Purpose: This study aims to explore team learning activities in nursing teams and to test the effect of team composition on team learning to extend conceptually an initial model of team learning and to examine empirically a new model of ambidextrous team learning in nursing. Design/methodology/approach: Quantitative research utilising exploratory and confirmatory factor analyses, and correlation and multiple regression analyses, were used for empirical validation. Findings: Principal component analyses of the team learning activities scale revealed a five-factor model, explaining 78 per cent of the variance on the team-learning scale. Being a nursing team in a community hospital, having high team longevity, and having a high percentage of female nurses explained 33 per cent of team learning. Research limitations/implications: Data aggregation in a cross-sectional design can be criticised for potential biases. However, statistical assumptions for aggregation were met, and the concepts used in this study were clearly formulated at team level. Thus, a valuable instrument is provided for further quantitative research on team learning in nursing. Practical implications: The team learning activities in nursing teams reflected the ambidexterity of teams in modern nursing practice. The findings provide a rationale for managers to create infrastructures that support both productive, as well as developmental learning tasks in teams. Originality/value: The study provides new insights regarding how team learning activities occur in ambidextrous teams in nursing. Contrary to prediction, the results show that team composition has little effect on team learning activities. This is valuable knowledge for researchers, trainers, teams and management in nursing. (Contains 6 tables.)