Every Portrait Tells a Story: Tracing Gender Disparities in Paintings, 1400–2024
Face Gender Coefficients on Representations
Wang, Yangyu, ongoing, single-authored paper.
Abstract: Paintings, unlike photographs, encode gender through pliable, low‑indexical symbols whose social meaning remains largely unmeasured. This study asks whether and how such images reproduce gender hierarchy. Drawing on 159,962 faces extracted from 139,910 paintings dated 1400‑2024, I combine RetinaFace and DeepFace with 3D pose estimation to classify depicted gender and measure two compositional cues—relative face size and obliqueness. Random-intercept regressions, period interactions, and subgroup analyses link these cues to artist gender and education, while targeted qualitative readings flesh out pivotal shifts. Results show that women occupy fewer than 30 percent of all faces yet, when present, are rendered larger and more frontal than men, supporting an “objectification” rather than “otherness” pathway. These gaps wax and wane across six centuries but narrow earliest in works by female and self‑taught artists, who disrupt canonical conventions decades before mainstream adoption. The findings demonstrate that computational analysis of low‑indexical art can expose hidden structures of visual power and enrich sociology of cultural production.