Prey experience a broad range of temporal patterns of predation risk and may face different problems when assessing risk level and deciding how to allocate costly antipredator behaviour across time. We simulated in the laboratory repeated predatory attacks of different intensity (low versus high risk) to Iberian rock lizards, Iberolacerta cyreni, with similar numbers of low- and high-risk attacks across a trial, but with different temporal patterns of presentation (i.e. a series of low-risk and then a series of high risk, or vice versa, or successive attacks where risk changed randomly). We measured activity levels, time hidden in refuges after each predatory attack, and propensity to hide in refuges when the predator is close but not actually attacking (‘false alarm’ flights). Antipredator responses were of higher intensity when risk was high, but responses also depended on temporal patterns of risk. When previous attacks were of low-risk or when risk was random, prey activity did not change between risk levels. False alarm flights in the low-risk situation were more frequent when risk was random or immediately after a series of high-risk attacks. When successive attacks were of similar risk level, lizards seemed able to predict risk level in the next attack and modified their antipredator behaviour accurately. However, when risk level was random, lizards did not seem able to predict risk level and adopted a conservative strategy, overestimating risk even though costs of the antipredator responses increased.