Introduction Non-literal language comprehension disorders are an integral part of the cognitive-communicative phenotype of individuals with acquired brain injury (ABI). Unfortunately, indirect requests have received relatively little attention among the various forms of non-literal language. However, unconventional indirect requests (or hints) are interesting speech acts to analyse. Indeed, several studies tend to show difficulties in understanding hints in ABI individuals (e.g., Champagne-Lavau & Joanette, 2009; Evans & Hux, 2011; Hatta, Hasegawa, & Wanner, 2004), although these results are not unanimous (McDonald, 2000; McDonald, Fisher, & Flanagan, 2016). Moreover, the fact that hints are coded in the mental lexicon under their literal meaning (Giora, 2002) suggests that an analysis of the context and additional inferential processes are necessary for their understanding. Some studies have thus focused on the cognitive processes involved in hints understanding - particularly the role of the theory of mind (ToM) and executive functions. However, again, the results are not consensual (Champagne-Lavau & Joanette, 2009; Muller et al., 2010). Several factors may explain these contradictions. First, the heterogeneity of ABI individuals and the nature of pragmatic errors are rarely taken into account in the analyses performed (Côté, Payer, Giroux, & Joanette, 2007). The results, therefore, do not make it possible to account for the diversity and fine nuances of the pragmatic disorders. Second, the measures are very heterogeneous, particularly in terms of cognitive load (Blake, 2017). To address these limitations, our study aimed to identify cognitive-pragmatic profiles in ABI individuals through a novel hint comprehension task manipulating cognitive mechanisms and processes of interest (i.e., context analysis, ToM, and executive function) (Cordonier, Champagne-Lavau, & Fossard, 2021). Methods We recruited 33 ABI participants from hospitals and clinical centres in the French-speaking part of Switzerland. They were between 20 and 65 years old, right-handed, and native French speakers. They had brain lesions consecutive to stroke in the right hemisphere or a moderate to severe traumatic brain injury occurring within at least six months. Thirty-three healthy controls (HC) were individually matched with the ABI participants for age and education level. We administered the ABI and HC participants neuropsychological tests assessing ToM (Faux-pas test, Mini-SEA, Bertoux, 2014) and executive functions (Hayling Test, Rouleau, 1998; verbal fluency task, Macoir et al., 2017 and Nasreddine et al., 2005; forward and backward digit spans (WAIS-III); reading span test, Desmette et al., 1995), as well as a hint comprehension task. This task included 36 written stories, which varied in terms of interpretation (literal vs hint), executive demand (low vs high), and contextual cues facilitating understanding of the hints (i.e., a hierarchical relationship between interlocutors; with vs without). In order to take into account performance heterogeneity, the k-means clustering method was performed to group ABI participants with similar pragmatic performance on the hint comprehension task. The silhouette and the gap statistic methods were used to determine the optimal number of clusters. Welch’s ANOVA and Welch’s t-tests for independent samples were then used to compare the performance of each ABI-subgroup to the group of HC on the pragmatic task and the neuropsychological tests. The p-values were adjusted with the False Discovery Rate method using the Benjamini - Hochberg procedure Results The silhouette and the gap statistic methods suggested that the optimal number of clusters is two. Therefore, these two subgroups were named a posteriori according to the problematic condition(s) evidenced in the statistical analyses: ABI-U (Unimpaired) and ABI-H (impaired in all the Hint conditions). One-way Welch ANOVAs and Post-hoc Welch’s t-tests revealed that the ABI-U subgroup performed similarly to HC participants in the hint comprehension task (see table 1). On the neuropsychological level, they had worse performance than the HC group on inhibition (Hayling inhibition: t(21.168) = -3.788, p <.05) and ToM tasks (Mini-SEA total: t(25.335) = -2.388, p <.05). On the other hand, the ABI-H subgroup performed worse than the HC participants and the ABI-U subgroup in all the hint conditions, regardless of the presence of cue or the executive demand (see table 1). On the neuropsychological measures, they demonstrated poorer performance than the HC group in short term memory (Digit Span forward: t(37.497) = -3.944, p <.01), working memory (Digit Span backward: t(36.715) = -4.474, p <.001; Reading Span Test: t(28.810) = -2.777, p <.05), flexibility (verbal fluency; (29.729) = -2.589, p <.05), inhibition (Hayling inhibition; t(24.714) = -6.375, p <.001; Hayling automatic: t(20.557) = -2.969, p <.05) and ToM measures (Mini-SEA total; t(17.929) = -4.289, p <.01; Mini-SEA faux-pas; t(17.923) = -3.943, p <.01). Compared to the ABI-U subgroup, they had poorer performance in short term memory (Digit Span forward: t(30.889) = 2.437, p <.05) and ToM measures (Mini-SEA total; t(24.244) = 2.523, p <.05) Discussion Our study has several implications. First, it confirms the heterogeneity of pragmatic disorders in the ABI population (Côté et al., 2007). Secondly, it clarifies the links between hints comprehension disorders and the mechanisms supposed to underlie this comprehension. It thus confirms the contextual insensitivity reported in some ABI participants (Champagne-Lavau, Cordonier, Bellmann, & Fossard, 2018), which leads them to infer the meaning of the utterance based on the literal meaning alone. It also confirms the association between pragmatic and theory of mind deficits (Bosco, Parola, Sacco, Zettin, & Angeleri, 2017; Champagne-Lavau & Joanette, 2009). However, it suggests that the severity or type of ToM deficit could lead to different pragmatic patterns (Abu-Akel, 2003). Finally, our results can be compared with studies that have used cluster analysis paradigms to strengthen the validity of the identified profiles (Champagne-Lavau & Joanette, 2009; Cordonier, Fossard, & Champagne-Lavau, 2020).