This study replicated a pile sort task involving 135 emotion terms originally presented by Shaver, Schwartz, Kirson, and O'Connor (1987) with a slight change in procedures, from an unconstrained single-sort task to a constrained successive-sort task, made necessary by methodological problems during their data analysis. Subjects (86 male and female undergraduates) were asked to assign names to their piles during the successive stages of sorting. This change produced a hierarchical cluster analysis solution supporting interpretations other than the prototype structure found by Shaver et al.. Decision criteria reported by subjects during sorting were described and revealed dimensions reported by previous investigators, suggesting that this is a viable method of determining dimensions important in distinguishing emotion terms.