Our brains consist of billions of neurons in densely-interconnected networks, and they control every aspect of our behavior, our movements, thoughts, and feelings. To explore their functioning, theorists have proposed neural network models in which every unit is connected to every other unit. Activity in one will spread to the others, depending on the strength of their connections. A key assumption is that when units are active simultaneously, their connection is strengthened; one formula used to calculate such changes is the delta rule, which is almost identical to the fomula of the Rescorla-Wagner model. These simple networks prove to be surprisingly powerful; they can account for many features of conditioning, concept learning, and memory. One recent development has been deep learning models that incorporate hidden units, between input and output units. This seemingly small innovation has dramatically increased the ability of these models to carry out sophisticated tasks; examples include beating world champions at chess and diagnosing skin cancer. One problem is that learning is slow, and new learning can result in the loss of older information (catastrophic interference). Whatever their ultimate fate, these models have demonstrated the power of even simple networks to perform tasks of astonishing sophistication.

The method of loci is one of the most effective techniques for memorizing lists. Its effectiveness reflects several memory principles, including elaboration, imagery, and the generation effect. The effectiveness of studying depends on the amount and spacing of practice; not just reading passively but actively thinking about material; and practicing retrieving what you’ve read: Practice, Think, and Review. The accuracy of eyewitness testimony can vary widely; important factors include the weapon-focus effect (distraction if a weapon is present), source confusion, and emotional intensity. Accuracy can be substantially improved by using the cognitive interview, in which witnesses are questioned in a manner designed to re-create the original scene as fully as possible, thereby maximizing possible retrieval cues. The accuracy of lineups can also be improved, by warning witnesses that the criminal may not be present. Recovered memories of childhood abuse can be false, as source confusion can make it difficult for people to tell whether their memories are based on real events or only ones that they imagined or heard described; therapeutic techniques such as hypnosis and guided imagery can exacerbate the problem. However, there is also compelling evidence that some recovered memories are true.

A wide variety of events can function as reinforcers. The Premack principle provides a useful guide; activities that individuals engage in when given a free choice will probably be effective reinforcers. One factor is the delay between the response and reinforcer; we discount, or give less value, to rewards that are delayed. Effectiveness also depends on the schedule of reinforcement; partial reinforcement (reinforcing only some responses) produces greater persistence in extinction. Effectiveness also depends on how the reinforcer contrasts with past reinforcers—we value a reinforcer less if we are accustomed to more attractive reinforcers—and how long we have been deprived of it. The stimuli present during reinforcement are also important; we are more likely to repeat a response if we are in the same environment where we obtained it previously. Which elements of that environment will be important, though, can be difficult to predict, depending in part on attention and perceptual learning (with practice we can become better at distinguishing elements of a situation). The chapter ends with an application involving shaping—when a behavior is difficult to train, start by reinforcing the response closest to it, and then gradually reinforce closer approximations.

Thorndike believed that reinforcement was an essentially simple process in which the reinforer stamped in an association between the preceding stimulus and response (S-R theory). Tolman disagreed, arguing that reinforcement was a more sophisticated process in which animals formed expectations that the response would produce the reinforcer. Behaviorists at first rejected the postulation of mental states, but both sides eventually agreed that it was acceptable to talk about invisible events (intervening variables), provided that they were couched in clear theories that led to testable predictions. This requirement proved difficult to satisfy, but it now looks as if Thorndike and Tolman were both right; evolution first produced a relatively simple associative system in which learning can occur without our awareness, but also to a more sophisticated system based on expectations. When several reinforced responses are available, the matching law says that we will distribute our responses to match the relative rates of reinforcement they have produced in the past. Kahneman and Tversky proposed an alternative analysis, that we use mental short-cuts or heuristics to make choices, partly based on emotions. Kahneman later proposed two systems, an unconscious system which generates intuitions, and a conscious system that may then modify them.