The resolution of filler-gap dependencies in aphasia: Evidence from on-line anomaly detection

Michael Walsh Dickey & Cynthia K. Thompson
Northwestern University

dickey@ling.northwestern.edu

Agrammatic aphasics have problems comprehending sentences with movement like (1) [see example 1 below].  In a number of off-line tasks, such as picture matching, agrammatic Broca's aphasics perform at chance in their comprehension of object relatives (Grodzinsky, 2000, and references therein).  This paper presents evidence from an on-line anomaly detection task regarding agrammatic aphasics' real-time comprehension of sentences with movement.  The results suggest that aphasics' capacity for gap-filling is relatively preserved for non-reversible object relatives like (1).  Such a result is surprising if agrammatics are unable to represent or accurately resolve movement dependencies.

8 previously diagnosed agrammatic aphasics and a group of young normal controls listened to sentences like (2) [see below] and were asked to monitor them for anomalies.  They were instructed to press a button as soon as the sentence started to "sound strange." If agrammatic aphasics are unable to represent movement dependencies (Grodzinsky, 1990, 2000) or unable to correctly resolve them (Mauner et al., 1993), their detection of the anomaly in the object-relative condition (b) should be significantly less accurate than their detection of the same anomaly in the non-movement conjoined condition (d).  Further, if aphasics rely on extralinguistic heuristics to guide their comprehension of movement constructions (Grodzinsky, 2000), they should show little evidence of on-line anomaly detection even when they accurately reject (b).

Subjects listened to 24 sentences like (2) interspersed among 48 fillers.  Accuracy in detecting the anomalies was recorded, as were reaction times in rejecting them.  The aphasic subjects rejected both (b) and (d) at levels well above chance, 73% for (b) and 81% for (d).  Further, they showed only small differences from the normal controls in their rejection rates for both (b) and (d).  In addition, they showed evidence of being able to detect the anomalies on-line: while aphasics rejected (b) significantly more slowly than they rejected (d), the majority of rejections for (b) came before the end of the sentence.

These results suggest that for semantically non-reversible sentences at least, agrammatic aphasics retain significant on-line processing abilities.  This in turn suggests that aphasics either retain the capacity to represent and resolve filler-gap dependencies or are able to apply extralinguistic heuristics to comprehending such sentences very quickly.  Time permitting, implications of these results for the nature of aphasic comprehension deficits and their treatment will be discussed, as will parallel results from a group of fluent Wernicke's aphasics.

Examples