This study focuses on differences between multiple-choice science tests and a learning-from-text (LFT) test, and how these tests predict success in basic medical studies. The subjects (N = 503) were applicants to the Helsinki University Medical Faculty. All of them had to take an entrance examination in order to be considered for admission to a 6-year study programme combining medical school and graduate studies. The entrance examination consisted of three traditional multiple-choice science tests and one LFF test, the latter designed to measure deep-level processing of text. A follow-up study was conducted in order to see how the different tasks were related to the grades and pace of studying of those who were accepted onto the programme.As hypothesized, there were very high correlations among the three multiple-choice tests, but no correlations between the LFr subtasks and the multiple-choice tests. LISREL analyses showed that the LFT Synthesis Task, designed to measure the ability to pull together the essentials of a text, was the best predictor of academic progress during basic science studies.