The quality of collaborative writing process to some extent depends on the verbal episodes occurring during the collaboration. This study aimed to describe English as a foreign language (EFL) students’ verbal episodes in proficiency pairings. A case study design was used in this study. It involved 40 EFL students of the English department in an Indonesian university. They were divided into two groups based on the types of proficiency pairings: heterogeneous (20 students) and homogenous (20 students). Therefore, this study was also intended to identify which of the two types of proficiency pairings produces more verbal episodes than the other. The verbal processes were audio recorded and then transcribed. The audio transcriptions were analyzed for common themes related to episode categorization. To ensure the reliability of the episode analysis, inter-coder, and intra-coder checks were employed. The results showed that Indonesian EFL students used three major types of verbal episodes: language-related episodes (LREs), text-related episodes (TREs), and scaffolding episodes (SEs). The study also revealed that proficiency levels determined the categories of LREs and SEs most frequently produced by the heterogeneous and homogeneous pairs. Both the heterogeneous and homogeneous pairs were likely to produce almost the same frequencies of categories of TREs more particularly in terms of organization and content. Further research might explore the link between the number of episodes and the learning gains.