Six golden hamsters (Mesocricetus auratus), which have not usually been used as subjects in operant conditioning research, were exposed to a spatial probability learning task in the apparatus with two levers differing in probabilities of reinforcement. Two pairs of probabilities of reinforcement (70% : 30% and 60% : 40%) were used in a discrete-trials procedure with a noncorrection method. A baseline condition in which subjects were required to respond to each lever equally often was presented prior to each of the first and the second probability conditions. In both probability conditions all the subjects showed maximizing : subjects allocated most of responses to the alternative with a higher probability of reinforcement. The findings reveal that golden hamsters responded in the same way as other species such as rats and pigeons.