning the metaphysical 'I' and the status of the causal principle.

51. If empirical knowability always knowability by a person here and now, whereas the scope of the knowable includes facts about the there and then; or if (to abstract from the specific conditions of human knowledge and move to the 'pure pragmatics' or 'transcendental logic' of empirical knowledge as such) knowability essentially involves a perspectival relationship between act of knowing and object known, must not the knowability of objects consist, in large part, of inferential knowability. (One can, of course, recognize the essential and irreducible role of inference without denying the existence of non-inferential knowledge of the here-now.) On the other hand we have learned from Hume that facts about the there and then are never logically implied by facts about the here and now, but, at best, by the latter together with the general facts captured by true law-like statements.

52. Now if we assume in the spirit of Hume that it is a contingent fact that such general facts obtain, it would seem to follow that it is a contingent fact that there-then objects are knowable. But if to be an object is to be a knowable, our conclusion would have to be that it is a contingent fact that there are there-then objects. But surely any here-now object is on its way to being a there-then object in the past and on its way from having been a there-then object in the future.

53. If so, then there would seem to be a logical inconsistency in granting the existence of here-now objects while denying that of there-then objects. A transcendental argument does not prove that there is empirical knowledge -- what premises could such an argument have? -- nor that there are objects of empirical knowledge. It simply explicates the concepts of empirical knowledge and object of empirical knowledge. Thus, to admit knowing that it now seems to me that there is a red and triangular object over there is to admit knowing that this was about to seem to me to be the case. If the skeptic (after making a similar move with respect to Space) attempts to replace the now the seeming by the semblance of a now, by putting Time itself into the content of that which seems, does it not reappear (at least implicitly) outside this content, -- thus: It (now) seems to me that there is such a thing as Time (an order of before and after) in the now of which (and as Space in the there of which) there is a red and triangular object, (and does it merely seem to me that there is such a thing as I?

54. What Kant bakes himself to have proved is that the concept of empirical knowledge involves the concept of inferability in accordance with laws of nature. To grant that there is knowledge of the here and now is, he argues, to grant that there are general truths of the sort captured by lawlike statements. As far as specifically human knowledge is concerned, he was convinced that the idea that knowable objects are located in Space and Time carries with it certain general commitments as to the form of these laws. These commitments could, he thought, be known a priori or non-inductively. Thus, the transcendental knowledge that spatio-temporal objects of knowledge must conform to certain generalizations which are themselves logically synthetic is itself, according to Kant, analytic.

55. Notice that the full expression of what is known in synthetic a priori knowledge has the form

Spatio-temporal knowables must conform to (synthetic) general truths satisfying such and such conditions
i.e.
If there is knowledge of spatio-temporal objects, then these objects conform to general truths satisfying such and such conditions.
This statement as a whole is an analytic or explicative statement belonging to transcendental philosophy. If we are willing to affirm the antecedent -- that is, if we are willing to grant, as even Hume does, that we do have knowledge of the here and now, then we can affirm the consequent, i.e.
Spatio-temporal objects in my world conform to general truths satisfying such and such conditions.
It is, however, essential to note that the latter, by itself, is not a necessary truth, except in the derivative sense that it is necessary relative to the antecedent. To construe it as intrinsically necessary is to commit a modal fallacy of a piece with
Necessary (if all men are bipeds then all Texans are bipeds)
All men are bipeds
So, necessary(all Texans are bipeds).

56. It is this conception of transcendental philosophy which distinguishes the critical rationalism of Kant (and the Wittgenstein of the Tractatus) from both dogmatic rationalism and the naive empiricism which thinks that empiricism is an empirical 'ism.'

Pittsburgh, December 12, 1968
March 1, 1969