The purpose of this study is to find the bi-polar adjectives for rural stream landscape evaluation by the semantic differential scale and to suggest the major determinants of visual preference in rural stream landscapes. For this, the bi-polar adjectives for rural stream landscape evaluation was found by the method of the reliability test, and the spatial image was analyzed by the factor analysis. The level of visual preference was measured by slide simulation test, and these data were analyzed by the multiple regression. The major findings of this study can be summarized as follows : 1) Of the bi-polar adjectives expressing psychological and physical characteristics, the hi-polar adjectives which demonstrated reliability and consistency run as follows : Bi-polar adjectives expressing psychological characteristics : 'calm-bustling', 'unfamiliar-familiar', 'still-active','depressing-brisk', 'discomfortable-comfortable', 'suppressed-free', 'lifeless-living', 'quiet-noisy', 'unpleasant-pleasant'. Bi-polar adjectives expressing physical characteristics : 'artificial-natural', 'narrow-wide', 'rocky-not rocky', 'desolate-fertile', 'dirty-clean', 'enclosed-open', 'flat-steep', 'not gravelly-gravelly', 'thicketed-not thicketed', 'not weedy-weedy'. 2) Two factors, the harmony and the movement, were derived from the factor analysis for the psychological variables. Three factors, the naturalness, the rock, and the vegetation, were derived from the factor analysis for the physical variables. 3) Rural stream landscape types were classified into four types by the multi-dimensional scaling method. Type III, IV obtained higher rank of visual preference and type I, II obtained lower. 4) For all types, the factors determining the level of visual preference were found to be the harmony, the naturalness, and the vegetation. The visual preference determinants of rural stream landscape need to be considered in improving or restoring the rural stream landscapes.