認知型テスト日本版に関する一研究

概要

論文の詳細を見る
We assumed that an accumulated exposure to TV and comics might have some effects on cognitive styles of children. Before confirming this assumption, we carried out a study to revise the. Japanese versions of GEFT (Group Embedded Figure Test), CST (Conceptual Style Test) and MFF (Matching Familiar Figures) which. were developed tentatively by the Tatsuno team, Tokyo Kyoiku University. The GEFT consists of twenty-five geometric figures which are employed from Witkin's Test and slightly simplified and coloreliminated for children version. The CST including twenty-six familiar figures and the MFF including twenty-one familiar figures are both similar to Kagan's Tests. Our principle to revise these Japanese versions and the findins by study are as follows. GEFT We rearranged the test items to make a spiral from due to their difficulties which were determined by a preliminary study, using eleven graduate students and eighteen children of: -a Sunday school. And we conducted the GEFT in the five classes of fifth graders at a public school located in the suburb of Tokyo metropolitan area, with the total number of two hundreds and five. The time limit of performance was seven minutes. The mean score gained from this test was 13.00 (SD=4.66), and. the range of scores was 1 to 25 (a girl gained the full mark) Table 1 shows the response pattern to each item. This analysis, was made with the date gathered from three classes out of the five, totaling one hundred and twenty two children. Response error ratios range from 1% to 56% and tend to form a spiral figure of difficulties, although they do not show regular distribution pattern. The number of non -responses tends to increase remarkably as items come later. The coefficient of reliability was rtt=0.88 (KuderRichardson's formula 20). CST Responses to CST are supposed to be much influenced by the stimulus itself. We made a new Japanese version on the basis of following standards. (1) Three familiar figures in a set must be paired in the two ways, either 'analytical vs. functional categorization' or 'analytical vs. inferential categorization'. (2) A functional categorization would be likely confined to a simple relation between two stimuli, for example, 'a match box & cigarettes', and a pair implying some fantastic relation would be avoided. (3) Similar parts of some two figures which induce an analytical categorization should be drawn less remarkedly. (4) An inferential categorization should be made with a mediating concept. The new Japanese version of CST including twenty-six items was conducted with total subjects. The instruction was the same as Kagan's. The mean analytical score was 8.54 (SD=4.80) and the range was 1 to 20. The result of item analysis (Table 2) with the same three classes also showed that responses were somewhat inclined to non-analytical direction (the mean of analytical responses by item was 35%, and the range was 3% to 70%). It is, however, better than the other case, because the test has been constructed to assess a tendency of analytical perception and a ceiling effect will not likely occur. The coefficient of reliability was r_<tt>=. 89 (Kuder-Richardson's formula 20). MFF We selected arbitrarily ten items from the Japanese version, and added the Kagan's twelve items to those. Then we split those twenty-two items into two parts, A & B. After a preliminary study which was conducted with two groups of children assigned respectively to the form A or B, we have adopted twelve items out of the both forms on the basis of the following criteria: to eliminate any item showing (1) too high response time mean, (2) almost non-errors at the first response, and (3) error tendency to concentrate in a few alternatives. We conducted the revised version of MFF with the two classes out of the five, totaling eighty boys and girls. The mean response time was 12.3 sec. (SD=6.4) and the range was 4.1 sec. to 37.4 Table 3 shows the means of the first response time by each item. Table 4 shows the frequency distributions of errors on alternatives by each item. Those which have too low error frequencies would not function as an alternative, while they would do as a facilitating response agent. From this point of view, almost all items seem to have substantially four alternatives at least. We can find that there is no tendency to indicate some sort of saturation, because. the error by item does not increase as it become later (see the right column of Table 4). We can find, however, a response bias stemming from the arrangement of alternatives (Figure 1). The alternatives of no. 2 and no. 5 arranged at the middle have more errors than those at the left or the right, and alternatives on the first line have more errors than those on the second line. This sort of response bias, however, would function to differentiate responses on the two dimensions of reflection and impulsivity. It is better to shorten performance time of the test without lowering reliability. We made a correlation matrix of respnse time by item, the total respose time, accumulated errors at each response step, and the total error. Then we found that all correlations were statistically significant at the 5% level. The association between the total response time and total error showed r=-.54. We found also a correlation between total error and accumulated error at the third response step was. 99, in other words, 98% of the total error variance can be explained by the errors at the third response step. Besides, response time to any item had significant correlations with the errors at the third step. Therefore even if we stop responding at third step, the reliability of MFF would not be changed. Relations among the Three Cognitive Style Tests. Any correlation among the three conitive style tests was not statistically significant (Table 7). This means they include independent factors.
国際基督教大学の論文

認知型テスト日本版に関する一研究

スポンサーリンク

概要

著者

関連論文

スポンサーリンク