Abstract
I think that some of the arguments in this article are themselves flawed, or are based on an understanding of linguistics that is too narrowly focused on certain versions of generative grammar. For example, the argument that in computational applications purely statistical approaches are in general more successful than rule-based approaches has to be qualified: It holds, or may have hold, for certain applications like machine translation, but not for others, like the generation of text to answer queries to databases. Furthermore, statistical methods have been integrated in certain linguistic theories themselves, like stochastic optimality theory (Boersma & Hayes 2001). The authors also claim that linguistics is not able to come up with leading questions for the cognitive and neurophysiological investigation of language processing. This statement is even more puzzling, as it is difficult to name serious research in psycholinguistics or in neurophysiological aspects of language processing that is not informed by theoretical notions rooted in linguistics, many of them derived from generative grammar. To cite just one case: Recursion has been proposed – perhaps unjustly so – as the single property that distinguishes human language processing from other animal communication systems; this has led to the identification of special brain regions and pathways of recursive language processing (cf. Friederici 2009).