This paper deals with the application of topic modeling to a corpus of 17th-century scholarly correspondences built up by the CKCC project. The topic modeling approaches considered are latent Dirichlet allocation (LDA), latent semantic analysis (LSA), and random indexing (RI). After describing the corpus and the topic modeling approaches, we present an experiment for the quantitative evaluation of the performance of the various topic modeling approaches in reproducing human-labeled words in a subset of the corpus. In our experiments random indexing shows the best performance, with scope for further improvement. Next we discuss the role of topic modeling in the CKCC Epistolarium, the virtual research environment that is being developed for exploring and analysing the CKCC corpus. The key feature of topic modeling is its ability to calculate similarities between words and texts. In an example we illustrate how such an approach may yield results that transcend a regular text search.