Expert finding systems support two distinct search modalities: name-based queries targeting known individuals, and topic-based queries exploring expertise areas, with users often moving between these modes as their search progresses. In this study, we analyze nearly half a million queries submitted to an institutional expert search system, using a lexicon-based approach to classify each query as "name", "topic", or "unclear". We find that most queries in multi-query sessions have the same mode as the previous query, but that more than a fifth of all such sessions begin and end in different modes. Moreover, mode switching is partially success-driven: queries followed by clicks then switch modes more often than do queries without clicks. Together, these findings show that expert search behavior is shaped by interactions between search modalities, session-level dynamics, and observed engagement outcomes, suggesting several directions for improving expert search systems.