My point of departure on this research path is the observation that some of us have different grammatical generalizations from others. For example, while some people would say (1), others would say (2):
(1) It is you who are confused.
(2) It is you who is confused.
A language critic could say that (2) is the wrong generalization: it fails to take into account that you is a second-person pronoun, and therefore requires the second-person singular form of the verb be. But a language critic could also argue that (1) is the wrong generalization: the subject of the second verb is not you but who, a third-person singular pronoun. In fact, (1) and (2) are equally valid solutions, and both are attested in data retrievable from google:
(3) It is you who are displaying distinct Talibanistic characteristics.
(4) You are wrong, Mr. Blair: It is you who is prejudiced about science.
The source of such grammatical indeterminacies is not faulty grammatical reasoning; the fault lies with the construction itself. The construction exemplified in (1-4), known as the cleft construction, fills a conversational need, but in the process creates a grammatical paradox: cleft sentences place the focus (you in 1-4) in the preferred position for new information—right after the main verb—but obscure the grammatical role of the focal expression: is it the object of the main verb or the subject of the following verb? Such grammatical trouble spots provide opportunities to explore the problem-solving abilities that speakers use during production—abilities that drive syntactic innovation and, in turn, grammar change. I assume that the primary agents of syntactic change are not children but adults. As skilled users of the grammar, adults know best how to extend its potential. This is not to say that adults’ solutions will be perfect or even elegant: speakers, like Lévi-Strauss’s (1966) bricoleur, will make do with whatever materials are ready to hand. Like Lass 1990, 1997, Traugott 2004 and Narrog 2007, I see syntactic bricolage as analogous to exaptation in evolutionary biology, although my focus is not on exapted structures per se, but on the communicative problems that those structures are designed to solve. Grammatical solutions like the cleft construction illustrate the principle that “functional systems must trade off conflicting demands” (Jackendoff and Pinker 2004: 26). Optimization in one sphere (say, discourse) may lead to suboptimal results in another sphere (say, syntactic parsing). For further development of these ideas, see the slides of my plenary talk, "Traces of Grammar Evolution: Protoconstructions, Amalgams and Mismatch Effects", given in July 2007 at a  on the statistical physics of social dynamics.