According to Gould and Lewontin, adaptive explanation in evolutionary biology suffers from an overabundance of riches. The problem is that it is so easy to come up with a plausible adaptive explanation for a given phenotypic characteristic that, at the very least, researchers never bother to consider alternatives to adaptive explanation. As a result, adaptive explanations come out as merely hollow, just-so stories.
I don't intend to here defend Gould and Lewontin's indictment of the adaptationist program (though I do think that many recent responses--especially by advocates of evolutionary psychology--just miss the point). What I do want to suggest is that a similar problem infects pragmatic explanations in linguistics and the philosophy of language. Specifically, it has become an increasingly popular strategy to deal with uncomfortable, seemingly semantic phenomena by invoking a pragmatic explanation (usually grounded in some sort of implicature) according to which the phenomena is not really semantic after all. And if the particular pragmatic explanation invoked can be shown to be mistaken, it is easy enough to cast about for another one. As a result, these pragmatic explanations have become little more than linguistic just-so stories.
The problem is pernicious because the ready availability of pragmatic explanations (based, as they so often are, on non-overt chains of practical reasoning) allows one to preserve one's favorite semantic theory "come what may".
Now Grice (writing in the shadow of ordinary language philosophers) proposed a principle of semantic parsimony: do not proliferate senses (meanings) beyond necessity. Being the philosophical grandchild of Grice, I think this is quite good advice--at least once we have a good grip on what counts as "beyond necessity."
Unfortunately, a second commonplace of contemporary philosophy of language, namely Frege's context principle, has conspired to muddle the issue. As Frege states it, the context principle is "to never ask for the meaning of a term in isolation, but only in the context of an entire sentence." And routinely this is understood as the thesis that the meaning of a sentence is nothing more (or less) than the contribution it makes to the compositional meaning of all the sentences in which it occurs.
Now, as someone who advocates a construction-based approach to grammar, I am skeptical of the context principle. But putting theoretical concerns to one side, the context principle has effectively devalued our lexical intuitions about word meaning. And, in conjunction with Grice's principle of parsimony and the availability of pragmatic explanations, this seems to me to be a very bad thing. For if we discount our intuitive sense of what various words mean, all Grice's principle requires is that we be able to get some basic sentences involving a given word up and running and any further "meanings" can be sloughed off into the waste bin of pragmatic just-so stories.
Well, even I admit that the preceding view is too strong--all of us accord some weight to lexical intuitions concerning, for instance, clear cases of ambiguity. And all of us do admit that in some cases a pragmatic explanation is just not acceptable. But the essential point remains.