Conversationally Attuned: Links Between Interpersonal Linguistic Synchrony and Social Functioning Across the Early-Psychosis Spectrum Authors: Kaley Angers, Melissa FV Kilicoglu, Josephine Luck, Bee Zaccaria, Kalif E Vaughn, Aubrey M Moe Published: 2026 ABSTRACT AbstractObjective:Social dysfunction and language abnormalities are core features of psychotic disorders that interfere with communication. Language studies in psychosis often utilize paradigms involving conversational exchange yet analyze only participant speech rather than understanding the broader participant-examiner discourse. Notably, humans naturally align their language to be more similar to their conversation partners during interaction—a process sometimes called language style matching (LSM). As individuals with psychosis face significant interpersonal challenges, understanding LSM in the context of social interaction is an important and unaddressed area of research.Methods:A total of 101 early psychosis-spectrum participants (early psychosis [EP;n = 24], high schizotypy [HS;n = 44], unaffected controls [UC;n = 33]) completed an audio-recorded social interaction task and an assessment of real-world social functioning. Transcribed speech from participant-examiner dyads underwent LSM analyses to assess interpersonal linguistic synchrony. Correlation and hierarchical regression analyses examined associations among LSM, performance on the social interaction task, and social functioning.Results:LSM was positively correlated with real-world social functioning and performance on the social task in EP and real-world social functioning in UC, but not HS. LSM further accounted for additional variance in social functioning, over-and-above performance on the social interaction task and relevant clinical and demographic variables (diagnostic status, age, sex).Conclusions:These results provide preliminary evidence that LSM during interpersonal interaction is linked to social dysfunction. Critically, LSM may measure features of interaction and social outcomes that are not captured by standard assessments. Collectively, conversational exchange is a promising area for future work assessing social dysfunction in emerging adults across the early psychosis-spectrum. Individuals with psychosis often experience broad impairments in social functioning that impact their ability to effectively interact with other people across contexts and in activities of daily living (i.e., difficulties in interpersonal relationships, maintaining employment, and diminished self-care; Blanchard et al., 1998; Mueser et al., 1991). Social difficulties are present early in the course of psychosis (i.e., early psychosis [EP]; Frawley et al., 2023), with some evidence suggesting social dysfunction is present even before the initial onset of illness (Carrión et al., 2021; Niendam et al., 2007). In fact, relative to unaffected controls, social dysfunction is also observed in schizotypy (e.g., Cohen & Davis, 2009; Fonseca-Pedrero et al., 2010; Gayer-Anderson & Morgan, 2013; Gooding et al., 2014; Wang et al., 2018), which refers to a grouping of personality traits that mirror positive, negative, and disorganized symptoms of schizophrenia (Lenzenweger, 2006), and when present in larger quantities, place an individual at-risk for developing psychosis (Barrantes-Vidal et al., 2015; Cohen et al., 2015; Debbané et al., 2015). Relative to individuals with a psychotic disorder, schizotypal traits are often present at a much lower threshold and are less interfering with respect to everyday functioning (Cohen et al., 2015). At the same time, individuals high in schizotypal traits share biological and psychosocial overlap with psychotic disorders (Barrantes-Vidal et al., 2015) and schizotypy samples are often comparable in age to typical onset for EP (Debanné et al., 2015). Notably, EP typically emerges during the late teenage years and into the 20s (Häfner et al., 2003)—years of life that intersect with a developmental phase recognized as emerging adulthood. Emerging adulthood is characterized by marked exploration of self-identity, relatedness to others, and navigation of uncertainty. The onset of psychosis and its associated social difficulties during a period of life already characterized by normative instability and change may have a particularly relevant and reverberating impact on critical developmental milestones (Moe & Breitborde, 2019), including problems with developing and maintaining interpersonal relationships. Critically, social support positively influences the early course of psychosis (e.g., lower rates of re-hospitalization and reduced psychotic symptoms; Norman et al., 2005). Lower levels of social support are consistently associated with poorer functional outcome in EP (Erickson et al., 1989; Gayer-Anderson & Morgan, 2013; Lim et al., 2018; Sündermann et al., 2014). Thus, understanding the mechanisms that contribute to interpersonal dysfunction among emerging adults with EP may provide critical insight on how to more effectively address social impairments early in the course of illness. Alterations of language and communication are also common features of psychotic disorders (Andreasen, 1979) that may have an especially important role in interpersonal dysfunction. Language disturbance can manifest in the speech of people with psychosis in various ways (e.g., tangentiality, derailment, poverty of content; Andreasen & Grove, 1986), leading to disruption in effective communication with others (Docherty et al., 1996) that is associated with poor social functioning (Kuperberg, 2010; Pienkos & Sass, 2016). Interestingly, some language abnormalities are also observed among individuals with elevated schizotypal traits (Minor & Cohen, 2010; Minor et al., 2018), further illustrating its relationship to the psychosis-spectrum. Recent application of computational tools has revealed more fine-grained language abnormalities (Ehlen et al., 2023), including alterations in self-experience, semantic coherence, pronoun use, linguistic cohesion, idea density, and analytical thinking in spoken language produced by individuals with psychosis (Chan et al., 2023; Lundin et al., 2023; Moe et al., 2016; Silva et al., 2021; Tang et al., 2021; Watson et al., 2012). Notably, existing studies of language abnormalities in psychosis have almost exclusively focused on speech generated by individuals with psychosis without consideration of the broader social context. More specifically, language paradigms typically elicit speech from individuals with psychosis from structured interviews, spontaneous life-story narratives, or picture description tasks (Corcoran et al., 2020) that are analyzed as monologues as opposed to considering speech as a conversational exchange. Thus, these approaches may neglect to consider how linguistic abnormalities emerge within more naturalistic conversations—and, in particular, how alignment of language in response to the language from others may be altered among people with psychosis. A rich body of literature supports the idea that humans exhibit a broad set of signs and behaviors intended to coordinate actions, emotions, thoughts, and even physiological processes when interacting with other humans (i.e., interpersonal synchrony; Ackerman & Bargh, 2013; Mayo & Gordon, 2020; Mogan et al., 2017). Interpersonal synchrony is generally considered to be a spontaneous, unconscious process that may have evolved as a mechanism for facilitating social connectedness (Launay et al., 2016). While investigations of interpersonal synchrony among dyads within the psychopathology literature have spanned across clinical populations and experimental methodologies (Bolis et al., 2023; Dean et al., 2021; Hendrix et al., 2018; Kupper et al., 2015), studies of interpersonal linguistic synchrony are less common. One promising approach for addressing this gap is language style matching (LSM). LSM is one metric of interpersonal linguistic synchrony that measures the subtle similarities of language use across two or more people throughout a conversational exchange. More specifically, LSM assesses the extent to which the function words (e.g., pronouns, articles, adverbs, etc.) from the language produced by one person aligns with the use of those same function words in the language of their conversational partner. Previous research has shown that LSM is associated with social cohesiveness among small groups of people (Gonzales et al., 2010), level of engagement during interactions (Niederhoffer & Pennebaker, 2002), and dyadic relationship stability over time (Ireland et al., 2011). Consistent with the interpersonal synchrony theory that posits adjusting our behavior to align with others fosters social bonding (Launay et al., 2016), higher levels of LSM are generally associated with better social outcomes. LSM is of particular relevance for understanding interpersonal dysfunction in psychosis as it extends the measurement of social processes as they unfold among two people within a more naturalistic context (i.e., a conversational exchange). Despite increased application of natural language processing and other computational tools to the speech of people with psychosis, LSM remains relatively unexplored. To the best of our knowledge, Abel and colleagues (Abel et al., 2025) published the only proof-of-concept study examining LSM across the psychosis-spectrum (i.e., individuals with psychosis, schizotypy, and unaffected controls), with findings that individuals with long-standing schizophrenia showed diminished LSM relative to controls as assessed in daily life interactions collected via experience sampling methodology (ESM). Importantly, the use of ESM to capture daily life interactions in this study provided critical insight into how LSM may unfold with repeated assessment of interpersonal interaction in the real-world setting. At the same time, a sizeable number of participants with schizophrenia (42%) did not generate interaction samples that met minimum word length thresholds and were thus excluded from the analyses. Though the use of ESM has powerful strengths in ecological validity, a more standardized approach to generating conversational speech across psychosis-spectrum participants may permit higher rates of data retention and allow for comparison of LSM to embedded metrics of social performance quality or effectiveness of the conversations. Additionally, as Abel and colleagues included only individuals with longstanding psychosis, LSM among individuals with EP remains unknown. Abel et al. (2025) did not find any difference in LSM within their high and low schizotypy groups; however, the depth of the conversational interaction predicted LSM across participant groups, which does support LSM as an important variable of study across varying levels of psychotic symptoms. Finally, the contribution of LSM to broader measures of real-world social functioning has not yet been investigated. As pointed out in this paper (Abel et al., 2025), the extent to which LSM is associated with validated behavioral measures of social processes—and the potential added value of LSM in understanding functional outcomes over-and-above these conventional measures—is an important next step in this area of research. The overarching goal of the present study was to delineate the impact of conversationally-derived language style matching (LSM) on social functioning among emerging adults across the early psychosis-spectrum. Given our aim of delineating language factors impacting social dysfunction within the context of interpersonal interaction, we include language generated by both the study participant and the trained researcher generated during a standardized, validated behavioral measure of social performance and competency (i.e., the Social Skills Performance Assessment (SSPA; Patterson et al., 2001). The use of an emerging adult sample spanning the early psychosis-spectrum (i.e., low and high schizotypy and EP) permits an examination of how conversational language is associated with social functioning during a phase of life characterized by key developmental milestones related to forming and maintaining interpersonal relationships. Finally, examining for the first time whether LSM contributes to social functioning beyond other conventional behavioral assessments provides a critical opportunity to clarify the relevance of conversational processes to real-world outcomes. We anticipated that i) language style matching would be positively correlated with both laboratory-based social competency and real-world social functioning; and ii) language style matching would account for unique variance in real-world social functioning, over-and-above laboratory-based social competency. METHOD Participants Study participants came from two separate larger study protocols examining neuro- and social cognition in the early psychosis-spectrum. Given our interest in understanding conversational and social processes during emerging adulthood, we included data from participants between the ages of 18 and 30 in the present analyses. High and Low Schizotypy Groups Undergraduates (n = 1,656) enrolled in introductory psychology courses were screened using the Schizotypal Personality Questionnaire—Brief Revised Updated SPQ-BRU; Davidson et al. (2016) (SPQ-BRU). Embedded validity items were included in the screening version (e.g., “Select strongly agree for this item.”); 417 individuals were excluded for partial surveys and 48 for failed validity checks. Individuals who completed all survey items, passed the validity checks, and were at least 18 years of age were invited to participate. Similar to prior studies (Angers et al., 2021, 2023) we defined schizotypy groups by SPQ-BRU score, where those who scored at or above the 70th percentile of scores were considered high schizotypy (i.e., the schizotypy group), and those at or below the 30th percentile of scores were considered low schizotypy (i.e., the unaffected control group; UC). All participants who accepted the study invitation completed the SPQ-BRU a second time during their study visit to confirm their participant group (n = 121). Individuals who did not continue to meet the SPQ-BRU parameters were excluded (n = 15), along with those who scored lower than 70 on a measure of estimated premorbid intelligence (n = 4), reported history of a neurologic disorder (n = 1), were older than age 30 (n = 7), or whose data on the social competency task could not be transcribed due to audio recording quality (n = 3). This resulted in a final sample of 44 in the schizotypy group and 33 unaffected controls. Early Psychosis Group EP individuals were recruited from the Early Psychosis Intervention Center (EPI-CENTER) at The Ohio State University, a coordinated specialty care service treating youth and young adults with early psychosis that is located in the same geographic region. Eligibility criteria for participation in the study included: i) meeting criteria for a primary psychosis-spectrum disorder as confirmed via the Structured Clinical Interview for DSM-5; ii) age between 18 and 30; iii) being within the first 5 years of psychosis onset; and iv) estimated premorbid intelligence of >70. Individuals who had previously consented to be contacted for research studies were invited to participate through recruitment phone calls or were provided information or a flyer by their clinicians during clinical encounters. The final sample consisted of 24 EP individuals. All study procedures were IRB approved. All participants provided informed consent after review of the study procedure and prior to completing any study activities, then completed self-reported demographics, the SSPA, and the GF-Social. All study participants received IRB-approved compensation commensurate with time and travel. Measures Schizotypal Traits The SPQ-BRU (Davidson et al., 2016) is a 32-item self-report measure of schizotypal traits that we utilized to categorize our schizotypy groups. Responses are made a 5-point Likert scale, and items are summed to create a total score. Higher scores indicate more schizotypal traits. Social Competency We assessed social competency via the Social Skills Performance Assessment (SSPA; Patterson et al., 2001). The SSPA is a brief performance-based, clinician-rated behavioral assessment of social skills that requires participants to take part in two conversational role plays with the examiner. The first role play involves meeting a new neighbor and requires the participant to initiate and maintain a conversation with a new person (i.e., rapport building). The second requires participants to negotiate a repair with their landlord (i.e., social problem-solving). In its traditional administration, participant performance is rated on multiple elements of effective social performance (i.e., social interest, fluency, clarity, focus, affect, grooming, overall conversation, social appropriateness, negotiation skills, and persistence) using a 5-point Likert scale for each domain. Higher total scores indicate better social performance. In the present study, we used the average performance score across both study role-plays. All SSPA raters underwent training and achieved adequate inter-rater reliability (ICC [absolute agreement] ≥75%). Each SSPA administration was recorded and later transcribed for linguistic analysis. Transcription was performed by trained research assistants under the supervision of the last author (AMM). Each completed transcript was double-checked by a second trained research assistant to verify accuracy prior to data analysis. Real-World Social Functioning The Global Functioning—Social (GF-Social; Carrión et al., 2019) is a semi-structured interview that assesses the quality of various social relationships over the past month. Overall scores can range from 1 (extreme social isolation) to 10 (superior social/interpersonal functioning). All study raters underwent training and met interrater reliability criteria (ICC [absolute agreement] ≥0.75) prior to administering or scoring any GF-Social interviews. Language Style Matching Interpersonal linguistic synchrony was assessed using the Language Style Matching (LSM) metric from the Linguistic Inquiry and Word Count (LIWC-22; Boyd et al., 2022) program. LIWC is a research software program that analyzes text files to generate various indices of word-use. LIWC has been most widely used to assess the frequency of word-use falling into a-priori dictionary categories that were developed for and exist within the LIWC program. For example, LIWC generates percentages of total words falling into selected categories, thus providing a metric that inherently accounts for the overall amount of text or speech that is being analyzed. LIWC has been applied to a variety of types of texts, including both written documents and transcriptions of speech, including speech that is generated in the context of interpersonal interaction. In the present study, we used LIWC to calculate an index score of language style matching (LSM). The LSM metric assesses how similarly two or more people use nine different types of function words (i.e., personal pronouns, impersonal pronouns, articles, adverbs, auxiliary verbs, conjunctions, pre-positions, negations, and quantifiers) in their writing or speech (Boyd et al., 2022). LSM utilizes function words as an indicator of matching, as these words are frequently used, context independent, and likely not consciously produced (Chung & Pennebaker, 2011). This approach minimizes the possibility that any observed dynamics are attributable to any particular topic within the conversation (Gonzales et al., 2010). To calculate LSM for each dyad, all transcribed SSPA samples were prepared in accordance with LIWC pre-processing guidelines (Boyd et al., 2022) and segmented by speaker. Prepared transcriptions were then analyzed using the LSM function of LIWC, which calculates matching using the following formula (Niederhoffer & Pennebaker, 2002): In this formula, fxword1 is the percentage of that function word used by the first person in the conversation, and fxword2 is the percentage used by the second person in the conversation. The denominator includes 0.0001 to prevent dividing by zero, should a specific type of function word not be used by either conversation participant (in which case the LSM score will be set to zero). LSM is calculated for each of the 9 types of function words, and then averaged to compute the overall LSM score. As we were interested in similarities of language between the study participant and the examiner within dyads, we utilized the pairwise comparison to calculate LSM scores. STATISTICAL ANALYSES All statistical analyses were conducted using IBM SPSS (29.0.1.0; IBM, 2020). As we conceptualize linguistic synchrony to be a continuous variable, we performed our primary analyses across the entire psychosis-spectrum sample to increase the likelihood of examining LSM across a wider range of interpersonal and social functioning. Bivariate correlations were computed among the social competency, social functioning, and language style matching variables to provide an overall understanding of associations. To examine whether LSM explained unique variance in real-world social functioning, over-and-above lab-based social competency, we utilized hierarchical regression analyses. In the first hierarchical regression model, participant group status (unaffected control = 0, schizotypy = 1, and early psychosis = 2) was entered at step 1 to account for the impact of group status on the dependent variable (i.e., real-world social functioning scores). Thus, any additional variance explained by LSM would not simply be attributed to group differences. We first examined whether LSM explained additional variance in social functioning over-and-above social competency. Therefore, social competency (SSPA) scores were entered at step 2 to examine the unique variance explained above-and-beyond group status. Lastly, LSM was entered at step 3 to examine the unique variance accounted for by this index over-and-above group status and SSPA scores. Finally, we conducted follow-up sensitivity analyses that also included key covariates (age and sex) in step 1 of the hierarchical regression model to assess whether any significant results remained after accounting for these variables in the analyses. RESULTS Demographic and Descriptive Statistics. Participant demographics appear in Table 1. Group comparisons revealed significant differences on age and sex. No significant group differences were observed for race or ethnicity. Descriptives and group comparisons for the language and social measures appear in Table 2. No significant group differences were observed for LSM or participant word count during the SSPA. Significant group differences were observed for both performance on the SSPA and GF-Social ratings. Post-hoc Tukey HSD tests showed that control participants had significantly higher SSPA scores than the schizotypy (p < .001) and EP (p < .001) groups, and similarly that control participants had higher GF-Social scores relative to both schizotypy (p < .001) and EP (p < .001); schizotypy and EP did not differ. TABLE 1.Demographics by GroupFEP (n = 24)Schizotypy (n = 44)UC (n = 33)StatisticsDemographics M SD M SD M SD Age 24.254.0720.452.1420.301.82F(2,98) = 19.55, p < .001*** Education 14.452.5414.301.4913.921.34F(2,98) = 0.723, p = .488 n % n % n % Sex χ2 = 13.15, df = 2, p = .001** Female833%3477%2267% Male1667%1023%1133% Race χ2 = 10.93, df = 8, p = .205 Asian14%00%00% Black313%49%13% White1875%3477%3091% Native American00%00%00% Multiracial00%511%13% Other28%12%13% Ethnicity χ2 = 2.31, df = 2, p = .315 Hispanic14%37%03% Non-Hispanic2396%4193%3397%*p < .05**p < .01***p < .001Note. Percentages listed by group represent the proportion of the group (i.e., first-episode psychosis [FEP], high schizotypy [Schizotypy], and Non-clinical control [NCC] group), not the total sample. Demographics by Group p < .05 p < .01 p < .001 Note. Percentages listed by group represent the proportion of the group (i.e., first-episode psychosis [FEP], high schizotypy [Schizotypy], and Non-clinical control [NCC] group), not the total sample. TABLE 2.Descriptive Data for Language Style Matching and Social MeasuresFEP (n = 24)Schizotypy (n = 44)UC (n = 33)Statistics M SD M SD M SD LSM.874.045.858.041.875.037F(2,98) = 2.352, p = .101Word Count566.17185.46506.98127.06581.36121.44F(2,98) = 2.954, p = .057SSPA4.34.4314.56.3244.91.134F(2,98) = 21.089, p < .001**GF-Social7.041.4617.11.9708.55.506F(2,98) = 20.934, p < .001***p < .05**p < .01***p < .001Note. LSM = Language Style Matching; SSPA = Social Skills Performance Assessment; GF = Global Functioning Scale Descriptive Data for Language Style Matching and Social Measures p < .05 p < .01 p < .001 Note. LSM = Language Style Matching; SSPA = Social Skills Performance Assessment; GF = Global Functioning Scale We used bivariate correlations to examine associations between LSM, SSPA, and GF-Social across groups. Higher levels of LSM were associated with greater real-world social functioning (r = .251, p = .012). As expected, social competency was also positively correlated with real-world social functioning (r = .412, p < .001). LSM and social competency scores were not significantly correlated (r = .151, p = .135). However, follow-up correlations showed differential associations between these variables at the group level (Table S1). More specifically, in the EP group, LSM was positively correlated with both SSPA and GF-Social scores. In the control group, LSM was positively correlated with SSPA scores. There were no significant correlations between these variables in the schizotypy group. Contribution of LSM to Social Functioning, Over and Above Social Competency Hierarchical regression analyses were performed to test the unique contribution of LSM to real-world social functioning, over- and-above social competency (Table 3). Notably, social competency scores did not explain significant variance in real-world social functioning after accounting for participant group. However, after accounting for participant group and social competency scores, LSM explained additional variance in real-world social functioning (ΔR2 = 5.9%, p = .006). Notably, group differences were observed on age and sex (Table 1). In a follow-up sensitivity analysis with age and sex as covariates in the first step of the hierarchical regression model, we observed that LSM still accounted for unique statistical variation in real-world social functioning (Table 4; ΔR2 = 5.9%, p = .007). TABLE 3.Hierarchical Linear Regression of Social Competency and Language Style Matching as Predictors of Real-World Social FunctioningVariable StatisticsModel StatisticsOutcomePredictorβ t P r 2 Δr2ΔF P GF-Social Step 1 Group−.46−5.10<.001***.20.2026.01<.001*** Step 2 Group−.35−3.01<.001*** SSPA.181.56.122.21.022.43.122 Full Model Group−.41−3.83<.001*** SSPA.131.12.244 LSM.232.70.008**.26.067.91.006***p < .05**p < .01***p < .001Note. LSM = Language Style Matching. SSPA = Social Skills Performance Assessment (average score); GF = Global Functioning Scale Hierarchical Linear Regression of Social Competency and Language Style Matching as Predictors of Real-World Social Functioning p < .05 p < .01 p < .001 Note. LSM = Language Style Matching. SSPA = Social Skills Performance Assessment (average score); GF = Global Functioning Scale TABLE 4.Hierarchical Linear Regression of Social Competency and Language Matching Style as Predictors of Real-World Social Functioning (Accounting for Age and Sex)Variable StatisticsModel StatisticsOutcomePredictorβ t P r 2 Δr2ΔF P GF-Social Step 1 Group−.53−5.81<.001*** Age.040.47.645 Sex.101.16.247.21.2410.10<.001*** Step 2 Group−.41−3.67<.001*** Age.050.52.604 Sex.091.02.313 SSPA.171.50.136.22.021.86.175 Step 3 Group−.45−4.12<.001*** Age.040.43.672 Sex.131.50.138 SSPA.110.99.323 LSM.252.89.005**.27.067.73.007***p < .05**p < .01***p < .001Note. LSM = Language Style Matching; SSPA = Social Skills Performance Assessment; GF = Global Functioning Scale Hierarchical Linear Regression of Social Competency and Language Matching Style as Predictors of Real-World Social Functioning (Accounting for Age and Sex) p < .05 p < .01 p < .001 Note. LSM = Language Style Matching; SSPA = Social Skills Performance Assessment; GF = Global Functioning Scale To better understand the specific relevance of LSM to social functioning in EP, we conducted an additional exploratory follow-up hierarchical regression model including only those participants. We found that LSM also accounted for unique variance in real-world social functioning (ΔR2 = 17.3%, p = .045), over-and-above social competency scores when age and sex were also included in the model (full results appear in Table S2). DISCUSSION In the present study, language style matching (LSM) assessed via conversationally-derived speech from emerging adults across the early psychosis-spectrum was positively associated with real-world social functioning. LSM further predicted unique variance in real-world social functioning independent of group status (i.e., high schizotypy, EP, or UC) and a standardized laboratory assessment of social competency. Finally, in sensitivity analyses also accounting for demographic covariates that were significantly different between groups (i.e., age and sex), LSM continued to predict variance in real-world social functioning, over-and-above social competency. Our observations related to LSM and real-world social functioning may have important implications. First, these results support that language generated at the level of a one-on-one conversational exchange are linked to the overall quality and quantity of social interactions in the real-world setting. Though previous studies have noted that language alterations in psychosis are linked to social functioning (Docherty et al., 2013; Kuperberg, 2010; Pienkos & Sass, 2016) and conversation depth ratings (Abel et al., 2025), to the best of our knowledge, this is the first study to empirically link LSM from conversationally-derived speech to validated measures of social functioning across the early psychosis-spectrum. This further highlights the importance of considering social context and dynamics when striving to understand social processes as they unfold in real-world settings. Next, we assessed the relevance of conversationally-derived language during a critical period of development (i.e., amongst a sample of emerging adults). As emerging adulthood is a notable period of social and interpersonal development (Arnett, 2000), inclusion of data derived from real-time interactions may be particularly impactful on more comprehensively characterizing social dysfunction and social needs that emerge during EP. Finally, LSM predicted significant variance in real-world social functioning, over-and -above social competency scores and when accounting for relevant covariates, which further strengthens our cautious confidence in our pattern of results. Contrary to our expectations, LSM was not associated with embedded measures of social competency (i.e., standardized scores on the SSPA from which LSM scores were calculated) across participants. However, examination of correlations at the group level (Table S1) revealed that LSM was positively correlated with SSPA and GF-Social in EP and SSPA in controls. In contrast, none of these variables were significantly correlated in the schizotypy group. Notably, social competency as assessed via the SSPA is consistently and strongly correlated with GF-Social in the psychosis literature (Angers et al., 2023; Kemp et al., 2025; Kurtz et al., 2016; Patterson et al., 2001)—an association we also replicated in this study in both the EP and control groups. Though unexpected, the divergence observed in the schizotypy group may reflect the inherent heterogenous, multidimensional nature of the schizotypy construct. Individuals with schizotypy often show substantial variability in social and functional impairments that may be less consolidated relative to clinical psychosis groups, and thus, associations among LSM, social competency, and real-world social functioning in schizotypy may be weaker or more diffuse. At the same time, the observed positive association between LSM, SSPA, and GF-Social in the EP and control groups provides preliminary evidence that the LSM reflects a meaningful interpersonal process that has some overlap with broader social functioning. As LSM is thought to represent a spontaneous and unconscious phenomenon in human communication (Launay et al., 2016) and we observed that LSM is associated with a unique variance in GF-Social, it is possible that it assesses features of dynamic interpersonal interaction that are relevant to broader, real-world social functioning that are not captured by or extend beyond the specific social demands of the SSPA itself. Collectively, these results suggest that LSM may capture a distinct aspect of interpersonal processes that contribute to social functioning and can be considered within the broader context of factors that also influence social outcomes. Importantly, we did not observe significant group differences in LSM in the present study. This partially aligns with Abel et al. (2025), which similarly found no differences between controls and schizotypy but contrastingly that individuals with chronic schizophrenia had significantly reduced LSM relative to controls. Importantly, Abel et al. included individuals spanning the 18–60 age range, whereas our participants were restricted to the emerging adulthood phase of life spanning ages 18–30. It is possible that developmental phase or stage of illness may influence the extent to which LSM is disrupted in psychosis, with more pronounced differences emerging later in the illness course. Despite the absence of group-level differences in the present study, LSM nonetheless demonstrated functional relevance in our hierarchical regression analyses—with similar patterns observed when assessed only within the early psychosis group. These findings suggest that LSM may be sensitive to subtle interpersonal processes (e.g., moment-to-moment attunement and reciprocity that emerge when two individuals dynamically apart to one another during an interaction) that are not fully captured by traditional performance-based measures, both within psychosis and across psychosis-spectrum groups. Thus, though mean levels of LSM do not differ across the early psychosis-spectrum, the degree of LSM in conversationally-derived speech remains an important indicator of interpersonal functioning in early psychosis. This study has several limitations to consider. First, we examined LSM derived from conversational exchanges between study participants and a trained member of the research team. Though this represents an important step toward approximating social processes as they unfold between two people, the level of familiarity or relatedness with a researcher is likely to be different than what would be experienced with a true friend or peer. The SSPA is also structured with specific prompts for how the researcher should respond to the participant during the task (e. g., using prescribed responses and prompts, being friendly but reserved in responses), and therefore additional research with more naturalistic samples is needed to understand how LSM may emerge in less structured interactions. Similarly, the SSPA is designed to assess participant performance within the context of a specific social demand or goal and thus, the extent to which LSM is associated with social functioning in more unstructured social exchanges or with specific types of conversation partners (e.g., friend, acquaintance, family member) remains an important area for further investigation. Furthermore, we utilized total composite SSPA scores in our analyses as we did not have specific hypotheses about individual SSPA items or domains. However, our findings support future research probing whether LSM relates more specifically to specific aspects of social competency (i.e., by role-play and/or criterion). Next, it was recently demonstrated that LSM in interpersonal conversations may fluctuate at the within-person level among schizophrenia-spectrum participants (Abel et al., 2025); thus, additional longitudinal studies are needed to examine the interplay of factors that influence LSM dynamics over time. Consistent with our previous work (Angers et al., 2021, 2023), we included participants in the high schizotypy group if they scored at or above the 70th percentile on the overall SPQ-BRU. This differs from other schizotypy studies which utilize higher cut-offs (often the 95th percentile on the positive, negative, or disorganized subscale) to define groups that are considered to be at elevated risk for developing psychosis, more akin to clinical or ultra-high risk groups. Importantly, the study from which these data were drawn was not designed to restrict schizotypy to a narrowly defined risk phenotype and rather to capture the broader, multidimensional construct to encompass a wide range of personality, cognitive, and affective characteristics along the broader psychosis-spectrum. Though prior work has emphasized the value of broader conceptualizations of schizotypy (Kwapil & Barrantes-Vidal, 2015), how our pattern of results may or may not generalize to a more restrictive sample of individuals with schizotypy remains unknown. Further, prospective participants who scored between the 30th and 70th percentile on the SPQ-BRU were not included in the study in order to maximize contrast between groups, which may limit representation of how social and interpersonal processes manifest across the broader schizotypy continuum. Thus, future studies are needed to clarify how different methods for defining schizotypy may influence associations between LSM and social functioning. As another limitation, other than restricting the age of our sample between 18 and 30 to align with emerging adulthood, we did not collect any additional measures of developmental stage. Furthermore, additional variables that are known to impact social functioning (e.g., social cognition, social motivation, negative symptoms) were not assessed in this study and should be considered in the future research. Last, some sociodemographic (e.g., socioeconomic and occupational status) that may influence our findings were not collected in the present study and should be considered in the future research. In summary, to the best of our knowledge, our findings provide the first preliminary evidence that LSM is associated with a gold-standard measure of real-world social functioning among emerging adults across the early psychosis-spectrum. Importantly, LSM accounts for unique variance in social functioning over-and-above the contributions of standard behavioral measures of social competency and other relevant clinical and demographic factors. Additional studies that integrate language generated from conversational exchanges within models of social dysfunction in psychosis may yield important insights into the underpinnings of interpersonal dynamics and difficulties—and relatedly, how to best intervene. Importantly, existing social skills training (SST) interventions for psychosis often target conversational behaviors like reciprocity, turn-taking, and staying on-topic (Corrigan, 1998). These constructs have important conceptual overlap with the interpersonal processes reflected in LSM. At the same time, LSM provides a quantifiable, automated index of moment-to-moment interpersonal coordination that is not typically measured in STT and may not be fully captured by observable conversational behaviors. Thus, LSM is likely best conceptualized as a potential complementary extension to current SST interventions that can be explored in future studies. Though there are currently no psychosocial interventions for psychosis that are designed to address social difficulties at the conversational level, insights from other literatures and areas of medicine and science that also focus on communication disruptions are likely underexplored. In particular, studies and collaborations that integrate perspectives from the speech and language pathology and communicative disorders fields may be especially warranted. Supplementary Material Supplemental data for this article can be accessed online at https://doi.org/10.1080/00332747.2026.2660514. FUNDING This project was supported by funding from the National Institute of Mental Health (K23MH131967 to A.M.M). The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies. Footnotes DISCLOSURE STATEMENT No potential conflict of interest was reported by the author(s). ETHICS APPROVAL STATEMENT This study was reviewed and approved by the Ohio University and The Ohio State University Institutional Review Boards. PARTICIPANT CONSENT STATEMENT All participants provided informed consent prior to their participation in this study. DATA AVAILABILITY STATEMENT Due to the nature of this study, participants did not consent for their data to be shared publicly. Therefore, the data that support the findings of this study are not publicly available due to privacy or ethical restrictions. REFERENCES Associated Data Supplementary Materials Data Availability Statement Due to the nature of this study, participants did not consent for their data to be shared publicly. Therefore, the data that support the findings of this study are not publicly available due to privacy or ethical restrictions.