The purpose of this study is to explore the associations among learning motivation, engagement and outcomes, and the moderating role of various traits in the relationship between deep approaches to learning and outcomes. Based on data from 2,340 students in multiple universities in Taiwan, this study proposes two alternative models, tested by Structural Equation Modelling (SEM). The findings suggest that deep approaches to learning play a critical role in improving college students' cognitive and non-cognitive gains. The moderating effect in learning outcomes could be attributed to student and faculty traits, suggesting that in order to understand students' learning process, each trait should be considered individually. Implications for theory and practice are also discussed in this paper.