Previously I detailed why T sentences alone couldn’t be an adequate theory of meaning. The specific objections that I raised to them were that for some languages no finite set of T sentences would give an understanding of meaning for every sentence, there were no valid T sentences for some perfectly meaningful natural language sentences, and that they had problems dealing with “subjective” sentences; basically that there were some parts of the language that the T sentences provide an understanding of. However, this is not the only appearance T sentences make in the philosophy of language. Another related proposal is that the meaning of a language is some set of rules that convert every sentence into a T sentence. In some ways this theory is reminiscent of the one that I proposed, although I argued that the transformation rules worked on words, not on sentences, and that they were transformed into a kind of “intending”, not truth conditions (see here). Such transformational T sentence theories do overcome many of the problems facing “pure” T sentence theories, but unfortunately they still fall short of being an adequate theory of meaning.
What is an adequate theory of meaning? That is a subject of much debate, which I won’t describe here. To evaluate the transformational T sentence theory of meaning I will rely on the simple, and I should hope undisputed, requirement that given a theory of meaning for a language we should be able to understand every correct utterance made in that language. The problem, then, for transformational T sentence theories of meaning is not that they are inconsistent, but simply that they don’t pull enough weight, or so I claim.
The first obvious gap for a transformational T sentence theory is exclamations, like “ouch!”. Strictly speaking it isn’t a sentence of the kind dealt with by the theory, since “‘ouch!’ is true if and only if ouch” doesn’t make much sense, and is even less informative. It does seem to have a meaning though, the same meaning as “I am in pain”, which is a proper sentence, and can be dealt with by the theory. What is needed then is an additional set of transformation rules that take utterances that aren’t the right kind of sentences (that don’t straightforwardly assert a claim) and convert them into sentences with the same meanings that can by transformed into a T schema. This might seem to be a problem with the transformational T sentence theory, but since these additional rules are in the same “spirit” as the original theory, and because everything ultimately still must be transformed into a T sentence, I think it is reasonable to consider these rules only a minor addition to the theory, and not an invalidation of the whole enterprise.
Another potential problem for a transformational T sentence theory is ambiguity in the natural language, either in the form of sentences that can mean more than one thing or sentences that don’t pin down the possibilities exactly (i.e. “it might be dark out”). Some proponents of these kinds of theories simply pass over ambiguity, arguing that if we simply carry it over into the meta-language it isn’t a problem. It really is problematic, though, for two reasons; because if the meta-language is supposed to be in terms of how the world really is it doesn’t seem like the world contains the necessary kind of ambiguity (except, perhaps, at the quantum level), and also because sentences like “it might be dark out” are strictly speaking always true, and thus being given their truth conditions doesn’t yield any understanding. However, both these worries can be resolved if we allow sentences with multiple meanings to be transformed into T schema that give their truth conditions in terms of a disjunction of the possible meanings, and sentences like “it might be dark” to be transformed into sentences that make claims about the knowledge of the speaker, e.g. “the speaker doesn’t have enough information to conclude definitely that it is dark out”.
A transformational T sentence theory of meaning may be able to deal with ambiguity, but questions are another matter. We might attempt to handle questions, such as “who is the murderer?” by transforming them into statements about the speaker’s intent, for example into “the speaker wishes to know the identity of the murderer”. Unfortunately that isn’t always the case, maybe the speaker really does know, and is trying to play innocent, so we might transform it into “the speaker wants someone else to respond with the name of the murderer”. But again, it is possible that the question may be rhetorical, ect. Essentially questions simply aren’t about states of affairs, and so no truth conditions can really capture the meaning of the question, and hence transformational T sentence theories are unable to help us understand them.
Another serious problem for transformation T sentence theories are sentences that are always true (tautologies). All tautologies have the same truth conditions since they are always true (in a sense they don’t really have truth conditions at all, since their truth isn’t conditional upon any other facts). However, not all tautologies seem to have the same meaning, for example “snow is snow” tells you something different than “(A or B) and ~A implies B”. And, in a transformational T sentence theory of truth, they would both be converted into something like: “‘snow is snow’ is true if and only if T” and “‘“(A or B) and ~A implies B’ is true if and only if T” (where T indicates that it is a necessary truth, not contingent upon anything). Assuming that you didn’t know what the sentences in the object language already meant, these T sentences will not further your understanding of them, and hence the transformational T sentence theory of meaning fails to give us an understanding of every sentence in the language.
So there are at least two kinds of sentences, questions and tautologies, for which transformational T sentence theories of meaning come up short. Obviously I think the problem lies in trying to force the results to be T schema, since that is how my own theory about meaning differs from it, but I will let the reader draw their own conclusions.