This paper proposes an evolutionary design methodology of multilayer feedforward neural networks based on the constructive approach. The authors elaborate an adjustable processing element as a primitive neuron model. The neural layer can be constructed by assembling several neurons. The multilayer neural network can be finally constructed through cascading several neural layers. The constructive approach facilitates substantially the extraction of design specifications from a multilayer neural network. Based on the constructive representation of multilayer feedforward neural networks, a genetic encoding method is used, after which the evolution process is elaborated for designing the optimal neural network. The results of these experiments reveal that this methodology is superior to the error back-propagation algorithm both for its executing efficiency and performance.