This paper develops a statistical procedure for estimating the HIV infection rate among needle exchange clients without using any self-reported information. Instead, data are accumulated by following the distribution and return of sequentially labeled syringes and by testing a sample of returns for the presence of HIV-1 proviral DNA using polymerase chain reaction. For each drug injector in the sample, a maximum likelihood change point model is constructed to determine if a statistically significant upward shift in the fraction of needles testing HIV positive is evident, as would occur if the drug injector in question became infected. A second maximum likelihood model is formulated to estimate the HIV incidence rate among needle exchange participants by aggregating the individual change point results. When these methods are applied to the syringe tracking and testing data collected to evaluate the legal needle exchange program in New Haven. Connecticut, the maximum likelihood incidence estimate equals zero, with a 95% confidence interval of 0–10.2 new infections per KM) drug injectors per year. Given these data, we cannot reject the null hypothesis that no new infections have occurred among needle exchange participants between November 1990 and May 1992.