Serial Dependence in face perception

The complete preregistered plan is provided in /preregistration/preregistration.pdf. In case of any discrepancy between the OSF form fields and preregistration.pdf, preregistration.pdf is the authoritative preregistered plan. This project (empirical aesthetics/visual perception) examines serial dependence (SD) in facial attractiveness judgments—systematic history-related biases in current judgments as a function of the preceding face—within the continuity field framework, where biases can be attractive (assimilative) but may attenuate or become repulsive under strong discontinuities (e.g., category changes). Using faces as high-dimensional, socially meaningful stimuli, we test five linked questions: (i) how objective structural similarity modulates SD, (ii) whether gender and ethnicity boundaries add effects beyond metric similarity, (iii) the contribution of response/decision requirements vs. perceptual processing, (iv) changes in extreme regions (very low/high similarity; very large |ΔAttr|), and (v) whether bisexual vs. non-bisexual participants differ in gender-boundary modulation. We run an online PsychoPy/PsychoJS (2024.2.4) experiment via Pavlovia with Prolific recruitment: 4 blocks (ABAB/BABA; 160 trials) of sequential face pairs (t-1, t); in Response blocks (A) both faces are rated, whereas in No-response blocks (B) only t is rated, using a continuous 1–100 slider. Stimuli are 298 neutral faces from CFD 3.0 and CFD-India (age-rated 20–30), covering five ethnicities (White, Black, Latino, Indian, Asian) and two genders (male/female), sampled randomly with replacement to form pairs. The overarching question is how attractiveness at t-1 biases judgments at t, and when SD is attractive, repulsive, or modulated by similarity, social categories, and response requirements. SD is operationalized as ΔAttr = SampleMeanRating(stimulus_t-1) − SampleMeanRating(stimulus_t) and tested with trial-level linear mixed models (LMMs), where positive ΔAttr slopes indicate attractive SD and negative slopes indicate repulsive SD; confirmatory expectations are overall attractive SD (H1), stronger SD for same-gender (H2) and same-ethnicity (H3) pairs, stronger SD when a response is required at t-1 (H4), and a positive similarity gradient (H5), with exploratory tests for extreme-region effects and subgroup differences, and Holm–Bonferroni control across H1–H5. Project materials created by the authors (e.g., preregistration text, appendices, and analysis scripts) are licensed under CC BY 4.0; this license does not apply to third-party materials (including CFD/CFD-India stimulus images and any restricted database fields), which remain governed by the respective providers’ terms.