The Psychology of Deductive Reasoning (Psychology Revivals)

Jonathan Evans

• Publisher: Psychology Press
• Year: 2013
• Print ISBN: 9781848723153
• eBook ISBN: 9781317820451
• Topic: Psychology
• Subtopic: Cognitive Psychology & Cognition
• Index: Psychology

Originally published in 1982, this was an extensive and up-to-date review of research into the psychology of deductive reasoning, Jonathan Evans presents an alternative theoretical framework to the rationalist approach which had dominated much of the published work in this field at the time.

The review falls into three sections. The first is concerned with elementary reasoning tasks, in which response latency is the prime measure of interest. The second and third sections are concerned with syllogistic and propositional reasoning respectively, in which interest has focused on the explanation of frequently observed logical errors. In an extended discussion it is argued that reasoning processes are content specific, and give little indication of the operation of any underlying system of logical competence. Finally, a dual process theory of reasoning, with broad implications and connections with other fields of psychology, is elaborated and assessed in the light of recent evidence.

1  Introduction

In one sense this book is only about deductive reasoning. In another sense it is about language comprehension, mental imagery, learning processes, memory organisation and the nature of human thought. The first sense is defined by the paradigms employed; the second by nature of the psychological processes which the paradigms evoke.

We will start with a brief consideration of the notion of deductive reasoning as a philosophical concept, and then proceed to psychological issues. One rough definition of deductive thought is that it leads from the general to the particular, e.g.

All swans are white,

This bird is a swan

Therefore, this bird is white.

An inductive inference, on the other hand, leads from the particular to the general:

All the swans I have ever seen are white,

Therefore all swans are white.

This inference is not valid, nor is any other inductive inference. An argument is valid if assumptions which are true cannot lead to conclusions which are false. However many white swans I have seen, there may be some black ones around.

The validity of a logical argument is not affected by whether its premises and conclusions are, in fact, true or false. The following argument is valid, even though its conclusion is obviously false:

All cats are fish,

All fish have gills,

Therefore, all cats have gills.

This valid argument permits a false conclusion because one of its assumptions is false. Invalid arguments may also lead to true conclusions:

If 2 + 2 = 4 then dogs chase cats,

Dogs chase cats,

Therefore, 2 + 2 = 4.

Deductive reasoning is tautological. It adds no new knowledge, but states necessary consequences of that which is already assumed. However, logical systems adopt certain principles as axiomatic, such as the law of non-contradiction. A contradiction is the simultaneous assertion that some proposition p and its negation not p are both true. The avoidance of such contradictions is fundamental to most systems of logic.

A logical system normally has a set of principles or rules of inference which provide techniques for making valid deductions of conclusions from assumptions or premises. Mathematics can be regarded as a set of logical systems, in which symbols are manipulated. For example, the solution of simultaneous equations starts with the equations (premises) and, by application of rules, deduces the necessary numerical values of the algebraic variables as the conclusion. A good example of a set of logical arguments is Euclid’s geometry, in which a set of theorems (conclusions) are deduced from a set of axioms (premises).

We are not concerned in this book with mathematical logic, but rather with the deduction of conclusions stated as verbal propositions. A number of philosophers, from Aristotle onwards, have been concerned with devising systems of logic to achieve this. Assuming that it is possible to analyse the logical structure of natural language arguments, we might ask why it is necessary to do so. One area in which logical argument seems to be essential is in the construction and testing of scientific theories. Logic facilitates this process in two ways. Firstly, it provides a check on the internal consistency of a theory; it should not be possible to deduce a contradiction from the assumptions of the theory. Secondly, deductive logic is involved in testing empirical predictions of scientific theories.

Philosophers of science traditionally viewed science as a process of making inductive generalisations from experimental observations. However, as observed earlier, such inductive inferences can never be valid, so is it not unsatisfactory for science to be based on an invalid type of logic? The modern philosopher Karl Popper (e.g. 1959) has resolved this dilemma by asserting that the purpose of science is not the verification but the falsification of theories. Falsification can be achieved by deductive logic. I can never prove the statement ‘All swans are white’ to be true, but I only need to observe one black swan to prove that it is false. The logical nature of scientific theory testing is as follows:

1  State theory.

2  Deduce logically necessary and empirically testable predictions.

3  Test predictions by experiment or observation.

4  Re-evaluate theory.

If a theory’s predictions fail, i.e. are not borne out by empirical observation, then we have a contradiction between prediction and observation. Assuming the experiment was properly conducted, this means that at least one assumption of the original theory must have been incorrect. Consequently, the theory must be either revised or abandoned. If the prediction is confirmed, this does not necessarily mean that the theory is correct. Popper argues that a theory must be open to empirical falsification in order to be called a scientific theory, and that a good theory will make risky predictions, i.e. predict things that would not be likely a priori.

Science, as Popper sees it, consists of alternating conjectures and refutations. Bad theories are weeded out, and good ones survive by a kind of natural selection. We may agree with Popper, then, that deductive logic is fundamental in science. We must be able to deduce conclusions and to eliminate contradictions. Is deductive logic a necessary and natural part of human thought, however? The answer depends upon whether or not one adopts a rationalist view of man.

In the rationalist approach (e.g. Kelly, 1955) man is seen as a kind of scientist in his everyday life. According to this model, people are continually involved in constructing theories, deriving predictions, collecting evidence and the like in all aspects of life, including, for example, the development of interpersonal relationships. A rational man would need, and be expected to possess, some system of deductive logic. The behaviourist view (e.g. Skinner, 1972), on the other hand, sees man’s behaviour as under the control of his environment, and determined by his personal history of reinforcement. This approach does not require the assumption of any internalised logical system. Behaviour will accord with logical principles if and only if the appropriate reinforcement contingencies have been applied.

There are, of course, many intermediate positions, but it will be useful to keep sight of these extremes. Rationalism, in particular, is manifest in many of the approaches to reasoning research that will be reviewed. This viewpoint has implications beyond the assumption of logical competence. For example, it may lead to the assumption that thought processes are available to introspection, an idea that will be critically examined in the later part of this book. It is even supposed by some that a system of logical thought is innately determined. Recent books on the psychology of reasoning have had a distinct bias towards the rationalist approach (Falmagne, 1975; Revlin and Mayer, 1978). In contrast, this book will emphasise non-rational aspects of human reasoning, in the context of a review of the recent literature in the field.

All the experimental research to be reviewed can be seen, in one sense, as assessing people’s competence to solve logical tasks. Part I of the book is concerned with relatively simple tasks where error rates are generally low. In the more complex tasks involving syllogistic reasoning (Part II), and propositional reasoning (Part III), however, logical errors abound in all studies. This might, in itself, lead one to abandon rationalism; but rationalists are not so easily deterred.

Rationality is seen as subjective to the individual. In the field of decision-making, for example, rational man is supposed to choose in such a way as to maximise personal gain (see Lee, 1971). Objectively, many decisions appear ‘irrational’, but can be explained as rational by assuming subjective distortions in the assumptions which the decision-maker holds. Some discussion of decision research and its applicability in the explanation of reasoning data is given in Chapter 11. In reasoning, as in decision-making, ‘rational’ explanations can be offered for apparently irrational, i.e. illogical, performance. Henle (1962) proposed that subjects interpret the premises of reasoning arguments in a personal way. They may alter, add or drop premises. She contends, however, that subjects’ conclusions follow logically from their reinterpreted version of the problem. This highly influential paper is discussed in detail in Chapter 5, and the merits of ‘Henleism’ are subsequently examined with reference to both syllogistic and propositional reasoning.

In this book, it will not be assumed, a priori, that ‘reasoning’ is necessarily going on in reasoning experiments. This brings us back to the point of the opening paragraph of this chapter. The book reviews research which has involved deductive reasoning paradigms. Our concern is with the psychological explanation of performance on such tasks, and the wider implications that follow. One disadvantage of the psychology of reasoning is the relative isolation of the field from other work in cognitive psychology. A further aim of the present book is to relate reasoning research to general issues in the study of cognition.

At this point, it would be helpful to define the criteria by which work has been included in the review that forms the major part of this book. A deductive reasoning task involves making an inference from information which is given. If the task requires access to memory of things which are not presented, then it is not simply a reasoning task. This means, for example, that problems which draw on semantic memory for their solution are inadmissible. For similar reasons, the interesting area of pragmatic inference (see, for example, Harris and Monaco, 1978) is not included. Many of the complex reasoning tasks discussed in Parts II and III employ abstract materials such as letters and numbers, although the use of thematic content in logically equivalent tasks is also considered in some detail. Even in the latter case, however, instructions to these tasks will always indicate that subjects should evaluate the validity of arguments on the basis of what is presented.

The limitation is, of course, one of paradigms not processes. Subjects may well be influenced by transfer of learning from other situations, and a number of the explanations to be considered are along these lines. Both rationalists and non-rationalists account for reasoning errors in terms of some learned tendencies, relating to the interpretation of sentences, or other factors. The non-rationalist, however, has both more to explain (right as well as wrong answers) and more devices with which to formulate explanations (e.g. response biases).

The tasks used do, of course, require some linguistic knowledge for their solution. For example, negation has the logical property of reversing truth value: if p is true then not p is false, and vice versa. Studies of sentence verification (Chapter 3) reveal that negation serves this function in language as well, but linguistic negation is more complicated. Other reasoning tasks involve sentences such as All A are B, If p then q and Either p or q. Obviously, subjects can only attempt to solve such tasks if they understand the meaning of connectives such as ‘If … then …’. Again, we shall see that the linguistic interpretation of such connectives is a good deal more complicated than the corresponding relations in standard textbooks of logic.

The aims of the present book are twofold. The first aim is to provide a comprehensive review of research involving deductive reasoning tasks. The expansion of the field in recent years necessitates this arduous exercise. It is assumed that the reader is familiar with basic work in experimental and cognitive psychology, but the book is otherwise self-contained. In Part II, a chapter is devoted to discussion of basic issues in the study of language and imagery, which are necessary for understanding of the work reviewed in Chapters 3 and 4. Parts II and III each contain an introductory chapter which explains the basics of the logical systems used in the reasoning tasks with which they are concerned.

The second aim is to provide a viable theoretical alternative to rationalist theories of human reasoning, and to relate the study of logical reasoning performance to the main field of cognitive psychology. The Discussion is concerned with this latter objective. In Chapter 11 it will be argued that a major change of approach is needed to the psychology of reasoning. From consideration of the material reviewed in Parts I to III of this book, it appears that there is little evidence for the influence of a general system of logical competence, and that the thought processes involved are highly content dependent. It is also argued that reasoning experiments are best viewed as specialised problem-solving or decision-making tasks, and explanations for various phenomena are offered in line with these considerations. In Chapter 12, attention is focused on the dual process theory of reasoning (Wason and Evans, 1975) and its subsequent development. In this respect, reasoning phenomena are related to diverse work in social psychology, memory theory and the differential function of the two hemispheres of the brain. Whilst some of these later suggestions may be rather speculative, I would maintain that it is for its contribution to cognitive psychology that psychology of reasoning – and this book – must be judged.

Finally, a point about style is in order. Like all writers of the English language, I was confronted with the problem that there are no generic pronouns which are neutral with respect to sex. For stylistic reasons I have retained the old-fashioned ‘he’, ‘his’, etc., in preference to such devices as ‘he/she’, ‘(s)he’, etc. This in no way implies that the people referred to (e.g. experimental subjects) are more likely to be male than female.

Part I

Elementary reasoning tasks

2 Theoretical background

In Part I the focus is on relatively simple tasks involving sentence verification (Chapter 3) and transitive inference (Chapter 4). These are considered first because of their comparative simplicity. Performance on the complex tasks discussed in Parts II and III is subject to h...

Table of contents

• Cover Page
• Half Title Page
• Title Page
• Copyright Page
• Contents
• Acknowledgments
• 1 Introduction
• Part I Elementary reasoning tasks
• Part II Syllogistic reasoning
• Part III Propositional reasoning
• Part IV Discussion
• References
• Name index
• Subject index