While self-assessments of welfare have become popular for measuring poverty and estimating welfare effects, the methods can be deceptive given systematic heterogeneity in respondents' scales. Little is known about this problem. This study uses specially-designed surveys in three countries, Tajikistan, Guatemala, and Tanzania, to study scale heterogeneity. Respondents were asked to score stylized vignettes, as well as their own household. Diverse scales are in evidence, casting considerable doubt on the meaning of widely-used summary measures such as subjective poverty rates. Nonetheless, under the identifying assumptions of the study, only small biases are induced in the coefficients on widely-used regressors for subjective poverty and welfare.