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

We first explored the differences in paraphasia proportion made during connected speech and naming. In the CS+PNT subgroup, we found that semantically related paraphasias occurred more often on the naming task than on the connected speech task, while unrelated paraphasias occurred more often on the connected speech task. An explanation for this finding is that the classification of verbal paraphasias in connected speech may be inherently difficult, as it is challenging to definitively know the intended target of a word when there are many possible targets. However, as demonstrated by the excellent inter-rater reliability on error coding across raters, this is unlikely to be the driving component behind the difference in the proportion of semantically related paraphasias across tasks. Further persuasive evidence toward the argument that verbal paraphasias were correctly classified in the connected speech task comes from the clear differentiation of lesion maps for each paraphasia type in both tasks (e.g. verbal paraphasias associated with ventral stream damage; sound paraphasias with dorsal stream damage) and the similarity in lesion maps for each paraphasia type across tasks (e.g. semantically related errors in both naming and connected speech associating with damage to left anterior temporal cortex). There are reasonable explanations for the difference in verbal paraphasia proportion across the tasks. Connected speech provides additional context in the form of prior word and future word selection, and the semantically related paraphasias that do occur during connected speech likely suggest a problem with semantic selection rather than a downstream problem of lexical selection. Therefore, the naming task may require higher lexical-semantic load than the connected speech task, a supposition that is supported by semantic interference demonstrated during blocked cyclic picture naming experiments (Schnur et al., 2006; Oppenheim et al., 2010) . There is another possible explanation for the reduced proportion of semantically related paraphasias made during connected speech. During picture description, one can avoid talking about certain objects, whereas in picture naming, one is required to come up with a name for the picture. This may lead to a significantly smaller proportion of semantically related paraphasias in connected speech. Therefore, while there may be a reduced lexicalsemantic load in connected speech, there are other strategies, such as circumlocution, that contribute to the reduced proportion of semantically related paraphasias made during connected speech. The overall greater . CC-BY-NC-ND 4.0 International license is made available under a The copyright holder for this preprint (which was not peer-reviewed) is the author/funder. It . https://doi.org/10.1101/544841 doi: bioRxiv preprint proportion of unrelated paraphasias (always real words) and larger proportion of phonemic paraphasias and neologisms during connected speech also suggests an increase in lexical-phonological load, in keeping with the evidence proposing that several lexical items are selected simultaneously during connected speech (Kempen and Huijbers, 1983 ). An additional contributor to increased lexical-phonological load, aside from stress imposed by prior and future word choice and multiple word selection, is that connected speech is produced at a faster rate than the words produced during a picture naming task. Dell (1986) suggested that, when speech rate is