On the basis of our experiments and simulations on pattern recognition and learning, we have extended the previous four reactor network to eight reactors which are electrically coupled via Pt-working electrodes in the fashion of a Hopfield network. This extension considerably improves the recognition processes and allows to encode three reactor patterns. Since each of the eight reactors can be either in a periodic (P) or a nodal (N) steady state using the Belousov-Zhabotinsky (BZ) reaction, there are 256 (=2(8)) dynamical patterns of which any three patterns may be encoded in the reactor net. We describe the recognition processes that successfully associate some of the remaining 253 patterns (including mirror images) as initial patterns to one of the encoded patterns to which it has the least number of errors. The advantages and limitations of electrical coupling versus mass coupling are discussed. Numerical simulations using the seven-variable Montanator by Gyorgyi and Field are in agreement with the experiments.