Legal abortions are authorized medical procedures, and as such, they are or can be recorded at the health facility where they are performed. The incidence of illegal, often unsafe, induced abortion has to be estimated, however. In the literature, no fewer than eight methods have been used to estimate the frequency of induced abortion: the "illegal abortion provider survey," the "complications statistics" approach, the "mortality statistics" approach, self-reporting techniques, prospective studies, the "residual" method, anonymous third party reports, and experts' estimates. This article describes the methodological requirements of each of these methods and discusses their biases. Empirical records for each method are reviewed, with particular attention paid to the contexts in which the method has been employed successfully. Finally, the choice of an appropriate method of estimation is discussed, depending on the context in which it is to be applied and on the goal of the estimation effort.