本研究は,青年期の思考を特徴づけるとされている論理的思考,その中でも特に,含意的関係に基づく条件推論能力の発達を明らかにしようとしたものである。その際,誤答分析をも含めたより詳細な条件推論能力の発達の理解を可能にするために,従来の方法を改善して多肢選択による課題を構成し,実施した。 その結果,条件推論能力の発達は推論形式によって大きく異なっており,前件否定(DA)・後件肯定(AC)は前件肯定(AA)・後件否定(DC)に比べて,大学生でも著しく困難であった。また,条件推論の難易は,そこに含まれる推論内容によっても異なることが明らかにされた。 このように,知的に成熟していると考えられる大学生においてさえ,前件否定・後件肯定の推論が困難であるのは,形式的操作によって青年期の思考を記述したPiaget理論に対する反証であると考えられるかもしれない(Roberge, 1971bなど)。しかしながら,誤答の分析及び個々人の反応パタンに関する分析からは,1)前件否定・後件肯定の推論が困難であるのは必ずしも論理的な操作ができないためではない,2)正答に至らなくても,論理課題に対する無理解を示す誤答は中1以降で殆ど見られなくなる,ことなどが示唆された(実験I)。これらを考え合わせると,先の結果だけからPiagetの考えを妥当でないとすることはできないであろう。 そこで,実験Iで得られた特徴的な誤反応パタンの意味を明らかにし,それらについて行った解釈の妥当性を吟味するために実験IIを実施した。その結果,直接条件推理の解法を教えなくても,前件否定・後件肯定での正答が有意に増加しうることが示唆された。これは,前件否定・後件肯定の推論が困難であるのは,単に条件文の誤った認知(A-Cパタン),あるいは結論の表現についての理解が不十分なため(A-Bパタン)であるという,実験Iでの解釈を支持するものある。 ところがここで,もしA-Cパタンの反応が論理操作の不可能なことを意味しないとすると,既に小3の段階から(あるいはそれ以前にも)こうした思考が可能であると考えねばならなくなる。この点に関しては,A-Cパタンの反応が,年齢段階により異なった意味を持つのではないかと予想された。実際,Knifong(1974)は,年少児では,論理的推論を行わなくても転導推理(trunsductive reasoning, Piaget, 1924)によって,A-Cパタンの反応が起こりうることを示唆している。転導推理とは,一種の知覚的・直感的な推理様式で,「もしAならば,Bである」とき,「Aである」から「Bである」を,「Aでない」から「Bでない」というように,それら2つがちょうど"釣り合っている"ように結論するものである。標本数の関係上分析結果は示さなかったが,実験IIにおいても年齢段階によって訓練効果の異なる傾向が見られている。この点については,今後,被験者数を統制したより詳細な検討を行う予定である。

The present study was to examine the development of conditional reasoning ability, which characterizes adolescent thinking. The problems consisted of four basic types of conditional reasoning, each type containing five contents of propositions. The four types of problems examined were: affirming the antecedent (AA), denying the consequent (DC), denying the antecedent (DA), and affirming the consequent (AC). The five contents Were: causal, factual, arbitrary, hypothetical, and abstract, inserted in every conditional sentence. Forty problems were used in all, giving two problems to the twenty combinations of four types and five contents. 405 subjects were picked up from 3rd and 5th graders of elementary school, 1st and 3rd graders of junior high school, and 1st year of college, and they made experiment I. Main results were as follows: 1. The four types of problems had distinctly different developmental traits; in AA, 80% of the 3rd graders of elementary school gave the correct responses, and a tendency for increasing ability was shown from this period onward. In DC, about 70% of all subjects gave the correct responses, However, a tendency for increasing ability was not shown. On the other hand, in DA・AC problems, only 30∼40% of the college students performed correctly. Thus, in terms of increasing number of correct responses, the order of problem types was AA, DC, and DA and AC. This finding was similar to previous findings on the development of conditional reasoning ability. 2. The coefficients of correlation between problem types were calculated in each grade. From this analysis, it was found that there were negative correlations between AA and DC, and DA and AC except for college students (-.15∼-.55). This indicated that subjects who were good performers on the former types (AA, DC) were poor performers on the later (DA, AC), and vice versa. But, with college students, these coefficients were positive (.12∼.18). 3. An examination was also made of the contents used for each problem type in order to determine whether some contents were more difficult than others. It was revealed that factual and arbitrary contents were easier than hypothetical and abstract ones, and causal content was easy in AA and DC, but relatively difficult in DA and AC, 4. The error analysis indicated that logically nonsense responses were no longer used by students in the 1st grade of junior high school. Thus, it was suggested that the period of transition from a lower to a higher stage in thinking occurred at about this time. Experiment II was designed to clarify the meaning of error patterns (A-B pattern, A-C pattern) found in experiment I; the A-B pattern indicated a probable inference, and the A-C pattern had been considered as a result of a biconditional inference. Two kinds of training, implication (I) and verbal representation (VR), were given to the Ss. Main results were as follows: 1. With the subjects classified in the A-C pattern, the training effect of VR was nil. But, in the I condition, the correct responses of DA and AC increased significantly. It therefore appeared that these subjects had a reasoning ability, but had mistaken conditional sentences. 2. With the subjects in the A-B pattern, the training effects of VR and I were obtained in DA and AC, and AC, respectively. Thus, it was suggested that this pattern showed an incomplete understanding of representation of conclusions, but it is believed that it was due in part to a misunderstanding of implication.