Recent years have seen a burgeoning of intergenerational programmes aiming to improve the well being of participants. However, very few programmes have been formally evaluated. In this paper we report results from a randomised controlled trial of a school-based intergenerational intervention undertaken in Brazil. Randomly selected samples of 253 adolescents and 266 elders aged 60 and over resident in the school's catchment area were administered a questionnaire including questions on cognitive components of social capital, family relationships, and self-rated health. Participants were then randomly allocated to control and intervention groups. The intervention comprised participation in a 4 month programme of intergenerational activities in which the elders shared their memories with the students. At the end of the intervention the questionnaire was re-administered to the samples. High proportions (85-95%) of both samples completed the questionnaires but compliance with the intervention was low in the elderly group. In the analysis of results from the elderly sample, we therefore compared the control group with the group assigned to the intervention (intention to treat analysis). Results showed that adolescents in the intervention group were nearly three times more likely to rate their health as good than those in the control group, but also more likely to judge that most people were selfish. In the elderly sample, those from the intervention group were over twice as likely as those from the control group to report positively on the helpfulness of neighbours; judge most people to be honest or consider their family relationships as good. There were no significant differences between groups on other outcome measures. These results suggest that structured intergenerational activities may have positive effects on some aspects of social capital for both adolescents and elderly people, although further research is needed to elucidate the processes involved and the extent to which the findings are context specific. This study is the first to use a controlled trial design to evaluate this type of intervention and is valuable in showing that this design can be used in interventions of this kind, and also the difficulties involved. (C) 2007 Elsevier Ltd. All rights reserved.