In this paper, two architectures of neural networks have been used to solve diagnosis problems. As a case study, they are used to diagnose 9 eye diseases knowing the symptoms common between the diseases and the individual signs of each one. The first architecture involves the application of supervised learning artificial neural network based on the back-propagation learning rule to the problem. The second architecture involves a combination of supervised learning rule (Widrow-Hoff) and a competitive learning rule (Kohonen) in a counter propagation network. A modification in the Kohonen learning has proved to overcome the problems encountered in the learning of counter propagation network. Both architectures proved to be capable of classifying the diseases. The counter propagation network is faster in learning and it gives more accurate results. But, it requires more accuracy in choosing the learning rules.