In protected areas around the world, wildlife habituate to humans and human infrastructure, potentially resulting in human-wildlife conflict, and leading to trophic disruptions through excess herbivory and disconnection of predators from prey. For large species that threaten human safety, wildlife managers sometimes attempt to reverse habituation with aversive conditioning. This technique associates people as a conditioned stimulus with a negative, unconditioned stimulus, such as pain or fright, to increase wariness and prevent the need for lethal wildlife management. Resistance to aversive conditioning by some habituated individuals often results in more frequent conditioning events by managers, but there are few studies of conditioning frequency with which to evaluate the usefulness of this management response. We evaluated the effect of conditioning frequency on the wariness of elk (Cervus canadensis) by subjecting marked individuals to predator-resembling chases by people over a period of three months. In that time, animals were subjected to conditioning a total of 3, 4, 5, 6, 7, or 9 times which we analyzed as both an ordinal variable and a binary one divided into low (3-5) and high (6-9) conditioning frequencies. We measured wariness before, during, and after the conditioning period using flight response distances from an approaching researcher. During the conditioning period, overall wariness increased significantly for elk in both treatment groups, although the increase was significantly greater in individuals subjected to high conditioning frequencies. However in the post-conditioning period, wariness gains also declined most in the high-frequency group, equating to more rapid extinction of learned behaviour. Across all treatment frequencies, rapid changes in flight responses also characterized the individuals with the lowest wariness at the beginning of the study period, suggesting that individuals with greater behavioural flexibility are more likely to habituate to both people and their attempts to change wariness via aversive conditioning. Together, our results imply that aversive conditioning may be most effective at intermediate frequencies and that its utility might be further increased with proactive assessment of individual personalities in habituated wildlife.