Aggregate convergence and constituent divergence in incel-related online communities: A mentalizing-theory lens on cross-platform linguistic patterns

The “incel” (involuntary celibate) subculture comprises an online ecosystem spanning mainstream-hosted, independent, and recovery-oriented communities. Using Fonagy’s mentalizing framework, we derived four community-level Linguistic Inquiry and Word Count (LIWC-22) patterns: rigidity–flexibility (Psychic Equivalence-inspired), goal–affect balance (Teleological Mode-inspired), cognitive-affective elaboration (Pretend Mode-inspired), and counter-positioning–curiosity (Epistemic Mistrust-inspired). Study 1 compared 2,261 incels.is discussion-thread documents with 15 LIWC-22 Test Kitchen reference corpora and the overall benchmark. Incels.is generally showed higher Psychic Equivalence-inspired scores and consistently higher Epistemic Mistrust-inspired scores; Teleological Mode-inspired differences were reference-dependent, whereas Pretend Mode-inspired results were heterogeneous across reference genres. Excluding text from the top 1% of identifiable contributors retained 61 of 64 comparison directions. Study 2 analyzed 28,980 platform-specific user aggregates across five communities. Although platform differences were supported, their magnitude was modest (η 2 = .037–.055), with most user-level linguistic variation remaining within platforms. Relative to IncelExit, active-incel platforms generally scored higher on the Psychic Equivalence- and Teleological Mode-inspired patterns and lower on the Pretend Mode-inspired pattern. IncelExit had the highest Epistemic Mistrust-inspired score despite high curiosity, showing that the composite did not isolate epistemic closure from counter-positioning. Across all four patterns, community means showed a consistent ordering, although adjacent contrasts were not uniformly supported. The three-platform portion of the ordering persisted in temporally matched and year-adjusted analyses. Evidential support was clearest for PE, partial for TM, limited for EM and PM. The latent ecological or psychological basis of the ordering remains unresolved, and the linguistic patterns are descriptive rather than validated measures of mentalizing.