Despite the efficacy of exposure and response prevention (ERP) for OCD, patients who primarily experience obsessions are at heightened risk of attenuated outcomes and relapse. The current study sought to translate laboratory research on inhibitory learning to the use of ERP for primary obsessions, with the ultimate aim of maximizing outcome and reducing the need for follow-up services for this group of at-risk patients. Although preliminary research suggests that learning to tolerate varying levels of fear during exposure therapy enhances long-term outcomes for some anxiety-related problems, no previous studies have examined this in the treatment of obsessions. In the current study, 30 participants with a moderately distressing obsessional thought were randomly assigned either: (a) the gradual exposure condition (EXP-G), emphasizing fear reduction, or (b) the variable exposure condition (EXP-V), emphasizing variability in exposure intensity (EXP-V). Both groups completed four twice-weekly exposure sessions in which subjective and physiological indices of fear were collected. Clinical interview, self-report, and behavioral outcome measures were evaluated by an independent assessor at pre-treatment (PRE), post-treatment (POST), and 1-month follow-up (1MFU). Both the EXP-G and EXP-V interventions were associated with significant decreases in interview, self-report, and behavioral measures of fear from PRE to POST, with no significant differences in PRE/POST changes between the two groups. Furthermore, there was no significant return of fear for either group from POST to 1MFU. Variability in subjective and physiological fear did not predict treatment outcomes, which is in contrast to previous studies suggesting benefits of variability in fear level during exposure for other anxiety-related problems. These results indicate that random/variable exposure warrants future study to better understand the mechanisms, moderators, and implications of this novel approach.