This study examined the nature of learning within an afterschool mathematics club established by the South African Numeracy Chair project. The study sought to establish what sort of progress in mathematical learning occurred in a grade 3 afterschool maths club, using assessment instruments associated with the Learning Framework in Number. The study also sought to understand the nature and effects of mentor mediation in the maths club, using Vygotsky’s notion of the Zone of Proximal Development (ZPD) together with the notion and practice of scaffolding. The study made use of a variety of data collection techniques, including one-to-one assessment interviews, task-based interviews and observations. In line with the case study approach adopted, four learners were selected for interviews. The assessment interview results revealed that, in terms of proficiency in early arithmetical learning, all four learners showed progress after spending four months in an afterschool maths club. Even though they were found to have advanced in their Strategies for Early Arithmetic Learning (SEAL), some of them were observed still using their fingers to support their counting. Such strategies were likely to mirror the teaching approaches used in their usual school mathematical lessons. The overall findings in terms of learners’ proficiency and progress give rise to concerns about current number teaching practices in their school, which emphasize the standard written algorithm in the lower primary grades. The study also made use of Vygotsky’s notion of the ZPD to analyse the nature of mentor-peer mediation. Witnessing the learners’ use of trial and error strategies during the task-based interview allowed both mentors to support learners through understanding their thinking, prompting them and encouraging them to reflect on their answers and develop more effective strategies. Learners progressed through the ZPD at different paces and in different ways, with ‘aha’ moments happening at different points for individual learners. Their progression in the ZPD seemed to depend on interaction among all participants, which varied according to what was contributed and what requested by each participant. The findings revealed that although there was evidence of learners achieving success at the tasks in task based interviews there were also there were also some learners who experienced difficulties. Additionally, in order to argue that learning was fully realised within the ZPD would have required follow up task based interviews to assess the extent to which learners were able to complete the tasks independently without the scaffolding of mentors. This was not possible within the scope of this research but would be useful in future research.