Beyond Accuracy: Inner Speech and Error Patterns Reveal Treatment-Based Learning in a Single-Subject Design Study

In our recent single-subject design study, we showed that a sense of "successful inner speech (IS)" at baseline (i.e., "I can say it in my head") predicted treatment-related improvements in naming accuracy in four of five individuals with aphasia. Here, we analyze error patterns from those data to determine whether error patterns also changed after treatment and whether self-reported IS predicted such changes.

Five adults with chronic poststroke aphasia completed 2 weeks of a cueing-based naming treatment targeting 40 personalized stimuli: 20 items reported prior to treatment as successful IS (sIS) and 20 items reported as unsuccessful IS (uIS). Before and after treatment, naming attempts were coded ordinally as follows: (0) no response, (1) neologism, (2) unrelated, (3) description, (4) semantic, (5) mixed, (6) phonological, and (7) correct. We used ordinal cumulative logit mixed-effects models to assess longitudinal changes in error types across all treatment items and in sIS versus uIS items. We evaluated individual-level changes by examining raw error type proportions across time.

Naming performance improved from baseline testing to immediate posttreatment, when accounting for error types alongside naming accuracy. Participants' posttreatment naming attempts had higher odds of being closer to the target for sIS items versus uIS items, after controlling for baseline scores. Finally, most individual participants' error patterns changed across time, including one participant who was a "nonresponder" in our prior accuracy-based analysis.

These preliminary findings suggest that error patterns are an important source of information about treatment-based learning in aphasia and converge with recent evidence that IS impacts naming treatment outcomes.