On the purpose of effective selection of the measuring sites for skin temperature, cluster analysis and principal component analysis were applied to a correlation matrix based on 32 skin temperature measurements. The measurements were taken on 27 healthy unclothed women by the thermography after 2 hr exposures at 22, 25, 28, 31 and 34°C in ambient temperature. The results were chiefly as follows : 1) Several groups of measuring sites, such as hand group, foot group and knee and leg group, which showed the same characteristics could be detected by both cluster and principal component analysis. The results suggested the possibility of decreasing the number of measuring sites, effectively. 2) The first two components obtained accounted for 88.00 and 3.75 % of the total variance, respectively and they were interpreted as the factor indicative of skin temperature level and the factor expressing the contrast of the skin temperature in head and trunk to those in extremity. The mean skin temperature was the best indicator of the first one.