Naming practice effects and inconsistencies relate to treatment outcome in people with aphasia

Intra-individual variability in language performance has been proposed as a factor associated with treatment outcomes in chronic aphasia. However, the nature of linguistic variability and the degree to which it informs therapeutic success remains poorly understood. In this study, we sought to (1) assess person- and item-level factors associated with naming variability, practice effects, and item-based inconsistencies, (2) determine the relation between intra-individual naming variability, practice effects, and item-based inconsistencies at baseline and treatment outcome in aphasia, and (3) determine the treatment impact on the change in variability and consistency of naming responses.

Seventy-eight participants with chronic (>6 months post-stroke) aphasia after a unilateral left-hemisphere stroke completed the Philadelphia Naming Test (PNT) twice over two consecutive days prior (at baseline) to receiving six weeks of lexical processing treatment. For each participant, we calculated the absolute difference in correctly named items between the two PNTs at baseline (which we termed "naming variability") and the change from the first to second PNT at baseline (which we termed "naming practice effect"). Further, we classified participants' naming responses for each item on the two PNTs as correct-correct, one-correct, and incorrect-incorrect. One-correct responses reflected intra-individual naming inconsistencies for the same item across the two baseline PNTs (which we termed "naming inconsistencies"). We assessed the relationship between naming variability, practice effects, inconsistencies, and person-level factors (aphasia severity, aphasia type, apraxia of speech, stroke severity, age, education), item-level factors (word frequency, length, phonological neighborhood density), treatment response (the change in the rate of correct responses from before to after treatment).

The rate of correct naming responses on the two baseline PNTs did not statistically differ across the 78 participants; thus, there was no significant practice effect. However, variability and inconsistencies were common, with a difference of up to 17 % in correctly named items and up to 45 % of the same items named once correctly and once incorrectly between the two baseline PNTs. Naming variability was significantly related to aphasia type and severity, and inconsistencies were related to aphasia type, severity, the presence of apraxia of speech, and target word frequency. While naming variability was not associated with treatment outcome, practice effects and inconsistencies at baseline were significantly associated with treatment outcome and explained 36 % and 6 % of the variance in the change in the rate of correct responses from before to after treatment, respectively. Using 5-fold cross-validation, practice effects and inconsistencies had a coefficient of determination (R²) of 0.32 and 0, respectively, for predicted vs actual responses. After treatment, naming variability and inconsistencies significantly decreased, counterbalanced by an increase in consistently correctly named items.

We observed greater treated naming improvement in participants with stronger practice effects or more inconsistencies at baseline than in participants with smaller practice effects or more consistent naming at baseline, independent of aphasia severity. However, the generalizability to new data to predict treatment outcomes was weak for practice effects and poor for inconsistencies. After treatment, participants produced less variable and more consistent responses, indicating that treatment led to an overall strengthening of lexical-semantic retrieval processes.