Foundational issues 1

Ray Jackendoff is well qualified to talk about the trend towards increasing specialization and fragmentation of the science of language. For over 30 years he has been a prolific contributor to a number of different topics in syntax and semantics, and has also made notable contributions to other areas of cognitive science, including music, consciousness, spatial cognition, and psycholinguistics. It’s hard to find many in the field who can match Jackendoff’s breadth. Therefore, all linguists should be interested in Jackendoff’s most recent book, Foundations of language, in which he lays out a series of objectives for the science of language, and describes some of the steps that he believes are needed in order to reintegrate theoretical linguistics with psycholinguistics and even computational neuroscience. We find a lot to like in Foundations of language, and we are in broad agreement with Jackendoff on the set of goals that he proposes. In particular, we agree with his claim that the time is ripe for linguists to pay more than lip service to the long-standing mentalistic commitments of the field. However, we disagree with Jackendoff’s core theoretical proposal, which is that key challenges for unifying linguistics, psycholinguistics and neuroscience are solved by his ‘ tripartite parallel architecture’, which he describes as replacing traditionally dominant ‘syntactocentric ’ models with independent generative systems for phonology, syntax, and semantics. We find the linguistic arguments for the architecture to be quite limited, and we are also not convinced that the architecture confers the psycholinguistic benefits that Jackendoff claims for it. We think that the real challenges for real-time