To quantify how frogs in terrestrial environments respond to recurrent fire, and to what extent this is mediated by isolation from breeding sites or vegetation structure. We used data from 8 years of pitfall trapping, collected via a random stratified design, to quantify frog occurrence at 110 locations. We then used an information theoretic approach to compare 13 logistic generalized linear mixed models, each of which related frog occurrence to a distinct combination of additive and interactive effects of fire, vegetation structure and proximity to known breeding sites. For all four species, the effect of one or more fire variables on frog occurrence depended on both the density of breeding sites in the surrounding area, or on the vegetation structure at the trap locality. A classic “fire averse” response of initial declines followed by post‐fire recovery did occur, but only in frequently burned, low‐quality terrestrial habitats (i.e., heath vegetation for Uperoleia tyleri, or locations with few available breeding sites for the remaining species), or in some cases, when suitable habitats were infrequently burned. However, a “fire‐dependent” result of negative effects of time since fire was also evident for some species and contexts. The effect of fire on frog occurrence can be mediated by environment. Therefore, a single species could be identified as either “fire dependent” or “fire averse” depending on the combination of isolation, vegetation types and fire histories in the study region. Failure to account for the context specificity of fire response curves could lead to incomplete conclusions regarding the effect of time since fire—or the cumulative impacts of multiple fires—on faunal assemblages.