conceivedof as theoutput of syntacticderivation, which has the feelof a relatively late innovation.It is of course hard
to bring evolutionary concerns to bear on the gritty details of language. Nevertheless, the central clai mof generative
grammar is that the language faculty is a cognitive specialization, and we claim that cognitive specializations in every
other species are the consequence of natural selection. So it behooves us to take responsibility for an evolutionary
scenario, even if at the moment the picture is far from clear.
The parallelarchitecture turns out topermita nicelyarticulatedstory ofincremental evolution.Theinitialinnovationis
permitting sounds to be used symbolically in a non-situation-specific fashion. This is admittedly a mysterious advance
over primate calls, but vastly less expressivethan modern language. From therewe are able to see the evolutionof the
language faculty as the successive addition of more and more“tricks” to the toolkit, in the interest of greater
expressiveness and precision in conveying thought. In particular, syllabic and then segmental structure within
phonology developedas a way of reliably increasing vocabulary size, and interface principles relating semantic roles to
linear wordorder suppliedafirst roughcutonexpressing relationsamongwordmeanings. Phrase structure, inflection,
and grammatical functions then developed successively as machinery for expressing more complex hierarchical
relations. The more recent innovations are the ones most subject to disruption, critical period effects, and above all
variation fro mone language to the next. And even within the gra m mar of modern languages it is possible tofind
remnants of more“primitive”stages that have not been completely overlaid by modern grammar, as well as a“basic
body plan”for syntax that has consequences throughout the grammar.
Finally, mainstream generative grammar has lacked a developed theory of semantics. In reaction, the two most
prominent (American) schools of semantics have rejected generative grammar, in different ways: psychological issues
play virtually no role in formal semantics, and issues of syntactic combinatoriality are downplayed in cognitive
grammar. The parallel architecture acknowledges an important insight of both schools: that meaning has its own
characteristic combinatorial structure, one that is not simply“read off” syntax. But our global approach to the
language faculty has yielded more: a valuable perspective on the relation between syntax and semantics. In particular,
the syntax–phonology interface, the interfaces among the phonological tiers, and the interfaces between phonology,
acoustics, and gesture give us afixon what to expect fro mthe syntax-se mantics interface. We should not expect an
isomorphism: syntaxshouldnot(alone)determinesemantics, as inmainstream generativegrammar and manyversions
of formal semantics; but neither should semantics entirely determine syntax, as oftenasserted in cognitivegrammar. It