In order to evaluate the dispute over the nature of intentionality which Chisholm and Sellars have expressed in their published correspondence [Chisholm and Sellars], it will be important to examine in some detail both the respects in which Chisholm's and Sellars' concerns diverge, and the sort of views which, independently of this dispute, Sellars brings to bear on their disagreements. It is to these two preliminary tasks which the current chapter is devoted. The largest part of this chapter will be concerned with an exposition of Sellars' views on the philosophy of mind, insofar as they have a bearing on his views on intentionality and the conflict of these views with those of Chisholm. It is therefore important to begin, in the first section of this chapter, by summarizing certain features of Chisholm's position and attempting to capture certain respects in which the divergence exhibited in their published correspondence is a result of their differing concerns and interests with regard to intentionality, rather than due to a substantive disagreement.

Much of the foregoing discussion of Chisholm's views has been confined to considering his proposals for marks of intentionality, in the sense of proposals for features which might be exhibited by all and only (or perhaps simply only) psychological expressions but are logically independent of the fact that the expressions so determined are psychological in subject matter. This is a result, in large part, of the fact that Chisholm's discussions of intentionality are themselves for the most part confined to such matters. It is not at all with regard to such a mark of intentionality, however, that Chisholm and Sellars disagree. Nor has Sellars made any proposals concerning such issues. As a consequence, the evaluation of their disagreements and the attempt to isolate those aspects of Sellars' position which have a bearing on this disagreement will involve becoming somewhat more clear about, given the hypothetical adequacy of some mark of intentionality envisaged by Chisholm and the resulting truth of an appropriate thesis of intentionality, Chisholm's views on the implications of this thesis. In particular, it will be useful to see what, if anything, Chisholm has to say by way of explanation of his conviction that some formal mark of intentionality can be found (in the sense of a mark which picks out expressions used to describe psychological phenomena independently of the subject matter such expressions have), and what Chisholm believes to be the philosophical significance of establishing some corresponding thesis of intentionality.

I. Chisholm begins to consider the philosophical significance of his thesis when he writes, toward the close of "Sentences about Believing":

Let us suppose for a moment that we cannot rewrite belief sentences in a way which is contrary to our linguistic version of Brentano's thesis. What would be the significance of this fact? [Chisholm (17) 519, emphasis original]

Chisholm points out that philosophers who have argued that, contra Brentano's thesis, psychological sentences can be analyzed in non-psychological terms have felt that that (putative) fact is philosophically significant. If we suppose, then, that Chisholm is correct in believing that any such program is unrealizable in principle, defending this conviction by appeal to some (adequate) mark of intentionality, then that fact should similarly have some philosophical significance. What does Chisholm believe to be the philosophical significance of his position on this matter?

Chisholm does not appear to answer this question directly in the course of his writings on intentionality, except to the extent to which his correspondence, to be discussed in Chapter IV, is concerned with this topic. He does, however, indicate his belief that certain sorts of answers to this question are mistaken, and he provides us with reasons for this belief. It is by considering these, then, that we may be able to form an adequate view of his position on this matter.

There are two explanations of the logical behavior of sentences about psychological phenomena as described by Chisholm's thesis in "Sentences about Believing" which, at the end of that paper, he considers and rejects. The first is that psychological sentences "are not factual; they are not descriptive; they don't say things about the world in the way in which certain non-psychological sentences say things about the world." [Chisholm (17) 519] The second explanation is that such sentences do not "say of the world what at first thought we tend to think they say of the world." [Chisholm (17) 519] The first explanation involves claiming that psychological sentences are misleading in the respect that while they appear to be sentences which are used to make assertions in the way in which 'There is a cat on the mat.' might so be used, they really are not. They must, if they are adequately to be understood, be construed as sentences which are used non-descriptively, and non-assertively, in an ordinary sense of 'assertive' and 'descriptive'. If this move is adopted as a means to explain the (putative) fact that psychological sentences are unanalyzable in non-psychological terms, then some account of how these psychological sentences are in fact used must be provided. For unless such an account is given, it is not at all clear what is involved with the claim in question.

Similarly, the second sort of explanation involves claiming that although psychological sentences are both assertive and descriptive, in some appropriate ordinary sense of the terms, they do not assert or describe what we might expect they do, if we consider them on the model of standard cases of non-psychological sentences: we are, we might say, misled by their "superficial grammar." Chisholm suggests that according to this move they are rather like 'The average carpenter has 2.7 children.', 'Charity is an essential part of our obligations.', and 'Heaven forbid.' (these are Chisholm's examples) "in that their uses, or performances, differ in very fundamental ways from other sentences having the same grammatical form." [Chisholm (17) 519] This sort of move, like the first sort, is incomplete, however, unless the move includes or involves an account of the difference between the sort of use or application which is peculiar to psychological sentences and that which is peculiar to, say, standard empirical sentences (as well as, perhaps, to other sorts of non-empirical or non-standard empirical sentences such as those suggested by Chisholm as examples above). For just as in the first case, unless we are clear about what is being said about psychological sentences, it is difficult to see how we might become clear on the nature of the putative difference. Chisholm expresses this incompleteness of both sorts of explanations, as they stand, by arguing that in order to offer one of these explanations one must, to make it stick, offer a positive account of the nature of psychological sentences.

The two sorts of explanations which Chisholm considers and rejects for the reasons mentioned above would be invoked to show that, in spite of the truth of the thesis Chisholm has defended, this thesis has no philosophical significance. For, the argument would run, even though the adequacy of Chisholm's thesis would involve concluding that we cannot analyze psychological sentences into non-psychological terms, this conclusion is to be understood as no more than a relatively trivial consequence of the (putative) fact that psychological sentences do not behave logically at all like the empirical sentences into which someone who disputed the linguistic version of Brentano's thesis might wish to analyze psychological sentences. In rejecting these two explanations, Chisholm conceives of himself as arguing against the position that the truth of his thesis would have no philosophical significance. We might wish to ask, then, given that we may accept for the purposes of the present discussion that some version of Chisholm's thesis is true, what positive account or explanation of its truth Chisholm might offer which does not, as the two accounts rejected by Chisholm do, deprive the thesis of philosophical significance.

It might be thought that Chisholm has, as a goal, no more than to show that psychological sentences make up a class of locutions which all have in common some feature which (a) is logically independent of the fact that they are psychological, and (b) is not exhibited by any non-psychological sentences. If this were so, then we might understand Chisholm's views as follows. We are, in fact, inclined (so we might envisage Chisholm arguing) to suppose that psychological expressions do exhibit some such feature, for we are inclined to believe (a) that their subject matter is irreducibly different (in some appropriate sense) from other subject matters and (b) that the logical behavior of sentences reflects, in some appropriate way, their subject matter. Thus, one might be envisaged as supposing, on an intuitive basis, that sentences which have radically different subject matters exhibit correspondingly different logical features. Thus Chisholm's concern to find an adequate mark of intentionality might be seen as an attempt to specify just how such sentences differ, aside from with regard to their subject matter. The attempt to specify this would be seen as philosophically important because unless the difference in logical behavior were exhibited, it might be false that there was any such difference at all.

If this were Chisholm's goal in propounding a thesis of intentionality, the necessity, from his point of view, to present an explanation for the logical behavior of psychological sentences would be obviated. Thus Chisholm could, conceivably, establish his point without ever be[eing] required to present such an explanation. Sellars' concern with intentionality, however, is in great part to provide such an explanation. While Sellars would be in agreement with Chisholm that if true, Chisholm's thesis is of philosophical importance, the importance would, Sellars believes, lie in the fact that it would corroborate certain views which Sellars wishes, on independent grounds, to hold. The correspondence between the two (to which we shall turn in the next chapter) involves Sellars in presenting these views, and in the course of presenting them, engaging in certain disputes with Chisholm. It is there that Chisholm appears to come the closest to giving his belief as to what, aside from defending the intuitive beliefs one might hold which were presented in the foregoing paragraph, the philosophical significance of the truth of his thesis would be. It is also interesting that it is on just this point which, we shall see, the major differences between Chisholm and Sellars seem to hang. It is important, then, for us now to turn to an examination of Sellars' position as it relates to these issues.

II. In very broad terms, Sellars characterizes what he calls the problem of intentionality as one of the major strands of the mind-body problem, specifically

the problem of interpreting the status of the reference to objects and states of affairs, actual or possible, past, present or future, which is involved in the very meaning of the 'mentalistic' vocabulary of everyday life. [Chisholm and Sellars 507]

This rather diffuse characterization of the problem of intentionality exhibits one feature which is crucial to an understanding of Sellars' views. Like Chisholm, Sellars appears to believe that what is involved in making reference to items of one or another sort is peculiar and involves intentionality--in Chisholm's view, this might correspond to his claim that sentences about the meanings and uses of sentences are, like psychological sentences, intentional; for on Chisholm's view, they are implicitly about psychological phenomena. Similarly, Sellars seems to hold that "the 'mentalistic' vocabulary of everyday life" is similar, in certain relevant respects, to language used in characterizing what is involved in reference to various sorts of items. This initial characterization also seems to suggest what appears to be an important difference between Sellars' and Chisholm's views, however. For while Chisholm begins with some intuitive notion of what it is for a sentence to have psychological subject matter, and argues that sentences about the meanings and uses of expressions have, in effect, such subject matter, Sellars' starting point in his characterization of intentionality is rather the idea of expressions being about this or that item--that is, expressions having reference of one sort or another. For having characterized the problem in such terms, he then goes on to claim that to consider such a problem is also to consider the peculiar nature of psychological locutions.

For our present purposes, however, it is more important to note that the problems which seem to face the two are, aside from the similarity noticed above, very divergent. While Chisholm is concerned to advance some criterion of intentionality in order to defend his thesis, and then asks what the significance of that thesis might be, Sellars is not, as we have noted above, concerned with such a thesis. For him the problem is rather one of interpreting "the status" or nature of the referring expressions we use, and the reference they make, given that the nature of sentences about the reference of expressions is similar to the nature of psychological sentences. Thus, although we shall see, in the course of Chapter V, that there is also disagreement on the nature of this similarity of the two sorts of sentences, Sellars' concern is initially to explicate the nature of such sentences, rather than to defend the view that there is some logically describable way in which, independently of difference of subject matter, such sentences differ from all others.

To make this point somewhat clearer, let us turn to the question of the similarity between psychological sentences and sentences concerned with the meanings and uses of expressions. In Chapter II we were concerned with sentences like 'George is talking about Pegasus.', which seem to be about speech acts of one sort or another. Following what appeared to be Chisholm's lead, it was argued that such a sentence cannot be true unless George uses the words he uses meaningfully, knowing what they mean and intending, in using them, to be talking about Pegasus. (The sense in which George might be talking about Pegasus by using words to which he had assigned meanings different from those commonly accepted may be ignored.) Thus, it was argued, our sentence entails a sentence which involves claiming that George exhibits certain psychological phenomena, and for this reason, we concluded, the original sentence was implicitly psychological in subject matter.

We may, by way of comparison, consider the following statement of Sellars':

It is important to distinguish between two senses of 'meaningless utterance' (a) An utterance is meaningless if it does not token a properly formed expression in a language. (b) An utterance is meaningless if it is uttered parrotingly by one who does not know the language [Sellars (8) 657]

If an utterance is meaningless in sense (a), this, we may imagine, is a matter which can be captured by means of specifying the formal requirements of meaningful utterances in a given language and noting that the utterance in question fails to satisfy the relevant formal requirements. 'Meaningless' in sense (b) cannot, however, be handled in this fashion. (We might wish to add that an utterance might be meaningless in this sense even if, although the speaker knew the language, he nonetheless uttered it parrotingly.) For meaningless in this sense is not a matter of utterance types, as it is in the first sense, but of utterance tokens produced by a given person on a given occasion. Using this distinction, one might recast the argument of the foregoing paragraph as follows. If what George utters is to be meaningfully uttered, if, that is, it is to be uttered in the performance of one or another speech act, then it must be uttered non-parrotingly. To specify what is meant, in this context, by 'non-parrotingly', however, involves us in saying things such as that George knew what the words he used mean, and used them with the intention of conveying such-and-such. These circumstances are, however, psychological facts, and, according to Chisholm's thesis, require psychological language to be described as such. If we take for granted, then, that 'George is talking about Pegasus' entails that George uses certain words non-parrotingly to say something about Pegasus, then our sentence must be understood as asserting something about psychological phenomena.

These are clearly cases which might be offered in support of Chisholm's contention that sentences "we use to describe the meanings and uses of words" are covertly psychological. [Chisholm (17) 516-517] For on Chisholm's view, such sentences involve claims about people's use of certain words "when they wish to express or convey something they know or believe--or perceive or take--with respect to so-and-so " [Chisholm (17) 517, emphasis original] Semantical sentences in general, sentences, that is, which describe the use, application and reference of certain words and sentences, are according to Chisholm psychological; for they assert something about people's use of expressions in connection with their beliefs, knowledge or perceptions. The fact, recognized by both Chisholm and Sellars, that semantic phenomena are similar in nature to psychological phenomena is thus accounted for by Chisholm on the basis of his contention that semantic phenomena are phenomena which involve certain psychological phenomena. For on Chisholm's view, a part of what is being described when semantical phenomena are being talked about is certain psychological phenomena. This contention will be important in examining the dispute between Chisholm and Sellars.

It is against such an account that Sellars' views appear to stand out most clearly. For according to Sellars, the intentionality of mental phenomena is simply the intentionality "of those features of mental acts by virtue of which they are analogous to conventional speech " [Sellars (15) 211] Put differently, although somewhat more cryptically, "the categories of intentionality are nothing more nor less than the metalinguistic categories in terms of which we talk epistemically about overt speech as they [the metalinguistic categories] appear in the framework of thoughts construed on the model of overt speech." [Chisholm and Sellars, 522]

There are several things which stand out in these claims, especially in contrast to the views of Chisholm which we have examined. First, the fact that psychological and semantical phenomena are similar in regard to certain aspects of their nature (in respect, that is, of the fact that they are both intentional) is not, on Sellars' view, to be accounted for on the basis of supposing that semantical phenomena involve psychological phenomena, in the sense that semantical phenomena are, in part, psychological phenomena. For unlike Chisholm, Sellars does not argue that part of what is being talked about when we talk about semantical phenomena are certain psychological phenomena accompanying episodes (utterings and mark-makings) which would not be intentional were they not so accompanied. On Sellars' view, the common nature of the two sorts of phenomena is rather to be accounted for by the fact that both simply do have properties which are similar.{1} Secondly, the way in which Sellars supposes that we obtain some grasp of this common nature shared by the two sorts of phenomena is not by supposing ourselves first to understand the nature of psychological phenomena, but rather by supposing ourselves first to have an independent grasp of the nature of semantical phenomena (Cf. [Sellars (13) 107].) On this view, semantical phenomena may, just as psychological phenomena are, be characterized as exhibiting properties of the some sort--properties which we may call intentional properties. A more detailed examination of the differences separating Chisholm and Sellars, however, will be undertaken in the following chapter. At this point then, we may turn to our account of what, beyond this initial characterization, Sellars' views consist in.

III. We have seen that Sellars' view involves a certain analogy between the non-parroting use of language and psychological phenomena of certain sorts. In order to sketch what Sellars believes is involved in this analogy, it will be useful first to examine, on an intuitive basis, what one might wish to say about non-parroting speech episodes which would not beg any of the questions involved in framing the analogy between them and psychological phenomena of some sort. The following sort of account may be imagined. It is clear that non-parroting speech episodes are tokens of certain expression-types belonging to a given language, and describable in terms which make reference only to the physical properties of linguistic expressions. Thus we should be able to describe occurrences of such episodes in straightforwardly empirical terms For we may say that a given such episode exhibits certain empirical properties in accordance with the requirements which any episode must satisfy in order to be classified as tokening a certain type. We may further describe the episode, in straightforwardly empirical terms, by specifying when and where it was produced, by whom (or by what), and by specifying peculiarities of the way in which the particular episode tokens the corresponding type--say, the loudness of the episode, the pitch, or the rapidity with which it was uttered.

By hypothesis, however, there is more than this to say about such an episode for we may at the least say also that it was non-parrotingly uttered, and, moreover, we may, in some intuitive terms, specify what it means. These ascriptions of properties to the episode we may think of as ascriptions of the intentional properties of the episode. Thus if we say, of a given episode, E, that E was non-parrotingly uttered, or that E means (that-) such-and-such, or that E is about so-and-so, we shall be ascribing to E certain intentional properties. In order not to beg the questions considered, for example, in section II of the current chapter, some care must be taken on this point. Thus without engaging in any attempted explanation of the distinction between parrotingly and non-parrotingly produced episodes, nor in attempting to give an account of what it is to say what an utterance means, or what it is about, or what it makes reference to, we shall simply suppose ourselves, at this stage, able to specify these matters.

It will be useful in what follows to be able to distinguish between these two ways of talking about intentional items, between describing them by reference to their empirical properties, on the one hand, and ascribing intentional properties to them, on the other. In order to mark this distinction, it will be useful to use the term 'determinate factual character', which is used by Sellars in discussing his views about psychological episodes (Cf. [Sellars (8) 663].), although the precise application which we shall give to this term is not intended necessarily to reflect the use which Sellars makes of it. Thus the 'determinate factual character of a thing' will be used, in what follows, to apply to the totality of the empirical properties of that thing which are not specified by means of semantical, metalinguistic, or other intentional locutions. Thus to say of a thing that it means this or that, or that it is about, or refers to, this or that, will be understood as saying something of that thing which is not a partial specification of its determinate factual character. Following the usage in the foregoing paragraph, to say such things will be to ascribe to the thing certain intentional properties. To say of a thing that it has a certain color, shape, sound, size, and so forth, or that it is causally (or spatially or temporally) related in a certain way to something, will be (at least partially) to specify the thing in terms of its determinate factual character.

It is important to note, in this connection, that the distinction between the intentional properties of an (intentional) item and its determinate factual character is a matter of how the thing in question is being specified. Thus it is not ruled out that certain (and possibly all of the) intentional properties of a thing might be specifiable not only by means of semantical or other intentional locutions, but also, say, in terms which are causal. In such a case, the intentional properties in question would be part of the determinate factual character of the thing in question--the difference being that in such a case, the same properties would be viewed as intentional properties when specified in one way, and as part of a determinate factual character when specified in another. On the other hand, there is of course nothing in the way in which the distinction has been drawn which suggests that there should be any properties which are specifiable in both ways. In any event, in order to settle such a question, it would be required that we have a way of determining when two linguistic expressions which are used to specify properties are used to specify the same property. As we shall see in the course of the last chapter, it is not clear that it is required, for the purposes of the current discussion, that we arrive at any conclusion concerning this question{2}

Let us, then, return to the analogy which Sellars wishes to draw between such episodes and certain psychological phenomena. It might be thought that the basis of such an analogy could reasonably consist in the facts that both mental acts and speech acts may be about things, and that both sorts of phenomena are commonly reported by means of indirect discourse--as, for example, George thought that . . . , or George said that . . . .{3} The analogy which Sellars constructs, however, takes its beginning point from a certain problem which, in "Empiricism and the Philosophy of Mind," he poses. The problem is the following. Let us imagine, for logical examination, "a stage in prehistory in which humans are limited to what I shall call a Rylean language, a language of which the fundamental descriptive vocabulary speaks of public properties of public objects located in Space and enduring through Time." [Sellars (5) 178] This language has a number of resources of ordinary natural languages, such as elementary logical operations, quantification, and the subjunctive conditional, and its empirical terms are characterized by what has been called, following Waismann [Waismann 119 ff.], vagueness and open texture. This language is not only "a behavioristic language, but a behavioristic language which is restricted to the non-theoretical vocabulary of a behavioristic psychology." [Sellars (5) 186]

Having imagined such a language, devoid of theoretical discourse, psychological discourse and semantical discourse, we are in a position to pose Sellars' problem, which Sellars raises by means of two questions (1) "What resources would have to be added to the Rylean language of these talking animals in order that they might come to recognize each other and themselves as animals that think, observe, and have feelings and sensations, as we use these terms?"{4} and (2) "How could the addition of these resources be construed as reasonable?" [Sellars (5) 179] The first question involves a concern with the idea that even if these talking animals cannot, by hypothesis, talk about their psychological states, they have them, and thus a concern as to how they might come both to be able to talk about them and recognize themselves as having them. The second question involves the requirement that whatever answer is given to the first should, judged by certain standards of reasonableness, not be ad hoc or implausible. It is of course not to be required that the answer be historically accurate, in the sense of tracing the development which lies behind the present state of human thought and language-using.

The first of the two resources which Sellars believes would be each necessary and jointly sufficient for these purposes is semantical discourse, the resources, that is, which are "necessary for making such characteristically semantical statements as ' 'Rot' means red', and ' 'Der Mond is rund' is true if and only if the moon is round'." [Sellars (5) 179] Given such resources added to our Rylean language, Sellars contends that its speakers would be enabled

to characterize each other's verbal behaviour in semantical terms; . . . they not only can talk about each other's predictions as causes and effects, and as indicators (with greater or less reliability) of other verbal and nonverbal states of affairs, but can also say of these verbal productions that they mean thus and so, that they say that such and such, that they are true, false, etc. [Sellars (5) 179, emphasis original]

"With the resources of semantical discourse," Sellars continues,

the language of our fictional ancestors has acquired a dimension which gives considerably more plausibility to the claim that they are in a position to talk about thoughts just as we are. For a characteristic of thoughts is their intentionality, reference, or aboutness, and it is clear that semantical talk about the meaning or reference of verbal expressions has the same structure as mentalistic discourse concerning what thoughts are about. [Sellars (5) 180, emphasis original]

It is clear that Sellars is here invoking the idea, discussed in section II, that psychological phenomena and semantic phenomena exhibit certain properties in common--what we have called intentional properties, and what Sellars wishes to capture by pointing out that both sorts of phenomena are about things, and refer to things. Thus, Sellars is arguing, if we imagine these talking animals as able to talk about the way in which semantical{5} episodes are about things and make reference to things, then we have come a long way to imagining them able also to talk about psychological phenomena being about and making reference to things.

It is important for the following discussion that it be recognized that the fact that semantical episodes exhibit intentional properties is, at least in part, a matter of their being able to be used in framing certain explanations of accompanying nonverbal behavior, viewed as intelligent or rational. For it is one way of coming to understand what is involved in such nonverbal behavior that it be explained by reference to accompanying verbal behavior, or, at least, by reference to such verbal behavior as we can envisage appropriately accompanying the nonverbal behavior in question. Thus, we come to understand certain courses of action by reference to the semantical episode which might be used to state the motivation or intention which is involved in such behavior; similarly, we come to understand nonverbal behavior by reference to semantical episodes which would express the perceptual state of the perceiver, or the frame of mind (as we might put it) of the perceiver. The relation of this sort of coming to understand bits of nonverbal behavior to other sorts of explanations is not at issue for the point which is important to Sellars' account.{6} What is important is simply that it is reasonable, on occasion, to accept as one kind of explanation of behavior the sort which makes reference to the subject's verbal description of his own behavior and the subject's own description of his perceptual states. Thus the fact that verbal reports of what sort of behavior one is engaging in and how one views one's perceptible environment constitute reliable indicators concerning the sort of nonverbal behavior which is taking place, allows us to frame the explanations in question.{7}

The only problem at this stage would be that although our fictional ancestors are envisaged as able to talk about certain properties which psychological phenomena, like semantic ones, exhibit, they are envisaged so far as having the idea of only one sort of item which is about things--namely semantical episodes. What is wanted, then, is a further enrichment of the linguistic resources of these speakers to allow them to form the idea of an item which is about something (henceforth the term 'intentional item' will be used to denote items which are about something--items, which in Sellars' terms, exhibit intentionality), but which is not a semantical episode

The additional resource which Sellars wishes to add to our semantically enriched Rylean language is that of theoretical discourse. Several aspects of Sellars' views on the nature of theoretical discourse are important for the use he proposes to make of such discourse as a requirement for talking about psychological phenomena. We shall turn to these in a moment, after having made some preliminary remarks by way of specifying more exactly the nature of the analogy Sellars wishes to defend. As we have noted, semantical episodes are characterizable in terms of their determinate factual character, and it is important for the analogy Sellars has in mind that the semantic phenomena are episodic in nature. This suggests that the psychological phenomena which Sellars believes are analogous to semantic phenomena are also episodic. It seems clear that at least some sentences which are clearly psychological as regards their subject matter are not about episodes, except, perhaps, indirectly. Thus sentences about beliefs are reasonably understood as typically about certain dispositions of a person. In framing his analogy, however, Sellars wishes temporarily to restrict consideration to sentences which are about episodic psychological phenomena--phenomena which may be thought of as beginning at a certain time and having a certain duration. For these purposes, we shall imagine the having of a thought as such an episode, speaking indifferently of such phenomena as havings of thoughts and as thought episodes (or, briefly, thoughts). Let us turn, then, to a brief consideration of Sellars' views about theoretical discourse, in an effort to see how the addition of such discourse to our semantically enriched Rylean language would allow us (1) to provide an answer to Sellars' problem, stated above, and (2) to understand the analogy between thought episodes and semantical episodes which Sellars wishes to defend.

"Informally," Sellars writes,

to construct a theory is, in its most developed or sophisticated form, to postulate a domain of entities which behave in certain ways set down by fundamental principles of the theory, and to correlate--perhaps, in a certain sense to identify--complexes of these theoretical entities with certain non-theoretical objects or situations. [Sellars (51 181]

Given this characterization, in general terms, of theory construction, there are two qualifications which are crucial for Sellars' purposes. The first is that such postulated entities may be formulated by analogy with certain other entities understood independently of the theory, these latter constituting what Sellars calls a model for the theoretical entities.{8} The second qualification consists in Sellars' adoption of what has been called a realistic view concerning theoretical entities: such entities are not, on Sellars' view, non-existent items about which it is, nonetheless, convenient to talk. [Sellars (5) 182-183] As Sellars expresses the point elsewhere,

whereas according to the 'positivistic' account of theoretical entities, 'Xs are theoretical entities' has roughly the sense of ' 'X' is a non-logical symbol in a calculational framework for ordering our knowledge about (observational) entities,' on my view 'Xs are theoretical entitles' has roughly the sense of 'Xs are entitles whose existence has been established by theoretical reasoning.' [Sellars (17) 12]

Thus on Sellars view, if the theoretical reasoning on which the existence of certain items is based is sound, then we have good reason to believe that the items exist, just as if the non-theoretical reasoning used to establish the existence of other sorts of items is sound, then we have good reason to believe that they exist.{9} [Sellars (17) 12]

The two resources which we have added to our Rylean discourse, semantical and theoretical discourse, are sufficient, Sellars believes, to generate discourse about thoughts. We have seen what function semantical discourse might be supposed to serve in constructing psychological discourse; it remains to ask how theoretical discourse is to be helpful in this connection. Recalling that our concern is with what we have called thought episodes, the move which Sellars makes is the following. We may, he suggests, view thought episodes as theoretical entitles, and discourse about them as a certain sort of theoretical discourse. The model for such entitles would be overt speech episodes, which, like havings of thoughts, are episodic and exhibit intentional properties.

It is useful to be specific concerning just what sort of theorizing is envisaged here. We may, writes Sellars, suppose there to have been a theoretician called Jones who developed, given our theoretically and semantically enriched Rylean discourse, a theory according to which overt [meaningful] utterances are but the culmination of a process which begins with certain inner episodes." [Sellars (5) 186] The point of developing such a theory lies in Jones' attempt to "account for the fact that his fellow men behave intelligently not only when their conduct is threaded on a string of overt verbal episodes--that is, as we would put it, when they 'think out loud'-- but also when no detectable verbal output is present." [Sellars (5) 186] Thus, as a part of the prepsychological resources of Jones and his contemporaries, we may imagine them able to distinguish cases of intelligent behavior from nonintelligent behavior, and that they are able to render patterns of behavior intelligible by relating them to concurrent overt verbal behavior. While such explanation of nonverbal behavior by reference to verbal behavior is adequate when there is accompanying verbal behavior, it is of course inadequate in cases in which there is none. Jones' theory therefore involves supposing that such overt verbal episodes are the causal result of certain intentional episodes which we may for convenience call inner episodes, and that they occur even when no verbal behavior does, and may therefore be used as verbal episodes had been to explain nonverbal behavior (as well as occurrences of verbal episodes, of course).{10}

If the way in which explanation of nonverbal intelligent behavior is made by reference to overt verbal episodes is to involve the fact that they are about certain things, that is, that they exhibit intentional properties, then it is to be expected that the inner episodes postulated by Jones should also be conceived of as exhibiting intentional properties. There are therefore three specifiable ways in which, on the basis of such a theory, we may suppose such inner episodes to be related to speech episodes. (1) Since the former are "but the culmination of a process which begins with certain inner episodes," [Sellars (5) 186] thoughts are conceived of as causally related to speech acts. (2) These inner episodes are theoretical entities modeled on the very items to which, according to the theory, they are causally related. In particular, although thoughts are not conceived of as exhibiting any of the determinate factual character of overt verbal episodes (such as sounding loud, or guttural, and so forth), they are conceived of as exhibiting the intentional properties which speech acts exhibit. Thus we may speak of both sorts of episodes as having reference, and as being about this or that. (3) Since we have imagined Jones' theory as allowing us to explain what used to be explained by certain semantic episodes having certain intentional properties by reference to thought episodes which cause such semantic episodes, it is reasonable to stipulate, as part of the theory, that the intentional properties exhibited by a given thought are just those exhibited by the speech episode which it tends to cause.{11} Whatever, that is, a given speech episode is about or has reference to, is just what the thought episode which is according to the theory causally related to it is about. Thus not only does the theory carry "over to these inner episodes the applicability of semantical categories" as provided in (2) above, but it carries them over in such a way that the thought and speech episodes which are causally related also exhibit the same intentional properties. That this is a part of the theory is important, as we shall see in the course of Chapter V, for providing a means for picking out cases of thoughts according to the intentional properties the various cases exhibit. {12}

IV. Having provided an account of what Sellars believes is the answer to his first question, namely, what resources need to be added to Rylean discourse in order to generate discourse about psychological phenomena, it will be useful to take note of a number of features of this answer before turning to examine Sellars' answer to his second question--the question concerning how these features might have been added.{13} A first feature of Sellars' account of the generation of discourse about thoughts seems fairly clear: given Sellars' notion of what is involved in theoretical discourse, one must not conclude that because thoughts are construed as theoretical entities, they therefore do not really exist. For if we accept that "the framework of thoughts developed as a theory develops, with semantical discourse about overt verbal episodes as its model," [Chisholm and Sellars 536] and if we accept the theoretical reasoning involved as sound, then we are, on Sellars' view, committed to the existence of thoughts. In particular, it is not to be supposed that thoughts are simply invented by Jones; for if his theory is a good one, then thoughts had already existed. "Pre-Joneseans have thoughts (though they are not aware of themselves as having them)." [Sellars (17) 4] The theory provides Jones' contemporaries with the ability to talk about the thoughts they have, and even the thoughts they had had before the theory was expounded and adopted.

A second feature of Sellars' account is crucial to the analogy which, we have seen, he wishes to draw between thought and speech episodes. It is clear that both sorts of episodes exist, that they may be conceived of as involving a given person, and that they both exhibit both causal and intentional properties. Thus just as we have seen that speech episodes can be characterized in terms both of their intentional properties and of their determinate factual character, so thought episodes are susceptible of these two sorts of characterization. There are two consequences of this fact which it is important to notice. (1) Just what determinate factual character is exhibited by speech episodes is clear, as we have remarked above. For their physical properties provide us, in fact, with a means of talking about such episodes independently of any reference to their intentional properties. For the purposes of his theory, however, Jones need not specify the nature of the determinate factual character of thought episodes beyond a certain limited point. Thus it is provided that whatever empirical features thoughts exhibit, they exhibit the causal properties of being suitably related to both verbal and nonverbal behavior.{14} Moreover, thoughts are characterizable as certain states of persons, and as having whatever empirical properties are required for them to be differentiated according to the intentional properties which they are thought of as exhibiting. As we shall see below, there are also other causal properties which thoughts come to be conceived of as having, as a result of a development of Jones' theory.

(2) Beyond this point, however, Jones' theory does not, and does not need to specify the nature of the empirical properties of thought episodes. It is compatible with Jones' theory, Sellars points out, that thoughts be conceived of as states of an unextended substance. [Sellars (5) 187] Similarly, since this question is left open, we might come to discover that the determinate factual character of thoughts consists in their exhibiting certain neurophysiological properties. "The fact that . . . [thoughts] are not introduced as physiological entities does not preclude the possibility that at a later methodological stage they may, so to speak, 'turn out' to be such." [Sellars (5) 187] What is important is that the entities which Jones postulates are modeled on overt verbal behavior not only in respect of having certain intentional properties, but also in respect of having, as states of persons, some empirical properties which constitute their determinate factual character. While some aspects of this factual character are specified by Jones' theory, others are left open to be discovered by the results of appropriate future scientific research.

While Sellars argues that the intentionality of thoughts is a result of the fact that the model used in constructing the notion of such inner episodes is semantical episodes, he does not wish therefore to be pictured "as tracing the aboutness of thoughts to characteristics which marks and noises can have as marks and noises." [Chisholm and Sellars 525] For as we have remarked above, semantical episodes can be picked out and discussed in terms of no more than their purely physical properties, the properties which Sellars calls their determinate factual character. In order to be able to discuss the properties of semantical episodes conveniently, we shall refer to them as the sign-design characteristics of the semantical episode. Thus the sign design-characteristics of a given semantical episode will be the physical properties of the episode, including any non-intentionally specified properties which can be used in identifying and reidentifying such an episode. Since the factual character of thoughts and speech episodes is conceived by the theory as being at least in part radically different, there is of course no reason to suppose that the intentional properties of thoughts can be understood as deriving from the sign-design-characteristics of speech acts. "Marks and noises," Sellars argues, "are, in a primary sense, linguistic expressions only as 'non-parrotingly' produced by a language using animal." [Chisholm and Sellars 575, emphasis original] That is, such episodes are linguistic episodes only if they are regarded as exhibiting certain intentional properties--being about this or that, or making reference to this or that.

There is, however, an important relationship between the intentionality of thoughts and the determinate factual character of semantical episodes. For the latter are distinguishable according to their intentional properties only because they are also distinguishable on the basis of their empirical properties; and thoughts are distinguishable according to their intentional properties only because they are causally related to semantic episodes which in turn are distinguishable on the basis of their sign-design-characteristics.{15} It of course does not follow from these considerations that the fact that thoughts are about something is traceable to the fact that semantical episodes are distinguishable on the basis of sign-design-characteristics; rather it is simply the case that if the latter were not so distinguishable, then Jones' theory would be unable to supply a means for telling when either a semantic episode or a thought episode exhibiting certain specifiable intentional properties had occurred.

These considerations raise a third point which is important to discuss, namely the fact that on Sellars' account we must distinguish between those features of thoughts which are theory relative, in the sense that they are features which must be assigned to thoughts if we view them as having been introduced by Jones' theory, and those features which are not in this sense theory relative. That thoughts are causally related to verbal and nonverbal behavior in certain ways, that they are states of persons, that they are episodic in nature, that they exhibit intentionality and that they exhibit some determinate factual character (just as semantic episodes exhibit sign-design-characteristics) are features which belong to thoughts in virtue of their having been postulated in the course of Jones' theoretical reasoning. But much in the way of the properties exhibited by thoughts is left open by Jones' theorizing, and is therefore subject to future research. The point seems entirely parallel to what we may envisage in the case of more standard examples of theoretical entities--say, electrons. For while many properties of electrons, as they are postulated in the course of any given process of theoretical reasoning, belong to electrons on the basis of the theory and its success in explaining certain phenomena, others are left open for determination by further research.{16}

In particular, these matters account for the requirement, on Sellars' view, that thoughts are intentional. For thoughts are introduced by Jones as items which may be described as about other items, as making reference to things, and by means of that-clauses. But to say this is simply, on Sellars' view, to say that they are intentional. If we use 'intentional' in this way, then on Sellars' view it is "absurd to say of anything that it is a thought but lacks intentionality." [Chisholm and Sellars, emphasis original] That this is so, we have argued, is for Sellars because "the aboutness of thoughts is to be explained or understood by reference to the categories of semantical discourse about language" [Chisholm and Sellars 530] and by reference to the nature of Jones' theoretical reasoning. Thus on Sellars' view, "intentionality is a necessary feature of thoughts," [Chisholm and Sellars 536, emphasis original] necessary in the theory relative sense explicated above.{17}

It is important in understanding the rationale for Sellars' specification of which properties of thoughts are conceived of as theory relative in the sense provided above, that is, features imputed to thoughts on the basis of the sort of theoretical considerations invoked by Jones, that we understand precisely what the purpose of Jones' theoretical reasoning is. As we have remarked, pre-Joneseans are conceived of as having no psychological discourse, but as being able to distinguish between (a) cases of parroting versus non-parroting talk, and (b) cases of intelligent as against non-intelligent behavior. "Pre-Joneseans do not have the concept of thought as contrasted with talk, though, of course, they can distinguish between real talk and parroting talk." [Sellars (17) 4] Similarly, "the philosophical situation . . . [which Sellars' account] is designed to clarify is one in which we are not puzzled by how people acquire a language for referring to public properties of public objects, but are very puzzled indeed about how we learn to speak of inner episodes and immediate experiences." [Sellars (5) 178-179]{18} Thus in introducing discourse about thoughts on the basis of his (presumably sound) theoretical reasoning, Jones has, on Sellars' view, introduced just those sorts of ways of talking which form the basis of our current psychological discourse.

If this is true, however, then it must be possible, given Jones' theory as a basis, to at least sketch the way in which current psychological discourse might reasonably be construed as having resulted from the developments following Jones' theorizing. For Sellars' aim is to show that it is possible, given only certain linguistic resources, to "save the appearances" with regard to what we regard as ordinary everyday talk about psychological phenomena. It is this issue which, in effect, constitutes Sellars' second question, noted above, concerning how his answer to the first question may be seen as plausible. It is to two issues on which the plausibility of Sellars' initial account seems to turn that the following two sections are devoted.

V. It is a feature of our views on the nature of thoughts that we regard ourselves as able to report on our havings of thoughts without recourse to considerations of the sort we have seen invoked by Jones' theory. Thus, while a person who had mastered what was involved in Jones' theory construction might accurately pick out cases of thought episodes, he would be doing so in a way different, we might wish to argue, from the way in which we normally do so. For while such a person would invoke considerations pertaining to the overt behavior, verbal and nonverbal, of others and of himself, we are able to report on our own thoughts (and possibly those of others) without running through inferences of the sort warranted by Jones' theory construction, and without invoking considerations pertaining to our own overt behavior.

Pre-Joneseans, as we have seen, "think, but they don't know that they think." [Chisholm and Sellars 526, emphasis original] They come to learn that they think, and have been having thoughts, by way of learning and adopting Jones' theory. Thus their means for telling when they think must be a matter of performing inferences of a nature dictated by the theory. By contrast, we might wish to argue, we are directly aware of our thoughts, aware of them, that is, independently of the theoretical inferences warranted by Jones' theory. The problem is, then, how is it possible that given the development represented by Jones' theory, we should be able to form a notion of a thought which in these respects resembles our own. Sellars is aware of this problem when he writes, for example, that

thoughts, of course, are not theoretical entities. We have direct (non-inferential) knowledge, on occasion, of what we are thinking, just as we have direct (non-inferential) knowledge of such non-theoretical states of affairs as the bouncing of tennis balls. [Chisholm and Sellars 522, emphasis supplied.]

Sellars goes on to argue, however, that "if thoughts are not theoretical entities, it is because they are more than merely theoretical entities." [Chisholm and Sellars 522, emphasis original] Now so far as the features assignable to thoughts on the basis of the way in which Jones' postulated them in the course of his theorizing are concerned, the facts that they are noninferentially reportable and that we seem aware of them independent of theoretically warranted inferences are independent of such theory relative features. Nothing in Jones' theory, Sellars is clear, required that thoughts should exhibit these properties. For the theory provides two ways of discovering whether a given person has a certain thought, indifferently whether that person is the person making such a determination or not. The first is on the basis of certain speech acts being performed [Sellars (5) 188], and the second is on the basis of certain non-linguistic behavior which would characteristically be accompanied by such speech acts, were the person to say anything at all relevant to his nonverbal behavior. [Sellars (17) 5] So far as what is provided by the theory is concerned, however, it is entirely open that other, perhaps radically different, ways of telling what thought someone has should be developed. That we in fact have these other ways is, so far as anything which is involved in Sellars' construction of the notion of a thought is concerned, a contingent matter.

In order to deal with these facts, however, Sellars describes a further development which is involved in the stage of teaching Jones' contemporaries to use the language of the theory. Having mastered the theoretical considerations involved in invoking thoughts to explain patterns of behavior, Jones' contemporaries are envisaged as trained to make reports on the thoughts of themselves and of others on the basis of overt behavior. As a matter of contingent fact, however, it also turns out that they can be trained "to give reasonably reliable self-descriptions, using the language of the theory, without having to observe . . . [their own] overt behavior." [Sellars (5) 189] Jones' contemporaries thus come not only to be able to infer what thoughts someone is having on the basis of behavior, but also to be able to report noninferentially what they themselves are thinking. "Our ancestors begin to speak of the privileged access each of us has to his own thoughts. What began as a language with a purely theoretical use has gained a reporting role." [Sellars (5) 189, emphasis original]{19} The privileged access which we speak of having of our own thoughts is, however, a contingent matter. Given Sellars' notion of a thought, there is nothing in Jones' theoretical reasoning which requires that if we are to count an item as a thought it must be subject to such special access.

That Jones' contemporaries can be so trained, however, provides good reason for supposing that while thoughts had been introduced, by Jones, as purely theoretical entities they come to be seen as "more than merely theoretical entities." [Chisholm and Sellars 522] For while Jones had reasoned to their existence on the basis of purely theoretical considerations involved in framing certain explanations, there is now an independent ground for believing them to exist. Their existence is also testified to by our noninferential reports of them.

It is important that this development, in addition to providing a new way of reporting on the occurrence of thoughts, also has consequences for the properties which we must suppose, given this development, that thoughts exhibit. For just as we supposed, on the basis of Jones' theoretical reasoning, that thoughts exhibit certain causal properties in virtue of which speech acts are the culmination of a causal process initiated by thoughts, so here it seems that there must be certain causal properties in virtue of which our non-inferential reporting of our own thoughts is made possible. Jones, however, is not presented by Sellars as having any explanation of this fact in terms of his theory. What is required, then, is simply that whatever the nature of thoughts, and, in particular, whatever their determinate factual character should turn out to be, it should be such as to provide for the possibility of offering such an explanation.{20} But this possibility is not imputed to thoughts on the basis of any of the theoretical considerations of the sort invoked by Jones.

It is worth, then, becoming somewhat clearer on just what is involved in our supposing that such developments in the learning of the language of Jones' theory might have taken place. In the first place, what is supposed is that we have the ability to learn to report our own mental states independently of inferences warranted by the theory, not those of others. Thus "the correct contrast between other people's mental states and our own is that between theoretical entities and theoretical entities plus." [Sellars (17) 12] For even if our reporting on the mental states of others does not, in most cases, involve actually making an inference of some sort, such reports are warranted in the way in which conclusions from the relevant theoretical inferences would be. Our own self-reports are warranted, however, in other ways as well--warranted, we may say, non-inferentially. While such non-inferential reports are not, Sellars claims, precisely like reports based on ordinary empirical observations,{21} they are like such reports in that both sorts are not to be, and need not be, justified by appeal to certain inferences.{22}

If our coming to be able to respond to our own thoughts, at least on occasion, by the making of the report that we are having such a thought, then, if this report is itself causally the culmination of a process beginning with some inner episode, and if this inner episode exhibits the intentional properties of the report itself, it follows that the inner episode in question is the thought that one is having a certain thought. Such meta-thoughts, as we may call them (Sellars uses this term in his correspondence with Castañeda, (Sellars (17) 17]), would be intermediate in the causal process which initiated with the original thought and culminated in one's reporting it. (Sellars (5) 189] Thus the determinate factual character of such thoughts as we are able directly to report would have to be such as to allow us, by means of future research, to provide an explanation of their being causally related to certain other thoughts, which in fact have the intentional property of being about the original thoughts. While it is merely a matter of fact that "a person who has a thought that-p can respond to it . . . with the thought that he has the thought that-p," [Sellars (17) 6] this is a fact which must be compatible with what is provided by Jones' theory and presumed susceptible of explanation of some sort. For although it could not have been predicted, on the basis of Jones' theory, that this would be the case, it must be compatible with Jones' theory in a way which does not rule out its coming to be explained by further theorizing.

Given that it is not problematic how, prior to Jones' time, the resources of semantical and theoretical discourse were added to what we have supposed was a Rylean language, it should not be problematic how Jones manages to teach his contemporaries the language of his theory and how, on the basis of certain theoretically warrantable inferences, discourse about Jonesean theoretical entities is to be applied. What might be seen as problematic, however, is how Jones' contemporaries are then trained (the word is Sellars' [Sellars (5) 189]) to report their thoughts non-inferentially. While Sellars speaks of this training as being "in the first instance a conditioning," he goes on to assure us that the having of the appropriate meta-thought as the conditioned response to the having of a thought "is only a necessary condition of direct, non-inferential self-knowledge.{23} It is not a sufficient condition" as well. [Sellars (17) 5] The fact that a person responds to his thought that-p by saying that he has the thought that-p is not sufficient by itself to insure that his statement "expresses direct self-knowledge." [Sellars (17) 6] The problem arises, then, as to what more than a conditioned response is required for 'I have the thought that-p.', non-parrotingly uttered, to express such direct self-knowledge.

Just as, on Sellars' view, the misleading thing about characterizing thoughts as theoretical entities is not that they are not such, but that they are more than merely such, so here the misleading thing about characterizing what is involved in such direct self-knowledge as a sort of conditioned response is not, on Sellars' view, that it is not a conditioned response but that it is more than merely a conditioned response. For if the conditioned response is to be a matter of such self-knowledge, then "the conditioning is itself caught up in a conceptual framework." [Sellars (17) 6] Roughly, what is involved is that the speech act used to report be recognized as a report, in the sense that it is recognized as tokening a certain reporting sentence type, say, 'I have a thought that-p.'. [Sellars (5) 167-168]

Among the necessary conditions for such recognition would be the preliminary recognition, by both the speaker and his linguistic community, of the fact that such reports are reliable symptoms of the speaker having the thought that-p.{24} [Sellars (5) 167-168] Another such necessary condition would be the ability to distinguish cases of reporting speech acts from other sorts, and possibly, to characterize the difference in appropriately semantical terms. We may ask, however, whether any more is involved in what Sellars calls meta-thinking than these two necessary conditions and the required conditioned response. Sellars' answer to this question appears to be the following. In addition to reporting the occurrence of a thought that-p and being a reliable indicator of such a thought, the utterance of 'I have a thought that-p.' must "give overt expression to a meta-thought . . . [having the intentional properties exhibited by the sentence 'I have a thought that-p.'] exactly as the candid assertion 'it is raining' gives overt expression to a non-meta-thought . . . [with the corresponding intentional properties]." (Letter from Sellars, November 8, 1965) The reporting of a thought must, that is, involve expressing a meta-thought if it is to constitute expression of direct self-knowledge. Just as such a report must be regarded as reporting something, it must likewise be regarded as expressing something. While in the case of an observational report such as 'This is green.' what is reported is the presence of a green object and what is expressed is a thought about such an object, so in the case of 'I have a thought that-p.', what is reported is a thought that-p and what is expressed is a thought about the thought that-p.{25}

VI. The second way in which it is important to see whether Sellars' account succeeds in "saving the appearances" stems from the fact that Jones is envisaged as having introduced only one sort of intentional item, namely the episodic occurrences we have been calling thoughts. These episodes are understood as accompanying not only verbal behavior, but items of intelligently performed nonverbal behavior as well. In particular, the thought accompanying a certain bit of nonverbal behavior is understood as being that thought which exhibits the intentional properties exhibited by the speech act which would be concurrently performed and allow pre-Joneseans to explain the nonverbal behavior.

We may consider, then, a case in which a person is engaging in a certain sort of habitual nonverbal behavior and is thinking, as we might put it, of something else. According to inferences warranted by Jones' theory, our person is thinking about two things: about what he is doing, and about something else. It appears to be a feature of our discourse about thoughts, however, that if we inquired of such a person in such circumstances what he was thinking, we would expect to be met by the true response that he was thinking only about that certain something else. If, moreover, we pursued the matter, we would expect to be met by the true denial that he had been thinking about what he was doing: for his having been doing it habitually seems to rule out that he had been thinking about it, at least in some cases. We are faced with the problem that it appears that in this respect, at least, Sellars' account fails to square with the facts of our discourse about thoughts.

The point at issue may be put somewhat more strongly as follows. It is of course compatible with Jones' theory, and thus with Sellars' account, that we could come to be aware of all our thoughts, in the sense that we could come to have meta-thoughts about all our non-meta-thoughts.{26} It is also compatible with Sellars' account, however, that we are continually having many thoughts of which we are not, and will not come to be, aware, and indeed of which it might be the case that we are prevented from becoming aware, in the sense that certain (as yet undiscovered) laws of nature rule out such awareness. [Sellars (17) 25-26] In fact, of course, it is this latter alternative which is the case: given Sellars' notion of a thought episode, it is clear that we fail to be aware of most of our thoughts. It might be thought, however, that our ordinary notion of what it is to have a thought involves us in believing that if a person has a thought and also has the ability to handle discourse about thoughts, then if he is asked what he is thinking, he is able to say. If this view were accepted, this would constitute another way in which it might seem that Sellars has failed to do justice to the phenomena at hand.

It is not required, in the context of defending Sellars' position, to examine the merits of such arguments. For what is required is that the distinctions involved in our ordinary talk about psychological phenomena be preservable on the basis of constructing such talk from the resources Sellars provides us. Thus as long as we are clear that some of what Sellars calls thoughts are non-inferentially reportable and some are not, we are at liberty, if we like, to regard only a proper subset of what Sellars calls thoughts as properly speaking thoughts.

It should be remarked that on Sellars' view there is an important difference between a person's being said to think that-p and a person's being said to have the thought that-p. It is the latter case which following the notion framed by Jones is to be understood as episodic. The latter, however, are often dispositional in nature, in the sense of having a disposition to have the thought in question.{27} Thus Sellars provides that we may speak of the havings of thoughts as mental states or as mental acts, "in the sense of actualities (as contrasted with dispositions or propensities).{28} [Sellars (8) 655] Sentences such as 'I am thinking that-p,',{29} 'I think that-p.', 'I believe that-p.' and so forth have, on the other hand, "a dispositional force." [Chisholm and Sellars 522] In short, " 'having the thought that p' is not equivalent to 'thinking that p'." [Chisholm and Sellars 521] Since it is a part of the resources of the Rylean discourse which pre-Joneseans are envisaged as having that the subjunctive conditional is included in it, it seems clear that post-Joneseans can, in a fairly unproblematic way, be imagined as coming to talk about thought episodes dispositionally, and thereby come to talk about thinkings, havings of beliefs and believings.

With this distinction before us, we are provided with an additional means for dealing with the problem of constructing something close to our ordinary psychological discourse on the basis of Sellars' account. For much of our discourse about thoughts is, in fact, discourse about dispositions to have thoughts and propensities to have thoughts. If this is the case, then it is of course not surprising that we should be able to report non-inferentially on only some of our thoughts. For much of what we ascribe to one another and to ourselves in the way of psychological phenomena is dispositional, and we quite reasonably may be expected to be aware of some, but not all, of our dispositions to have this or that thought. Moreover, if our discourse about psychological phenomena is, as it seems to be, typically about dispositions to have certain thoughts and not about the occurrences of such thoughts themselves, then it seems quite reasonable to expect that we might not feel the need to talk about most of the thought episodes we actually do have, nor, in consequence, would we feel the resulting lack. In these ways it appears that Sellars' account is, indeed, able to do the job of saving the appearances regarding our ordinary mentalistic discourse.

VII. Much of what we believe to be the nature of psychological items such as thoughts has been accounted for, on Sellars' view, by an appeal to the way in which the inner episodes of Jones' theory are causally related to other phenomena. This appeal rests, of course, on the idea that thoughts exhibit some determinate factual character. For without such empirical properties, it does not seem possible that such inner episodes could exhibit the requisite causal properties.{30}

By contrast, comparatively little has been said in the present chapter about the intentional nature of thoughts, beyond the idea that their intentional nature is to be thought of, on Sellars' view, as consisting in the fact that thoughts, like overt verbal episodes, are about things and have reference to things. In particular, beyond this fairly general characterization, we are not yet in a position to characterize thoughts in virtue of their intentional properties. We may say that a given thought is a thought that-p, or that it is about this or that, but we have not examined just what, on Sellars' view, is involved in ascribing such intentional properties to thoughts. Whatever such ascriptions amount to, however, we do know that on Sellars' view it will be much like characterizing semantic episodes in the corresponding terms. Before turning, then, to an examination in Chapter V of the differences in Sellars' and Chisholm's views, it will be important to discuss, in this last preliminary section, these matters.

The approach which Sellars adapts to this problem resembles his approach to specifying what is involved in psychological discourse in at least the following way. For just as Sellars' account of how we might construct psychological discourse rests on considering what is involved in one particular sort of standard locution--'have the thought that such-and-such'--so in the case of semantical discourse Sellars directs his consideration to one sort of standard semantical locution, namely what he calls the 'means'-rubric:

(1) '. . .' means ---.

(Cf. [Sellars (5) 161-164], and especially [Sellars (5) 163].) It is important, for Sellars' purposes, to distinguish such locutions from another sort of semantical locution, namely:

(2) '. . .' means the same as what '---' means.

For whereas sentences of the form of (1), correctly understood, tell us what the expression '. . .' means, sentences like (2) tell us simply that it means the same as whatever '---' means. The force of this point may be brought out as follows. We may imagine sentences (1) and (2) translated into some other language, say German, as follows:

(1') '. . .' bedeutet ***.

(2') '. . .' bedeutet derselbe wie '---'.

where '***' is taken as the German equivalent of '---'. Since the translation into German of (1) and (2) will affect only those portions of the sentences not enclosed within single quotes, the translation (2') will fail to convey to a German speaker with no command of English anything beyond the fact that two English expressions mean the same. (1'), however, will also convey just what is meant by '. . .', for the right-hand expression in (1) and (1') does not, as it does in (2) and (2'), fall within the scope of single quotes. Because (1) is therefore richer in what it conveys, Sellars argues that sentences like (2) may be constructed given that we have an understanding of sentences like (1), but not vice versa. It is for this reason that Sellars concentrates his attention on the latter.

It seems clear that (2) expresses a relation which is asserted as holding between two things, namely a relation of meaning-equivalence holding between two linguistic expressions. It also seems clear that (1) and (2) entail one another. The obvious question to raise, then, is whether (1), like (2) also expresses a relation. It is an important part of Sellars' views concerning semantical discourse that the means rubric does not, like sentence (2), express such a relation. Cf. [Sellars (5) 163] and [Sellars (8) 658-659].) The following argument may be advanced as capturing the motivation of Sellars' conviction on this point. Let us suppose that (1) did express a relation. We might then wish to rephrase (1) so as to read as follows:

(3) What '. . .' means is identical to ---.

The question arises, then, just what is this item, ---, which '. . .' bears the relation of meaning to. If we suppose that it is a linguistic expression, then we are back with a sentence which has the force of (2), and the consequent limitations of such sentences. But there does not seem to be any plausible alternative. We cannot read such sentences as involving reference to ordinary physical things, as for example understanding 'horse' in

What 'Pferd' (in German) means is identical to horse.

to refer to some horse; nor does it seem plausible to construe 'horse' in such a case as referring to the concept of a horse. For it is not clear what sense we are to give to talk about concepts independently of having first become clear about what is involved in the semantical locutions in question. The only acceptable alternative seems to be to deny, as Sellars does, that (1) expresses a relation at all. For if this is denied, then the paraphrase of (1) into (3) is no longer warranted, and we are enabled to avoid the objectionable consequences of such a move.

If we are prevented from adopting a relational construal of our 'means'-rubric, it is clear that we must search for another. The suggestion of what would be involved in an acceptable construal is provided by Sellars when he writes that

The rubric ' ". . ." means ---' is a linguistic device for conveying the information that a mentioned word . . . plays the same role in a certain linguistic economy . . . as does . . . [a certain other word] which is not mentioned but used--used in a unique way; exhibited, so to speak--and which occurs 'on the right-hand side' of the semantical statement. [Sellars (5) 163, emphasis original]

Thus the 'means'-rubric is to be thought of as conveying at least the information which the corresponding sentence of the form of (2) would convey, but conveying it in such a way as to provide that the expression occurring on the right hand side of the rubric (a) is part of the language which speakers of the language in question have at their command, and (b) is exhibited as such a part.

A notational device which Sellars adopts in later articles captures what is being sought for perhaps more clearly than the above account. The notational device consists in the introduction of a special sort of quotation mark--what Sellars calls dot-quotes--which function in the following way. An expression flanked by dot-quotes "is a common noun which applies to items in any language that play the role played in our language by the sign design that occurs between the dot quotes." [Sellars (8) 658] Thus if we assume that 'red' in English, 'rot' in German and 'rouge' in French all play the same roles in the behavioral economy of the speakers of these respective languages, then the name of any item playing such a role in a language would be .red. (in English), .rot. (in German) and .rouge. (in French). Each of these expressions would be general terms true of all and only such intentional items.

This notation innovation allows Sellars to provide the following sort of paraphrase for the 'means'-rubric:

(4) '. . .' belongs to the class of .---.s.

On this construal, the 'means'-rubric is an assertion of class membership, and is, of course, expressible in quantificational terms as

(5) (x)(x epsilon A only if x epsilon B)

This construal therefore allows one to avoid the undesirable consequences of viewing sentences like (1) as relational. [Sellars (8) 658-659]{31} On the other hand, this construal also appears to satisfy the conditions intuitively set down in "Empiricism and the Philosophy of Mind." For the expression on the right-hand side of the paraphrase is translatable into whatever language (4) itself is translatable into, and the expression is therefore exhibited, in sentences like (4), as an expression which is part of the linguistic economy of the speakers of the language at hand.

To specify that a certain linguistic expression is an instance of the expression type '---' is to specify it, at least in part, as a certain sign design in terms of some determinate factual character, that is, in terms of its sign design-characteristics. To specify a linguistic expression as an instance of a .---., however, is to specify it in terms of its intentional properties alone. For such a dot quotation names any linguistic expression which exhibits the intentional properties exhibited by the expression contained within the dot quotes in our language. Thus our notational innovation may be viewed as a way of specifying a class of expressions, in any language whatever, according to their intentional properties.

It will be recalled that in characterizing the sameness of intentional properties exhibited by both expressions flanking the 'means' of the 'means'-rubric, Sellars spoke of their playing the same role in the linguistic economy of certain speakers. Thus we may speak of certain intentional roles as the roles specified in specifying the intentional properties of an intentional item,{32} and think of our dot-quote notation as a device for classifying expressions in any language on the basis of the intentional role which they play.

Since thought episodes, like semantic episodes, are conceived of as playing intentional roles ton Sellars' account, in virtue of their causal relation to corresponding semantic episodes),we may imagine a parallel way of talking about thought episodes. Thus, just as we use single quotes to form the names of expression types in a given language, so we may adopt French quotation marks to form names of corresponding thought episodes. We may, that is, "use '<<Es regnet>>' to stand for the kind of mental act that occurs in the minds of the German-speaking people and finds its overt expression in candid utterances of 'Es regnet'." [Sellars (8) 662] General terms so formed would be true of all and only such thought episodes as are had by speakers of a given language, and which play the same intentional role as the sign design contained in French quotes plays in their language. Thus so to specify a thought episode is to specify it, as in the case of applying single quote names to linguistic expressions, by reference to both the intentional role and the determinate factual character of the thoughts. For to specify that the thought episode is causally related to certain linguistic role-players in a certain language is to specify an aspect of its (causal) empirical properties.

A convention applicable to thought episodes but parallel to dot quotes rather than to single quotes is also possible. Thus we may adopt diamond quotes according to the following convention. An expression flanked by diamond quotes (such as '<>---<>') is the name of any thought episode which plays the intentional role played by the contained expression in our language. Thus any thought episode which is a <>red<>, or a <<red>>, a <<rot>> or a <<rouge>> is thereby a <>red<>, and just as '.rot.' is simply the German translation of '.red.', so '<>rot<>' is simply the German translation of '<>red<>'. (The diamond quote convention is suggested by Sellars in a letter of January 4, 1966.) Thus we have a means for specifying a thought episode in terms of the intentional role it plays, independently of its determinate factual character.{33}

What is accomplished by such conventions is that we are provided with a means of discussing thought episodes in a way suggested by Sellars' account of psychological discourse. Thus in saying, for example, that a certain thought episode is a <>not<>, we are saying that it is a thought episode which plays an intentional role analogous to that played by 'not' in English, and by .not.s in other languages. Thus just as we paraphrased (1) by means of (4), so we might paraphrase

(6) George has the thought that it is raining.


(7) The thought episode which George has is a .<>It is raining.<>.

In asserting (7), we are asserting that George has a certain inner episode which is causally related in the way specified by Jones' theory to whatever semantic episode plays the role in George's language played by 'It is raining.' in ours, and which is conceived of as itself playing an analogous intentional role by virtue of that causal relation. The fact that it is expressed by some .It is raining.. is simply the fact that it plays the analogous intentional role in question. By contrast, to say that

(8) The thought episode which George has is a <<It is raining.>>.

would be to say not only what (7) asserts, but also that the thought episode is the sort which in addition to satisfying the conditions specified above, also satisfies the condition that it is an episode which exhibits a suitable relation both to the English 'It is raining.', which expresses it, and to other thought episodes expressed by English verbal episodes.

With this apparatus for talking about both thought episodes and semantic episodes in a way suggested by Sellars' account of psychological and semantical discourse at our disposal, we are in a position to turn to a consideration of the respects in which Sellars' account differs from that defended by Chisholm.

[[Table of Contents][Go to Chapter 4]


{1} Sellars does not claim that the two sorts of phenomena exhibit the same properties, but rather that they exhibit similar, or analogous, properties. Cf. on page 120, below. Vide also [Sellars (13) 105-106] and [Sellars (13) 120-121] in this last location, Sellars seems to claim that it is open to us to talk not of similar properties exhibited by the two sorts of phenomena, but of properties which are exhibited by both sorts. [Back]

{2} Sellars' resistance to the idea that the intentional properties of a thing are a part of its determinate factual character seems to be based on the idea that to specify an item solely in terms of its intentional properties is not to describe it, in the sense in which to specify a part of its determinate factual character (such as the pitch or constituent phonemes of a speech episode ) is to describe it. Thus Sellars writes that the locution " ' " . . . " means ---' is the core of a unique mode of discourse which is as distinct from the description and explanation of empirical fact, as the language of prescription and justification" [Chisholm and Sellars 527, emphasis original] As Chisholm urges in response to this point, unless the term 'description' can be given a rigorously determined application which warrants this distinction, it is not clear what is to be gained by drawing the contrast in question. Thus Chisholm argues that to deny that 'hund' means dog in German would be to make a mistake in just the way that to assert that Berlin is part of Warsaw would be. [Chisholm and Sellars 529] The force of this claim seems to be the idea that in both cases, it is reasonable to suppose that the mistake is a factual one, and therefore a mistake in describing the world.

It is important to note that while the way in which the distinction has been drawn above is intended to avoid the difficulty which Sellars faces in invoking what seems to be a restricted and technical application of the term 'description', that in an important respect, the distinction as drawn seems to reflect what Sellars has in mind. Thus whether a property is to count as part of the determinate factual character or as an intentional property of a thing depends, both for Sellars and as we have drawn the distinction, on the sort of locution which is being used to talk about the thing. The difference which it seems crucial to note is that Sellars' characterization of the distinction seems to suggest that just as to ascribe ethical properties of a thing cannot be reduced to ascription of empirical properties, so with ascription of intentional properties to a thing. Questions of this sort will be raised in the course of the final chapter, and for this reason, it is important to avoid predetermining the answers to such questions by no more than a terminological stipulation. [Back]

{3} In his book Mental Acts, Geach presents an analogy similar in certain respects to that defended by Sellars, basing it primarily on such considerations. [Geach 75-96, 101-107] Geach, however, adopts certain non-standard views concerning direct discourse constructions and the use of quotation marks in such constructions, and frames his analogy on the basis of this sort of locution, rather than on the indirect discourse construction. Sellars acknowledges the similarity between his account of the nature of psychological discourse and that presented by Geach in [Geach], pointing out that both he and Geach construe discourse about thought episodes as an analogical extension of discourse about candid, non-parrotingly produced overt verbal episodes (Vide [Sellars (6) 150] [Back]

{4} It is worth noting that for our purposes the having of thoughts is sufficient for the problem at hand; for whether, for example, feelings and sensations are psychological phenomena in the sense in which Chisholm and Sellars are concerned with such phenomena is an open question in the following ways. (a) Sellars, while characterizing thoughts, beliefs and so forth as intentional, denies that feelings and sensations exhibit intentionality. [Sellars (3) passim] This denial will not need to concern us, however, in the discussion that follows. (b) That Chisholm did not have sensations and feelings in mind in framing his thesis of intentionality is suggested by both the fact that sentences about such phenomena, if psychological, would be clear counter-examples to his claim (Cf. [Heidelberger 530], where the example 'I itch ' is used), and the fact that whereas Chisholm in certain cases hopes to bring all psychological phenomena under a given mark of intentionality by invoking the idea that such phenomena all involve such element of believing, this does not seem to be the case with sensations and feelings. [Back]

{5} In the following discussion, the expression 'semantical episode' will be used to apply to episodes which may appropriately be characterized in semantical, or metalinguistic, terms. Thus typical cases of semantical episodes will be non-parrotingly produced utterings which properly token well-formed expression-types in a spoken language. The term 'speech act' will also occasionally be used in the same way. It is important to notice, however, that applying 'speech act' to an episode is not intended to convey that the episode is an action, in the sense of being something which can be "done" on purpose. Thus Sellars points out that "not every linguistic episode is an act in the sense of performance or piece of conduct." [Sellars (13) 109, 110] Similarly, the use of the term 'mental act' to apply to psychological phenomena is intended to specify that the phenomenon is episodic, and not that it is necessarily an action in the sense specified above. (Cf. [Sellars (13) 110] and [Sellars (8) 655].) [Back]

{6} Thus the compatibility of this sort of explanation with other sorts is not here at issue. Cf. [Taylor (2)], [Malcolm (1)], and [Hampshire 11-33, 53-102]. What is at issue is simply that the sort of explanation of nonverbal behavior which is envisaged here is a plausible alternative among others. [Back]

{7} Cf. [Sellars (8) 657], where Sellars discusses overt speech episodes as "speech understood in terms of the uniformities and propensities that connect utterances (a) with other utterances (at the same or at a different level of language), (b) with the perceptible environment, and (c) with courses of action (including linguistic behavior) " It is these uniformities and propensities which allow us to frame the sorts of explanations which are in question here. [Back]

{8} For a detailed account of how Sellars believes theory construction involves reference to a model, Vide [Sellars (11) 178-184 et passim]. [Back]

{9} It is of course not the intention of the present discussion to evaluate the adequacy of Sellars' views concerning theory construction, but rather to present these views with a view to examining the role they play in Sellars' views concerning the nature of psychological discourse. [Back]

{10} In his correspondence with Castañeda, Sellars suggests the following way of viewing the matter. We may suppose that pre-Jonesean people had, at a certain stage in their development, talked constantly whenever they were engaging in any sort of nonverbal behavior, but that this tendency to talk became inhibited at a later stage. While constant speaking was normal, all nonverbal behavior could he explained by reference to speech episodes; when such verbal behavior became inhibited, the need then may be envisaged as becoming felt that some other explanation of such nonverbal behavior be provided. [Sellars (11) 4)]

A different way of viewing the matter, which does not involve suppositions about stages in the development of our mythical pre-Jonesean conceptual ancestors, may be put as follows. We may suppose that (a) pre-Joneseans can distinguish between items of intelligent nonverbal behavior and items of nonverbal behavior which are not; (b) pre-Joneseans are straightforwardly able to explain instances of intelligent nonverbal behavior when these instances are accompanied by non-parrotingly produced verbal behavior; and (c) pre-Joneseans recognize that there are (presumably frequent) instances of intelligent nonverbal behavior which are unaccompanied by items of verbal behavior. Given that pre-Joneseans have, as Sellars specifies that they do, recourse to talk about dispositions and to the use of the subjunctive conditional, it seems reasonable for us to imagine them supposing that items of intelligent nonverbal behavior which are unaccompanied by items of verbal behavior may partially be explained by the idea that such instances of intelligent nonverbal behavior would be accompanied by appropriate non-parrotingly produced items of verbal behavior if certain circumstances were to obtain. (Vide [Sellars (6) 153], and especially [Sellars (6) 153, fn. 8].) [Back]

{11} If we are to imagine Jones' theory as attributing to thoughts the sort of intentional properties which are exhibited by semantical episodes, then it is important to raise the question whether the intentional properties exhibited by, say, the thought that-p, are the identical properties which are exhibited by the semantical uttering that-p. For it might be argued that one way in which Sellars' account breaks down is that on Jones' theory these are identical intentional properties; but, it might be argued, the way in which the thought that-p exhibits intentionality is different from the way in which the semantical episode that-p exhibits its intentional properties. Sellars does not, however, specify that these properties are identical, but rather, to use his word, that they are analogous. (Cf. [Sellars (8) 657] and [Sellars (3) 44 et passim].) Thus we might imagine a more accurate account of the matter involving the claim that the properties in question, although not identical, are similar (in the sense that, say, different shades of purple are similar properties, albeit, not identical). This issue seems to be squarely faced, if not entirely satisfactorily dealt with, in connection with certain features of Sellars' position regarding the use of models in theory construction. For while it is clear that theory construction, thought of as involving reference to entities of a model, must include a commentary [Sellars (5) 182, 187] which restricts the sorts of properties which the particulars of the model and those of the theory have in common, there is another qualification which is important here. For Sellars rejects the view that aside from the restrictions of the commentary, the two sorts of entities have identical sorts of properties, arguing instead that their properties, although not identical, are nonetheless similar. (Cf. Sellars' rejection of Miss Hesse's notion of a model, in [Sellars (11) 180-184].) Thus the particulars of the theory are intended to have attributes similar to those of the particulars of the model, but not identical.

The idea that two particulars might have attributes which are similar but not identical is, on Sellars' view, to be explicated in terms of the idea that the properties exhibited by such particulars have, themselves, certain second order properties in common, but, presumably not all, for the first-order properties are to be thought of as nonidentical. [Sellars (11) 180-184] The unsatisfactory nature of this sort of move seems to lie in specifying what is involved in supposing properties to exhibit second-order properties. Sellars' example concerns the idea of using points of a line as a model for moments in time, which he elaborates by advancing the idea that both the properties of points and moments (namely, their relational properties} share the identical properties of transitivity, and, perhaps, other mathematically specifiable second-order properties as well. [Sellars (11) 180-184] As applied to the case at hand, the situation might look something like the following. Both thoughts and semantical episodes exhibit certain properties, and these properties themselves exhibit certain second-order properties, such as playing a role in the behavior patterns of persons, and allowing certain sorts of explanations to be framed. Thus the thought that-p and the semantical episode that-p exhibit certain properties which may be thought of as analogous, or similar, in the sense that these properties exhibit certain second-order properties. Thus when we come, below, to speak of thoughts and semantical episodes as playing the same (or similar) intentional roles, this is to be understood in the sense that both sorts of items have attributes which themselves exhibit the second-order attributes of occupying a certain position in the behavioral patterns of persons and the explanations of intelligent behavior which we frame about persons. In particular, it is in just this sense which we may reasonably speak of the two sorts of items as playing analogous roles, and of the structural properties which they share; for these structural properties are, on Sellars' view, simply the role played by the various sorts of items in virtue of their second-order properties. For convenience in discussing Sellars' views, however, these refinements will be skirted in the discussion which follows [Back]

{12} On the account examined above, semantical items are envisaged as employable by pre-Joneseans in framing explanations of accompanying nonverbal behavior, and by analogy thought episodes are envisaged as introduced as items which can be used in explaining not only nonverbal behavior in the way in which verbal episodes might be so used, but also in explaining nonverbal behavior unaccompanied by speech behavior and the occurrence of speech episodes themselves. These explanations are to be thought of as explanations of behavior thought of as intelligent, in the informal sense that we can come to understand what sort of nonverbal behavior is occurring by reference to accompanying verbal episodes (including both self-descriptions and other verbal episodes)

The psychological items which we have envisaged Jones as postulating are such as to have their existence established, as we have seen, on the basis of theoretical reasoning. There is, however, an ambiguity in the term 'psychological item' which is important to notice in this connection. A psychological item might be the sort of item which (a) plays a certain role in explaining behavior, and (b) is postulated by Jones on the basis of theoretical reasoning. It is in this sense that the term has been used, and will be used in what follows. By contrast, a psychological item might be seen as any item which can be used in framing the sorts of explanations of intelligent behavior invoked in Jones' theoretical reasonings. On this second reading, pre-Joneseans would of course have the ability to speak about psychological items, for on this reading semantical items are just one special sort of psychological item. In order to avoid ambiguity in what follows, and, in particular, to undercut the argument which would say that pre-Jonesean discourse about semantical items essentially involves reference to psychological items on the basis of relying on this second reading of 'psychological item', it is important to note that it is not this second sort of reading which is to be understood in what follows. This ambiguity was pointed out to me by Sellars in a letter dated September 3, 1965. [Back]

{13} It is important to notice that Sellars' argument, as sketched in section III, is in support of the claim that adding theoretical and semantical discourse is sufficient to generate psychological discourse, not that these are the only linguistic resources which would do the job. The fact that these two sorts of linguistic resources are sufficient to do the job, however, enables us to view psychological discourse, as it is present in our language, as in principle constructible from these linguistic resources, and it is Sellars' view that regarding current psychological discourse as in principle constructible from such resources will provide us with an understanding of the nature of current psychological discourse. [Back]

{14} That they are causally related to nonverbal behavior is presumably involved in the idea that we should be able to provide adequate explanations of nonverbal behavior by reference to thoughts. Thus whereas pre-Joneseans might be conceived of as having thought that their explanation of nonverbal behavior by reference to verbal episodes presupposed that such episodes caused nonverbal behavior, Jones' theory may be thought of as providing an alternative explanation of nonverbal behavior and therefore an alternative aetiology. [Back]

{15} That this is so is a remark about how we are able to distinguish among thought episodes on the basis of as much information as is available given Jones' theory. The possibility of coming to be able to distinguish among them in other ways will be examined below, in connection with the fact that we appear to be able to report noninferentially on the occurrence of these inner episodes. [Back]

{16} A point which is worth considering in this connection is one in effect raised by Putnam. For while we may envisage discourse about certain theoretical entities being introduced in some way which determines the properties of such entities in certain respects, so that we should imagine ourselves unprepared to admit that any items which did not exhibit those properties were the entities which had been so postulated, we might also, following Putnam, want to argue that further scientific research might lead us to say, in effect, that the theory was mistaken in the assignment of certain properties to the entities even though the entities under consideration were exactly those originally postulated by the original theory. (Cf. [Putnam (6) 378] and [Putnam (1) 5], and especially [Putnam (1) 5, fn. 1].) To argue this position effectively seems to involve becoming clear on just what entities, whose existence is established by theoretical reasoning, count as the same. In order to avoid this problem, Putnam's position on this matter will not be considered, although a similar point, made in much the same spirit, will be defended in Chapter V under the heading of what is there called the revised thesis of physicalism. [Back]

{17} It is important to contrast this use of 'intentional' with that invoked by Chisholm. Whereas Chisholm uses 'intentional' as a general term which is true of all and only expressions which exhibit a certain logical characteristic (that characteristic which will be adequate to show the truth of a given thesis of intentionality), Sellars' use is widely divergent. According to Sellars, we may speak of an item as intentional if, and only if, it is the sort of thing which is about something, or makes reference to something. By extension, we may then speak of such properties exhibited by intentional items as intentional properties. Thus we might relate these two uses as follows. Sentences are, in Chisholm's sense, psychological if, and only if, they are about intentional phenomena, using 'intentional' now in Sellars' sense. If Chisholm's thesis as presented in "Sentences about Believing" is true, it would follow that sentences are intentional (in Chisholm's sense) if, and only if they are about intentional phenomena (in Sellars' sense). [Back]

{18} By immediate experiences, Sellars has in mind the feelings and sensations which we shall not be concerned to discuss. The issues involved in such immediate experiences will be touched on in the first footnote of Chapter V, section III. [Back]

{19} Cf. [Sellars (13) 108], where Sellars speaks of the requirement that we be able to give an account of "the acquisition of the avowal role" of discourse about mental episodes. Sellars argues here that a parallel requirement must be imposed concerning discourse about overt behavior: one must be able to account for "how language pertaining to behavioral dispositions and propensities can acquire the use by which one's possession of such dispositions and propensities is avowed." This requirement would seem to be of particular importance in giving an account of our ability to give self-descriptions of our own verbal behavior without making inferences from self-observations. Cf. also [Sellars (6) 158] where substantially the same point is made. [Back]

{20} If the determinate factual character of thoughts were adequately representable in terms of neurophysiological properties which they exhibit, which as we have seen is left a viable possibility by the theory, then this fact would suggest that our explanation could be made in terms of the establishment of certain neural connections. For since the fact that we can report non-inferentially on our own thoughts would appear to be a matter of our responding verbally to our thoughts, we should wish to search for neural connections which made this direct response possible. [Back]

{21} "We have direct non-inferential knowledge of our mental states, we do not observe them. We have them, and we know that we have them." [Sellars (17) 12, emphasis original] Sellars is not explicit concerning the differences which he believes there are between the non-inferential reporting of non-theoretical states of affairs, which we ordinarily call observing, and the non-inferential reporting of thought episodes. (Cf. [Chisholm and Sellars 522] where his remarks seem to lead us to the conclusion that there are no important differences: "We have direct (non-inferential) knowledge, on occasion, of what we are thinking, just as we have direct (non-inferential} knowledge of such non-theoretical states of affairs as the bouncing of tennis balls." (Emphasis supplied).) The differences Sellars appears to have in mind when he makes the remark quoted above from his correspondence with Castañeda might be understood in the light of a suggestion by Rorty. Whereas in the case of ordinary observable state of affairs we seem to be clear on how to draw a distinction between a person's making a mistaken report on the basis of misjudging what is going on, on the one hand, and on the basis of misusing the language used in making the report, on the other, Rorty suggests that no such clear distinction can be drawn in the case of reporting non-inferentially on our own sensations. [Rorty (2) 52-53] If we apply Rorty's suggestion to non-inferential reports of our thoughts, as it seems we are warranted in doing, then there does appear to be a significant difference between noninferential reports of our own thoughts, and non-inferential reports of ordinarily observable facts. [Back]

{22} Cf. a similar characterization of this distinction by Rorty: "By 'non-inferential report' I mean a statement in response to which questions like 'How did you know?' 'On what evidence do you say . . . ?' and 'What leads you to think . . . ?' are normally considered misplaced and unanswerable, but which is nonetheless capable of empirical confirmation." [Rorty (2) 43] [Back]

{23} Since conditioning is required, however, in order that each of us come to have the ability non-inferentially to report our own thoughts, it seems to follow that each of us must initially learn to apply discourse about thoughts in an inferential way. For only if we are able to apply discourse about thoughts inferentially, could we come to be conditioned to apply such discourse independently of any such inferences. [Back]

{24} We may compare reports of one's own thoughts to reporting observation statements such as utterances of 'This is green.', in the following way: both reports are ordinarily taken as non-inferential in the sense discussed above. Thus Sellars writes: "no tokening by [a person] S now of 'This is green' is to count as 'expressing observational knowledge' unless it is also correct to say of S that he now knows the appropriate fact of the form X is a reliable symptom of Y, namely that ( . . . [to] oversimplify) utterances of 'This is green' are reliable indicators of the presence of green objects in standard conditions of perception." [Sellars (5) 169, emphasis original] Cf. also [Sellars (6) 158] where substantially the same point is made. [Back]

{25} Thus pre-Joneseans, lacking the concept of a thought as it is later introduced by Jones, are substantially restricted in the scope of their thinking. [Sellars (17) 17] For among the sorts of thoughts that they do not have are meta-thoughts. If one required that in order to count someone as having thought she must at some time have thoughts about his thoughts, then one might suppose that in a certain sense pre-Joneseans did not have thoughts. This may be what Sellars has in mind by writing that "pre-Joneseans do not think 'in the full sense' " [Sellars (17) 17] [Back]

{26} Becoming aware of all our thoughts in the sense of having meta-thoughts about all our thoughts would involve, of course, having meta-meta-thoughts about our meta-thoughts, and so on. Thus the infinite regress would seem to preclude becoming aware of all our thoughts in this sense. Compare [Ryle 197-197]. [Back]

{27} Cf. [Chisholm and Sellars 521]. It is important that these sentences are to be construed, on Sellars' view, as describing dispositions to have certain thought episodes, and not, as on views such as that of Ryle, as describing dispositions to engage in certain sorts of behavior. (Vide [Sellars (5) 180].) Cf. also [Sellars (6) 154], where he writes that "very roughly, to believe that-p is to be disposed to have thoughts that-p, rather than thoughts whether-p, let alone thoughts that-not-p." [Back]

{28} It is worth pointing out that Sellars also wishes to deny that havings of thoughts are mental acts in the sense of being some sort of action--that is, in the sense of states which we could decide to bring about, or bring about voluntarily. Cf. [Sellars (6) 157]. That this is so follows, on Sellars' view, from the fact that the candid, non-parrotingly produced, overt verbal episodes on which we are to think of thought episodes as being modeled are themselves not actions in the sense of items of behavior which are brought about as the result of a decision. Cf. [Sellars (6) 151-152] and [Sellars (13) 110-111]. That this is so, however, does not rule out that we should have some control over our propensities to have certain thoughts rather than others, and Sellars claims that as a matter of empirical fact, we do have such control [Sellars (6) 157, fn. 10], and that our ability to exercise such control has as a necessary condition our ability to respond non-inferentially to a certain range of our thoughts with the appropriate meta-thoughts. [Sellars (6) 154] [Back]

{29} In response to this point, Sellars has written the following: "It is 'I think . . .' rather than 'I am thinking . . .' which I count as a dispositional cousin of 'I believe . . .'. 'I am thinking . . .' is a variegated locution which implies a sequence of episodes (a process?), and can also be used to refer to processes-cum-dispositions ('I am thinking much, these days, about Vietnam')." (Letter dated September 3, 1965.) The point at issue is that we do have, granted discourse about thought episodes in our enriched Rylean language, the resources required to construct the great variety of sorts of psychological locutions found, say, in natural languages such as English. [Back]

{30} It will become important, in the course of the next chapter, to become somewhat clearer on what is involved in the supposition that the inner episodes postulated by Jones' theory must exhibit some determinate factual character. Thus as we have seen, Sellars points out that it is not incompatible with anything in Jones' theory that thoughts should be states of some unextended substance--say, of a Cartesian soul--and thus exhibit no empirical properties of the sort exhibited by physical objects. (Vide [Sellars (5) 187].) What is required in order that thoughts exhibit the intentional properties which allow their use in explanations of human (verbal and nonverbal) behavior in a way analogous to the use of semantical episodes in such explanations is that they exhibit certain causal properties. In particular, it is required that they be causally related in a suitable way to episodes which make up, and dispositions and propensities which lie behind, overt behavior which we classify as intelligent. Since these causal properties are empirical, they are (at least) part of the determinate factual character of thoughts. What is involved in the imputation of further empirical properties to thoughts will be examined in the following chapter. [Back]

{31} Cf. also [Sellars (6) 151], where Sellars writes that the function of statements about what linguistic expressions mean is to "classify rather than relate." (Emphasis original.) [Back]

{32} This notion is developed at some length in "Some Reflection on Language Games," but for the present purposes we may accept talk of intentional roles as simply a more perspicuous way of talking about intentional properties. It is of course important to specify just what is involved in the specification of the intentional role played by a certain sort of item, as Sellars attempts to do, in the case of semantical entities, in the article cited above. Thus it is important, for our purposes in Chapter IV, that the specification of the intentional role of a semantical episode does not necessarily involve reference to thought episodes. It is not clear why one should believe that such reference would be required in so specifying the intentional role of a semantical item; for on the line adopted by Sellars in the article cited above, this role can be captured by speaking in terms of no more than (a) the role which such items play in explaining behavior, and (b) the role such items play in the overall behavior economy of language-using persons. [Back]

{33} As Sellars points out, given that we accept the idea that thoughts do exhibit some descriptive characteristics, there is, for example, "every reason to suppose that Japanese inner speech differs systematically from English inner speech in a way which reflects the differences between these two languages." [Sellars (8) 669] [Back]

[Table of Contents][Go to Chapter 4]