i'm not interested in an assessment of this position, because that would take effort, and i am a lazy son of a bitch. i just want to present an argument in favor of the latter proposition, that the form of language is not well-suited for communication. more particularly, i'll point out that one of joseph greenberg's universals, if in fact universal**, shows that no language tailors its morphology to a communicative function that it could do very straightforwardly.
the universal i'm talking about is # 37 in greenberg's famous "some universals of human language..." paper:
A language never has more gender categories in nonsingular numbers than in the singular.to illustrate, greenberg mentions hausa, which distinguishes masculine from feminine gender in the singular, but not in the plural***. the universal means that the reverse is never the case: you don't get languages in which there is a masculine/feminine distinction in the plural but not the singular, for example.
there's no reason in principle that this should be, other than the architecture of language. plenty of languages distinguish masculine from feminine plurals, and plenty don't distinguish them in the singular. it is therefore surprising that no language combines these two facts.
the reason this shows that linguistic form doesn't follow function is as follows: gender distinctions have a function: they disambiguate. for example, the gender morphology in "i saw john and mary and kicked him" makes it clear who i kicked. but there are more actual "real world" genders in the plural than in the singular. in the singular, you have masculine, feminine and neuter. in the plural, you have masculine, feminine, neuter, and mixed gender to refer to a plurality consisting of both men and women. if form followed function, you might expect the universal to go in the opposite direction.
but it doesn't. and i'm out.
* to which generative linguists say "duh!"
** some of his claimed universals turn out not to be so once languages are added to his sample. for example, his universal 31 - "If either the subject or object noun agrees with the verb in gender, then the adjective always agrees with the noun in gender" - does not hold of tamil, although it does hold of closely related malayalam.
*** the universal uses "nonsingular", which encompasses plurals, as well as duals, trials and whatever other nonsingular morphology might exist.