A model approach to computational vision.

repeatedly says that he is explaining how our brains make us who we are, and tries hard to make the meaning of this mantra clear. He defines the self as 'the totality of the living organism', but also proposes that 'the self is created and maintained by arrangements of synaptic connections'. I was unable to find any clear coherent meaning linking these statements, or LeDoux's many other attempts to convey his conception of the self. More plausible to me is William James's pluralistic interpretation of the various concepts of self, including bodily, social, and 'spiritual' versions, and extending to those that are remote and potential, as well as to those that are immediate and actual. That said, I do not regard my sceptism concerning LeDoux's concept of the self as a major criticism. I am enthusiastic about the broader project of building the discipline of molecular psychology, with synaptic interactions at its centre. My optimism for its prospects is greatly increased by the quality of the work reviewed by LeDoux in this valuable contribution to the task of building a better understanding of the relation between brain and mind, and of presenting it to as broad an audience as possible. A more accurate title would be Commonsense Reasoning and Rationality, and, indeed, that is the phrase used in the title of the editor's introductory chapter. The 19 authors, mostly philosophers, along with several psychologists and a couple of computer scientists, are engaged in a debate about the nature of human commonsense reasoning and the applicability of a normative notion of rationality to the study of human commonsense reasoning. The prime contenders, here, for a normative notion of rationality are deductive logic and Bayesian probability. We might first wonder what a normative notion of rationality is. The answer provided is a standard of reasoning against which a person's reasoning might be judged to be right or wrong, whereas commonsense reasoning is however people actually reason. For example, consider, as several of these authors do, the well-known findings of Kahneman and Tversky (e.g. see [1]), where people consider the likelihood of a conjunction to be higher than that of either conjunct. If you think these subjects are making mistakes in their reasoning, you must be comparing their performance to some normative notion of reasoning, such as Bayesian probability theory. If, however, your goal is to understand or model how the subjects actually …