The role of affective correspondence between signal and outcome in associative learning was studied in 2 experiments with 4-month-olds. In Experiment 1, an infant-directed (ID) speech segment, categorized by adults as having consoling acoustic properties for infants, was either paired or unpaired with smiling or sad face reinforcers. In other groups, a sad adult-directed (AD) speech segment signaled the smiling or sad faces. Later, all speech segments were tested for their ability to increase infants' visual interest in a checkerboard. Positive summation was observed only when the consoling ID speech segment had signaled the sad face. In Experiment 2, an arousing ID speech segment signaled the smiling or sad face. Positive summation was evident in both conditions. These results show that infants sometimes selectively associate speech segments and faces. Discussion centers on the basis for the selective association.