The need for effective interventions to reduce sexual risk behaviour and prevent HIV among adolescents remains critical. In South Africa, which is enduring the world’s most severe HIV epidemic, approximately 30% of new infections occur among 15-24 year-olds. Prevention programmes delivered in schools are seen as an important strategy to increase knowledge and effect self-reported behaviours. The effectiveness of such programmes is often evaluated in randomised controlled trials, but process evaluations also play an essential role. Process evaluations can add understanding and depth to trial findings and advise on how the intervention can be adapted to enhance its likely effectiveness. This thesis presents results from a process evaluation of a split-sex, discussion-based curriculum delivered in 46 South African schools by local facilitators called “coaches”. It was implemented by the NGO Grassroot Soccer, and used sport themes, activities and metaphors in its activities. The curriculum primarily focused on reducing multiple sexual partners, age disparate sexual relationships and gender-based violence, and was one of two curriculums evaluated in the GOAL Trial, an RCT conducted amongst 4,500 high school participants in Cape Town and Port Elizabeth. The trial showed that the intervention was not effective in achieving its primary behavioural objectives, despite demonstrating increases in knowledge. Four key domains were assessed in this study; 1) the fidelity and quality of implementation, 2) facilitator training and experiences delivering the intervention, 3) participant engagement with, and perceived relevance of the programme, and 4) contextual factors affecting delivery and use of the programme. Data collection within these four domains consisted of: structured observations (n=49) of the delivery of intervention sessions, unstructured observations of the week-long training of facilitators in both sites, focus group discussions (n=3) with twenty-five facilitators, and an additional fifteen focus group discussions with students (n=99). The study found that the intervention was implemented with varying degrees of fidelity across sessions; that the intervention was initially received more favourably during training by female coaches compared to males, and that coaches engaged variably with aspects of the programme; and finally, the programme was perceived slightly differently between different groups of participants, but generally students appreciated the split-sex, discussion-based approach and found the topics relevant. Many said, however, that the programme needed to be longer, include more discussion around certain topics, and allow more dialogue between males and females. Based on findings emerging from this process evaluation, I suggest that the results of the GOAL Trial can be attributed, in part, to a combination of limitations in how the intervention was able to be implemented in practice, and an incomplete understanding of the functioning of intervention components, which lead to weaknesses in the theory underpinning the intervention.