(No abstract. The following are extracts from the review.)
Donahoe and Palmer's Learning and Complex Behavior presents a carefully reasoned case in favor of hierarchical integration based on a single, small set of organizing principles. The book gives serious attention to research findings that have been generated in the traditions of cognitive psychology and psycholinguistics. In doing this, however, it attempts to show explicitly how those findings can be interpreted as products of learning histories involving relatively simple and general processes of behavioral selection. Moreover, the book attempts to show how these basic learning processes can be derived from even more fundamental principles that have been established through neurophysiological research. These derivations are made, in part, through a kind of computer model known as adaptive networks, instances of which are built from simple, undifferentiated elements rather than from the highly modularized units of traditional [artificial intelligence] models. The book thus promises an integrated account of complex human behavior that is strongly grounded in biobehavioral science and connected with recent developments in computer modeling.
A number of noteworthy themes are developed in the book. I will comment on just four of them: selectionism and historical science, interpretation, the process of reinforcement, and the treatment of memory phenomena.