Although aerial surveys are an effective and commonly used method of monitoring wildlife populations, variable detection probability may result in unreliable indices or biased estimates of absolute abundance. Detection probability can vary between sites, sampling periods, species, group sizes, vegetation types and observers. These variables were examined in helicopter surveys of a suite of medium-sized mammals in a hilly environment in central eastern New South Wales. Maximum-likelihood methods were used to investigate the effects of these variables on detection probability, which was derived using the double-count technique. Significant differences were evident between species in the overall analysis, and group size, vegetation, observer pair and sampling period for various individual species when analysed separately. The implications for monitoring wildlife populations between sites and across time are discussed. This paper emphasises that aerial survey indices may be effective in detecting large differences in population size but can be improved by quantifying detection probabilities for a range of variables.