Despite the continued growth of research demonstrating that marriage promotes desistance from crime, efforts aimed at understanding the mechanisms driving this effect are limited. Several theories propose to explain why we observe a reduction in offending after marriage including identity changes, strengthened attachments, reduced opportunities, and changes to routine activities. Although mechanisms are hard to measure, we argue that each proposed mechanism implies a specific change process, that is, whether the change that ensues after marriage is enduring (stable) or situational (temporary). Drawing on a medical model framework, we cast the role of marriage as a treatment condition and observe whether the effect of marriage is conditional on staying married or whether the effect persists when the “treatment” is taken away (i.e., divorce). We use 13 years of monthly level data from the National Longitudinal Survey of Youth (NLSY97), a nationally representative sample containing close to 3,000 individuals with an arrest history, to examine changes in relationship status and arrest from adolescence into young adulthood. Estimates from multilevel within‐individual models reveal greater support for situational mechanisms in that divorce is detrimental particularly for those in longer marriages; yet they also reveal important caveats that suggest a closer examination of the marriage effect. This research adds to the growing body of knowledge regarding the marriage effect by redirecting desistance research away from asking if marriage matters to asking how marriage affects desistance. A better understanding of this change process has important implications for criminal justice policy.