Response to Seidenberg and Hoeffner

DISCUSSION NOTES Response to Seidenberg and Hoeffner Jeri J. JaegerRobert D. Van Valin Jr.Alan H. Lockwood SUNY Buffalo Seidenberg & Hoeffner (hereafter S&H) have presented a critique of our PET study of regular and irregular verb morphology in English1, and have faulted it on five main points. We will address each of these points briefly in this reply. (1) Brain regions activated. S&H state that 'Jaeger et al.'s data do not provide strong evidence that rule-governed forms and exceptions are processed in separate brain regions '. Our PET data provided clear evidence that different but overlapping brain areas were activated by our three past tense tasks, Regular, Irregular, and Nonce. S&H agree that during the regular and nonce tasks, our subjects were appending the regular suffix to each stem, so there seems to be no disagreement that our PET data do in fact show that this regular process activated different areas from those activated by the irregular task. S&H fault us for our interpretation of the patterns of activation found in each task, saying that our 'account turns on attributions about the functions associated with different brain regions that are highly questionable; they lack independent motivation and they are not internally consistent'. In fact, we supported our functional analysis with approximately 35 references to the neurolinguistic literature, documenting that our functional interpretation is consistent with findings from other studies. The two interpretations particularly singled out for criticism are our functional analysis of left dorsolateral prefrontal cortex and left superior frontal gyrus. They offer an alternative hypothesis on the function of DLPFC, which is that it maintains in working memory the information that 'the regular suffix was required' . This is an interesting suggestion (although supported by only one reference to the primate literature), and in fact it would fit very well with our hypothesis that during this task this area ofthe brain was activated to facilitate regular past-tense formation. As far as the left superior frontal gyrus is concerned, we documented that it is usually activated during novel linguistic tasks and not during practiced tasks; we suggested that this general function would be manifested differently in different tasks, depending on which aspects of the task were 'novel', which is why the details of performance would differ in the Irregular Past vs. Nonce Past tasks. Overall, we presented a unified theory as to what the areas of the brain activated in each task could be contributing to the performance of that task, and the S&H critique makes no attempt to address the entire pattern of activation. Finally, S&H point out that in other neurolinguistic studies which have also found dissociations between regular and irregular processes, each study has found a somewhat different pattern of association between brain areas and function. This suggests that much work remains to be done in order to determine the exact contributions of each brain area to the processes, but it also highlights a crucial fact: S&H report no neurolinguistic studies 1 Many linguists will be surprised to hear that 'most linguistic accounts of these phenomena have used the same principles to capture regular and irregular forms' (n. 1). Spencer (1991) provides a rich historical overview of the diverse ways in which irregularity has been treated in morphological theory during the past 60 years, documenting many theories in which regular and irregular morphology are treated as separate phenomena. S&H apparently disagree with Plunkett (1995) who, in his discussion ofconnectionist approaches to language acquisition, says that 'appeals to dual mechanisms thus proliferate through much of current linguistic theory' (1995:38). 123 124LANGUAGE, VOLUME 74, NUMBER 1 (1998) on this issue that found no differences between the processing of regular and irregular forms. In the Ullman et al. 1997b and Marslen-Wilson & Tyler 1997 studies of brain damage cited by S&H, damage to different cortical areas caused damage differentially to regular vs. irregular inflectional systems. Similarly, Ullman et al. (1997a) report on an fMRI study in which subjects produce regular and irregular past tense forms from a mixed stimulus list, and find a different but overlapping pattern of brain activation, with the dissociation being most robust in left prefrontal region vs. left temporal region. Finally, Penke et al. (1997) performed an ERP study of regular and irregular German...