In this research, a computational model that simulates the mental function of multicolor aesthetic evaluation through backpropagation neural networks was constructed. Helmut Leder's psychological model served as the theoretical framework. We determined the macro‐architecture of the computational model through two psychological experiments using the semantic differential (SD) method. The aesthetic score of a multicolor stimulus is defined as the inverse of its factor score on the factor “Pleasure” extracted in the first experiment, and each of the three factors extracted in the second experiment—i.e., “Stability,” “Heaviness,” and “Presence”—is regarded as a simple perceptual feature. The genetic algorithm was then employed to optimize the hidden layer node number, the learning rate, and the momentum constant of each neural network. In two simulation tests, this computational model exhibited some predictive power, implying that the model can be regarded as a relatively successful approximation of the psychological mechanism of multicolor aesthetic evaluation. In addition, the results of the second simulation also show that the perceptual feature “Heaviness” possesses the principal impact on the aesthetic evaluation of multicolor objects, whereas the other two perceptual features “Stability” and “Presence” have a minor influence. The heavier and/or more stable a multicolor object is perceived to be, the less aesthetically pleasing it is. Conversely, the stronger the sense of matter presence a multicolor object elicits, the more aesthetically appealing it is. © 2016 Wiley Periodicals, Inc. Col Res Appl, 42, 216–235, 2017