We propose a novel theoretical and empirical approach to studying group-level social functions of emotions and use it to make new predictions about social consequences of gratitude. Here, we document the witnessing effect: In social groups, emotional expressions are often observed by third-party witnesses-family members, coworkers, friends, and neighbors. Emotional expressions coordinate group living by changing third-party witnesses' behavior toward first-party emotion expressers and toward second-party people to whom emotion is expressed. In 8 experiments (N = 1,817), we test this for gratitude, hypothesizing that third-party witnesses will be more helpful and affiliative toward a first party who expressed gratitude to a second party, as well as toward the second party, and why. In Experiments 1-3, participants who witnessed a "thank you" in 1 line of text, expressed to someone who previously helped the grateful person, were themselves more helpful toward the grateful person. In Experiment 4, witnesses of gratitude expressed to someone else via video recording subsequently self-disclosed more to the grateful person, and in Experiment 5 wanted to affiliate more with the grateful person and with the person toward whom gratitude was expressed. Experiments 6-8 used within-subjects designs to test hypothesized behavioral and social-perceptual mechanisms for these effects, with videos of real gratitude expressions. Gratitude may help build multiple relationships within a social network directly and simultaneously. By specifying proximal interpersonal mechanisms for reverberating consequences of 1 person's communicated emotion, the present work advances theory on the group-level functions of emotions. (PsycInfo Database Record (c) 2020 APA, all rights reserved).