In our study it was questioned to what extent a response to a certain personality questionnaire item would be stable in the measurement of personality. We gave a test of 20 items selected from the standardized personality tests four times at three day intervals to the pufils in the first year class of junior high school. About one-half of the students showed substantially the same responses in all of four tests. And 18 items from the same questionnaire were given twice to each of three different groups. One group had the retest after 20 minutes another after one seek and the third after five weeks. These groups were second year class pupils of junior high school. Sixty-85 percent of subjects responded to items in the same way in the first and second times when the tests were given at intervals of 20 minutes, 55-80 percent at intervals of a week, and 50-70 percent at intervals of five weeks, respectively. From this result, we may conclude that the response to questionnaire items is not very stable but rather variable, even though the time intervals are so short that no change might occur on personality. Then we investigated from what factors these response variabilities resulted. Among the factors related to the items' quality, we examined the response similarity in such factors as distribution of responses, contents of items, varieties of meaning, and whether the item has a relevance to every day life. There were the items to which responses were assimilated in high ratio and some were not so high in response similarity in all of these factors. Some pupils responded variably to the same items and almost all of them were found inferior in scholarly attainments or personality characteristics according to their teachers' evaluation. On these factors further research is needed.