To be evolutionarily stable, sexual signals need to convey information regarding the quality or motivation of individuals. This can be achieved through direct or indirect signal costs that guarantee honest communication or through signal designs that facilitate the assessment of quality. Here, I present the case for an alternative hypothesis: that signalling exposes informative mistakes. Signalling mistakes range from occasional to frequent and from slight distortions of typical signals to grossly atypical signals. Their occurrence may be enhanced by disease or stress, thus revealing individual quality or motivation, and receivers typically respond negatively to them. By this mechanism, honest communication is due to costs of developing resistance to mistakes. Therefore, the hypothesis can function independently of signal design costs, although it can also be enhanced by signal design costs when those increase the occurrence of mistakes. This hypothesis widens the scope of signals expected to be sexually selected and creates new approaches to research in sexual selection and animal communication.