Objective:Social impairment is observed across psychotic-spectrum disorders (PSDs). Prior work demonstrates that neurocognition may account for social impairment in chronic PSDs. Concerning specific neurocognitive facets, there is support for a relationship between language/verbal performance and social outcomes in chronic PSDs. However, few studies have investigated this relationship in at-risk and early intervention samples, despite the clinical importance in these populations. The present study aimed to identify whether language is related to social functioning across the psychosis-spectrum, utilizing a sample comprised of individuals low in schizotypal personality traits, at-risk for psychosis (high in schizotypal traits), and those who recently experienced a first episode of psychosis (FEP). As an exploratory analysis and guided by findings from general studies of neurocognition and functional outcomes in the chronic PSD literature, we also investigated potential mediating mechanisms (i.e, negative traits; social cognition). We hypothesized that language is related to social functioning, and language is indirectly related to social functioning through negative traits and social cognition.Participants and Methods:We recruited 42 participants low in schizotypal traits and 44 high in schizotypal traits from undergraduate courses, and 15 FEP individuals were recruited from an early psychosis intervention center. On average, participants were 21.55 (4.39) years old and completed 14 (1.57) years of education. A majority were female (62%) and White (82%). Participants completed the Schizotypal Personality Questionnaire – Brief Revised Updated, which was used to categorize the schizotypy groups and sum negative traits; and measures of language/verbal ability (Similarities, Proverbs Test, semantic fluency, Digit Span), social cognition (Hinting Task, Affect Naming), and examiner-rated (Global functioning [GF]: Social) and performance-based social functioning (Social Skills Performance Assessment; SSPA). We also measured verbal processing speed and COVID-19 distress as covariates. Standardized scores were used for neurocognitive variables, and we used raw scores for most other variables. We utilized hierarchical linear regression models to examine whether specific language/verbal skills accounted for unique variance in examiner-rated and performance-based social functioning. For our exploratory analyses, we created averaged z-scores for language, social cognition, and social functioning and then, employed PROCESS Macro Model 4 to examine whether negative traits or social cognition were significant mediators in two separate mediation models.Results:Controlling for verbal processing speed and COVID-19 distress, language accounted for a significant portion of variance in SSPA performance, p = .008, ΔR2 = .12. Specifically, better Proverbs Test performance was uniquely associated with better SSPA performance, β = .33, p = .002. Controlling for study covariates, language was unrelated to GF: Social ratings, p = .31, ΔR2 = .038. In exploratory analyses, language was significantly indirectly related to social functioning through social cognition, β = .15, SE = .04, 95% CI [.04, .27], but not through negative traits, β = .08, SE = .06, 95% CI [-.001, .17].Conclusions:Our findings suggest that executively-mediated language tasks (e.g., Proverbs Test) and social cognition may be beneficial treatment targets for social impairment. Limitations include generalizability of the present findings, small FEP sample, and cross-sectional design. Future work should replicate these findings in longitudinal models.