Over the past 20 years or so, some of the most important insights into the nature of human cognition have come from studying people with acquired disorders of cognition following brain damage. After brain damage, patients can lose the ability to perform certain cognitive tasks, such as reading or speaking words, recognising objects, orienting attention appropriately within visual scenes and so forth — tasks that are normally performed effortlessly by intact cognitive systems in the brain. The disorders can inform us about the nature of these cognitive systems, particularly about which processes can operate independently of others, and so which processes can be selectively spared when others are impaired. Cognitive neuropsychologists attempt to use data from cognitive disorders in order to test and refine normal models of cognition; they also attempt to understand cognitive disorders in terms of these models (Coltheart, 1984). Connectionist modelling is useful in this last regard, because the effects of brain lesioning can be simulated by removing “neurons” (processing units) from the models, by reducing the weights on connections, or by adding noise to either activation functions or to connection strengths. Also the patient data can provide strong tests of the validity of the simulated models, as the models need to break down in a way that matches human performance (see Olson & Humphreys, 1997). In this chapter we review attempts to simulate cognitive disorders in such models.