Neural organization of speech production: A lesion-based study of error patterns in connected speech

Brielle C Stark; Alexandra Basilakos; Gregory Hickok; Chris Rorden; Leonardo Bonilha; Julius Fridriksson

A handful of prior studies have evaluated neurogenic communication disorders, such as post-stroke aphasia, to compare the distribution and properties of word-level errors (paraphasias) during connected speech and single word retrieval. It seems intuitive that the underlying processes involved in producing a paraphasia on the two tasks should be similar, and for the distribution of such errors on one task to predict the distribution on the other. However, there are reports of patients with aphasia demonstrating superior word retrieval during confrontation naming compared to connected speech (Manning and Warrington, 1996; Schwartz and Hodgson, 2002; Wilshire and McCarthy, 2002) and vice versa (Zingeser and Berndt, 1990; Ingles and Mate-Kole, 1996; Pashek and Tompkins, 2002) . It is therefore becoming increasingly evident that connected speech requires dynamic changes in the linguistic system, resulting in differing distributions of linguistic components compared to naming. An early study evaluated word retrieval deficits in picture naming (the Boston Naming Test) and word-finding difficulty for nouns during connected speech production and found a ~58% overlap in variance between the two variables (Nicholas et al., 1989) . A later study evaluated naming performance and connected speech in 14 participants with mild and moderate aphasia, finding that there was superior word retrieval and overall more self-corrected errors during the connected speech task and that naming ability was not significantly correlated with connected speech paraphasias (Mayer and Murray, 2003) . Further evaluation by other groups suggested that naming ability may relate more to the ability to retrieve nouns, in particular, during everyday conversation (Herbert et al., 2008) . Further, a recent study in 98 participants with aphasia found that picture naming ability was not predictive of the proportion of paraphasias made during connected speech, finding that paraphasias produced in discourse in isolation were not well predicted by performance on confrontation naming tests (Fergadiotis and Wright, 2015) . Taken together, these studies encourage the direct comparison of the distribution of paraphasias between connected speech and naming for improved understanding of the dynamic lexical retrieval system employed during natural speech. Explanations regarding the reason for the discrepancy of paraphasia distributions between connected speech and single word retrieval have revolved around the differential context-dependent, non-linguistic demands required by connected speech as well as by linguistic factors such as the effects of lexical neighborhood and phonemic similarity (Foygel and Dell, 2000; Penn, 2000; Pashek and Tompkins, 2002) .