This research consists of three parts: (1) the analysis of error responses of teacher-made tests,(2) the analysis of scientific knowledge from the point of view of its developmental aspect and (3) the analysis of the retention of scientific knowledge from the point of view of its contents.Each part will be explained in detail as follows(1) Fifteen problems referred to “science of water” were selected from various teacher-made problems. Errors to all items of the problems were classified into the several degrees from trifling degree to serious degree, according to the distance from correct response, and scored as -1, -3, -3, etc., As the results, many sorts and various degrees of errors were revealed. In comparison of the rate of errors, the rate of non-response and the average minus scores of a error between the pupils who were good in science and those were poor, the good pupils tended to get correct response and even if they mistake, they made only trifling errors. The pupils on the medium made various degrees of errors, some of them relatively trifling but some considerably serious or unusual. Many poor pupils had no response or made so serious errors that we could not find out the relationship to the correct response at all.(2) The processes in which the scientific concepts developed were considered as follows.i) the primitive experiences in daily life,ii) the higher experiences by observation or experiment,iii) the lower abstraction, e.g. the qualitative knowledge, etc.,iv) the fixated conception, e.g. the quantitative law, etc.Eight tests were made to diagnose the structure of knowledge of- the pupils of the 1st and the 2nd grade of secondary school. Four of them were made referring to “science of water” and the others to “temperature and heat”. Each test consisted of about 30 items which were arranged according to the developmental process of the scientific concept. Besides, two scientific vocabulary tests were made and administrated. As the results of these testings, the rates of correct responses to items decreased gradually from the items which needed only the primitive knowledge to those which needed high abstract knowledge. Especially the decreasing tendency was significant in the rates of the pupils poor in science and of the 1st grade pupils.(3) The tests for the 4, 5, and 6th grade were made and each of which contained the problems of six units. The problems of each unit were made in order to evaluate four aspects of ability. The 4, 5, and 6th grade pupils were given all tests which contained only the problems that the pupils had already learned in the class. The results showed general increase of the rate of correct responses as the grade became higher and the variance of the rate within the grades, four aspects of ability and six units were all significant. As to the items of which the pupils had already learned and of which the rates of the correct responses were lower than the other items, the higher the grade was the more improved tendency appeared in the rates of correct responses.However, there were items which gave the lower rate of correct response as the grade was higher. As for these items, the higher the intelligences of the pupils were, the slower the rate decreased.