Are There Two Kinds of Reasoning

Are There Two Kinds of Reasoning? Evan Heit (E.Heit@warwick.ac.uk) Department of Psychology, University of Warwick Coventry CV4 7AL, UK Caren M. Rotello (caren@psych.umass.edu) Department of Psychology, University of Massachusetts Amherst MA 01003-7710, USA people use a common set of reasoning processes for both deductive and inductive arguments. For example, Chater and Oaksford (2000) have applied an account of probabilistic reasoning, explicitly non-deductive in nature, to a range of deductive problems. Likewise, Harman (1999) has argued that people reason in an essentially non- deductive way, and bring these same reasoning processes to bear on both inductive and deductive reasoning problems. Taking a related approach, Johnson-Laird (1994) has extended the mental models account, more frequently applied to deductive problems, to a range of inductive problems as well. Finally, some researchers have proposed accounts that focus mainly on reasoning about inductive arguments, and have treated deductively correct arguments as special cases that would be covered by the same accounts (Heit, 2000; Osherson, Smith, Wilkie, Lopez, & Shafir, 1990; Sloman, 1993). In contrast, other researchers have emphasized a distinction between two kinds of reasoning (e.g., Evans & Over, 1996; Sloman, 1996; Stanovich, 1999). In these two- process accounts there is one system that is relatively fast but heavily influenced by context and associations, and another system that is more deliberative and analytic or rule-based. Although these two systems do not necessarily correspond directly to induction and deduction, it is plausible that induction would depend more on the first system whereas deduction would depend more on the second system. In addition there is some neuropsychological evidence, based on brain imaging, for two anatomically separate systems of reasoning (Goel, Gold, Kapur, & Houle, 1997; Parsons & Osherson, 2001). These one- and two-process proposals are mainly aimed at accounting for a range of phenomena rather than drawing a sharp line between deduction and induction. In contrast, the proposal by Rips (2001) does not aim for a detailed description of reasoning processes but instead focuses on a key commonality and a key difference between deduction and induction. In his account, there is a single scale for evaluating arguments. This account will be referred to as the criterion-shift account, and it is illustrated in Figure 1. Here, the unitary scale of argument strength is shown, with different points on the scale corresponding to arguments of different strengths. Criterion 1 indicates the dividing line between arguments that are inductively weak, or implausible, and arguments that are inductively strong, or Abstract Two experiments addressed the issue of how deductive reasoning and inductive reasoning are related. According to the criterion-shift account, these two kinds of reasoning assess arguments along a common scale of strength, however there is a stricter criterion for saying an argument is deductively correct as opposed to just inductively strong. The method, adapted from Rips (2001), was to give two groups of participants the same set of written arguments but with either deduction or induction instructions. Signal detection and receiver operating characteristic analyses showed that the difference between conditions could not be explained in terms of a criterion shift. Instead, the deduction condition showed greater sensitivity to argument strength than did the induction condition. Implications for two-process and one-process accounts of reasoning, and relations to memory research, are discussed. Keywords: reasoning; deduction; detection theory; memory; modeling. induction; signal Introduction How do convincing arguments differ from non-convincing arguments? Rips (2001) has referred to the intuitive case for a single psychological dimension of argument strength, in which arguments can range from utterly worthless to completely compelling. Hence, the convincingness of an argument could be judged by assessing its position on the scale, in a similar manner to how judgments of loudness or brightness would use a psychophysical scale. This intuition of a unitary scale needs to be reconciled with the notion that there are different kinds of reasoning. In particular there is the textbook distinction between deduction and induction, with deduction being concerned with drawing logically valid conclusions as opposed to induction which involves drawing plausible inferences. Strictly speaking, there are different kinds of arguments, such as deductively correct arguments, with respect to a well-defined logic, and inductively strong arguments (Skyrms, 2000). It is still an open question whether there are different kinds of reasoning, such as deductive reasoning and inductive reasoning. Some researchers have suggested that rather than having specialized cognitive processes for each kind of reasoning,