Assessing agrammatic speech: Insights from perceptual ratings, statistical metrics, and lesion-symptom mapping

The perceptual recognition of expressive agrammatism in aphasia is rapid and readily apparent by seasoned and novice researchers alike. However, in recent years much work on agrammatism has relied on statistical quantification of agrammatic speech through variables such as mean length of utterance and proportion of grammatical elements. This study aims to examine the insights of perceptual rating and statistical linguistic analysis of discourse in identifying and understanding agrammatism. Additionally, lesion-symptom mapping (LSM) is employed as a supportive tool to assess the effectiveness of these measures in capturing meaningful neurobiological correlates. 53 Participants with left-hemisphere ischemic stroke were part of retrospective analyses. An agrammatic/non-agrammatic perceptual classification based on audiovisual recordings of discourse samples was drawn from previously published analyses. Behavioral analyses showed significant correspondence between the perceptual groupings of agrammatic and non-agrammatic individuals and several linguistic variables, including mean length of utterance, use of past tense, and percent nouns. Despite a general correspondence between perceptual and statistical linguistic measures, only the perceptual approach revealed robust lesion correlates. LSM of perceptual classification revealed a strong lesion correlate primarily in the frontal lobe, spanning the middle frontal gyrus and the precentral and postcentral gyri, consistent with previous research, in both voxel and ROI analyses. In contrast, voxel-based analyses using individual linguistic variables and PCA-derived components did not reveal any significant clusters, and ROI-based analyses targeting Broca's area identified significant group differences only for past tense and present participle usage. These findings highlight the challenges using isolated, surface-level linguistic measures without identifying common underlying processing mechanisms that, when impaired, result in agrammatic speech, which may be more identifiable through perceptual experience.