| Summary: | An unresolved question in research on executive control in language production is whether the processes responsible for inhibiting a dominant, prepotent response in order to comply with task goals is the same or different from control processes that bias intrinsic competition during lexical selection. Heretofore, these processes have been studied with different paradigms, such as the Stroop task and semantic blocking paradigm [1], respectively. The present study introduces a new paradigm to study both mechanisms as they impact word retrieval in neuropsychological and neurotypical populations. The task included several blocks of trials, where within a block two pictures were named repeatedly in random order. Two manipulated factors were: (1) relatedness of the pair of names, which bore either a semantic (duck/pig) or phonological relationship (ball/bag); or were unrelated (map/gun); (2) canonicity, where participants named each picture either with the canonical name (e.g., say “pig” for pig, “duck” for duck) or reversed the labels (e.g., say “duck” for pig, “pig” for duck). The names were closely matched for length, frequency, and other variables. Crossing the factors created six conditions (semantic-canonical, semantic-reverse, phonological-canonical, phonological-reverse, unrelated-canonical, unrelated-reverse), with each condition administered in one block of 16 trials (8 trials per picture). The task was administered to 12 participants with aphasia (PWA) with mild to severe naming impairment and 25 neurotypical controls. The dependent variables were naming latency (calculated for correct naming trials only) and naming accuracy, defined as a binary variable (correct versus error), which were analyzed with mixed linear and logistic regression analysis, respectively. For each dependent variable in each participant group, contrasting each related condition with the unrelated condition permitted measurement of three effects--a main effect of relatedness, canonicity, and their interaction. For example, focusing on the semantic and unrelated conditions, a main effect of relatedness (collapsing across canonicity) provided an index of difficulty in resolving intrinsic semantic competition. A main effect of canonicity (collapsing across relatedness) provided an index of difficulty in inhibiting a prepotent response. Presence of an interaction (unrelated/semantic X canonical/reverse) would suggest that the two control mechanisms are interdependent. Parallel effects were measured in a comparison of the phonological and unrelated condition as well. An a priori expectation was for poorer performance in the semantic condition versus unrelated, similar to semantic blocking effects in prior work [1]. This expectation was confirmed with a main effect of relatedness, with the PWA group showing lower accuracy (p<0.05) and controls showing marginally lower accuracy (p=0.059) and reliably longer latencies (p<0.05) in the semantic versus unrelated condition. Neither group showed a main effect of relatedness in either measure when comparing the phonological condition to unrelated, aligning with prior work showing that phonological relatedness does not consistently induce competitive interference [2]. All main effects of canonicity were reliable in both dependent measures for both groups (all ps<0.01), with poorer performance in the reverse versus canonical condition. Scant evidence of an interaction across the analyses suggests that the two control processes exert independent effects on performance.
|
|---|
