Objective: To determine the relationship between length of time spent expressing highly aroused emotion and therapeutic outcome. Method: Thirty-eight clients (14 male, 24 female) between the ages of 22 and 60 years (M = 39.5, SD = 9.71), treated for depression with experiential therapy, were rated on working alliance and expressed emotional arousal (with the Client Expressed Emotional Arousal Scale) in their three highest arousal sessions. Among the clients, 34 were of European ethnicity, 2 were of Asian ethnicity, 1 was of Latino ethnicity, and 1 was of Caribbean-Canadian ethnicity. Clients were administered the short form of the Working Alliance Inventory following their 4th therapy session and also completed, pre- and posttherapy, the Beck Depression Inventory (BDI), the Global Severity Index (GSI) of the Symptom Checklist-90-Revised (SCL-90-R), the Inventory of Interpersonal Problems, and the Rosenberg Self-Esteem Scale. Results: Hierarchical regressions showed that a nonlinear pattern of expressed emotional arousal predicted outcome significantly above the alliance. This combination predicted 30% of outcome variance on the BDI and 24% on the GSI (p less than 0.01). An optimal frequency (25%) of highly aroused emotional expression was found to relate to outcome, with deviation from this optimal frequency predicting poorer outcome. Conclusions: Too much or too little emotion was found to be not as helpful as a moderate amount. It was concluded that expressed emotional arousal in experiential therapies has a more intricate relationship with therapeutic outcome than has previously been shown and that it is moderate amounts of heightened emotional arousal that improve predictions of therapeutic outcome. (Contains 3 tables.)