The transfer of knowledge is considered to be a fundamental goal of education; therefore, knowing and understanding the conditions that influence the efficiency of the transfer from learning activity to problem solving play a decisive role in the improvement of science education. In this article, the results of a study of 196 high school students' ability to transfer knowledge in explanatory model construction are present. Three test groups were formed, traditional, prediction and lab groups, in which students were involved in three different learning activities. A week after instruction, students were tested with a foil test and Lawson's Classroom Test of Scientific Reasoning. According to the results, little knowledge transfer from learning activities to the foil test occurred. Among the three tested learning methods, the one asking for prediction seems to best improve the transfer of knowledge. Time spent on activities had little or no effect on the transfer of knowledge. Some possible reasons for the observed results are presented, and the importance of correct scientific explanation during the learning process is considered.