Can Assertional Practice Explain Representation?
January 19, 2011 § Leave a comment
One of my main interests is the connection between language and conceptual content. I belong to a class of philosophers who are convinced that there must be an essential connection between the existence of language and the existence of conceptual content. If language is to explain the existence of conceptual content, however, then there must be some principled reason for why conceptual content is made possible only with language.
I think the default view on this matter certainly is language has meaning because mental states have content, rather than the other way around. This is not the least made plausible by the unavoidable practice to attribute many of the same sort of attitudes to animals as we do to ourselves.
The burden of the langauge-first proponent then, is to point to what crucial difference that the existence of language should make for the existence of conceptualization. My own tentative answer to this challenge is the is the following simple thought. What language provides is the existence of a peculiar form of social practice, and that is to use sounds or inscriptions as virtual perceptions for features of the surroundings. The social equivalent of perceptions are found in the practice of assertions, which serve as virtual substitutes for sensory experiences.
Each of us is a biological organism which have sense that have an ability to discriminate features of the surroundings in a way that matches our biological interests. The causal nexus of these discriminating features are our senses, that instantiates this relationship. The problem with this discriminating sort of ability is that it does not extend to the past, the future, the general, the hypothetical or to far away places beyond immediate causal effect. It is also essentially private, confined to each individual’s localization in time and space.
My idea is that language greatly expands on the biological causal nexus by instituting the practice of creating an intersubjective world that exists in langauge use and is shared by all users of language. This intersubjective world exists in the practice of assertion in virtue of taking on a dual role. On the one hand operates as an regulative ideal in which every assertive action takes on the responsibility to uphold this intersubjective world by satisfying various a priori constraints on objectivity. On the other hand the posited world is treated as substantial in that assertions are taken to reveal something that is really there. This intersubjective world is not something that comes in addition to the real world, rather it is the world we already know through conceptual thinking, made visible to us by our participation of linguistic practice. Seen from this perspective, there is every reason insist that qualitative difference is introduced with the participation of linguisitic practice. The language-less animals does not relate to the same world as those with a language. Indeed, since the notion of a world is a part of language, the language less individual does not relate to a world at all.
This simple idea touches upon a large of issues, and one of them seems to me especially central to the existence of conceptual content: it is the feature of conceptual content that it constitutes representations of the surroundings. It is my idea that it is possible to provide a fairly naturalistic and causal theory of representation with the help of this notion. The connection between a causal theory of representation and this idea of assertion as virtual perceptions is actually not hard to see, since it may simply taken to be a version of a reinforcement theory of linguistic behavior in the tradition of Skinner, Quine and Davidson.
A causal theory of representations is based on the notion that what a representation represents can be given by its cause. The bond must furthermore be law-like, so that the the representation tracks what is being represented. One intuition behind the theory can be assiated with Grice’s notion of natural meaning, according to which certain causal relations are indications of what causes them. Hence, to use an example from the granddaddy of causal theories, Dennis Stampe, the tree rings of tree trunk represents age of the trunk, since the seasons causes the rings.
Lingusitic behavior differs from natural indications in that there is no inherent connection between cause and what they represent. By themselves, the expressions are meaningless. However, what a linguistic community provides is reinforcement, so an organism can be tuned so that it is caused, say, to utter the expression “dog” when a dog occurs in the field of vision of the individual. This simple thought, famously exploited by Skinner in “Verbal Behavior” and criticized by Chomsky, is obviously a very long way from providing a theory of representation, but indicates how a causal theory may be in the offing.
The actual details of providing a causal theory based on the notion that assertions involve taking a stance on a number of difficult issues that relate to the possibility of meaning. The aim of a causal theory representation is to explain how it is possible that some physical feature of the world, a representation, can be about some particular state of affairs that is expressed by that representation.
It is important to be clear about two different tasks that is involved in giving an explanation of conceptual content. One task is the problem of individuation which is the problem of explaining that a representation has one particular content rather than another. The other task is the problem of representation, which is to explain that a content can be about the features of the world that it expresses. It is possible to pursue the first questions, without having the pretensions to explain the other. However, the natural way to address the last question is to show how individuation is in principle possible in at least some cases without in any way relying on the phenomenon of representation in the explanatory resources. The task of a causal theory that attempts to explain representation is to explain content individuation by causal connections in a way that yields a substantial explanation of representation. The the question is what is involved in such a substantial explanation.
As an approximate generalization we can say that a causal theory be a specification of the following general schema.
(R) r expresses s being F if “S’s being F cause R’s under conditions C” is counterfactual supporting.
According to this proposal, the a particular tokening of a representation expresses a particular content when caused by a stimulus token, when the types they instantiate exhibit a law-like connection. If the idea that assertional behavior is to provide a theory of representation, in other words, then the tokening of assertion types must be lawfully correlated with the states of affairs that they represent.
To outline an account of assertional practice can provide an account like this, it is neccesary to say more about what is involved in providing a causal theory of representation.
There is one problem in particular that is so important that it can be taken to be the central problem to be solved, and that is the problem of error. It easiest to see the formal constraints a causal theory by considering a very simple proposal, such as a pure disposition theory, perhaps a bit like the theory that Kripke consideres as a possible answer to the rule-following paradox.
(SD) r expresses s being F if S’s are disposed to cause R’s when S’s are F
In (SD), S and R are supposed to stand for candidates for stimuli and the representations they cause.
The problem of error is that there is a natural ambiguity in the assignment of the representation types. Because the representations represent the same states of affairs even when they are false and caused under other circumstances, then a number of assignments are possible. This is the indeterminacy problem that tends to infect every version of (R). There are both the asynchronous possibilities as well as the synchronous possibilities. The former gives rise to disjunctive assignments, and the problem may for that reason be called the disjunction problem. A number of proposals more or less conforming to (R) have been proposed, and and we can summarize a number of related challenges that is involved in attempting to solve the disjunction problem.
The explanation must provide a way to exclude the disjunctive possibilities, and we here run up against the problem what counts as a sufficiently substantial explanation. The task of explaining error has two central challenges, given that we can look away from whatever cognitive requirements is placed on the representation candidates in order to count as representations. On the one hand it must provide an identifying cause that avoids the indeterminate possibilities, and on the other hand it must provide a way to explain that the representation types retain their meaning in the cases of error. The success will depend on the nature of the causal account in question.
The most straightforward way to offer an account is to provide some sort of reduction of the causal types in question. The most likely form this reduction is going to take is by what Fodor calls a “Type 1” account, which distinguishes between the cases in which the causes of the representations are meaning determining and those that are not.
A natural objection to such an account would be either to allow some sort of ceteris paribus clause on the specification or to deny the possibility of reduction. The problem with these move is that such moves may be equally applicable to the disjunctive possibilities. There is a matter here of what it means to assign a meaning determining cause. The problem is that of all the tokenings of the representations are compatible with different ways of “reading off” the meaning determining causes. It would seem that the assignment of a meaning determining cause is deep down a matter of metaphysical connection between the causes and the the meanings. In some sense to be specified, the causes metaphysically determine the meaning in question. The matter of reading off off the meaning determining cause concerns the justification for holding that there is one meaning determining cause over another. The most straighforward way to provide such a justification is to offer a straightforward specification of the circumstances. Another way is to rest on a scientific explanation of the causes in questions, claiming perhaps that the assignment of causes are those that best explains the assignment of a cause. Yet another option is found Fodor’s assymmetry thesis, which provides a sort of criterion of picking out the right causes, rather than actualy providing a specification of how it is satisfied.
This is not the place to survey all the different proposals of nomological relations that has been proposed, and the variety of explanatory issues and difficulties that involved in their respective ways of reading off the identifying causes, but the fundamental problem is that they run up against problems with providing sufficient determinacy. From my perspective the famous problems outlined in Wittgensteins rule-following considerations. According to these considerations, there is no fact of the matter that can be pointed to that makes makes it correct to prefer one continuance rather than another, and that the correctness is found in what we actually do as a matter of lingusitic practice. The conclusion is that we in some sense is walking on thin air when are using words the way we do and that the actual use is all the support we have. What this discussion shows is that that attempts to point to some objective feature merely reveals an indeterminacy, which can be taken to indicate that meaning therefore exists in virtue of a lingustic practice. This is theme that is also echoed in McDowells understanding of the rule-following considerations.
I will now indicate how I think that the practice of assertion can provide a causal theory along the lines of (R). The basic idea is that the cause of assertions, understood as dispositions, can identify the content of the tokened assertion type. The proposal is as follows:
(AP) r expresses s being F if a competent assertion practice practitioner I would sanction that that interpretee P ought to utter r in response to s in order for R to serve as assertion practice virtual indicator for S’s being F’s.
This is undoubtedly quite a mouthful, so to outline the proposal I will proceed in the following fashion. I will first outline the motivation and an explanation of the nature the dispositions to assertion. Secondly I will explain the role (AP), and finally I will justify that (AP) provides a robust explanation of representation.
In order to understand the nature of the dispositions that arise from this proposal it might be best to look some early versions of this account. As already explained the basic idea is that certain sounds or inscriptions may be reinforced by the community to serve as indicators of relevant features of the surroundings. A couple of central problems with the theory of reinforced dispositions is on the one side the fact that the solution fails to solve disjunction problem. If solitary dispositions are indeterminate with regard to their identification because of the distribution of the false tokening, then so is the use of learning. A part of this is due to the qua problem, according to which it is going to be ambiguous what is being learned or reinforced. This is a problem for any attempts to use learning to solve indetermincy, and the problem is the same for for example Dretske as well. The other problem is concerns the triggering of the dispositions. In Skinners theory of linguistic behavior the the dispositions where triggered in the presence of the objects in questions, which is slightly ludicrous, since people don’t go ahead saying “dog” just because a dog is present. In the far more sophisticated version of linguistic reinformcement found in Quine, the relevant circumstances are those that concern “prompted assent”, which goes some way to solve that part of the problem. Another part of the solution is also found in Quine, because he notes that there exist an “objective pull” in the social reinforcement, although he probably has a sort of realist picture about what this pull amounts to, or in other words that the role of the reinforcement is to close in on objective properties that they serve to manifest. Davidson has another interesting clue in that he likely hold an anti-realist picture of the causally relevant features in which such dispositions ought to be understood, since they are supposed to be intrinsically intersubjectively available features of the surroundings.
These observations sets the stage for the kind of dispositional story that might fill in picture to provide us with an entirely alternative account to the kind of reinforcement found in writers like Skinner, Quine and Dretske, and that is a version in which the causally relevant types are not natural kinds, but practice determined causal types, much in the same way that “expensive” is an interest dependent and practice determined causal type.
To look for an interest dependent and practice determined causal type fits with an important interpretation of the the rule-following considerations, since these can be taken to show that there is no underlying realist fact about the world that we can point to that explains meanings. That does not amount to a skeptical position, only the epistemological thesis that access to these facts are essentially those of a participant of these practices, and the metaphysical lesson is that “meaning is use”, meaning that it is features of practice that explain the determinateness of meaning.
What is missing in the Quine/Davidson picture of reinforcement is what this use amounts to if if cannot be understood realist terms. The burden of this strategy to account for dispositions to linguistic behavior that does not depend on interest independent causal types, which means that one must find some interest dependent source of the type in question. That such a type exist is probably hinted in Davidson who clearly sees that these similarities are driven by the practice of communication. What there is no hint of, however, is what purpose is served by communication that gives rise to these intersubjective causal types. (There is no of course no point to suggest that the purpose of communication is to convey meanings, since that blatantly misses the challenge outlined here).
It might seem that there is traditional alternative here, and that the kind community theory outlined by Kripke as a solution to the skeptical challenge, a version of which certainly could be recruited for the purpose of explaining the “correctness” of utterances. According to this solution correctness consists simply in the agreement in “assertability conditions”. with other participants of a community in continuing a in a certain way, which is simply way of going on. There is no need to rehearse the different versions and the different criticisms of this sort of proposal here, but it should be noted how its shortcomings are indicative of how little sense it makes of assertional practice. According to such at theory, it easy to say what a term or a sentence should be used on, it is simply to say that “red” should be used on red things. Disregarding deflationary or skeptical accounts of use based correctness, there are now two ways to explain why this is correct. One can either say that there external causal types that explains why a particular use constitute the “red” type, or one can say that the red type is simply what “red” is used on, whatever that is. In the former case we get an account that is dependent on solitary dispositions to derive communal disposition, which only shows that the communal contribution is empty.
With regard to the second “idealist” solution, we run up against a version of the problem of error, since the solution requires that actual use on a communal basis is incorregible. This is a well known objection to this account, but what is more important is that it misses the point of assertion, since the asserter is in virtue of asserting contributing to a practice in which he is potentially corrigible. Such an assertion would amount to a stipulation, something one simply cannot regard most assertions as being. The proposal misidentifies the nature of communal authority. It seems clear that the correct way to construe the nature of communal authority is not as a determination, but as an equal right to question the assertion when it is put forward. The problem with the idealist version is that conceives of the world as a mirror image of the propositions, and is in that way infected with the same problem as the realist solutions, since it does go anyway toward closing in on the point of the game of assertions in the first place. This is in fact seen in the works of Wittgenstein, which in “Philosophical Investigations” clearly attempts to analyse the social point of assertions as regional practical interests with no sui generis point, a kind of investigation that might seem well motivated by practice dependent features of meaning, but which is implausible thesis on the face of it, since we so easily recognize and play along with the sui generis featurs of assertion.
These observations about the reinforcement model and the community approach respectively, finally sets the stage for an alternative account of the point of assertion. The point of of assertion is to present that “is the case”, and we should understand what the point is of this feature.
The proposal at hand is that this feature of assertion, which is to posit a world or a fact, stems from the need to virtualize the original causal nexus that the senses in a social setting. This must be practice that must have started with some simple features of language and which have been so successful that the practice and humans as as species have coevolved to the present status at this point in time. Whatever the details of this development, it can be characterised by the following general features. It is no doubt an extension of using sounds as natural proxies for perceptually relevant features of the surroundings, such as a scream or a sound to indicate some causally relevant feature. The problem with such natural proxies is that they are very localized. The trick that is the source of assertion is perform a Copernican turn with regard to the causal types in question. Rather than finding expression types that fit causal types, the trick is to construct causal types that fit linguistic types. This is the source of the “what is the case” that that is exhibited in the point of assertion, since the trick is to posit the existence of a world that is available to be matched with sounds. The language users and the world is in this way become a correlated system. The burden of the trick is that the practioners of his practice must carry the entire world inside language and must be biologically fit for this purpose. Assertion thus comes to serve a dual role. At the one hand it is the source of a calibration of the intersubjective causal types that all language users must share and accept, and on the other hand, it offers the possibility to call up this world in assertion just like it is perceived. The calibration sets up a guarantee for a match and thereby a bridge to non-perceptual systems that may be called up by a the practitioners along the way.
Language as we know it has a lot of structure, and and facts like “s is F” is only the simplest versions of what can be expressed. Central to the practice is no doubt the recognition of objects or substances and their attributes, the feature of compositionality to gain easy acess to any part of the posited world, and logial and inductive relations to be able to transverse the posited world without actually walking around in it. The features of generalization is inherent in the project of an intersubjective world, since the the very positiong of a world can be duplicated from particular features to general features of features. The connectives of logic are devices to expand the inter subjective worlds in various ways, and of these the feature of conditionalization is especially important because it allows perceptions of how the world might be in addition to how it is.
What is important for the present project is that these structural features are internal the practice of virtual perceptions and not something that exist in the world. The causal types of the world that are relevant to dispositions to assertions are simply those that satisfy this structure, and is imposed on the world must like the causal type “expensive” is imposed on the world. There is hence no need to look for interest independent causal circumstances that matches a particular aspect of a causal event, since it will be projected by assertional practice.
Bearing this outline of the nature of assertional practice we can finally return to the issue of whether this account of verbal dispositions can yield a causal theory. I think that it does, but it involves looking into what it means to provide a substantial explanation on several important accounts.
We can list the several requirements that must be addressed by a causal theory. First, there is the problem of identifying cause, which is tied the methodology of “reading off ” this cause, and secondly there is the problem of preserving meaning across all representation types, and thirdly there are the problems of objectivity that must be raised vis a vis the community theories.
Compared to the naturalistic theories, the causal types are essentially “minded” as they are picked out by their relation to practioners of this practice, much in the same way intimated by McDowell. There is a traditional distinction between production and consumer theories in naturalistic theories, but the distinction does not fit exactly, since the consumer terms are objective categories described by disinterested science. This theory places the criteria for the causal types for satisfaction in culture not in nature.
This leads us to the notion of ” reading off”. In naturalistic theories these are metaphysical dependencies that are to be revealed by scientific method. In this theory they are read off by the procedures of an interpreter, by a fellow practioner of the practice of assertion. The interpreter identifies the relevant causal types and use them to perform an interpretation the utterances the interpreted. This practice is certainly fallible, but its correctness is what constitutes reading off the relevant dispositions.
It might be objected that this does not count as substantial explanation on a number of grounds, but they should all be rejected. As has been mentioned, it is probably illegitimate for naturalistic theories to make use non-reductionism or ceteris paribus clauses to hide from the disjunction problem, since these devices are compatible with alternative ascriptions. It is this sort of problem that means that strict reduction in non-intentional descriptions of the “Type 1” sort is most likely called for to provide a legitimate explanation of the meaning determining cause. The situation is different with regard to (AR) because the point of calibration provides an a priori guarantees that the meaning determing causes are satisfied. To calibrate a causal type is simply to calibrate a type that makes assertions play the roles they are assigned. The ultimate criterion for success is not simply the fluency and efficiency of communication. It might further be objected, as it has been with Davidsons reinforcement version, that it is inherently circular it depends on the kinds concepts it is to explain. In Davidson’s version the calibration process is described as a piece of interpretative methodology, which requires mental states like beliefs on behalf of the interpreter. However, in this version bedrock in the explanation is the actual execution of the practice of assertion. The beliefs of the interpreter is to be explained in terms of this practice, rather than the other way around. It is a constitutive norm for assertion that the asserter claims only what is true, and driven in reverse, this norm determines the assignment of beliefs to the individuals.
Finally, it might be considered whether the account runs up against the problems of objectivity that community accounts do. It does not run up against the problem that agreement makes for truth, since according to this proposal they can be revised at any point. The agreement is rather found in the positing of an intersubjective world serve as a regulative ideal. Nor does it mean that the world is “constructed” in the sense of being mirror images of the assertions themselves. Naturalistically, they are nothing more than causal types that are relational, that occur as a conjunctions of an organism with its surroundings, in the same way as a great deal of other natural relations. The intersubjective world being a regulative ideal, these causal types are not set in stone and are dynamically expanded to fit the needs of communication.