The number of Boolean functions that can be represented by neural networks is limited by their connective structure, which is kind of local in space. This imposes constraints on training, especially if it involves recognizing particular patterns of visual inputs.