For over two centuries, the Holbein drawing RCIN 912189 in the Royal Collection Trust has been identified as Anne Boleyn on the basis of an unverified eighteenth century inscription. This study reassesses that attribution by combining computational facial recognition with documentary and art-historical analysis. We propose a Working Likeness Methodology that restricts biometric comparison to Holbein’s life-drawn sketches, reducing the stylistic mediation. Using a deep learning model based on AdaFace embeddings, we compare RCIN 912189 with RCIN 912190 (Unidentified Woman). RCIN 912190 shows stable similarity to the authenticated portrait of Elizabeth I by William Scrots and aligns closely with contemporary descriptions of Anne Boleyn, whereas RCIN 912189 conflicts with eyewitness accounts. Validation across Tudor kinship networks shows the method distinguishes documented relatives from unrelated courtiers, suggesting computational facial analysis can provide probabilistic support for attribution where traditional evidence remains limited.
Davies et al. (Tue,) studied this question.