The legal needle exchange in New Haven, Connecticut, was signed into law in July 1990. As part of a rigorous effort to evaluate this program, all distributed syringes receive unique tracking codes, and a sample of returned needles are tested for the presence of HIV proviral DNA via polymerase chain reaction. These data, in conjunction with "back-of-the-envelope" statistical models, allow the estimation of HIV prevalence among those intravenous drug users participating in the needle exchange. We present four new techniques for calculating prevalence estimates: the majority rule cutoff model, the random sharing model, the sharing network model, and a version of Kaplan's "needles that kill" model. To our knowledge, these estimates are the first that attempt to infer HIV prevalence among intravenous drug users via syringe tracking and testing. All four techniques suggest a prevalence of infection of approximately 60%.