In a model, conflicts of interest between communicating individuals are shown to have an important influence on the cost and form of signals that evolve. Two types of conflict are considered: competition between senders to obtain a response from the receiver, and conflict between the sender and the receiver. The receiver system is modelled as an artificial neural network whose `resistance' to signals is represented as a motivational factor that varies independently of the signal. Biases in the receiver system act as the selective force on signals, causing them to become more costly and conspicuous as the intensity of conflict increases. There is some evidence that competition between senders and sender-receiver conflict may have qualitatively different outcomes. We give examples of some situations to which the model might be applied and point out some predictions that could be tested empirically.