Evaluating Deficit Patterns of Broca Aphasics in the Presence of High Intersubject Variability

AbstractA hypothesis about the existence of a computation for which Broca's area is indispensable would take the following form: identify two classes of sentences such that one involves the relevant computation and the other does not, then form a motivated prediction on the way a deficit in this computation should affect the performance of the Broca aphasics. The concluding step would be to evaluate the prediction in light of the empirical record of performances of Broca aphasics. This chapter is mainly concerned with this last step. It looks for a way of modeling the data that leaves room for sharp statistical claims, such as: “The performances of Broca aphasics on sentences of type A are significantly different from their performances on sentences of type B” while taking into account the high variability of the data. The chapter reports on the principles that have guided the compilation of performance data to which the methodology will be applied. It then briefly explains the trace deletion hypothesis (TDH) and its prediction concerning patterns of performance of Broca aphasics. It also presents a very simple stochastic model for the performances of a given subject and deduce a way of summarizing one individual's performance on a set of sentences. The chapter then illustrates the high individual variability of performance data viewed through the above statistics and presents a model for evaluating group performances (the beta binomial). Finally it applies the described framework to the evaluation of the TDH, showing a contrast in accordance with the prediction of TDH.