Recent experiments have shown the importance of statistical learning in infant language acquisition. Computational models of such learning, however, often take the form of corpus analyses and are thus difficult to connect to empirical data. We report a cross-situational learning experiment which demonstrates robust individual differences in learning between infants. We then present a novel generative model of cross-situational learning combining two competing processes habituation and association. The models parameters are set to best reproduce each infants individual looking behavior from trial-to-trial in training and testing. We then isolate each infants word-referent learning function to explain the variance found in preferential looking tests.