Research on nonstatistical heuristics has been criticized on several grounds. Some critics have maintained that evolution should be expected to produce highly efficacious and generally correct principles of reasoning and that the research may therefore be misleading in some way (Cohen, 1979; Dennett, 1978, 1981, 1983; Lycan, 1981). Others have maintained that the research does not demonstrate that people fail to apply correct inferential rules but rather that (1) it is the researchers themselves who are mistaken about the correct inferential rules (Cohen, 1981), (2) subjects have been misled by illusionary circumstances of little general significance beyond the laboratory (Cohen, 1981; Lopes, 1982; Dennett, 1983), or (3) people's general inferential goals are such that at least some violations of statistical principles should be regarded as a form of satisficing or cost-effective inferential shortcuts (Einhorn & Hogarth, 1981; Miller & Cantor, 1982; Nisbett & Ross, 1980)….

In this chapter, we first summarize the work establishing failures to reason statistically. We then review anecdotal and experimental evidence indicating that people do sometimes reason statistically. Next we present original experimental work indicating some of the factors that influence statistical reasoning. Then we summarize research suggesting that people's ability to reason statistically about everyday life problems is affected by training in formal statistics. Finally, we speculate on the normative implications of people's ability and trainability for statistical reasoning.