Laboratory investigations of gambling are sometimes criticized as lacking ecological validity because the stakes wagered by human subjects are not real or no real monetary losses are experienced. These problems may be partially addressed by studying gambling in laboratory animals. Toward this end, data are summarized which demonstrate that laboratory animals will work substantially harder and prefer to work under gambling-like schedules of reinforcement in which the number of responses per win is unpredictable. These findings are consistent with a delay discounting model of gambling which holds that rewards obtained following unpredictable delays are more valuable than rewards obtained following predictable delays. According to the delay discounting model, individuals that discount delayed rewards at a high rate (like pathological gamblers) perceive unpredictably delayed rewards to be of substantially greater value than predictable rewards. The reviewed findings and empirical model support the utility of studying animal behavior as an ecologically valid first-approximation of human gambling.