Tacit knowledge of theories of meaning?

I’ve been thinking about the use of the concept of tacit knowledge in the context of compositional semantics. What is the relation between the kind of theories of meaning articulated in formal philosophy of language or linguistic and the linguistically competent speaker? It would be attractive to explain their competence in speaking a language, no less than their competence in making a bechamel sauce, as an expression of some sort of knowledge. But it seems that since the project has proved so difficult and since ordinary speakers typically do not recognise formal theories of meaning as modelling their competence – there’s no hitting the forehead and saying “Yes of course” – it is implausible to say that this is a matter of explicit, person-level knowledge. And hence the idea that it might be a form of tacit knowledge.

But there’s an influential objection to this line of thought as one might first construe it. The output of a theory of meaning is usually taken to assign meanings to whole sentences of an object language via a statement of their truth conditions. (Knowing the condition under which a sentence is true looks to be a good way of stating its meaning and working with truth is more straight forward than working with meanings directly.) These sentence by sentence meaning theorems might well be objects of knowledge. But the theory that yields them is usually assumed to have axioms governing names and predicates and it is these – together with other upstream aspects of the machinery of the theory – which do not seem to behave like the objects of knowledge. Or at least, they do not seem to behave like ordinary intentional states or propositional attitudes.

The objection stems from Evans but this is Wright’s endorsing summary considering belief status.

A rat may acquire the disposition to avoid a kind of foodstuff which is poisonous and has caused it sickness in the past. and we might casually ascribe its unwillingness to eat this material - or one that looked/smelled similar - to the belief that it was poisonous. But we should not, Evans urges5, let casual language induce casual thought. Beliefs are essentially things which interact with desires and intentions in the production of behaviour. They are also essentially involved in the production of other beliefs. To ascribe a belief is significant only as part of the ascription of a system of beliefs. And what behaviour is expressive of a certain belief depends, in general, upon the ingredients in this system and in the system of the subject’s intentions and desires. Thus, my belief that a certain substance is poisonous may manifest itself in a literally indefinite variety of ways. I may, like the rat, avoid the substance. But I may also take steps to ensure my family avoid it, or take steps to ensure they don’t! I may take small but daily increasing quantities of the stuff in the belief that I can thereby inure myself against its effects, and .that background circumstances are such that it may stand to my advantage to have done so. I may take a large quantity if I wish to commit suicide, and a smaller one if I wish to incapacitate myself so as to avoid an obligation. My belief that the substance is poisonous is thus, as Evans puts it, at the service of indefinitely many potential projects corresponding to indefinitely many transformations in my other beliefs and desires. With the rat, in contrast, concepts like the desire for suicide, or malign intent, can get no grip. The ‘desires’ which we are prepared to attribute to it are restricted, in the present context, to avoidance of distress; and its ‘belief’ that the substance is poisonous has consequently no other expression than in shunning it. [Wright 1986: 33]

A belief should be at the service of many potential projects. But a belief that encapsulated the axiom of a theory of meaning would be only that the service of derivations of theorems giving the meaning of sentences. Hence it looks the wrong sort of thing to count as doxastic, or an intentional state, or a propositional attitude.

So if we are to deploy some use of the notion of tacit knowledge in this case, it cannot be thoughts of as any form of attitude for the speaker. Nevertheless, Evans does wish to do this. He thus summarises the discontinuity thus:

Tacit knowledge of the syntactic and semantic rules of the language are not states of the same kind as the states we identify in our ordinary use of the terms ‘belief’ and ‘knowledge’. Possession of tacit knowledge is exclusively manifested in speaking and understanding a language; the information is not even potentially at the service of any other project of the agent, nor can it interact with any other beliefs of the agent (whether genuine beliefs or other tacit ‘beliefs’) to yield further beliefs. Such concepts as we use in specifying it are not concepts we need to suppose the subject to possess, for the state is inferentially insulated from the rest of the subject’s thoughts and beliefs. There is thus no question of regarding the information being brought by the subject to bear upon speech and interpretation in rational processes of thought, or of making sense of the subject’s continued possession of the information despite incorrect performance, due to his ‘not thinking’ of the rule at the appropriate time, etc. [Evans 1981: 133]

So what is meant by tacit knowledge if it doesn't connect to other states of/for a speaker? Evans starts his positive account from this position:

I suggest that we construe the claim that someone tacitly knows a theory of meaning as ascribing to that person a set of dispositions – one corresponding to each of the expressions for which the theory provides a distinct axiom. [ibid: 124]

The challenge that then seems to be picked up Evans, Wright, Davies and Miller, among others, so to try to tighten that claim so that some, at least, extensionally equivalent theoretical codifications of competence are not ascribable as tacit knowledge, so understood. I will mention two aspects.

I’ll assume that everyone is familiar with the agreed examples and I’ll talk about meaning for ease. There are three theories of meaning.

T1, comprises a simple list of 100 combinations of 10 names and 10 predicates to yield 100 sentences with assignments of truth conditions for each.

T2 has 10 axioms giving the meanings of the 10 names and 10 axioms giving the meanings of the predicates and one compositional principle explaining how predicating a predicate of a name works (ie yields a truth condition for the whole).

T3 in which the last axiom is apportioned into the 10 predicate axioms. The axioms giving the meanings of the predicates do this by saying stating the truth condition for varying names.

Given that T1-3 all yield the same meaning theorems covering the 100 sentences, what are the grounds for ascribing tacit knowledge of one rather than another?

The answer Evans suggests and Davies and Miller develop is that patterns in the development or loss of the capacity can serve as evidence for the underlying causal structure that underpins the dispositions codified in the theory to distinguish T1 and T2. Davies adds the idea that the pattern of revision of an axiom distinguishes between T2 and T3. Hence there is more in play than just the derivation of the meaning theorems known by a competent speaker. There is a further causal structural pattern that can be mirrored better or worse by different extensionally equivalent theories. Hence there isn’t an objection from overly much indeterminacy to the idea of calling the ‘attitude’ to the meaning theory ‘tacit knowledge’.

It is worth recalling how this bit of Davies’ argument works. He considers a speaker (C) who has a kind of internal language of thought with a theory of meaning represented internally as T1 - this we just assume - but where the theorems that store the output theorems of meaning are stored in a 10x10 matrix memory but which is supplied by two types of nutrient. Loss of nutrients a memory cell produces loss of memory of the theorem in that cell. The idea is that despite the fact that the theory of meaning is T1, it is stored in a causal structure that has the internal logic isomorphous with T2 and so failures of nutrient flow will produce patterns of language loss akin to patterns of language loss had the speaker been following T2 on their inner blackboard and lost knowledge of axioms, whether of a name or a predicate. Nevertheless, the causal structure is not a structure of semantic theory. Davies suggests that the problem is that the causal structure is not sensitive to the information stored. That is why it is not sufficient for tacit knowledge of a theory of meaning. So then the idea of revision of axioms - and it’s consequent pattern - comes in to suggest a tighter constraint between causal structure and information. But note that revision is merely a tighter connection than the previous versions of the mirror constraint. They too had looked to link the right sort of theory with the right sort of causal structure. Why think that there is no equivalent of the nutrient story for the new version? Davies has not proposed a full-blown account of what information is in this new theory and how information is essential to understanding whatever causal transactions are involved. (Let me here register doubt that he can pass the crucial job description test of making the system informational as opposed to merely causal.)

Wright additionally complains that there is no justification to think of the mirroring theories as trading in meaning or semantics. He suggests that the mirroring of causal structure could be thought of as merely showing a sensitivity to syntax not semantics. Then, additionally, further text could be added to the theory to connect this to competence undermining the basic ascription of tacit knowledge of meaning axioms. While the mirror constraint would provide reason to favour T3 over T2, say, it would not be reason to ascribe tacit knowledge of axioms concerning reference. Miller responds that whatever the theory states, if its satisfactoriness is constrained in such a way as to yield meanings, then it’s a theory of meaning. (This reminds me of how McDowell suggests that Davidsonian truth theories serve, despite that, as finer grained theories of meaning.)

I don’t want to dispute any of this but it does seem to me that there is a sleight of hand. With the assumption that there would be something wrong with ascribing tacit knowledge of any particular theory of meaning if there were no way of eliminating different theories that yielded the same meaning theorems, this literature has come up with plausible ways of narrowing things down. Additional evidence of gaining, losing and revising understanding serves as further constraints on what the ascription of tacit knowledge of a theory amounts to. Since explicit knowledge couldn’t tolerate the indeterminacy – there would be all the difference in the world in explicitly knowing the different theories T1-T3 – narrowing things down looks to be a plausible move to justify the ascription of tacit knowledge. But we haven’t really moved very far from the starting point. Evans plausibly argued that the bodies of theories themselves were not the objects of intentional states or propositional attitudes, and that claim remains. He suggested that we could use the phrase ‘tacit knowledge’ for dispositions codified in theories of meaning and this has been tweaked. But the subject who has linguistic competence need have no conception of the upstream elements of the theory. The conceptual articulation within the theory need not be something the subject has any grasp of. While either the statement of the theory, or a description of the derivation of the theory (to get round Wright’s point), by a theorist will be intentional and semantic in nature, there is no reason to think that this articulates intentional states of/for the speaker. And hence there is something misleading about linking the articulated theory to the speaker through the relation of tacit knowledge. This use of ‘tacit knowledge’ has no connection to reasons, justification or concepts. It is not a standing in the space of reasons at all.

I want to say: it isn’t any form of knowledge.


Davies, M. (1987), "Tacit knowledge and semantic theory: can a 5 % difference matter?", Mind 9 6, pp.441-­‐62.
Evans, G. (1981), "Semantic theory and tacit knowledge", in S. Holtzmann and C. Leich (eds), Wittgenstein: To Follow a Rule. London: Routledge, pp. 118-­‐37.
Miller, A. (1997). ‘Tacit knowledge’. In Hale, B. and Wright, C. (eds). A Companion to the Philosophy of Language. Oxford: Blackwell
Wright, C. (1988), "How can the theory of meaning be a philosophical project?", in Mind and Language 111, pp. 31-­‐44.