WITH ALL OUR EXTENSIVE KNOWLEDGE OF SENSORY CODING, sensory-

motor feedback loops, and motor control, we are still largely ignorant about what enables animals to be such successful autonomous agents. Even in arthropods with their tiny brains, we do not fully understand how behaviorally relevant information is acquired, filtered, and processed to guide behavior. This poses a fundamental challenge to neurobiology (e.g., see Pflüger and Menzel 1999): Most of what we know about the neural basis of visual processing, for instance, comes from experiments using well-defined simple patterns, such as black and white stripes or ran-dom dots, which were chosen to facilitate our interpretation of the neu-ronal responses. In addition, in most cases, we have no alternative but to study neurons out of context, in stripped-down organisms and in iso-lated or sliced-up brain preparations, in order to be able to measure their response properties. Neurophysiological studies have revealed a wealth of information on neurons and many of their interesting properties, but these properties are not necessarily the only ones that are relevant under natural operating conditions. Nervous systems are not general information transmission and processing devices, but they have evolved to solve specific computational tasks that are relevant to the survival and the fitness of an animal in a given environment. Visual processing mechanisms are adapted to the properties of the signals they encounter under these natural situation