抄録
The apparent similarities (or dissimilarities) of all possible pairs of twenty-four English consonants were estimated by two independent panels of eighteen Japanese and twenty-four American listeners by using a psychophysical method of magnitude estimation. The normalized responses were analyzed by a statistical technique of cluster analysis (SPSS, V-6: RMODE, AVERAGE DISTANCE METHODE), the program of which was made available through the Computer Center of the University of Tokyo. The results of analysis were obtained in the forms of tree-diagrams and correlation matrices. The apparent interconsonantal differences of all possible pairs of consonants were examined by observing these diagrams and matrices. The results suggested a strong tendency of the listeners to be affected by the manners of articulation in determining the apparent similarities among the consonants they heard. A closer look at the results may give some valuable information to predict the systems of distinctive features for both American and Japanese sound systems.
