We developed a lecture demonstration experiment to improve students’ scientific reasoning skills. In this experiment, students roll two objects (e.g. a metal cylinder, ping-pong ball etc.) down a slope, compare their velocities, and study the factors that affect the motion of the objects. The goal given to the students is to predict the result of the experiment with a pipe and a spool. In order to make a certain prediction, students repeatedly conduct preliminary experiments with various objects. This practice has two features. First, students design the procedure of the experiments. Second, students conduct scientific reasoning (e.g. hypo-deductive reasoning, control of variables etc.) with a worksheet. We evaluated the extent to which students’ scientific reasoning skills were improved through this practice by means of Lawson’s Classroom Test of Scientific Reasoning (CTSR). We found that the average CTSR score increased in the class with a low average score (df=17, p<.01, d>0.7), but did not increase in the class with a higher average score. One of the causes could be that the students with low scores followed the high scoring students’ assertions and did not conduct the scientific reasoning themselves.