People with autism and schizophrenia have been shown to have a local bias in sensory processing and face recognition difficulties. A global or holistic processing strategy is known to be important when recognising faces. Studies investigating face recognition in these populations are reviewed and show that holistic processing is employed despite lower overall performance in the tasks used. This implies that holistic processing is necessary but not sufficient for optimal face recognition and new avenues for research into face recognition based on network models of autism and schizophrenia are proposed.