Cleveland et al. (2005, Environ. Biosafety Res.4: 197–208) offer useful suggestions for monitoring transgenes in landraces of maize, but we disagree with their statement that the scientific conclusions of our paper (Ortiz-García et al., 2005, Proc. Natl. Acad. Sci. USA102: 12338–12343) are not justified. First, contrary to their perception, our survey was not designed to evaluate transgenes in the whole State of Oaxaca, but rather to monitor a specific portion of the District of Ixtlán de Juárez where the presence of transgenes had been reported previously by Quist and Chapela (2001, Nature414: 541–543). Second, our paper described two methods for estimating frequencies of undetected transgenic seeds, while Cleveland et al. recommend a third approach that explicitly estimates effective population size. They argue that the effective population size of our seed samples is smaller than we assumed, leading to false claims about our detection accuracy. However, we employed a robust statistical approach to compensate for possible bias by using numbers of maternal plants, in addition to numbers of seeds, to provide a conservative estimate of the minimum number of independent samples. When we re-analyzed our 2004 data using effective population sizes, our conclusion that transgenic seeds were “absent or extremely rare” did not change, nor did the general range of possible frequencies of undetected transgenic seeds. Unlike Cleveland et al., we advocate using combined probability tests to analyze data across localities. Third, our critics argue that we accepted the null hypothesis that transgenes were absent. Actually, we assumed that transgenes were present in local landraces, and we used parameter estimation methods to calculate the probability of failing to detect transgenic individuals at a range of frequencies. In agreement with Cleveland et al., we reiterate that there is a clear need for additional surveys with rigorous sampling methods to provide estimates of transgene frequencies over broad geographic areas in Mexico.