The measurement of psychological constructs depends upon clear definitions and demonstrated relationships among items, scales, and relevant theories. In this study, we examined these relationships for the measurement of affect as reflected in the popular Bradburn Affect Balance Scale (ABS). Results of confirmatory analyses of data from 187 older Canadian adults (Mean age, 69.7 years) showed that a two-dimensional structure fitted the item data much better than the unidimensional, bipolar model suggested by the original scoring key. The two dimensions showed parallel patterns of correlations with two measures of morale and with a measure of social desirability, but these patterns differed from that of the conventional unidimensional ABS score. Results suggested that the ABS should be used with caution and with scoring for two distinguishable dimensions.