Transcription-less analysis of five discourse tasks in Laurentian French persons with post-stroke aphasia: Adaptation and reliability.
L’essentiel
Discourse analysis is essential for aphasia assessment and treatment monitoring, but transcription-based methods are often too time-intensive for clinical use. Transcription-less approaches, including Main Concept Analysis (MCA), Thematic Units (TU), and coherence/reference (CoRe), offer practical alternatives, yet their psychometric properties remain underexplored. Guided by the Linguistic Underpinnings of Narrative in Aphasia framework, this study evaluated the reliability and construct validity of these measures in Laurentian French. Twenty-three French-speaking adults with chronic post-stroke aphasia completed five monologic discourse tasks across two sessions (mean interval = 11.7 days). Transcription-less measures were scored directly from audio recordings, and thirteen linguistic variables, including correct information units and moving average type-token ratio, were extracted from transcripts. Intra- and inter-rater reliabilities were assessed using intraclass correlation coefficients; test-retest reliability was examined using intraclass correlation coefficients, Spearman correlations, and Wilcoxon signed-rank tests. Construct validity was evaluated through correlations between transcription-less and linguistic (transcript-based) measures. Intra- and inter-rater reliability were good to excellent for most transcription-less measures. Test-retest reliability varied by task and measure, with CoRe showing consistently high stability across conditions. Six linguistic variables demonstrated excellent test-retest reliability. CoRe scores were strongly associated with MC scores across tasks, and transcription-less measures correlated with information content across all tasks. MCA procedures were also adapted for three tasks in Laurentian French. This study provides the first psychometric validation of transcription-less discourse measures in Laurentian French aphasia, supporting their use as reliable, feasible, and ecologically valid tools for clinical and research discourse assessment.
Synthèse détaillée
Résumé original
Discourse analysis is essential for aphasia assessment and treatment monitoring, but transcription-based methods are often too time-intensive for clinical use. Transcription-less approaches, including Main Concept Analysis (MCA), Thematic Units (TU), and coherence/reference (CoRe), offer practical alternatives, yet their psychometric properties remain underexplored. Guided by the Linguistic Underpinnings of Narrative in Aphasia framework, this study evaluated the reliability and construct validity of these measures in Laurentian French. Twenty-three French-speaking adults with chronic post-stroke aphasia completed five monologic discourse tasks across two sessions (mean interval = 11.7 days). Transcription-less measures were scored directly from audio recordings, and thirteen linguistic variables, including correct information units and moving average type-token ratio, were extracted from transcripts. Intra- and inter-rater reliabilities were assessed using intraclass correlation coefficients; test-retest reliability was examined using intraclass correlation coefficients, Spearman correlations, and Wilcoxon signed-rank tests. Construct validity was evaluated through correlations between transcription-less and linguistic (transcript-based) measures. Intra- and inter-rater reliability were good to excellent for most transcription-less measures. Test-retest reliability varied by task and measure, with CoRe showing consistently high stability across conditions. Six linguistic variables demonstrated excellent test-retest reliability. CoRe scores were strongly associated with MC scores across tasks, and transcription-less measures correlated with information content across all tasks. MCA procedures were also adapted for three tasks in Laurentian French. This study provides the first psychometric validation of transcription-less discourse measures in Laurentian French aphasia, supporting their use as reliable, feasible, and ecologically valid tools for clinical and research discourse assessment.