Most vector-borne zoonotic pathogens are transmitted among several host species, but different species vary considerably in their importance to pathogen transmission, at least partially because they vary in their propensity to infect feeding vectors. This propensity is often called realized reservoir competence. Realized reservoir competence is the product of 1) the probability the individual host is infected, i.e., infection prevalence, and 2) the probability that if the host is infected, it will transmit the infection to a feeding vector, or infectivity. Prevalence varies in space and time, whereas infectivity may be a property of the host species. Both prevalence and infectivity are ecologically and epidemiologically important, but measuring them simultaneously is difficult. We present a probabilistic model that separately estimates host infection prevalence and infectivity from data on the infection status of vectors collected from individual hosts, data generally used to measure realized reservoir competence. We then consider how imperfect diagnostic tests (i.e., false negatives and positives) influence these probabilities—estimates of prevalence and infectivity are fairly robust to false negatives, but not to false positives. We thus extend the model to estimate the rate of false positives in order to improve estimates of prevalence and infectivity. We illustrate these methods by reanalyzing data from LoGiudice et al. (2003; Proc. Natl. Acad. Sci. U.S.A. 100: 567–571) on the reservoir competence of ten vertebrate hosts of Borrelia burgdorferi, the agent of Lyme disease. We find that these vertebrate hosts vary both in prevalence and infectivity and that both values are highly, positively correlated among species.