You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

 
Poster A27, Wednesday, November 8, 10:30 – 11:45 am, Harborview and Loch Raven Ballrooms

Lexical Selection and Multiword Speech in Acute Stroke

Tatiana Schnur1, Randi Martin2;1Baylor College of Medicine, 2Rice University

Selecting a word for production is generally understood to be competitive, where the word to be produced is selected by comparing its activation to other activated words that are not the intended target (Levelt, Roelofs, & Meyer, 1999). Although some deficits in multiword speech because of stroke are caused by damage to the meanings, phonological representations, and articulation of individual words, we hypothesize that beyond these factors, a deficit in the ability to reduce interference from competing words during selection (decreased selection capacity) causes poorly formed sentences. We predicted that decreased selection capacity should create increased semantically inappropriate word substitutions, omissions, and the omission or substitution of function words (decreased proportion of closed-class words) resulting in grammatically ill-formed sentences (decreased proportion of well-formed sentences). We also predicted decreased selection capacity should correlate with decreases in general measures of discourse productivity (numbers of words per minute and narrative words) under the hypothesis that failures in word selection result in general word-finding difficulties. Methods. As part of an ongoing project, we recruited 74 monolingual English speakers from multiple hospital stroke units in the Houston area who were diagnosed with acute left hemisphere ischemic cerebrovascular accidents. Subjects named 15 high-name agreement (low selection conflict; e.g., FORK) and 15 low-name agreement pictures (high selection conflict; e.g., COINS/MONEY; one item removed due to visual confusability and >50% error; Kan & Thompson-Schill, 2004). We measured selection capacity as the proportion correct naming difference between high- and low-name agreement conditions, where the more positive the NA difference score, the larger the assumed selection deficit. Subjects’ narrative production was scored per the Quantitative Production Analysis of the freely recalled Cinderella story (Rochon et al., 2000). Forty-six subjects confirmed with a unilateral left hemisphere stroke and no previous stroke or other neurological disorder completed name agreement and 36 subjects completed narrative story-telling within eight days after stroke (mean= 3 days, range= 1-8 days). Subjects named an additional set of unrelated pictures (n=47), where 12 subjects could only complete 17 items. Results. Consistent with predictions, as NA difference scores increased, accuracy of picture naming on the unrelated list decreased (r= -.40, p = .01) and the number of semantically related and description errors increased (144 errors; r= .53, p < .001). Subjects produced other error types too infrequently for statistical comparisons. As predicted, as NA difference scores increased, the proportion of well-formed sentences decreased during narrative story telling (r= -.38, p= .02). However, correlations with other QPA measures were not significant (p’s > .28). Conclusion. Selection deficits as measured in single word production contribute to semantically related and description picture naming errors and failures to produce grammatically well-formed sentences in spontaneous speech, but not to general discourse measures. This discrepancy may be because general discourse relies more heavily on the ability to generate message-level representations, while our selection task reflects processing at a lexical-grammatical encoding level of language production. We need to further explore the lack of relationship between selection deficit magnitude and production of closed-class words during narrative production.

Topic Area: Language Disorders

Back to Poster Schedule