The present paper analyzes the semantic features of positive and negative emotion names. In a pretest the 12 most typical positive rated emotion names were selected by means of a free listing of 70 Ss. These 12 items were presented to 42 Ss as stimulus words for a free association test. From this association data overlap coefficients were computed and analyzed by means of nonmetric multidimensional scaling. It was possible to fit the features "Soziale Nähe-Soziale Distanz" (social distance) und "Körperhaftigkeit" (Quality of body feeling) in the two-dimensional euclidian solution. In the same way as described above, an experiment was carried out with 11 negative emotion names. In this case it was possible to fit in the same two features and the feature "Aktivation" (activation).