Cognitive Neuropsychology Of Reasoning

In the other chapters on thinking, we have seen that there has been a considerable amount of new activity in the cognitive neuropsychology area. Similar developments have occurred in reasoning research. This research has confirmed the importance of the frontal lobes, and has gone further to distinguish hemispheric specialisations in deductive reasoning (see e.g., Goel, Gold, Kapur, & Houle, 1997).

In this chapter, we have reported marked differences in people's responses to logical problems when they are couched in abstract versus thematic content. There is evidence to suggest that this distinction has a grounding in hemispheric differences. Golding (1981) used a circles and colours version of the selection task to test subjects that had either no cerebral brain lesions, right hemisphere lesions, or left hemisphere lesions. This abstract version of the task was found to be particularly difficult for those patients with left hemisphere damage, whereas the right-hemispheredamaged patients did much better (better indeed than controls). The left hemisphere has also been implicated in studies showing that aphasics, with left posterior lesions, have severe difficulties understanding simple logical statements, whereas right-hemisphere-lesioned subjects have no general difficulties with logical reasoning (Wharton & Grafman, 1998).

Johnson-Laird (1995) has emphasised the right hemisphere's contribution to deduction based on the proposals of mental models theory. He reports an experiment by Whitaker et al. (1991) on conditional reasoning in group of patients who had undergone a unilateral anterior temporal lobectomy to relieve epilepsy, either to the right or left hemispheres. Those with right hemisphere damage were poorer at reasoning from false conditional premises than those with left hemisphere damage. Thus given: "If it rained the streets will be dry" and the categorical assertion: "It rained" the right-hemisphere-damaged group had a reliable tendency to conclude: "The streets will be wet". In other words, these patients were unable to carry through the process of deduction in isolation from their knowledge of reality.

Finally, Waltz et al. (1999) have used deductive and inductive problems to show that the underlying basis for the localisation in the prefrontal cortex is specifically the integration of multiple relations. They gave three groups of subjects who were matched in terms of their IQ—patients with prefrontal damage, patients with anterior lobe damage, and controls—versions of problems that could be solved by using one relation or two relations. They found that the prefrontal patients showed a very specific, catastrophic deficit in their ability to deal with the multiple-relation problems, although they paralleled people in the other groups on single-relation problems. On the multiplerelation problems they exhibited a level of performance normally seen in pre-school children. In short, the picture emerging from these studies and others on problem solving and analogy is a very consistent one (see Chapters 14 and 15).


Given the evidence we have seen on people's reasoning abilities, one might be forgiven for thinking that we are all massively irrational. From the literature we have reviewed there are perhaps three main stances on this question.

The view that emerges from the long-standing theories in the area—the abstract-rule and mental models theories—is that people are rational, that they operate in accordance with a rational principle, in trying to deduce valid conclusions from the premises of an argument. Johnson-Laird and Byrne (1991, Johnson-Laird, 1999) frame this as a modicum of rational competence, which is important to everyday goals and essential for the development of expertise in logic. This is not to rule out irrational responses altogether; such responses clearly occur and will continue to occur.

A second, more recent, view changes the question slightly, in proposing that people are adaptive to their environment. This is clearly the view motivating domain-specific-rule theories and probabilistic theory. The proposal is that people act in accordance with another rational principle— maximising goal achievement or maximising information gain—rather than one derived from logic. For instance, on evolutionary grounds Cosmides (1989) argues that there are specialised inferential mechanisms for "checking for cheaters" which were specifically adaptive for our hunter-gather forebears. Oaksford and Chater argue for the maximisation of information gain and the reduction of uncertainty as the guiding principle. More widely, this view is taken by Anderson (1990) in his "rational analysis" technique to characterise many diverse areas of cognition from memory to problem solving.

The final position, which is more a metaanalysis of the problem, is to distinguish between different forms of rationality. Evans and Over (1996, 1997) distinguish between the unfortunately named rationality! and rationality2. Rationality! is the adaptative goal-directed form of reasoning and problem solving that helps us achieve our goals in the world. Rationality2 is a more reflective form of reasoning that is consistent with the normative models of logic. It is possible for someone to be make an inference that is fine from the rationality! perspective but completely at variance with the rationality2 view (for commentaries see Cummins, 1997; Hertwig, Ortmann, & Gigerenzer, 1997; Noveck, 1997; Ormerod, 1997; Stanovich, 1999). However, beyond labelling two interpretations of rationality, it is not clear how this distinction brings us closer to any conclusions about human rationality. Ideally, we should have a definition of rationality that is both adaptive and normative.

In summation, it is clear that there is general agreement that people are, for the most part, adaptive or rational in some way. What is at issue is the exact rational principle at work.

Assessing competing theories

In this chapter, we have examined empirical work that centres on reasoning with conditionals. It is important to note that this is just one area of research in deductive reasoning. There are large literatures on syllogistic reasoning (see Evans et al, 1993; Eysenck & Keane, 1995, Chapter 17; Manktelow, 1999), on spatial reasoning (Johnson-Laird & Byrne, 1991), and reasoning with logical connectives other than if... then (e.g., and, or, not). We have only peeked through a small window at the theories of concern. With this in mind, we can tentatively broach the question "What is the current best account of reasoning?". Usually, science seeks a single unified theory, rather than a cluster of alternative theories. So, we will assume that the answer to this question should be a single candidate theory.

The domain-specific rule and probabilistic theories probably come off the worst in terms of the generality of their coverage. Domain-specific rules really only apply to selection task problems and rely on some other theory (e.g., abstract rules) to deal with a wide set of phenomena. Probabilistic theory has also been applied to the selection task and more recently to syllogistic reasoning (Chater & Oaksford, 1999b) but, as we saw earlier, has a way to go in developing a performance theory of reasoning. The remaining theories—the abstract-rule and mental model theories—have the best overall coverage of the field.

Abstract-rule theories have been applied to propositional reasoning. They can generate precise predictions about subjects' judgements of validity (Braine et al., 1984; Osherson, 1974, 1975; Rips, 1983), about the reaction times of subjects on tasks (Braine et al., 1984), and about intersubject differences on problems (Rips & Conrad, 1983). They have also been applied to syllogistic reasoning (Braine & Rumain, 1983; Osherson, 1976; Rips, 1994). The results of research on the selection task are more embarrassing. Some abstract-rule theorists have proposed that formal inference rules can be supplemented by domainspecific rules or by the addition of modal logics and non-logical operators (see Braine & O'Brien, 1991; Rips 1989). However, the problem with this proposal is that it begins to move away from a unified account of a variety of reasoning phenomena (see also Manktelow & Over's, 1991, criticisms of this proposition).

Finally, the model theory scores highly on gen-erality. It has well developed accounts of conditional and syllogistic reasoning. Johnson-Laird and Byrne (1991) have proposed an account of all of the effects found on the selection task, although this proposal requires some fleshing out. The theory has also been spread wider to explain aspects of imagination in counterfactual reasoning (Byrne, 1997; Byrne & Tasso, 1999), spatial reasoning (Johnson-Laird & Byrne, 1991), creativity (Johnson-Laird, 1989), and probabilistic reasoning (Johnson-Laird et al., 1999). At present, it looks like the most complete theory of human reasoning, although its opponents continue to challenge many of its foundations.

Stop Anxiety Attacks

Stop Anxiety Attacks

Here's How You Could End Anxiety and Panic Attacks For Good Prevent Anxiety in Your Golden Years Without Harmful Prescription Drugs. If You Give Me 15 minutes, I Will Show You a Breakthrough That Will Change The Way You Think About Anxiety and Panic Attacks Forever! If you are still suffering because your doctor can't help you, here's some great news...!

Get My Free Ebook

Post a comment