This study aims to investigate the academic achievement of students who were admitted to a college from three different college entrance channels in Taiwan. The three channels are personal application, admission by examination and placement, and the Star Program. Longitudinal data composed of 1053 students during eight semesters over 4 years from one university were examined. Beyond basic descriptive statistics, linear growth curve modelling was utilized to capture the dynamic extent of the exploratory variables for predicting students' learning performance. The descriptive statistical results showed that, prior to college, students' learning performance followed the order in which the college entrance channels were listed. During their time at college, however, the pattern of ranking of the students' performance was totally reversed. The results from the growth curve modelling indicated that prior knowledge, that is, students' standardized test scores, had a positive influence on initial grades but a negative impact on the slope of students' learning achievement. Different patterns of learning performance of students from the three channels were observed. Implications and suggestions for evaluating the effectiveness of college entrance channels and assessing students' learning performance from a longitudinal perspective are discussed.