Definition

Quine looks at the notion of definition to see whether it will do this job. Where do the definitions we need for this approach come from?

The first reaction to a request for a definition is to look in a dictionary for the definition of any word; we will call this the lexical definition. However natural the use of dictionaries is, it won't work for the task at hand for two reasons. First because dictionaries rest upon a prior understanding of synonymy. Lexicographers engage in empirical investigation of the uses of words to find out what should be encoded in the dictionary. They are trying to find out what is synonymous within the language they are studying. Second, it won't work for reasons which don't appear until Word and Object: the lexicographer can hope only to discover what Quine calls ``stimulus synonymies.'' Apart from the fact that they are ``discoverable'' only on the lexicographer's analytical hypotheses, these synonymies are available only for observation sentences, and they are not strong enough to support the sort of definition or meaning which is needed to sustain Kantian analyticity.

There is another sort of definition which may prove more suitable for the task: explication. Explication isn't the purview of the lexicographer or philologist alone, philosophers routinely engage in it (indeed it is the life-blood of the analytic philosopher).

In explication, the aim is not simply to paraphrase a term in another set of, hopefully, more easily understood terms which are synonymous with the original term. Instead explication aims to improve upon--go beyond--the synonyms, the current definition. The idea behind explication is not to create synonymies, but to ``preserve the usage of these favored contexts [i.e., the contexts in which the use of the term being explicated is generally agreed upon] while sharpening the usage of other contexts'' (25/66b)

Once this is said, the problem is obvious: explication, just like lexicography, rests upon a prior understanding of the synonymy relations, it rests ``nevertheless on other pre-existing synonymies'' (25/66a).

And indeed there is yet a further problem. If the goal of explication is to clarify and expand the usage of a term, it is not only possible, but highly likely that two theorists with different interests may produce explications which succeed in the goal of explication, but do so in ways that are incompatible with one another. One simple way in which this may occur is for the theorists to address different subsets of the ``other contexts.'' However, as Quine notes, even if the theorists are addressing the same subset.

Two alternative definitions may be equally appropriate for the purposes of a given task of explication and yet not be synonymous with each other; for they may serve interchangeably within the favored contexts but diverge elsewhere (25/66-7c)

There is one more sense of definition which may save the day: stipulative definition. Stipulative definitions do ``not hark back to prior synonymies at all'' (26/67). These definitions are ``explicitly conventional'' and are introduced ``for purposes of sheer abbreviation'' (26/67). Here `A' becomes synonymous with `B' simply because the speaker says so. According to Quine, this works insofar as it frees us from pre-existing synonymies, but there are two problems:

1. it's a very small sub-set of definitions; and
2. it tells us nothing whatsoever about the semantic relationshipof synonymy.
   

In sum, there are three ways in which a definiendum may be related to a definiens.

i.

definiens is a faithful paraphrase of the definiendum, preserving a prior relationship of synonymy;

ii.

definiens improves upon the usage of the definiendum by explicating the antecedent usage of the definiendum--either by sharpening the existing meaning or by sharpening and redirecting existing synonymies;

iii.

the definiendum is a newly created notation; the definiens is attached to it by an act of fiat.

Only the last escapes the charge of employing prior synonymies, but it doesn't address our targets: synonymy, meaning and analyticity.