In this thesis, I evaluate how human receivers respond to signallers displaying a two-component signal, when each component differs in its probability of being associated with a binary outcome (desirable/undesirable). I conducted tests under a broad range of conditions under which neither, one or both multicomponent signals were predicted to be attended to according to a simple signal detection model. I also considered a complementary modelling approach that predicts the same long-term response but uses exploration/exploitation theory to identify the optimal tradeoff between learning more about the nature of the signaller and using current information to reject it. I found that receivers frequently attended to both forms of signal even under conditions when they were not predicted to do so by the signal detection model. Exploration/exploitation was more successful in accounting for the observed behaviour and provides a promising starting point for future work on multicomponent signal evolution.