Predators confronted with an unfamiliar prey must decide whether or not to attack it. This decision is dependent on the predator's previous experiences with the prey type. In this thesis, I tested the ability of Jamaican field crickets, Gryllus assimilis, to learn a novel binary food choice between a rewarding and unrewarding prey using visual cues. Evidence of learning was confirmed across trials. Moreover, the colour of the prey significantly affected the probability of crickets choosing the palatable option, with green prey more likely to be attacked than blue prey. I then developed a model that formalized prey selection in terms of an exploration-exploitation trade-off. Here, I identified the optimal sampling strategy for a predator with Bayesian learning. I demonstrate that a predator's prior beliefs (Bayesian priors), and the certainty it has in its beliefs affect the optimal sampling strategy, and hence the nature of selection it places on prey.