In this paper we present the data of an experimental study on semantic space, which is based on models of neural networks. The neural network models implemented in the experiment use the evoked potentials of the human brain as the basis of their work. In experiment we showed two movies for 44 subjects. The first 22 subjects, we showed the one series, and the rest of the 22 other series. We trained mathematical models of neural networks, its semantic distances between concepts from films, for all subjects. Trained mathematical models of semantic spaces in two groups of subjects reliably differed among themselves.