``Cognitive neuroscience'' is the name of the game, which Harvard psychologist Kosslyn and Univ. of Geneva colleague Koenig equate to ``wet mind.'' The ``wet'' alludes to understanding how the brain really works (equating brain function to mind), and not, as with ``dry mind,'' to designing computers or models of artificial intelligence to perform visual perception or reasoning tasks. The computer designs and experiments the authors describe, then, are based on neuronal networks and parallel-processing systems rather than the sequential mode of operation of your ordinary PC. Their constructs of how we see or hear or read or remember are based on breaking the process into component subsystems and relating these systems to anatomical sites and pathways in the brain. Finally, as a partial test of their hypotheses, they use the data from patients with brain lesions to relate loss of function to damage in a particular subsystem (or systems). This is heady stuff, embodying at the outset five principles: the brain works by a division of labor; it expresses ``weak'' modularity (its components are not independent and, while they may be anatomically close, may include distant elements as well); the systems operate under ``weak'' constraints (they may be tuned to some inputs but accept others as well); parallel processing is the mode; and the brain is opportunistic, so that, for example, parts controlling fine movements may be applied to doing serial arithmetic problems too. The authors provide detailed accounts of how computer networks have been used to model visual perception and thinking, reading, language, movement, and memory, and how they may apply to emotion, consciousness, and other mental phenomena. Exciting, important research that's on the right track: It's about time that neuroscience, cognitive psychology, and computers got together. Now if they could only clean up the language (``The property lookup and categorical-to-coordinate conversion subsystems are probably implemented in the frontal lobes....'').