AbstractVarious philosophical definitions of free will are first considered. The compatibilist definition, which says simply that acts are freely willed if they are not subject to constraints, is identified as much used in the legal system and essentially impervious to scientific investigation. A middle‐ground “incompatibilist” definition, which requires that freely willed acts be consciously initiated, is shown to be relevant to the idea of mens rea and in the author's view not actually incompatible in principle with a fully scientific worldview. Only the strong libertarian definition, which requires that freely willed acts have no physical antecedents whatsoever, makes the existence of free will very hard to swallow scientifically. However, with regard to the middle‐ground “incompatibilist” definition, three different lines of scientific experimental evidence are then described, which suggest that, in fact, consciousness is not the real cause of much of what is generally considered as voluntary behavior. Many voluntary actions are initiated preconsciously, with consciousness kept informed only after the neural events leading to the act have begun. It is suggested that a reasonable way of integrating these experimental findings with the idea that persons do have a somewhat more than compatibilist version of free will is to acknowledge explicitly that a person is a mixture of conscious and unconscious components. In this scenario, the mind in mens rea would have to be judged guilty if it contained either conscious or unconscious intentions to perform the guilty act. Copyright © 2007 John Wiley & Sons, Ltd.