In this chapter I hope to demonstrate that answering the question “who does what in reasoning experiments?” can also help us answer fundamental questions about the nature of thought. Because it is nearly always the case that some experimental participants display phenomena of interest and others don't, almost every experiment run by cognitive psychologists produces data on individual differences. But, as most psychologists do not wish to take differences between their participants as the starting point of their investigations, individual differences tend to be ignored. Thus, the means that experimental psychologists report when they describe their data abstract across differences between individuals, with only the standard deviation or standard error indicating the extent to which participants varied in their responses. In this chapter I will address what the study of differences between individuals might tell us about a range of issues in inductive reasoning, including what constitutes a good inference, what processes underlie induction, and how induction differs from deduction.

My primary focus will be on the general question of how individual differences data support dual-process theories of thinking (Evans & Over, 1996; Stanovich, 1999; Sloman, 1996). Dual-process theories have been applied to deduction, decision making, and induction. I will outline the dual process approach and describe evidence in its favour, paying particular attention to individual differences. I will also describe how individual-differences methodology has been used to arbitrate between different normative accounts of specific thinking tasks (see Stanovich & West, 1998b, c). With this background covered I will consider what individual differences might tell us about category-based induction.

Books on induction are rare; in fact, you are holding in your hands the first book devoted solely to the psychology of inductive reasoning in twenty years. And yet induction is a central topic in cognitive science, fundamental to learning, discovery, and prediction. We make inductive inferences every time we use what we already know to deal with novel situations. For example, wherever you found this book – in your university library, online, or in an academic bookshop – before picking it from the shelf or clicking on a link, you will have made some inductive inferences. Amongst other things, these inferences will have concerned the book's likely content given its title, its editors, its publisher, or its cover. Common to all of these inferences will have been the use of what you already know – about us, or the topic suggested by our title, or Cambridge University Press – to make predictions about the likely attributes of this book.

It is not only in publishers' catalogues that attention to induction has been scant. Despite its obvious centrality to an understanding of human cognition and behaviour, much less work has been carried out by psychologists on induction than on deduction, or logical reasoning. As a consequence, although there have been several edited collections of work on deduction and on decision making (Connolly, Arkes, & Hammond, 2000; Leighton & Sternberg, 2004; Gilovich, Griffin, & Kahneman, 2002; Manktelow & Chung, 2004), to the best of our knowledge there has never previously been an edited collection of papers on the psychology of inductive inference.

In this chapter we examine research that uses well-defined laboratory problems requiring hypothetical thinking and reasoning for their solution. By “well-defined” we mean that all information required to solve the problem according to the instructions is explicitly presented. For this reason, psychologists have traditionally regarded any influence of prior knowledge or belief about the problem content or context to be normatively irrelevant to the definition of a correct answer. Consequently, where such beliefs exert an influence this has often been termed a “bias” by the investigators concerned. The effects of prior belief, however, turn out to be so pervasive in these studies that reasoning researchers in the past decade or so have begun radically to reexamine their assumptions about the nature of rational reasoning.

This reassessment has been no where more visible than in the study of deductive reasoning, one of the major paradigms in this field. Typical experiments involve presenting participants with the premises of logical arguments and asking them to evaluate a conclusion presented, or draw one of their own (for reviews, see Evans, Newstead, & Byrne, 1993; Manktelow, 1999). The deduction paradigm has its origins in logicism – the belief that logic provides the rational basis for human reasoning (Evans, 2000a). The modern study of deductive reasoning dates from the 1960s, where it was motivated by the writings of psychologists such as Henle (1962) and especially Jean Piaget (Inhelder & Piaget, 1958), who proposed that adult human reasoning was inherently logical.