To investigate whether and how structured feedback sessions can increase rates of appropriate antimicrobial prescribing by junior doctors. This was a mixed-methods study, with a conceptual orientation towards complexity and systems thinking. Fourteen junior doctors, in their first year of training, were randomized to intervention (feedback) and 21 to control (routine practice) groups in a single UK teaching hospital. Feedback on their antimicrobial prescribing was given, in writing and via group sessions. Pharmacists assessed the appropriateness of all new antimicrobial prescriptions 2 days per week for 6 months (46 days). The mean normalized prescribing rates of suboptimal to all prescribing were compared between groups using the t-test. Thematic analysis of qualitative interviews with 10 participants investigated whether and how the intervention had impact. Data were collected on 204 prescriptions for 166 patients. For the intervention group, the mean normalized rate of suboptimal to all prescribing was 0.32aEuroS +/- aEuroS0.36; for the control group, it was 0.68aEuroS +/- aEuroS0.36. The normalized rates of suboptimal prescribing were significantly different between the groups (PaEuroS=aEuroS0.0005). The qualitative data showed that individuals' prescribing behaviour was influenced by a complex series of dynamic interactions between individual and social variables, such as interplay between personal knowledge and the expectations of others. The feedback intervention increased appropriate prescribing by acting as a positive stimulus within a complex network of behavioural influences. Prescribing behaviour is adaptive and can be positively influenced by structured feedback. Changing doctors' perceptions of acceptable, typical and best practice could reduce suboptimal antimicrobial prescribing.