Measuring proximal outcomes of sentence production therapy: probing treated and untreated constructions with a story completion test

Background

Many of the methods devised to assess aphasic sentence production involve spoken narratives (e.g. retelling a story, explaining a procedure) with high functional validity. However, when used to measure change following therapy, such unconstrained production tasks may fail to elicit the structures targeted by intervention. An alternative approach in targeting production of treated constructions is a story completion test (SCT), where the speaker completes a story stem, designed to elicit a specific sentence. In this study, we used a SCT to explore the outcomes of a new sentence therapy for aphasia (Unification Therapy Integrating LexIcon and Sentences – UTILISE), an intervention motivated by usage-based Construction Grammar. This approach allows creation of sets of frequency-matched treated and untreated sentences, permitting an exploration of generalisation, analogous to probing treated vs untreated pairs in lexical therapy studies.

Aims

To determine (1) whether speakers with aphasia produce more treated constructions and (2) more grammatically well-formed responses in a SCT following UTILISE therapy; and (3) whether therapy effects generalise to untreated sentences. In addition, we explored (4) whether baseline SCT performance was associated with behavioural profiling measures, and whether aphasia severity could predict therapy response on the SCT.

Methods & Procedures

Participants with chronic aphasia (N = 39) were recruited to a two-arm randomised control trial, with 31 completing the SCT at both baseline and outcome timepoints. Participants heard a story stem and attempted to produce a final sentence to complete the story, with stories probing treated and matched untreated sentences. UTILISE therapy involved two auditory tasks and one spoken sentence production task, all of which targeted 20 high-frequency constructions. Therapy was delivered across 12 sessions over a four-week period. Audio-recordings of SCT responses were rated by masked scorers on two scales: whether (a) the target (expected) construction and (b) a grammatically well-formed response were produced.

Outcomes & Results

Participants showed significant increases in both production of target constructions and grammatically well-formed responses. In both cases, improvement generalised to untreated sentences. A significant association between sentence comprehension and SCT expected structure scores was found, and regression analysis indicated that severity may be a predictor of SCT change.

Conclusions

Applying usage-based principles to a sentence-level intervention facilitated the creation of a SCT with treated and untreated sentences to explore generalisation effects. Results revealed change in sentence production following UTILISE therapy, with generalisation to matched constructions of similar structure, length and usage frequency.