In this paper we explore the nature and causes of individual differences in the use of feeling thermometer items to evaluate social groups. Some respondents have a general tendency to assign higher overall temperature readings and some tend to use a wider range of the temperature scale. These tendencies are partially understandable in terms of the respondent's evaluations of socil groups, but other important predictors are identified. Methods of adjusting feeling thermometer scores to account for individual differences are explored, using data from the 1984 American National Election Study. The conclusion is that some sort of adjustment is probably useful for those studying supporters ofconservative groups, though such a procedure may not be needed for those who focus on liberal groups.