Isaac Beeckman (1588-1637) is a self-learning man. He learned medicine by his reading medical books (contemporary and classic). In this paper I study how Beeckman read and understood them. He did not merely memorize them. But he gave some supplementary explanations to their (he thought) insufficient passages, sometimes criticized them and gave mechanical explanation that was based on atomism with hydrostatics. We can find similar ways of reading in the works of Lucretius and Cardano which young Beeckman read repeatedly. Beeckman learned the way of explaining natural phenomena with atomism from Lucretius' De rerum natura, and the way of explaining mechanics with natural philosophy and of demonstrating the principles of natural philosophy with machines from Cardano's De subtilitate. Beeckman's interactive reading is a good style of self-learning, but to avoid some bad effects of self-learning, he had to talk actually to a good respondent such as young Descartes.