I haven't read that Pinker book.

I mean that these type of models are very restrictive and don't describe the full spectrum of data available. There are more phenonmenon taking place then the model describes. So looking at all the available data seems more optimal to me.