Increasing rates of natural disturbances under a warming climate raise important questions about how multiple disturbances interact. Escalating wildfire activity in recent decades has resulted in some forests re-burning in short succession, but how the severity of one wildfire affects that of a subsequent wildfire is not fully understood. We used a field-validated, satellite-derived, burn-severity atlas to assess interactions between successive wildfires across the US Northern Rocky Mountains a 300,000-km2 region dominated by fire-prone forests. In areas that experienced two wildfires between 1984 and 2010, we asked: (1) How do overall frequency distributions of burn-severity classes compare between first and second fires? (2) In a given location, how does burn severity of the second fire relate to that of the first? (3) Do interactions between successive fires vary by forest zone or the interval between fires? (4) What factors increase the probability of burning twice as stand-replacing fire? Within the study area, 138,061 ha burned twice between 1984 and 2010. Overall, frequency distributions of burn severity classes (low, moderate, high; quantified using relativized remote sensing indices) were similar between the first and second fires; however burn severity was 5-13% lower in second fires on average. Negative interactions between fires were most pronounced in lower-elevation forests and woodlands, when fire intervals were <10 yr, and when burn severity was low in the first fire. When the first fire burned as high severity and fire intervals exceeded 10-12 yr, burn-severity interactions switched from negative to positive, with high-severity fire begetting subsequent high-severity fire. Locations most likely to experience successive stand-replacing fires were high-elevation forests, which are adapted to high-severity fire, and areas conducive to abundant post-fire tree regeneration. Broadly similar severities among short-interval 're-burns' and other wildfires indicate that positive severity feedbacks, an oft-posited agent of ecosystem decline or state shift, are not an inevitable outcome of re-burning. Nonetheless, context-dependent shifts in both the magnitude and direction of wildfire interactions (associated with forest zone, initial burn-severity, and disturbance interval) illustrate complexities in disturbance interactions and can inform management and predictions of future system dynamics.