Command hallucinations (CH) have recently become the focus of research into positive symptoms of schizophrenia. The importance of CH has become clear for theoretical and practical reasons, because CH are regarded as potentially the most dangerous symptoms of schizophrenia. The aim is to critically review research that has attempted to better understand CH and develop theories that may predict behavior and shape psychological treatments. This review draws together, presents, and critically discusses the current disparate body of literature produced (1990-2000) considering compliance with CH. The literature can be grouped broadly into three overlapping classes. (1) Studies testing for a relationship between CH and compliance. (2) Studies considering factors associated with acting on CH (mediating variables). (3) Studies that look for a relationship between CH and dangerous behavior. The body of literature is in its infancy and marred by methodological difficulties. Researchers have asked differing questions about different samples while hoping to provide similar answers. While acknowledging difficulties in interpreting the literature, themes do emerge. The weight of the evidence is that some individuals who hear commands will act on them. However, some studies show no link or only a weak link and although methodological errors account for some of the inconsistent results, there are also clearly individual differences. The relationship between command and action is more complex than many researchers have assumed. One conjecture is that the relationship is established via psychological processes or mediating variables which include beliefs about the voice and content of instruction. Prediction of compliance with commands is of further interest when dangerous behavior results. It is postulated that the issue of dangerousness is subject to the same mediating variables as other commands. It is clear however, that CH are not sufficient to produce action in isolation and that psychological processes mediate the process.