With the development of environmental psychology, many impression evaluation experiments using semantic differential method were being conducted. However, research question rose that there could be a system to organize those words, or selected basic words that could be used throughout those studies.

 This research aimed to organize impression evaluative words used in interior evaluation. First, words were gathered from previous researches, bibliographic research and interior related websites. Keywords such as “impressive evaluation” and “atmospheric evaluation” were used to search previous researches, and about 40 studies were gathered, but only one of related researches were listed in reference since the aim is to pick up the evaluative words. Words were coordinated in following ways. In a monopolar case of semantic differential method, such as natural vs. unnatural, positive evaluative word was picked up, and in a bipolar case, such as natural vs. artificial, both evaluative words were picked up. When gathering the similar words, obvious negative nuance words were avoided. 85 words from previous researches, 104 words from bibliographic research and 49 words from interior related websites were taken. After gathering, words were divided into 7 groups; subjective words, ambiguous words, stylistic words, space-limited words, behavioral words, non-visual words and unambiguous words. Within those groups, unambiguous words (91 words) were targeted for organize.

 Next, impression evaluation experiment was conducted targeting variation of toilet spaces of retail buildings. 15 subjects were asked to evaluate the space using 4 rating scale of suitability of 91 evaluative words. Using the results, multi-dimensional correspondence analysis was conducted. The contribution rate shows that, 83% can be explained by 3 dimensions, 88% by 4 dimensions, and 91% by 5 dimensions, which harmonizes with the results of many previous studies that results in 3 to 5 factors of factor analysis. Using those results of factor analysis of the previous studies, correlation analysis was done to see the similarity of the plotting behavior. In most cases, correspondence was seen with one of the factors, which showed the validity of the system. Close comparison was done with the language image scale proposed by Kobayashi, and the test result of crosstabs showed the validity of 4 dimensions. By the plotting of the words, 4 dimensions were named accordingly; warmness axis, activeness axis, heaviness axis and gender axis. Cluster analysis was done using 4 dimension values, resulting in 22 clusters. In order to characterize the plotting, each axis was divided into three areas, and each cluster was plotted according to the coordinate of the average of the cluster. Closest word to the average was selected as the representative word of the cluster. Using 4 dimension, better explanations of the impression evaluative words became possible.

 This research aimed to organize the impression evaluative words used. Through bibliographic research, impression evaluative words were gathered and grouped. Unambiguous words were targeted since the aim was to characterize the impression of the space. Through impression evaluation experiment and its analysis, 4 axis was extracted. As a result, 8 adjectives (warm, cool, active, calm, heavy, light, masculine, feminine) could be the possible sets of words that could characterize the impression of the space. Further study that relates spatial components with the impression axis could be done, which leads to design aid of creating aimed impression.