Frontiers in Psychology (Apr 2014)
Prediction during sentence comprehension in aphasia
Abstract
Much recent psycholinguistic work has focused on prediction in language comprehension (Altmann & Kamide, 1999; Federmeier, 2007; Levy, 2008). Unimpaired adults predict upcoming words and phrases based on material in the preceding context, like verbs (Altmann & Kamide, 1999) or constraining sentence contexts (Federmeier, 2007). Several models have tied rapid prediction to the language production system (Federmeier, 2007; Pickering & Garrod, 2013; Dell & Chang, 2014). Evidence for this link comes from that fact that older adults with lower verbal fluency show less predictive behavior (Federmeier, et al., 2010; DeLong, et al., 2012). Prediction in aphasic language comprehension has not been widely investigated, even though constraining sentence contexts are strongly facilitative for naming in aphasia (e.g., Love & Webb, 1977). Mack, et al. (2013) found in a visual-world task that people with aphasia (PWA) do not predict upcoming objects based on verbs (cf. Altmann & Kamide, 1999). This finding suggests that prediction may be reduced in aphasia. However, it is unclear whether reduced prediction was caused by language-production impairments: all the PWA in their study had non-fluent aphasia. The current study examined whether PWA show evidence of prediction based on constraining sentence contexts (e.g., Federmeier, 2007). Specifically, it tested whether they exhibited facilitation for highly predictable words in reading, using materials that have previously demonstrated strong predictability effects for unimpaired adults (Rayner, et al., 2004). In addition, it tested whether differences in language-production ability among PWA accounted for differences in predictive behavior (viz. Pickering & Garrod, 2013; Dell & Chang, 2014). Eight PWA read sentences adapted from Rayner, et al. (2004) in a self-paced reading task. The materials crossed word frequency with predictability: high- vs. low-frequency words (bottle/diaper) were preceded by contexts which made them either highly predictable or unpredictable (Before warming the milk/To prevent a mess, the babysitter took out/checked the …). The PWA varied widely in their verbal fluency, as measured by their verbal-fluency scores on the Comprehensive Aphasia Test (Swinburn, et al., 2004; verbal fluency T-score range: 43-71). Reading times at the critical word (bottle/diaper; see Figure 1) showed a large effect of predictability, with both high- and low-frequency words being read faster in constraining sentence contexts (F[1,7]=13.66, p.1). This pattern is different from findings for older and younger unimpaired adults, who show similar-sized effects for predictability and frequency (Rayner, et al., 2004; Ashby, et al., 2005; Rayner, et al., 2006). However, it is consistent with previous results for aphasia, where effects of word frequency are often missing (e.g., DeDe, 2012). There was no relationship between verbal fluency and the size of the predictability effect (r=.06, p>.4). These findings indicate that PWA can show very robust prediction in comprehension, at least for constraining sentence contexts (Federmeier, 2007). However, prediction appears to be dissociated from variation in verbal fluency/language-production impairment. This finding casts doubt on theories which ground prediction in comprehension in the language production system (Federmeier, 2007; Pickering & Garrod, 2013; Dell & Chang, 2014).
Keywords
