The assessment process of contingency between two binary events was examined in the present experiments using university students. Some researchers have obtained a learning curve in judging contingency and have thereby applied an associative model to an explanation of human contingency judgment. Other researchers, however, claimed that the task structure did not adequately reflect the structure of 2×2 contingency tables and failed to obtain learning curves. After having resolved methodological problems of task structure and procedure (Experiments 1 & 2), we demonstrated little evidence of learning curve in judging contingency (Experiment 3). These results were discussed in terms of associative viewpoints and rule-based models.