We examine independence in retrospective evaluations of positive and negative affect (PA and NA, respectively). Study 1 shows that the estimation of weekly PA and NA frequencies depends on the congruent daily affects but not on the incongruent daily affects (i.e., weekly PA [NA] depends on the daily PAs [NAs], but not on the daily NAs [PAs]). Study 2 replicates and extends the results of Study 1 by using a within- and between-subjects design as well as both frequency and intensity measures of affect, and by also examining the relationship between momentary and daily affect. Study 3 shows that daily PA and NA are affected by the existence of subjectively defined congruent peak affect and less by subjectively defined incongruent peak affect. We suggest a focusing bias account, according to which asking about positive [negative] affect focuses attention on positive [negative] experiences.