The purpose of this paper is to outline the fundamental concept of wavelet networks (WNs) and to demonstrate its specific advantages in a clinical discrimination task. A group of 25 boys with attention deficit hyperactivity disorder (ADHD) should be separated from a control group of 25 healthy boys by auditory evoked potentials. Because of the high variability of the recorded EP signals quantification of the averaged sweeps by peak latencies and amplitudes failed. However, with wavelet networks a maximum classification rate of 80% was achieved by crossvalidation. A WN is basically described as a multilayer perceptron which consists of two parts for feature extraction/parametrization and classification. These essential steps of a pattern recognition task are not separated in different tasks but linked together by the clamp of the learning algorithm. Because no user interaction is necessary we call this procedure a self-learning method.