Abstract Signal reliability is a major focus of animal communication research. Aggressive signals are ideal for measuring signal reliability because the signal referent – attack or no attack – can be measured unambiguously. Signals of aggressive intent occur at elevated rates in aggressive contexts, predict subsequent aggression by the signaler, and elicit appropriate responses from receivers. We tested the ‘predictive criterion’ in smooth‐billed anis, Crotophaga ani , by broadcasting one of two playback types (‘ahnee’ calls only or ‘ahnee + hoot’ calls), presenting a taxidermic mount, and observing the animals’ behavior. Based on the hypotheses that ‘hoot’ calls and ‘throat‐inflation’ displays signal aggressive intent, we predicted that they would be associated with attack, and that signaling rate would increase over the time period leading up to an attack. Indeed, both hoots and throat‐inflation displays reliably predicted attack. The second prediction, that signaling rate increases in the time leading up to attack, was strongly supported for throat‐inflation displays, which increased over the pre‐attack period in both treatments. Hoots increased over the pre‐attack period in ahnee playbacks but not in ahnee + hoot playbacks. Hierarchical signaling systems are characterized by early, less‐reliable predictors of attack, and later, more reliable predictors of attack. During both natural and simulated interactions, the more‐reliable throat‐inflation display tended to precede the less‐reliable hoot call, suggesting that this signaling system is not hierarchical. In a comparison of 11 putative signals of aggressive intent in birds, the throat‐inflation display had the second highest mutual information (reduction in uncertainty) among visual signals and non‐passerine signals while hoots had below‐average mutual information. Natural observations indicate that both hoots and throat‐inflation displays occur in the context of aggressive between‐group encounters, and hoots also occur during within‐group interactions. Throat‐inflation displays appear to be reliable indicators of aggressive intent, but the function of hoot calls is less clear.