Altered fire regimes are a driver of biodiversity decline. To plan effective management, we need to know how species are influenced by fire and to develop theory describing fire responses. Animal responses to fire are usually measured using methods that rely on animal activity, but animal activity may vary with time since fire, potentially biasing results. Using a novel approach for detecting bias in the pit-fall trap method, we found that leaf-litter dependent reptiles were more active up to 6 weeks after fire, giving a misleading impression of abundance. This effect was not discovered when modelling detectability with zero-inflated binomial models. Two species without detection bias showed early-successional responses to time since fire, consistent with a habitat-accommodation succession model. However, a habitat specialist did not have the predicted low abundance after fire due to increased post-fire movement and non-linear recovery of a key habitat component. Interactions between fire and other processes therefore must be better understood to predict reptile responses to changing fire-regimes. We conclude that there is substantial bias when trapping reptiles after fire, with species that are otherwise hard to detect appearing to be abundant. Studies that use a survey method based on animal activity such as bird calls or animal movements, likely face a similar risk of bias when comparing recently-disturbed with control sites.Electronic supplementary material The online version of this article (

No abstract

1. Testing the extent to which traits act alone or in combination with other traits to influence responses to fire informs the trade‐off between increased generalisation using single traits and increased predictive power using interactions. This study investigated the following question: do four traits (body size, trophic group, dispersal ability, and stratum of the ecosystem), alone or in combination, best explain changes in beetle occurrence with time since fire? 2. The data from 4 years and 15 independent fires in southern Australia were analysed using generalised linear mixed models. The study also assessed whether detectability depends on time since fire using multi‐year detection models, because detectability has the potential to confound occurrence patterns. 3. The best model included the three‐way combination of size, flight, and trophic level interacting with time since fire and with year. The relationship between detectability and time since fire was similar to the occurrence relationship in six of the 10 trait–combination groups, with flightless species generally showing reduced detection probability as time since fire increased. Detectability did not confound occurrence responses for four trait groups, with three increasing with time since fire and one decreasing. 4. Generalisation using main effects of traits risks oversimplifying animal responses to fire, because combinations of traits influence the direction and magnitude of the response. Also, taking detectability into account is critical to correctly interpretating occupancy data. Three‐way trait combinations that differ by just one trait, particularly dispersal ability, can result in either negligible effects of disturbance on detectability or strong effects that influence observed occurrence.