The recticular formation as a multi-valued logic neural network

A description is presented of the first of several multivalued logic neural network simulations undertaken by the author. He demonstrates a supervised trial-and-error learning scheme with motivation modulation meant to represent the instrumental conditioning which is proposed to take place in the reticular formation. As such, it is the second layer in the nonmammalian pattern-recognition system where the first layer (the tectum) learns without supervision to associate each new pattern with randomly chosen output line. This line feeds into the reticular formation values proportional to the degree of pattern recognition. The reticular formation in turn learns to associate each activated tectal line with the appropriate motivation-triggered action. As a multivalued logic neural network, it revives and extends the logic-based version of W.S. McCulloch and W.H. Pitts (1943) and rests on a foundation completely different from the mathematically based systems neural networks