Addendum to Ivan Handler's comments, which I pretty much agree with:
Believing in genetics and evolution --not ONLY cultural evolution--
and seeing their relevance to the appearance of social behavior
including language, I find myself faced with a huge conceptual
challenge which I have not met.  Nobody else has met it either so
I shouldn't feel bad.
        Namely, when we get beyond the silly superficial models
which the genetic determinists tolerate --when we understand that
there can't be a gene for the passive voice any more than there can
be a gene for altruism-to-first-cousins-by-marriage-- then we see
that it is hard to hypothesize traits relevant to the topic which
ARE the kind of trait a gene could conceivably enable.  At least
to me this seems exceedingly knotty.  Without having accomplished
any progress on it, I venture to guess that enabling syntax is not
disjoint from enabling other kinds of organization of activity, so
that genes which do it could be selected for even in a population
which was not relying on symbolic communication.  End of guesswork
for today.
                                Chandler Davis