A Retrospective Replacement for Meaning Rules and Rule-Following

John Dilworth, draft only, November 11, 2005

Powerful skeptical arguments by Wittgenstein, Kripke and others strongly suggest that there are no mental rules of meaning that we follow while thinking, believing and perceiving. Nevertheless, virtually all commentators have attempted to resist such far-reaching skeptical conclusions. But I argue that a more promising strategy for naturalists, in pursuit of a purely dispositional theory of meaning, is simply to accept the skeptical conclusions. Then, as a substitute for the eliminated rules and rule-following, an appropriate dispositional replacement can be provided.

The suggested replacement is retrospective, in that it is based on prior cases of behavioral conformity to conventional applications of a supposed rule, rather than on prospective mental envisaging of a rule to be followed, as in standard non-skeptical accounts. It is also shown how adequate replacements for concepts of normative rule-following, perception, and propositionally based true or false belief can be developed in this skeptical retrospective framework.

Any putative theory of meaning has to come to terms with standard criticisms of the possibility of such accounts. One such is provided by Kripke 1982, who presents a view of Wittgenstein's discussions of rule-following according to which Wittgenstein concludes that there is no fact of the matter as to what anyone ever means by a term, or as to which particular rule they follow in using it. In particular, naturalistic dispositional theories of meaning come in for special criticism. Also, to reinforce Wittgenstein and Kripke's own attacks, Paul Boghossian 1989 has drawn out of them several apparently fatal flaws in such views. If Wittgenstein, Kripke, Boghossian and others were correct in such criticisms, then apparently, dispositional or other naturalistic analyses of meaning would be fundamentally useless for philosophical or scientific purposes. (For a comprehensive collection of articles on these issues see Miller and Wright 2002).

However, there is an alternative strategy available for dispositional accounts of meaning. It is to a) simply accept the skeptical conclusion that there are no rules of meaning, or genuine cases of rule-following; and then b) provide a minimal kind of naturalistic dispositional replacement for these discredited, meaning-related concepts. This is the strategy that will be followed in this paper.

As initial support for this replacement strategy, if skepticism about meaning rules and rule-following is justified, then attempts to give reductive dispositional accounts of these concepts must be a waste of time, since they are, according to the skeptic, fundamentally incoherent concepts, which need to be eliminated rather than reduced. Hence, to the extent that the skeptical attacks on rules and rule-following themselves are successful, the ancillary attacks on dispositional accounts of meaning by Kripke, Boghossian et al are beside the point, and it would be equally pointless to try to defend reductive dispositional analyses of meaning rules and rule-following under those conditions. So dispositionalists about meaning should either adopt some form of the present replacement strategy, or undertake, as a necessary preliminary, the herculean task of defending traditional intuitions about rules and rule-following from the skeptical attacks--an unpromising strategy indeed.

1. Inner Processes and the Incoherence of Rule-Following

Many of the skeptical issues raised by Wittgenstein, Kripke and others have their origin in conceptions of meaning or meanings as inner mental processes of various kinds. For example, it is commonly assumed that in using a concept of addition, one is mentally following the rule for that concept of 'plus', rather than for a related concept of 'quus', which diverges from 'plus' under certain conditions (Kripke 1982).

But questions can be raised as to how one knows that one is following one rule rather than the other--an epistemic rule indeterminacy or rule incoherence problem. Since such indeterminacy skeptical objections would apply whether or not inner rule-following were naturalized, one could not resolve it simply by naturalization, no matter how convincing one's account might be. Also, such skepticism concerning rule-following inevitably leads to skepticism about the existence of rules themselves. For if it makes no sense to suppose that there are any knowable, fully determinate mental rules that anyone ever follows, then any ontological grounds for postulating the existence of such fully determinate rules, and of well-defined mental activities of following them, is also undercut.

In addition, there is another standard problem with simple dispositional attempts to naturalize mental rule-following. For if following a rule R were simply to have a disposition to follow or conform to R, then the distinction between correct versus incorrect following of the rule would be obliterated, since any causal manifestation of the disposition would automatically count as correct (Kripke 1982, Boghossian 1989). Hence, though dispositions--or some related, purely causal construct--would inevitably have to play a central role in any naturalization of meaning, clearly their role would have to be a more complex and indirect one than could be provided by this initial simplistic account.

One standard attempted solution to the current rule indeterminacy problems is to move to a social, rather than inner, account of rule-following. (Kripke 1982 calls this the 'skeptical solution' to indeterminacy problems). On such an account, rule-following is explained in terms of social institutions and social agreement as to what counts as correct versus incorrect rule-following. However, several writers have objected than any such account is circular, since essentially the same rule indeterminacy problems would recur at the social, rather than at the inner individual, level (e.g., Blackburn 1984). So some different way of dealing with the problems is required.

2. A Retrospective Replacement for Rules and Rule-Following

One little-tried method of coming to terms with the skeptical problems is to simply accept the implied conclusion that strictly speaking, there are no rules, nor any rule-following. If so, one must also accept that consequently, there cannot be any mental content that involves the following of rules either (on which see section 6). However, I shall show that a substitute conception of rules and rule-following can be constructed, which can avoid many of the potentially dire consequences of accepting the skeptic's conclusions.

The basic idea is that an alternative kind of social practice can be defined, which is not rule-following as such, but which has enough similarities to the eliminated rule-following to often provide an effective substitute for it. The practice will be called retrospective rule-following (RRF), with the understanding that RRF is not a species of genuine rule-following--of which there is none--but instead an ersatz, substitutive replacement for it.

As a preliminary, though there are no rules, there are conventional representations of rules. For, just as we can have pictures of Santa Claus without Santa Claus having to exist, so also we can linguistically, diagrammatically etc. represent rules without there actually being any rules. Hence we can still refer to a rule R, as long as it is understood that any such reference is only to a representation of R. Also, there can be kinds of behavior that are related to rules, without there being any rules. Behavior conventionally described as 'adding' numbers is related to the addition rule, in the sense that cognition involving conventional representations of the addition rule often results in characteristic addition-like kinds of behavior--even though, strictly speaking, there is no genuine activity of addition that is the following of an addition rule. Thus, just as we can recognize pictures of Santa Claus, so also can we recognize Santa-related kinds of behavior, such as dressing in Santa suits, talking about Santa, and so on, with similar points applying to rule-related behavior in general.

As a second preliminary, generally a represented rule R is conventionally associated with supposed examples of applications of rule R. For example, '2+2=4' is such an example for the addition rule. Each socially accepted rule R will have, associated with it, a set AR of such conventionally accepted applications A of R.

The basic idea of retrospective rule-following (RRF) can then be stated as follows. To begin, there is no positive conception of what it is to follow a rule as a prospective guide. Instead, the idea is that to follow a rule R, in the new retrospective sense, is to be willing to attempt to change one's rule-R-related behavior in case that there is some application A of the rule R, belonging to the conventional set AR of such applications, to which one's previous R-related behavior did not conform. Or in other words, the rule does not act as a future-oriented or prospective guide to behavior, but instead one is willing to make changes in further cases of one's previous or retrospective R-related behavior that did not conform to applications A of the rule R.

The concept of conformity used here is specifically not a concept of conformity or non-conformity to a rule. Instead, it is involves only a raw behavioral difference between a particular case of a person's R-related behavior B, and a particular application A from rule R's conventional application-set AR. For example, if person P says that 'two plus two equals five', the sounds he makes during that sonic behavior are behaviorally distinct from sounds conventionally associated with the conventional application '2+2=4' of the addition rule. Then the basic idea of RRF is that person P could qualify as a rule-follower in the RRF sense if he is willing and capable of attempting to adapt his future R-related behavior so that this specific behavioral mismatch or non-conformity of his behavior B with application A will not recur--or at least that the mismatch will be lessened.

Or, in a more strict behavioristic idiom, to be an RRF (retrospective rule-follower) with respect to rule R is to be such that exposure to members A of rule R's application-set AR tends to minimize the behavioral differences between one's own R-related behaviors and corresponding behaviors associated with such members A of set AR. Hence there is no suggestion here of a kind of second-order rule that one is following, namely one such that one ought to make ones future R-related behaviors conform to members of set AR. To be an RRF (retrospective rule-follower) is simply to be such that one's actual behaviors do in fact tend to exhibit the relevant behavioral changes under the stated conditions.

For example, to follow the mathematical rules for 'plus' is to be willing and able to alter one's future plus-related behavior in the light of any previous failures in conforming to standard applications of the 'plus' rule, such as that 68+57=128--rather than 5, which would instead be a particular application of Kripke's 'quus' rule (Kripke 1982, e.g. p. 37).

Or, a person counts as following the rules of English grammar if, after reviewing anything that he has written, he is willing to alter those sentences which fail to match particular applications of the rules for English grammar, so that they conform to those applications.

Thus the retrospective conception of rule-following (RRF) can provide a sparse replacement for both rules and rule-following--which is also different enough from standard conceptions to avoid both mentalistic objections, and standard circularity objections to social accounts of rules. In effect, the theory sketch provided is one part of a functionalist theory of mind, according to which kinds of cognition related to rule-following have the functional role of minimizing behavioral mismatches between an individual's own rule-related behavior, and corresponding kinds of behavior that would instantiate standard, socially accepted applications of supposed rules.

This kind of account can also be extended to cases of supposed rules governing correct perception or true propositions, or any other cases where there is usually considered to be some independent factual source on which rules, and human rule-following, depend. For instance, mathematicians typically assume that there is an independent world of mathematical fact, such as facts about the operation of addition, to which any rules governing human activities of addition must correspond. Similar assumptions are made about supposed rules involved in perception and language. An advantage of the RRF account to be provided is that such broadly realist assumptions about worldly correspondence can be fully accepted, even though the RRF account specifically of rules and rule-following is an eliminativist one (see sections 7 and 8 for details).

Also, the resulting RRF view is substantive enough so that it can accept the traditional distinction between following a rule, versus behavior that just happens to accidentally conform to a rule (Searle 1995). The distinction on the RRF view is that mere rule-conformity behavior with respect to a rule R is unaffected by exposure to standard applications of R, whereas the characteristic signs of rule-following are that such exposures do tend to lead to some behavioral minimizations of any mismatches.

3. Why People Retrospectively Follow Rules

Why do people follow rules, or at least seem to? Traditional mentalistic views of rule-following explain it in terms of mental content involving rules that we, as rational beings, recognize that we ought to follow, in order to further our own interests, and those of our society. An alternative, eliminative theory of rules and rule-following, such as the current retrospective rule-following (RRF) view, also needs to provide some overall explanation as to why people engage in RRF-related behavior.

The RRF view, as part of a broadly naturalistic theory of mind and cognition, can appeal to evolutionary considerations in place of mentalistic ones. We retrospectively follow rules because species and societies that accidentally happened to do so thereby gained enough evolutionary fitness so that they were more likely to survive than other species or societies that did not happen to use such methods. This evolutionary kind of explanation need not involve a denial that we have beliefs, desires or reasons, for they too could potentially be explained in terms of appropriate naturalistic substitutes that would be compatible with such a broadly evolutionary explanation of cognition. This evolutionary mode of explanation is, in my view, a prime factor in explanations of apparent rule-following in naturalistic dispositional terms, as will become clear in subsequent sections.

4. Reforming Dispositional Accounts of Rule-Following

As noted in section 1, there is a standard problem with simple dispositional attempts to naturalize mental rule-following. For if following a rule R were simply to have a disposition to follow or conform to R, then the distinction between correct versus incorrect following of the rule would be obliterated, since any causal manifestation of the disposition would automatically count as correct. Also, attempts to escape such problems by appeal to rule-following dispositions as manifested in optimal circumstances also cannot succeed (Boghossian 1989).

Now it might be thought that the current retrospective rule-following (RRF) replacement for rule-following could avoid those standard problems by instead invoking dispositions, with respect to a rule R, to minimize behavioral differences between one's own R-related behaviors, and corresponding behaviors exemplifying standard applications of rule R, as discussed in section 2. If there were any such dispositions, they would apparently not be rule-following dispositions, and so standard criticisms would not immediately apply. Or in other words, in place of standard rule-following dispositions, the putative strategy would be to invoke retrospective rule-following dispositions instead, whether in simple or optimal form.

However, arguably that strategy could not succeed either, and for reasons analogous to those applying to standard conceptions of rule-following dispositions. We still need to be able to distinguish between attempts to minimize behavioral differences, and any successes or failures achieved by such attempts. Hence, if we are to explain the apparent normativity of the replacement RRF theory, distinctions between correct versus incorrect attempts to minimize behavioral differences must be preserved. Or, to put the matter in terms of a standard criticism, there is, in effect, a second-order rule R' that one ought to minimize relevant behavioral differences with respect to one's R-related behaviors vis-a-vis standard applications of R. Following that rule R' cannot be explained simply as the having of a disposition to do so, whether or not it is done in optimal conditions, or the apparent normativity of rule-following would be lost, in RRF terms just as much as in traditional terms with standard rule-following.

To be sure, the RRF account must, as mentioned in section 2, deny that there genuinely is any such second-order rule R', of the kind invoked by the standard criticism just given. (The RRF account must also deny that there is any RRF substitute for such a rule, in that such an assumption would be the start of an explanatory infinite regress). To be an RRF (retrospective rule-follower) is simply to be such that one's actual behaviors do in fact tend to exhibit the relevant behavioral changes under the stated conditions.

Nevertheless, though there hence cannot be, strictly speaking, any dispositions to minimize R-related behavioral differences, there could be dispositions to attempt to do so, as long as only a suitably minimalist conception of an 'attempt' is invoked. Here is where the evolutionary perspective briefly explained in section 3 becomes relevant. Evolutionarily successful species and societies will tend to have some dispositions such that their behavioral manifestations, for some sufficient number of their members, exhibit rule-related minimizations of relevant behavioral differences--i.e., the RRF substitute for rule-following. Also, closer scientific examination would tend to show that behaviors associated with attempts to minimize relevant behavioral differences also occurred in all three categories of habitual rule-followers, occasional rule-followers, and some non-rule-followers who at least attempted to produce relevant behaviors.

Then the category of dispositions to attempt to follow rules would include all of the highly various dispositions, whose behavioral manifestations in fact tend to exhibit minimization-related behaviors of relevant sorts in the relevant individuals. For example, someone who wishes to pass their driving test could provide evidence of attempts to learn and follow standard driving rules by their behavior of repeatedly taking the same exam, even if they always fail and so never do retrospectively follow any of the rules. Or in other words, in addition to straightforward behavioral evidence of successful or unsuccessful RRF minimization behaviors, there can also be ancillary evidence that relevant attempts have occurred, whether or not they are successful.

5. Facts and Values

What is the relation between facts and normative standards (norms or values)? Insofar as all normative concepts and discourse are integrally connected with issues of rule-following, an initial outline of the fact/norm relation is already derivable from the current RRF replacement for supposed rule-following. On the RRF account, the supposed normative force of a rule R is not reduced to, but instead eliminated in favor of, the actual factual incidence of cases of rule-R-related minimizations, namely future minimizations of any previously occurring behavioral divergences between rule-R-related behaviors and corresponding behavioral versions of standard rule-R-related applications. To repeat, this is not a reduction of rule-related normative force but an elimination of it, in that facts about rule-related behavioral minimizations are simply facts having no intrinsic normative force or normative properties.

Nevertheless, the evolutionary argument that explains the wide incidence of RRF rule-following in contemporary species and societies does provide a non-intrinsic, purely instrumental normative account of such rule-following. Such actual practices in general presumably would not have become endemic unless they contributed to the survival of groups adopting them, so to that extent they have an instrumental survival value.

This point would also be important in any putative RRF-related explanation of semantic or perceptual normativity involving concepts of truth and falsehood, or correctness and incorrectness. To the extent that evolutionary considerations make it likely that a significant percentage of our beliefs are true, or that perception generally tends to be correct rather than incorrect, such instrumental kinds of semantic value--along with corresponding disvalues or negative values such as falsity or incorrectness--could constitute the purely instrumental normative force of such concepts, insofar as they are normative at all.

However, perhaps somewhat paradoxically, this straightforward, purely factually based model of instrumental normativity arguably presupposes a robustly realistic account of worldly facts, including facts about behavioral correspondences and minimizations, facts about perceptual correctness or incorrectness, and facts of truth or falsity of beliefs. For example, we cannot instrumentally justify some true beliefs as promoting survival unless there is a genuine fact of the matter that at least some of them are true. But then there must be more to their truth than their instrumental value, or the explanation would fall apart. So something akin to a traditional correspondence theory of truth is required to validate that there are indeed basic facts involved in true beliefs, which are factually distinct from whatever facts are involved in false beliefs. What this means is that we must carefully distinguish purely factual aspects of semantic values, such as truth and falsity, from their instrumental normativity. The succeeding sections will show how this could be done in RRF terms, after a preliminary discussion of mental content.

6. Dispositional Replacements for Mental Content

Recall from section 2 that there is a price to be paid for accepting skeptical arguments that lead to a denial of the existence of rules and rule-following. If the conclusion is accepted, one must also accept that consequently, there cannot be any mental content that involves the following of rules either. Hence, arguably, any putative kinds of mental content that have normative aspects, including the putative contents of all perceptual and belief states, is thereby also eliminated by such an account.

For example, veridical perception of an item as being green rather than grue (i.e., green up to now and blue thereafter) would presuppose that one is following a rule to perceive only green things as being green in such cases (Goodman 1983). But since there is no such rule on the current account, and no genuine disposition to perceive all and only green things as green--whether as traditionally or retrospectively interpreted--then a purely naturalistic dispositional account of perception must deny that there is any such perceptual content.

As it happens, there are also independent arguments leading to the same conclusion. Arguably a purely dispositional account of perception would have to explain any subsequent behavioral manifestations of the relevant dispositions purely in dispositional terms, hence leaving no remaining role for content-based explanations of the perception or subsequent behavior (compare Worley 1997). Hence even if mental rules and rule-following did exist, they would have no explanatory role in any purely naturalistic account of mental and perceptual causation. Thus, far from acceptance of the skeptical rule-following arguments having catastrophic consequences for naturalists, instead their methodology arguably already commits them to a rejection of any explanatory relevance for mental content as such--or at least, for any rule-related kinds of mental content.

Nevertheless, in spite of our acceptance that there is no rule-related mental content, we still want to maintain that there is a legitimate sense in which perception of an item X as being green is either correct or incorrect. My suggestion is that we can do so by extending to perceptual and semantic cases the same retrospective rule-following (RRF) account as we used to explain social rule-following in general. Recall that on the RRF account, there is a fact of the matter as to whether someone is retrospectively following a rule R, depending on whether or not minimizations of behavioral divergences from standard applications of rule R have occurred.

In the perceptual case, an example of such a relevant fact about retrospective perceptual rule-following would be as follows. If there is behavioral evidence of at least partially successful attempts by an individual to minimize divergences between individual green-rule-related perceptual behaviors and standard, socially accepted perceptual applications of a green-related perceptual rule, then the individual is generally following the perceptual rule; otherwise not.

Or, in more detail, any particular perceptual episode of an item X being perceived as being green can fail or succeed in conforming to some standard application of an appropriate green-related-rule in the following manner. Standard applications of a green-related rule can be provided by scientifically controlled perceptual tests, such as by exposing a test subject to a well-lit green surface, and prompting him to verbally identify its color. The standard behavioral outcome of such a test would be the behavior of saying 'green' after exposure to the color. If the individual perceiver's behavioral response is also 'green', then this provides evidence of his having correctly perceived the color as being green, in that his response matched the standard response. But if he provides some different response, that would instead be evidence of incorrect perception of the green surface.

To be sure, such experiments provide evidence only, because there might have been non-perceptual reasons as to why a person responded with the word 'green' when prompted, though good experimental design can minimize such extraneous factors. Also, perceptual correctness as thus defined itself depends on a prior conception of correspondence with worldly facts, to be discussed in section 8 in connection with truth.

However, since strictly speaking there is no mental content, the evidence provided in such standardized scientific experiments is not evidence of correct or incorrect states of perceptual content. Instead, the evidence is evidence only of correct or incorrect green-related perceptual retrospective rule-following (RRF) behavior, which is, on the current naturalistic account, the only full and literal kind of evidence that perceptual cases are correct or otherwise.

In order to complete our naturalistic account of perception, a purely dispositional theory of perception itself would be required, to complement this RRF account of perceptual correctness or incorrectness. A suitable, purely causal theory has been provided elsewhere. The basic idea is that a person Z perceives an item X to have property F just in case X causes Z to acquire F-related dispositions. This account, as required, makes no use of a concept of mental or perceptual content (author articles 1, 2, 3).

7. Truth and Propositional Content

At this stage in our theoretical exposition, the most attractive option for a theory of truth would be one which used the same reflexive rule-following (RRF) approach as has already been developed for miscellaneous and perceptual kinds of apparent rule-following, as supplemented with an account of correspondence with worldly fact. Belief-states as such would be purely dispositional, and not involve any rule-related mental content. Also, just as evidence of perceptual correctness and incorrectness were explained in terms of minimization--or failure of minimization--of relevant behavioral divergences, evidence of truth and falsity would be explained along similar lines.

However, there is an initial roadblock to any such account. Even if belief states as such are purely dispositional, the concept of the propositional content of a belief, or of the proposition expressed by a belief, seems to be a theoretically central one. If Joe believes that the cat is on the mat, then surely there must be a proposition p = 'the cat is on the mat' that Joe believes. Also, that same proposition could be stated or expressed in a mind-independent form by a linguistic expression in some natural language, such as the English sentence 'the cat is on the mat'. Thus at a minimum, we need some account that could reliably correlate purely dispositional particular belief states with particular sentences that express the relevant propositions that belief states would normally be taken to express. Or, in more succinct terms, we need an explanation of how a purely dispositional belief that p could be related to the proposition p, as expressed by some sentence in a natural language. Also, the explanation must be consistent with our rejection of standard rule-following, and consequently of mental content as well.

Fortunately, there is a way to do this that is also congenial to cognitive naturalists. As mentioned in the previous section, an explanation in purely dispositional terms of perception of worldly states of affairs--such as that of the cat being on the mat--is already available. The relevant account is a causally reflexive one, according to which the fact of the cat being on the mat causes perceiver P to acquire appropriate dispositions toward that same worldly state of affairs. In addition, the behavioral manifestation of those dispositions would provide evidence of the correctness or otherwise of the perception.

Given this basic perceptual account, we already have everything we need for an account of propositional content. A person can learn to associate a sentence such as 'the cat is on the mat' with her perception of the cat being on the mat, so that she acquires some of the same dispositions from the sentence as she would otherwise acquire directly via perception of the cat on the mat. In this manner the sentence can come to represent the relevant cat-related situation, and hence to have a propositional content. Nevertheless, this postulation is fully consistent with our rejection of mental content, because the propositional content is the content of a worldly physical representation--a sentence--rather than being a specifically mental content.

As for the issue of which perceptual dispositions are thus propositionally acquired, arguably they involve that subset of perceptual dispositions which are relevant to the correctness of the perception, and hence to the truth or falsity of the proposition. This point also further links dispositional belief states with propositions, in that presumably it is the same correctness-related subset of perceptual dispositions which are involved both in a dispositional belief-state and in the corresponding proposition.

Thus we have explained propositional content as a side effect of cognitive processes of association, in which people come to acquire from linguistic sentences a subset of the same dispositions that they would normally acquire perceptually. But arguably this process of association is itself a perceptual one, in which perceivers learn to acquire, from perception of the relevant sentences, some of the same dispositions that they would otherwise acquire from perceiving cats on mats etc. Hence this whole process is a purely dispositional one, in that it involves only perceptual processes that are themselves dispositional.

Also relevant to this explanation is a familiar distinction between original versus derivative intentionality and content (Searle 1992). Original intentionality is associated with the traditional conception of human mental content, whereas derivative intentionality--such as that of pictures or sentences--always involves some interpretation by humans of such worldly physical artifacts. In effect, my claim is that there is no content-based original intentionality at all, but that nevertheless we can still have derivative intentionality, and derivative representational content, via human perceptual associative interpretations of sentences or pictures.

Our explanation sketch is now complete. A purely dispositional mental belief state, which is a belief that p, for some proposition p, is associated with no original mental content. But the relevant belief dispositions are also dispositions that could be perceptually acquired, via processes of association, from conventional linguistic representations of the proposition p, which linguistic sentences thereby have a purely derivative propositional content as explained above. Hence we can explain how it is possible for beliefs to 'have' a propositional content, in spite of the fact that there is no original mental content as such.

8. A Retrospective Account of Truth

The previous section explained how a purely dispositional belief state could nevertheless be said to have a propositional content--which derivative content belongs, not to the relevant dispositional belief state as such, but instead to sentences conventionally expressing such a proposition, via an associative perceptual interpretation of the sentence. Hence, with that road-block having been surmounted, we can now proceed with the promised account of truth and falsity in retrospective rule-following (RRF) terms.

But as a preliminary caution or reminder, on the account to be given of truth and falsity--as with perceptual correctness and incorrectness--it is particular behavioral manifestations of appropriate dispositions that will primarily count as being true or false, rather than propositions as such. Thus the functional role of the concept of a proposition, on the current RRF account, is one that involves at least two kinds of indirection, as compared with traditional accounts. First, as in the previous section, propositions aren't directly the original content of belief states, since there are no such rule-related original content states. But in addition, the current point is that propositions are only true or false in another kind of indirect sense, in that, strictly speaking, it is only the behavioral manifestations of associated dispositional belief states that directly count as true or false.

Here is an overview to help motivate and clarify this account. Suppose a person P perceives a green item X, and then linguistically reports her perceptually acquired belief that X is green. Now on a traditional correspondence view of truth, the expressed proposition 'X is green' is true just in case there is a fact of X being green that it corresponds to. The RRF account can preserve a variant form of this robust traditional concept of truth as follows. To begin, the original skeptical concerns about meaning rules for the predicate 'green' arise because applications of it to some worldly green-related property are potentially multiply ambiguous or indeterminate. Though we intend to apply the predicate 'green' to X just in case X is actually green, the relevant green-related property of X might instead be e.g. a version of Goodman's 'grue', that is, green up to now but blue thereafter. So, though we think we mean the property green by 'green', there are any number of other worldly properties that might be meant. Hence--if we accept the skeptical conclusion--there are no such meaning rules for application of our predicates or concepts to worldly properties.

The retrospective approach accepts this skeptical conclusion, but instead makes use of conventional applications of the supposed rule, such as that in any particular case we assume that the predicate 'green' applies to a green property-instance rather than to some other one, such as grue property-instance. There is no rule that justifies this use, so that the concept of a 'conventional application' is not itself a rule-governed one--its only justification is a broadly evolutionary one, namely that cultures which in fact rely on such conventional applications of property terms tend to survive and prosper.

Next, we must carefully distinguish realism about worldly properties from skepticism about meaning rules. Accepting skepticism about rules does not commit one to denying that there is a genuine property of being green, another of being grue, and so on. Instead it just involves a denial that there are any meaning rules connecting our predicates or concepts with whatever relevant worldly properties there actually are. Hence we can also be realists about there being genuine worldly facts, such as the fact of object X being green, even though we are meaning skeptics. Hence also a correspondence theory of truth also remains an open possibility, in which truth consists in correspondence to such worldly facts.

For the meaning skeptic, truth concerns only the limited subclass of possible correspondences that are provided for in conventional applications of supposed rules. The claim 'X is green' will be true just in case the property conventionally picked out by applications of the predicate 'green' is actually instantiated by object X. Now in addition, a substantive correspondence theory of truth should be able to identify two items independently of each other: 1) the relevant putative truth-bearer, and 2) the relevant worldly fact or property, correspondence to which would make 1) true.

Since we want the relevant truth-bearer to be an item of behavior, a useful model of its structural relation to the corresponding fact would make use of the following logical equivalence: item X is an F, for some property F, just in case X is correctly classifiable as a member of the equivalence-class of Fs. Or in other words, for a worldly item X to possess or instantial property F is for it to be correctly classifiable with other items that also are F's. Hence we can have a substantive correspondence theory of truth, in which independently identifiable behaviors of classifying X as an F correspond with X actually being F. For example, the behavior of putting an object X in a bin reserved for green objects would be a behavior describable as one of classifying X as being green, and it is true just in case it is independently the case that object X is actually green.

To be sure, ordinarily we would not describe an action or event of classification behavior as being true, but rather as being correct or incorrect. There are at least two reasons for this mismatch. First, since semantic folk psychology assumes the existence of rules and rule-following, it is not surprising that some conceptual adjustments are required in order to achieve an adequate, eliminative retrospective rule-following (RRF) theory of truth. Normally we assume that propositions are truth-bearers, but the RRF dispositional analysis of propositions makes that view inappropriate.

Second, arguably the concept of correctness is a more inclusive concept than that of truth. It covers cases of behavioral conformity, such as when addition behavior is correct because it conforms to, or is behaviorally identical with, standard behavioral applications of a supposed rule. But it is also used to apply to cases of perceptual or semantic correctness, in which correspondence with worldly facts is involved. It is this latter more restrictive category in which the more specific concept of truth, rather than just generic correctness, finds its main application.

To summarize, a belief, such as a belief that item X is green, is just a dispositional state, with no mental content. Second, the concept of a believed or expressed proposition only arises in cases when representations, usually linguistic, are used as perceptual substitutes, via associative learning, for actual perceptions of X being green. Such propositional uses of sentences involve only a proper subset of the dispositions that likely would be perceptually produced by someone perceiving that X is green, namely those relevant to true or false classification of X with other green things. Beliefs also involve only this same proposition-related proper subset of likely perceptual dispositions associated with seeing that X is green--which point at least partly explains the integral connection of a belief with a proposition.

As for truth itself, on this RRF account it is a matter of whether the behavioral manifestations of the relevant subset of propositional classification dispositions correspond with the actual facts or not. Also, even though the corresponding belief is associated with a proposition or propositional content p--which is the derivative content of an associatively used linguistic sentence--that proposition is not, strictly speaking, itself true. Instead, it in turn is only 'true' in a secondary or indirect sense, derived from whether or not its behavioral manifestations correspond to the relevant facts.

9. Conclusion

It generally seems to have been assumed that if Wittgensteinian or Kripke-style skepticism about meaning rules and rule-following is justified, then there cannot be any legitimate remaining role for a concept of meaning in the cognitive and linguistic sciences, nor for related concepts such as those of truth, belief, propositions and propositional content, and so on. However, in effect the claim of this paper is that such an assumption is premature. As long as some naturalistic, dispositionally based replacement for concepts of rules and rule-following can be provided, such as the retrospective rule-following (RRF) account provided here, then it is possible to reconstruct plausible substitutes for all of these concepts, including a substantive concept of truth that has some significant advantages over more standard conceptions.

To be sure, any kinds of original mental content that would involve rules must also be given up, but arguably a purely naturalistic, dispositional theory of cognition should not make any appeal to such concepts in any case. Thus, far from the skeptical rule-following arguments being a threat to naturalistic accounts of meaning, instead their acceptance is a liberating factor that finally can enable dispositional accounts to break free of standard criticisms of them.

REFERENCES

Blackburn, S. 1984: 'The Individual Strikes Back.' Synthese 58: 281-302.

Boghossian, P. 1989: 'The Rule-Following Considerations.' Mind 98: 507-549.

Goodman, N. 1983: Fact, Fiction and Forecast. Cambridge, Mass: Harvard University Press.

Kripke, S. 1982: Wittgenstein on Rules and Private Language. Cambridge, Mass.: Harvard University Press.

Miller, A. and C. Wright 2002: Rule-Following and Meaning. Montreal: McGill-Queens University Press.

Searle, J. 1992: The Rediscovery of the Mind. Cambridge, Mass: MIT Press.

Searle, J. 1995: The Construction of Social Reality. New York: Simon and Schuster.

Worley, S. 1997: 'Determination and Mental Causation.' Erkenntnis 46: 281-304.