Abstract: Three experiments were conducted to explore which stages within the decision process make decision performance better by developing a modified Iowa Gambling Task (IGT). Experiment 1 and 2 were mainly used to confirm whether the modified IGT realized the function of the classical IGT, whether the decision process could be separated, and the influence of knowledge about four decks of cards on task performance. Experiment 3 is an improvement upon Experiment 1 and Experiment 2. The results show that the modified IGT can realize the basic function of the classical IGT and the separation of the IGT decision process and can identify the more accurate time of reaching the conceptual level. Healthy adults reached the conceptual level after 200 trials (starting at the sixth block), much later than the after-80-trials for the classic IGT. This study indicates that the knowledge of four decks of cards did not affect participants' performance on the task but learning to skip and give up adverse options made participants' decision performance better.