The capacity to experience guilt is assumed to benefit individuals, as the rewards of repeated, cooperative interactions are likely to exceed the rewards of acting selfishly. If that assumption is true, the extent to which people experience guilt over interpersonal transgressions should at least partly depend on the utility of another person for the attainment of personal goal(s) through social interaction (relational utility). Three experiments confirmed the relational utility hypothesis by showing that people felt guiltier (a) over excluding someone from a fun game if this person could subsequently distribute more money in a dictator game, (b) over hypothetical social transgressions toward a person who was instrumental to the attainment of a salient goal than toward a person who was not instrumental to the attainment of that goal and toward the same person when no goal was salient, and (c) over a low contribution in a social dilemma game if they were more dependent on their group members for performing well in a subsequent debating contest. Closeness with the other person, differences in severity of the transgression, and strategic motives for expressing guilt were consistently excluded as alternative accounts of the effects. By showing that relational utility may affect guilt, these findings (a) provide support for the individual level function of guilt; (b) extend research on the antecedents of guilt in social interactions, which mainly focused on retrospective appraisals; and (c) bear implications for the status of guilt as a moral emotion.