This project addresses three interrelated problems: (i) developing a rich biologically grounded account of intelligent agents, (ii) understanding the basis of the transition from reactive to anticipative forms of adaptive behavior, (iii) developing an ethologically plausible account of the developmental learning processes involved in complex cognitive behavior. i. A biologically based account of intelligent agents. Despite recent widespread interest in bringing biology and cognitive science into contact, the terms of engagement remain highly controversial. Evolutionary psychology has been justifiably criticised for being adaptationist, but there are also deeper problems inherent in the nature of the cognitivist paradigm, which historically has dominated psychology and most other branches of cognitive science. The recent development of dynamical embodied approaches to cognitive science are inherently much more suited to integrating cognitive science and biology, but they face a key problem in accounting for the nature of high order cognition. The project seeks to address this issue by developing a theory of self-directed agents that builds on existing dynamical embodied research, and focuses on explaining the emergence of fundamental cognitive abilities such as choice, anticipation and evaluation. ii. Investigating the transition from reactive to anticipative forms of adaptive behavior. In developing the theory of self-directed agents, the project will examine the basis of the transition from the largely reactive type of adaptive organization often associated with insects (and mimicked by behavior-based robotics), to more sophisticated forms of cognitive organization. Understanding the transition from reactive to flexible anticipative behavior is a crucial issue both for cognitive science and for a biological understanding of adaptability. The central hypothesis is that integrative processes for managing multiple streams of information are required to produce flexible, coordinated behavior in complex task environments. The project will examine current research on integrative architectures in neuroethology and biomimetic robotics, and analyze the neuro-architectural requirements for self-directedness. iii. Modeling the developmental learning processes involved in complex cognitive behavior. Complex agents face what has been termed the ‘degrees of freedom’ problem: the ecological problems they must deal with have many variables, and there are many possible strategies, only some of which will be successful. Arguably, the only way for agents to manage problems of this kind is through extended developmental learning processes. The project will develop an ethologically plausible, task-oriented approach to modeling bootstrapping learning processes called self-directed anticipative learning (SDAL). SDAL processes provide a means for solving open problems, in which the agent initially lacks the ability to recognize or act upon the factors that are responsible for ultimately achieving success. Open problems are solved by a progressive transformation of the representational and action capacities of the agent.