Predictive processing in aphasia under memory interference conditions: A partial replication of Mack et al. (2013) using webcam-based eye-tracking

Purpose: The anticipation of upcoming linguistic information is a crucial component of sentence processing. However, such predictive processing has been critically understudied in post-stroke aphasia. Patients with aphasia (PWA) often present with working memory deficits, and given that predictive processes have been closely tied to memory pre-activation in past literature, PWA present a highly intriguing case for studying the memory underpinnings of prediction. This experiment was run online and recorded eye movements through participants’ own webcams to avoid small- n confounds. Method: A partial replication of Mack et al. (2013; 10.1016/j.jneuroling.2013.04.002) was designed. Participants (30 PWA, 32 age-matched controls) viewed arrays including one target picture (e.g., “cake”) and three distractors, and listened to sentences including semantically restrictive or unrestrictive verbs (e.g., “Tomorrow Susan will eat/see the ...”). Memory interference was manipulated by three-noun memory load sets presented prior to each trial, which either interfered semantically with the target word (e.g., icing, candle, birthday) or were semantically unrelated. Eye movements were recorded using WebGazer.js . Results: PWA exhibited smaller behavioral prediction effects and some predictive processing declines on eye movement data. Surprisingly, interfering memory words resulted in marginally higher accuracy and higher target gazes in PWA. Conclusion: Mack et al. (2013) was mostly replicated: PWA showed reductions in predictive processing efficacy compared to controls, although conditional differences were still robust. Memory manipulations may have resulted in semantic priming rather than semantic interference. The case is made for the widespread application of online research methods, and webcam-based eye-tracking in particular, with PWA and other patient groups.