Preface<br >As in the previous editions of Learning: A Survey of Psychological Inter-<br >pretations， a major focus of the fifth edition is on theories that combine<br >the best features of connectionist and cognitive approaches to learning. In<br >any revision， there is always the feeling that more new or newly popular<br >theories should be added. In the interest of keeping the book manageable<br >for the reader， I have restricted my choices to two: Anderson s ACT theory<br >and the set of network models based on parallel distributed processing.<br >The former begins with cognitive assumptions and moves via response<br >compilation to processes with which a traditional connectionist would be<br >almost comfortable. The latter appropriates the term connectionism but<br >proceeds to develop models that might be most pleasing to a gestalt<br >psychologist. Both， therefore， contribute to the search for a theory that<br >will deal optimally with learning across the range， from classical condition-<br >ing to logical deduction.<br > Bandura s social learning theory， which represented a "liberalization"<br >of the behaviorist approach in a cognitive direction， has now moved fur-<br >ther in that direction and become social cognitive theory. Treatment of<br >Bandura has been correspondingly expanded. Other topics that receive a<br >little more attention than in the previous edition are Pavlovian theory，<br >biological constraints on learning， and the relationship between behav-<br >ioral regulation and reinforcement.<br > The cumulative effect of the three previous revisions has been to<br >increase both the length and the difficulty level of the book. In an effort<br >to stop this trend and to keep the book both short and accessible to a wide<br >range of readers， I have made a number of deletions. The most conspicu-<br >ous is in mathematical models. One such model of learning is enough to<br >X<br ><br >