Summary In social interactions among mammals, individuals are recognized by olfactory cues, but identifying the key signals among thousands of compounds remains a major challenge. To address this need, we developed a new technique, component-activity matching (CAM), to select candidate ligands that "explain" patterns of bioactivity across diverse complex mixtures. Using mouse urine from eight different sexes and strains, we identified 23 components to explain firing rates in seven of eight functional classes of vomeronasal sensory neurons. Focusing on a class of neurons selective for females, we identified a novel family of vomeronasal ligands, steroid carboxylic acids. These ligands accounted for much of the neuronal activity of urine from some female strains, were necessary for normal levels of male investigatory behavior of female scents, and were sufficient to trigger mounting behavior. CAM represents the first step toward an exhaustive characterization of the molecular cues for natural behavior in a mammalian olfactory system.