V. The Identity Thesis (2): The "Categorial" Objection

There are certain aspects of this objection to any form of the identity thesis which we have considered which may be dealt with by pinning down the precise form in which Sellars wishes to advance the identity thesis. For example, one might argue that the following two sorts of properties are properties which are exhibited by physical phenomena but not by psychological phenomena, and, in particular not by thoughts. (a) While physical phenomena are, within the limits of experimental error, precisely locatable as regard to place, thoughts are not. (b) While physical phenomena are, within the limits of experimental error, precisely locatable as regard to time and precisely specifiable with regard to duration, psychological phenomena, and, in particular, thoughts, are not.

The first claim has, in fact, been made as an objection to the identity thesis by both Coburn and Shaffer. [Coburn 91-92], [Shaffer (1) 813-816] In this connection, Shaffer argues that we could, of course, adopt a convention specifying how to apply location with regard to place to psychological phenomena, and suggests that if this were done, it is not clear that there would be any objection outstanding to the identity thesis, at least in certain forms.{26} The second of these claims has not been advanced as a specific objection to the identity thesis, but is worth considering on its own right. The point in question is considered by Geach [Geach 104-106], although not in connection with a discussion of the identity thesis.

Sellars' construction of a notion of a thought by means of his Jonesean theory involves the claim that thoughts are to be understood as certain states of persons. He is clear that the Jonesean theory does not require that we make a choice between, say, some form of materialism and some form of Cartesian dualism. [Sellars (5) 187] Thus it will depend, so Sellars contends, on what view one might adopt concerning the nature of persons whether it is open to suppose that thoughts are states of a certain non- material, non-spatial part of a person. Similarly, the Jonesean theory allows for a view of persons as simply material substances, in which case thoughts would have to be physical states of physical things, namely persons. It is this second possibility which, we have seen, requires Sellars to accept a certain form of the identity thesis, construed as a thesis not that certain identifications are in fact warranted, but rather that their warrantability is not ruled out on logical or philosophical grounds.

Now if thoughts are certain states of persons, and if persons are certain sorts of physical substances, then thoughts will be certain states of physical substances. Thus, to the extent to which we are able to locate persons as regard to place, we should be able thereby to locate states of theirs as regard to place. In particular, if the physical phenomena with which scientists might, according to the identity theorist, come to identify thoughts, are also states of persons, for example, neurophysiological states, then we should be able to locate the physical phenomena in question just as well as, but possibly no better than, we are able to locate thoughts. Thus we should be able to locate neurophysiological states on the basis of the location of the neurophysiological fibers and tissues which exhibits the states in question. Similarly, if we take thoughts to be states of persons, then it seems appropriate to locate such states not merely on the basis of the location of the person who is having the thought, but on the basis of the location of those parts of the person which, it might be established, are required if the person is to be able to have the thought in question.{27} Thus we might come to discover that certain parts of the central nervous system satisfy the condition that without them, no person could, as a matter of empirically discovered fact, have thoughts, and that no other parts of the biological human being satisfy this condition. On this basis, we reasonably locate thoughts as states of just these parts of the CNS of a person. Thus, we may think of the psychological phenomenon to be identified as the having of a certain thought and the physical phenomenon in question as a person's being in a certain physical state. So construed, the identification does not seem, in point of localizability in regard to place, to be a case of cross-categorial identification.{28}

The second of the claims mentioned above rests not upon the idea that certain predicates are applicable to physical phenomena but not to psychological ones, but rather upon the idea that the precision with which we can, in principle, apply certain predicates to physical phenomena is not matched by a corresponding precision in the case of applying these predicates to psychological phenomena, or, at least, to psychological phenomena in general. Thus, while we can delimit in an arbitrarily precise way the duration of a certain physical state of a thing, we do not seem able to delimit in anything like a precise way the duration of thoughts. And while we can, in some cases, specify (at least roughly) when thoughts occur, we do not seem able to do so with most thoughts, nor to do so with any thoughts in nearly the precise way we can with physical occurrences.

It is useful in this connection to consider again how, according to Sellars, we are able in the first place to tell when a given person is having a certain thought. Since thoughts are, in the first instance, theoretical entities which are understood as being causally related to certain occurrences of patterns of observable behavior, we are able to tell, of others and of ourselves alike, that we are having certain thoughts on the basis of inference founded upon such observable behavior. Since the theoretical entities in question acquire, in the course of the use of discourse about them, a reporting role, since, that is, we come to be able to report our own thoughts without recourse to such inferences, there is a second means of telling when certain persons are having certain thoughts.

According to Sellars' model, these two ways of telling about the occurrence of thoughts are both logically and causally independent. For while the first rests upon our construal of thoughts as entities which are, on a theoretical basis, causally related to certain other occurrences, the second rests upon the fact that persons having certain thoughts happen to be so put together that the theoretical entities in question, thoughts, come to be causally related to class of occurrences different from either the thoughts in question or the associated overt behavior, namely, thoughts that one is having a certain thought, or meta-thoughts.

Now let us consider a case of a given person, George, who is having certain thoughts, and ask how we are to tell, and how George is to tell, when these thoughts occur and what their duration is. Like George, we are able to observe George's behavior, and to make inferences, on the basis of Jones' theory, concerning when George has a certain thought. George, however, has an additional way of telling, namely reference to his own meta-thoughts. Thus we may suppose that George has a certain thought, say T, and that he exhibits overt behavior, say B, of the sort which, on the basis of Jones' theory, warrants the inference that he is having T. And we may suppose further that as a result of his having T, he has the meta-thought that he is having T, namely M. This provides George with two different ways of telling when T occurs; for he may temporarily locate T on the basis of when B occurs, or on the basis of when he feels tempted to report that he has T, his feeling so tempted being a result of his having M.

Now T is supposed, according to Sellars' model, to be causally related both to M, and thereby to George's feeling tempted to report his having T, and to B. Since the causal relationships which T bears to M and to B are independent, however, it is easily seen that it is possible that M should occur at a time different from the time at which B occurs. Unless, then, George is clear concerning how long the causal processes which we are supposing relate T to M and T to B take, George might wish to assume that T occurs, as we might say, "just before" both B and M. If, however, M and B occur at different times, these two ways by means of which George is supposed able to tell when T occurs will generate conflicting results. That we should be unable, on the basis of no more empirical knowledge concerning these causal processes than contemporary science possesses, to make precise the time at which thoughts occur in the way in which we can ordinarily make precise the time at which physical phenomena occur, seems to be just what we should expect on the basis of Sellars' model.

Similar arguments can be constructed in an attempt to account, on the basis of Sellars' views, for the lack of precision which we meet with in any attempt to determine the duration of thoughts. For while it is our conviction, in the case of certain, typically sudden thoughts, that we can with a relatively high degree of precision state when one occurs (this presumably, on Sellars' model, due to the strong temptation we feel to report such thoughts and thus to the presumed strong causal linkage between the thought and the corresponding meta-thought), the occurrence of meta-thoughts does not seem to provide us with any means for determining when we cease to be in the psychological state in question. Similarly with the behavior patters which are supposed to be causally related to thought episodes. For the occurrence of the relevant behavior patterns may signal the onset of a given thought episode, but does not seem to require, by its own cessation, that we conclude that the duration of the thought episode has come to an end. Typically, we might be imagined as determining when we are no longer in the state of having a certain thought by noticing that our behavior has come to indicate the presence of a new and different thought, or by the occurrence of a meta-thought that we are having a certain thought which is different from the preceding one, but these means for determining the endpoint of the duration of a thought episode will not provide any high degree of precision, and the fact that there are two means for so determining which are both causally and logically independent allows for the same incompatibility of the results generated which we noticed above.

It will be useful, then, to turn to the general form of the objection to the identity thesis to which this section is devoted, and to an argument on which the objection is frequently claimed to be founded. If we are able to pick out occurrences of thoughts, so the argument might run, independently of our observing anything in the way of our behavior on the basis of which we would be warranted in inferring the presence of such thoughts, then we must do so on the basis of noticing certain characteristics of the thoughts. For if we are able to pick out occurrences of thoughts, or, more generally, of psychological phenomena, then it must be on the basis of noticing some features which they exhibit. And if the features are not the causal properties of resulting in certain behavior patterns, then they must be peculiarly psychological features. Since we are able to pick out such occurrences independently of noticing anything about our behavior, it seems to follow that we must notice peculiarly psychological features which psychological phenomena exhibit.{29}

If this argument is sound, then, it seems to follow, it must be the case that psychological phenomena, in particular, thoughts, exhibit certain properties which no physical phenomena exhibit, namely, those properties in virtue of noticing which we are able to pick out occurrences of thoughts when we do so independently of inference from overt behavior. For it seems to be the case that if the causal properties which thoughts have of resulting in certain behavior patterns are excluded from consideration, then any remaining properties by means of which we should be able to identify thoughts must be non-physical.

It is an important feature of this argument that we should be required to notice something about thoughts if we are able to pick out occurrences of them. (Compare [Rorty (1)].) Similarly, it is important that we should be agreed that we are able to pick out occurrences of thoughts independently of noticing anything taking place in the way of observable behavior. Sellars' views on what is involved in our being able so to pick out occurrences of thoughts diverge sharply from the account implicit in the argument above. For the way in which we are able, according to Sellars, to pick out occurrences of thoughts independently of reference to overt behavior involves the occurrence of meta-thoughts, and the causal connection between the meta-thoughts in question and our tendency to report the having of the corresponding thought. Now there is a sense in which the noninferential reporting which the having of meta-thoughts makes possible involves certain properties of the thoughts on which we are able noninferentially to report. For the fact that the thoughts on which we can noninferentially report do result in such meta-thoughts and in turn in the tendency to report the corresponding initial thoughts is a causal feature of those thoughts. It seems clear, however, that as Sellars views the matter, no noticing of these causal properties is required for our noninferential reporting. In any event, it is unclear why any properties which were noticed in the course of such reporting would be counted as non-physical properties, as against straightforwardly physical properties. For it seems clear that the properties which were noticed in connection with such reporting, given that any properties are in fact noticed, might be simply the properties of thoughts which make it possible that we should be able to make noninferential reports, namely, the causal properties of thoughts which link them to other thoughts and to certain forms of overt behavior such as the making of the noninferential reports in question. Unless there is some reason to believe that these causal properties of thoughts are non-physical properties, it does not seem that an objector to the identity thesis would be warranted in concluding that they are; for they might equally naturally be regarded as physical properties, just as we regard many other cases of causal properties with which we are familiar in the natural sciences as physical properties. It does not seem clear, however, that there is any satisfactory reason to regard them as non- physical.{30}

The idea that we should be required, if we are to report on a given sort of occurrence, to notice certain properties of that occurrence, seems, on this model, to be superfluous. For if the noninferential reporting of thoughts is construed as a causal result of the occurrence of those thoughts, the noninferential reporting seems to be understandable independently of any assumption about noticing features of thoughts. If one wishes, one can, compatibly with this model, view the tendency to report on a certain thought which results from the occurrence of the corresponding meta-thought as a noticing of certain features of the thought, and one may so construe this tendency preserving the required sense of 'notice'; for certain causal properties are, in this tendency, being responded to by the reporter. If this move is made, however, the properties do not seem clearly to be peculiarly psychological in any sense which rules out their being physical properties. If responding to these causal properties of thoughts is argued not to be what is involved in noticing features of thoughts which allow us noninferentially to report on them, then it is not clear why such noticing should be required, for noninferential reports can be accounted for without reference to such noticing.

There is one final argument which it is important to consider in defense of the claim that the identity thesis does not suppose it possible to make certain identifications which are, in fact, cross category identifications. The argument, considered by Smart in his original article, rests on the idea that it is not possible to construct sentences which describe psychological phenomena in the ways in which we wish to without invoking expressions which ascribe to those phenomena properties which are not ascribable to physical phenomena, and, in particular, which are not ascribable to the physical states with which we might come to identify psychological states. Smart's means of resisting this move, with regard to the identity thesis as applied to sensations,{31} is to propose translations of such psychological sentences into what he calls topic-neutral language, language, that is, which captures what is being asserted by the original sentences without invoking properties that could not be construed as physical properties just as easily as they could be construed as psychological ones.

It is worth spelling out just how this move might be supposed to apply to the case of thoughts, for Smart's own argument rests upon his own professed views concerning the nature of so-called secondary properties, in connection with the thesis as applied to occurrence of after-images. (Cf. [Smart (4) 148-149].) Smart's proposal concerning the translation of sensation sentences (in particular sensation reports) into language which is intended to be topic neutral with respect both to physical and psychological categories involves an important use of the locution 'something going on which is like':

When a person says 'I see a yellowish-orange after image', he is saying something like this: 'There is something going on which is like what is going on when I have my eyes open, am awake, and there is an orange illuminated in good light in front of me, that is, when I really see an orange.' . . . Notice that the italicized words . . . are all quasi-logical or topic neutral words. [Smart (4) 149-150, emphasis original]

Smart claims that this last fact, that the italicized words are topic neutral, explains why the ancient Greek peasant's reports about his sensations can be neutral between [psychophysical] dualistic metaphysics or . . . [his] materialistic metaphysics," [Smart (4) 150] and suggests further that this account explains why it is that the task of characterizing sensations by assigning properties to them has, in the course of past philosophical investigation, proved fruitless; for on this account, sensations can be, independently of empirical investigation, assigned properties only to the degree of saying that they are occurrences which resemble occurrences which take place under certain specifiable conditions.{32} It will be important to examine, in connection with the discussion of physicalism in sections VII and VIII, the respects in which this account (a) does not square with Sellars' account, and (b) may, to that extent and in ways suggested by that divergence, be profitably criticized; but at present our concern is first to see how this account might be recast to apply to the case of the identity thesis for thought episodes, and second to examine the merits of such a defense of the identity thesis so applied.

What is wanted, in order to apply this move to the case of thought episodes, is an indication of how to paraphrase sentences normally used noninferentially to report thoughts into language which is topic neutral with respect to psychological and physical categories. If such a paraphrase is possible, then either (a) the closed family of terms which figures in our revised version of Chisholm's thesis as expressed by sentence (2) must be construed as including at least some topic neutral terms, or (b) that thesis is mistaken.{33} What is wanted, then, is a topic neutral rendering of sentences like:

(10) I have the thought that-p.

We might imagine the translation taking the following form:

(11) There is something going on which is like what is going on when condition C is satisfied.

where condition C is the analogue, for the case of having a thought, of seeing an orange in the case of having a certain sensation. Our problem, then, is how to spell out condition C.

One might imagine attempting to provide such a condition as follows. One might capture what is involved in the having of a thought by saying that it is like what goes on when a person meaningfully and non-parrotingly utters a certain sentence. This characterization could be expanded by providing other patterns of behavior such that when a person engages in such behavior we suppose him (perhaps on the basis of the Jonesean theory) to be having a certain thought. And the condition could be further specified by adding to the foregoing that what is going on is also like what goes on when certain things happen to the person--say, he is asked a certain question, understands it and is responsive, as a typical language speaker with certain past experiences would normally be expected to be.

It is worth considering two sorts of objections to this move, as we have imagined it applied to the case of thoughts. The first is that the characterization of condition C is not restricted to terms which are, with respect to a psycho-physical dualism, topic neutral. For at least part of the characterization of condition C involves talking about persons' meaningfully and non-parrotingly uttering certain sentences, thereby invoking the semantic categories involved in talking about overt verbal behavior which both Chisholm and Sellars, at least, consider as in effect psychological phenomena. And, along the same lines, it is arguable that to characterize something in terms of certain actions requires that we talk about certain patterns of behavior which are, say, intentionally engaged in, and that to talk about behavior as intentional involves invoking psychological notions. Finally, to specify that a person is affected by a certain question which, we wish to require, he understands, is, it is arguable, to invoke psychological notions, at least implicitly. Thus it might appear that such a paraphrase of (10) as we have provided by (11) involves going beyond our topic neutral restrictions.

It seems clear, however, that this objection misses the mark. For if this were an adequate objection, the same would apply to Smart's program as applied to sensation reports. For Smart's paraphrase, cited above, uses both physical and psychological terminology in specifying the condition invoked. This does not, however, in either case mitigate the topic neutrality of the paraphrase. For in each case one is simply specifying that the occurrence of the mental condition in question is (at least) something like something which is going on when certain other occurrences, howsoever these must be specified, take place. Just what it is which is going on which counts as the mental occurrence in question is not in any way, in either case, characterized either by psychological or by physical terminology.

VI. The Identity Thesis (3): Further Objections

There is a second sort of objection to this program, which has been advanced by Cornman as a criticism of Smart's proposal as applied to sensation reports. In outline it runs as follows. [Cornman (4) 489-490] We may suppose that P1 is our original sensation report, and that M1 is the first approximation of a paraphrase, say, the one suggested by Smart in the passage quoted above. Cornman's claim is that there is a sentence which entails M1 but not P1, which we may call P2, and that as a result, since in at least one case, P1 and M1 are characterized by differing entailment relations, they cannot mean the same thing. In particular, since P2 entails M1 but not P1, P1 may entail M1, but the reverse is ruled out: P1 is supposed, on this line of argument, to be richer in meaning than M1. Cornman's example of a sentence P2, relative to the translation proposed by Smart in the passage quoted, is

(P2) I see a roughly spherical shape.

Cornman's argument continues by claiming that even if the initial paraphrase, M1, is revised so that P2 entails neither P1 nor M1, there is still a sentence, call it P3, such that P3 entails the revised paraphrase, M2, but not the original sentence P1:

(P3) I have a colored after-image.

Cornman's claim, then, is that this process can be generalized as follows. Each proposed paraphrase of a given sensation report, if it is topic neutral, will be entailed by a sentence which does not entail the original, and thus even if the original entails the topic neutral translation, the reverse cannot hold. It follows, concludes Cornman, that no paraphrase, if it is to be topic neutral, can be adequate.

It might be argued that the force of this criticism can be minimized if it can be shown that even if any proposed topic neutral paraphrase of P1 fails to entail P1, nonetheless a procedure exists by means of which the difference in meaning between P1 and the members of a series of proposed paraphrases can be, although not eliminated altogether, at least made arbitrarily small. To construct such a series, the following would be required. We may begin with P1 as our original sensation report, and M1 as our initial paraphrase. The latter, although not the former, is entailed by P2. So M1 is revised into M2, which is entailed by P3, although P3 does not entail P1. If M2 is to be closer in meaning to P1 than M1 is, however, it must also be the case that M2 entails both P2 and M1, for otherwise M2 might merely differ in meaning from M1 without more closely approximating P1. So the construction of such a series of successive approximations to P1, in regard of meaning, will require that each revised paraphrase of P1 entail all paraphrases occurring earlier in the series. If it is possible to construct such a series, then it might be argued that the difference in meaning between P1 and each successive paraphrase being made increasingly small minimizes the persuasiveness of the objection to Smart's proposal.{34}

Even if the force of Cornman's objection is minimized by such considerations, it nonetheless does not seem that such considerations overcome the objection. For if Cornman's objection is sound, it remains the case that the requirement of the topic neutrality of the paraphrase rules out the possibility that any paraphrase should ever mean the same as the original report. And the difference in meaning will be significant, in the sense that it should be possible for a speaker with a command of the locutions involved to construct a sentence driving the wedge between the original sentence and any proposed translation.

Now it seems to be the case that this form of objection applies equally well to Smart's proposal as modified to fit the case of thoughts. For no matter how we specify the condition C used in framing our paraphrase (11), the possibility is still open that there is more than one sort of thing which we can imagine going on when condition C is satisfied, one of which is the thought mentioned in (10), and one of which is either some different thought episode, or not a thought at all. If, in defending Sellars' position, then, we are required to defend the form of the identity thesis advanced above, we shall have to construct some means other than Smart's for dealing with the claim that the identity thesis involves the false assertion that certain cross category identifications are possible.

The means of dealing with the objection that is suggested by Sellars' treatment of discourse about thought episodes seems to attack the assumptions on which the objection is founded. It will be useful, in developing this line of thought, to return once more to that sort of intentional item with which we may feel more familiar, overt semantical episodes. In particular, we may consider the overt semantical episode of George meaningfully and non-parrotingly uttering a given sentence, S. Now what has happened can be described in various ways. It can be described merely as a physical occurrence, as the event of George's producing certain sounds; it can be described in terms of a speech act performed, as George's saying that S, or as George's meaningfully and non-parrotingly uttering S in accordance with the rules governing the use of utterances of a certain language. And we suppose that the occurrence mentioned in the first way of talking about what has happened is the same as (is identical to) the occurrence mentioned in the second. This notwithstanding, the second way of talking involves invoking expressions belonging to semantic discourse, while the first does not; and simply to say that George said that S, although it entails that something happened of a physically describable sort, says nothing about what that something was.

This can be exhibited more clearly as follows. We may suppose that in fact the following two sentences

(12) George said that-p.

(13) George uttered S.

are about the same occurrence, in spite of the fact that the first specifies a certain occurrence in terms of a person and certain semantical notions, and the second in terms of a person and certain physicalistic notions. And there is clearly a sense in which neither entails the other; for each specifies something which the other does not--the first that what happened was done as part of some form of non-parrotingly language using, the second that what happened involved a certain sort of physical occurrence. Thus, we might say, framing a particular identity statement, that

(14) George's saying that-p at time t is identical to George's uttering S at t.

While different categories, or sorts of discourse, are involved in framing the expressions flanking the identity sign in (14), this does not lead us to claim that (14) is a case of a cross category identity statement.

Now it seems, on this basis, that it is certainly possible, objections involving claims of cross category identifications notwithstanding, to frame general identity statements of the sort suggested by (14). If, for example, we had a catalog of all and only those utterances, the utterance of which could be used, as things stand, to perform a certain speech act, then it seems that the following sort of general identification would be warranted:

(15) (x)(If x is an occurrence of a saying that-p then x is identical to the occurrence of an uttering of sort S1 or of sort S2 . . . or of sort Sn) and (x)(If x is an occurrence of an uttering of sort S1 or of sort S2 . . . or of sort Sn then x is an occurrence of a saying that-p)

The identification expressed by (15) would be a general identity involving occurrences which are pickable-out either as utterances exhibiting certain empirical properties (in Carnap's terminology, certain formal features) or as occurrences playing a certain intentional role in overt language. This identity statement is exactly parallel to the sort exemplified by (7), which, according to the version of the identity thesis simpliciter which we have been considering, is not ruled out on the basis of logical or philosophical considerations alone. It appears, then, that no difficulties involving cross category identification face the identity thesis as it might be applied to this sort of intentional item.

The obvious move to make, then, is to question why such a difficulty should be raised when the identity thesis is applied to a different sort of intentional item, namely thought episodes. If we are to accept the view that any intentional item, of any sort, must exhibit some determinate factual character if it is to be able to play its intentional role, then there must be some way of identifying it independently of the fact that it does play that role. If we can so independently identify it, then it seems clear that we can say that all and only occurrences of that empirically specifiable sort play such- and-such a specifiable intentional role. To say this, however, seems to amount simply to asserting that identifications of the sort considered by the identity thesis are, indeed, possible. Our willingness to entertain the possibility of such identifications in the case of overt semantic episodes should, it seems, lead us to be willing to accept such a possibility in the case of those intentional items which, following Sellars, we may conceive of on the model of overt semantic episode.

If considering the case of speech episodes is to lead to our rejecting the force of the objection to the identity thesis which rests on the claim of cross category identification, it will be useful to see how this objection fares as applied to just this sort of case. Just as the objector argued, in the case of thoughts, that thoughts exhibit certain properties which are not ascribable to physical phenomena, and that certain properties exhibited by physical phenomena are not ascribable to thoughts, by parity of reasoning we might imagine the argument to run as follows. Physical phenomena exhibit certain properties which cannot be exhibited by overt speech episodes per se, for such episodes, being essentially intentional items, cannot be characterized in terms of any physical properties; similarly, physical phenomena do not, although speech episodes do, exhibit intentionality. Thus, the argument continues, it cannot be the case that certain speech episodes are identical to certain physical phenomena, for instances of the two sorts of items cannot have all their properties in common. Thus any identity thesis as applied to speech episodes must be rejected.

In defense of this argument, one might imagine applying a modified version of Cornman's objection. One might, we may imagine it argued, try to get around the objection of the preceding paragraph by arguing that what we say when we talk about speech episodes can be equally well put by using topic neutral language. Thus a sentence like

(16) George said that-p.

could be paraphrased, in topic neutral language as

(17) George produced sounds of description D, and something was going on which is like what goes on when condition C is satisfied.

for an appropriate choice of description D and condition C. Such topic neutral translation cannot, it would then be urged, succeed, for no matter what paraphrase into such language is offered, the paraphrase can always be shown not to entail the original by finding some sentence which, although it entails the paraphrase, does not entail the original.

If the parallel nature of thoughts and semantic episodes which Sellars advances is not misleading, then this line of reasoning should be no less persuasive than the corresponding line of reasoning applied to thoughts. The inadequacy of such argument, seen more clearly as applied to overt speech episodes, is that there seems no reason to believe that certain physical phenomena cannot exhibit intentionality if what is involved in exhibiting intentionality is the playing of a certain intentional role--a role, that is, which is analogous to that played by non-parrotingly uttered meaningful tokens of speech. Similarly, there seems no reason to believe that intentional items cannot exhibit certain physical characteristics, and if we accept the requirement that if something is to play an intentional role it must have certain empirical properties, it seems reasonable to believe that intentional items do exhibit such characteristics. In any event, it seems clear that this much is so in the case of overt speech episodes, which, both Chisholm and Sellars agree, are cases of intentional (or implicitly psychological) phenomena. If this is so, then it seems that if one is to reject this possibility in the case of thought episodes, one must show some relevant disanalogy between the two sorts of cases, some disanalogy, that is, between thoughts and speech episodes which would lead us to discriminate between them with regard to the applicability of this argument.

In the absence, then, of any further objections to the form of the identity thesis we have been considering, then, we may accept the idea that so far from being a disadvantage of Sellars' position that it requires our adopting that thesis, this requirement seems to point to an important strength of that position. For Sellars' views not only seem to point to a version of the thesis which seems acceptable, but seem also to provide us with a strategy for defending the thesis against certain recent objections.

VII. The Thesis of Physicalism (1)

It will be useful, in moving on to consider the revised thesis of physicalism which we left behind in section III, to consider a certain sort of objection to the identity thesis which has not been raised in the foregoing three sections. Sentence (7), taken as expressing the sort of general identification wanted by the proponent of the identity theory, involves the claim that no matter what instance of a person's having the thought that it is raining we may consider, we shall find that that occurrence can also be described in certain neurophysiological terms. Now it might be objected that although this claim is certainly possibly true, it is also certainly a rash claim to make. For it seems plausible to suppose that there are beings, located in some far corner of the universe, who like us think (in the sense of having certain states which exhibit the relevant causal relations to items of behavior which in turn play certain intentional roles), but are radically unlike us in respect of composition and function. Thus such beings might be composed of chemical substances analogous to our organic compounds but differing in regard of containing silicon atoms in place of any carbon atoms. And this might be imagined to result in radically different "neurophysiological" functioning. It seems clear, however, that these beings could be sensibly said to think, say, that it is raining, and that we could, upon encountering them, learn to pick out cases of their having such thoughts independently of our knowing anything at all about their makeup and biological mechanisms. Thus, it might be urged, even if all the cases of having the thought that it is raining which we have thus far encountered do seem to satisfy (7), unless we have good reason to suppose that cases such as that envisaged do not, in fact, exist, we are not warranted in asserting (7). Since, quite to the contrary, one might conclude, we have good reason to think that such cases do exist, and certainly no reason to rule them out, we are unwarranted in advancing the identification expressed by (7).{35}

Clearly, as this objection stands, it is beside the point. For the identity thesis involves claiming not that the identifications in question are as a matter of empirical fact plausible, but rather that they are not ruled out by logical or philosophical considerations. The objection we have just considered, however, rests on the idea that such identifications are empirically unwarranted, not that they conflict with logical or philosophical results. While such an objection might deprive the identity thesis of its interest by successfully arguing that the identifications in question, although logically possible, are nonetheless unacceptable, it cannot be advanced as an argument against the thesis itself.

It is worth examining, however, what reaction a scientist who had propounded identifications such as that expressed by (7) might have to such a case as we have envisaged. We may suppose our scientist to have advanced a number of identifications on the basis, say, of correlations of certain sorts between psychological and neurophysiological phenomena, and on the basis of acceptance of such identifications allowing the scientist to derive certain psychological laws from certain neurophysiological ones. He is then faced with the appropriate data concerning these beings whose composition involves, instead of carbon atoms, atoms of silicon.

One reaction we say imagine him to have is, of course, simply to regard this as evidence that any identification of the requisite sort is hopeless. A more natural one, however, appears to be that he will work towards suitably revising the identity statements to accommodate the new data, much as we might go about revising an identity of the form of (15) if faced with data concerning speakers of a language hitherto unknown to us. And one result of coming to be able so to revise such identity statements as to accommodate such new data might be the discovery of laws linking the makeup of our silicon beings to the makeup of human beings as we know them; thus we might envisage coming to be able to provide a suitably broad description of the neural occurrences identified in our case with the having of certain thought and of the corresponding state so identified in other cases. Since all this seems an acceptable projection regarding the possibilities open to scientific research in the face of such discoveries, the warrantability of some identifications of the sort envisaged by the defender of the identity thesis seems acceptable.

We are now in a position to return to further consideration of the final formulation of the thesis of physicalism which was first presented in section III. This formulation involves the claim that

(18) Every psychological sentence is nomologically equivalent to some nonpsychological sentence which contains no terms not required for the language of physical science.

where two sentences are nomologically equivalent if, and only if, each conjoined with some natural law entails the other. Since this version of the thesis of physicalism was used, in section III, to give substance to the idea that we can say all we might wish to about the world with recourse only to the language of physical science, we may regard the thesis of physicalism as also involving the claim that

(19) If S is a psychological sentence and S', although nonpsychological, is nomologically equivalent to S, then we can dispense with S in favor of S' whenever we wish to describe whatever is described by S.{36}

To evaluate these claims, let us consider a particular case, say, involving our having established that the identity statement (7) is warranted in the way in which a natural law is. It would follow, according to (18), that

(20) George is having the thought that it is raining.


(21) George's CNS exhibits a pattern of neural firings satisfying condition C.

are nomologically equivalent, and, according to (19), that we can dispense with (20) in favor of (21) whenever we wish to describe whatever is described by (20). It is clear that it could be the case that neither of these claims were true even if the general identity statement (7) were true (for some appropriate condition C); and thus that in spite of our argument that Sellars' position on intentionality, so far from conflicting with the truth of sentences like (7), entails the possibility that such sentences are true, nonetheless Sellars' (or Chisholm's) views might conflict with the truth of one or both of these claims.

We have seen that the possibility that there are beings which have thoughts and which do not exhibit the neurophysiological characteristics involved in (7), while it does involve the possibility that sentence (7) (and others restricted to human neurophysiological considerations) is false, nonetheless does not rule out the possibility that sentences like (7) are true. It might be argued, however, that this possibility does rule out the truth of the claims advanced in the preceding paragraph. The argument might proceed as follows. Let us suppose that (7) is true, and that the way in which it is warranted is like the way in which natural laws are established. It would follow that (20) and (21) are, in fact, nomologically equivalent. Would it follow that we can dispense, in the required sense, with (20) in favor of (21)?

It will be useful in considering this question to return to our scientist, whom we have envisaged faced with data concerning these silicon beings providing evidence counter to the claims of (7). As we have seen, the scientist may be imagined, in fairly straightforward ways, to go about suitably revising (7) to accommodate the new data. The question arises, then, as to just how he is to go about making such revisions, and on what basis he may reasonably judge a given revision as being acceptable.

One fairly obvious restriction on any such revision is that the revised sentence (7) (we may call it (7')) will have to be such as to continue to take account of the identifications which (7) asserted. In the simplest case of revision, this will be accomplished by revising the "neurophysiological" side of the identification so that the condition to be satisfied is disjunctive between the original condition and the condition(s) suggested by the new data. In the more sophisticated sort of revision our requirement will be met if an entirely new condition specified in place of the original neurophysiological one is satisfied by whatever satisfied the original neurophysiological one. Our concern, then, is how the scientist may go about framing new conditions. The straightforward answer to this problem seems to be the following. Since the scientist is able to frame the original identification expressed by (7), he must be able somehow, independently of any knowledge of the neurophysiological states of his subjects, to tell when his subjects are, and are not, having the thought that it is raining. If he can do this, he should be able to tell, independently of any knowledge about the anatomy and "physiology" of our silicon beings, when a given one of them is, or is not, having the thought in question. He can therefore go about making just the sort of correlations, and projecting just the sort of subsumption of laws that led to the establishment of (7), when he wishes now to make suitable revisions of the physical condition mentioned in (7).

If this is how the scientist is to be envisaged as making the appropriate revisions of (7) to accommodate the new data on silicon beings, our argument which purports to rule out our being able to dispense with (20) in favor of (21) on the basis of their nomological equivalence might proceed as follows. While it is clear that the mere possibility of our discovering silicon beings of the sort imagined does not, of course, require us to reject the nomological equivalence of (20) and (21) (what natural law could be retained if its rejection could be warranted on such a slender basis?), we must retain (20) as having an application independent of that of (21), for unless we do, the scientist has no recourse when he comes even to envisage the possibility of such silicon beings. For it is only reasonable to require, in the face of the possibility that such beings exist, that we retain some means for the scientist effectively to deal with the revisions in natural laws which would be called for if such a possibility were realized. And if, in any strong sense, we dispensed with (20) in favor of (21) on the basis of their present nomological equivalence, on the basis, that is, of their nomological equivalence given the current state of scientific knowledge, then we would be depriving the scientist of just what is needed to deal with the possible new discoveries. For unless (20) retains an application independent of the application which (21) has, the scientist will be unable to frame his revision. And if (20) does retain such an independent application, then it is hard to see what might be meant by saying that, in spite of this, we have dispensed with (20) in favor of (21) on the basis of their (current) nomological equivalence.

If this argument is sound, then in spite of our accepting (20) and (21) as nomologically equivalent on the basis of the nomologicality of (7), we should be debarred from making the final step--applying (l9)--and dispensing with (20) as applicable independently of (21). It will be useful to make three points before turning, in the penultimate section of this chapter, to an evaluation of this argument.

(A) The notion of independent application of two sentences which is invoked in the argument does not turn on the idea that, as things stand, (20) would be applied on the basis of certain noninferential reports of persons and of certain patterns of behavior, whereas (21) would be applied on the basis of the examination of neurophysiological states of persons. For as things stand, it is of course the case that these two sentences have independent application. What is envisaged as our dispensing with (20) in favor of (21) is rather that, on the basis of our establishing their nomological equivalence given some current state of science, we should come to apply (21) on the basis of the sorts of ways in which we formerly applied (20). We would, that is, come to use not only results concerning neurophysiology to apply (21), but also the noninferential reports that someone is thinking that it is raining, or certain patterns of overt behavior, in our application of (21), and we would further imagine that persons might be trained (as it seems clear that they could be) to utter sentences like

(22) My CNS is in a state satisfying condition C.

whenever they now, as things stand, are tempted noninferentially to report that they are thinking that it is raining. For at the very least, the possibility that persons should be trainable in such a way seems to be one which must be settled empirically: there seems to be nothing in our current knowledge concerning the acquiring of speech dispositions which would rule out this possibility.

Our dispensing of (20) in favor of (21), involving these steps, would result in (20) simply disappearing from use as a result of the nomological equivalence in question. The force of our so dispensing with (20) would be that we could do whatever we had previously been able to do with (20) by means of using (21) whenever we had, formerly, used either (20) or (21). The force of the argument against our being able to make this move on the basis of a (current) nomological equivalence rests on the idea that although we might be able to do all our reporting, describing, arguing, explaining and so forth without using (20), we nonetheless would be required to retain an independent application for (20) if we wished also to retain the possibility of dealing with what seem to be counter-examples to (7).{37}

(B) The idea, invoked in the argument presented above, that we are able to frame the original identity statement (7) on the basis of our being able to pick out cases of havings of certain thoughts and occurrences of certain neurophysiological episodes independently from one another, was one of the important respects in which Sellars' views seemed to involve the truth of the form of the identity thesis defended above. For it was in just this way that the possibility of framing general identifications such as that expressed by (7) was defended. This possibility rested upon Sellars' insistence that intentional items such as thoughts exhibit some determinate factual character that, independent of the specific intentional role played by a given item, allowed us to identify and reidentify the item in question; as well as upon his insistence upon the converse, that since the learners of Jones' theory, and indeed Jones himself, are supposed not to be able to specify the determinate factual character of thought episodes save that they provide the causal efficacy of such items which is required by the theory, it follows that such episodes are identifiable and reidentifiable on the basis of the intentional role played, independently of their specific determinate factual character.

The argument that sentences like (20) cannot be dispensed with in favor of sentences like (21) seems therefore, to reflect just the sorts of convictions which we have seen that Sellars holds. For the argument rests upon the conviction that there are some ways in which it can be seen that sentences like (21) cannot be enough; we must also have independently applicable sentences like (20). For while it is possible, of course, to revise the speech dispositions of speakers of a given language (or class of languages), to revise them in the way envisaged by the second part of the thesis of physicalism requires that we give up the possibility of dealing effectively with certain sorts of new scientific data, so the argument runs.{38} It is therefore a natural step to take on the basis of Sellars' insistence on the independent identifiability and reidentifiability of thought episodes and neurophysiological occurrences to insist that in spite of the possibility of establishing, nomologically, general identifications such as that expressed by (7), this requirement must be preserved and the dispensing of (20) in favor of (21) envisaged by the revised thesis of physicalism must therefore be ruled out.{39}

(C) It might be argued that we can, our argument notwithstanding, accept a revised version of (19) which specifies the respects in which and the purposes with regard to which sentences like (20) can be dispensed with in favor of sentences like (21). Thus one might wish to specify that one need retain an application of (20) independent of that of (21) only for the purpose of required revision of the original identity statement (7). To retain such independent application, however, in any case and for any purpose, seems to run sufficiently counter to the spirit of the thesis of physicalism under consideration, that we may regard such a revised version as tantamount to the rejection of the thesis altogether. For while the thesis in its unmodified form projects the modification (or, in any event, the in principle modifiability) of speech dispositions such that sentences like (21) will come to perform all the linguistic tasks formerly performed by sentences like (20), the modified version clearly involves a limitation on this alteration of speech dispositions. For if the modified version is to accommodate what seems to be the conclusion of our argument, then at least some of our speakers must be imagined (say, at least the scientists who would deal with the problems arising from our encounter with silicon beings) as retaining speech dispositions involving application of (20) and (21) on independent grounds. But it is just this which the thesis of physicalism wishes to say is not so. Let us move on, then, to a consideration of the adequacy of the argument which has been presented.

VIII. The Thesis of Physicalism (2)

One initial reaction to the argument stems from the idea presented in the third point made at the end of the foregoing section. One might wish to argue that even if all the speech dispositions are altered in the way envisaged by the thesis of physicalism, we would, in fact, be able to handle effectively the problems encountered in the data concerning silicon beings. For we would be able, in such a case, in effect to recapture the speech dispositions required to deal with such data. One may imagine the process going as follows. Our scientist, having no use for sentences such as (20) and no speech dispositions relating to the application of such sentences, encounters the silicon beings. He discovers that many of his reports about them are false, for when on behavioral grounds he applies, say, (21) to these beings, he discovers that they are not in the neurophysiological state mentioned by (21), and vice versa. He is, moreover, confused by their use of sentences like (21): he takes such sentences as involving an assertion which would be made, indifferently, on behavioral or neurophysiological grounds as well as noninferentially, and he discovers that this is not the case.{40} (Since our scientist is envisaged as having no speech dispositions governing the application of sentences like (20), we may imagine the same for sentences like (7), for the expressions contained in sentences like (20) which resulted in their being dropped from usage also, by hypothesis, figure in sentences like (7). He will therefore not encounter problems in discovering, pure and simple, that the identification expressed by (7) which warranted dispensing with (20) in favor of (21) is simply false.)

We may then suppose that our scientist, noticing these phenomena and feeling the resultant scientific discomfort, frames what are to him new notions to cover cases of intentional items playing certain intentional roles and pickable out both noninferentially and behaviorally, on the one hand, and certain neurophysiological occurrences on the other. He is then in a position to see that sentences like (7) are true for human beings, and were, at the time he encountered the silicon beings, in effect taken for granted by all human beings. For the speech dispositions of human beings at that point were such as to result in the same application for both sorts of new notions which he has just framed. He is then in a position to go about framing a new identity statement, (7'), which will be true of not only human beings but silicon beings as well. And we may imagine, finally, that not discovering any new language users for an appropriately large span of time, the same collapsing of intentional talk about thought episodes and talk about physical states meeting the conditions specified by (7') takes place as had previously taken place with the appropriate talk about neurophysiological states of human beings.

Our proposed answer to the argument would then run as follows. Since we can envisage the speech dispositions of human beings being modified, as envisaged by the thesis of physicalism and, in particular, by the truth of (19), without its being a result that scientists would be unable effectively to deal with data on silicon beings, the argument against the thesis must fail.{41} For if the speech dispositions of human beings can, in principle be so modified as a result of scientific research, then we can, in fact, dispense with (20) in favor of (21).

It will be useful to consider what may, for the purposes at hand, be viewed as a parallel case. Let us ignore a possible distinction between the strict identity invoked by the identity thesis as discussed above and some form of theoretical identity which, it has been argued (by Cornman and Nagel) the identity thesis might invoke instead. Let us further suppose, contrary to the supposition of Cornman and Nagel, that

(23) Water is identical to collections of H(2)O molecules.

is a case not of some sort of theoretical identification but of the strict identification supposed in the foregoing discussion. If we suppose, finally, that (23) expresses a contingent truth, then, it seems, it must be possible to pick out cases of water on a basis independently of that used to pick out cases of collections of H(2)O molecules--that is, that 'water' and 'collection of H(2)O molecules' have, as things stand, applications which are independent and, to some extent at least, governed by different speech dispositions of English speakers.

We may now imagine that the scientific results surrounding the truth of (23) become so well known that it happens that reports of water simply drop out of use, and become replaced by reports of collections of H(2)O molecules. The speech dispositions of human language users become altered so that instead of ever using 'water', 'collection of H(2)O molecules' is used in its stead. Parallel to what we envisaged in the case of psychological phenomena, we may then imagine that somewhere in the universe there is discovered a substance which has all the properties which, before the advent of chemical analysis, would have led us to classify the substance as water, but is composed (contrary to what we currently believe as laws of nature) of two atoms of tin and one of nitrogen-- Sn(2)N.

Before the "disappearance" of 'water' which we have envisaged, we might imagine scientists discovering such a substance engaging in argument as to whether Sn(2)N should or should not be considered water. Such a debate would be set in a context of scientific reaction to the necessity, on the basis of this discovery, of revising a fair number of chemical laws, including, possibly, (23). Our question is, however, what the scientists' reaction would be should such a substance be discovered after the projected disappearance from linguistic use of the expression 'water'. Two possibilities seem to present themselves. (a) Scientists may, parallel to the case of psychological phenomena presented above , decide that it is best to revise linguistic usage so that 'collection of H(2)O molecules' no longer is true of anything which either exhibits the phenomenal properties of water or exhibits certain chemically specifiable behavior, and, in effect, return to using two expressions. On this basis, Sn(2)N (if its phenomenal behavior is sufficiently like that of water--this can be granted for our case) might be classified as water, and then (23) would be accordingly revised.

This reaction to our discovery of the water-like Sn(2)N may strike one as somewhat implausible. For something along the lines of (23) seems to be so well embedded in the speech dispositions of current speakers that it seems unclear where one should draw the line concerning which properties belong to water as water, and which to water as a collection of H(2)O molecules. One might therefore imagine the following as the scientists' reaction to our projected discovery. (b) Instead of revising any linguistic usage (such as that of 'collection of H(2)O molecules'), scientists might simply maintain that the discovered substance just is not a collection of H(2)O molecules and leave it at that. Clearly, they would be saddled with a substance which behaved in many respects, perhaps in all easily observable respects, like collections of H(2)O molecules, but the substance clearly would not, by hypothesis, consist of such a collection.

The plausibility of alternative (b) over (a) seems to rest, in large measure, on the fact that it seems now to be the case that 'water' and 'collection of H(2)O molecules' are, to some extent at least, no longer applied independently of one another. This fact, partially a fact about our speech dispositions, also makes plausible the idea that (in this case without too great an inconvenience) 'water' might be dropped from usage and 'collection of H(2)O molecules' used in its stead. How do these considerations apply to what we have imagined as an analogous case, namely the disappearance of psychological locutions in favor of certain neurophysiological ones on the basis of certain nomological equivalences?

If we imagine the dispensing of sentences like (20) in favor of those like (21) to be a gradual one, then it seems reasonable to expect that before the complete disappearance of psychological expressions, language speakers should first arrive at a stage much like that we are in now with regard to (23). That is, before the complete disappearance of such expressions but after the establishment of the nomological equivalences in question, we may imagine a developing lack of clarity concerning what features of thoughts are the ones thoughts exhibit in virtue of being thoughts and what features are the ones they exhibit in virtue of being certain neurophysiological items. Put more precisely, we may imagine speakers (and scientists) becoming unclear about where to draw the line between phenomena which warrant (independently of other knowledge) our applying sentences like (20), and phenomena which warrant (independently of other knowledge) our applying sentences like (21). If this is so, however, then if our analogy is in the relevant respects an accurate one, it seems that the scientists' reaction to the discovery of data concerning silicon beings would be parallel to alternative (b) above, rather than to (a).

To spell this out, we may see the case as follows. Having discovered silicon beings and having made some fair success of translating their language into ours, the scientists simply deny that it is the case that silicon beings have thoughts. After all, silicon beings share none of the neurophysiological mechanisms which are in question when our scientists, in their language atrophied of psychological locutions, describe human thoughts; while their behavior is much like ours, and they even emit language-like sounds which are translatable (in some fairly adequate way) into our human languages, this does not warrant us in assuming that these beings think as well. For the having of thoughts is simply, for our scientists, a matter of having certain neurophysiological states which are, as a matter of contingent fact, related to overt behavior in certain ways and are, also as a matter of contingent fact, reportable noninferentially. The conclusion of this line of argument is, then, as follows. If we were exactly to imagine ourselves dispensing, in accordance with the thesis of physicalism, with sentences like (20) in favor of those like (21), we would not resurrect the independent applicability of these two sorts of sentences like (21) as true of them, and, having no independently applicable sentences like (20), we would be restricted simply to describing their physical behavior and their physical makeup. If this conclusion is sound, then the argument, presented in section VII against the acceptability of the thesis of physicalism, is, we may conclude, also sound.

Our conclusion seems to depend on the idea that faced with data concerning silicon beings, our scientists with their modified speech dispositions would have no more motivation to construct (for them) new notions applying independently to psychological and neurophysiological phenomena than they would have motivation in the corresponding case to construct (for then) new notions independently applicable to water and to collections of H(2)O molecules. It is this idea which we may imagine the defender of the thesis of physicalism would wish to attack. For if the cases are in such a way disanalogous, then it could not be argued that our dispensing with sentences like (20) in favor of those like (21) would result in our being unable effectively to deal with the data presented us in the discovery and investigation of silicon beings. It is important, then, for us closely to examine this assumption.

The disanalogy between the two cases seems to rest with the fact that in the case of our silicon beings, we are, by hypothesis, dealing with language users of some sort. (If this were not an aspect of the situation we have projected, it is difficult to see how the problem we are facing would be importantly different from the familiar problem concerning whether so-called higher animals, such as monkeys and dolphins, have thoughts.) If we regard silicon beings as language users, however, we are in so doing already attributing to them what we attribute to ourselves by saying that we have thoughts. Thus the fact that our scientists are envisaged as treating silicon beings as language users might be thought as providing a prima facie case for projecting the scientists' reaction along the lines of alternative (a) above.

This point may perhaps be put more aptly as follows. Whereas in the case of water and collections of H(2)O molecules we may become unclear concerning just what their current independent applicability consists in, the case of psychological phenomena fails to be analogous in just this respect. For the applicability of sentences like (21) on the basis of certain forms of overt verbal behavior is sufficiently radically different from their applicability on the basis of neurophysiological phenomena that this radical difference would be motivation enough for the scientists to adopt an alternative like (a) above rather than like (b). And this would be so even if the dispositions of all human language users had, prior to the discovery of silicon beings, become such as to involve the use of sentences like (21) where we currently use, independently, such sentences and sentences like (20) as well. For unlike the case involving the identification expressed by (23), the basis on which sentences like (21) would then be applied would include two easily distinguishable sorts of phenomena (as well as the "noninferential basis").

If the foregoing argument is sound, then we must reject the argument presented in section VII, and, in the absence of other arguments to the contrary, we may accept the thesis of physicalism originally presented in section III. Moreover, since our attempt to find reasons for rejecting that thesis has been governed by considerations suggested by Sellars' views on intentionality and psychological discourse, it seems that there is no incompatibility between Sellars' views and the thesis of physicalism which we have examined. It remains, then, to inquire into just what would be involved in accepting such a thesis, and, in particular, in accepting it subject to the arguments presented on its defense.

Briefly put, accepting the thesis as it has been defended here involves accepting the idea that we could, as a result of certain scientific discoveries and developments, come to speak a language which differed from ours in that it contained no expressions that were applied to psychological phenomena but not to certain neurophysiological phenomena. These expressions would be applicable indifferently to certain phenomena on the basis of neurophysiological facts, behavioristic facts and noninferentially. Our thus dispensing with current psychological locutions would not, however, warrant the claim that we could not encounter phenomena which required us to revise the language from which psychological expressions had disappeared so as, in effect, to reintroduce them.{42}

The thesis of physicalism as we have defended it, then, is of a form which does not project the linguistic alterations as permanent or irreversible; in fact, in order adequately to defend the thesis, we have seen it necessary to allow explicitly for the possibility that the linguistic modifications in question should, in effect, at least partially reverse themselves. For just as the linguistic modifications projected by a defender of the thesis of physicalism result from certain scientific discoveries and developments, so the thesis must allow that in the course of even further scientific discoveries and developments a modification of our linguistic usage should at least partially and temporarily be required in the reverse direction.{43}

If this restriction on the acceptability of the thesis of physicalism is taken note of, then it seems that an important part of the force of the argument against the thesis presented in section VII is preserved. For that argument rested on the idea that there were certain tasks which our current language, with its independently applicable psychological and neurophysiological locutions, could perform, but which a language devoid of peculiarly psychological expressions in the sense projected by the thesis could not perform. If our restriction on the thesis is accepted, then we have in effect admitted the truth of this claim. Our defense of the thesis of physicalism involved, however, the idea that which tasks could adequately be performed by means of this or that language was, in at least large measure, a result of the state of scientific knowledge (taken broadly) at a given time. Thus the claim that there are data (say, concerning silicon beings) which the language as modified in accordance with the thesis of physicalism could not effectively deal with, is a claim which must be relativized to our current state of scientific knowledge. Since the thesis of physicalism is itself so relativized, that is, it is a claim concerning what could happen as a result of certain scientific developments, this restriction does not appear to run counter to the spirit of the thesis.

It is a fairly straightforward assumption, however, that a thesis such as that expressed by the conjunction of sentences (18) and (19) does involve envisaging a change in linguistic use and speech dispositions which is in principle permanent. Thus it seems reasonable to suppose that the often expressed resistance on the part of philosophers to such a thesis does result from so construing it. There is of course a sense in which it seems true that such a modification of our language as the result of certain scientific developments could be permanent without thereby resulting in any inability to do the linguistic tasks which we could perform with our present unmodified language, for there might be no data on silicon beings of the sort to collect, or by the time they were encountered our state of scientific knowledge might be such as to handle the data without recourse to expressions applicable on the basis of intentional roles independently of physical characteristics. (Cf. p. 327, fn. 1.) The important point is that no amount of scientific data, short of being exhaustive of the sort of phenomena in the universe (past, present and future) which we might encounter, could warrant us in accepting such modifications in the certainty that reverse modifications would never be required.

It is not clear that Sellars' rejection of the identity thesis under certain interpretations is very much different from a rejection of the thesis of physicalism which has been examined. For Sellars seems to stress that intentional discourse about thought episodes is not, in principle, replaceable by talk about neurophysiological states. And in some passages at least (Cf. [Sellars (5) 193-194]. ) he seems to reject what he is rejecting on the basis of the requirement that we retain the categorial framework of intentional talk about thought episodes. (Cf. [Sellars (14) 27-30]. ) Such an insistence is similar to the basis of the argument against the thesis of physicalism presented in section VII. What has been argued, however, is that it does not appear, given the restriction mentioned above, that there is any adequate reason to reject the thesis. For the thesis-cum- restriction appears not only to be defensible in itself but also to be compatible with the views of Sellars examined in the present chapter and in the two foregoing ones. If Sellars is to reject our thesis of physicalism, then if the foregoing arguments are sound, he must do so on grounds independent of the considerations presented in those chapters.

In seeing this it is particularly useful to make reference to the closing paragraphs of Sellars' "The Identity Approach to the Mind-Body Problem." Although the so-called Identity Approach which he examines and rejects in this paper is concerned not with thought episodes but with what Sellars regards as nonintentional items--sensations--the paper is useful in suggesting Sellars' position with regard to the thesis of physicalism we have examined. For Sellars argues that the reduction of the psychological to the physical is unlike the reduction of chemistry to physics, thereby suggesting that what is at stake in his argument may be more something like our thesis of physicalism than like the version of the identity thesis we have examined and (on Sellarsian grounds) defended. His position seems to involve that the "logical space" of thoughts--the "determinate logical grammar" [Sellars (18) 291] of discourse about thoughts-- must be relocated in any modified language which results from such scientific advances as we have been projecting.

The requirement that something which is somehow equivalent to our current way of talking about thoughts independently of their determinate factual character reappear in any such modified language would seem to be unwarranted unless it is specified why this reappearance is required. In order to specify this, however, it would seem necessary that we claim that unless the modified language contains something which is somehow equivalent to our current discourse about thoughts, then we will be unable to perform certain tasks with the modified language which we can with our present way of talking. In this way it appears that what Sellars may have in mind is certain respects in which such a modified language would, unless it contained some version of the "logical space" of thoughts, be inadequate.{44} It is, however, just this sort of claim which has been examined and rejected in the course of the last two sections. For if the argument of the present section is sound, then such a claim must involve not only specifying the sorts of inadequacies which a language which did not include the "logical space" of thought episodes would have, but also the relativity of any such claim to the state of scientific knowledge at a given time.

In closing, it is worth suggesting an alternative reading of Sellars' dissatisfaction with a view which seems roughly the same as our thesis of physicalism. For it might be argued that we have misconstrued what is involved in his insistence that the "logical space" of thought episodes be relocated in the modified language. Our restriction on the thesis of physicalism involves claiming that the dispositions governing the use of sentences in the modified language such as (21) provide that such sentences be applicable (roughly) indifferently on the basis of neurophysiological facts, and on the basis of behavioral facts and noninferentially. For it must be allowed that scientists could, if necessary, so revise usage as to provide us with expressions which were applicable on the former basis independently of the latter, and vice versa. If this is what is required by Sellars' insistence upon the relocation of the "logical space" of thought episodes, then the foregoing arguments go to support his contention.{45} In the absence of a clearer formulation of what is involved in Sellars' rejection of the thesis he examines in "The Identity Approach to the Mind- Body Problem," however, it is difficult to evaluate whether that rejection is, in fact, warranted

IX Summary

In the foregoing eight sections, three philosophical theses have been discussed, with a view both to examining the positions which Chisholm and Sellars adopt (or may reasonably be regarded as adopting) with respect to these claims, and to presenting some evaluation of the merits of these claims. Because the conclusions which have been defended in these foregoing sections have been distributed throughout a rather lengthy and involved discussion, it may be useful, in closing, to present a recapitulation of these conclusions, making some reference to the sections in which these conclusions have been defended.

The first two sections were devoted to a discussion of a thesis of logical behaviorism, formulated by sentence

(1) Every sentence about the thoughts and propositional attitudes of persons means the same as some sentence about the observable behavior of persons or dispositions or propensities to such observable behavior.

It was argued in section I that Chisholm's claims in "Sentences about Believing" are jointly incompatible with such an assertion, but since the claims made in "Sentences about Believing" were found to be indefensible as they stand, this incompatibility does not warrant us in rejecting the truth of (1). It was also argued that Chisholm's later claims (in [Chisholm (3)], [Chisholm (4)], [Chisholm (9)], [Chisholm (l0)], and [Chisholm (11)], which were found acceptable in Chapter II) seem to provide us with good reason to reject (1) as false, given that we accept the idea that if two sentences are equivalent in meaning then they are about the same thing(s); but that since if we grant this latter claim we are thereby warranted in rejecting (1) as false without making reference to Chisholm's more recent claims, these more recent claims are superfluous in an attempt to find an argument in defense of rejecting the truth of (1). It was argued, finally, that the idea that if two sentences are meaning-equivalent then they are about the same thing(s) is itself defensible, and that for this reason we are in fact warranted in rejecting the thesis of logical behaviorism as false

In section II, a distinction was made between the thesis expressed by (1) and a different thesis of nomological behaviorism, expressed by sentence (3). It was argued that this latter thesis could not reasonably be viewed as the thesis against which Chisholm was concerned to argue, and that nomological behaviorism, while it seemed plausible, would require a detailed examination if we were to conclude that we are in fact warranted in accepting its truth. It was argued further that nomological behaviorism is not the thesis which Sellars defends under the heading of what he calls methodological behaviorism, and that this latter thesis seems warranted because its truth seems to follow from the truth of those features of Sellars' position which were defended in Chapter III. It was also argued that Sellars is concerned to reject the thesis of logical behaviorism, and that while his arguments on this point are somewhat obscure, they seem to reflect an appeal to the same sorts of considerations which were used, in section I, in making plausible our rejection of that thesis. Thus it was argued that although Sellars' claim that discourse about the intentional properties of intentional items is irreducible to descriptive discourse, in a way parallel to the irreducibility of normative discourse, to descriptive discourse was indefensible without far greater spelling out of the issues than Sellars has in fact offered us, nonetheless Sellars seems, in making these claims, to be making implicit appeal to the idea that two sentences cannot mean the same unless they have the same subject matter and are therefore about the same thing(s).

In section III, it is argued that a thesis of physicalism which is formulated in a way parallel to the thesis of logical behaviorism cannot be viewed as defensible, for roughly the same reasons that led us to reject the latter. For this reason, section III is devoted to the attempt to formulate a thesis of physicalism which is not susceptible to the sort of attack used against logical behaviorism in section I, and which at the same time seems to capture the intuitive appeal which physicalism has had for a number of philosophers. Using the notion of the nomological equivalence of two sentences, a revised thesis of physicalism is formulated as follows:

We may say all we might wish to say about the world by using no more than sentences belonging to the language of the physical sciences; and the sentences which, although belonging to the language of the physical sciences, we may use to say what we wish to say about psychological phenomena, will be nomologically equivalent to psychological sentences which do not belong to the language of the physical sciences.

It is argued that while those of Chisholm's claims which were found to be defensible are incompatible with such a revised thesis of physicalism if a strong notion of saying all we might wish to about some sort of phenomenon is invoked, there is no reason to suppose that a notion which is strong enough to yield such incompatibility need to be adopted in defending the revised thesis of physicalism. Thus it is argued that this revised thesis can be defended in spite of the incompatibility which would result from the acceptance of a sense of 'say all we might wish to say' which is unnecessarily strong.

The notion of nomological equivalence invoked in formulating the revised thesis of physicalism involved the idea that two sentences would be nomologically equivalent provided that (1) they are alike except that where one contains an expression denoting a given thing, the other contains a different expression denoting that thing; (2) the two expressions occur in purely referential position; and (3) there is a true identity statement asserting that what the two expressions denote is the same, and this identity statement is well established in roughly the way in which a well established law of nature is established. The nomological equivalence of psychological sentences with sentences of the physical sciences would therefore involve the acceptability of statements asserting that certain psychological expressions denote the same as what certain expressions of the physical sciences denote, and that these statements be established roughly as laws of nature are established. For this reason, it is argued that the revised thesis of physicalism cannot be true unless some appropriate version of the so-called identity thesis is true. (It is argued that the converse does not hold, because the identity thesis, unlike the revised thesis of physicalism, could be true even if the identity statements were, although true, not well established in the way in which natural laws are.) For this reason, sections IV, V and VI are devoted to a critical examination of the identity thesis.

Section IV begins by undertaking a formulation of the identity thesis, as follows:

There are no logical or philosophical considerations that rule out the possibility that it should be a result of future scientific research that psychological phenomena, in particular the occurrence of thought episodes, should be strictly identical with certain physical phenomena.

It is argued that there is nothing in any of Chisholm's claims which would conflict with the truth of the identity thesis, so formulated, and, in particular, that the objection to the identity thesis which involves claiming that the identifications in question are cross category identifications, is independent of considerations which would follow from the truth of Chisholm's claims. Thus although the merits of the so-called cross category objection are not discussed in section IV, it is argued that whether or not such an objection can be met, there is no conflict between the identity thesis and Chisholm's claims.

In spite of certain writings in which Sellars has espoused a rejection of what he has called the identity thesis, it is also argued in section IV that not only is there no conflict between the identity thesis as we have formulated it and Sellars' views, but that certain of Sellars' views on intentionality require that something like the identity thesis be true. It is further argued that if Sellars' views are to be accepted, then not only must we accept the identity thesis, but we must accept what we have called the identity thesis simpliciter, as opposed to the particular identity thesis; we must accept, that is, that there are no logical or philosophical reasons for rejecting the idea that certain general identifications can be framed of the sort envisaged by the identity thesis. For it is argued that a feature of Sellars' position is that if an item is to play a certain intentional role, then there must be empirical features which all and only items playing such a role exhibit, in virtue of which they play that role, and in virtue of which they can be picked out independently of the fact that they do play that role. Since this feature of Sellars' position has been discussed and defended in Chapter III, the fact that this feature of his position presupposes acceptance of the identity thesis as we have formulated it lends some plausibility to our accepting that thesis as true.

Sections V and VI are devoted to an examination of the plausibility of the identity thesis in connection with certain standard sorts of objections which have been raised against it. Thus section V discusses the merits of the objection that the identity thesis cannot be true because the identifications which are in question are ruled out on the basis of categorial considerations. In particular, section V is devoted to a consideration of the claim that the possibility of locating psychological items in place and time is restricted in a way in which locating physical items is not; and that this restriction provides reason for adopting the view that expressions applying to psychological items are categorially different from those applying to physical items. It is argued against this view that any restrictions of this sort do not constitute reason for inferring any such categorial difference of the relevant expressions, for such restrictions can be plausibly explained on the basis of limitations which depend on our ordinary ways of picking out psychological items, and are in principle eliminable, granting that we are not in principle restricted to picking out psychological phenomena in the ways we ordinarily do.

It is also argued in section V that the objection that we must be able to pick out psychological phenomena on the basis of our noticing some peculiarly psychological properties which they exhibit is founded on a mistake. For we need not have recourse to noticing any properties whatever when we pick out our own psychological states noninferentially, and even if there were peculiar properties exhibited by such phenomena which we noticed in the course of picking them out, there seems to be no reason for supposing them to be nonphysical properties. It is argued, finally, that it is possible to formulate expressions which apply to all and only psychological phenomena of a certain sort without having recourse to language which is peculiarly psychological. Section VI, then, is devoted to a particular objection to the identity thesis, formulated by Cornman, that argues that any attempt at providing topic neutral translations of psychological sentences must fail, and that since this is the case, we may identify psychological items with physical items, but we shall still be committed to the peculiarly psychological properties ascribed by psychological sentences. It is argued that while Cornman's claim that we cannot provide topic neutral translations of psychological sentences is correct, this in no way diminishes the plausibility and acceptability of the identity thesis.

Sections VII and VIII are chiefly devoted to a defense of the revised thesis of physicalism advanced first in section III. The idea is considered that there might be beings in the universe which behave in ways which warrant our describing them by means of the intentional language we use in describing one another. It is argued that the discovery of such beings, while not affecting the truth of the identity thesis, would lead us to revise any tentative identifications which we had made, of the sort envisaged by the identity thesis. It is argued that the revised thesis of physicalism, however, might be affected by the possibility of such a discovery in the following way. If we are to accept that the thesis of physicalism is a claim that we are warranted in replacing our current psychological discourse by nomologically equivalent discourse of the physical sciences, then it seems that the possibility of such a discovery provides reason for rejecting the revised thesis of physicalism. For if we replaced our psychological discourse by discourse which, on the basis of our empirical study of human beings, we take to be nomologically equivalent to psychological discourse, then, it is suggested, we would be unable accurately to formulate our discoveries about any beings which, although they behaved intentionally, had a radically different kind of composition and physiology from our own.

Section VII is chiefly devoted to setting out this objection to the revised thesis of physicalism, and section VIII is chiefly devoted to the attempt to formulate an adequate response to this objection. Thus it is argued that even if scientists, equipped only with a language in which psychological expressions had been dropped in favor of expressions which were their nomological equivalents given a certain body of empirical knowledge about, say, human beings, were to discover beings such as those envisaged, such scientists would not be prevented from making fairly straightforward linguistic accommodations to enable them to describe and explain the behavior of such beings. Since the argument advanced in section VII against the acceptability of the revised thesis of physicalism does not seem to bear scrutiny, and since it does not appear that there are any other arguments which would lead us to reject that thesis, it is argued that the revised thesis of physicalism is in fact acceptable. It is argued, however, that acceptance of that thesis must involve the proviso that because the linguistic equivalences which are in question are based on what we may envisage as laws of nature, the equivalences in question must not be understood as independent of possible future scientific discoveries, and any claims that we may replace psychological discourse by discourse which is, on the basis of a certain stage of empirical research, nomologically equivalent to current psychological discourse must be understood as subject to reversal or alteration in the face of yet further scientific research. Thus the revised thesis of physicalism must be accepted as relativized to certain stages of scientific and empirical inquiry.

In conclusion, it is argued that there are at least two possibilities as regards what position Sellars might adopt with respect to the revised thesis of physicalism. Since Sellars' espoused rejection of what he calls the identity thesis involves the claim that psychological discourse is in principle not replaceable by physicalistic discourse alone, it appears that this concern of Sellars' may, given the way the issues have been formulated above, be reasonably seen as a concern to reject the revised thesis of physicalism. But given the restriction on that thesis introduced in section VIII, that any replacement of psychological discourse by nomologically equivalent discourse of the physical sciences must be regarded as warranted only relative to a certain stage of scientific research, it is not clear why Sellars' would wish to reject this thesis. In particular, it does not seem that there is any reason for him to reject such a thesis basing a rejection on features of his position which have been discussed and defended. Thus it seems that if Sellars is to reject the thesis, he must make appeal to considerations beyond those discussed and defended in Chapters III and IV.

For this reason, the first possibility as regards Sellars' response to the revised thesis of physicalism would be to claim that it is indefensible, but it is not clear what sorts of considerations Sellars might appeal to in defending such a conviction. The second possibility is that Sellars' conviction that something like the revised thesis of physicalism is indefensible might be the result of his viewing the thesis as making a claim which is independent of any stage of scientific research. Thus the claim which Sellars is rejecting might be that a replacement of psychological discourse by nomologically equivalent discourse could (in principle) be made, such that no future scientific research could upset that replacement, or result in a partial reversal. In this case, however, Sellars would not be in disagreement with the thesis as it has been defended above, and the restriction and defense of that thesis offered in section VIII would, in that case, satisfy Sellars' objections. In any event, it does not seem clear that the reasons which lead Sellars to take the position he does on this issue have been made sufficiently precise to allow a definite conclusion to be drawn concerning Sellars' position regarding the revised thesis of physicalism.

Since, therefore, the revised thesis of physicalism seems compatible with Chisholm's claims, and since, even if Sellars would be inclined to reject that thesis, it is not clear what sorts of considerations he would appeal to in defending such a rejection, it does not seem that this thesis is incompatible with the positions adopted by either Chisholm or Sellars. Since, moreover, the thesis seems to be defensible on its own account, it is argued that we may regard it as acceptable.

[Table of Contents]--[Return to Main Page]


{26} Shaffer so argues in the first of two papers devoted to a discussion of Smart's version of the identity thesis. [Shaffer (1) 816-822] His arguments in the second of these papers are, in large measure, independent of this contention. [Shaffer (2)] [Back]

{27} Thus, in connection with his discussion of the identity thesis as applied to sensations (or so-called "raw feels"), Sellars introduces the notion of a core person, as the thing which, in the first instance may be said to have both raw feels and thoughts. Such a core person would be, on the basis of currently warranted projections of scientific results, the empirically known central and peripheral nervous system, devoid of the flesh, organs and bones which, along with the nervous system make up a biological person. [Sellars (14) 20] [Back]

{28} A similar point is made by Nagel, in discussing the identity thesis; he, however, speaks of identifying a person's having a certain sensation, and a person's body being in a certain physical state. [Nagel 343] What is important, however, is that precisely this sort of argument is justified by the way in which Sellars constructs his notion of a thought. [Back]

{29} This, in rough outline, is the argument presented by Shaffer [Shaffer (2) 163], in connection with the identity thesis as applied, however, to sensations. Since it is not clear why, if this argument is sound in connection with the identity thesis as applied to sensations it should not be likewise sound when the thesis is applied to thoughts, it is important to consider it here. The role played in the argument by the word 'notice', which is used above in the same way as it is used by Shaffer, is crucial. [Back]

{30} It might be argued that we can conclude that the causal properties relating thoughts to certain physical states of affairs should be viewed as physical properties just because they relate certain items to physical states of affairs. This argument does not seem to be compelling, for it begs the important question whether there could be non-physical items, having no physical non-causal properties, which are causally related to physical items having only physical properties. If one were inclined to give an affirmative answer to this question, then one might argue that the causal properties linking the first sort of item to the second sort might reasonably be classified as non-physical. What is being urged above is rather that if we accept that the properties exhibited by thoughts which make possible our noninferential reporting of them are simply causal properties of thoughts, then there seems no good reason to suppose that these (causal) properties are non-physical, as the objection to the identity thesis required they be. [Back]

{31} Once again, the objection seems just as applicable to the thesis as applied to thoughts. Cf. [Smart (4) 149-150]. [Back]

{32} It is important to note that, as Smart points out, the adequacy of this account "depends on the possibility of our being able to report that one thing is like another without being able to state the respect in which it is like." [Smart (4) 150] If this presupposition is understood as involving that we be able to report the similarity or dissimilarity of items of certain sorts without it being required that we notice the respect in which the items are similar or dissimilar, this presupposition of Smart's appears to represent an important disagreement with the presupposition of Shaffer's, noted above, which was crucial to his criticism of the identity thesis. [Back]

{33} It is important that it is not similarly clear that if such a paraphrase as Smart advances, as applied to sensations, is correct then the thesis is incorrect. For this would only follow if sentences reporting sensations are instances of what Chisholm would consider (or we should consider) as psychological sentences, that is, if sensations are psychological phenomena. But this, as we have noted, is a controversial view in itself. (Cf. fn., p. 257) It is important, in evaluating answers to this question, to consider the plausibility of a distinction between perceptions and sensations, perceptions understood as involving some conceptual component, that is, as involving some thought episode or propositional attitude, and sensations as not. In any event, the question of the incompatibility of Smart's move with Chisholm's thesis is more controversial with the move as applied to sensations than with the move as applied to thoughts. [Back]

{34} This line of argument is adopted by Rorty in [Rorty (1)]. In particular, Rorty argues that since a similar limitation is encountered in the attempt to translate reports of supernatural entities, for example, reports of succubi, into language which is topic neutral with regard to a natural-supernatural dualism, and since this limitation does not appear to provide good reason for accepting such a dualism, such a limitation with respect to paraphrasing sensation reports cannot provide, by itself, good reason for accepting a psychophysical dualism. [Rorty (1) 17-18] The argument which Rorty is rejecting, from the limitation on any series of topic neutral paraphrases to the requirement that we accept some dualism, appears to rest on the idea that if we cannot eliminate the special sort of language required to characterize a particular sort of thing which we want science to take account of, then we are forced to adopt the relevant dualism. But the disanalogy between succubus reports and sensation reports becomes evident if we consider this assumption. For while we do suppose that science should account for succubus reports and sensations reports alike, we do not, as we do with sensations themselves, suppose that science should deal with succubi. For we believe that science has shown that there are no such things. Thus, it can be argued, the requirement that science deal with succubus reports does not, as the requirement that science deal with sensation reports does, result in the necessity for dealing with the items mentioned in the report. For a counter to this sort of argument, however, cf. [Rorty (2)]. Rorty's response in this paper involves what seems, however, to be a modified version of the identity thesis as advanced by Smart. [Back]

{35} A similar point is urged by Putnam in [Putnam (7) 190-194]. Putnam's argument differs in two important respects, however. (a) He takes the identity thesis to involve identification of properties, such as the property of having the thought that-p and the property of being in a certain physically specifiable state; and (b) his argument rests on a form of the identity thesis as applied to Turing Machines, with the intention of showing that the way in which the thesis goes wrong in this case sheds light on how it goes wrong in the case of persons. Putnam's point resembles the one presented above in that he argues that any identification of "psychological" states of Turing Machines (states, that is, which are suitably analogous to human psychological states) with any sort of physically specifiable state of Turing Machines runs up against the difficulty that Turing Machines are, by hypothesis, capable of being physically realized in an indefinitely large number of different ways. [Putnam (7) 187] It is not clear why Putnam believes that if the identity thesis is not to be restricted to particular identifications, then it must involve the identification of properties.

The contention that an acceptable version of the identity thesis must involve identification of properties is also advanced, although for rather different reasons, by Brandt and Kim. [Brandt and Kim 518] Their argument for this contention appears to rest on the idea that unless the identity thesis involves such identifications, then there is no way to distinguish the claims which the identity thesis asserts as possibly true from the claim that there is a pervasive correlation of the two sorts of phenomena. [Brandt and Kim 518] Since Brandt and Kim do not go into any detail in presenting this argument, it seems difficult to evaluate whether such a conclusion could be established in the way they hope to do so. [Back]

{36} It will be remembered that in Chapter I the notion of a sentence being expendable in favor of another was invoked in the course of discussing Chisholm's thesis in "Sentences about Believing." Towards the beginning of section II of that chapter, it was argued that certain passages in Chisholm's article seemed to warrant our accepting the view that a sentence is expendable in favor of another if, and only if, the two sentences are equivalent in meaning. Given this understanding of the notion of expendability, that notion is clearly not the same as the notion invoked here of being able to dispense with a sentence in favor of another. For whatever condition we should wish to impose if we are to be able to dispense with one sentence in favor of another will, as we have seen, be weaker than the meaning-equivalence of the two sentences.

Since, as it was pointed out in section II of Chapter I, the particular understanding of the notion of expendability which was adopted there was not relevant to the issues discussed in that chapter, the current notion of being able to dispense with a sentence in favor of another could have been used instead of the notion of expendability, as we construed it there, without altering any of the conclusions or arguments of Chapter I. Our reason for adopting the construal we did rested only on a consideration of the passages in "Sentences about Believing" which seemed to justify such a reading.

A reasonable reading of the notion of being able to dispense with one sentence in favor of another would require that we can do at least as well with the second sentence as we can with the first in explaining, describing, reporting and predicting the phenomena which the first sentence is intended to explain, describe, report and predict. (Such a notion seems implicitly to be invoked in [Rorty (2) 35-47].) On this account, it seems clear that we need not require that two sentences be equivalent in meaning in order that we should be justified in dispensing with one in favor of the other. [Back]

{37} The idea of our dispensing with (20) in favor of (21) on the basis of their (current and presumably well-established) nomological equivalence is similar to the idea, advanced by Rorty, of what he calls a "disappearance" form of the identity thesis. [Rorty (2) 33] Rorty's version of the identity thesis involves (a) that certain results of empirical science might entitle us to account for all occurrences of certain sorts of mental reports (both inferential and noninferential) as being really reports of certain neurophysiological occurrences, and thus (b) that we would thereby be entitled to say, the monstrous inconvenience of such a move notwithstanding, that what we used to call mental phenomena of the sort in question are really nothing but neurophysiological occurrences of the requisite sort. [Rorty (2) 32-39] Aside from the fact that Rorty's version of the identity thesis is restricted to non-intentional items such as sensations, there seems to be the following difference between what he is defending and the thesis of physicalism understood as involving the truth of sentence (19). Whereas Rorty's thesis involves a claim to the effect that we might come to claim that sensations do not exist, in the sense in which we now claim that witches do not exist, the thesis of physicalism which has been advanced has been framed deliberately to skirt that question. The strong resemblance of the two theses results from the insistence of both that sentences like (21) could come to perform all the linguistic functions currently performed both by such sentences and by sentences such as (20). Cf. [Rorty (2) 39-49].

The rationale for skirting the question which is crucial to Rorty's thesis is the following. Rorty's position seems to involve the belief that there is a reasonable sense of 'Xs are nothing but Ys.' in which this sentence is compatible with both 'There are no Xs.' and 'There are Ys.'. If 'Xs are nothing but Ys.' is held to be an instance of a generalized strict identification, then it seems that such a sentence entails that there are Xs if, and only if, there are Ys. (For example, one might construe the sentence as reconstructible as '(x)(x is an X if, and only if, it is a Y)', and this in turn entails that '(Ex)(x is an X)' and '(Ex)(x is a Y)' entail each other. A weaker construal, perhaps peculiarly suggested by Rorty's example involving witches, might be '(x)(x is an X only if x is a Y)'. Since this is entailed by there being no Xs, it is compatible with both of the existential claims mentioned above. But this weakened version seems to deprive Rorty's thesis of the strength which, it seems, he would wish it to have. A third reconstruction of 'Xs are nothing but Ys.' might be 'What have been called (incorrectly so) Xs are, in fact, identical to Ys.'. This reconstruction seems to be that suggested by the arguments offered by Rorty. But this reconstruction makes no reference to items which actually are Xs; for it is no more than a statement about the items which have been believed to be Xs. Thus our revised thesis of physicalism seems to be a satisfactory way of avoiding the problems involved in providing a satisfactory reconstruction of 'Xs are nothing but Ys.'. [Back]

{38} A related argument would thus rest upon the following sort of consideration. We may imagine that we have accepted (7) as nomologically well-established, and that we have effectively dispensed with (20) in favor of (21). We may further imagine that prior to our encounter with them, our imagined silicon beings have framed corresponding general identity statements, done the research required to establish them firmly on the basis of their current knowledge, and thereby dispensed with their thought reports in favor of certain physical-state reports. Our encounter with them may be imagined as resulting in the need for framing some means of translating their (possibly radically divergent) language into ours, and vice versa. One result of this attempt would seem to be that, in the absence of sentences applicable in either language on the basis of considerations such as those involved in Jones' theory, we should be unable to avoid translating our reports of being in certain neurophysiological states into their reports of being in certain, by hypothesis radically different, physical states. Unless some means for overcoming this difficulty can be offered, so this related argument concludes, our need to be able to frame such translations would rule out our being warranted in dispensing with (20) in favor of (21). [Back]

{39} It does not seem at all clear that on the basis of Chisholm's thesis of intentionality, there is any conflict between his views and the revised thesis of physicalism. For if this revised thesis rests on the notion of nomological equivalence while Chisholm's thesis rests on the notion of meaning-equivalence, it is clear that there is no incompatibility between (18) and Chisholm's thesis. Chisholm's views do not, moreover, appear to involve anything beyond this thesis which would clearly represent a disavowal of (19), save the idea, somewhat imprecisely expressed, that physical sentences cannot, if Chisholm's thesis is correct, be used to describe psychological phenomena. Cf. [Chisholm and Sellars 533], where Chisholm argues that because "thoughts are a 'source of intentionality'," it follows that "thoughts are peculiar in that they have an important characteristic which nothing else in the world has." On this basis, it seems, Chisholm might argue that no nonpsychological sentence can fully describe thoughts. Since, as has been remarked, Chisholm has earlier in his correspondence suggested that 'describe' stands in need of further explication [Chisholm and Sellars 529], it does not seem reasonable to construe Chisholm's views as conflicting with the revised thesis on physicalism in the absence of such explication. [Back]

{40} We may imagine, for convenience in exposition, that we may ignore the difficulties of translating their language into ours mentioned in the first footnote on p. 314. [Back]

{41} The considerations on which this proposed answer rest seem similar to those suggested by Feyerabend. (Cf. [Feyerabend (3) 295-296].) Feyerabend appears to be arguing that the idea that there is anything privileged in the apparent categorical distinction between discourse about physical phenomena and discourse about psychological phenomena should be construed as at least in part a matter of the current state of scientific (and informally scientific) knowledge of the world. Thus, the proposed answer rests on the idea that even if the independent identifiability and reidentifiability of psychological and physical phenomena is dropped as a result of certain scientific discoveries and developments, it can, in effect, be resuscitated on the basis of still further ones. Cf. also [Feyerabend (1) 90-91] and [Feyerabend (4) 256-259, fn. 26]. [Back]

{42} Even if our scientists, speaking a language devoid of peculiarly psychological expressions, did adopt an alternative like (b) above, we could imagine their being able effectively to deal with the phenomena. For as we have seen, it remains open in the face of data concerning silicon beings as viewed by speakers of our language, to discover descriptions of our physical states and their physical states (other than simply disjunctive descriptions) which could warrant accepting identifications of psychological with physical phenomena for both human and silicon beings. Thus a scientist faced with such data but speaking a language devoid of peculiarly psychological expressions could simply search for physical descriptions under which to subsume the appropriate physical states of both human and silicon beings. To do so, however, would require his selecting certain sorts of bases for the applicability of sentences like (21) (to both sorts of beings) which are, in fact, those concerned with the intentional roles played by the physical states. To do this, however, would be in effect to have reintroduced psychological terminology into the language, even if only heuristically and temporarily. [Back]

{43} That certain beings like the silicon beings we have imagined might be discovered which did not require such reverse modifications is, of course, possible. For scientists might be able to correlate our neurophysiological functioning given our general makeup with the physical functional of silicon beings, given their general makeup, in such a way as to obviate the necessity of making such correlations through the mediation of the intentional roles played by certain physical states. It seems clear that this possibility is, like the others we have projected a matter of the then current state of scientific knowledge. What is claimed here is simply that such beings might be discovered which, given the then current state of our knowledge (and of their knowledge), would require reference to intentional roles played by physical states independent of the determinate factual character of such states, if we are adequately to describe the psychological behavior and states of such beings. [Back]

{44} Cf. [Sellars (11) 190 ff.], where Sellars argues that the framework of common sense must be retained because of certain methodological considerations, in spite of its acknowledged inadequacy in providing an accurate account of the world. In particular, Sellars seems explicit that it is not on the basis of any ontological considerations, say, concerning whether the mental items presupposed by the common sense framework really do exist or not--as against only items presupposed by a certain theory--that he disagrees with the position advanced by Feyerabend, and presumably the positions advanced by Rorty and in the text above. [Sellars (11) 190] It is somewhat difficult to evaluate the force of Sellars' argument on this point, because it seems to rest on the contentions that (1) the common sense framework has only what Sellars calls an internal subject matter, while scientific theories involve both an internal subject matter and the external subject matter provided by the common sense framework [Sellars (11) 179-180]; and (2) that the common sense framework, unlike the frameworks of physical theories, involves principles of a categorial sort in addition to substantive beliefs about the world [Sellars (11) 172] Thus the so-called methodological considerations which, on Sellars' view, lead us to need to retain the common sense framework in the course of developing scientific theory seem to rest on these features on the common sense framework which are not shared by scientific frameworks. [Sellars (11) 189-190] Thus Sellars concludes that unless a scientific framework can be provided which reflects these features of the common sense framework, the latter is methodologically--although neither ontologically nor epistemologically--indispensable.

Feyerabend has criticized the first of these two points, arguing that general relativity provides a framework which, like that of common sense, has no external subject matter. [Feyerabend (4) 248] Since it does not seem clear just what distinction Sellars is invoking in making this first point, it is difficult to evaluate the force either of his position or of Feyerabend's rejoinder. As for the second point, while it seems plausible to suppose that the idea that the framework of common sense contains certain categorial principles would be where Sellars' objection to the revised thesis of physicalism would stand or fall, it is again not clear what is involved in such categorial principles, nor why they should be seen, granted they exist, as indispensable. For it is unclear that it is possible to draw any illuminating distinction between categorial principles embodied in a body of knowledge and ordinary beliefs contained therein. Thus the disanalogies between scientific theories and the framework of common sense which lead Sellars to search for the "core of truth in the concept of 'the observation framework'" [Sellars (11) 187], while suggestive, do not seem to be firmly grounded in argument or sufficiently clearly stated to allow for any useful evaluation. [Back]

{45} That Sellars' views are perhaps not at odds with the conclusions reached on the basis of the foregoing arguments may be suggested by his writing, in a footnote added to the published version of one of his letters to Chisholm, of a sense in which he "would wish to maintain that the world can 'in principle' be described without mentioning either the meaning of expressions or the aboutness of thoughts." [Chisholm and Sellars 539] The only illumination that he offers concerning the claim he has in mind here is that it is parallel to a claim that there is a "sense in which the world can 'in principle' be described without the use of either prescriptive or modal expressions," and a reference to certain passages of his article "Counterfactuals, Dispositions, and the Causal Modalities," in which he says that this latter claim is discussed. For this reason, it remains obscure as to whether such remarks might accurately be used to support the idea that Sellars is not in significant disagreement with the conclusions reached above. [Back]

[Table of Contents]--[Return to Main Page]