As indicated by the preceding description of conduction aphasia, impaired repetition is often considered of prime importance for identifying this syndrome. However, it is not clear that this behavior should be the focus when rendering a diagnosis of conduction aphasia. Various chapters in this book argue that the frequent production of phonemic paraphasias is the key (i.e., phonological distortions of identifiable words that are free from phonetic error). That is, the production of phonemic paraphasias in essentially all speech contexts (e.g., spontaneous speech, repetition, oral reading, picture naming) is viewed by many as the most prominent "defining symptom" of conduction aphasia (i.e., necessary for diagnosis; see Introduction). Moreover, characteristics of these paraphasias help distinguish this aphasia from others (Kohn).

Confusion about the role of impaired repetition versus paraphasic speech in conduction aphasia existed from the inception of this syndrome. More recently, work by Shallice and Warrington (1977) presented a possible solution to this issue by distinguishing "repetition conduction aphasia" and "reproduction conduction aphasia." They argued that a deficit in auditory-verbal short-term memory was responsible for impaired repetition in some conduction aphasics and that an independent deficit in speech programming was responsible for the paraphasic output in other conduction aphasics. However, why should aphasics with such divergent underlying deficits be considered members of the same general diagnostic category? Because of the restricted effect of the memory deficit on speech (i.e., only impairing repetition), it is reasonable to exclude patients with isolated repetition deficits from the category of conduction aphasia. Then, in Shallice and Warrington's terminology, we would consider only reproduction conduction aphasics as true conduction aphasics.

This latter decision does not rule out the presence of impaired repetition in conduction aphasia, but merely specifies that output based on other input modalities must also be impaired in conduction aphasia. Moreover, one should not expect repetition to be the most impaired modality. In contrast to picture naming and oral reading, repetition involves input that is already in a phonological form, so that conversion into an articulable form (i.e., from acousticphonetic to articulatory-phonetic information) requires fewer transformations (Kohn & Smith, 1991; Kohn, Smith, & Alexander). This notion is supported by cases of conduction aphasia in which noun production during picture naming, oral reading, and repetition has been systematically compared, and repetition has been least impaired (Kohn, 1989; Kohn & Smith, 1991). . .