In a recent issue of this journal, Edward Gibson published a review of ‘The
crosslinguistic study of sentence processing’ (CSSP) -- a book which contains several
chapters summarizing our recent crosslinguistic work and those of our collaborators in
several countries. The primary goal of CSSP was to explore the predictions that derive
from a minimalist model of language processing and acquisition which we have called the
Competition Model.
Gibson’s review of CSSP exposes a set of deep fracture lines separating two
competing approaches to psycholinguistics. One one side of the fault zone, the
Competition Model focuses on the construction of a minimalist model designed to predict
exact numerical values in controlled studies. In this approach, an initial candidate model is
not responsible for predicting all levels of all variables for all sentences in all languages.
Instead, the model is developed inductively on the basis of a constrained set of sentence
types and certain limited predictions. On the other side of the psycholinguistic fault zone,
the UG model that Gibson espouses avoids minimalism and emphasizes the complexity of
the theory of formal principles of language structure motivated by binary acceptability
judgments generated by small numbers of professionally-trained theoreticians