A model of a neuronal network is presented which, like a significant proportion of the neurons in the visual cortex, identifies not only the orientation of a single band, but also high selective sensitivity to cross-like figures. The model was used to study the properties of a "cross detector" constructed on the basis of convergence of excitatory connections with different weightings from elements with different orientational tuning. It is shown that a cross detector with monomodal orientational tuning needs an amplifier mechanism (a reverberator).