Second language acquisition (SLA) is complex and multidimensional. Using the framework of the unified competition model (UCM), the current study explores how robust learning and testing of Chinese Pinyin are fostered by optimal integration of different kinds of feedback in an intelligent computer-assisted language learning (CALL) environment offered by the Pinyin Tutor at TalkBank. The findings demonstrated that the repeated feedback-embedded training with the Pinyin Tutor significantly boosted the learners' proficiency in all aspects of Pinyin knowledge for second language (L2) learners of Chinese whose first language (L1) backgrounds were varied and whose initial proficiencies in Chinese were elementary. Furthermore, there was a strong increase in Pinyin knowledge, as evidenced in the delayed posttest administered 3 months after finishing the training sessions. The results further showed that diagnostic feedback led to greater improvement than basic feedback. The significance of the results is attributed to the design of the Pinyin Tutor, which implements principles from psycholinguistic theory as well as corpus data on the speech production by L2 learners. The study sheds fresh light on improving the Pinyin Tutor, and CALL in general, by incorporating up-to-date findings in educational psychology.