Rules for Reasoning

The book argues that people possess abstract rules of inference—statistical, conditional, causal, and cost-benefit—and that these rules can be taught through brief instruction and transferred across content domains. It rejects the Piagetian and American psychological assumption that abstract reasoning rules cannot be taught, as well as reductionist models from behaviorism to connectionism that deny people operate with abstract rules at all . The book's framing is deliberately optimistic: where Human Inference documented the shortcomings of social judgment, Rules for Reasoning asks what conditions promote the effective use of sound rules. The evidence comes from training studies showing that instruction in the law of large numbers, pragmatic reasoning schemas, and cost-benefit analysis improves performance on unfamiliar problems .

The two books are complementary, not contradictory, despite their different tones. A contemporary review by Deuser and Anderson explicitly addressed the apparent tension: Human Inference focused on the cognitive processes that produce error and bias, while Rules for Reasoning focuses on people's ability to reason well across situations . The reviewer insisted that the later book is not a repudiation of the earlier one—both begin with the same broad question of how people make complex judgments, but use different heuristic questions to investigate it. Human Inference asks what conditions lead to error; Rules for Reasoning asks what conditions promote effective rule use. The later book extends the earlier program by showing that the same cognitive system capable of systematic bias is also capable of learning and applying abstract rules.

The Contemporary Psychology review by William E. Deuser and Craig A. Anderson (1994) was largely positive, praising the introductory chapter's five interrelated goals as targeting important philosophical, theoretical, and practical questions . The reviewer noted that Nisbett assembled twelve empirical journal articles published between 1983 and 1992, supplemented by three original chapters for coherence, and judged the research to be of high quality . The book's media reviews emphasized its broad appeal to cognitive and social psychologists, educational psychologists, and educators, and its success at taking basic science psychology research and applying it to education . The reviewer's main criterion for judgment was whether the book met its stated goals, and on that measure it was judged successful.

The book's empirical core includes several training studies. Fong, Krantz, and Nisbett trained subjects on the law of large numbers and found they sometimes applied it to cases where it was inappropriate—evidence that the rule had been learned abstractly and overextended . Cheng, Holyoak, Nisbett, and Oliver compared pragmatic versus syntactic approaches to training deductive reasoning, finding that abstract logic training alone was largely ineffective while training on pragmatic reasoning schemas (permission and obligation rules) produced marked improvement on selection problems . Larrick, Morgan, and Nisbett taught cost-benefit reasoning and found it could be applied to everyday life . These studies collectively demonstrate that abstract rules can be taught and that transfer occurs, though the effectiveness varies by rule type and training method.

Pragmatic reasoning schemas are abstract rules that people acquire through everyday experience with permission, obligation, and causation—rules like 'if one is to take action A, then precondition P must be satisfied.' Cheng and Nisbett argued that these schemas are more natural and more readily applied than purely syntactic rules of logic. In the book, the chapter by Cheng, Holyoak, Nisbett, and Oliver presented evidence that a semester of standard logic training improved performance on arbitrary problems only 11%, whereas students without formal training solved 62% of permission problems correctly; brief training on the obligation schema markedly improved selection-problem performance . This finding supports the book's broader claim that people possess abstract rules, but that the rules most readily used are those with pragmatic content acquired through experience rather than formal instruction.

The book's educational implication is that relatively brief, abstract training can broadly influence problem-solving across content domains—contrary to the assumption that transfer is rare and education must be content-specific . Part VI of the book, 'Implications for Education,' includes chapters on teaching reasoning and the effects of graduate training on reasoning . The review by Deuser and Anderson accepted the importance of these questions without disputing the evidence . However, the educational implications are more suggestive than definitive: the training studies were conducted in laboratory and classroom settings with specific populations, and the book does not offer a comprehensive curriculum. The claim is that abstract rules can transfer, not that any particular educational intervention will reliably produce broad improvements.

The book is organized into seven parts: establishing the existence of rules for reasoning, teaching statistical rules, rules for conditional reasoning, rules for causal reasoning, rules for choice, implications for education, and implications for cognitive science . It consists primarily of previously published journal articles (twelve in total) bracketed by three original chapters that provide coherence . This structure has advantages and limitations: it gives readers access to the empirical papers in one volume with contextual framing, but the chapters vary in focus and the book reads as a research compendium rather than a unified monograph. Nisbett's introductory and concluding chapters do the work of synthesis, but readers seeking a single sustained argument should recognize that the book's unity is programmatic rather than narrative.