Andrew Chrucky, Critique of Wilfrid Sellars' Materialism, 1990 |
CHAPTER 6
PHILOSOPHICAL PSYCHOLOGY
A defense of Materialism should give an account of mental (or psychological) phenomena. And in order to give such an account, these phenomena must first be identified. But this is complicated by the number of criteria offered for their identification. {1}
To begin with, a distinction can be made between ontological and epistemological criteria. Of the ontological criteria, there are Brentano's criterion of intentionality {2} and Descartes' criterion of non-extensionality or non-spatiality for mental phenomena, and Chisholm's criteria for psychological sentences. Of the epistemological criteria, there is "direct awareness" which has as instances such phenomena as privileged access, incorrigible awareness, indubitable awareness, private awareness, and so on.
Sellars' approach is epistemological. However, it ties in to the ontology via the necessary condition for the possibility of knowing. This ontological condition is the possession of a conventional language. Given this approach, Sellars can agree with Brentano's thesis that the mark of the mental is intentionality. But since conceptual intentionality requires language, Sellars can also agree with Chisholm that conceptual intentionality is to be found in language. However, such an agreement with Chisholm is superficial. Chisholm searches for intentionality in kinds of sentences; whereas Sellars finds intentionality in the very function of sentences, as will be explained.
Given Sellars' assumption of the primacy of a public language for the conceptual awareness of any kind, this creates a problem of accounting for how introspective awareness is possible. Specifically, Sellars has to account for our introspective awareness of sense impressions and thoughts. Although the strategy for accounting for introspection is similar for both sense impressions and thoughts, there is a difficulty in lumping them together. This stems from the differences in the concepts of sense impressions and thoughts. In this chapter I will concentrate on the concept of thoughts, their intentionality, and how awareness of (private) thoughts is possible. Sense impressions will be covered in the chapter on perception.
I. TAXONOMY OF PHILOSOPHICAL PSYCHOLOGY
I will begin by locating Sellars' views on philosophical psychology among those popularly held today. This exercise is valuable for taking stock of the range of plausible alternatives, for assessing the merit of Sellars' position in relation to them, and for suggesting as yet untried alternatives.
Contemporary debate about philosophical psychology is split along several lines. The major divisions, as I see it, are between what are called these days peripheralists and centralists. Peripheralists limit their investigation to either introspection and/or gross publicly observable phenomena. Their distinguishing trait is in either the exclusion of theoretical claims altogether (e. g., later Wittgenstein, "bracketing" phenomenologists) or the exclusion of irreducible theoretical claims (e. g., Skinner, logical positivists). Centralists, by contrast, extend their sphere of interest to include neurophysiological variables and irreducible theories. Now if we abstract from the peripheralists and centralists the introspective approach, we will be left with what are called by Sellars Philosophical (Logical) and Methodological (Scientific) Behaviorists{3}--a demarcation between those who allow only reducible theoretical terms and those who allow also irreducible theoretical terms, respectively.
Another division which can be abstracted from the peripheralist-centralist dichotomy is between "bottom-up" and "top-down" approaches. The top is the psychology of common sense or, as it is currently called, "folk psychology" (roughly, a component of Sellars' Manifest Image). Its distinctive trait is in explaining human behavior in terms of intentional states, such as desires and beliefs. The bottom is neuroscience. Working from the bottom means that one tries to connect the laws of neuroscience with the laws of higher level functioning. Working from the top means working in the opposite direction.
Cognitive science{4} occupies a middle position in this debate. It posits a representational system which mediates between the stimulus and response, and which may be constituted by some (as yet unknown) neurophysiological structures. The claim of those working in the field of artificial intelligence is that these representational structures can be modeled by a Turing machine such as the computer.{5}
Within these lines of division, I would call Sellars a cognitive scientist. This is so because he posits thoughts as irreducible theoretical states and uses the computer as a model.{6}
Once we start to talk seriously about representational systems we can get bogged down over the word 'representation', especially if we want to talk about the 'cognitive' life of a planarian (a type of worm), the 'cognitions' of a plant, or even the 'cognitions' of a computer. If we want a common term for these activities or states, 'cognition' or 'representation' seems to connote too much sophistication; we would do better to switch talk to 'information processing', as is commonly done.
A great deal of current debate is about the nature of the posited information processing. These debates can be viewed from the perspective of (i) the types of information processing subsystems, (ii) which information processing systems are innate, (iii) the extent to which these information processing systems are analogous to a public (conventional) language, and (iv) the ontological reality of propositional attitudes.
Concerning (i), it is clear that Sellars would distinguish between sensorial information processing systems, innate (instinctual) animal representational system, a learned animal representational system, and a learned human conceptual system (public language). He would also obviously recognize an information storage system or systems (memory) and some kinds of information retrieval, creation, and anticipation systems.
Within these informational systems Sellars distinguishes sensorial states (sensa, qualia) and functional states. The cogency of such a dualistic stance is one of the hotly disputed issues. Some theorists try to ignore or eliminate talk of sensorial states altogether and make functional states be exhaustive of all information processing. Such a view is called functionalism, and is probably the dominant view in cognitive science.
Ned Block, in his review and critique of functionalist theories,{7} makes a number of distinctions within functionalist theories. The most basic, it seems to me, is between what he calls Functionalism and Psychofunctionalism.{8} The former he describes as an a priori functionalism which is related to the logical behaviorist's endeavor to define mental terms by physicalist terms. The difference is that the physicalistic terms in Functionalism become physicalistic functional terms. Psychofunctionalism, by contrast, is described as empirical functionalism in which mental states are understood to be functional states posited as scientific hypotheses. Using Block's classification, Sellars is a Psychofunctionalist about thoughts.
Turning now to the innate perspective (ii), and considering only the status of thoughts, the extremes are Jerry Fodor's posit of an innate language of thought, the position of Noam Chomsky who posits an innate universal "grammar", and the position of Sellars for whom (conceptual) thoughts are not innate but learned.
As concerns animal representational systems (which humans presumably share with animals), Sellars considers them to be innate capacities.
As concerns the sense organs, the positions of James Gibson, Richard Gregory, and Patricia Churchland are that they contain innate sensorial computational devices. Here one can make the Aristotelian distinction between 'proper sensibles' and the 'common sensibles'. It seems that Sellars could identify the common sensibles as components of animal representation, and it seems that Sellars would not have any objections to describing them as analogically intentional.
Turning finally to the problem of how analogous the information processing systems are to a public language (iii), one extreme position for understanding thoughts is to understand conceptual thoughts as isomorphic to sentences. This view is found in most functionalist analyses of thoughts. It is the view of Sellars, Geach, Aune, Castañeda, Rosenberg, Fodor, Chomsky, Clark, Putnam, and Armstrong, among others.
As far as I am aware, there are only a few writers who have taken the option of making non-verbal human thoughts less than isomorphic to sentences; among these are Paul and Patricia Churchland{9} This is also where I will differ from Sellars. The position which I am proposing is an extreme one, namely, that non-verbal human thoughts are limited to an animal representational system. To offset a possible ambiguity in the notion of thought, what I am proposing is not to be confused with having or expressing thoughts by talking to others or to oneself either loudly or silently, and it is not to be confused with the idea of thinking as in imagining or dreaming that one is talking about the passing show. In these cases it seems to me one is talking and using language, and hence the thinking is conceptual. I propose that when there is no talk of any sort under any guise, then one is in a non-conceptual animal representation system frame of mind. It may also be the case that all our unconscious thoughts are of this nature as well. I have two reasons for introducing this hypothesis. The first is Occam's principle. If we can account for mental life by positing at most only animal representations, why introduce conceptual thoughts? The second reason stems from reflecting on the difficulty (examined a little later) which Sellars' account has for learning to introspect conceptual thoughts from a perspective of Methodological Behaviorism.
Sellars' own position is a middle one. The posited states are called thoughts and they are claimed to be analogous to sentences of a public language. However Sellars is cautious here. The analogy is strong in the case of adult mature speakers of a public language and weakens as we descend down the chain of living creatures from language learners to brutes. The dividing line between human beings and other creatures, as was discussed in the previous chapter, is in the possession of a (broadly understood) logical vocabulary by the former and its absence in the latter.
Turning to the problem of ontological commitment to propositional attitudes (iv), this is the problem of determining whether folk psychology has an ontologically irreducible reality or not, and why. An excellent schematism of where the major philosophers stand on this issue is provided by Jerry Fodor's article "Fodor's Guide to Mental Representation: The Intelligent Auntie's Vade-Mecum."{10} He divides philosophers into two major divisions: in one division are the realists about propositional attitudes; in the other are the anti-realists. The anti-realists are then subdivided as to whether they are instrumentalists or not about the attitudes. Non-instrumentalists are further subdivided into those that are functionalists or not about the attitudes.
Turning to the realist side of this division, the subdivision is into functionalists and non-functionalists. The functionalists are subdivided on the basis of whether they take the propositional attitudes to be monadic or relational. Each of these branches is then further subdivided according to whether or not the propositional attitudes are given a non-denotational semantical functional role (FR) or not.
Here is Fodor's outline with the names of major philosophers appended to the position Fodor ascribes to them.
- propositional attitudes
- realist
- functionalists
- attitudes are monadic
- accepts FR semantics (Loar, Burge?, Stalnaker?)
- rejects FR semantics
- attitudes are relational
- accepts FR semantics (Harman, Block, Sellars, McGinn, Lycan)
- rejects FR semantics (Fodor)
- non-functionalists (Searle, Dreyfus, Haugeland)
- anti-realist
- instrumentalists (Dennett)
- non-instrumentalists
- functionalists (Stitch)
- non-functionalists (Churchlands)
I find this classification very valuable for orienting oneself in the literature on propositional attitudes. And, for all I know, Fodor's placements of most philosophers may be accurate. Here, however, I will focus on Fodor's placement of Sellars. According to this classification, Fodor claims that Sellars' stance on propositional attitudes is the following. He is a Realist, Functionalist, Relationalist, and an FR Semanticist.
I disagree. On my understanding, Sellars is an Anti-Realist, Functionalist, Non-Relationalist, and an FR Semanticist.
Sellars in an Anti-Realist because he recognizes the existence of sentence tokens only. Talk of propositions is a way of referring to sentence token roles or functions in the linguistic economy.{11}
As to the logical form of propositional attitudes, Sellars adopts an adverbial reconstruction. Let me explain. Suppose we have the sentence 'Fodor believes that Sellars has a relational analysis of propositional attitudes'. We can distinguish the person 'S', the psychological verb 'V' (e.g., 'believes that'), and the propositional construction 'p'. A relational reconstruction would have it that 'believes that' is a relational verb relating the person to the proposition, i. e., S-V-p, which has the form 'xRy'. Sellars does not accept this analysis. Rather than being a relation, believing is a state of a person, and the proposition is an adverbial specification of this state; so that the correct logical reconstruction is generically 'SVs' and specifically 'SV(p)s', which is to say that S is Ving in the p manner, or 'SVs-p-ly'.{12}
II. HOW IS INTROSPECTION POSSIBLE?
Sellars' intent in speculating about cognitive psychology is to provide an account of how it is logically and empirically possible to have the concepts of thoughts and sensations as inner episodes (to which we have a privileged access) on the basis of what Sellars calls Methodological Behaviorism.
This formulation to be understood requires a knowledge of two contrasting theses. The first contrast is that between behaviorism as restricting itself to the domain of publicly verifiable phenomena; in which case it contrasts with an introspective approach which purports to report about private phenomena. The second contrast is between an interpretation of theoretical terms which is reductive and an interpretation which is non-reductive.
Take the behaviorism of B. F. Skinner. As far as I understand him, his behaviorism is programmatic in the sense that he wants to see to what extent psychology can be done by considering only public phenomena and those theoretical states which are entirely definable in terms of public phenomena. Now whether in fact his own writings are faithful to such a program is another question; specifically: Does Skinner use only such variables which are reducible to things which are publicly ascertainable? It seems to me that he does not. And some semblance of verisimilitude enjoyed by Skinner's analysis is due in part to his utilization of the very terms he wants abjured.{13}
As to the status of private phenomena, Skinner is simply silent on their status while doing psychology. When he does make reflective comments on private phenomena he seems to be an epiphenomenalist.{14}
To understand the program of behaviorism requires an understanding of the alternative eschewed. This is the alternative of methodological solipsism{15} (it could also be called 'radical introspectionism', 'Cartesianism', or 'Mentalism'). Such a view presupposes that we have access to the media of private languages by which to express whatever is known through (private) perception and introspection.
By contrast, behaviorism is a position which assumes the methodological and epistemic priority of a learned public language.
Assuming that these two programs are viable alternatives, the problems that face these programs are reciprocal. For the methodological solipsist the problem is to explain the possibility of knowing an external, public world; for the methodological behaviorist the problem is to explain how private introspective knowledge is possible.
In giving his behavioristic account of how introspective knowledge is possible, Sellars adopts a genetic approach similar to the classical contract theories in politics.{16} He posits a mythical model in which it is shown by a thought experiment how it is possible to go sequentially from a public language which has no reports of private episodes to a language which does. The 'state of nature' in this myth is posited to be a community of Ryleans whose descriptive language is about publicly observable things and properties. With the resources of the Rylean language (i.e., Behaviorese or Verbal Behaviorism--it could also be described as Carnap's physicalistic thing-language), Sellars tries to introduce a theoretical language of private thoughts and sensations.
From the perspective I have given, Sellars prefers one kind of program over another, and it may seem that it is also possible to give an alternative account of public knowledge on the basis of introspective knowledge. Sellars thinks that this alternative program begs the question. His objection is that any conceptual ordering of experiences presupposes the possession of concepts. And the possession of concepts--as has been argued in the previous chapter--requires a conventional language. But presumably any account on the basis of introspective knowledge is an account of conceptual knowledge, and such an account begs the question by presupposing the very thing it is to explain, namely, the possession of conceptual knowledge. To put it otherwise, methodological solipsism must assume that there is a private language which is used to explain the existence of a public language. Such a view, for example, is attributed to St. Augustine by Wittgenstein:
Augustine describes the learning of human languages as if the child came into a strange country and did not understand the language of the country; that is, as if it already had a language, only not this one. Or again, as if the child could already think, only not yet speak. And 'think' would here mean something like 'talk to itself'.{17}J. Fodor quotes the above passage from Wittgenstein and adds: ". . . Augustine was precisely and demonstrably right . . ." {18} Fodor's thesis is that people possess an innate language of thought.
As formulated by Fodor, his thesis is too strong both for Sellars and me. His thesis relies on two objectionable assumptions. The first assumption is that the private language contains a logical vocabulary. The second assumption is that there is an equivalence between the predicates of any public (natural) language and the private language of thought, such that "for every predicate in the natural language it must be possible to express a coextensive predicate in the internal code [i. e., private language]."{19} This can be put as the acceptance of the formula:
'Py' (in L) is true iff Gx (in RS). ['L' stands for some public language and 'RS' for the innate representational system]The first assumption that a pre-linguistic creature possesses a logical vocabulary is rejected by Sellars for reasons given in the previous chapter. The second assumption has been criticized for implying the paradoxical conclusion that a public language cannot formulate predicates not translatable to the predicates of the private language. As Patricia Churchland has pointed out, Fodor's position commits him to the thesis that theoretical predicates of a public language are reducible to the primitive predicates of the private language. To which she replies:
. . . there is not the slightest reason to believe that theoretical predicates are reducible to elementary predicates . . . Quite the reverse; there is good reason to believe that theoretical predicates are not so reducible.{20}It seems to me that Fodor's program can be amended to sidestep these two criticisms. First, it is possible to substitute for a private language which contains a logical vocabulary a representational system which does not. Second, Fodor's equivalence thesis can be either weakened to an implication, such that
If 'Py' (in L) is true, then Gx (in RS).Or the original equivalence could be retained but the predicates involved could be restricted to second-order predicates, such as, for example, the relations of transitivity, symmetry, reflexivity. In either case, we can leave the relation between 'Py' and Gx unspecified, but allow for a broad relation of analogy between them. However, by withholding a logical vocabulary from the private language, we can grant that the (austere) private language is a necessary condition for a public language, but not sufficient; and therefore cannot account for the existence of a public language.
If we assume that pre-linguistic introspection occurs, then Sellars' problem could be couched as the problem of meshing this pre-linguistic awareness with publicly available concepts. So the question of conceptual introspection takes two forms, depending on whether or not an allowance is made for pre- linguistic introspection. Sellars writes in his classical "Empiricism and the Philosophy of Mind" as if the alternative of pre-linguistic awareness were not available. But this is in variance with his admission of animal representational systems in "Mental Events."{21} An alternative approach to Sellars' which is behavioristic in a broad sense is one which posits more innate factors than are usually used by behaviorists. Such a position for Sellars would be coherent; but which if not restricted, could lead to an appearance of circularity in explanation. Sellars has in mind the postulation of an innate language-like structure to account for the possibility of learning a public language. His position seems to be that if there is no better way to account for language acquisition than by positing an innate language-like structure, then this is what should be done:
There is nothing self-contradictory about the idea of an innate language-like structure (Mentalese), and there might just be no other way of explaining language acquisition.{22}The question we should raise in following Sellars' behavioristic account of the possibility of introspective knowledge is: first, to find out what constitutes Methodological Behaviorism; and, second, whether Sellars' specific account is physically possible. As is suggested by the reference to innate structures, Sellars' conception of Methodological Behaviorism is quite liberal. I would agree with Quine's remark that "His [Sellars'] moderate behaviorism is exactly to my taste."{23} Nonetheless, I will argue that Sellars' peculiar strategy of accounting for language acquisition, and specifically the acquisition of introspective knowledge does not show how it is possible.
A. THREE STAGES FOR THE POSSIBILITY OF INTROSPECTION
Our conceptual abilities to introspect our thoughts and sensorial states (which are included in the Manifest Image) are a problem for Sellars because of his behavioristic foundations. He must provide an account of the possibility of introspection on the basis of publicly accessible data. Sellars' approach is intended to be a reconciliation between the classical view of thoughts and sensations as inner episodes (to which we have a privileged access) and a behavioristic public foundation. To do this, he adopts an approach similar to the classical political contract theories by constructing a possible genetic account beginning with a (fictitious) Rylean community. The language of this community is based entirely on intersubjectively available data. It has the conceptual vocabulary for talking about physical objects and persons. But it lacks the vocabulary for introspective reports. Sellars believes that the resources available in Behaviorese are sufficient to account for the possibility of introspection. To provide an account of how it is logically and empirically possible to have the concepts of thoughts and sensations as inner episodes (to which we have a privileged access) on the grounds of Methodological Behaviorism, Sellars conducts a thought experiment to show that it is possible to go sequentially from a public language which has no reports of private episodes to a language which does. Sellars' behavioristic account has, or should have, three stages. In the first stage he posits a community which speaks Behaviorese: the Ryleans. However, this first stage should also have an account of how Behaviorese is possible. But Sellars does not present such an account explicitly; it has to be reconstructed from his writings. In the second stage he introduces semantical discourse from the resources of Behaviorese; while in the third stage he gives an account of the language of introspection of thoughts and sensations using the theoretical resources of Methodological Behaviorism.
I will describe and criticize these stages. Sellars, unfortunately does not argue for the first stage, at least not explicitly. What is needed for the first stage is an account of how people could have learned a Behaviorese language, and whether such a language is possible. At best we get a thesis about this: language is a product of social evolution through natural selection. The mechanism by which this takes place is evidently some form of conditioning. But this is not clearly explained.{24} Sellars' arguments are concentrated on the defense of his second and third stages.
The conclusion I reach about the first stage is that Sellars is implicitly committed to the ontological priority not only of rs-thoughts in animals and people, which has been granted, but also possibly of ur-concepts in pre-linguistic people and possibly in animals. As to the second and third stages, my conclusion is that although Methodological Behaviorism is adequate for these stages, Sellars has failed to show how Methodological Behaviorism gives an account of introspective language, and that another explanation along behavioristic lines is available. However, there is one aspect of Sellars' discussion which is insightful, namely the claim that thoughts are to be understood by analogy to overt speech.{25} But that insight is compatible with alternative psychological accounts.
1. 1ST STAGE: BEHAVIORESE
a. PRINCIPLES OF BEHAVIORISM
The first task is to get clear about what Behaviorism claims. Ausonio Marras has listed the following principles of a behavioristic methodology, which probably reflects B. F. Skinner's approach:
(i) The basic laws of conditioning are the laws of reinforcement (law of effect) and/or the classical laws of association (contiguity and frequency, stimulus generalization and stimulus discrimination, etc.)
(ii) S-R connections are essentially functional, i.e., they have the form R=F(S), where this is a 'gross function' usually derived from elementary functions relating stimuli and responses to other 'environmental' variables (state of deprivation, histories of reinforcement, etc.)
(iii) An S-R connection is acquired only if the probability of a response Rj given Si (under conditions C) is greater than .5.
(iv) Si and Rj are physical states describable in the observational (molar) language of the experimenter.
(v) Non- observationally definable states of the organism may not be postulated in a functional analysis of behavior. {26}That a language can be understood and explained on behavioristic assumptions is, of course, vehemently denied by Noam Chomsky who, in his review of Skinner's Verbal Behavior presented a series of plausible counter-arguments. His main objection was that Skinner had rejected theory construction as a program, and, furthermore, had adopted a crude version of the verifiability theory of empirical meaning, such that all legitimate theoretical terms had to be definitionally reduced to empirical terms.
Chomsky believes that the grammar of a language is too complex a phenomena to be understood without a theory, and, in addition, rejects Skinner's theory of empirical meaning. So, inasmuch as Skinner's behaviorism rests on a methodological rejection of theory construction, on the one hand, and a reductionist theory of empirical meaning, on the other, his behavioristic foundations for language are inadequate. Specifically, an explanation of linguistic abilities, according to Chomsky, must include a theoretical postulation in the language user of an innate knowledge of a universal grammar--a knowledge which is not derivable from any finite learning history. On his theory, particular grammars of the natural languages simply utilize some of the resources of the universal grammar. It's as if the mind had a propensity for learning only a range of possible grammars.{27} Chomsky's approach is thus in two respects incompatible with Skinner's version of behaviorism about language. It uses theory construction, and it does not reduce theoretical terms to observational terms.
b. SOPHISTICATED BEHAVIORISM
However, as many empiricists point out, a behavioristic program need not shy from theory construction, nor need it be shackled to a reduction thesis about the meaning of theoretical terms. For example, Marras himself objects to the behavioristic program as sketched by postulates (i)-(v). He writes that "Postulate (v), for example, must be rejected in favor of a more liberal postulate which allows for the introduction of theoretical constructs . . ."{28}
Sellars' view of behaviorism is another such sophisticated version which does not have these objectionable tenets. He writes:
Today it is obvious at a glance that in a sense the Behaviorist methodology was unnecessarily restrictive. . . . Why not introduce hypothetical constructs into Behavioristics (compare the postulational concepts of physical theory) instead of restricting yourself to intervening variables (the mark of which -- though the terminology was misleading -- is definability in terms of observables)? {29}In fact, Sellars introduces another methodological allowance in his version of behaviorism: he allows for a conditioned response (i. e., associative) connection to take place even between theoretical states. In a spirit of criticism, he writes that in the "Skinnerian tradition"
conditioned responses are talked about in such a way that the overt is conditioned to the overt. This is bound up with the reluctance of many psychologists to speak in terms of establishing connections between postulated central processes; and, in general, with an anti-theoretical bias. {30}Thus, taking Chomsky's criticism of Skinner's behaviorism seriously, it is possible to introduce into behaviorism exactly those elements Chomsky argues are missing: theory construction, irreducibility of theoretical terms, and even connections between theoretical entities. So there is a viable form of behaviorism, defended by Sellars, which is not taken into account by Chomsky.
It would seem that a language, which for Sellars is basically pattern governed behavior, could be learned through stimulus-response (S-R) conditioning:
To learn pattern governed behavior is to become conditioned to arrange perceptible elements into patterns and to form these, in turn, into more complex patterns and sequences of patterns. Presumably, such learning is capable of explanation in S-R reinforcement terms, the organism coming to respond to patterns as wholes through being (among other things) rewarded when it completes gappy instances of these patterns. Pattern governed behavior of the kind we should call "linguistic" involves 'positions' and 'moves' of the sort that would be specified by 'formation' and 'transformation' rules in its metagame if it were rule obeying behavior. {31}Since this passage may be a rare clue to what Sellars has in mind, let's take note of what Sellars is proposing here. First, Sellars has in mind Skinner's "operant conditioning" rather than classical Pavlovian conditioning. The relevant distinction is that in the latter the stimuli to behavior may not always be known; what is known is that it is possible to reinforce emitted behavior, including emitted verbal behavior.{32} And this is the kind of conditioning that suffices. Second, in learning sentences, the learner learns how to complete 'gappy sentences' such as '____ danced with _____ . Such a suggestion reminds me of Frege's characterization of a function or predicate as "unsaturated,"{33} and Ryle's notion of category as determined by "sentence frames."{34}
Skinner in his book Verbal Behavior{35} made a similar claim that sentences are learned through the learning of how to complete sentence frames. And if we focus on this thesis in Skinner which is shared by Sellars (and Quine), what objections does Chomsky have to this approach? In his review of Skinner's book, Chomsky comes near to considering this approach in Skinner when he comments on what he takes to be a common strategy: "The idea that sentences consist of lexical items placed in a grammatical frame is of course a traditional one, within both philosophy and linguistics." {36} I said "comes near" because judging by Chomsky's example, he does not have gappy sentences in mind, but rather grammatical structures. He compares the two sentences:
(A) Furiously sleep ideas green colorless.(B) Friendly young dogs seem harmless.
claiming that they have the "same" frames. And that (B) and (A) (read from back to front) are sentences of English. I am puzzled by what Chomsky has in mind in calling (A) and (B) the same frames. As far as I can discern, they both have the frame
(C) ______ly _____ ____s ___________lessIf this is what Chomsky has in mind, then I can readily agree with his comment that
It is evident that more is involved in sentence structure than insertion of lexical items in grammatical frames; no approach to language that fails to take these deeper processes into account can possible achieve much success in accounting for actual linguistic behavior.{37}It seems then that Chomsky ignores the approach to learning sentences through gappy sentences of the forms:
________ young dogs seem harmless.
______ly young dogs seem harmless.
Friendly _____ dogs seem harmless.
Friendly young _____ seem harmless.
Friendly young ____s seem harmless.
Friendly young dogs ______ harmless.
Friendly young dogs seem _________.
Friendly young dogs seem ______less.It seems to me that a considerable amount of grammatical and semantical knowledge could be acquired through inductive generalization (the sort that is presupposed by behaviorist learning theory) from piecemeal learning of how to fill these kinds of sentence frames.
But once behaviorism is expanded in the manner described to include theory use and principles of induction, someone may object that behaviorism is no longer distinguishable from the rationalism Chomsky wants to defend. The empiricist-rationalist distinction becomes blurred.
Let us pause for a moment to reflect on the requirements for a behavioristic theory of language. A sophisticated behaviorism, such as Sellars', will not shy from theory construction, but it will prescribe standards for theory construction. Specifically, Chomsky introduces a theory to the effect that humans have an innate knowledge of universal grammar which manifests itself in the course of maturation and of learning a conventional language. Likewise a behavioristic theory may also allow for a stage of maturation, not because maturation elicits as innate knowledge of grammar, but because on maturation the brain reaches a particular stage of development where learning a language is possible. And what is the difference in saying that the brain is now in a state of readiness (as the behaviorist will say) and saying that the brain now has the universal grammar? The difference between a behavioristic and an innatist approach such as Chomsky's is whether anything has to be postulated for brain endowment which cannot be accounted for by the postulation of mechanisms in the brain that allow for conditioned learning, especially through sentence frames.
c. METHODOLOGICAL BEHAVIORISM
Sellars assumes a version of behaviorism in the psychology of language learning. Behaviorism, as understood by Sellars, is the thesis that "all concepts [in psychology] must be introduced in terms of a basic vocabulary pertaining to overt behaviour." {38} All forms of behaviorism, by this characterization, initially assume a "a behavioristic language restricted to the non-theoretical vocabulary of a behavioristic psychology." {39} Sellars calls this common language of behaviorism the Rylean language (or Behaviorese).
Since Behaviorese is intended to be an empirical language, it has to satisfy the necessary and sufficient conditions of being an empirical language generally. Here we can incorporate all of Sellars' early writings on the a priori conditions for the possibility of an empirical language. {40} Such a language, as has already been discussed, must have language entry, intra-language, language departure transitions, and auxiliary positions. And Behaviorese satisfies these requirements.
Of course Behaviorese must include "the elementary logical terms for conjunction, disjunction, negation, and quantification, but especially of the subjunctive conditional." {41} I believe that Sellars means to include other (broadly) logical terms as well: terms for identity, class membership, and various modal terms, including those of deontic logic.
The peculiarity of Behaviorese is in how it satisfies the language entry requirement: ". . . the fundamental descriptive vocabulary speaks of public properties of public objects located in Space and enduring through Time."{42} The alternative being rejected by Behaviorese is an initial descriptive vocabulary about private objects and properties. Behaviorese resources allow the expression of empirical generalizations and counterfactual conditionals, which in turn allows talk about the natures of physical objects and persons, including dispositions and short-term propensities of persons.
And as becomes clear from Sellars' (unpublished) letter to Marras, Behaviorese may include a theoretical language about inanimate things: "pre-Joneseans were able to give theoretical explanations of purely physical phenomena."{43} And though it is clear that Behaviorese must initially exclude an introspective awareness of thoughts and sensations, it is not clear to what extent Behaviorese is committed to excluding other theoretical entities. We are only told that
the philosophical situation it [the Rylean language] is designed to clarify is one in which we are not puzzled by how people acquire a language for referring to public properties of public objects, but are puzzled indeed about how we learn to speak of inner episodes and immediate experiences.{44}This is the case because the Manifest Image is a language which includes introspection, and the problem is to give an account of the Manifest Image using the resources of the Rylean language (Behaviorese).
Sellars opposes the thesis of Philosophical (Logical or Analytical) Behaviorism -- the view that concepts about inner episodes "must be built by explicit definition -- in the broad sense -- from a basic vocabulary pertaining to overt behaviour."{45} By contrast, Scientific or Methodological Behaviorism -- the view which Sellars favors -- holds that "some behavioristic concepts are to be introduced as theoretical concepts."{46} Using the terminology of MacCorquodale and Meehl,{47} Analytic Behaviorists restrict themselves to the introduction of "intervening variables", while Sellars also introduces "hypothetical constructs" (or what in quantum theory speculation are referred to as "hidden variables"). From a semantical point of view, intervening variables are terms which can be translated into a non- theoretical Behaviorese language, while hypothetical constructs are terms which cannot. Specifically, Sellars' thesis is that introspective language of thoughts and sensations can be introduced as "hypothetical constructs" by analogy to the introduction of theoretical entities in the natural sciences.
Opposed to both these types of behaviorists are the views of psychological Cognitivists who are not only puzzled by whether and how we learn to speak of inner episodes and immediate experiences, but are also puzzled by how people acquire a language for referring to public properties of public objects in the first place. Cognitivism, in this context, is the view which tries to explain the learning of a Rylean language (Behaviorese) by introducing psychological "hypothetical constructs" in contrast to the view which tries to explain the learning of a Rylean language without this recourse. The latter is Sellars' position. The former position was held by, among others, Descartes, and is currently held by, among others, Jerry Fodor, Hector-Neri Castañeda, and Noam Chomsky, who contend that the learning of a public language requires innate linguistic endowments. Fodor speaks about an innate (private) language of thought, Castañeda about a private language (Privatish), while Chomsky of an innate competence for generating a grammar.
d. THE INNATE IN BEHAVIORISM
Before we can make a judgment about the adequacy of behavioristic resources, we should be able to list them. To say that behaviorism admits only learning from public experience is misleading. After all behaviorists presuppose all sorts of innate mechanisms in an organism. As Quine, who calls himself a behaviorist, put it: "the behaviorist is knowingly and cheerfully up to his neck in innate mechanisms of learning-readiness."{48} He gives as an example of an innate mechanism the ability to distinguish between the qualities of stimuli. In fact, for Quine, the very possibility of learning requires an innate ability to distinguish similarities and differences of form as well as content, and the ability to generalize these through analogies. When Quine talks about the learning of a language he appeals to such abilities.
Since Sellars' view is based on behavioristic learning theory, it must presuppose a similar conception of the human organism and its innate and developmental capacities. I have in mind such things as innate needs, innate reflexes, including various tendencies to respond, avoidance behavior, orientation reflex, instincts, imprinting, and capacity to learn. Behaviorism must also presuppose such innate capacities as 'stimulus discrimination' and 'stimulus generalization', 'spread of effect', 'differentiation of response' (which Quine acknowledges to be innate); principles of stimulus- response association: Thorndike's law of effect, various principles of classical and Skinner's operant conditioning. Judging by his comments on RS organisms, I don't think Sellars needs to or would want to dispute any of these. Since he grants to animals the capacities to represent (map), and even meta-represent their environment, such abilities, it seems to me, require considerable innate mechanisms. This behavioristic apparatus, or something like it, must be presupposed by Sellars, as it is by Quine. But, as far as I know, Sellars does not systematically discuss the foundations of behavioristic theory, though he does make some remarks about it.
Sellars seems to agree on the need to posit innate factors: "People obviously have innate psychological abilities."{49} But the question is: which ones? We know that he grants to people innate representational systems: "Undoubtedly a primitive form of representational system is also an innate endowment of human beings."{50} Judging by the fact that he attributes to animals the capacities to represent (map), and even meta-represent their environment; such abilities, it seems to me, are also innate human endowments. But, as far as I know, he does not discuss in detail behavioristic theory, though he does discuss the logical methodology of behaviorism. His most significant observation is to allow the introduction of hypothetical constructs in psychology. Such an admission allows Sellars to say: "There is nothing self-contradictory about the idea of an innate language-like structure (Mentalese), and there might just be no other way of explaining language acquisition."{51}
As a Methodological Behaviorist, he should apply the principles of Methodological Behaviorism to the learning of Behaviorese itself. This is to say that he should utilize theoretical posits for the learning of Behaviorese. But all he mentions are principles of conditioning -- both classical and operant. Theoretically he allows for an innate grammar of the sort we find in Chomsky as a possible theory, but resists appealing to such a theory in practice. In fact, I suspect that Sellars would side with Quine's criticisms of Chomsky, and even agree with a great deal of Quine's explanations of language learning in his Word and Object.{52}
Instead of the detailed account of language learning found in Quine, we get from Sellars a transcendental reflection on the necessary conditions for empirical knowledge. These are various psychological associations between non-cognitive experiences and language, between different language forms, and between language and actions. He passes over the problem of the more detailed structures of such associations, and the problem of how such associations are possible. It is at this point that cognitive psychology makes it's entrance. Chomsky, for example, is bothered by the fact that
Grammar and common sense are acquired by virtually everyone, effortlessly, rapidly, in a uniform manner, merely by living in a community under minimal conditions of interaction, exposure, and care . . . Knowledge of physics, on the other hand, is acquired selectively and often painfully, through generations of labor and careful experiment, with the intervention of individual genius and generally through careful instruction.{53}One puzzle is how it is possible for a child to learn a language so quickly. To account for this he postulates an innate knowledge of a 'universal grammar'.{54} In one sense of this phrase, it is equivalent to the necessary conditions of any language.{55} These we can grant to exist analytically wherever there is a language.
Chomsky is assuming that a knowledge of language -- specifically grammar -- is underdetermined by the evidence. Knowledge of a language becomes analogous to a scientific hypothesis. However, scientific hypotheses can be either empirical generalizations or theoretical postulates. Is the acquisition of language like the former or the latter? As I see the dispute between the empiricists and the rationalists, the empiricists see the situation of language learning as a matter of empirical generalization, while the rationalists see it as resembling theoretical postulation or, to use a term from Charles Peirce, abduction.
The rationalist-empiricist controversy can be put as a controversy about the cognitive status of abductions. I take it that a rationalist claims that abduction cannot be explained on empiricist principles; the empiricist denies this.
But no matter which way we look at hypothesis formation, there is the difficulty that a hypothesis is formulated in a language. Therefore if learning a language is a matter of hypothesis formation, then in learning language-1 we must assume the existence of metalanguage-2 in which the hypotheses about language-1 are formulated. Obviously this view will result in either an infinite regress, a vicious circle, or an innate language.
The solution is either to embrace one of these consequences (as does Jerry Fodor who follows the tradition of positing an innate "natural" language of thought as, for example, in St. Augustine and William Ockham) or reject the assumption of a literal understanding of "hypothesis" formation. The alternative is to take "hypothesis" formation as a figurative way of talking about cognitive structures. But these cognitive structures could be understood in two ways as either the generated representational systems or the structures for generating representational systems (including a conventional language). Put in these terms the issue between empiricists (behaviorists) and rationalists (cognitivists, mentalists) is the nature of this generating capacity. The question is: What are the human endowments which make the learning of a conventional language possible? To answer this question we must focus on those characteristics of a human language and its acquisition which are problematic. Chomsky cites two problematic features. The first is the creative aspect of language; the second, the effortless ease of acquiring grammar and common sense. Chomsky's solution is to posit an innate endowment for generating particular grammars. The generating capacity is (infelicitously) called a "universal grammar."
e. ORDER OF BEING AND ORDER OF KNOWING
Sellars sometimes makes use of the Aristotelian distinction between what exists and what is known to us. And he applies this distinction to the relation of thoughts to knowledge. He states that thoughts exist without being known; knowledge of them requires the conceptual apparatus of a language. If this is all that Sellars claims, then there would be no essential controversy with innatism.
This formulation, however, blurs the distinction between conceptual thoughts and rs-thoughts. It is clear that Sellars would grant human beings rs-thoughts which are independent of language, but it is not equally clear that he would grant conceptual thoughts independently of language. His claim that thoughts are ontologically prior to language may mean no more than that after a language is learned, a person acquires conceptual thoughts, and these, as internalized, are prior in the order of being from then on. On this interpretation we have to assume that the Ryleans in learning a public language at the same time (concomitantly) are learning to think conceptually. From this it would follow that there is no conceptual thinking prior to the learning of a language. The distinction between a prior ontological being of thoughts is at best the claim that thoughts (qua neural patterns) are causally (though not genetically) prior to overt utterings, and that such neural patterns embody the dispositions to overt utterings. What Sellars would not be claiming, on this interpretation, is that in the order of being conceptual thoughts exist prior to the acquisition of a public language.
From this interpretation many of Sellars' claims follow. His claim that "the individual as a conceptual thinker is essentially a member of a group,"{56} and that "It is no accident that one learns to think in the very process of learning to speak."{57}
This interpretation may also shed light on the Sellars-Chisholm correspondence. Chisholm wrote to Sellars that:
Thoughts would be intentional even if there were no linguistic entities. (This is a sentence about psychology. I concede that if we had no language, our thoughts would be considerably more crude than they are.){58}Chisholm may have meant by this no more than that rs-thoughts exist independently of language learning. That this is what Chisholm did mean, or was taken to mean by Sellars, is suggested by Sellars' letter to David Rosenthal several years later, where he wrote:
My guess is that Chisholm so uses 'psychological' that it includes items which I would deny to be intentional (e.g., sense impressions, tickles, pains, etc.) -- cf. my discussion of the pseudo-intentionality of sense impressions in EPM ["Empiricism and the Philosophy of Mind"]. Unless I am very much mistaken, he thinks that they have intentionality, or confuses a broader sense of 'intentionality' in which a non-conceptual items can have intentionality, with a narrower sense in which they cannot.{59}It is not entirely clear what Sellars is alluding to by 'intentionality in the broader sense'. My best guess is that he has in mind rs-concepts. But despite this clarification, if we introduce a distinction between conceptual and non-conceptual representations, it is still not entirely clear to me what kind of thinking Sellars in fact grants to pre-linguistic people and animals.
This finishes my examination of the first stage.
2. 2nd STAGE: SEMANTICAL DISCOURSE
For the Ryleans to reach the stage where sensations and thoughts are responded to without verbal expression, several things are needed. Sellars points out that two things are needed for the Ryleans: semantics and postulational theory.
The first addition needed for the possibility of introspective reports is semantical discourse. Semantical discourse consists of using such words as 'means', 'true', 'synonym', 'denotes', 'refers'. Sellars believes that the move to a semantical language, which requires the introduction of a metalanguage, is necessary for reflectivity, self-knowledge, and, ultimately, for becoming a person.
Sellars' position is that to be truly human, i.e., to be a rational person requires metathinking. In fact, he writes: "To lose the tendency to have appropriate meta-thoughts is to cease to be rational or to lose one's mind."{60}
The addition of semantical resources enables us to describe the intentionality of thoughts:
it is clear that semantical talk about the meaning of reference of verbal expressions has the same structure as mentalistic discourse concerning what thoughts are about.{61}Ausonio Marras in "The Behavioristic Foundation of Sellars' Semantics" found it a problem to account for the acquisition of semantical discourse with the Rylean resources. He reasoned:
Since the acquisition of semantical discourse by Sellars' Rylean ancestors is surely not to be viewed as a gift from the gods, but must, instead, be regarded as a natural development from within the primitive Rylean language itself, we must suppose that the conceptual and methodological resources of the primitive language must have been rich enough to permit this development.{62}Marras sees a problem for Sellars because he has focused his attention on the following characterization by Sellars of the Rylean language: "as not only a behavioristic language, but a behavioristic language that is restricted to the non-theoretical vocabulary of a behavioristic psychology."{63} Given this restriction, Marras can only see two alternatives for this kind of logical behaviorism: either semantical terms are introduced by explicit or implicit (contextual) definitions. And since Sellars has disavowed this kind of analysis for semantical terms, Marras sees only one way out for Sellars: to introduce semantical terms as theoretical terms. But to introduce theoretical terms, the Ryleans, according to Marras, must have had that ability to begin with. So, Marras concludes that the original characterization of the Rylean language as being restricted to a non-theoretical vocabulary must be given up. This is Marras' alleged correction of Sellars on this point.
To this, Sellars replied (in a personal letter to Marras) that
The reference to theoretical discourse is pertinent, but not of the essence, for I could have made an explicit commitment to the idea that pre-Joneseans were able to give theoretical explanations of purely physical phenomena. Indeed, I would have done so had I been on my toes.{64}According to Sellars, "Jones . . . was already familiar with theoretical explanation in non-mentalistic contexts."(65) So Marras' narrow reading of Sellars turns out to be a misinterpretation.
Still Marras' point about the logical character of semantic discourse has to be answered. Sellars' answer is that semantical sentences are not analyzable. So neither explicit nor implicit definitions are to be sought. Nor is semantical language like a theory. Semantical discourse is classificatory discourse, as Sellars explained to Marras:
Our Ryleans are presumably able to classify a wide variety of things and events. They also play chess, and use sortals, like 'pawn,' the senses of which are bound up with rules. Thus, as far as semantics is concerned, Jones' chief invention is the introduction of what I have called 'illustrating quotes' to form sortals which classify linguistic expressions in terms of their functions in the language to which they belong.{66}Sellars' point is that semantical discourse is simply an analogical extension of discourse that is granted to the Ryleans to begin with.
a. NATURE OF SELLARS' SEMANTICS
A major theses of Sellars' philosophy is the distinction between the order of signification (the logical order) and the natural order (the order of reality). Such a bifurcation leads to difficulties -- difficulties which could be avoided by distinguishing within the order of signification between the logical order and a non-logical order -- what I have been calling a Representational System. I will first formulate Sellars' position, then the difficulties, and finally my resolution of these difficulties.
Words and artifacts in general have their feet in the logical and the natural order. For example, a hammer qua material composite of metal and wood of a particular shape is just a physical object; describable and explainable by the principles of physics, biology, metallurgy, etc. -- it is in the natural order. However, this same physical object can be used as a tool. Relative to what it was designed for, and relative to what it is normally used for, it is a hammer. Still it could be used for other purposes: as a paper weight, a projectile, a miniature crow-bar, whatever. In this aspect as a role player, the hammer is in a different order -- the order of means for actions.
It is similar with words. In the natural order, words are normally formed either from a combination of vocal sounds or ink-marks (inscriptions) on paper. In the logical order these vocal sounds and ink marks play a role in the syntactical-semantical-pragmatical dimension. And if we think of thoughts as analogously composed of "words", then these mentalese words too are in the logical order.
(1) QUOTING CONVENTIONS
For the sake of clarity, Sellars introduces various quoting devices to facilitate talking about (mentioning or illustrating) the occurrences of words in these various capacities. Our common practice is to mention words by using either scare-quotes or double quotes. But this practice conceals a six- fold ambiguity. There is first a distinction between word tokens and word types, and this distinction is applicable to three orders: the real order, the logical order (the order of signification), and the mentalese order (the order of thoughts).
When Sellars wants to talk about (mention) words as tokens of ink-marks in the natural order, he writes the word "cat" using either hyphens (e.g., c-a-t) or slurs (e.g., c^a^t). He uses asterisk quotes to mention words in the natural order as types of designs. The word "cat" written in either italic, long hand, Gothic, and so on, placed between asterisks refers to the types as illustrated between the asterisk quotes; thus, *cat*.{67}
To mention words as tokens in the order of signification, he places them in scare (single) quotes (e.g., 'cat') or double quotes (e.g. "cat"). To mention words as types in the order of signification (role, function), he places them in (illustrating) dot quotes,e.g., .cat..{68} He explains that dot-quotation is "a mode of quotation which spans different languages, as familiar quotation spans the difference between written and spoken forms of the same language."{69} In his correspondence with Rosenthal he adds:
to classify an utterance as a .___. is to classify it in terms of its role in a language-life game, i.e., in terms of its place in a system of language entry transitions, language departure transitions and intra-linguistic moves.{70}The tokens of mentalese are neurophysical states of the brain and are mentioned by single French quotes; e.g., <cat>.{71} Mentalese types are mentioned by using double French quotes; e.g., <<cat>>.{72} An alternative terminology for what seem to be mentalese types is given in the Rosenthal-Sellars correspondence. Mentalese types are here called Ockhamite items, and are mentioned by the use of diamond quotes; e.g., <>cat<>.{73}
In his correspondence with Rosenthal, Sellars also introduces a disjunctive quoting device which mentions either a Rylean or an Ockhamite item. He calls this disjunct a Fregean item, and it is mentioned by putting it in dot- diamond quotes, e.g., .<>cat.<>.{74}
The following chart summarizes these quoting conventions:
ORDER TOKENS TYPES WRITTEN
(NATURAL ORDER)c-a-t or c^a^t *cat* SIGNIFICATION
(LOGICAL)'cat' or "cat" .cat.
(Rylean)MENTALESE <cat> <<cat>> or <>cat<>
(Ockhamite)FREGEAN .<>cat.<>
(Rylean or Ockhamite)[Note: Sellars' use of various quotation devices has to be distinguished from his informal use of double quotes and single quotes to either mention, illustrate, or emphasize a linguistic item.]
Sellars is not always consistent in distinguishing dot-quotes from double French-quotes. For example, in "Being and Being Known,"{75} he introduces dot- quotes for "mental words." But if he is to be consistent with his conventions in the above table, this should be done with double French-quotes.
Sellars claims that dot-quoting has an ordinary language analogue in the following suffixes for forming common abstract nouns: "-hood", "- ity", "-ness", "-tion", "-ship"; and for forming the common noun for propositions it uses the "that _____" construction.{76} The following identities are consequences:
Triangularity = the .triangular.That it is raining = the .it is raining.{77}
(2) MEANING
In Sellars' view, classical semantical discourse (involving the use of such words as 'means`, 'denotes', 'refers', 'truth', and their cognates) construes semantical words as relation words, linking together a verbal expression and some referent (object, state of affairs, or a thought). But, according to Sellars, this is a mistake. For example, consider the formula: "rot" means red.
The rubric ' "..." means ---' is a linguistic device for conveying the information that a mentioned word, in this case 'rot', plays the same role in a certain linguistic economy, in this case the linguistic economy of German-speaking peoples, as does the word 'red', which is not mentioned but used -- used in a unique way; exhibited, so to speak -- and which occurs 'on the right-hand side' of the semantical statement.{78}Semantical statements of the form: ' "..." means ---' are not to be understood as expressing a relation between "..." and --- ; rather they are to be understood as using variants of the copula to express sortal classifications of linguistic functions (uses, roles) having the form: "..." is a(n) ---.
A more detailed formulation relativizes the subject term to a language and dot-quotes the sortal predicate: 'D' (in L) means .G. (where "L" names a specific language). And a still more perspicuous formulation reformulates the subject term as a distributive singular term, substitutes a copula for the 'means' term (and cognates), and expresses the grammatical form for the sortal predicate: the 'D' (in L) is a .G..{79} Sellars gives the following explicit directions for expressing such formulas:
(a) 'Means is a specialized form of the copula.
(b) What follows 'means' is to be construed as a metalinguistic sortal.
(c) The subject of a 'means' statement is a metalinguistic distributive singular term.{80}{80}This treatment entails that the expressions 'means', 'denotes', 'stands for', 'refers', and their cognates are variants of the copula. The left side mentions a word or phrase whose meaning is sought, while the right side- functions as a predicate expression, using a familiar word or phrase by which to illustrate and to convey the meaning-as-linguistic function.
The concept of a proposition is a concept of a functional role of a sentence (abstracted from any particular linguistic vehicle). It is not a concept of a thought or anything non-linguistic; rather, it is the concept of a translinguistic functional role (i.e., as exemplified in various languages). Thus, "Das ist rot" and "This is red" express the same (or similar) propositions by playing the same (or similar) role in the observation-inference-volition transitions in their respective linguistic mediums.{81}
(3) POLEMIC WITH CHISHOLM
One objection to this semantical view of intentionality is that it evades the real issues. Chisholm insisted on the need for more resources than those provided by Sellars. The objection is that these meager resources do not account for the intentionality (or meaning) of thoughts. Sellars' way out is to analyze intentionality through meta-linguistical resources of the Ryleans.
Chisholm, in their correspondence, argued that Sellars could not explain the concept of meaning without presupposing the concept of a non- linguistic thought. The Chisholm-Sellars correspondence is ultimately the question of what distinguishes a "meaningful" language from the sound productions of a parrot. Sellars' position in this debate was to argue that:
the metalinguistic vocabulary in which we talk about linguistic episodes can be analyzed in terms which do not presuppose the framework of mental acts; in particular, that". . ." means pis not to be analyzed as". . ." expresses t and t is about pwhere t is a thought.{82}It seems to me that Sellars was arguing on the basis of two presuppositions. The first was that for the proposition q to be the analysis of the proposition p for a person S, S must have the concepts contained in q. The second was that we are dealing with a Rylean situation in which the person has a concept of a meaningful sentence, but does not have a concept of a thought, i.e., other than as a non-parrotingly made sentence. Given these presuppositions, it is obvious that the quoted formula cannot be an analysis in Sellars' sense; though the formula can serve as an explanation. At the end of their correspondence, Chisholm concedes relative to the presuppositions, but objects to Sellars' requirement for an analysis.
Chisholm should have conceded that the above so called "analysis" is really what Sellars calls an "explanation", and expressed his disagreement relative to this terminological concession. Of course, Sellars would have agreed that the above formula does express an explanation. But that would have been only a superficial agreement. The root disagreement concerns Chisholm's claim that there could be thoughts without language. And here Sellars, if he were more discerning, could make a distinction between conceptual and non-conceptual thoughts, and concede that there could be non- conceptual thoughts without language but also deny that there could be conceptual thoughts without language. And possibly Chisholm would agree. But neither Sellars nor Chisholm made such a distinction and their exchange ended on an impasse.
(4) IMPOTENCE OF EXTENSIONAL SEMANTICS
Given the various mentioning quotes, Sellars' major thesis is that semantical statements are in the order of signification: they classify words as role players. It is a mistake to try to relate words as mentioned role players with the real order, as is done, for example, in the Tarski-Carnap semantics. The formula
'feline' denotes catwhich has the form
(linguistic item) R (non-linguistic item){83}is ill-conceived in this semantics. The formula presupposes that a perception (or observation) of a cat is different from a thought of a cat (as expressed in a linguistic token of "cat"). To this Sellars responds: "My view, on the contrary, is that 'observations' are themselves thoughts; they are thought-tokens which are correct responses to the objects which caused them."{84} Since a cat is not a role player in the order of signification, the formula should be recast in the following way:
For some S, 'feline' (in E) stands for S, and S is materially equivalent to .cat.['S' is a variable taking dot-quoted expressions as substituends.] This formula is materially equivalent to:
(x) x is a feline iff x is a catSellars presents other extensional formulas in terms of such intensional formulations.{85} These reconstruction are provided as evidence for the claim that "extensions are limiting cases of intensions."{86}
The main reason Sellars rejects extensional semantics is because such a semantics is incompatible with his rejection of the myth of the given. Extensional semantics presupposes that two types of entities can be identified: an item in a language and an item in the world. But if there is no awareness independent of a language, then the awareness of an item in the world must be, for Sellars, mediated by a conventional language. Suppose a person knows an item T in the world through a word 'D' of language L. If he is to be aware of this T it must be through a metalanguage ML which expresses the functional classification of 'D' (in the 'D'-T complex) as a .D..
In extensional semantics the presupposition is that on the one hand there is the language L and on the other hand the world; both standing in some relation R to each other. But on the Sellarsian model the situation must be such that on the one hand there is the language L and on the other hand an L-World complex formed by a (causal) conditioned association. Apparently the relation between a language L and a complex L-World is a relation that hooks up between the Ls. All semantical assertions are ultimately language-language comparisons.
(a) TRUTH
Sellars approach to the meaning of truth is such that it is also a language- language relation. His account is a variant of Tarski's. Tarski analyzed truth by the formula:
'p' is true iff pSellars considers this formula to be inadequate. It requires some revision. The first revision is to change the scare-quotes to dot quotes in order to make the formula apply to inter-linguistic roles, thus:
the .p. is true iff pSecond, Sellars views the truth statement as itself authorizing an inference to p. He puts it this way:
'True' then means semantically assertible (S-assertible) and the varieties of truth correspond to the varieties of semantical rule. From this point of viewthe .snow is white. is truehas the sensethe .snow is white. is S-assertible:and the implicationthat snow is white is true --> snow is whiteis not an element of an extensional definition of 'true', a recursive listing of truth conditions, as, in effect it is on Carnap's account, but is rather a consequence of the above intensional definition of 'true', in the sense that the assertion of the right hand side of the implication statement is a performance of the kind authorized by the truth statement on the left.{87}On this rendition, to say that p is true is simply to permit saying p.
(b) CRITICISM OF LINGUISTIC IDEALISM
An apparent consequence of viewing semantics in the Sellarsian way is that we seem to be trapped in language, and cannot talk about a language-world relation. Any attempt to talk about a world independent of language involves a dependence on language. Let me call this 'linguistic idealism'. The so-called 'hermeneutical circle', I believe, is a variant of this seeming inability to get beyond a language. Cornman, following Sellars here, expresses this predicament succinctly in the following way:
What I shall try to show is that designation rules of the form of:I. 'p' (in L) refers to qand sentences of the form of:II. 'p' (in L) refers to what 'q' in this language,i.e. the language I am now using, refers to are logically equivalent and therefore that designation rules merely relate 'p' to another linguistic expression, 'q'.{88}This piece of reasoning had repercussions on Cornman's overall view of metaphysics. Following Carnap, he distinguished internal from external questions about conceptual frameworks, and identified metaphysics with the attempt to ask external questions about the adequacy of any particular conceptual framework. But reflection on the above predicament, led him to conclude that at best metaphysics can only work with internal questions.{89} This predicament has its radical expression in the writings of Jacques Derrida who sees language as an absolute free play with signs, and less so in the writings of Richard Rorty who talks about culture as a (foundationless?) conversation. These consequences follow if ultimately there is no way available to anchor signs to extra-linguistic foundations, origins, or "givens". If this is the conclusion we are forced to, then I take this to be a reductio ad absurdum of Sellars' position. However, I take it, Sellars too is troubled by such an outcome.
Sellars attempts to escape linguistic idealism through the Tractarian picturing strategy.{90} The claim here is that if we regard language from a naturalistic point of view as consisting of a concatenation of objects (ink-marks, sounds, or whatever), on the one hand; and the world as consisting of another grouping of objects, on the other; then the linguistic objects can be viewed as "picturing" the world more or less adequately through a complex manner of projection.{91}
Sellars is cautious in this formulation of picturing. He is not offering a correspondence theory of truth, which would presumably have the formulation:
(statement) pictures (fact);instead, he writes:
The schema for picturing is not(statement) pictures (fact)but rather(object) pictures (objects).{92}I see Sellars' strategy here as expressing a necessary condition for any kind of representation. This seems to be conceded by Sellars when he writes to Harman: "The truth of basic sentences, however, involves, indeed is grounded in, the existence of a relation of picturing between tokens of the sentences and the objects to which they refer."{93} Later in his letter to Harman he expands to make the picturing relation also a sufficient condition for semantic assertibility of truth: ". . . correspondence to the relevant objects is a necessary and sufficient condition of the semantical assertibility of a basic statement . . ."{94} In a subsequent letter to Harman, Sellars expressed the intention (in a future letter, which never was written){95}
to satisfy you that it follows from my account of picturing that (leaving dot-quote niceties and tenses aside)"a"s which are concatenated with an "f" [f*"a"s] picture a (correctly) iff fa(not, as you put it,"f"s picture a iff fa).{96}I find the idea of the picturing relation as providing the sufficient condition for the semantic assertibility of fa puzzling, unless the notion of picturing (i. e., projection) is so understood as presupposing a relevant psychological state of a person, a sizeable chunk of (or even the whole) language, and a knowledge of the contexts of utterances. These presupposed conditions seem to be demanded by Sellars' holism about language.
The appreciation of which conditions for picturing are necessary becomes clearer once we try to describe a computer simulation of human language functioning.{97} From an ontological programmatic point of view, I see nothing objectionable to Sellars' picturing strategy.
However, if the picturing strategy were viewed from an epistemic point of view, it would beg the question. This would be so because each of the relata in the picturing relation has to be identified, and this identification requires epistemic and linguistic resources. Suppose I focus on the relation of
T-h-i-s i-s a-n i-c-e c-u-b-eto an ice cube. For me to focus attention on these relata requires that I use the appropriate concepts, i.e., I must be able to say something like:
(T-h-i-s i-s a-n i-c-e c-u-b-e) R (ice cube). ('R' is the picturing relation)And how am I to determine the truth of this relation? Am I to go to a meta-picturing relation, and so on to infinity? I don't see how Sellars can extricate himself from this predicament.
However, by using the distinction between an rs- and a conceptual representational system, I can see a way of escaping linguistic idealism, and a way of defending a modified version of extensional semantics and a picture (read that 'correspondence') theory of truth.
What I need first is a way of referring to non-conceptual thoughts. I have been doing this by prefixing 'rs-' to the appropriate concepts. The domain of rs-concepts must be restricted to experiences with a sensorial content. Moreover, our vocabulary about these sensorial contents must have a dual association: one with rs-concepts and one with "logical" concepts. And we can say that at least for perceptual claims that part of their meaning consists in an association with an rs-meaning. We can even exploit Sellars' strategy by claiming that for perceptual reports 'p' (in L) means .rs- p..
A possible objection to this strategy may be that I am resurrecting a denoting theory of meaning, i. e., what Ryle had called the "Fido"- Fido theory of meaning. This is not the case. The theory of meaning that I am advocating is the same as Sellars'. What must not be overlooked is, first, that in the denotational theory "Fido" is a mere noise, whereas on my construal it is already a mentioned word in some language L. Second, the denotational theory uses a second occurrence of "Fido" to refer to an object; whereas on my construal the second occurrence of "Fido" (more properly .rs-Fido.) is also a mentioning use of a symbol in a representation system. My approach, like Sellars, is a translational view of meaning -- a translation between two representational systems; whereas the denotational theory makes no such assumptions.
3. 3rd STAGE: THEORETICAL DISCOURSE
a. THE TRUNCATION PROBLEM
However, there are other sorts of problems for Sellars' account of introspection at the rudimentary level of consideration. I have credited Sellars with the equivalence thesis (EQ), which is the idea that the Ryleans thinking is simply talking-out-loud.{98} And, of course, Ryleans can talk about having the disposition to talk-out-loud. So the Ryleans have the concept of non-parroting intelligent speech, and the concept of dispositions for such speech. But they do not have the concept of a thought as distinct from speech episodes or propensities for such speech episodes. To have these concepts they require the possession of meta-speech episodes. And there is no difficulty in granting this further ability to the Ryleans as well.
The problem is to provide an account of a concept of thought in an episodic sense which is distinct from either a concept of a disposition to talk or the concept of talk itself. In what follows, I have tried to reconstruct through a sequence of steps Sellars' account of how the concept of an episodic thought is learned. My misgivings are with steps 5 and 9.
Sellars' account can be viewed as consisting of the following theses:
(1) In the Manifest Image we have actual experiences of unexpressed conceptual thoughts(2) What is actual is possible.
(3) Language and thought are connected for the Ryleans at time t 1.
(4) Language and thought are severed for the post-Ryleans at time t 2.
(5) Language and thought are severed by . . . (?)
I am leaving (5) as an ellipsis because I don't know what Sellars would say here. But it is clear that he needs this as a step in his account.
although 'T --> V' is the natural unit with respect to acquiring a conceptual apparatus . . . actually the process can be truncated. Behavior which was originally conceptual as built on 'thinking out loud' (T --> V), can become conceptual as being built on T without V. Behavior which was originally a function of (T --> V) becomes a function of T.{99}In order to get from the Rylean to the Manifest framework, there must be a mechanism by which the thought-language tie is severed and a thought-(non- linguistic) behavior tie substituted. And this must be something that occurs without human intervention if it is a phenomenon which needs ex-planation. But as far as I can tell, Sellars never gives an account of how such a severance is likely or even possible. So we must assume such a severance. But how exactly is the truncation phenomenon to be described? At best the Ryleans can suppress loud talk to a kind of mumbling to oneself, and perhaps suppress it to a stage of having Watsonian laryngeal movements. But what will cause a suppression of these? Sellars never gives an account of the suppression of speech to the extent that only conceptual thoughts occur. He simply assumes that some mechanism of suppression is available so that there are thoughts without verbal expression.
In addition to this, there is another phenomenon that bothers Sellars, and which needs an explanation.
(6) Ryleans are sometimes observed to behave intelligently even when there is no overt talking going on.Ryleans must be able to act intelligently without this being the outcome of consciously entertained thoughts; otherwise if they are conscious of their thoughts as causing intelligent behavior, there is nothing for Jones to explain. Thoughts would already be introspected.
(7) To explain non-verbal intelligent behavior, a genius, Jones, formulates a theory that there is something like talking going on whenever people act intelligently.By 'theory' in this context Sellars means the introduction of unobservable entities by analogy to observable models. The Ryleans are introduced to a theory that thinking is to be understood on the model of overt talking. The analogy between thoughts and language is a second-order, formal analogy. It cannot be a first-order material analogy for the simple reason that the vehicles of thought are for Sellars neurological states. Sellars explains this theory introduction:
According to our philosophical myth, a proto-scientific member of the community, Jones by name, develops the hypothesis that people's propensities to think-out-loud, now this, now that, change during periods of silence as they would have changed if they had, during the interval, been engaged in a steady stream of thinkings-out-loud of various kinds, because they are the subjects of imperceptible episodes which are:
(a) analogous to thinkings-out-loud;
(b) culminate, in candid speech, in thinkings- out-loud of the kind to which they are specifically analogous;
(c) are correlated with the verbal propensities which, when actualized, are actualized in such thinkings-out-loud;
(d) occur, that is, not only when one is silent but in candid speech, as the initial stage of a process which comes 'into the open', so to speak, as overt speech (or as sub-vocal speech), but which can occur without this culmination, and does so when we acquire the ability to keep our thoughts to ourselves.{100}
Note that the assumption introduced here is that
(8) People undergo a change in propensity about what they will think-out-loud.Jones intends to provide an explanation for this change of propensity.
It seems that we need an explanation of something more basic, namely, thinking-out-loud. And whatever explanation is given for this, it seems, could be given for the change in thinking-out-loud. This is a complicated matter because the Ryleans make observational reports, inferences, and express intentions. We can also assume, I take it, that they express desires such as hunger or wanting to find something. How are these various thought- out-loud to be explained? Well, the observation reports will be made on occasions of observing, inferences on occasions of hearing thoughts-out-loud, and volitions on the occasion of expressing desires. And how is the expression of desire to be explained within the Rylean context? I suppose by something like the Skinnerian correlation between a state of deprivation and an expression of that state.
If these are the normal types of explanations for the Ryleans, then in the case of a change in propensity to think-out-loud, this can be explained by the same means. The Ryleans can refer to unobserved observations, unobserved thoughts-out-loud, and unknown deprivations.
But this is not the sort of explanation that Sellars puts in Jones' mouth. Jones' explanation is that
(9) People who undergo a change in propensity to think-out-loud are subjects of imperceptible episodes.Why doesn't Jones posit the occurrence of unperceived perceptible episodes? This seems the likelier course to pursue.
But even prior to the need to explain private thoughts, it seems that Ryleans would be confronted with a need to explain anomalous perceptions, such as hallucinations, abnormal desires, and abnormal emotional upheavals.{101} It seems to me that Jones would need to introduce something like a privately perceptible world rather than private thoughts.
Sellars, however, bypassing this route, makes Jones give explanations which in effect are modeled on Rylean inferences. Inferences in the Rylean community are verbal responses to verbal promptings; so in the Rylean theory all verbal responses are the outcomes of non-verbal-like promptings.
(10) Jones' theory will serve the function of making the unconscious thoughts conscious by training people to respond conceptually to their thoughts.Assuming that the other steps in the reasoning are acceptable, there is nothing objectionable to the idea that people can be trained to respond to their theoretical states. This is aptly demonstrated by the techniques of biofeedback training.
Still, Sellars' position has a gap. The Rylean myth leaves us with an unsolved problem of how thought is severed from language; so his account of how his position is possible is incomplete. According to Sellars, to have a thought is to say and to have the propensity to say p. But it makes little sense to identify the thought with the propensity. The propensity is to have a thought (when verbalized). In this context to talk about suppressing the verbalization is simply to suppress having a thought. What Sellars needs is not a propensity to talk, but an episode which is a thought. But what can this episode be? It is simply the episode of talking.
Does this account help to explain how people can recognize their non-linguistic thought episodes? I don't believe it does. Sellars' assumed without argument that the original tie between language and thoughts can be severed. But the nature of this severance is totally mysterious. It is in fact incoherent from Sellars' assumptions. Secondly, he has assumed that thoughts once severed from language can have a causal tie to non-verbal behavior, so that the behavior is intelligible. Thirdly, he has assumed that such intelligible behavior can be unconscious (i.e., unrecognized). Fourthly, he assumes that his genius Jones by teaching a theory of thoughts which (per assumption) are unconscious will become conscious.
I find Sellars concept of a thought as something analogous to speech perfectly acceptable. But accepting this thesis does not aid in trying to settle the question whether thoughts can exist independently of language. Thinking becomes understood as silent talk. And in fact there is a neurophysiological process taking place when we talk. When however this brain activity is responded to without overt talking, the Ryleans have learned to recognize their thoughts directly. So learning to recognize thoughts amounts to responding to the brain activity which took place while learning to speak and suppressing the speech.
The reason Sellars' account has troubles is because it conflates non-conceptual with conceptual thought. Sellars says that a child "learns to think in the course of learning a language." But such a thesis is misleading. In one sense, as Jean Piaget argues, a child learns to think non-conceptually independently of language; on the other hand, it learns to think conceptually in the course of learning a language. Sellars does not always keep these distinction in mind.
Given this distinction, it is not evident to me that step (1) is true. Perhaps all our unexpressed thinking is non-conceptual thinking. In that case steps (4) and (5) are not needed. The claim would be that all conceptual thinking is and remains overtly linguistic. In that case, what about step (6)? The non-linguistic intelligent behavior can be explained by the theory of non-conceptual thinking. It is at this point that we can accept step (7), as the thesis that non-conceptual thought is analogous to conceptual thought. (8) is acceptable, while (9) is offered without considering the more plausible explanatory hypothesis which appeals to perceptible experiences. What about (10)? This seems to be superfluous. Instead I propose the thesis that in the course of learning conceptual thought, one learns to conceptualize one's non-conceptual thoughts.
But let me be cautious here. Sellars is not describing our present conceptual framework, but an antecedent framework from which to generate our present framework. Before he introduces the mechanisms for introspection, Sellars needs to get agreement from us about the causal relation of thoughts (as brain states) to their expression in language. From a theoretical (ontological) point of view, according to Sellars, the relation between thought and talk is a causal one: talk is a culmination of thought. Let us call this the Causal Thesis:
(CT) T -> V ['->' is a causal relation]He puts it this way:
thoughts originally occurred as initial stages of the unitary process T -> Vwhere 'T' stands for thought and 'V' for overt verbal behavior. Doing, seeing, intending, recognizing, etc. were conceptual or rational as threaded on what we would call 'thinking out loud'.{102}>To learn any language, Sellars assumes that humans have the ability to be trained to respond to situations with sounds. Since he does not allow for conceptual awareness prior to having a language, the learner cannot be conceptually aware of the situation to which he is responding; though he must evidently be impressed non-cognitively by the situation. That is, Sellars assumes at this stage that there are non-cognitive states eliciting a response through conditioning.
But surely this ontological dependence of verbal expression on conceptual thinking from a genetic point of view must be that a conceptual thought (as a neuronal patterning) is brought about only by learning to think-out-loud: "the ability to have thoughts is acquired in the process of acquiring overt speech."{103} This amounts to the claim that the necessary causal condition for thoughts is learning to speak, which we can formulate as the Genetic Causal Thesis that
(GCT) V -> TIf this were a mitigated claim concerning some thoughts, as expressed in
(GCT') For some thoughts, V -> Tthen I could accept the claim. But the problem is that (GCT) as well as (CT) conflate the distinction between non-conceptual representational systems and conceptual thoughts. Let's refer to them as thoughts-1 or T-1 and thoughts-2 or T-2, respectively. And let me also make a distinction between verbal expressions as non-conceptual or V-1 and as conceptual or V-2. Now given these distinctions, it isn't even clear what Sellars' claims are. What is not clear is the nature of the thoughts which pre-linguistic people are said to have. If Sellars' system is to have the kind of coherence that it seems to have, Sellars must say that pre-linguistic people have thoughts-1 and not thoughts-2. And it is clear to me that he would agree that (CT) and (GCT) apply to thoughts-2, such that
(EQ) V-2 <--> T-2This is simply a reformulation of the idea that all Rylean conceptual thinking is thinking-out-loud.
What is not clear is the status of thoughts-1 in his framework. Is it the case that
(CT-1) T-1 -> V-1?This expresses the idea that a non-conceptual representational state causes some state (possibly verbal) which is a symptom or expression of T-1. Did Wittgenstein have something like this in mind when he wrote "the verbal expression of pain replaces crying and does not describe it?"{104} I see no reason for Sellars to reject (CT-1), so I will assume he accepts it. The controversial thesis, however, is
(CT-2) T-1 -> V-2This is the thesis that we are able to respond conceptually to non- conceptual states. Here we must be careful in our interpretation so as not to get the myth of the given. Sellars' position is that the types of V-2 available as means of responding to T-1 are restricted by our linguistic resources. Specifically the restriction is in the categorial kinds which we are able to express by V-2.
But I find an ambiguity here. On the one hand T-1 is a primitive state, supposedly unaffected by conceptual refinements such that there is a part of V-2, call it V-e, that is correlative to T-1. And there is a part of V-2 that is superfluous--it could be theoretical-- to T-1. And it may also be the case that V-2 has no part which is correlative to T-1 (I am thinking of fictitious entities as named by V-2).
To make this discussion connect with the philosophical literature, think of 'V-2' as a 'perceptual report',' V-e' as an 'expressive judgment', and 'part of' as 'phenomenological reduction of', 'T-1' can be viewed as the 'given'.{105}
Sellars' position, I believe, is that though V-e is correctly understood as a phenomenological reduction of V-2, there is a conceptual restriction to the type of V-e available. This brings us to the problem of whether phenomenal reports are available for the justification of empirical knowledge; so I will postpone any further discussion along these lines to the chapter on Foundationalism.
[Go to Chapter 7]