Different psychotherapy theories describe process patterns of emotional arousal in contradictory ways. To control both treatment and therapist responsivity, this study sought to test dynamic patterns in the arousal of negative affect using a controlled experimental study of expressive writing. There were 261 participants (78% women; M = 21 years old; 56% White) who suffered unresolved traumas who were randomly assigned to an expressive writing task and asked to write about their deepest thoughts and feelings, or to a writing control. Participants wrote for 15 min on three consecutive days, completing the Positive Affect and Negative Affect Scale before and after each visit. Data across 6 time points were subjected to hierarchical linear modeling (HLM) and pattern analyses. Session-by-session (24 hr periods), the expressive writing group showed an overall linear decrease in negative affect (β = −2.273, p < .001). However, in pre- to post-session ratings (15 min periods), the expressive writing group also demonstrated increases in negative affect (β = 6.467, p < .001). Neither of these patterns were observed in the control group. Pattern analysis demonstrated 69.8% of cases in the expressive writing group perfectly or almost perfectly followed a predicted zig-zag pattern (p < .01). No control cases showed this pattern. Findings demonstrate how the habituation/inhibition hypothesis (“it gets easier as one gets over it”) and the meaning-making hypothesis (“it gets worse before it gets better”) can both be supported, each at different scopes of analysis. Implications clarify the role of emotional arousal in change.