Pattern feedback is a new concept in learning networks of adaptive microcircuits. The net is trained to produce a model output pattern. During use, the output patterns are fed back to the input. Recognition of an input pattern takes place if a stable model output pattern is generated.