Syringe-exchange attenders (n = 86) and a comparison group of injectors not attending exchanges (n = 121) were recruited and interviewed in 1989 to examine HIV risk behaviour and HIV antibody status. In 1990, 59% of attenders and 50% of non-attenders were re-contacted and re-interviewed to examine changes over time. Re-interviewed subjects did not differ demographically or in drug use from those not re-interviewed. At first interview, all were injecting, and 35% of attenders and 42% of non-attenders reported recent syringe-sharing. One year later, levels of HIV-risk were similar for both groups, but risk reduction had been achieved by different routes. At Time 2, 16% of attenders were injecting and sharing (IS), 70% injecting and not sharing (INS), and 14% not injecting (NI), compared to non-attenders, 22% IS, 47% INS and 31% NI. Attenders were more likely to continue injecting (86%) with low sharing (18% of those still injecting). Non-attenders were less likely to continue injecting (69%), but continuing injectors had higher sharing rates (32% of those still injecting). More non-attenders than attenders stopped injecting, but non-attenders who continued to inject were at a higher risk of HIV infection than attenders who continued to inject. Baseline HIV prevalence was 2.5% and there were no sero conversions. The levels of sharing for both groups were lower than levels in the UK before HIV prevention programmes commenced and are lower than 1990 sharing rates in the USA. The study indicates the importance of tracking individual change over time, and of measuring outcomes additional to changes in sharing.