Levels of Canonicity

In this paper we investigate the dynamics of canon formation in Danish Golden Age paintings using techniques of visual embeddings and network analysis. Canon formation is an intricate phenomenon, especially in that time frame and medium. Thus, this case study deeply intertwines art-historical theory and historical contexts with computational methods and statistical measures. Our dataset consists of 1,656 digitized paintings from SMK (The National Gallery of Denmark, SMK), annotated with metadata including artist, date, and canonical status. We address the context and the challenges of this dataset, before we move on to the methodology of visual embedding and network construction. By correlating network measures – centrality, community position, and cluster specificity – with canonical status, we assess whether canonized artworks occupy distinctive structural positions within the corpus. Situating the analysis within art-historical and museological debates, we highlight how both 19th century curatorial practices (e.g., N.L. Høyen’s nationalist selection criteria) and 21st-century digitization initiatives (notably SMK Open) contribute to ongoing processes of canon formation. Our findings enable us to discuss on a broader scale the different reasons a painting becomes canon and the distinct types of canonicity that emerge within the corpus, while also considering how processes of digitization and curation in contemporary museum practice actively shape and reinforce these hierarchies.