Background: Paired reading (PR) is an application of peer tutoring. It has been extensively researched, and its efficacy across a range of outcomes has been established. Benefits include improvements in key reading skills, and also in affective aspects of learning. Several studies have shown gains in self-esteem, although measurement methods have varied, and the model of self-esteem has rarely been clearly articulated. Aims: To investigate the changes in self-esteem of children participating in a randomized trial of PR over a 15-week treatment period. To investigate the relative contribution of self-worth and self-competence to any gains in self-esteem. To investigate whether the pattern of change differs in children who take on different roles in the PR process. Participants: The participants comprised a subset of a large-scale randomized trial of peer learning (The Fife Peer Learning Project). Four schools were randomly selected from schools allocated to the same-age PR condition, and four schools from those allocated to the cross-age PR condition. The same-age group consisted of 87 primary 6 children (10-11 years old). The cross-age group consisted of 81 primary 6 children. The controls, from schools randomly selected from a neighbouring authority, consisted of 92 primary 6 children. Method: A pre-post design employing self-report measures of self-esteem. Rosenberg's Self-Esteem Scale was used, with scores analysed for worth and competence. The treatment period was 15 weeks, with the participants following a prescribed PR process. Results: Significant pre-post gains were noted in self-esteem, driven predominantly by improved beliefs about competence, in both same-age and cross-age conditions, but not for controls. Gains were also seen in self-worth in the cross-age condition. Further analyses of the influence of organizational condition (same-age or cross-age) and role played (tutor vs. tutee) showed significant differences between same-age tutors and cross-age tutors in relation to self-worth. Effect sizes were generally small or moderate. Conclusions: The findings provide further support for the belief that PR can enhance self-esteem. Importantly, the use of a two-dimensional model provides extra information about self-perceptions in PR contexts: first, the central role of self-competence; and second, the gains in self-worth which are associated with tutoring younger children (but not same-age peers). This new information has educational significance for schools considering the potential of peer tutoring and the benefits of different organizational conditions.