This article discusses the challenges of estimating levels and patterns of heroin use in a setting where there were no official records. Ethnographic fieldwork, carried out in a Kenyan Coastal town, utilised a range of qualitative research methods in an attempt to estimate numbers of male and female users and the proportion of them who were injectors of heroin. In the town of at least 85 000 people, it was estimated that there were perhaps about 600 heroin users, of whom about 30 were women. The ratio of male to female users was estimated to be 20 : 1. Fifty per cent of users in the town were estimated to be injectors of heroin. They were found to have poor injecting techniques, to share equipment from time to time and to have low awareness of the link between injecting drug use and HIV infection. An urgent need for harm reduction strategies was identified.This article reports on how levels and patterns of heroin use were estimated in a Kenyan town where there are no available 'official' records regarding drug deaths, arrests or of users in services. The research was carried out at the beginning of a study into the lives of women heroin users focussing on risk strategies and reproductive health. In this article we report how the population of heroin users in the town and the ratio of male to female users were estimated. In addition, we describe how patterns of heroin use, either by injection or smoking, were identified and the estimated proportion of injectors.Heroin use in Kenya, as elsewhere, is illegal, criminalised, stigmatised and therefore, hidden. As such, its study poses particular challenges for researchers. Indeed, the difficulties of researching hidden populations of illicit drug users are well known (Stimson and Choopanya, 1998;Lambert, 1990). Researchers, usually working in European or North American cities, have developed methods to count populations of illicit drug users and to validate the findings by cross referencing to other data sources (Hartnoll et al., 1985;Domingo-Salvany, 1998). Yet, there remains uncertainty as to