--- - |2+ Abstract Objective Compassion has been associated with eudaimonia and pro‐social behavior, and it has been regarded as a virtue, both historically and cross‐culturally. However, the psychological study of compassion has been limited to laboratory settings and/or standard survey assessments. Here, we use an experience sampling method (ESM) to compare naturalistic assessments of compassion with standard assessments, and to examine compassion, its variability, and associations with eudaimonia and pro‐social behavior. Method Undergraduate students (n = 200) took a survey that included standard assessments of compassion and eudaimonia. Then, over 4 days, they were repeatedly asked about their level of compassion, eudaimonia, and situational factors within the moments of daily life. Finally, pro‐social behavior was tested using the Dual Gamble Task and an opportunity to donate task winnings. Results Analyses revealed within‐person associations between ESM compassion and eudaimonia. ESM compassion also predicted eudaimonia at the next ESM time point. While not impervious to situational factors, considerable consistency was observed in ESM compassion in comparison with eudaimonia. Further, ESM compassion along with eudaimonia predicted donating behavior. Standard assessments did not. Conclusions Consistent with virtue theory, some individuals' reports were indicative of a probabilistic tendency toward compassion, and ESM compassion predicted ESM eudaimonia and pro‐social behavior toward those in need. - Journal of Personality, EarlyView.