Teaching entails many demands of an emotional and interpersonal kind. For the current study, emotional job demands were conceptualised as comprising three components: exposure to emotionally demanding situations, emotional labour (use of deep and surface acting) and work focused on the emotional well-being of others. Both emotional job demands and ‘non-emotional’ job demands (that is, general demands such as those pertaining to workload, time constraints and curriculum issues) were hypothesised to predict emotional exhaustion. Two resources, social support and confidence in one’s own teaching practices (teaching self-efficacy [TSE]), were expected to have main and buffering effects. Primary school teachers (N = 556) completed an electronic questionnaire measuring all study variables. Consistent with hypotheses, general (non-emotional) job demands, emotional demands, social support and TSE, each uniquely predicted exhaustion. In addition, TSE buffered the negative effect of deep acting on emotional exhaust...