The attribution of two portraits of Rembrandt revisited: a replication study in art history

This article presents the results of a replication study in the humanities, more specifically in art history. The initial study was carried out in 1998–99 and concerned the attribution of two similar painted portraits of Rembrandt in the collections of the Mauritshuis in The Hague and the Germanisches Nationalmuseum in Nuremberg. In the replication study, the initial research questions were readdressed: Is the Mauritshuis version painted by Rembrandt or not, is the Germanisches Nationalmuseum version painted by Rembrandt or not, and how do these paintings relate to each other? Two types of replication were utilised. First, a reproduction – a repetition that stayed as close to the initial study as possible. Second, a conceptual replication – a repetition that used an improved study protocol including new technical research methods. As part of the conceptual replication, the paintings were brought together and compared in real life during an expert meeting. A new protocol to structure expert meetings – the Attribution Expert Consensus Meeting (A-ECM) – was introduced to increase future replicability and mitigate bias related to group dynamics. The reproduction and conceptual replication corroborated the conclusions of the initial study. The A-ECM method revealed differences and similarities in the experts’ argumentation and their valuation of evidence in support of these conclusions. The study demonstrated how replicating this attribution study not only enhanced the trustworthiness of the initial findings, but also revealed broader epistemic implications. In particular, the replication process proved instrumental in identifying avenues for refining attribution methodologies. These include enhancing transparency, promoting equitable knowledge exchange, mitigating biases, and improving future replicability of expert assessments. Collectively, these improvements contribute to more robust and well-substantiated attribution practices. The introduction of the A-ECM further exemplifies how formalised consensus methods can increase scholarly transparency, efficiency, quality, and future replicability of attribution processes. As such, replication can contribute to pathways for adapting art history to current demands of Open Science.