Philosophy and the Vision of Language (routledge Studies in Twentieth-century Philosophy)

4. Ryle and Sellars on Inner-State Reports ==========================================

As we saw in chapter 3, Wittgenstein’s use-based theory of meaningfulness in theTractatus already conceived of the sense of propositions as defined by the regular possibilities of their significant use, including their inferential relations with other propositions in a language as a whole. Over the decades following the publication of theTractatus , developments of this holist, inferentialist program of analysis would come to exert an ever broader and more widespread influence over the methods of analytic philosophers. It would play a central methodological role, indeed, in the single development most characteristic of midcentury analytic philosophy. This was the radical critique undertaken by Austin, Ryle, and Sellars of the various subjectivist, empiricist, or Cartesian theories of mind that had placed the “givenness” of private sense-data or other immediate contents of consciousness at the center of their accounts of knowledge and understanding. Against these earlier theories, the midcentury philosophers emphasized the essential linguistic articulation of even the most basicperceptual concepts and judgments.[^172] Such judgments, they emphasized, are applied, first and foremost, to the description of objective facts, phenomena, and events, and only secondarily to the “private” phenomena of first-person experience.

In this chapter, I shall explore the historical and methodological implications of this appeal to the “publicity” of linguistic use over against traditional theories of the privacy of experience. When Ryle wroteThe Concept of Mind in 1949, his goal was to employ reflection on the “logical geography” of the ordinary concepts of mind and mentality against the claims of the “official doctrine” tracing to Descartes. This doctrine, with what Ryle characterized as its central dogma of the “ghost in the machine,” presented what to him seemed a strangely divided picture of the mental and physical departments of a human life, treating these as though they were the subject of two largely separate and independent biographies. In response, Ryle suggested a logically unitary analysis of the bearing of “mentalistic” terms on the description of actions and events of ordinary life. A key element of this suggestion was Ryle’s analysis of the terms ordinarily taken to refer to perceptions or sensations as having a “dispositional” logic. That is, rather than referring to special items or object immediately present to consciousness, Ryle suggested that they could be taken simply to attribute various kinds of tendencies, liabilities, and abilities to behave in ordinary perceptual situations. In this way analysis could bring the crucial recognition of the public and intersubjective character of language to bear against the subjectivist theories of mind that treated perception as grounded in the presence of immediately “given” mental objects such as sense-data. These theories could then cede to one that placed the possibility ofattributing dispositions and capacities to perceive at the center of ordinary linguistic practice.

When Sellars took up his own analysis of the language of “inner episodes” in what would become his most famous work,Empiricism and the Philosophy of Mind, he inherited some of the most significant methodological components of Ryle’s inferentialist analysis. Familiarly, Sellars’ largest aim was to dispel what he called the “Myth of the Given,” the myth of unconceptualized and non-linguistic deliverances of experience at the basis of our knowledge of the world. The myth, Sellars thought, had been a central component of empiricist theories of perception and knowledge, both in their classical forms and, more recently, in the early explanatory projects of the analytic tradition. His criticism of it involved both a decisive appeal to the publicity of language as well as, I shall argue, a determinative critical appreciation of the problems to which the attempt to describe first-person experience can lead. But although he shared Ryle’s anti-phenomenalist and anti-subjectivist motivations, Sellars nevertheless saw reason to criticize, in detail, Ryle’s dispositionalist account of sensation and perception. Understanding the reasons for this criticism, I shall argue, helps us to see the broader critical implications of a reflection on language for the specific problems of subjectivity and experience, problems it must encounter in taking up a more general inquiry into the significance of language for the form of a human life.

I

Readers ofThe Concept of Mind have long been familiar with Ryle’s anti-Cartesiandispositionalism about the meaning of many terms of mentalistic description. The core of Ryle’s suggestion is that concepts like those of knowing, believing, intending, and perceiving, can be analyzed in terms of verbal and nonverbal behaviors and capacities, tendencies, and abilities to behave in particular ways. This provides an alternative to the Cartesian assumption that they must as refer to occurent states, processes and events in an inner, mental realm. A good example is the notion of “intelligence.” On the traditional Cartesian picture, Ryle suggests, intelligence seems to be a property of inner acts of thinking or conceiving. Thus, for instance, the Cartesian analyst understands someone’sintelligently playing chess as involving two essentially different kinds of actions: first, an inner, mental act of calculation or intellection (the act properly described as “intelligent,”), and second, a separate physical act of carrying out its result.[^173] Ryle’s suggestion is that the intelligent playing ought to be analyzed, instead, simply as an instance of playing by somebody possessing the familiar background of skills and abilities (dispositions) that make for what we call intelligence in chess-playing. We can investigate the origin of these general skills, and even investigate their physical or neurophysiological basis in the brain. But there is no need to describe the performance as involving a separablemental act which itself has the property or feature of “intelligence.”

Ryle supported his dispositionalism about mentalistic terms with a sophisticated semantic account of the logic and grammar of disposition-ascriptions in intersubjective discursive practice. Here as well, Ryle’s grammatical account steadfastly aims to avoid invoking the existence of esoteric private or inner mental events, items, or structures, even those that can be understood purely physicalistically and neurophysiologically. It can be no part, Ryle reasons, of the ascription of an ability to play chess intelligently or speak French competently that reference is made to any underlying mental or neurophysiological structure. For we need have no knowledge of such structures in order to ascribe these dispositions as we do in ordinary intersubjective practice.[^174] Indeed, in a chapter ofThe Concept of Mind devoted to the nature of dispositionalist analysis, Ryle clarifies that it is no part of his style of dispositionalism to require the ascription of any sort offacts at all (behavioral, neurophysiological, or otherwise). Instead, he offers what can be called anon-factualist account of disposition-ascriptions. On the account, to ascribe a skill, tendency, liability, proclivity (or any other of a variety of specific types of dispositions) is not to report the obtaining of any set of facts, but rather tooperate among fact-statements.[^175] Ryle likens the role of disposition-ascriptions to the role of statements of physical law; both kinds of statements do not, he suggests, state facts but ratherlicense certain patterns of inferenceamong statements of them:

At least part of the point of trying to establish laws is to find out how to infer from particular matters of fact to other particular matters of fact, how to explain particular matters of fact by reference to other matters of fact, and how to bring about or prevent particular states of affairs. A law is used as, so to speak, an inference-ticket (a season ticket) which licenses its possessors to move from asserting factual statements to asserting other factual statements.[^176] (p. 121)

The point of the statement of a causal or logical law, Ryle argues, is simply to allow inference from certain factual statements to other factual statements. Given this analysis, he goes on to suggest, we can see that the thought that the adoption of a law requires the recognition of a new entity (for instance a “causally necessary connection” between the states of affairs connected by a causal law) must be confused. For any such entity could only do the semantic work that the law-statement already does, namely issue a warrant for predictive and explanatory inference from one set of states of affairs to another.[^177]

Like law-statements, Ryle argues, disposition-statements should be understood as having the logical job, not of reporting facts, but of licensing particular kinds and patterns of inference among fact-statements:

Dispositional statements about particular things and persons are … like law statements in the fact that we use them in a partly similar way. They apply to, or they are satisfied by, the actions, reactions and states of the object; they are inference-tickets, which license us to predict, retrodict, explain and modify these actions, reactions and states …

Dispositional statements are neither reports of observed or observable states of affairs, nor yet reports of unobserved or unobservable states of affairs. They narrate no incidents. But their jobs are intimately connected with narratives of incidents, for, if they are true, they are satisfied by narrated incidents..[^178]

On the Rylean account, then, to say of someone that they know French is just tolicense certain inferences, for instance from their being presented with a French telegram to their reading it correctly; to say that Doe knows French is just to say if the antecedent of the inference is fulfilled, its consequent probably will be as well. Ordinary use of disposition-terms does not require that the inferences licensed by these inference tickets be exceptionless; they need only be likely to hold, under normal conditions.[^179] Indeed, the general inference-patterns allowed by particular disposition-statements are, Ryle argues, usually onlypartially satisfied by particular statements of fact. For instance, somebody may be called “irresponsible” after committing a particular error, although he has not in fact committed many of the other errors which the censure predicts he will or may commit, in particular circumstances. Ryle calls dispositional statements that are only partially satisfied by particular occurrences “mongrel-categorical statements,” and suggests that these include most of the dispositional and categorical statements that function in our ordinary language of mental life.[^180]

Ryle’s analysis of the logic of disposition-statements exemplifies particularly clearly the bearing of analysis of the categorical structure of language on questions about what an earlier discourse would have described as our “inner” life. In the particular case of the language of sensations, Ryle makes the dispositional analysis central to his argument against the traditional empiricist or phenomenalist sense-datum theory of perception. His aim is to dispel the thought that the meaning of sensation-language involves its referring to a set of private or proprietary inner objects. To this end, Ryle aims to show that the standard perceptual cases on which the sense-datum theory trades can be analyzed along other lines than the sense-datum theorist suggests. For instance, where the sense-datum theorist is inclined to interpret the sense of “looks” in which a tilted plate “has an elliptical look” as showing theexistence of an “elliptical look” of the round plate (conceived as an inner item or set ofsensa ), Ryle proposes that the case be analyzed along dispositionalist lines:

In other words, the grammatically unsophisticated sentence ‘the plate has an elliptical look’ does not, as the theory assumes, express one of those basic relational truths which are so much venerated in theory and so seldom used in daily life. It expresses a fairly complex proposition of which one part is both general and hypothetical. It is applying to the actual look of the plate a rule or a recipe about the typical looks of untilted elliptical plates, no matter whether there exist such pieces of china or not. It is what I have elsewhere called a mongrel-categorical statement … The expressions ‘it looks …’, ‘it looks as if…’, ‘it has the appearance of …’, ‘I might be seeing…’ and plenty of others of the same family contain the force of a certain sort of open hypothetical prescription applied to a case at hand.[^181]

Where the sense-datum theorist takes the familiar language of perception to involve commitment to the existence of epistemologically primarysensa , Ryle proposes to analyze it instead as involving hypotheticals connecting the ways things look in various conditions of perception to the ways they are. These include, for instance, the rule that tilted round plates often “look like” untilted elliptical ones. Indeed, Ryle argues that the language of looking is itself essentially dependent on the language with which we describe the public properties of publicly ascertainable objects.[^182] To say that somethinglooks a certain way just is to say that it seems as if itis that way, while also recognizing that, owing to nonstandard perceptual conditions or error, it may not be. Our ability to employ the language of looking, then, is just our ability to comprehend the logical and inferential relations among the ways things can be and the ways they can seem to be, given nonstandard perceptual conditions or cases of perceptual error. This language, significantly, stops short of introducing any such entities as “looks,” “appearances,” “seemings,” or “sensings,” hypostatizedevents that begin the traditional theorist on the path toward substantial object-like sense-data.

In fact, Ryle goes on to suggest, the traditional theorist has misunderstood the nature of the logical gap between “looks”-talk and basic reports of how things actually are. According to Ryle, proponents of the traditional model discern rightly that, although knowledge about a public state of affairs will depend in part on simply observing that state of affairs, it will also depend on the satisfaction of further conditions, which may in some special cases fail to obtain. But they mistake thesestanding conditions for occurent processes which are said toaccompany the observation as it happens, or very soon afterward:

When a person is described as having seen the thimble, part of what is said is that he has had at least one visual sensation, but a good deal more is said as well. Theorists commonly construe this as meaning that a description of a person as having seen the thimble both says that he had at least one visual sensation and says that he did or underwent something else as well; and they ask accordingly, ‘What else did the finder of the thimble do or undergo, such that he would not have found the thimble if he had not done or undergone these extra things?’ Their queries are then answered by stories about some very swift and unnoticed inferences, or some sudden and unrememberable intellectual leaps, or some fetching up of concepts and clapping them upon the heads of the visual data. They assume, that is, that because the proposition ‘he espied the thimble’ has a considerable logical complexity, it therefore reports a considerable complication of processes.[^183]

For Ryle, then, talk of perceptions and sensations is to be analyzed as involving the application of learned rules within a logically prior descriptive language. The ability to apply these rules is theoretically inseparable from our mastery of an ordinary language, a kind of mastery we ascribe to anyone we consider to be perceptually and linguistically competent. The normal assumption of perceptual competence is itself essential to our understanding of what is said when agents report the ways things look or seem to them. But this assumption is just the attribution of a disposition, an attribution that we make to anyone who has mastered ordinary perceptual and observational concepts. The most important precondition for justification in issuing the attribution, and the most important component in the judgment that an agent is perceptually competent, is entitlement to suppose the agent’s perceptual and observational reports, when made in standard conditions, accurate. Like the other disposition-attributions that Ryle discusses, the attribution of perceptual competence operates as an inference-ticket, allowing the ascriber to infer from the agent’s observational report to the probability of things being as they are reported to be.

Ryle’s theory, therefore, iseliminativist with respect to at least some of theapparent commitments of sensation-reports and other seeming reports of inner states and processes. According to Ryle, the forms of language that appear to give support to a conception of the inner life can uniformly be analyzed as involvingonly commitments to public and publicly observable facts and their grammatical interrelations. The suggestion of the eliminability of the commitments of the language of the ‘inner’ is itself motivated, most of all, by Ryle’s claim to trace theactual commitments of ordinary language by systematically reflecting on the use of its terms. In thus turning reflection on the systematic structure of language to the criticism of those earlier theories that place individual, subjective experience at the center of their accounts of content, Ryle both continues the methods, and deepens the results, of the critique of psychologism that Frege and Wittgenstein had already pursued. Here, indeed, this critique drives to what is perhaps its most radically formulated bearing against traditional accounts of the ‘interiority’ of experience. Whereas earlier practitioners had been content to criticize philosophically or scientifically specialized psychologistic theories of content, Ryle adduces grounds for thinking that the entire metaphor of ‘interiority,’ as it is used in ordinary discourseas well as specialized philosophical discourses, is without foundation.

At the same time, however, its revisionist suggestions with respect to the apparent commitments of ordinary usage to the description of inner life invite the objection that Ryle’s theory has, in the end, actually failed to capture some of the most ordinarily significant features of this usage itself. In particular, on Ryle’s account, my description of my own sensory state embodiesnothing that could not equally well be recognized from the perspective of another observer. The state of affairs it identifies - including my recognition of the possible nonveridicality of my own perceptual state - is in no sense particularly private or even first-personal. It is a perfectly public, objective matter of the configuration of one’s perceptual devices and abilities. Accordingly, on Ryle’s theory, what appears to be the report of a sensation has no special claim to be true if issued in the first person; it is simply the description of a perceptual state of affairs, and may as well be taken to be true fromany perspective.

But it is a familiar feature of our ordinary language of sensation, and indeed of all first-person reports of experience, that the reporterdoes enjoy a special epistemic and semantic privilege in making the report. Such reports are routinely entitled, in ordinary intersubjective linguistic practice, to a default assumption of accuracy; indeed, it is not even obviously coherent to assume that onecan be mistaken about one’s own present sensations.[^184] And even if Ryle can reduce the first-personuses of sensation language to correspondent first-person uses of perception language involving talk of “looks,” the Rylean theory has no account, in either case, of the special authority - the default claim to be taken true - that characterizes them. For all Ryle says, the locutions that seem to report on the existence of perceptions and sensations might as well be empirical descriptions of one’s own perceptual states, enjoying no greater antecedent claim to truth than any other empirical description. Ryle’s theory fails, in other words, to construe sensation-reports genuinely asreports : declarative utterances that are, if true,caused in part by the states of affairs that make them true. Instead, on Ryle’s theory, sensation-reports are actuallydescriptions of the objective perceptual situation of the perceiver. This, however, seems to badly mischaracterizes the semantics of these locutions, with precipitous consequences for the place of first-person experience in Ryle’s theory.

II

At first glance, Sellars’ theoretical aims inEPM with respect to sensations and sensation-reports seem strikingly similar to Ryle’s. Both philosophers want to dispel the theory of sense-data, both in its Cartesian and contemporary forms, and the empiricism in which it figures. They both aim to defeat the sense-datum theorist’s conception of sensations as epistemologically ultimate by emphasizing the role of conceptual training as a precondition to even the simplest sensation-reports. Such reports can no longer be treated as epistemologically foundational, both philosophers argue, once thelinguistic preconditions for their use are appreciated. Both recognize as central to their own projects, moreover, the analysis of our ordinary language of observation, perception, and sensation. In particular, the actual epistemological significance of seeming perceptual and sensory reports is to be determined by their ordinary linguistic roles, as this is shown in everyday use. But on at least one centrally important point, Sellars’ theoretical aim is different from Ryle’s, and indeed arises, in part, from a direct criticism of the limitations of Ryle’s program. For Sellars is centrally concerned topreserve , rather than dispute, the meaningfulness of the forms of language with which we seem to refer to inner processes and episodes such as sensations.

If we insist, as Sellars and Ryle both do, that only public objects (and never sensations or impressions) literally have properties like being red or triangular, we can easily, Sellars suggests, be led to think that we never directlyrefer to such inner items at all. On this line of argument, we can only characterize them indirectly, via definite descriptions, in the language of public events and properties. But in following this line, Sellars notes, “we would scarcely seem to be any better off than if we maintained that talk about ‘impressions’ is a notational convenience, a code, for the language in which we speak of how things look and what there looks to be.”[^185] And in explicit criticism of Ryle, Sellars argues that any theory that follows this line will fail to account for important features of our ordinary discourse:

Indeed, once we think this line of reasoning through, we are struck by the fact that if it is sound, we are faced not only with the question “How could we come to have the idea of an ‘impression’ or ‘sensation?’ but by the question ‘How could we come to have the idea of something’s looking red to us, or,’ to get to the crux of the matter, ‘of seeing that something is red?’ In short, we are brought face to face with the general problem of understanding how there can be inner episodes - episodes, that is, which somehow combine privacy, in that each of us has privileged access to his own, with intersubjectivity, in that each of us can, in principle, know about the other’s.[^186]

Sellars’ criticism of Ryle is subtle and far-reaching. Ryle’s dispositionalist account of the language of “impressions” and “sensations” treats this language as a shorthand, a notational replacement for discussion of how things look or seem to us. But even if Ryle can reduce the language of “sensations” and “impressions” in this way, the special logical features of the supposedly anterior language of ways of looking or seeming themselves stand in need of explanation. Reducing language about sensations and impressions to language about ways of looking and seeming simply pushes the problem back. We still lack an account of the distinctive kind of authority that inner-state reports, whether reports of sensation or of perception, can have. Sellars insists that there will be no way to solvethis problem without confronting the issue of inner episodes: without, that is, discovering how there can be items or events that be, at the same time, bothdescriptive objects of public language andepisodic sources of our reports:

We might try to put this more linguistically as the problem of how there can be a sentence (e.g. ‘S has a toothache’) of which it is logically true that whereas anybody can use it to state a fact, only one person, namely S himself, can use it to make a report. But while this is a useful formulation, it does not do justice to the supposedly episodic character of the items in question. And that this is the heart of the puzzle is shown by the fact that many philosophers who would not deny that there are short-term hypothetical and mongrel hypothetical-categorical facts about behavior which others can ascribe to us on behavioral evidence, but which only we can report, have found it to be logical nonsense to speak of non-behavioral episodes of which this is true. Thus, it has been claimed by Ryle that the very idea that there are such episodes is a category mistake, while others have argued that though there are such episodes, they cannot be characterized in intersubjective discourse, learned as it is in a context of public objects and in the ‘academy’ of one’s linguistic peers. It is my purpose to argue that both these contentions are quite mistaken, and that not only are inner episodes not category mistakes, they are quite ‘effable’ in intersubjective discourse.[^187]

Without an account of how seeming reports of “sensations” and “impressions” can genuinely be reports of inner episodes, Sellars suggests, we will be unable to capture the logical features of these reports that account for their functioning, in intersubjective discourse, as they do. In particular, we will lack an account of how these seeming reports can be reports of happenings that seem in a certain way proprietary to their bearers, a status which is recognized in the default assumption of truth that reports of them enjoy when issued in the first person.

Sellars mentions his aim of preserving the meaningfulness of discourse about inner episodes often enough to show that it is one of the main theoretical goals of his account inEPM . He cites the explication of the “logical status of impressions or immediate experiences,” for instance, as the main purpose of the famous Myth of Jones. With this reconstructive story, Sellars aims to show how a group of people initially limited to a “Rylean” language capable only of referring to the public properties of public objects could, once given the resources to discuss thesemantic properties of their own language, develop a mode of discourse about inner episodes and states with all the logical features of our own inner-state language.[^188]

Indeed, Sellars suggests that his attack on the Myth of the Given will not really succeed unless it can preserve at leastsome of the theoretical motivations, implicit in ordinary language, that provided support for the traditional empiricist picture of sensation-reports as representing a semantically and epistemologically special stratum of knowledge. For Sellars as for Ryle, the meaningfulness of observation-reports depends upon their being generated by a reporter with the ordinary perceptual and conceptual abilities of a competent adult observer. What is ascribed in ascribing these abilities is itself at least partially comprehensible in terms of their inferential articulation.[^189] To judge someone’s perceptual report meaningful, then, is, at least in part, just to ascribe them the normal suite of perceptual and conceptual abilities, which in turn is just to issue the kind of inference-ticket that such ascription involves. But for Sellars, the ascription of competence that figures in the authority of perceptual and observational reports is notsimply the issuance of an inference-ticket. For in addition to the intersubjectively ascertainable reliability that perceptual competence involves, Sellars insists that the meaningfulness of an agent’s perceptual and observational reports depends, as well, on the agent’sknowledge that her reports are normally reliable.

This additional requirement of knowledge goes beyond anything that Ryle’s theory demands or suggests. Sellars insists on it - what I shall call Sellars’knowledge requirement - at several points inEPM . He formulates it most directly in section 35:

For if the authority of the report ‘This is green’ lies in the fact that the existence of green items appropriately related to the perceiver can be inferred from the occurrence of such reports, it follows that only a person who is able to draw this inference, and therefore who has not only the concept green, but also the concept of uttering “This is green” - indeed, the concept of certain conditions of perception, those which would correctly be called ‘standard conditions’ - could be in a position to token ‘This is green’ in recognition of its authority. In other words, for aKonstatierung ‘This is green’ to ‘express observational knowledge,’ not only must it be a symptom or sign of the presence of a green object in standard conditions, but the perceiver must know that tokens of “This is green” are symptoms of the presence of green objects in conditions which are standard for visual perception.[^190]

The requirement might seem innocuous, but actually it represents an important divergence from Ryle’s account and a key element of Sellars’ own argument against the Myth of the Given. Indeed, Sellars goes on to say that the requirement is essential to showing that observation reports are not epistemically basic in the sense in which traditional empiricism takes them to be:

Now it might be thought that there is something obviously absurd in the idea that before a token uttered by, say, Jones could be the expression of observational knowledge, Jones would have to know that overt verbal episodes of this kind are reliable indicators of the existence, suitably related to the speaker, of green objects. I do not think that it is. Indeed, I think that something very like it is true. The point I wish to make now, however, is that if itis true, then it follows, as a matter of simple logic, that one couldn’t have observational knowledge of any fact unless one knew manyother things as well. And let me emphasize that the point is not taken care of by distinguishing between knowing how and knowing that, and admitting that observational knowledge requires a lot of “know how.” For the point is specifically that observational knowledge of any particular fact, e.g. that this is green, presupposes that one knows general facts of the formX is a reliable symptom of Y . And to admit this requires an abandonment of the traditional empiricist idea that observational knowledge “stands on its own feet.”[^191] (pp. 75-76)

Even if the meaningfulness of observation reports can be analyzed in terms of their issuers’ possession of standard perceptual and conceptual abilities, it is essential to Sellars’ story that it additionally involves the reporter’s possession of general knowledge about the reliability of particular sentence tokens in reporting particular states of affairs. This knowledge is not, as Ryle would have it, explicable simply as “knowledge-how” in contrast to “knowledge that.” That is, it cannot be explained simply as a matter of our possession of various kinds of ability or dispositions. It is this that shows that,contra the epistemological foundationalist’s theory of them, “basic” observation-reports already presuppose a substantial amount of general knowledge, and so cannot be the ultimate basis of empirical knowledge in the way foundationalism takes them to be.

To understand the reason for the requirement, it is helpful to reflect on its setting within Sellars’ critical discussion of a specific form of epistemic foundationalism. Sellars states the requirement in the course of his reconstructive discussion of the logical empiricist view according to which observation reports are immediate reports on a stratum of inner and private experiences, non-verbal episodes which are held to beself-authenticating in that their authority does not rest on anything but themselves.[^192] On the view, which was held most closely by Schlick, basic observation reports are the immediate expression of the content of more primary experiential episodes calledKonstatierungen . TheKonstatierungen themselves were taken to be infallible, and to have a non-verbal or pre-verbal logical form. The authority of basic observation-reports was then thought to derive, as is suggested by their usual inclusion of indexical or token-reflexive expressions (terms like “I” and “now,” as in what was taken to be the standard form of a sense-datum report, “I am having a red sensation now”), from their being made in thepresence of the experiences on which they report. The kind of authority, or claim to be believed, that these reports have would then be essentially different from the kind of authority possessed by other propositional claims. For whereas the authority of most sentence-tokens can be understood in terms of theinferential role of the contents they express - the kinds of evidence that support them, and the kinds of inferences that can be drawn from them - the authority of observation reports would depend instead on theconditions under which tokens of them are issued.

Thus, whereas propositional authority ordinarily flows fromsentence-types tosentence-tokens (so that a token sentence is authoritative in virtue of its being a token of the sentence-type that it is a token of) the epistemic use of observational reports would involve a kind of authority that flows in the opposite direction, fromsentence-tokens authorized by theirconditions of utterance (and involving indexicals) tosentence-types expressing nonindexical observational contents. This would give them a kind of ultimate credibility, a credibility that does not depend on their relation to other propositional contents or tokens. Their having this kind of credibility, it is then reasonable to think, would be essential to their claim to express observational knowledge at all, knowledge that is the direct outcome of observational processes rather than inference from other propositions. At the same time, it would be comparable to the ultimate credibility of analytic statements, statements that are true simply in virtue of the conventional rules of linguistic usage. This parallel suggested to Schlick and others that the correctness of an observational report like “This is green” ought to be thought of as depending simply on one’s following the “rules of use” for the term “green” and the indexical term “this,” rules of use that call for the utterance of the observational report when, and only when, the requisiteKonstatierung is present.

Unsurprisingly, Sellars rejects Schlick’s view itself as an instance of the Myth of the Given. The postulatedKonstatierungen would comprise an ineffable stratum of immediate and self-authorizing, linguistically ineffable but somehow semantically contentful episodes.[^193] They are, in other words, typical instances of the unexplained Given; and the consistent point of Sellars’ insistence on the social and linguistic preconditions for these basic reports is that these reports do not rest, conceptually or evidentially, exclusively on any suchineffable episodes.[^194] But though his diagnosis of the Myth involves his showing that ineffable inner episodes cannot be the foundations of empirical knowledge, Sellars emphasizes that his account nevertheless aims to make room foreffable inner episodes that are in a sense non-linguistic.[^195] As Sellars’ subsequent remarks clarify, in fact, its ability to do so actually arises from its preservation of one of the most important components of Schlick’s view, the suggestion that observation-reports have the authority that they do in virtue of their being made under the particularconditions that they are.

Immediately after rejecting Schlick’s view as an instance of the Myth of the Given, Sellars considers what kind of theoretical view of observation reports, though purged of Givenness, might still capture their claim to express genuine observational knowledge. He begins by contemplating a standard reliabilist account, according to which

An overt or covert token of ‘This is green’ in the presence of a green item is aKonstatierung and expresses observational knowledge if and only if it is a manifestation of a tendency to produce overt or covert tokens of “This is green” . if and only if a green object is being looked at in standard conditions.[^196]

The reliabilist account shares with Schlick’s account the guiding idea that observation reports have the authority that they do have, an authority that flows from sentence-token to sentence-type, because of the conditions under which they are issued. The difference is just that the reliabilist account replaces Schlick’s inner, experiential episodes with the presence of a general disposition - a reliable tendency - to produce tokens of the right types in the right circumstances. On the reliabilist account, it is in virtue of their being an instanceof such a tendency that observation-reports have authority. But Sellars concludes that the unaugmented reliabilist account still fails to preserve the claim of observation reports to express genuine observational knowledge. For this claim to be preserved, the account must be supplemented with two additional conditions.[^197] First, Schlick’s understanding of the credibility of observation-reports as involving the correct following, by the reporter, of semantic rules, has to be replaced with an essentiallysocial account of the kind of correctness that authoritative observation-reports exhibit.[^198] In other words, the authority of observation-reports must be seen as deriving not only from an individual’s following linguistic “rules of use” but from their (in so doing) reflecting reliable tendencies that are intersubjectively ascribed in a linguistic community.

As they were for Ryle, these tendencies are intelligible as dispositions, demonstrated and attributed in a social context. But second, in addition to this social requirement, Sellars also adds the requirement ofknowledge of reliability that we’ve already discussed. This knowledge of reliable connections between reports and the inferences they make possible that Sellars refers to in accounting for the authority of first-person experiential reports plays a central role inEPM , and particularly in the concluding Myth of Jones. The most decisive chapter in the story is the invention of atheory of internal processes by the genius Jones on the model of semantic discourse. The Rylean ancestors become able to talk about each other’s thoughts by internalizing the language with which they formerly talked about the meaning and truth of each other’s public statements.[^199] Given these semantic resources, Jones invents a “theory” of inner episodes on the model of overt verbal behavior, semantically characterized.[^200] It is, in fact, essential to his ability to invent the theory that the inner states characterized by it can be described using the same semantic predicates as can already be used to talk about overt utterances. It is essential, in other words, to theidentity of the postulated inner states that they can be described as*“* meaning” this or that or being “about” this or that.[^201] Without this possibility of description, they would not be intelligible as the states that they are; but the possibility of reporting their identity is itself dependent on the application to them of the concepts ordinarily used to describe objective phenomena and events. They inherit their content from the common content of the public utterances that exemplify the perceptual and observational reports that can be made in the various situations where we describe things as looking or being thus-and-so, and therefore depend constitutively on the subject’s ability to issue those reports in standard and nonstandard perceptual conditions. And that ability, together with all the conceptual and semantic knowledge it implies, is just what is required by the knowledge requirement.

The knowledge of reliable usage that Sellars requires as the background of our use of perceptual concepts, if we are to be counted as competent users of these concepts at all, thus plays a central role in enabling his theory to give a plausible account of the authority of first-person reports. In earlier accounts such as Schlick’s that did not share Sellars’ commitment to a wholly public and intersubjective account of the acquisition and use of first-person language, the reference and authority of such reports was explained by reference to the immediate deliverances of Givenness. For Sellars, though, a speaker’s knowledge that a semantic token such as “green” can (reliably) be used in referring to green objects is itself sufficient to ground the authority that the earlier theories relied on the givenness of greensensa orKonstatierungen to explain. This knowledge about reliable use issemantic knowledge; where present, it counts as conferring on its possessor knowledge of themeaning of the term “green” in ordinary discourse and practice. It is knowledge that we can be expected to have, inasmuch as we speak a language at all, and which would be inaccessible to us if we did not. Our attribution of it to an agent expresses, in other terms, the judgment that that agent is a member of the linguistic community in which we, ourselves, live or can live.

III

We have seen that Sellars’ insistence on the knowledge requirement figures essentially in his claim to defeat traditional empiricism by showing that even basic perceptual reports already logically require a substantial amount of general knowledge. This explanatory connection between semantic knowledge and inner discourse has its home, beyond and beforeEPM , in a broader Sellarsian project ofpure pragmatics . This project understands semantical concepts as supporting logical structure in virtue of theirlinguistic roles and aims to solve traditional philosophical problems, including the traditional “mind-body” problem, through a characterization of the “pragmatics” of their use. Sellars articulated the project in a variety of articles over the first half of the 1950s; but it is sketched only partially and elliptically inEPM itself.[^202] We shall see that an appreciation of the contours of the program clarifies the role of the questions Sellars addresses in that work within the larger history of the methods and practices of analytic philosophy’s reflection on ordinary language, and of the problems to which the characterization of its logical structure is prone.

Starting in the late 1940s, Sellars offered “pure pragmatics” as a supplement to the existing formal characterizations of syntactic and semantic notions. The supplement would bepure in that it would retain the non-factual anda priori character of existing formal analyses. But it would bepragmatic in that it would give a formal analysis of semantic predicates like “meaningful” and “verified,” predicates whose adequate analysis would require a formal explanation of how anentire language or the large subset of one that is “empirical” can be meaningful at all.[^203] Thus, pure pragmatics would comprise a “pure theory of empirically meaningful languages,” and its formal analysis would display the pragmatic conditions that are required for any speech behavior to amount to empirically meaningful language at all.[^204]

Sellars’ clearest pre-EPM application of the program of pure pragmatics to the problems of philosophy of mind is the 1953 article “A Semantical Solution to the Mind-Body Problem.” The article is historically significant in its own right, for it offers what can be understood as the earliest suggestion in the philosophical literature of afunctionalist theory of mental-state terms. On such theories, these terms are understood as meaningful in virtue of the patterns of use that define theirconceptual roles and thereby identify the semantic “place” of the mental events to which they refer.[^205] The core of the article’s analysis is a consideration of the possibility and implications of a “behaviorist” analysis of mental terms like “thinks” and closely connected semantic terms like “means.”[^206] As inEPM , Sellars does not defend a (Rylean) “logical behaviorism,” according to which mentalistic and semantic discourse would be logically reducible to discourse about the behavior of bodies.[^207] Instead, he sketches a “scientific behaviorism” that would uphold truth-functional or material - but possibly empirical anda posteriori - equivalences between mentalistic statements and statements characterizing only bodily behavior.[^208] And as inEPM , Sellars begins with the thought that these equivalences, if they obtained, would analyze mental terms by identifying them with dispositions to behave and episodesqua expressing such dispositions. Smith’s thought that it is raining outside might, for instance, be identified by a behavioristic psychology with Smith’s tendency to behave in particular ways, for instance his behavior of reaching for his umbrella and putting on his raincoat, along with, of course, the tendency to utter the linguistic expression “it is raining” and other suitably connected expressions.[^209] But it is immediately clear that the last-mentioned kind of tendency poses additional problems for the behaviorist analysis. For the requirement of it is not just the requirement that Smith tend to utter a particular set of noises (it could be just as well satisfied, if Smith were a German speaker, by his tendency to utter the completely different set of noises “es regnet”) but that he utter a set of noises which meanit is raining .[^210]

This difficulty marks the essential difference of Sellars’ semantically based account from the behaviorist theories he criticizes. Such theories, insofar as they are genuinely behaviorist, are limited to describing verbal behavior in terms of the actual utterances issued and the normal occasions of their utterance. But a genuinely explanatory semantics, Sellars realized, would have to describe not only the utterances themselves but also what an earlier age of philosophical reflection would have characterized as their “meanings.” That is, it would have to characterize the abstract features (as it seemed to Sellars) of theiruse that make it possible for two tokens of a language on different occasions, or two wholly different utterances in different languages, to mean “the same thing.” Within the broader program of pure pragmatics, this required that the analyst give an account of the semantic and pragmatic functioning of the predicate “means,” and Sellars now undertakes to provide a description of this functioning. He exploits the central idea that to describe a sentence asmeaning thus-and-so is to characterize it as occupying a particularrole in the cognitive economy of the speaker. The predicate “means” has the metalinguistic use of gesturing at this kind of role, a role which may be shared by several different sentence-types across different particular languages. The problem, as it now stands, concerns the implications of a behaviorist analysis of sentences of the form “Smith utters ‘es regnet’ where ‘es regnet’ meansit is raining ” into sentences purely about behavior. If the behaviorist analysis is possible, Sellars argues, it will issue in equivalences of the form

“Es regnet” uttered by b meansit is raining <-> Ψ (“es regnet”, b)

where the right side of the biconditional “says of b that it has certain habits relating its utterances of ‘es regnet’ to other utterances, to other habits, and to sensory stimuli.”[^211] As the predicate “means” is generally used, “es regnet” can mean the same thing when uttered by a German-speaker as “il pleut” means when uttered by a French speaker; so we can take it that the habits of the German-speaker with respect to “es regnet” share a “common generic feature” with the habits of the French-speaker with respect to “il pleut.” Thus, if the behaviorist identification is possible, we can write the general schema:

“…” uttered by b meansit is raining <-> K(“…”, b)

where K(“…”,b) says that b has the particular habits concerning “…” that qualify it, when uttered by b, to meanit is raining .[^212] In other words, the right-hand side of the biconditional says that “…”occupies the particular pragmatic and conceptualrole in b’s cognitive economy that makes it an utterance meaningthat it is raining. Its occupying this role can only be understood as its instancing the generic tendencies and habits which qualify b as a competent user of the phrase “it is raining.” Viewed from another direction, of course, these generic tendencies and habits are just those that qualify b as a competent verbal reporter ofrain .

The suggested analysis of the semantic term “means,” then, analyzes the assertion that an utterance has a particular meaning as the assertion that it occupies a particular semantic role in the cognitive economy of a speaker, or, equivalently, that it is a manifestation of particular behaviorally comprehensible habits and dispositions. But it is essential to the pragmatic character of Sellars’ suggestion about the nature of mentalistic terms that one cannot, in general,specify the semantic role in questionexcept by issuing, in one’s own language, a token utterance that occupies it. Equivalently, one cannot, in general,specify the habits and dispositions that a meaningful token of a particular content must manifest, without issuing a token utterance that itselfmanifests those very habits and dispositions:

Now we are all familiar with the fact that when we say ‘Jones’ utterances of ‘es regnet’ means it is raining’ we are mentioning ‘es regnet’ and using ‘it is raining’ to convey what is meant by ‘es regnet’ as uttered by Jones. According to Scientific Behaviorism, if what we say of Jones’ utterances is true, then the utterance ‘it is raining’ which weuse is the manifestation of habits generically identical with Jones’ habits with respect to ‘es regnet’. Thus, when I utter

“Es regnet” uttered by b means it is raining <-> K (“es regnet”, b)

the “it is raining” of the left hand side is a manifestation of the habits mentioned by ‘K (“it is raining”, Sellars)’, and when I utter

“It is raining” uttered by Sellars meansit is raining <-> K(“it is raining”, Sellars)

the unquoted “it is raining” on the left hand side is amanifestation of the habits mentioned by the right hand side.[^213]

Sellars thus emphasizes that a description of a speaker’s utterance as meaning thus-and-so canconvey information about the semantic role of the utterance by comprising an utterance thatoccupies the same general role, without involving anydescription of the role itself. Indeed, this possibility of showing without saying is essential to the ordinary functioning of the predicate “means” itself. The most typical and basic use of this predicate is in connection with judgments of the form “ ‘a’ means that b.” Such judgments can compare propositions in two different languages, or they can compare two different formulations in the same language of (what is thereby asserted to be) the same fact. When we issue them, we are, as Sellars says, essentiallydemonstrating a semantic role by instantiating it. Our assertion of identity of meaning will be understood only by someone who already possesses the habits and dispositions thereby instantiated.

This point is central to Sellars’ suggestion of a kind of behavioristic analysis of mentalistic terms that stops short of the logically necessary identities that would be required by “logical” (as opposed to “scientific”) behaviorism. On the analysis, the ascription of mental states in ordinary discourse (for instance the determination of the character of an agent’s thoughts) depends, in general, on the possibility of characterizing those states as having propositional meanings, and thus on the pragmatic possibility ofconveying withoutspecifying the conceptual roles of the utterances they are modeled on. The identities of meaning in which the analysis issues do not, then,reduce mental states to behaviors and dispositions. Rather, the analysis shows how discourse about mental states can be understood as discourse ‘about’ behavioral dispositions and habits, in the special pragmatic sense of ‘about’ in which a semantic sentence can be ‘about’ a semantic role by exemplifying without specifying that role.[^214] This kind of exemplification cannot be understood, in general, except through a pragmatic description of the capability of semantic discourse to show or exhibit what it does not explicitly state. Accordingly, it is a consequence of the suggested analysis that a particular mental state can be construed as meaningful only by an interpreter capable of employing utterances with thesame linguistic role as that occupied by the expression of that state:

While we can convey how Jones uses ‘es regnet’ by the use of “’es regnet’ uttered by Jones means it is raining’ only to someone who shares our habits with respect to ‘it is raining’, we can convey this information even though neither of us has a ‘clear and distinct’ idea of what these habits are, and even though neither of us is able to characterize these habits without the repeated use of statements of the form “S means”, and indeed of the form “in Jones’ mind there is a thought about”.[^215]

Because semantical pragmatics is (loosely put) a matter ofshowing rather than saying, only someone capable of meaningfully making an utterance can describe the same utterance, when issued by another agent, as meaningful. And only someone capable of occupying the particular meaningful mental state at issue can describe someone else as being in that very same (type of) state.

With this “semantical solution” explicitly in mind, we can understand just how the kind of semantic knowledge embodied by Sellars’ knowledge requirement inEPM makes possible a description of the logic of inner-episode descriptions that avoids the logical reductionism of Ryle’s account. Sellars explicitly employs the semantic solution, and to its special pragmatic sense of ‘implication’, in explaining how the Jonesian myth can show that semantic discourse about inner states need not be reduced to a definitional shorthand or notational variant of the original Rylean language limited to the description of behavioral states and dispositions:

And let me emphasize … that to make a semantical statement about a verbal event is not a shorthand way of talking about its causes and effects, although there is a sense of ‘imply’ in which semantical statements about verbal productions do imply information about the causes and effects of these productions. Thus when I say “’Es regnet’ means it is raining,’ my statement ‘implies’ that the causes and effects of utterances of ‘Es regnet’ beyond the Rhine parallel the causes and effects of utterances of ‘It is raining’ by myself and other members of the English-speaking community. And if it didn’t imply this, it couldn’t perform its role. But this is not to say that semantical statements are definitional shorthand for statements about the causes and effects of verbal performances.[^216]

The Rylean analysis commits itself to a dispositionalist understanding of inner-state descriptions in terms of the semantic roles occupied by their typical verbal expressions, and thus, at least in principle, to an ultimate reduction of the language of inner-state descriptions to the language of causal descriptions of behavioral dispositions. For instance, Ryle’s analysis understands the attribution, to Jones, of the thought that it is raining as simply the attribution of a particular disposition to Jones. This attribution is itself simply the judgment that Jones will issue a token utterance with a particular semantic role under particular conditions. On the Rylean analysis, this semantic role can be characterized simply in terms of the kinds of situations which causally elicit the utterance, so the Rylean analyst commits herself, at least in principle, to the view that descriptions of inner states must be reducible to the purely causal description of the typical causes and effects of particular verbal utterances. But Sellars’ semantic solution shows how the semantic role of an utterance might be pragmatically identified in semantic discourse without anyspecification of its typical causes and effects. It thereby makes room for the possibility of a language for the description of inner episodes thatarises (as in the Myth of Jones) from the Rylean causal language when that language is supplemented with semantic discourse, without beingreducible to causal or causal-plus-semantic language. On the Sellarsian story, this inner-episode language will betheoretical in the sense of having been introduced for the purposes of explaining and predicting overt verbal and semantic behavior. But it will also exhibit, in its positing of inner states as new entities, a descriptive autonomy that makes it irreducible to those earlier strata of descriptive language.

On Sellars’ account, then, the issuance of an inner-state report is not simply the issuance of a redescription of one’s own behavior or behavioral dispositions. It is, instead, the issuance of a piece of semantic discourse, discourse that essentially exploits the special logical features of the predicate “means”. And as such, it is a piece of discourse which, if understood, must be understood asshowing (or “implying,” in Sellars’ special pragmatic sense of “implies”) the existence of a complex semantic disposition that the listener herself possesses. To be able to understand such a report, then, implies knowledge not only of the normal occasions of its production in an individual’s habitual behavior, but also of the ordinary circumstances of the use of its constituent terms in the linguistic practice of a community as a whole; this latter kind of knowledge marks its understanding as a manifestation of the ability to speak a language that one shares with such a community, insofar as one is a member of it. Their essential exploitation of semantic discourse gives first-person inner-state reports (like first-personal semantic discourse generally) a kind of authority that flows from token to type rather than type to token; the comprehension of the token essentially involves the recognition thatthat token was produced in the right sort of way, and thus endows it with a default presumption of truth. The token sentence itself thus bears the presumption of its truth in the conditions of its comprehension.[^217]

In comparison with Ryle’s account, as well as other applications of the methods of inferentialism and holistic analysis to the problems of “philosophy of mind” and subjective experience, Sellars’ semantically based account therefore goes some way to restoring something like a theory of the authority and privilege of the subject. But what is most remarkable about Sellars’ account, in the perspective of a broader history of the methods of linguistic analysis and reflection, is not simply its capacity to restore some of the ordinary logical features of first-person description and reporting by means of an appeal to our knowledge of a language. It is, rather, the pervasive and essential ambiguity it demonstrates in the form of this knowledge itself. Indeed, with Sellars’ semantic account, our ordinary knowledge of the language that we speak is shown to be capable of grounding ordinary attributional practiceonly insofar as it is opaque to theoretical description. Not only the authority of first-person reports, but indeed the entire possibility of semantic discourse on which it is based, depends on our ability toinstantiate ordisplay our knowledge of the regularities of a language without further describing them. The semantic roles underlying the use of ordinary terms might indeed be describable within a total structuralist description of the language as a whole, but such an account, like the Rylean one that Sellars criticizes, would make the distinctive authority of first-person accounts inexplicable. In place of those theories that appealed to an ineffable subjectivity to ground first-person authority in what were conceived as the deliverances of pre-conceptual givenness, therefore, Sellars appeals to the capacity of language to refer to, byinstantiating , its own regular structure. But the account makes the structure of language, again, essentially ineffable on the level of its explanatory theoretical description. The Given mental objects or sense data whose mute presence earlier theories placed at the foundation of the possibility of knowledge are thereby replaced, as promised, with the understanding that we can be taken to have, and regularly appeal to, insofar as we speak a language at all. But the mystery of what is involved in this understanding is by no means cleared up. Rather, the linguistic roles whose demonstration is, according to Sellars, the essential basis for any possibility of semantic discourse now themselves amount to theoretically ineffable objects of a recurrent and essential appeal.