Exposure-based therapy is effective in treating anxiety, but a return of fear in the form of relapse is common. Exposure is based on the extinction of Pavlovian fear conditioning. Both animal and human studies point to increased arousal during immediate compared to delayed extinction (>+24 h), which presumably impairs extinction learning and increases the subsequent return of fear. Impaired extinction learning under arousal might interfere with psychotherapeutic interventions. The aim of the present study was to investigate whether arousal before extinction differs between extinction groups and whether arousal before extinction predicts the return of fear in a later (retention) test. As a highlight, both the time between fear acquisition and extinction (immediate vs. delayed) and the time between extinction and test (early vs. late test) were systematically varied. We performed follow-up analyses on data from 103 young, healthy participants to test the above hypotheses. Subjective arousal ratings and physiological arousal measures of sympathetic and hypothalamic pituitary adrenal axis activation (tonic skin conductance and salivary cortisol) were collected. Increased pre-extinction arousal in the immediate extinction group was only confirmed for subjective arousal. In linear regression analyses, none of the arousal measures predicted a significant return of fear in the different experimental groups. Only when we aggregated across the two test groups, tonic skin conductance at the onset of extinction predicted the return of fear in skin conductance responses. The overall results provide little evidence that pre-extinction arousal affects subsequent extinction learning and memory. In terms of clinical relevance, there is no clear evidence that exposure could be improved by reducing subjective or physiological arousal.