A Computational Evaluation of Sentence Processing Deficits in Aphasia

Area of Science:

• Computational linguistics
• Neuroscience
• Psycholinguistics

Background:

• Individuals with agrammatic Broca's aphasia exhibit significant challenges in comprehending reversible non-canonical sentences.
• Existing theories attribute this deficit to either impaired syntactic representations (Trace Deletion Hypothesis) or processing deficits (e.g., slowed lexical activation, memory limitations).

Purpose of the Study:

• To computationally test competing theories of sentence processing deficits in agrammatic Broca's aphasia.
• Specifically, to evaluate the Trace Deletion Hypothesis (TDH) against processing accounts of slowed processing and intermittent deficiency.

Main Methods:

• Utilized a computational framework for sentence processing (ACT-R) to model different theoretical accounts.
• Operationalized slowed processing as slow procedural memory and intermittent deficiency as increased noise in procedural memory.
• Operationalized TDH as the absence of trace information in parse trees.
• Tested model predictions against offline (sentence-picture matching) and online (eye-tracking) data from a German study.

Main Results:

• The model incorporating both slowed processing and intermittent deficiency provided the best fit to the observed sentence processing difficulties in aphasia.
• Modeling individual differences indicated that patients exhibit varying degrees of both slowed processing and intermittent deficiency.

Conclusions:

• The findings support a dual-deficit model, suggesting that sentence processing impairments in agrammatic Broca's aphasia stem from a combination of slowed processing and intermittent deficiency.
• Individual variability in aphasia can be explained by differing levels of these two processing deficits.