[Schizophrenia and the theory of neural networks]

Artificial neuronal networks increasingly serve the interpretation of neuro-biological problems. Typically, neuronal networks store "associations". By a "store-matrix", which works like holography because it sums up many memories, a new impression is assigned to the most similar memory satisfying the criterion of minimal Hamming distance. Schizophrenia-like reactions arise in the case the neuronal network is overloaded with too many memories. Aside from the fascinating power of the model it needs to be criticized and possibly improved. "Ball" is both a spherical object and a dance. Although the semantical levels are at a great distance from each other, the Hamming distance of the binary coded chains of symbols "ball"-"ball" is zero. The investigation of schizophrenic disturbances shows that a few of them-inversing the model-assumptions-prefer associations by using the criterion of minimal Hamming distance.