In recent years, more and more interactive features have been incorported into information retrieval systems, but an evaluation paradiam for highly interactive systems has yet to emerge. This study reports the results of an experiment conducted to test the effectiveness and usefulness if a Term Suggestion Interface for a Buddhism study collection. Terms co-occurring with users' original queries were extracted and presented to users in order to refine the search. The results suggest that the performance of the experimental interface is influenced by the number of potentially relevant records in the collection. The methodological implications for interactive information retrieval evaluation are also discussed.