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

 
Poster B29, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Similarity-based interference effects in reflexive binding: Empirical evidence from aphasia

Maria Varkanitsa1,2, David Caplan1;1Massachusetts General Hospital - Harvard Medical School, 2University College London

Existing models of sentence processing have argued that establishing dependencies relies on cue-based retrieval mechanisms that often lead to interference effects. Here we report data from a visual-world eye-tracking experiment investigating processing of antecedent-reflexive dependencies in stroke patients with aphasia (PWA). Four PWA (mean age = 56, mean education = 15 yrs) have participated to date. Stimuli included auditorily presented short stories introducing characters (“mother”, “bride”, “teacher”), critical sentences presented along with four pictures, and comprehension probes. Condition 1 included critical sentences and probes of the type “The bride with the spectacular heels covered herself with the veil./ Did the bride cover herself/the teacher?”, and condition 2 “The mother of the bride with the spectacular heels covered herself with the veil./ Did the mother of the bride cover herself?”. Critical regions for fixation analyses were the reflexive region and the final sentence region. A target-advantage score compared to the discourse distractor was calculated, by subtracting the proportion of fixations to the distractor element (“teacher”) from the proportion of fixations to the target (“bride” in condition 1, “mother” in condition 2). For condition 2 we also calculated a target-advantage score compared to the intervener, by subtracting the proportion of fixations to the intervener (“bride”) from the proportion of fixations to the target (“mother”). Correctly and incorrectly comprehended trials were examined separately. In condition 1, the PWA had an accuracy of 77.5% (range: 65%-100%). In correct trials, they exhibited a target preference in the reflexive and final region, whereas in incorrect trials they exhibited a distractor preference in the final region. In condition 2, their accuracy dropped to 47.5% (range: 10%-90%). In correct trials they exhibited a distractor and intervener preference in the reflexive and final region, whereas in incorrect trials they exhibited a target preference again in both regions. Fixation patterns in condition 1 indicate that failure in processing antecedent-reflexive dependencies in PWA may be due to interference from a distractor item established in the discourse (“teacher”) that is similar to the target (“bride”). Fixation patterns in condition 2 seem to suggest that the PWA exhibit an incoherent fixation pattern. However, given that critical sentences in the two conditions are only minimally different, we argue for an alternative interpretation; the PWA often misinterpreted the comprehension probes the same way they misinterpreted the sentences, that is, they erroneously retrieved the intervener or the distractor as the antecedent of the reflexive in the probes. Therefore, in trials with an intervener/distractor preference they gave a YES response, whereas in trials with a target preference they replied NO. In other words, their correct trials could actually be considered as incorrect and vice versa. In this case, their fixation pattern confirms the effect of similarity-based interference in processing antecedent-reflexive dependencies. This needs further exploration with simpler probes (“Did the mother cover herself?”). We conclude that our results are consistent with interference effects from elements in the sentence and context on on-line processing of reflexives in aphasia, and provide evidence for cue-based retrieval mechanisms.

Topic Area: Language Disorders

Back to Poster Schedule