Objectives: Youth mentoring involves a supportive relationship between a youth and a volunteer adult with the aim of promoting a range of positive developmental outcomes. The predominant paradigm in mentoring research involves large-N, pre- post designs. The present study examined a small number of child-mentor pairs over time in order to gain a more detailed understanding of patterns of change. Method: A multiple case, longitudinal design was used, looking in depth at four children, aged 9 to 10 years. Data were collected at regular intervals over the first 15 months of the mentoring relationship from the child, their mother, and their mentor, using self-report measures of key domains of psychosocial functioning. The quality of the mentoring relationship was also measured. Results: At baseline, all four children were functioning in the clinical range in at least one domain. Two of the children reported modest improvement over the course of the mentoring relationship, with reliable changes occurring mostly in the behavioural domains; however, this improvement was not reflected in the data provided by the parent. The scores for the other two children, as reported both by themselves and their mothers, were relatively stable over the 15 month period. Conclusions: The wider literature suggests that more consistent improvements would have been expected; it may be that 15 months was not long enough for the influence of mentoring to be borne. A multiple case design offers a detailed view of patterns of change, providing the opportunity to identify potential moderating factors for particular individuals, which could improve the theoretical understanding of mentoring as well as its practical application. Further research using this paradigm could help define the role that mentoring should play within the helping services and which children would benefit most.