["Journal of Personality, EarlyView. ", "\nABSTRACT\n\nIntroduction\nSelf‐presentation behaviors are associated with self‐esteem. However, it remains unclear how verbal behaviors during self‐presentation are associated with self‐esteem. This uncertainty stems from the lack of fine‐grained linguistic analysis. In addition, existing findings on the relationship between verbal behaviors and self‐esteem are inconsistent, possibly due to variations in the content of self‐presentation. The present research aims to address these limitations.\n\n\nMethod\nIn Study 1, 211 participants (178 female) completed two self‐presentation tasks based on the social‐cognitive dimensions of agency and communion. Twelve linguistic features across lexical, syntactic, and semantic levels were extracted using computational linguistic techniques and analyzed via ridge regression. Study 2 recruited an independent sample of 63 participants (50 female) to replicate Study 1's findings using machine learning methods.\n\n\nResults\nStudy 1 showed that linguistic features (e.g., adverbials, positive and negative emotion sentences) were consistently associated with self‐esteem across both tasks. Specifically, linguistic features in agency‐based self‐presentation predicted self‐esteem more effectively (R2 = 0.106) than those in the communion‐based task (R2 = 0.042). Study 2 suggested stronger generalizability in the agency task.\n\n\nConclusions\nVerbal behaviors are associated with self‐esteem and are moderated by the content of self‐presentation. Agency (relative to communion) may function as a more expressive channel for self‐esteem.\n\n"]