In the L2 fluency literature there is considerable debate over how best to operationally define pauses and how different pause measures relate to L2 proficiency. This creates a challenge for researchers interested in L2 fluency, and particularly those who are working with groups that vary in L2 proficiency. This article addresses these issues by examining oral production data from 48 learners of Spanish (L1 English) to evaluate which pause measures best serve as markers of fluency. Different pause measures were analyzed -- per 100 syllables/per minute, duration, type (filled/unfilled) and location (in the middle or at the end of a clause) -- along with their relationship to L2 proficiency. Proficiency was operationalized as a composite language knowledge score from L2 grammar and vocabulary tests. Correlational and regression analyses were then carried out to determine how different pause measures account for variability in the proficiency scores. The findings show that longer pauses (of 1 second or longer) are better predictors of L2 proficiency than shorter pauses, unfilled pauses are a stronger predictor of L2 proficiency than filled pauses, and mid-clause pauses distinguish between proficiency levels more clearly than do end-of-clause pauses. The results will serve to guide researchers in determining which pause measure best corresponds to their needs.