We investigated the processes of making inference in binary choice tasks. We proposed statistical models for two simple heuristics and two knowledge-based inferences, and com-pared how well these models could explain the observed pat-terns of inferences. It was found that the best model for explaining choice patterns varied depending on the inference problem. In particular, results suggested that participants used a simple heuristic when they had difficulty in retrieving clues pertaining to the inference problem. In contrast, when participants could retrieve enough clues pertaining to inference problems, they made inferences based on these clues.