The functions of two visual signals of the mandrill [silent bared-teeth face (SBTF) and crest-raise (CR)] were investigated by quantifying the probability of each signal occurring across a set of distinct contexts. The motivation for the investigation was twofold: (1) SBTF had been interpreted in diverse and sometimes contradictory ways, and (2) CR had been interpreted as a distinct signal from SBTF, despite indications that the two signals grade into one another. In our investigation we considered four functions for both SBTF and CR (threat, submissive, conciliatory, and ambivalent), and we made specific predictions about the relative probability that a signal with each of these functions should occur in different contexts. To determine if SBTF and CR represent a single graded signal, we analyzed them separately and together. We predicted that if they represent a single graded signal, then they should exhibit similar patterns of occurrence across contexts when they were analyzed separately, and that these patterns should be strengthened when they were analyzed together. The results showed that both SBTF and CR met the predictions for conciliatory signals, occurring most often in non-aggressive, non-hostile contexts and least often in aggressive, hostile contexts. The results were also consistent with the hypothesis that SBTF and CR represent a single graded signal.