Preproceedings 2006 - Austrian Ludwig Wittgenstein Society
Preproceedings 2006 - Austrian Ludwig Wittgenstein Society
Preproceedings 2006 - Austrian Ludwig Wittgenstein Society
Create successful ePaper yourself
Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.
Natural Language and its Speakers: Davidson’s Compositionality Requirement - Isaac Nevo<br />
set of semantic primitives that is constructed along these<br />
lines, i.e., by aggregating the primitive vocabularies of all<br />
actual speakers, need not be attributable to any single<br />
speaker. It may be too large, or too counterfactual, for that<br />
purpose. Just uniting the various sets of primitives that<br />
different speakers use need not work up to a workable<br />
theory, even if we settle for a finite set of such sets<br />
according to the finite number of speakers there actually<br />
are.<br />
Secondly, the set of finite vocabularies of finitely<br />
many speakers may not exhaust the language either, for<br />
given any such enlarged vocabulary, the language may<br />
still grow and change. The latter set is, of course, finite,<br />
and no language will ever have more than a finite number<br />
of speakers. But this de-facto finitude of language is not a<br />
consequence of its truth-theoretic form, and it is<br />
compatible with the language being potentially infinite, i.e.,<br />
with it’s not being exhausted by any finite theory. As to the<br />
set of primitives that would exhaust the language in terms<br />
of its potentiality of expression, there is no guaranteed that<br />
it is a finite set. The set of all finite sets of primitives that<br />
would be needed to derive any sentence that might be<br />
formulated in the language need not be a finite set. This is<br />
not to say, of course, that any natural language is in fact<br />
infinite, only that it is potentially so (and the infinite<br />
potentiality of a natural language is what gives it its<br />
prospects of development and growth).<br />
But why is potentiality of expression relevant for the<br />
empirical study of language? Admittedly, no empirical<br />
theory should be held accountable to such a notion of<br />
potentiality. An empirical theory need not be made to<br />
account for all the different ways in which a language may<br />
develop and change. The point is only that it should not<br />
preclude such changes and developments, by imposing –<br />
non-empirically – a theoretical individuation of languages.<br />
Finite and empirical theories are all very well as long as<br />
they are not taken as exhaustive of the phenomenon<br />
under discussion, i.e., as long as they do not replace their<br />
own principles of individuation for the more customary<br />
ones.<br />
3. The Context Principle<br />
In a famous passage, Davidson extends Frege’s context<br />
principle from sentences to whole languages. His<br />
argument is as follows:<br />
We decided a while back not to assume that parts of<br />
sentences have meaning except in the ontologically<br />
neutral sense of making a systematic contribution to the<br />
meanings of the sentences in which they occur . . . One<br />
direction in which it points is a certain holistic view of<br />
meaning. If sentences depend for their meaning on their<br />
structure [WCP, I add], and we understand the meaning<br />
of each item in the structure only as an abstraction from<br />
the totality of sentences in which it features [WCP,<br />
again], then we can give the meaning of any sentence<br />
(or word) only by giving the meaning of every sentence<br />
(and word) in the language [SCP. Notice how “language”<br />
slips in to replace the lesser totality of sentences in<br />
which a word appears]. Frege said that only in the<br />
context of a sentence does a word have meaning; in the<br />
same vein he might have added that only in the context<br />
of the language does a sentence (and therefore a word)<br />
have meaning (1984: p. 22. Bracketed comments<br />
added).<br />
However, the passage from sentence to language<br />
as the minimal unit of meaning is more slippery then<br />
Davidson would have it. The totality of sentences in which<br />
a word occurs, or even the totality of sentences in which a<br />
particular finite vocabulary occurs, need not be the same<br />
as the whole of language, at least not if “language” is<br />
meant to apply inclusively to natural, as well as formally<br />
constructed languages. And while it is plausible to assume<br />
that the meanings of the words are abstractions from the<br />
roles that they play in the sentences they occur, it is much<br />
less plausible to assume that these various totalities must<br />
all merge into a single formal unity. Vocabularies and the<br />
sentences that can be expressed by them hang together,<br />
but different vocabularies (and different totalities of<br />
sentences) need not be unified under a single theoretical<br />
structure. In this respect natural languages are very<br />
different from formal systems. They evolve continuously<br />
without loosing their identity, and they constitute “wholes”<br />
that are greater than those captured by any particular set<br />
of rules and words. Hence, to the extent that Davidson’s<br />
holism consists in imposing a single theoretical structure<br />
on the whole of language, it freezes a dynamic human<br />
activity into an abstract and a formal entity.<br />
In ‘Reality without Reference’ (1984[1977])<br />
Davidson poses a dilemma for the theory of meaning. It is<br />
the following. No theory of meaning could be adequate<br />
unless it supplied an account of truth (since ‘truth<br />
conditions’ are such a central feature of what a theory of<br />
meaning is a theory of). Hence, also, no theory of meaning<br />
could be adequate unless it supplied an account of<br />
reference (since the determination of reference must enter<br />
into the determination of truth). However, any reductive<br />
theory of reference, i.e., a theory of reference in nonsemantic<br />
terms, is bound to lead to results that are<br />
unsatisfactory in various respects. First, such ‘building<br />
block’ theories tend to be metaphysical in nature, reducing<br />
semantics to non-semantic facts. Secondly, no such<br />
reduction can do as well as reference in point of underlying<br />
the compositionality of language. Only the semantic<br />
feature of satisfaction has in fact been given a<br />
compositionality-supporting account (Tarski’s). Thus, an<br />
account of meaning requires an account of truth, and the<br />
latter account requires an account of reference, but no<br />
account of reference that could be independently given<br />
satisfies other requirements of the theory of meaning,<br />
particularly the requirement of compositionality.<br />
Davidson goes on to formulate what he calls the<br />
‘paradox of reference.’ It is the following. Although the<br />
‘building block’ method of semantics, which starts by<br />
accounting for the meanings (references) of constituent<br />
terms and builds upwards to the meanings (truth<br />
conditions) of the sentences in which they occur fails as a<br />
theoretical account, the alternative, namely, the holistic<br />
approach, which views the semantic features of subsentential<br />
components (predicates) as mere abstractions<br />
from the meanings of sentences in which they occur, is<br />
incapable of giving ‘a complete account’ of the meanings<br />
of these sub-sentential components, and, consequently,<br />
no account of truth either. Davidson’s resolution is ‘to<br />
defend a version of the holistic approach, and urge that we<br />
must give up the concept of reference . . .’ (1984: p. 221).<br />
The point is that relative to a finitely-based theory which<br />
makes possible a derivation of truth conditions to all<br />
sentences of the language, reference and satisfaction<br />
could be seen as mere theoretical constructs, for which<br />
there should be no non-theoretical role (and no reductive<br />
account). Given that semantic theory could find its<br />
empirical basis in the connection between sentences and<br />
extra-linguistic facts, no further evidence is needed<br />
regarding the semantic features of sub-sentential<br />
components. Words and their various connections with<br />
223