The variation in behavioral traits and the adaptive significance behind such variation has been a classic question in behavioral ecology. Traits that enhance while simultaneously impose high fitness costs are particularly suitable for addressing this fundamental question, as their expressions are likely under strong selection. In this dissertation, I investigate the variation in a costly antipredator behavior and the underlying cost-benefit mechanisms. The trait of interest is the voluntary shedding of the tail, or tail autotomy, in lizards. Tail autotomy allows lizards to survive close-range encounters with predators but also has severe fitness consequences, including increased energetic demand for regeneration and wound repair, impaired locomotion, lower mating success and reduced long-term survival. The propensity for tail autotomy exhibits remarkable variation both within and among populations, but whether and how this variation reflects the outcome of intricate cost-benefit dynamics has remained poorly understood. The fist chapter uncovers individual variation in locomotor costs following tail autotomy in green anole lizards. Losing the tail affects jump performance in some individuals more than others. In addition, some individuals are able to improve their jump performance over a relatively short period of time whereas others are unable to do so. The second chapter investigates the roles of risk-taking tendency (bold-shy personality) and body condition in explaining the variation in the latter within populations. Using the brown anole lizards, I discover that bolder individuals compensate for their risky behavior by losing their tails more readily. Interestingly, however, such compensatory dynamics only exists when individuals are well-nourished. The third chapter integrates theoretical modeling and field data to examine how ecology (predation, food availability and the intensity of male-male fighting) might drive the variation in tail autotomy across populations. The model predicts that higher predation and food availability both favor higher propensity for tail autotomy, whereas more intense male-male fighting has a stronger yet opposite effect. The model also successfully predicts the pattern of variation in tail autotomy based on the ecological information. The last chapter is a literature review on the interactions between life-history and performance traits in light of the modern ecomorphological paradigm.