ABSTRACT
Two decades ago, the general assumption was that aphasic subjects from the classical diagnostic categories (Broca’s aphasics, Wernicke’s aphasics, conduction aphasics, and so on) would have similar deficits in the processes of word production. As we have seen, such assumptions are clearly not tenable. For instance, patients with deficits at any level from lexical phonological representations to the assembly of an articulatory programme could all legitimately be described as reproduction conduction aphasics (cf., Kohn, 1992; Shallice & Warrington, 1977), and yet they might have very different kinds of impairments within these processes. In recent years, the performance of individual patients has been much more carefully characterised, using experimental manipulation of the factors affecting their accuracy, as well as analyses of the properties of their errors. However, both of these sources of evidence are, at least partly, ambiguous. Variables can have their effects at many different levels; for instance, as we have seen, decreasing accuracy with increasing stimulus length-a characteristic feature of reproduction conduction aphasia-is a natural prediction of every postlexical production impairment. Speech errors from both normal subjects and patients are similarly overdetermined; as Cutler, Howard, and Patterson (1989) note, many “errors are ‘imperfect’ by virtue of having more than one possible source” ( p. 70).
