<p>This paper investigates the existence of a simplification force in neural networks, which can appear in different forms depending on the perspective, suggesting the presence of multiple simplification forces. As a first approximation, two types are considered: internal and external simplification, realized by internal and external prototypes, respectively. The internal prototype represents the simplest network within given resources, while the external prototype extracts simplified features common to many inputs. The internal prototype is obtained through potentiality minimization, which reduces the number of necessary components, whereas the external prototype is realized through potentiality maximization, aiming to increase the latent representational ability of components by using as many of them as possible. The method was tested on an artificial dataset with a small number of input variables. While all methods could extract the internal prototype, the min–max potentiality method improved generalization by representing input patterns with many different weights. Moreover, this method could detect the prototype even during intermediate learning stages, indicating that external prototype learning is strictly grounded in internal prototype learning.</p>