Linguistics and the parts of the mind : or how to build a machine worth talking to 9781527502925, 1527502929
415 93 1MB
English Pages 173 [194] Year 2017
Polecaj historie
Citation preview
Linguistics and the Parts of the Mind
Linguistics and the Parts of the Mind: Or How to Build a Machine Worth Talking To By
C.L. Hamblin
Foreword by Douglas Walton
Linguistics and the Parts of the Mind: Or How to Build a Machine Worth Talking To By C.L. Hamblin Edited by Phillip Staines This book first published 2017 Cambridge Scholars Publishing Lady Stephenson Library, Newcastle upon Tyne, NE6 2PA, UK British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library Copyright © 2017 Mrs Rita Hamblin All rights for this book reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without the prior permission of the copyright owner. ISBN (10): 1-5275-0292-9 ISBN (13): 978-1-5275-0292-5
TABLE OF CONTENTS
EDITOR'S PREFACE FOR THE EDITOR WALTON'S FOREWORD AUTHOR'S FOREWORD INTRODUCTORY CHAPTER ONE MACROLINGUISTICS SOME ELEMENTARY LINGUISTICS OF A NEW KIND CHAPTER TWO QUESTION AND ANSWER CHAPTER THREE COMMITMENT CHAPTER FOUR IMPERATIVES CHAPTER FIVE EMOTIONS AND ATTITUDES THE CENTRAL THESIS CHAPTER SIX THE PARTS OF THE MIND SOME ELABORATIONS CHAPTER SEVEN EPIMENIDES THE CRETAN CHAPTER EIGHT MODELLING OTHER MINDS CHAPTER NINE THE PARADOX OF THE MORALISER CHAPTER TEN PERCEPTION THE NATURE OF THE THEORY CHAPTER ELEVEN WHO MAKES THE RULES? BIBLIOGRAPHY AUTHOR INDEX SUBJECT INDEX
EDITOR'S PREFACE Following the publication of his celebrated Fallacies in 1970, Charles Hamblin completed the typescript for this book early in 1971 with light handwritten editing. It did not find a publisher then but material from its fourth chapter "Imperatives" was used in the full length book on the same subject completed just before he died in 1985. His Imperatives (Basil Blackwell) was published in 1987. A considerable time after that the typescript for this book came to light and for a considerable time it seemed it may remain unpublished again. But no longer. In this Preface I aim to indicate the scope of the book by giving a sense of some of its highlights and its overall direction and then acknowledge, with thanks, those who have contributed to its appearance now. This remarkable book is an extraordinary achievement. The subtlety and breadth of Charles Hamblin's thought is on show throughout. His Foreword reveals he had been thinking about a theory of dialogue (dialectic) for at least two decades. The upshot is that this book is a major contribution to philosophy of language and despite his denying being a linguist it also makes a major contribution to pragmatics. Attractively written in an easy conversational style, the book's central theme is engaging in conversation: what goes on as participants speak to each other. In particular, what general kinds of things may they say to each other and to what effect. In his first chapter Charles Hamblin introduces the term macrolinguistics for the branch of linguistics that studies this process. As he says there " this book, I shall say almost nothing about sentence structure. I shall be concerned, instead, with what kinds of sentences there are, and with how they may be put together to make up larger pieces of discourse, particularly dialogue." And he argues there for the primacy of pragmatics over both semantics and syntactics in the study of language. The first major section of the book, "Some Elementary Linguistics of a New Kind", instantiates this approach and consists of four chapters: two focused on indicatives followed by a chapter on imperatives and another on the language used to convey emotions and attitudes (emotives).
Editor's Preface
vii
Charles Hamblin's key starting idea is that in order to be engaged in conversation participants need to be able to keep track of what has been said. He calls the memory devices needed to do this commitment stores (and sometimes tallies) and asks us not to confuse them with belief (or fact) stores. The former are the subject of Chapter 3 on Commitment. In broad strokes, let's sketch what happens with indicatives. A typical use of indicatives is to impart information, to tell people things or make statements to ones conversational partners. Each of the participants is taken to have a belief (or "fact") store and an indicative commitment store. The latter, unlike the former, is public: it keeps a public tally of who has made what statements to whom during a conversation. In an indicative commitment store the contents will be statements a speaker has made or agreed to or acquiesced in. Questions may be asked to prompt statements and there is a macrolinguistic rule that questions be answered. One special kind of question, a Why-question, can be used to ask for a rationale of a statement: asking what reason there is to think it is true. If a conversational contribution can be questioned in this way it is said to be "accountable". During a conversation, the contents of a commitment store (i.e. a tally) need not always increase. Statements may be withdrawn or retracted and a participant can issue a retraction demand to another to encourage this. All of the things participants can say are called locutions: e.g. indicatives, imperatives, emotives, questions, retractions and retraction demands. A significant and to some, surprising, feature of engaged conversations is hearer commitment. Speakers' statements go into their own commitment stores and if those they are engaged with in conversation (hearers) do not object, into their commitment stores as well. Failure to object (perhaps by not saying "No, I'm not so sure about that.") gives consent. In other words, it will in general be presumed in the absence of demurring that hearers have agreed with what has been said. So there are two stores related to this use of indicatives: a more permanent fact store and a much more transient public indicative commitment store that starts afresh, as do all the commitment stores, with each new conversation. Analogous pairs of stores are proposed in Chapters 4 and 5 for imperatives and emotives: act and imperative commitment stores, and sentiment and emotive commitment stores respectively. It is worth drawing attention to how much more comprehensive this framework is than the one used in Fallacies. In that work, there was only one kind of store: indicative commitment stores. Here we have six.
viii
Editor's Preface
I want now to highlight just three instructive discussions made in this section of the book to give the flavour of what is achieved by our author and then comment on something significant woven through the text. Very early in Chapter 2 Question and Answer he displays, in passing, a remarkable example of the macrolinguist's craft in discussing a line from the Pirates of Penzance, Continue fine I hope it may. He writes that it "does rather more than report an internal state of desire whose object is the prolongation of the absence of rain, but expresses a feeling in such a way that someone who agrees may be said to share the feeling and not merely appreciate that the speaker has it." In the context of a conversation someone might agree to this and he points out here one of the consequences of doing so. In Chapter 3 on Commitment he says: "A curious shift of meaning sometimes takes place when a speaker makes a statement but refuses to provide a rationale for it. The statement changes from being the statement that such-and-such is true, to the statement that the speaker thinks, or believes it is true. Often enough the speaker will make this revision himself; and, if he does not, others will do it for him." The latter kind of statement has been called an avowal and has quite unusual properties. It sounds like a description of his mental state but it is stronger than this and commits him to such-and-such being true, yet it does not commit the hearer to this. It lacks hearer commitment to such-and-such being true and also accountability. So failure to provide a rationale when asked can have the effect of weakening the statement you are taken to have made. In Chapter 5 Emotions and Attitudes when discussing the difference between an emotive locution (an example might be a human saying Ow! in a dentist's chair) and a natural cry, (say, the yelp of a dog whose tail has been stood on) after a nuanced discussion and dismissal of a number of differences, he writes "A natural cry, emotive or any other, may be the result of emotion in the being that cries, but cannot be said to commit that being to having the emotion. He or it is not regarded as under any dialectical pressures or obligations, cannot contradict himself or itself or be asked to provide a rationalisation of the cry, as distinct from being asked for its motive or cause. Hearers of the cry, though they may be emotionally affected by it, are not regarded as committed by it to any belief or action, and their failure to react will certainly not be regarded as recognition of or agreement with the emotive reaction. But in the case of a true locution all this is different. Your stylised cry of pain Ow! is intended to have me understand that you intend me to understand that you intend me to understand ... that you are in pain. It goes into your emotive commitment store, and something relevant goes into mine."
Editor's Preface
ix
"This is, then, the difference between a natural cry and a locution: locutions commit, and natural cries do not. Locutions need to be tallied, and there are dialectical rules that take into account the tally." Finally, an especially interesting feature of the book is the recurrent discussion of what would be needed to program a computer to participate in conversations. This addresses the book's subtitle: How to Build a Machine Worth Talking To. It does not get a chapter to itself but appears in most chapters. Alan Turing writing in 1950 made a famous 50 year prediction about how well machines would do in conversations with humans. Writing just 21 years later with 29 years to go Charles Hamblin held a contrary view. (See Chapter One) Acknowledgements First and foremost of the people to thank are Rita Hamblin, Charles's widow, who gave permission to offer this book for publication and Julie Hamblin, their daughter, who provided much appreciated encouragement, advice and support for the project as it proceeded. Thanks also to Douglas Walton for his foreword. And thanks to Cambridge Scholars Publishing's Victoria Carruthers, Theo Moxham, Sophie Edminson and Amanda Millar for their help in this venture. Thank you, to the many people who helped convert Charles's edited typescript to the book you now have. These include David Kinloch for providing optical character recognition (OCR) and in roughly temporal order and helping mostly with word-processing: Alana Cerkesas, Chris Leaney, Laurence Rosier Staines, Shameer Sheik, Jason Chan, Sophia Rosier Staines and Amber Cross. Joining in nearer the end, but with much to do, was Roxane McDonald, deserving a special thank you for her herculean efforts not only on word processing but giving the book further shape by compiling the Bibliography and the active Author and Subject indexes. Thanks as well to Gavin Gayle for putting us in touch. And, finally, thanks to my wife Penny Rosier for her help with the editing and much else. Historical Note: This book is true to the original typescript. This means that some passages contain gendered language more consistent with the time in which it was written than the date of eventual publication. Phillip Staines
FOR THE EDITOR
In May 1985, my father died from a brain tumour at the age of 62. At that time, the ubiquity of computers today was barely imaginable, and the concept of artificial intelligence was unknown except among computer scientists. One of the many sadnesses of my father's premature death was that while he worked for decades on philosophy, computing and linguistics, and their relationship with each other, he did not live to see them come together as they have now done to shape the world. He would have been exhilarated by the advent of the internet, and the current explosion of artificial intelligence would have offered a whole new outlet for his decades of work on language and meaning. The fact that this manuscript from the early 1970s is being published today, so many years after my father's death, is testament to the vision and tenacity of Phillip Staines, his former student who has gone on to become a successful academic in his own right. Following my father's death, Phillip agreed to become the custodian of his papers, and has performed this task of stewardship wisely and conscientiously over many years. It was Phillip who recognised that this manuscript, despite its age, has continuing value and relevance, possibly more than my father could have imagined when he wrote it. The notion of a machine worth talking to (which at the time reflected his quirky sense of humour as much as it did the actual concerns of computer programming) is now a reality of everyday life as we explore both the excitement and the limits of artificial intelligence. My father would be grateful indeed to Phillip for having brought this manuscript back to life in the 21st century at a time when talking with machines is a central preoccupation of the modern world. Were he alive today, he would enjoy the conversation. Julie Hamblin Sydney July 2017
WALTON'S FOREWORD Charles L. Hamblin was Professor of Philosophy at the University of New South Wales until the time of his death in 1985. I do not personally know the circumstances that explain the delay of the publication of Linguistics and the Parts of the Mind for 47 years. I was struck, having read the unpublished manuscript for the first time in April 2017, how amazingly prescient it was. It not only anticipated current foundational developments in so many fields, but many of its ideas and arguments are still challenging and insightful today. Hamblin's writings have had a revolutionary impact on how we think about thinking, arguing and reasoning in multiple fields, most notably linguistics, philosophy, especially informal logic, computer science, and especially artificial intelligence. His thinking has affected these fields so fundamentally by shifting to a different paradigm of how humans and artificial agents can think more rationally and productively together when reasoning with each other in orderly dialogue give-and-take. As Hamblin's ideas began and continue to be incorporated into many fields, especially technical fields such as computing, they came to be expressed in more complex and practical ways. For example, in multiagent computing and artificial intelligence, we now have many formal argumentation systems that model many of his ideas in technically sophisticated ways, and have even in many instances built them into implemented computational systems (Rahwan and Simari, 2009). Hamblin's formal models of dialogue in his book Fallacies (1970) and in his paper on formal dialogue systems (1971) show considerable subtlety, sophistication and technical skill in how they have been presented, but their novel technical aspects pose a barrier for the general reader. An important aspect of Linguistics and the Parts of the Mind is that these revolutionary ideas are explained in a disarmingly simple manner using commonplace examples. Hamblin situates his main contribution nicely in chapter 1, in relation to the current state-of-the-art in philosophy and linguistics at the time of writing of this book around 1971. Both subjects at the time more narrowly focused on syntax and semantics and payed little attention to pragmatics. Hamblin took a breathtakingly forward-looking practical view of language study, holding that pragmatics is an area of it that the theorist
xii
Walton's Foreword
cannot postpone until work on syntax and semantics has made more progress. He looked at things in a different way, seeing pragmatics as being of the essence of language (p. 8). The clue to this different kind of orientation on the study of language, he writes (p. 16) in the introduction, is the observation "that language is "a give-and-take activity". It was this observation that led Hamblin to be primarily concerned with dialogue in his writings. He saw Alan Turing as being more realistic than many of the linguists of his time, because Turing realized that the most important problems of the study of language resided in how one might mechanize dialogue. This puts linguistics in a different light, both at that time and even to the present day, because it had concentrated on the solitary use of language, focusing on what he called monologue, a focus that ignores the pragmatic viewpoint. One of the channels through which Hamblin's theory began to receive greater and greater attention was the book Commitment in Dialogue (Walton and Krabbe, 1995). A quick look at this book shows how much it owes to Hamblin, and how its formal argumentation system was built on the foundations laid by Hamblin. It was mainly through the influence of this book (especially on computer science) that Hamblin's ideas achieved their wider impact on artificial intelligence. As that book showed, Hamblin's seminal ideas reached fertile ground throughout the many strands of argumentation studies which burgeoned in Europe just after 1970. Much of the initial work in this area was on informal fallacies, and it was here that Hamblin's work was widely used in philosophy and especially philosophical logic, leading to greater recognition in these circles of his pioneering application of formal dialogue models to this area (Rescher, 1977). As the field of argumentation studies continued to grow, through the many conferences, workshops, papers and academic books, it has continued to flourish in Europe, and even started to gain a significant foothold in North America, China, and many other countries. A new interdisciplinary field has emerged from this research in computer science, especially in artificial intelligence, multi-agent systems and cognitive science. As well, researchers in informal logic who have concentrated on the structure of argument by analyzing natural language arguments found new ways to model argumentation in real examples. There is a growing army of researchers not only in Europe and North America but all over the world now who are now actively participating in this kind of research. Because this work arose from coalitions between different disciplines, a single widely accepted key word categorizing it has not yet been settled, but it includes such areas as informal logic, computational dialectics, formal argumentation systems, argumentation
Walton's Foreword
xiii
technology and computational linguistics. Research topics in the area are theories of argument, and of dialectic in particular, the design of computational systems for multi-agent communication, rhetorical argumentation models for argument invention tools for debating, the use of formal and computational models for autonomous reasoning in artificial intelligence, and the development of software tools for helping to improve argumentation in domains such as education, medicine, law and political domains, most notably deliberative democracy and egovernment. Many of those familiar with Hamblin's work on fallacies and other topics of interest in philosophy, formal logic, linguistics and informal logic (Hamblin, 1963; 1970; 1966; 1987) may not be aware that he was also a pioneer researcher in the field of computer science. Peter McBurney (2003) chronicles the story, here briefly outlined. The University of New South Wales purchased a DEUCE computer in 1956 that was manufactured by the English Electric Company. Hamblin had a radar background from his service in World War II and took up the problem of computing mathematical formulae using the limited memory of this machine. He proposed a solution of using the so-called Polish notation, which eliminates the need for brackets (Hamblin, 1962). He implemented these and other ideas in a programming language for the DEUCE machine he called GEORGE, for general order generator. Hamblin continued his research in computer science, resulting in a number of published scientific papers (Hamblin, 1985). But after 1960, Hamblin returned to the work in philosophy which had begun with his PhD thesis at the London School of Economics entitled Language and the Theory of Information. Hamblin's book Fallacies was only a small part of a larger project that he was working on prior to 1970, and that continued until the time he died. This larger project is outlined and explained in Linguistics and the Parts of the Mind. Its aim was to find practical technologies that could be used to identify, analyze and evaluate arguments. The method used was to reframe all logical problems about the evaluation of arguments by putting them into a dialogue structure in which arguments are viewed as linguistic interactions between two or more participants who take turns putting forward speech acts such as asking questions and answering them (van Eemeren and Grootendorst, 1992). Of course, as the fallacies book made clear, this dialogue approach seemed highly radical and irregular at the time to many other formal logicians and philosophers of the time even though it was by no means entirely new. It was in fact a development of the dialectical framework of the ancient Greek philosophers (Krabbe,
xiv
Walton's Foreword
2013; 2016), most notably including Plato and Aristotle, as Hamblin made clear in Fallacies. Fallacies was nevertheless seen as a departure from the widely accepted approach to logic (and philosophy generally) in 1970. It was the only comprehensive academic book on the logical structure underlying the formal and informal fallacies since the Middle Ages, and arguably the first book with this level of depth of theory since Aristotle's On Sophistical Refutations (Mackenzie, 2011, 263). But Hamblin's book went against the dominant intellectual paradigm established since the time of the Enlightenment of admitting only two respectable methods for evaluating logical reasoning, deductive logic and probability theory of the statistical kind stemming from the era of Pascal. Now that this dogma is finally starting to fall away, more and more academics are open to looking at Hamblin's achievement with fresh interest. Until 1970, the subject of fallacies had been sadly neglected over the history of logic since the time of Aristotle, with only a few exceptions that failed to excite any concerted interest in the importance of this field. But not long before Hamblin's book appeared, the subject of fallacies had begun to be much more widely treated in introductory level logic textbooks, and seen to be important for helping students to acquire better thinking skills. However, prior to Hamblin the textbooks treated fallacies in a fragmented and superficial way. The Standard Treatment, as Hamblin called it, lacked the coherence of an approach that could be provided by a theory. That started to change just after the time of the publication of Hamblin's book Fallacies (1970), and that in turn led to the rise of the new field now called informal logic or argumentation. The problem was that even approaching the study of fallacies required a new theory, since many of the types of arguments treated under the heading of "fallacies" had instances that seemed from a point of view of common sense to be quite reasonable when placed in the right context. The textbook authors did not apply any such theory (for there was none) to real examples of arguments in a way that could help their students try to figure out which examples of arguments were fallacious and which were not. This state of ignorance (lack of knowledge) abruptly changed when Hamblin's book offered a clear and formally developed theory of a kind that clearly had this application, at least potentially. When the theory was formulated and published (Hamblin, 1970; 1971), although it was widely read and esteemed for its evident depth of scholarship, it departed too far from the conventional approach to logic favored by the dominant philosophy graduate schools in North America and Europe for it to have much impact on the influential curricula of the time. For many, its ideas
Walton's Foreword
xv
were too far outside the mainstream to be acceptable. Also because of their technical nature, these ideas can be hard for beginners to grasp or accept very easily. A main aim of Hamblin's Linguistics and the Parts of the Mind was to explain the theory in a clear and easily comprehensible way by showing how to apply it to the logical and linguistic problems that were current at the time. Interestingly, these problems are still of central interest to those of us working in philosophy, argumentation studies, artificial intelligence, cognitive science, linguistics and related fields. They are perennial problems of rational cognition, and that is why Linguistics and the Parts of the Mind is still useful. According to the current argumentation models used in artificial intelligence, argumentation is defined as having two basic characteristics. The first one is that it contains inconsistencies and needs to tolerate working with inconsistencies, because it is used in a context where there is a conflict of opinions of the kind often called the issue (Prakken, 2007). The issue is made up of a claim put forward by one side, and a situation where the other side either doubts this claim or puts forward an opposed claim that is the negation of the original proposition advocated by the first. Logic in the past has not been very friendly to inconsistency, often identifying it as a kind of error (van Eemeren and Grootendorst, 1992). But dialectical argumentation systems need to at least work with inconsistency, even though inconsistency of commitments is the basis of many kinds of important criticisms of arguments. The second characteristic is that as a sequence of argumentation proceeds the arguments both for and against a claim are considered and the outcome of the dialogue is evaluated by allowing the Pro arguments to interact with Con arguments in a rule governed dialogue exchange. This second defining characteristic could be called the pro-con characteristic of argumentation. These two characteristics precisely represent the notion of argument that Hamblin advocated not only in his published works in 1970 and 1971 but also in Linguistics and the Parts of the Mind. Now the fields of artificial intelligence and multiagent systems, along with other subfields of computer science, have adopted many of Hamblin's key concepts and methods of his dialectical framework and built on them with many advances. Now formal dialogue models of argumentation are being used in agent system negotiation protocols, implementation of software systems to model legal procedures, persuasion dialogue models to structure various kinds of debates, computer-based learning tools that use dialectical argument structures of the kind advocated by Hamblin, for rational deliberation, and models of knowledge engineering (Rahwan and
xvi
Walton's Foreword
Simari, 2009). In particular, there has been recent rapid growth in the fields of artificial intelligence and law (Gordon, 1995; Prakken, 2006), and computerized systems for diagnosis and treatment in medicine (Uphsur and Colak, 2003). There are many regular conferences each year on computational models of arguments, and in many of them argumentation is treated as a standard research topic for submitting papers. This gives you an idea of how widely and deeply Hamblin's ideas have spread, affecting the fundamental logical basis of computer science, now current research efforts in scientific fields have applied these dialectical models. They are still doing that. References Gordon, T. F. (1995). The Pleadings Game: An Artificial Intelligence Model of ProceduralJustice. Dordrecht: Kluwer Academic Publishers. Hamblin, C.L. (1962). Translation to and from Polish notation. Computer Journal, 5, pp. 210-213. (1963). Questions Aren't Statements. Philosophy of Science, 30(1), pp. 6263. (1966). Elementary Formal Logic, a Programmed Course. (Sydney: Hicks Smith). Republished, London: Methuen, 1967. (1970). Fallacies. London: Methuen. (1971). Mathematical Models of Dialogue. Theoria. 37, pp. 130-155. (1985). Computer Languages, The Australian Computer Journal, 17(4) 195-198. Reprinted from (1957) The Australian Journal of Science, 20, pp. 135-139. (1987). Imperatives. Oxford: Blackwell. Krabbe, E. C. W. (2013). Topical Roots of Formal Dialectic, Argumentation, 27(1), pp. 71-87. (2016). The Formalization of Critical Discussion, Argumentation, to appear. DOI: 10.1007/s10503-016-9401-y Mackenzie, J. (2011), What Hamblin's Book Fallacies Was About, Informal Logic, 31(4), pp. 262-278.
Walton's Foreword
xvii
McBurney, P. (2003). Charles L. Hamblin: Computer Pioneer, This Month in Automated Decision-Making, 3(3), March. http://web.archive.org/web/20031224215629/http://www.csc.liv.ac.uk/~ peter/this-month/this-month-3-030303.html Prakken, H. (2006). Formal Systems for Persuasion Dialogue. The Knowledge Engineering Review, 21(2), pp. 163-188. Rahwan, I. and Simari, G. (2009). Argumentation in Artificial Intelligence. Dordrecht: Springer. Rescher, N. (1977). Dialectics. Albany: State University of New York Press. Upshur, R. E. G. and Colak, E. (2003). Argumentation and Evidence, Theoretical Medicine and Bioethics, 24(4), pp. 283-299. Van Eemeren, F. H. and Grootendorst, R. (1992). Argumentation, Communication and Fallacies. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Walton, D. and Johnson, R.H. (2011). Editor's Introduction to the Special Issue on Charles Hamblin, Informal Logic, 31(4), pp. i-iv. Walton, D. and Krabbe, E. C. W. (1995). Commitment in Dialogue. Albany: State University of New York Press.
AUTHOR'S FOREWORD
This book, or at least my preoccupation with its subject, has a long history. I still have a copy of a note I wrote in 1950, in which I told myself that utterances are classifiable by mood, into indicative, interrogative imperative and perhaps others, and by point, according as they raise topic points, points ad hominem or points of order. I have amended these classifications; but I thought then, and still do, that classifications of these kinds, and their consequences for language, have been neglected. Basically what was required was a theory of dialogue. In the following few years, although I worked on communication theory and the logic of questions, the main project was shelved. Then about 1960 I wrote a very bad book on it, called The Testing of Reasons. Its style was a studied chaos of a typically twentieth-century kind, along the lines of Finnegan's Wake or Dali pictures: the rules of dialogue were supposed to be exhibited by being discussed in a dialogue which constantly broke them; but the few people whom I induced to read it found it inscrutable altogether. It was rejected by two publishers, rewritten, and rejected by two more. By this time I had rightly begun to lose faith in it myself as well. Some of the content, though not the style, of that effusion is here reproduced; most of it has been reworked and much altered. None of this material has been published, but about 1965 I wrote a paper entitled "Formal Dialectic and the Parts of the Mind" which I have read on two occasions and circulated privately. It is partly reproduced here in Chapter 6. My interest in the project was sustained throughout the writing of Fallacies (Methuen, 1970), when I found that the conception of "dialectic" as a paralogical study was not merely helpful but in some sense necessary to the correct characterisation of the logical fallacies of the traditional list. This book, one might in fact say, starts where Chapter 8 of that one left off; though its eventual aim is different. I have written down the whole thing here as I now conceive it, without pretence either to logical completeness or to rigour, because it seems to me that there are important points to be made and the matter has gone on long enough. Those who would like greater rigour may find it, over a small part of the field, in my paper "Mathematical Models of Dialogue"; and those who are worried about some of the lacunae may find at least
Author's Foreword
xix
one of them, the dialectic of tenses, discussed in my "The Effect of When It's Said". (Both papers are forthcoming in Theoria. (August 1971, Vol.37(2), pp. 130-155 and December 1970, Vol.36(3), pp. 249-263, resp., Ed)) These, at least, will serve as promissory notes. So many people have helped or good-naturedly hindered this project that I would not know where to start in enumerating them. The influence that it especially occurs to me to acknowledge is one of the temporally most remote, that of my former mentor D.A.T. Gasking. There is not much of him here, but the philosophically acute will recognise it and may even judge it as crucial. I hope he does not shudder at what it has become. Sydney February 1971
INTRODUCTORY
CHAPTER ONE MACROLINGUISTICS
Linguistics is so many different things to so many different people that the ordinary reader must be pardoned if he has doubts whether they all know what they are talking about. To the field linguist, to start with, it is the patient accumulation of grammar and vocabulary, the tape-recording of phonetic nuances, the invention of alphabets and of teaching aids. To his archaeological counterpart it is the decipherment of inscriptions and scrolls. We may remember Browning's grammarian: He settled Hoti's business - let it be! Properly based Oun Gave us the doctrine of the enclitic De Dead from the waist down.
But even Browning thinks these accomplishments "loftier than the world suspects". These classes of linguists are, in their different ways, practical men, and their combined efforts have given us an impressive picture of the range and diversity of human languages. On the other side there are what it would be ungracious to call unpractical linguists, and more polite to call theoretical linguists. (Some of them, these days, call themselves structural linguists, but this is a rather restrictive characterization.) This kind of linguist studies not languages but Language; and often enough his ultimate quest is not even a theory of language but rather a theory of the meaning of life or of human society, or the answers to certain philosophical problems, or the nature of thought or of the human mind. The idea that the study of Language has something to contribute to these apparently irrelevant purposes has been going around for a long time now; not only in this century, but in Descartes and Locke, in the middle ages, and even in Plato and the Stoics. The recent upsurge of interest in linguistics, though it has incidentally contributed to the study of grammar, seems in most of its practitioners to have this kind of source and
Browning, R.B. "A Grammarian's Funeral".
4
Chapter One
motivation. The other unpractical linguists are the philosophers, many of whom make no pretense of linguistic competence but the relation between linguistics and philosophy at present must be very confusing for the casual reader who had understood all along that twentieth-century philosophy was linguistically-oriented, but now finds that linguists have apparently staged another revolution and are making the pace. Actually, what passed as "linguistic philosophy" was not very closely allied to the study of language, but was concerned rather with the application of various preconceptions of language to the study of philosophical problems. This book is itself a contribution to unpractical, and in fact philosophical, linguistics. But I hope it will be found a little less unpractical than some. The two wings of the study are currently too far apart, and theory needs to be tested with practice. If, however, we are to advance the theoretical study of linguistics we must first cure linguists, and the philosophers who advise them, of a myopia that leads them, like schizoid patients looking at Rorschach inkblots, to see only rare detail and be blind to the larger patterns into which this detail fits. And I must hasten to say that I exempt the practical linguist of this charge. The nature of his task inevitably keeps him in touch with reality. But amongst those who have recently written about the theory of language it is hard to find any synoptic appreciation of the processes and purposes of human communication. The reasons for this we shall explore in due course; in short, to concentrate on the grammatical structure of sentences, without considering their use in broader discourse-patterns, is like studying Hamlet not only without Hamlet himself, but without distinguishing one player's utterance from another's, or taking into account speaking-order, place, time and set. And this is what most writers currently undertake to do. For all they tell us, the interchange of linguistic noises by two-legged humans might be a random and rationally inexplicable phenomenon, like the dripping of a tap.2 Consequently, in this book, I shall say almost nothing about sentence structure. I shall be concerned, instead, with what kinds of sentences there are, and with how they may be put together to make up larger pieces of discourse, particularly dialogue. I do this not because I consider grammar totally unimportant, but because I think priority should be given to an attempt to extend the linguist's field of view. The word macrolinguistics
Since first writing this I have read Fries, who does some part of what I here say no one has done. But he is the exception, and he does not do it all. Fries, C.C. The Structure of English, New York, Harcourt Brace and Company, 1952.
Macrolinguistics
5
will serve as a name for this study.3 Now macrolinguistics, as it happens, has its own light to throw on various concerns outside linguistics proper. In fact it is almost axiomatic that so soon as attention is given to the larger purposes linguistic utterances serve, issues and interests beyond those of language become involved. In a later chapter I advance a cautious thesis about the relationship of macrolinguistic categories to theories of the human mind; and in others I shall sketch attitudes to various more-or-less important philosophical issues. But the first part of the book will, I hope, stand on its own feet. And, to start with, something must be said about the present state of the theoretical, "structural" study. Structural linguists often write as if it were possible to give a syntactic and semantic analysis of language in isolation from pragmatic aspects. This is understandable, since if successful they would thereby mark out for themselves a private specialism. Unfortunately it can be done, if at all, only at great strain to the theory they erect. More practical-minded linguists would never in practice agree to this, since they know the limitations of grammatical theory and the extent of the exceptions to it. Linguistic facts are recalcitrant by nature, and forcing only distorts them. (And this is a lesson we must remember throughout this book. My attempts are at best attempts to accommodate linguistic facts better, and are to be judged by their success at this, not by their shortfall from perfection. My criticisms of current linguistic theory are to be read in the same spirit). Thus most of the sins of the structural linguist are of omission, not commission. But there is one respect in which considerable naivety has been displayed, namely in the matter of meaning. The modern distinction between syntactics, semantics and pragmatics is due to Rudolf Carnap, and was conceived as a distinction between three levels of rules for a system of logic: syntactics gave the rules of sentence-construction, semantics the rules determining logical truth and falsehood (in what were predominantly "indicative mood" sentential formulations), and pragmatics the rules of use of these sentences in supposed actual communication. In fact pragmatics was never very clearly defined compared with the other two, since the question of use was one that Carnap never particularly got round to considering. Even as an approach to a formal system of logic, this
I owe it to Dr Robert Dakin of the University of Papua and New Guinea, who used it in introducing a paper I read. 4 Carnap, Rudolf The Logical Syntax of Language (1934), translated by Amethe Smeaton. London, Routledge & Kegan Paul, 1937.
6
Chapter One
has its limitations; and when we turn to the study of natural languages it is not at all clear that to ignore pragmatics - that is, to ignore the rules of actual use of the expressions of the language - is not to throw out the baby with the bath-water. It could in any case be doubted whether it is possible to treat syntactics and semantics separately. Many writers on structural linguistics profess to do just this: their books are wont to contain block-diagrams of the process of generation of a grammar, in which one block is labelled "semantic element", perhaps associated with a stage of the process called "lexical insertion". Different writers favour different block-diagrams but the general idea is the same: rules of grammatical structure give us sentenceblanks which can be filled in in various ways with meaningful words to make fully-meaningful sentences. Thus consider the analysis of a sentence such as John hit the ball:
Fig 1-1
(I ignore the tense of the verb.) The upper part of the tree - if it is considered as a scheme for the generation of the sentence - can be constructed independently of what words are finally inserted at the ends of the branches. I could have written Joan instead of John, threw instead of hit, and so on. The answer to this sort of theory of grammar is that it fails to deal with what is usually called "idiom". In fact an idiom is commonly defined as a group of words whose meaning cannot be deduced from the meanings of the individual words. Thus consider the sentence Bill kicked the bucket:
I owe some enlightenment to a paper by Alan Healey entitled "English Idioms", in Kivung, vol.l, No.2 (August 1968); and to discussion with its author.
Macrolinguistics
7
Fig 1-2
The structure is apparently the same as in the previous example, and "lexical insertion" produces it from what is otherwise the same tree; but the semantic properties of this sentence are not to be equated with those of the previous one. The phrase hit the ball refers to someone's act of hitting something, but kicked the bucket does not refer to anyone's kicking anything. (And it does not permit the same transformations: it cannot, for example, be put into the passive, The bucket was kicked by Bill.) In saying that an idiom has a meaning different from the one its structure might indicate we mean that it has a use that deviates in appropriate ways, from the uses of its component words. Deviance implies a norm, and it may well be that structural linguists more-or-less correctly describe the norm. Nevertheless, it is important here to emphasise that use, not structure, is the criterion of meaning. To take even simpler examples: if I meet you and say Hallo no amount of structural analysis of this unstructured utterance will tell you what it means; and if, alternatively, I say How do you do?, a structural analysis could be positively misleading even concerning the category of the utterance, suggesting that I have asked you a question. (How do I do what? you would answer.) Examples of idiom could be multiplied. Noam Chomsky, one of the fathers of modern structural linguistics has repeatedly used an argument against this contention, to the effect that since the potential of language is infinite - and, in particular, because a speaker of a language is apparently capable of using new sentences, meaningfully, without previous experience of their meanings - knowledge of a language must be knowledge of a set of structural rules rather than of a number of prototype utterances. Properly understood, this need not be inconsistent with what is here said: it tends to miss the point, which was made succinctly for us by Lewis Carroll through the mouth of Humpty Dumpty. The question is, who is to be master, that's all. It is true that, when a
8
Chapter One
sentence is used for the first time, one's tendency is often to allot it a meaning in accordance with a set of grammatically oriented habits. A foreigner or a child hearing Bill kicked the bucket for the first time might do this, and ask "Which bucket?" But, first time or not, the use of the sentence could have been idiomatic, in departure from grammatical expectations; and if so, that would have been the determining factor in its meaning. Idiom, whether it is actually the rule or the rare exception, could permeate language, and to see whether any given locution obeys the meaning-rules of the structural linguists we need to examine how it is used. A modern structural linguist, particularly one of the "general semanticist" school would, perhaps, be quite unperturbed by our account of How do you do? and Bill kicked the bucket. He would say that How do you do?, properly considered, is a one-word indivisible greeting; and that kicked the bucket is a special kind of irregular verb that differs from most verbs in having spaces in its spelling. That is, he would make ad hoc exceptions in all the difficult cases. But I think this is cheating. Already in calling How do you do a greeting he would be making a point about its use; just as he would be if he were to call it, incorrectly, a question. And he is making a point about use even if he calls John hit the ball a statement, for he would be saying that it is, or is properly, used to state a fact. Statement, question, greeting these are, at least in typical cases, categories that indicate features of use of the utterances concerned. I do not want to suggest that there are no complications to this picture: that we never, for example, use a question to make a statement ("rhetorically"), and so on. But by origin, and ultimately, even grammatical distinctions are distinctions of use. Consequently pragmatics, far from being a study the theorist can postpone until he has finished with syntax and semantics, is of the essence of language. And one effect of the tendency to ignore the practical use of language has been the tendency to concentrate on smaller linguistic units rather than larger. If I think that the difference between a statement and a question is that the first is designed to impart information and the second to elicit it, I shall be encouraged to look at questions and statements as units in a situation of linguistic interchange between speakers. If, on the other hand, I think that the difference is mainly one of word-order and special idiom, I shall tend to spend my days studying minutiae of
Generative semantics is expounded in McCawley, J.D. "The Role of Semantics in a Grammar", in Universals in Linguistic Theory, edited by Emmon Bach and Robert T. Harms. New York, Holt, Rinehart and Winston, 1968. For a comparison with his own theory see Katz, Jerrold J., "Interpretative Semantics vs. Generative Semantics", Foundations of Language 6 (1970), pp. 220-59.
Macrolinguistics
9
grammar. Most present-day linguists, it seems, take the latter view and course. This leads them to neglect an important and interesting study, that of macrolinguistics. Now let us return to the philosophers. Twentieth-century philosophy has been concerned above all with theories of meaning. Here I should remark that of the two theories I have been contrasting - the "use" and "structure" theories - the second is not really a complete theory as it stands, but needs to be supplemented with a theory of the meanings of individual words; for example, a realistic "naming" theory, or a theory that sees the meanings of words as ideas in the mind of the speaker. But I shall not enter on discussion of these alternatives. The important division between philosophers, for our purpose, in the past seventy years has been between (a) those who saw language as perfectible, and the philosopher's task, in part, as promoting its perfection Russell, the early Wittgenstein, Carnap and most philosophers interested in formal logic; and (b) what came to be known as the "ordinary language" philosophers, who generally believed, with Wittgenstein after about 1930, that everything in language "is in order as it is" . Proposals for the reform of language, or even for the construction of a "logical" language, were regarded as misplaced. Although philosophical problems arise through misunderstandings about language we must blame not language, but ourselves. Even logicians may contribute to these misunderstandings, by importing fixed ideas of the job language ought to do. In reality, language is a diversified tool with a variety of purposes. "Colloquial language is a part of the human organism, and is not less complicated than it."
And when we look at logical tradition - or even at recent logical writings - there is no doubt that the nature of language is gravely and strangely misrepresented. For centuries it had generally been assumed that the business of language was to state "propositions", namely, indicativemood tenseless fact-stating true-or-false-but-not-both entities capable of standing in a limited number of logical relations to one another and of figuring in inferences. The growth of modern logic, in the work of Boole,
There was no sudden reversal of Wittgenstein's views. What he thought about ordinary language was already stated in the Tractatus, which he wrote in 1917. But in the Tractatus he had also tried to define a "perfect language", and this element disappears from his later work, in a major change of emphasis. Wittgenstein, Ludwig, Tractatus Logico-Philosophicus (1922), translated by C.K. Ogden. 2nd ed., London, Paul, 1947; and Philosophical Investigations, with a translation by G.E.M. Anscombe, Oxford, Blackwell, 1953.
This quotation is also from the Tractatus, 4.002.
10
Chapter One
Frege and Russell, though it enlarged the range of forms these propositions were permitted to have, did not basically challenge our conceptions of them. In the past thirty or forty years there has been some whittling away of these conceptions by writers who have produced logics of questions, or of imperatives, or of other more esoteric forms of language. But even yet there does not seem to have been any logical writer to whom it occurred to survey the forms language may take, and to produce a synoptic logical theory of those forms. Instead, there have been attempts to minimise the diversity of the forms, by reducing them to one another or to a common form. Thus theories of questions have been put forward that present questions as equivalent to statements; that is, in effect, to propositions. 9 Or ink has been expended to show that the logic of questions and imperatives is really just the logic of statements and that the person who asks a question or issues an order is still dealing with propositions but, as it were, doing something different with them.10 Russell once asserted that all language is really imperative, since when I speak to you it is because I want you to do something about it; overlooking the variety in the kinds of thing I may want you to do. And even J.L. Austin, who would normally be classed as an ordinary-language man par excellence, tried at one time to argue that all utterances are "performative", a category he had invented himself primarily to deal with the special idiom of promises. There are other examples, too numerous to mention. Wittgenstein had made short work of dismissing these tendencies: If you do not keep the multiplicity of language-games in view you will perhaps be inclined to ask questions like: "What is a question?" - Is it the statement that I do not know such-and-such, or the statement that I wish the other person would tell me ...? Of course it is possible to substitute the form of statement or description for the usual form of question: "I want to know whether ..." or "I am in doubt whether ..." but this does not bring the different language-games any
See Chapter 6, below. See, for example, my discussion paper "Questions Aren't Statements", Philosophy of Science 30 (Jan.1963), pp. 62-3. 11 Russell, Bertrand, An Inquiry into Meaning and Truth, London, George Allen and Unwin, 1942, pp. 26-7. And a variety of cognate idioms: I return to performatives in a later chapter. Austin's earliest treatment of performatives is in "Other Minds", Proc. Arist. Soc., Supplementary Volume XX, 1946; reprinted in Austin, J.L., Philosophical Papers (J.O. Urmson and G.J. Warnock, eds.), Oxford University Press, 1961.
Macrolinguistics
11
closer together.
The contention is clear: attempts to reduce linguistic forms to one another, or to a few simple forms, are nugatory because of the infinite diversity of possible uses of language. But although I have cheerfully subscribed to Wittgenstein's "use theory of meaning" I shall equally cheerfully part company with him on the question of the non-perfectibility - or it might be better to say, the noncharacterisability - of language. This is as much on methodological grounds as on doctrinaire ones; for in the long run, ordinary-language philosophy is negative in effect (not to say self contradictory) if it denies the possibility of any theory of language. The attempt to see ordinary language clearly, in all its diversity, was overdue; but if, having made the attempt, we provide no description of what we see, the cynic might be justified in suspecting that we have seen nothing. Language, in short, is not infinitely various, and even Wittgenstein did not really think so. There is, at least, a family resemblance between the various kinds of behaviour we would be prepared to call linguistic; and it should be possible more or less exactly to survey the extent of the family. It is for this purpose that macrolinguistics needs to be invented. Among the various studies that have contributed to present thinking about linguistics we should now consider one other, namely, automatic computing and the growth of mechanisation. In 1950 Alan Turing, who had contributed "Turing's theorem" to the theory of automatic computing some years before, wrote an article in which he envisaged computers taking part in ordinary conversation with human beings. He claimed that: In about fifty years' time it will be possible to programme computers ... to make them play the imitation game so well that an average interrogator will not have more than 70% chance of making the right identification after five minutes of questioning.14
According to Turing, machines will eventually be able to have so complete a command of English and other natural languages as to be able to speak them like natives. Whether or not we equip them with realistic body, voice and sense organs, they will be able to converse with us by some means, such as by teleprinter, in such a way as to be indistinguishable from men
Philosophical Investigations, §24. Turing, Alan M., "Computing Machinery and Intelligence", Mind 59 (1950) pp. 433-60.
12
Chapter One
similarly conversing. They may even, he suggests, write poetry. Since that time machines have become much bigger and faster than they were, and have taken over many complicated tasks for us. But their performance in the field of linguistics has been unimpressive compared with what Turing might have hoped. In fact existing machines are probably large enough and fast enough to speak the English language, if only we knew how to program them! Language is so subtle a tool, its laws so complicated, that we have no idea what shape an overall theory of it would be. And research in linguistics cannot be hurried as engineering developments can. Some thirty years remain to us before we can definitely say that Turing's prediction has been falsified, but it is just worth venturing the opinion that this is, in due course, what we shall have to say. Turing, though a mathematician by background, had studied under Wittgenstein at Cambridge. In his later years he undertook a number of projects aimed at producing signs of human intelligence in machines. He committed suicide, after a period of depression in 1954. I do not know whether the lack of co-operation of the machines he was working with contributed to his death. Modern linguistic studies are considerably more abstract and machineoriented than older ones. Yet there is another kind of machine-oriented work that is still more practical and down-to-earth, and that is the day-today work of machine programmers themselves. When someone has a problem he wants solved by a computer or more usually not a problem but a set of routine operations he wants carried out on data he has a practical linguistic problem of a sometimes interesting and difficult kind, that of how best to get down in precise and unambiguous language the specification of what he wants the machine to do. The precision and the freedom from ambiguity are the central requirement here: the fact that the actual language he uses must be one that can be fed directly into the machine is relatively unimportant theoretically, since once a problem has been stated precisely in some language or other it is usually a simple matter to translate it into a different one this is simply one of the things we mean by "precision". Consequently computer programmers have a more practical outlook on language than computer theorists. No doubt many of them like us all are prisoners of their limited experience but, such as it is, they know what they are about. There is, moreover, a moral in this: theories of language, these days, ought to be not merely programmable which only says that they should be clearly formulated but such that we can see some precise and preferably useful aim in programming them. A good test for a linguistic theory is: would a programmer use it? I do not pretend that it is the only test.
Macrolinguistics
13
Computer-oriented studies of language, however, tend to be infected with a tendency partly parallel with the one we have been stigmatising in logicians and structural linguists to see language too formalistically. A sentence, says the computer-theorist, is primarily a sequence of symbols; and grammar is the study of the properties of symbol sequences. The process of communication is the process of generation of a sequence of symbols by some device, their transmission to some other device, and their absorption, with some consequent effect, on arrival. Translation between languages is a process of coding, whereby one of a family of symbolsequences is replaced by a corresponding one of another family. The information contained in a sentence can be measured in terms of the statistics of the symbols in which it is expressed. And all these linguistic concepts can be studied formally and mathematically, without reference to meaning. In objecting to this trend, one must not commit the contrary error of minimising the engineering importance of some of these concepts. But object we must. Language is for people, and computers are only an extension of people: people have all along used various extensions of themselves, such as pens and paper, as aids to linguistic communication, and computers are glorified pens and paper. So soon as we begin to think of the goings-on inside computers as linguistic processes themselves, apart from what they mean to people, we are talking not about language but about something else that should not be confused with language. This does not mean that this other thing should not be studied, or that no good can come from its use; and it does not mean that what is dealt with inside computers is not language, for it well may be. But if it is it is because it means something, or is capable of meaning something, to the computer operator or his client. And what that means is that it must be ultimately capable of being translated into the operator's or the client's own natural language. We must always come back to natural language in the long run. We are sometimes tempted to speak of the existence of a language used only by machines. This temptation is particularly strong in the case of machines which ape human beings in their behaviour and, perhaps, perform tasks hitherto performed by humans. For example, an automatic weather observatory each day at an appointed time sends radio messages to its master station, to which is attached a computer that collates the readings. We say it "sends messages", and perhaps that the master station "interrogates" it or "sends instructions", but these terms are used not so much because there is anything intrinsic in the radio waves that marks them as informative, interrogative or the rest, but because these are the functions the radio transmissions ultimately serve for us. The weather
14
Chapter One
station is a reaction device that undergoes various changes of state while relays click, currents pass in wires, and so forth in accordance with known scientific laws; but all sorts of things undergo changes of state without thereby being regarded as purveyors of language. It is quite possible that the configuration of the water-droplets in the thundercloud that is at present hovering over my house could be regarded as a code of the United Nations Charter - and, in fact, in a trivial sense, it must be possible so to regard it - but that is not enough for it to be reasonable for us to see the cloud as expressing or conveying internationalistic sentiments. A certain kind of formal structure or diversity is at most a necessary condition of linguistic symbol-sequences, not a sufficient one. As technology gets more and more complicated we shall probably build communication devices that are ever more automatic and remote from human operation, so that anthropomorphic language will become ever more natural to the engineer; but none of this alters the fundamental fact that they would not be communication devices at all if it were not for our own interest in their signals. This stipulation can be carried a step further. Language, as we ordinarily understand it, is for communication between people, and is not properly at home when used by only one person. This does not mean that a person is not really using language if he speaks to himself, but it does mean that the language he uses on such occasions is what it is because it is in principle capable of being used for communication within a social group. This point was originally made by Wittgenstein15 and has been much debated by philosophers, not all of whom would agree with what I say. However, I think the most debatable part of the contention concerns how we should describe the activities of a man who keeps a secret diary consisting of a set of marks of a meaning known only to himself. Wittgenstein would say that the secret marks, unless they are merely a code for a known language such as English, are strictly speaking without meaning and serve only as reminders to the writer to remember something, as if he had tied a knot in his handkerchief. What he remembers is up to him. Marks in a book may, of course, have another function, namely, as aids to calculation, as when we use them to help us do complicated arithmetic; and it is perhaps an unfortunate red herring that the modern electronic machine is called a "computer" and was originally used primarily for this purpose. But let us, then, consider the case of a programmer who invents a language for his own sole use in communicating with a computer. Is it really a language, and does he really communicate? We might be disposed,
Philosophical Investigations § 256ff.
Macrolinguistics
15
even if we agree that this is not a true case of communication, to insist that he does use a language. But that is because what he feeds into the computer and accepts out of it could in principle be understood by other programmers and, presumably, since computer operations are all very precisely defined, be easily translated into other possible computer languages. On the other hand, there is no contradiction in supposing that some part of what he feeds into the computer has a meaning only for him, and that he uses the computer in the same way as one might use a secret diary; but then we can hardly say that the computer "understands" the language either, and the analogical "communication" between programmer and machine breaks down. If a mechanical language carries "information", that must be because, in some remote sense, it has the purpose - or one should rather say, the diversity of purposes of "informative" expressions and idioms of ordinary, natural languages such as English. And if it conveys "instructions", that must be because it does duty for instructions that could, in principle, be expressed in English; and so on. It follows, in one sense, that an adequate theory of natural languages is also a theory of all possible artificial languages. This is a statement that must be made with caution, since it is possible that artificial languages might be designed that are markedly different from natural languages in all sorts of surprising respects. The effect their invention had might be to superannuate prevailing theories that were satisfactory in their application to natural languages. Nevertheless, for an adequate theory of natural languages, the statement holds. The project of designing a good, universal theory of language that would apply to all natural and artificial languages has something in common with attempts to design a universal language for human use. There have been some hundreds of these over the years, at least if we include partially-constructed ones like Leibniz's characteristica universalis: the best known is, of course, Esperanto, designed about 1887 by L.L. Zamenhof. There is one function such a language could have, if it were properly designed: it could be an aid to mechanisation. In some sense we must design such a language in the process of developing a satisfactory linguistic theory applicable to machine languages, and those who are concerned with the day-to-day use of computers are doing so in rudimentary ways every time they use a new programming technique. Vannevar Bush said as long ago as 1946 that it was curious that no one concerned with the design of a universal language had particularly had machine use
16
Chapter One
in mind, and that "mechanisation may yet force the pace".16 As in the problem of realising Turing's imitation game, mechanisation has been singularly unable to force the pace as might have been wished: what is standing in the way is our ignorance of the structure of language. But again, it is not the grammatical detail of language that especially needs attention here, but rather the filling in of the picture of how language is used and acquires its meaning; that is, of macrolinguistics. I must now begin to descend from generalities to the kind of detail that will substantiate and fill out this claim. The clue is, as I have already said in other words, that language is a give-and-take activity. It is for communication between people. It has a job to do, and to understand it we must understand the job. And the best way of doing this is to see our theory of language primarily as a theory of how this job is done. This means that grammatical features of the kind that determine sentence structure are of only secondary relevance to linguistics. I shall, in fact, ignore them almost completely in this book, and concentrate instead on features that are nearly ignored in the other books. My principal concern will be to see pieces of language as a whole, in the context of their use, and to relate them in their context to one another. This means, primarily, that I shall be concerned with dialogue. Here Turing was more realistic than many modern linguists, since he saw that the generality of the linguistic problem resided in how one might mechanise dialogue. At first sight it seems, of course, that there are linguistic contexts that do not depend on dialogue at all: the text of this book, for instance. But we shall see that, just as the solitary use of language by one speaker is, at best, a limiting case of its general use for communication between members of a family of speakers, so that what we might call "monologue", the use of language in a long, connected passage in a speech or in writing a book, is a limiting case of its use in conversation. The writer has, at least in principle, a reader, and he cannot write without considering and, in fact, knowing something about that reader. Moreover, it would be a mistake, at least of method, to make too much of a mystery of the communication bond, such as by dismissing it with an abstract label like "empathy". Up to a point, I can be perfectly definite about the aims and conditions of satisfactory communication and the functions of the individual locutions concerned in it. Perhaps we shall never succeed in mechanising the sophisticated dialogue that is possible
16
Bush, Vannevar, "As We May Think", Atlantic Monthly, 176 (July 1945), pp. 101-8.
Macrolinguistics
17
between humans, but it would be silly to be too sure either way; and there is much communication even of a day-to-day nature that is not of this character. We start, as usual, with the easiest cases but they are easy cases of a different kind from those that are studied in most of the books on linguistics. If these investigations have any success then, besides those who are pleased at the prospect of making machines behave more like people there will be those who are offended by what they take to be yet one more step towards exhibiting people as behaving like machines. Perhaps, they fear, we shall eventually all be compelled to speak in mechanical dialogue and accused of solecism if we do not conform. I do not think they need worry. In the human field, as fast as we invent theories we find that people manage to behave in ways inconsistent with them, simply by understanding and climbing above them. This means that theories in the behavioural sciences, amongst which we must in the long run include linguistics, have a special kind of lack of finality.
SOME ELEMENTARY LINGUISTICS OF A NEW KIND
CHAPTER TWO QUESTION AND ANSWER
The unit of meaning or of logical discourse is often thought to be the statement of fact. I state a fact when I say The cat is on the mat or Roses are red. Some such statements of fact are true and some false, though when we call something a "statement of fact" we usually imply that it is true. But although we commonly make statements of fact, true or false, to one another as we go about our daily business, we also say many things that are not statements of fact, such as Good morning, How much are the apples? or Could I have the salt, please. Statements of fact really occur in quantity and undiluted only in rather special circumstances, in encyclopaedias, books of statistics, telephone directories and timetables, and even then they are commonly in some sort of reduced code rather than in plain language. Perhaps there are sometimes dialogues consisting mainly of an exchange of facts by two or more speakers. But we should be careful to see that we know what is involved in this. Some broadcasting stations have news programmes in which two news readers alternate in the reading of items: one says This morning 26 people were arrested during a riot outside the Transalbanian embassy, and then the other says Gloom descended over Stock Exchange circles with the announcement of a rise in the price of imported tungsten. These may or may not be facts, but it is easy to see that the two speakers cannot be regarded as being engaged in a dialogue. At a low level of dialogue we often do exchange facts about, for example, the weather when we have nothing better to do. W.S. Gilbert, in The Pirates of Penzance, has Major-General Stanley's daughters get over an embarrassing moment by chattering together How beautifully blue the sky! The glass is rising very high. Continue fine I hope it may, And yet it rained but yesterday ...
This is about as inane a level of fact-stating as we can imagine; but even here, it should be noted, non-fact-like elements creep in. How beautifully
22
Chapter Two
blue the sky! is not merely the statement The sky is blue but an exclamation, and Continue fine I hope it may does rather more than report an internal state of desire whose object is the prolongation of the absence of rain, but expresses a feeling in such a way that someone who agrees may be said to share the feeling and not merely appreciate that the speaker has it. Perhaps a Pinter play would provide better examples. Even monologue, as in a printed book, though it often consists mainly of sentences in the indicative mood, is always more than just a string of statements of fact. The first requirement on the string of locutions in a connected discourse, monologue or dialogue, is that the locutions should be mutually relevant. They should, broadly, be on the same subject or topic. The topic of an actual discourse may change from time to time, but there is a flow or development from one topic to another. Of course, there may be sudden interruptions too, as when someone says Oh, that reminds me, I haven't seen Fred recently or Look out, there's a spider on your shoe! But these are exceptions, and we can even say of them in extreme cases that the discourse itself has been interrupted. In any case the subject or topic of a given phase of discourse is one of the things that binds it together. But what is a "subject" or "topic"? We usually specify it either by saying that it is about a particular sort of thing, or that it is concerned with answering a particular question. The Walrus and the Carpenter talked Of shoes and ships and sealing-wax, Of cabbages and kings
but also Of why the sea is boiling hot, And whether pigs have wings.
A discussion about shoes or ships or sealing-wax is a discussion about a particular sort of thing; a discussion of why the sea is boiling hot or whether pigs have wings is concerned with answering a question. On reflection we may come to think that the second kind of topic is perhaps the more fundamental, since even a discussion about a particular sort of thing may not be very homogeneous unless it is aimed at answering a particular question about it. So we should start by considering questions. But as soon as we consider questions it is clear that we are, at least obliquely, considering dialogue, even when the discourse we are concerned with is a connected piece of fact-stating monologue. If the text of a book is a unity, it is because it aims at answering a particular question,
Question And Answer
23
or group of questions, in the mind of a reader. Often enough the author, it is true, introduces his own questions, and they are such as not previously to have occurred to a given actual reader; but then, the author is conversing with himself, or with an imagined reader. Questions, at all events, are really at home in dialogue, and if we want to consider the simplest cases we must go to dialogue to find them. So let us consider question-andanswer dialogues. I do not want to suggest that the question a whole discourse is aimed at answering is to be equated with any individual question that occurs in the discourse. We are playing with two different meanings of the word "question", a lower-level and a higher-level one. Nevertheless even the introduction of the lowest-level kind of question into our considerations puts us some way towards providing an account of relevance, since it gives us one low-level model of it - the relevance of a statement to the question to which it is an answer. Logicians have written a lot about the classification of questions. The simplest questions are the yes-no variety, like Is it raining this morning? or Have we any coffee left? Often enough the verbal form of yes-no questions is little different from that of statements. In English we usually (1) change the order of subject and verb, and (2) put a question-mark at the end of the sentence or, in speech, introduce a characteristic raising of the voice on the last word or two. But we sometimes dispense with these special signs. In some languages, such as Italian, there is no change of word-order, and in others, such as Chinese, which has a regular tonal system of pronunciation anyway, rises in pitch are not specially characteristic of questions. The Chinese and Japanese indicate a question by putting a special particle, na or ka, at the end of the sentence, like an audible question-mark. But the important thing for us to notice here is that the main part of a yes-no question is no different from the main part of a statement. It is its use that is different. Before going on to discuss use, let us look at the other main kind of question. This is the W-question, such as What happened about yesterday's lecture?, When do you expect Jean? or Where's the poker? These use the special interrogative words enumerated in the rhyme: I have six honest serving men. They taught me all I knew. Their names are What and Where and When And How and Why and Who.
Not to mention Which. W-questions generally start with one of these. It is a little disturbing for the English purist that the w in how is at the end of
24
Chapter Two
the word rather than at the beginning: Germans, as usual, are more thoroughgoing with was, wo, wann, wie, warum, wer and welcher. Wquestions are not only different in verbal form from statements and yes-no questions but usually cannot be answered in any so simple form as by Yes or No. Instead, they demand a choice between an unspecified number of alternatives of a kind indicated by the interrogative word: times in the case of When, places in the case of Where, people in the case of Who, and so on. How forms important compounds such as How much and How many, which demand quantities. Chomsky, in Syntactic Structures (Ch.7) gave a transformational account of questions which has had some currency. According to this account, yes-no questions are generated by transforming the corresponding indicative statement, namely, by inverting subject and verb, if the verb is one of a certain short list, or otherwise by introducing do, does or did as an auxiliary. And W-questions are generated by means of two transformations carried out in sequence: (1)
An attachment-transformation, which permits us to attach "Wh" to any sentence, thus Wh + Sentence;
and (2)
A question-forming transformation, thus Wh Who Wh + non- What.
The account is incomplete, since it does not enable us to form W-questions with why, where, when, how and which. Not that that is a serious objection to it since it could be supplemented. A more important objection is that it does nothing to elucidate the logical relations of different questions with one another, or with statements. And it is conceptually wrong, if it seeks to reduce questions to statements or to suggest that a question can mean the same as some "underlying" statement. At best it demonstrates that there are grammatical connections between questionforms and statement-forms. The clearest general characteristic of questions is that they set out a family of alternatives. If I ask Whose sandwich is this?, when there are
Question And Answer
25
five of us, A, B, C, D and E, at a picnic, there are five possible answers, It is A's sandwich, It is B's sandwich, It is C's sandwich and so on, and the true answer to the question is to be chosen from among these five. Of course, it is possible that it is no-one's sandwich, or that Billy and Doris have been sharing it, but these are deviations we shall do better to deal with in another way. Consequently questions are logically different from statements of fact. The yes-no question Is it raining this morning? differs from the statement It is raining this morning not merely in details of structure and pronounciation but because it sets up a choice-situation between two statements It is raining this morning and It is not raining this morning. And the person who asks a question, unlike the person who makes a statement, hovers (at least for the moment) between the alternatives instead of coming down definitely on one of them. Some structural linguists, noting similarities between the grammatical accompaniments of questions and those of negative. sentences, conclude that questions "have an element of negation in them"17. For example the word any can occur both in the question Did you find any coffee? and in the negative I did not find any coffee, but is ungrammatical in I found any coffee Similarly constructions with do, does or did (except for emphasis). But of what importance is this observation? Let us accept that it may be useful in psycholinguistics, or to the student of linguistic history. It applies rather narrowly to English and some closely related languages; not clearly to French, not to Scandanavian languages. And it is capable of being quite misleading - a blind alley - for the logician; or macrolinguist! But even this is not the most important difference between a statement and a question. Questions do not merely set up families of answers; they are also, properly, addressed to someone who is expected to answer. At least if a question does not demand an answer it is of a special and atypical kind known as "rhetorical". If I ask you Did I lend you my copy of Roget? you may shrug or change the subject, but my intention in asking you was
17 The facts in question though not this expression of them are in Chomsky, Syntactic Structures, see p. 64.
26
Chapter Two
that you should answer me with a Yes or a No. This, at any rate, is the linguistically proper thing for me to intend, and for you to do: we shall distinguish later between the standard or explicit function of a locution and the various non-standard functions it may serve on particular occasions. There is, as it were, a linguistic rule: Questions should be answered. When A addresses a question to B, B should forthwith address an answer to that question to A. So expressed, the rule is a little too tight, since there are situations in which it would be proper for B to do something other than provide one of the answers laid out by A's question: he may have to say I don't know, or he may have to point out some feature of A's question that makes it impossible or misleading to answer directly. But these are the only exceptions. If you ask me a question and, for no reason, I say nothing and walk away, or make a remark about something else, I have broken one of the rules of language. I have, as it were, behaved ungrammatically. It is interesting in passing to note that logicians have sometimes praised the virtues of silence. Silence, it has been said, is the perfect logic. Presumably this stems from the theory that all language is imperfect and that, in consequence, everything anyone says is in some measure logically perverse. The view is confused since it does not distinguish between logicality and truth. However, we can now see that it is also quite unsound. Silence is the perfect logic only so long as no one asks you a question. Another thing that typically distinguishes statements from questions is the state of mind of the speaker. Here, and for the next few paragraphs, I shall adopt another simplifying assumption about questions, namely, that they are all of the kind that might be called "information-seeking". Some kinds of questions are not like this, such as the questions written on an examination paper, which are designed as a test of ability; or, sometimes, questions that are intended to force the answerer to admit something that the questioner perfectly well knows he knows. So far as informationseeking questions are concerned, the questioner and answerer are generally in quite opposite positions as regards their state of knowledge on the matter concerned in the question. The questioner does not know the answer to his question, or he would not ask it; but he thinks the answerer knows it. If he is right, and the answerer does know it, the answerer gives it because he knows or thinks the questioner does not know it, and wants it. And under these circumstances it is normally proper for the questioner to accept the answer given. If he says That's not the answer I want, you are answering me incorrectly the answerer may reasonably be peevish and say You shouldn't ask questions if you are not going to like the answers you get. At least this is so in the case in which the original question is of a sufficiently simple form, so that the range of possible answers was or
Question And Answer
27
should have been understood between questioner and answerer. We are working up to a model of a two-person question-asking-andanswering situation, which will involve specifying not merely a grammar of questions and their answers, but also how the questions and answers are related to the states of mind of the participants and their influence on one another. At the moment, we are doing this in a very rudimentary way, but we shall achieve greater logical sophistication in due time. Let us suppose that we have a group of people, whom I shall call P, Q, R in an information-seeking question-and-answer conversation. I shall suppose that each has in his possession a store of information, and that each is interested in adding to this store by drawing on information possessed by the others. I shall also suppose that the total store of information cannot increase, being, say, on some subject such that there is no possibility, during the conversation, of the participants' discovering new facts not previously known to any of them. That is, we have an empirically closed situation. We can think of the various participants' stores of information, if we wish, as something like the stores of electronic computers; or, it might be better to say, as sheets of paper with sentences written on them in precise ordinary English.
Fig 2-1
One of the things that studies of the concept of information have shown us in modern times is that there is no difference of principle between these suppositions, in the sense that a precise language can be reversibly coded into all manner of different forms. It is even necessary, if we are to be able to answer questions in the way we do, for our own brains to contain some remote analogue of the pieces of paper or the computer store. The information, in a precise technical sense, must be there somehow, and we can
28
Chapter Two
say fairly precisely what parts of the brain might act as information storage and how much information they must be capable of storing. P, Q and R may exchange information by making statements. When P, for example, says Cambridge University was founded in the thirteenth century, what happens is that a fact obtained from his information store is transmitted by way of nerves, his larynx, sound waves, the ears of Q and R and more nerves, to the information store of Q and R where, if it was not there previously, it is written together with the information Q and R already possess. Q and R are themselves each now capable, in an appropriate situation, of making the statement Cambridge University was founded in the thirteenth century. This process is, in essentials, quite mechanical, and all the processes postulated in this book are intended to be mechanical in the same way. Although, as I have already indicated, I want to maintain a sharp contrast between communication by people and communication by machines, it would be foolish to pretend, against all the evidence, that people and machines do not both completely obey the laws of physics. It is not to my purpose here to give full details of how the various processes discussed might be mechanised; but they are intended to be mechanisable in principle. When P wants a piece of information from Q or R he may ask a question, which is no more than an information-specification. However, a question does not specify a precise piece of information, but rather a range of pieces of information between which the addressee is obliged to choose. If P asks Q and/or R Was Adam of Balsham an Englishman? he is asking Q, or R to give him either the piece of information Adam of Balsham was an Englishman or the piece of information Adam of Balsham was not an Englishman. Q may or may not have one or other of these pieces of information in his store, and may answer by making a statement which is one or other of these two. When this happens, the piece of information, say Adam of Balsham was an Englishman, is transmitted, by way of larynx and sound waves as before, to P's store - and perhaps also to R's. Commonly an answer to a question like Was Adam of Balsham an Englishman? is given in a short form, as either Yes or No. These do not look like whole statements, but it is important to realise that, for our purposes, they must be regarded as statements. This is because, just like
Question And Answer
29
statements in their full form, they result in the copying of information from one person's store to another's. Dialectically speaking, we shall say, they are statements. The word dialectic, which has had a long and involved history, originally referred to the study of the pursuit of truth through dialogue, and I shall here restore it to this use or one very like it. Yes and No are abbreviations for statements - or sometimes for other things - when they occur in appropriate contexts. When P asks a question some part of his brain "holds" this question, at least until an answer comes in, and if this answer is Yes or No he converts it into a statement in full form by a grammatical manipulation: reordering the words of the question if a Yes answer was received, incorporating not (or does not, etc.) in the case of a No. In the same way R who overheard the original question and also Q's answer to it, is enabled to store away a whole statement. Q and R, of course, may not know the answer to P's question, and may have to say so. This means that it is not sufficient for a question-andanswer language to contain just questions and statements. It must also contain at least one other kind of locution, I don't know. We are driven to this conclusion by the exigencies of satisfactory communication. I don't know, of course, may itself be a statement of a kind, but it is of a different kind from Adam of Balsham was an Englishman and might be better regarded as a special and different kind of locution altogether. When both Q and R answer I don't know to P's question the question is regarded in one sense as "answered", but no information on the subject of the question is transmitted or copied up. We can imagine that P, Q and R go on making statements or asking and answering questions indefinitely. The long term result is that their information stores get more and more alike in their contents. Eventually, in the absence of new information entering the situation from outside, the contents of the three stores may be all identical. However, there are various mechanisms which may aid or hinder the process of information-sharing. Before we go on to consider complications of the simple picture, let us look briefly at some of them. First, P if he is intelligent, will never ask the same question twice; since it will always be the case either that he will get the same answer (and hence no new information) the second time as the first or, if Q's and R's stores have changed in appropriate respects in the meantime, that he would have been in a position to deduce the nature of the change if he had kept his ears open. (I assume that all three have been present throughout the conversation. My point would not apply if P, for example, had met Q, then Q had gone away and met R and subsequently P had met Q again.) But this means that P must remember something about what has gone on earlier in the conversa-
30
Chapter Two
tion. It is not actually necessary that he should remember which questions have been asked, but he should store up, as far as he can, information about which information is already in the possession of Q and R. Sometimes this will be positive information in the form Q knows that such-andsuch sometimes it will be negative in the form Q does not know the answer to such-and-such a question. We can imagine that among the sheets of paper making up P's store there is one headed "State of Q's store" and another headed "State of R's store", and that on these sheets are listed various statements known to be in Q's and R's stores and, below a horizontal line, various others known not to be. In the version of the dialogue so far described the contents of the sheet for Q would be identical, with those of the sheet for R, and the information known to be known to Q or R would also be known by P himself, so that a wholly simpler plan would be to have a public sheet of paper headed "State of shared knowledge". However, this would not be satisfactory in general. Generally, too, P will not ask any question such that he himself knows the answer. Questions, in this conversation, are assumed to be genuinely information-seeking. If P wanted R to know something that he knew and thought Q also knew, it would be simpler to tell R directly than to ask Q a question. Similarly, P will not make any statement that has already been made and is hence known to be in the stores of both Q and R. Statements, we assume, are in intention information-giving, and no information is given to someone who has it already. In practice, when we repeat statements it is usually because we are unsure whether the information was properly imparted on the first occasion, or whether it may have been forgotten in the meantime. It is easy to design, from these requirements, a mechanical realisation of them, a set of precise rules for the "mechanism" within P's brain. It must have an input, an output, a processing unit and a store, capable of handling questions or statements in some suitable language - we may suppose, a logical language of as simple a form as will be consistent with the diversity of the statement-forms and question-forms needed. P may speak provided no one else is speaking. If the last thing said was by Q or R and was a question, the only thing P may say is an "answer" (in the enlarged sense in which I don't know counts as an answer) to the question; otherwise P may make any statement or ask any question, subject to certain restrictions. We may suppose (for the present) that questions and statements are produced at random and checked for adequacy before outputting. A statement is adequate if it is already in P's store as a piece of knowledge and is not in the stores representing P's knowledge of the
Question And Answer
31
knowledge of Q and R. A question is adequate if no answer to it is in P's store as a piece of his knowledge, and if it is not the case that P's stores representing the knowledge of Q and R contain items indicating that neither of them knows any answer to it. We may make various more precise stipulations of detail; but those given are sufficient to ensure that P does not "waste" statements or questions. We must now turn to some complications of the picture. Yes-no questions clearly indicate what the supposed set of answers to them are: they are Yes and No. The same applies to certain other simple kinds of question with small numbers of possible alternatives, such as Did she wear her red hat, or her blue one, or her white one? But W-questions are often a different matter. If I am asked In which continent is the country of Ecuador? I shall have a general, but not at all a definite, idea of just which pieces of the earth's surface my questioner regards as continents. Does he include Antarctica, for example? Are the Americas two or one? Is Oceania a continent? Is Australia a continent, or should it be Australasia? Is Iceland a part of Europe, and where is the precise boundary between Europe and Asia? It is true that most of these questions are irrelevant when it comes to giving the true answer to the question In which continent is Ecuador? but they are not irrelevant to the question of what counts as an answer. Commonly an answerer resolves ambiguities, when they matter, by giving, if he is able, more information than may be strictly necessary. Ecuador is in the northern part of South America, on the equator. That's what its name means, you see. Moreover some questions, particularly Why and How questions, have much less clearly defined families of answers than this one. Consider Why did Milton write Paradise Lost?. Fame, fun, piety, money suitably filled out, these are all possible answers. If we are to believe the old joke, it was as a dramatisation of his missed bachelor freedom. We can imagine all manner of mixed explanations, from the simplest practical to the most abstruse psychoanalytic ones, none obviously absurd and not all even mutually exclusive. They are of somewhat different kinds causal, motivational, rational. It would be a ridiculous oversimplification to say that such a question specifies a family of possible answers. At best it sketches one. Practically all that the answers have in common is that they are of the form Milton wrote Paradise Lost because .... In passing we might notice a curious way of giving an answer to a Why question, that is nevertheless very common, namely to point out that the fact to be explained is of a kind that is frequent or usual. I answer Why is John in London? with Oh, he's there very often, and Why are you so happy? with I'm always happy. Sometimes, perhaps, these answers are
32
Chapter Two
more like a rejection of the question: they say There's nothing to be explained. But they usually satisfy the questioner. It follows that what is offered and accepted as an answer to a given question is to some extent itself open to disagreement and discussion. This means that the appropriate reply to a question is sometimes the raising of an objection That is much too vague or the asking of a clarification question Do you mean...?; and that sometimes an answer to a question may also be rejected as unsatisfactory or, at least, as not being what the questioner wanted. We shall return later to replies of this kind when we come to treat "points of order". One other kind of complication, very common in questions, is that they presuppose in their formulation some piece of information which may or may not be already in the possession of the various participants. Logicians have sometimes objected to questions such as Have you stopped beating your wife? on the grounds that presuppositions, such as You used to beat your wife, should not be smuggled into questions surreptitiously. But it is a common experience that presuppositions of questions may be quite open and harmless. When I ask In which continent is the country of Ecuador? I may be said to presuppose that there exists a country called Ecuador, and that it is a continental country unlike, for example, Tonga. It is difficult to say anything at all without presuppositions, though most of the ones we make are true or, at least, agreed to beforehand, and cause no trouble. If P, Q and R all agree that there are just six particular continents and that Ecuador is a country in one of them, the question In which continent is Ecuador? is harmless on all counts: it merely requires to be answered. If P, say, knows that Ecuador is a continental country and Q and R do not, it is somewhat more doubtful whether P should be allowed to ask the question. If he does, one of two things may happen: either Q (and/or R) accepts Ecuador is a continental country as a piece of extra information implicitly conveyed by the question, or he treats it as an objectionable presupposition. In the first case he writes the piece of information into his store; in the second he objects to it or questions it or otherwise calls attention to it. In particular, if he goes ahead and gives a straight answer (true or false) to the question, he has accepted the presupposition; and under these circumstances both P and Q gain information from the interchange. Now let us consider the question of the truth of the information in the various participants' stores; or rather, its consistency, since there is nothing yet in the system as a whole to mark the distinction between true and false, provided the participants' several beliefs are consistent. We have tended to beg the question by using the word "information", which generally implies
Question And Answer
33
truth: the phrase "false information" has a slightly uneasy ring, since falsehoods might be held not really to inform but to misinform instead. What are we to say about the possibility of inconsistencies in the stock of information? If inconsistency is to be discoverable at all we shall have to assume some kind of logical faculty in the participants. We shall have to assume, that is, the ability to draw logical consequences and detect incompatibilities. Let us assume that P, Q and R are all first-rate logicians and always draw, in a flash, any conclusions there are to be drawn. We can now relax one assumption we have tacitly made so far, namely that a participant can make a statement only if that precise statement was already present in his information store. It is enough that he should have statements in his store from which the statement he makes follows. Now P will not ask any question whose answer follows from any statement, or any combination of statements, in his own store; nor make any statement that follows from what his store tells him to be contained in the stores of Q and R. When a statement is made to P and added to his store, all its logical consequences, and its joint consequences with what is already there, are in effect - though not necessarily in fact - also added. And if a statement is made to P that already follows from others in his store, he need not add it at all. But what will happen if P is told something that conflicts with what is in his store? He may, of course, reject it, a possibility for which we have not yet provided him with an appropriate form of locution. Or he could accept it and "change his mind", but we have not allowed yet for this possibility either. Or he could just accept it. But the consequences of his simply accepting it are far-reaching and hardly to be tolerated. If P accepts a statement inconsistent with other statements he has already accepted, the result is that his information store is internally inconsistent. It is a fact well known to logicians that, as we might put it, once you believe a contradiction you will believe anything. Formally, any statement whatever can be deduced from a contradiction. It is as if, in mathematics, one were to start out with the axiom "l = O". Multiplying both sides by "A-B" we get "A-B = 0" and hence "A = B", independently of what A and B are, so that we shall be able to prove anything equal to anything else. The proof of the corresponding logical theorem is in every logic book: given two premisses S is true and S is false you can deduce T is true, whatever T is, by two easy steps. From S is true deduce One or other of S, T is true. From this and S is false deduce the conclusion. There have been various attempts to challenge this paradoxical result, but none
34
Chapter Two
clearly relevant to the present application.18 Now that P "knows" a contradiction to be true he can deduce every other fact whatever, and so he "knows" everything. Unfortunately, he "knows" all the false facts as well as all the true ones and has no way of distinguishing the one from the other. P's dilemma will be greatest when he is asked a question. Let us suppose that he has been told, and has accepted, both that brachiopods are bivalves and that they are not; and is then asked Are kangaroos carnivorous? He will find even if he has never previously learnt anything about kangaroos that Kangaroos are carnivorous is implied by the contradiction in his store, and that Kangaroos are not carnivorous is implied also, whence he might equally answer Yes or No. He might be tempted to give both answers at once, if it were not for the fact that it is also implied that both are false. He can no longer act rationally at all. This artificial situation would not be at all worrying if it were not for the fact that it is not entirely unrealistic. We often do get ourselves into contradictions, yet manage to get ourselves out of them without complete mental breakdown. You and I pass, on the whole, for rational people: we are not, of course, of infinite logical wisdom as we have assumed P to be, but we have enough logic to spot contradictions now and again. There have certainly been periods in my life at which I have been tempted to subscribe to each of two opposed beliefs, and to draw strange consequences from the conjunction of them. But, even when this has been so, it has not made ordinary conversation impossible for me. I have been able to go about my business, asking and answering questions, at least on topics unconnected with the subject of the contradiction, as before; and even on this subject my irrationality has not, so far as I have been aware, gone beyond bounds. A limitation on our logical abilities is imposed in practice by the necessary compartmentalisation of our store of information. We know so many things that - as any computer programmer could soon calculate - it is humanly and even mechanically impossible that we should be capable of surveying them all even occasionally to try to draw logical consequences or spot contradictions. If someone asks me a question about kangaroos, the
18 The best known challenge is that of Anderson and Belnap, whose "System E" is founded on the principle that no deduction is valid unless the conclusion is "relevant" to the premisses, in a certain precise sense. But even the assumption that the mental processes of P, Q and R conform with the logic of System E will hardly enable us to rest content with the consequences of having them believe contradictions. Anderson, Alan Ross and Belnap, Nuel D., Jr. "The Pure Calculus of Entailment", Journal of Symbolic Logic, 27 (1962), pp. 19-52.
Question And Answer
35
most I can do in answering it is to survey my knowledge of kangaroos, or of some other relatively small relevant parcel of my knowledge. It follows that a contradictory belief about, for example, the freedom of the will will be very unlikely to have any influence on my answers to questions about kangaroos. But this only imposes limitations on the scope of the problem: it does not solve it. Our answer will be that we need to re-examine the basis of the system of information stores, and of reference to them, and make some changes in it. This will also involve the introduction of some subsidiary kinds of locution. It is interesting that the pressures come from the dialectical side, being dictated by the requirements of satisfactory communication. It is also interesting that our model of the internal workings of the participants in the dialogue is developing hand-in-hand with our development of the language. We have, so far, been spending a lot of time on fundamentals, but the development will accelerate. The reader might appreciate an example of dialogue on which to try out his intuitions regarding the analysis of questions and their answers. It would be possible to invent a symbolic language that would embody this analysis but it is probably more profitable here to use ordinary English idiom. Brown: What can you tell me about the history of the study of questions? Jones: That's rather a vague question. What would you like to know in particular? Brown: Well, has much been written about it? Jones: Oh yes. Quite a lot. Brown: Whom by? Jones: Aristotle. And one medieval writer was Adam of Balsham. Brown: Balsham? Isn't that near Cambridge? Jones: Yes. Brown: Was he English, then? Jones: His family had come over with William the Conqueror. I suppose that makes him English. Brown: Did he go to Cambridge? Jones: No, Cambridge wasn't founded then. I don't know where he studied. He taught mainly in Paris, in the middle of the twelfth century. Why are you so interested in this? Brown: Oh, I am generally interested in the history of logic. Jones: Adam was called "Parvipontanus" because he taught near
36
Chapter Two
the little bridge over the Seine. Brown: I didn't know there was one. Is it still there? (Jones's first reply is a point of order. In precising his demands Brown asks first a yes-no question and then a W-question: both are answered more fully than is strictly called for. To Was he English, then? Jones gives an answer that fills the deficiencies of a mere Yes or No. Brown's reply to Jones's Why-question is of a kind we noted in an aside. Several of the later replies take up presuppositions of the questions preceding them.)
CHAPTER THREE COMMITMENT
Can I believe a contradiction? It seems at first as if one should be able to resolve this question by conducting a mental exercise, for example saying to oneself first There are lions in Africa and then There are no lions in Africa, to see if one can generate an appropriate state of mind. If I personally try this exercise all I get is a sort of two-level map of Africa, with a symbolic lion at one level and none at the other. I expect your attempt will give you a similar image irrelevant to the question. There are certain fields in which it does not seem at all impossible to believe a contradiction; for example, mathematics. I can easily imagine believing that the opposite angles of a cyclic quadrilateral are equal rather than merely supplementary, or even that the sum of 63 and 14 is, say, 79. However, this seems to me to show that such statements should not really be described as contradictions, but rather as mathematical falsehoods, I am aware that this contention raises a problem and will offend many who think that "contradiction" is a term of formal logic and that mathematics and formal logic are the same thing. However, this is a long story and we need not embark on it. The fact is that there is a lot of difference between believing a mathematical falsehood and believing each of a pair of clearly contradictory statements. To believe a contradiction, then, is difficult or impossible. But we should be careful to distinguish the question of believing a contradiction from the question of assenting to one. Someone says There are lions in Africa and I say Yes; a little later, someone else says There are no lions in Africa and I say That's right. It does not matter whether I say this vacantly or enthusiastically, with or without thinking: I have, literally, assented to a contradiction. It is true that this is regarded as dialectically objectionable behaviour on my part, and that an onlooker will be very much in order if he says Hold hard! You said both that there are lions in Africa and that there aren't. Which do you really mean? But the fact that a certain kind of behaviour is objectionable is not evidence that it is impossible: rather, it assumes its possibility. My assent to two incompatible statements was a simple dialectical fact.
38
Chapter Three
Some people would say that assent to contradictory statements is objectionable because one cannot believe both of them. Whether this is reasonable we shall consider later. So far as rules of dialogue are concerned, however, it would be desirable on the whole to leave belief out of the picture, at least at first. We ought to be able to state rules such as One should not contradict oneself, and see their role in the conduct of dialogue, without getting involved in the much more difficult questions surrounding words like "knowledge" and "belief". It should be possible to drive a wedge between these psychological considerations and purely dialectical ones, and study the latter at first in isolation. Now consider another difficulty raised by our model in the previous chapter. I adapt this exposition from a recent paper by Colin Radford.19 Adrian is having an affair with Beatrice, and their friend, Celia, who lives in an adjoining flat, finds out about it. Celia then knows that Adrian and Beatrice are having an affair; but she says nothing about it, and they do not know she knows. But the next day they read her secret diary, which she has left lying about; they now know she knows, but she does not know they know she knows. However, she overhears them talking about it and now knows they know she knows, though they do not yet know she knows they know she knows. And so on. It seems that the chain of knowing about knowing can grow without limit. Adrian's and Beatrice's information stores contain not only sections headed "State of Celia's knowledge", but also sections within them headed "State of Celia's knowledge of our knowledge of her knowledge", and so on without limit. And the same for Celia. But now, Adrian, Beatrice and Celia meet for drinks and bring the whole affair into the open. Adrian and Beatrice tell Celia and she congratulates them. They now know she knows and she knows they know she knows, and so on, and nothing is hidden. And now it seems to follow that they know she knows they know ... without limit! But the trouble with this is that it seems to follow that each of them has a brain of infinite capacity. For there must be subsections, and subsections of subsections, without limit, to record all the various pieces of mutual knowledge imparted - by the conversation in which Adrian and Beatrice came clean! Since it is clear that virtually every case of informing someone of something is a case of "coming clean" like this, and secondly that we do not have brains of infinite capacity, something is again wrong with our
19
Radford, Colin, "Knowing and Telling". Philosophical, Review 78 (1969), pp. 326-36.
Commitment
39
model. 20 A third argument for doing something about our model is that there must often, for a number of reasons, be a divergence between what we know or believe and what we commit to language. In the first place, of course, we cannot and do not say all the things we believe to be true, if only because they are very numerous and generally irrelevant. Secondly, even if we have said them in the past, we often change our minds. In periods of calm we sift our beliefs, perform inductions, reject inconsistencies of various sorts (not always logical), speculate. And we have conversations with different people, so that when I meet you again I may have changed my mind on something I said to you previously And thirdly, we sometimes tell lies. In the model sketched in the previous chapter lying would be impossible since there is no mechanism for accumulating the body of facts P has actually asserted or accepted, independently of the body of facts he is supposed to know or believe. Perhaps the moralist who condemns lies could wish that our linguistic behaviour could be so represented, but this should not be allowed to blind us to the fact that it cannot. And actually, the moralist, if reality were as implied, would be out of a job since there could be no special virtue in truth-telling either. If the moralist wishes us to be able to build a virtuous computer, which is capable of telling the truth, he will have to allow that the machine we build must be one of which it could be meaningful to say that it had told a lie. We shall return to the question of the morality of lying in a later chapter. The answer to all these objections to our model of the preceding chapter is to separate each participant's store of facts into two stores of different functions. I shall call them the fact store and the commitment store. Since, as we shall see in later chapters, there may be commitments of different kinds, I shall eventually particularise the latter to indicative commitment store. One's commitments, in this sense, are not necessarily beliefs at all, but simply statements to which one has committed oneself in
20 Must the chain really be infinite? It has been argued to me by Neil Harpley that it is not even in principle meaningful for the chain to go beyond about two steps. There may be a story, Harpley thinks, of successive interpersonal discoveries without limit, but it is not meaningful to sum these up with a long chain of "knowing that so-and-so knows that ...". If he is right, this argument for my thesis fails (though I have plenty of others). But I do not think he is right. Although, as Radford says, "death or loss of interest would stop [the chain] developing, and an understandable confusion on the part of either or both parties as to where he or the other is at may deform or blur the complex but regular structure of the later situations", I can see no reason of principle against an indefinitely long chain.
40
Chapter Three
conversation. If I say to you Last summer I visited Sweden this statement becomes one of my commitments and is written in my commitment store; and this takes place independently of whether it is true, whether I believe it or know it to be true, whether you believe me when I say it, or anything else of that kind. If I am telling the truth when I say Last summer I visited Sweden then the statement must have come from my fact store. This does not mean, perhaps, that it must be actually true, for telling the truth is an ambiguous phrase that fails to discriminate between saying what is true and saying what one thinks to be true; similarly telling a lie fails to discriminate between saying what is false and saying what one believes to be false. When we are in a truthful mood, in the sense of not being inclined to deceive our hearers, the things we say are things that are in our fact stores, and the upshot of our saying them is that they are copied into our commitment stores. When we are untruthful, in the sense of saying things we do not think, what we say is written into our commitment stores anyway, but is not in our fact stores. There may or may not be entries in our fact stores actually inconsistent with what we say. Now let us confine ourselves to the case of a dialogue between two people, P and Q. We shall suppose that P and Q have never met before, and that they know nothing about one another, and know that they know nothing; so that P has never committed himself to anything in conversation with Q or vice versa. Under these circumstances we should say that both commitment stores are empty: the commitment stores will act, as it were, as score sheets for the dialogue, and play has not yet started. But P and Q could nevertheless be very knowledgeable people whose fact stores are crammed full. It will be seen that people carry the contents of their fact stores about with them from birth to death, but that their commitmentstores sometimes have to be wiped clean and restarted, depending on whom they are talking to. In practice there is some carry-over from one situation to another since we learn of people's earlier commitments by hearsay and are often, in fact, differently committed to different people all present. Nevertheless the case of a brand-new two-person dialogue is an important ideal case, and I shall frequently consider it. As P and Q converse, the various things they say and agree to are entered in their respective commitment stores, and these stores reflect the course of the dialogue. Their commitment stores can be regarded, at any time, as giving a resumé of relevant features of the dialogue: relevant, that is, to its future conduct And each will know what is in the other's commitment stores, which unlike the other fact stores, are public, and could in principle be kept on slates with chalk, subject to inspection by both
Commitment
41
participants and by onlookers. They represent objective and public features of the dialogue. In practice, of course, each participant "remembers", in his own head, the state of the other's store and his own; but, provided all goes right, their records correspond. In this fact lies the solution to the problem of the infinite chain of knowing about knowing. So long as knowing was a partial and essentially private affair, P's knowing something did not imply Q's knowing that P knew it. But completely public knowledge is another matter. Let us suppose P, knowing that Adrian and Beatrice are having an affair, has said so and that this fact has been inscribed in his commitment store in the presence of Q. On the assumption that this is a normal communication situation and that Q has no reservations about accepting what P says, we shall be able to say that Q now knows too, and knows that P knows. And P knows that Q knows that P knows. And Q knows this, and so on. This does not mean that there is an infinity of entries chasing one another in their fact stores, but merely that the information is in their commitment stores, which are public. It is true that there is an element of presumption in the stipulation that this is a "normal communication situation", but this is just what most communication situations are. Now let us look in detail at the operation of commitment stores in a normal dialogue. There are four symmetrically related kinds of locution that seem to me to be basic to this kind of situation. First we have the statement, in the indicative mood, which states a fact and leads to the insertion of an entry in the speaker's commitment store. Next we have the question, which is a device for eliciting a statement from the person to whom it is addressed. Questions set out a specification, detailed or vague, of the kind of statement - that is, answer - required. A question does not necessarily involve insertion of an entry in the questioner's commitment store, though it will do so if it contains a hitherto unstated presupposition; but the answer to it, being a statement, will so involve the answerer. Thirdly, however, there will sometimes be occasions on which a speaker wishes to delete an item from his commitment store. This happens when he "changes his mind", and "retracts" what he has previously said. Retraction is to be compared with negation, but the two are distinct, since I can take back something I have said without actually denying it: I can say No, I am not so sure now it was the postman: I don't know whether it was or not, and this is different from saying No, it wasn't the postman. All that necessarily happens to a speaker's commitment store when he retracts something is that the appropriate entry is deleted from his store: the insertion of a contradictory entry in its place is another matter.
42
Chapter Three
And fourthly, we have locutions related to questions in the same way as the ones we have just been discussing are related to statements. These are designed to elicit retractions. A locution of this kind is particularly appropriate when the person to whom it is addressed has become committed to a number of statements that are in contradiction: either to a pair such as Roses are red and Some roses are yellow, or a larger set such as the trio Public transport is terribly inefficient, If you haven't a car there's only public transport and You can get about quite efficiently without a car. Under these circumstances it can reasonably be demanded of him that he retract one of the statements. The kind of locution that makes such a demand is considerably less common in practice than questions are, and less easily identified grammatically, but it is a theoretical counterpart of a question. Questions are normally such that at least one of the indicated family of answers is true, and each question carries a presupposition to this effect. The locutions we are now discussing, which I shall call retraction demands, are normally such that at least one of the indicated statements is false, and each such locution can be regarded as also carrying such a presupposition. The hearer may, of course, want to reject the presupposition in either case. We have, then, four kinds of locution: (1) Statements. (2) Questions. (3) Retractions. (4) Retraction demands. These are not necessarily exhaustive, but they are basic. The existence of retraction demands clearly could not be justified by consideration of the commonly used forms of sentence: their existence in principle, rather, is revealed by logical symmetry, namely, by the necessity of an idiom which will enable one speaker in a dialogue to place a demand for retraction on another, of the same general character as the demand - the word is not quite right, but we have no closer one - placed by a question. Otherwise, there would be nothing one could say to a speaker who had contradicted himself. To put this point another way: someone who contradicts himself does not break a rule of dialogue, in the way someone who utters irrelevancies, constantly repeats himself or fails to treat questions as requiring answers does. The "Law of Non-Contradiction" does not forbid you to contradict yourself, any more than the "Law of Excluded Middle" forbids you to have no opinion. If someone does (apparently) contradict himself by uttering each of two (apparently) opposed statements, the
Commitment
43
proper course is simply - if one wishes to ask him which he wishes to withdraw; just as, if he had stated neither, one might have wanted to ask him which he wished to affirm. There is, in fact, a complete duality. In passing, it is interesting to notice that we could proceed, if we wanted to, by supposing that, at the beginning of a dialogue, everyone is committed to everything; and then retractions would have the function that statements do at present. That is, retracting not-S would have the function of stating S and vice versa. But then to retract both of two statements S and not-S would be as much or as little objectionable as contradiction is at present - by the Law of Excluded Middle! Then questions would also have the function of retraction demands, and vice versa. This brings us to hearer-commitment. The addressee of a question is under an obligation to give an appropriate answer either one of the family of indicated answers, or Don't know which is a sort of advance retraction of all of them and could be so represented, or a challenge to the question's presupposition, which could also be represented as a retraction, namely, of the presupposition. He will be under a similar obligation in respect of a retraction demand. But is he under no other? If P and Q both confine themselves to the making and retraction of statements, may they do so independently and at random? Or is there some way in which they are obliged to take notice of each other? It seems highly appropriate to say that there is. We sometimes say Silence gives consent. This is not a very strong principle, but it embodies an idea we can sharpen up. In the ordinary course of events we do not expect much disagreement from people we talk to, and it is hardly going too far to say that we presume agreement. If I say It's twelve o'clock to Smith in the course of an ordinary conversation, and he neither explicitly agrees nor disagrees but goes on to say something else generally relevant like I wonder what will be on for lunch or Have you got through much work this morning?, I shall be in order in assuming not merely that he has heard and understood me but that he agrees. The onus, we might say, is on the hearer to object to any statement he disagrees with. If he does not he will be bound by it. What this means in terms of rules for operation of the commitment stores is that statements are always entered in the hearer's store as well as the speaker's - unless, that is, the hearer objects. He may do this by means of an immediate retraction. Perhaps, on a fine point, we should say that he ought to be given a reasonable chance to retract before he is regarded as committed at all; but this is splitting hairs. Let us say simply that every statement commits both speaker and hearer, and that the way for a hearer
44
Chapter Three
to avoid unwanted commitment is to issue an immediate retraction.21 Let us consider a short example. Brown: Aristotle wrote about the logic of questions much earlier than Adam? Jones: On reflection I don't really want to say that. Aristotle wrote about how to conduct a question-and-answer dialogue, but not about the logic of questions. Brown: But the medieval logicians went further than Aristotle, did they? Or did they just copy him as usual? Jones: They often didn't just copy him. In this case several of them went further. Adam and some other French scholars were really interested in the logic. Brown: But you said Adam was English! Jones: Yes, all right. He wasn't French. Do you understand about the origin of the logical studies? Brown: Yes. I understand that Aristotle started them. Jones: No. As I said, and as you agreed, he didn't go deeply into it. Amidst the complications of the ordinary-language dialogue we discern that Jones's first locution involves a retraction; that Brown's third is a retraction demand in respect of Jones's earlier statement Adam was English and the presupposition of the one just made, that Adam was French; and that Jones's last both rejects and contradicts Brown's preceding locution, and objects that Brown is in contradiction, owing to having implicitly agreed to one of Jones's earlier statements. Now let us turn to another feature of any dialogue involving statementmaking, the question of accountability. We often disagree with things that
21 There could be (and has already been) a lot of argument about hearercommitment. Often, it will be said, we let a remark pass, although we disagree with it, because it isn't central or because we have other preoccupations. This must be admitted; but the very fact that the qualification that the remark must be noncentral or unimportant is necessary is some vindication of the existence of hearercommitment in the standard or typical case. If a remark gets conceded, as it were, by default which is what the hearer-commitment principle implies in these cases there is a case for saying that a lapse (albeit minor) of communication has taken place. But there is such a case anyway. To the extent someone need not be regarded as having accepted a statement addressed to him, to that extent he is not engaged in a conversation of which this statement is part.
Commitment
45
other people say, and say so: but we seldom disagree flatly without following up our difference to its source. In some sense flat disagreement is bad logic: it is certainly bad manners. Broadly, it is not open to us to commit ourselves to any statement we please, and under most circumstances we shall be expected to be able, if challenged, to provide a justification or rationale of what we have said. There are, however, two subtly different things that can be meant by "providing a rationale of a statement". By way of justifying a statement I can either (1) give reasons for thinking it to be true, or (2) give reasons for having made it. I may, of course, have made it for no other reason than that I think it to be true; but even this apparently excellent justification of a statement is not normally sufficient for sense (2), since there are many things I know to be true that could not possibly be of the slightest interest to you, and which it would be quite inappropriate for me to tell you. How would it be, for example, if I met you in the street and said I say, ice melts at 32° F or The city of Nagoya has a population of two million? You would think I was mad. But the capacity to be irrelevant, surprising or in need of justification - in sense (2) - is not peculiar to acts of statementmaking. You would be just as surprised if I suddenly stood on my head, or even, if you were a stranger, if I gripped and shook your hand. The kind of accountability that is peculiar to statements is that associated with sense (1). If I say The population of Nagoya is two million and you ask me Why? I shall reply by giving you the evidence I have for thinking it, or a set of premisses from which it may be deduced. That is, I shall give you some other statements that serve as a logical justification or rationale for it: I may say That's what it says in this pamphlet, and it's an official Japanese government one, or It has doubled in the last thirty years and, as you no doubt know, it was then a million. Moreover, all statements, with the exception of two special classes I shall come back to, are supposed to be accountable in this way. One's readiness to account for a statement is part and parcel of one's readiness to make it, and of its ability to impress a hearer. If P says Modern plastics are an abomination and Q says Why? meaning What reason is there for thinking that? - Q's question is related to his readiness to accept what P has said. It might be said to contain an implicit rejection of it: at least it denotes suspension of acceptance. P may succeed in giving reasons - that plastics are unaesthetic, cause skin disease, create a waste disposal problem, or some such - in which case Q if he accepts the reasons, is under commitment in respect of the statement itself, and may escape this only by an explicit rejection. On the other hand, if P cannot give reasons, it seems completely in order for Q to challenge him to
46
Chapter Three
retract. Accountability means that a commitment store is not just a list, but a structure. The evidence I give for my assertions consists of assertions that go into the commitment store themselves; but the fact that I suppose it to be related to my original assertion, as evidence or rationale for it, also commits me. Consequently a kind of tree of statements may get built up in a commitment store. Not every statement needs to be connected with others, since I may either refuse, or not be asked, to give a justification of a particular statement, yet remain committed to it myself. However, connection to a tree is the norm. Accountability is important because it is one of the things that distinguishes statements from other kinds of locution. Questions are not generally accountable, because their role as questions is not dependent on their having any particular kind of rationale. If I ask you a question you may wonder what prompts me to ask that particular question on that particular occasion - this is sense (2) - but you will not invalidate my question or relieve yourself of the obligation to answer it if you ask What reason is there for wondering that?, for I am more-or-less allowed to ask questions for no reason, if I want to. Retractions are not accountable, since it is open to anyone at any time to have lingering doubts, however irrational, and to say so. Retraction demands are not accountable either. Anyone of these locutions may have other consequences and involve other commitments, but accountability, which is related to the building of trees within a commitment store, is so far peculiar to statements. (Though we shall actually find later that there do exist accountable locutions of other kinds.) A curious shift of meaning sometimes takes place when a speaker makes a statement but refuses to provide a rationale for it. The statement changes from being the statement that such-and-such is true, to the statement that the speaker thinks, or believes it is true. Often enough the speaker will make this revision himself; and, if he does not, others will do it for him. If I say That's the funnel of a ship on the horizon and you say Why?, I shall say, if I can say nothing else, Well, that's what I think it is anyway or Well, that's what it looks like to me. Or if you say Silver melts at a lower temperature than lead and I say Why?, you will say If my memory is right, it's so The explicit case of I believe that so-and-so is of especial interest, both because it is a weak statement and because it is sometimes thought to be a strong one. We are often told that it is important to have beliefs - either to have beliefs of certain sorts or, at least, some beliefs. Or to have faith, or a faith. These words can have several meanings, and to have faith in a concept, or a person, is not the same thing as having faith in the truth of a fact.
Commitment
47
Nevertheless we are also sometimes told that we should have certain beliefs, or faith, in the sense that we should believe the truth of certain facts. But let me suppose, now, that I am told that I should believe that a certain kind of God exists, and am forced to intone I believe God exists. If I had just said God exists, then God exists would have been inscribed in my commitment store; but is it now? All I believe God exists seems to do is report, truly or falsely, on the existence of an item God exists in my fact store, and place in my commitment store an item I believe God exists, that is, My fact store contains an item 'God exists' This is not only a strange kind of affirmation to be forced to intone, but an affirmation that, in any case, has a rather curious logic. It seems at first appropriate to say that whenever anyone says I believe that so-and-so, not merely this statement but also the so-and-so itself is inscribed in his commitment store. That is, when I claim that I believe something, I commit myself to it. The idioms I believe, I think and some others have special properties and commit those who utter them in special ways. Now it would seem that these idioms give statements that are really stronger than other statements since they have double commitments rather than single ones. An affirmation of faith assumes an especially important role, both as reporting on ones state of faith, and committing or re-committing one in respect of the content of that faith. But this analysis overlooks two features of ordinary statements, namely (1) hearer-commitment, and (2) accountability. First consider hearercommitment. If I say God exists I commit not merely myself but also you, and if you say No he doesn't you contradict me, But if I merely say I believe that God exists you do not contradict me if you say No he doesn't or even if you say I don't, but only if you say No you don't. The only hearer-commitment my statement I believe God exists has is in respect of my having this belief, not in respect of God's existing, and if you accept the statement all you have accepted is a fact about the contents of my fact store. Now consider accountability. The statement that I believe somethingor-other is not a statement of a kind I am normally expected to provide evidence for: it is one of the exceptions, of a kind known to philosophers as an avowal. I can, on the other hand, be asked to provide a rationale of my belief; that is, of its content, not of the fact that I subscribe to it. If I say I believe there is a blue-tongued lizard living in my garden hedge it would be very odd for you to ask me what evidence I have that I believe this though quite normal to ask me what evidence I have that it is true. On the other hand it is clear that if I fail to provide such evidence, or refuse, no
48
Chapter Three
further weakening of my statement takes place. Told, even taunted, that one of my beliefs lacks evidence, all I need do is reaffirm it. Typically, the firm believer does just this, with emphasis: Nevertheless I believe it. I just do believe it. I have faith that it is so.22 It is as if he thinks emphasis could overcome the statement's logical weakness, or commit hearers in some additional way. The fact will usually be that hearers already agree with all the statement calls on them to agree with, namely, that he does believe it. There is only one dialectically effective way of stubbornly maintaining something, namely, by maintaining the fact itself and not merely the fact that you believe it; and by meeting the demands of accountability to whatever degree is called for. By this view, I believe that S is to be regarded neither simply as a· statement about the speaker's own state of mind - since it does commit him in respect of its content, S - nor as equivalent to S itself. It is a weakened version of S, lacking hearer-commitment and accountability. We shall return in a later chapter to the question of statements about one's own state of mind. The theory of statement-making and question-asking sketched in this chapter is oversimplified at many points and needs much filling out in matters of detail. It should be clear that we have been dealing with what might be called the "standard" use of statements and questions, and that there are deviant uses in which one or another of the features of the standard use may be lacking. Very many non-standard situations, however, do not so much affect details of the system as transfer the system as a whole into some new dimension. Let us consider the case of actors on a stage. Their statements, questions and so on do not commit them in the way they would in ordinary use, and are not meant to; but then, the persons represented in the play, fictitious though they may be, are supposed to be the ones who are asking and answering the questions and making the statements, and if we wish to understand the play we need to think of these
22 Tertullian is reported to have countered the charge that his belief in God was an absurd one with the statement I believe it because it is absurd. Perhaps what he meant was that it was so outlandish a hypothesis that unless it were true no-one would ever have entertained it. I shall not pursue this interpretation. Taken more literally, it is ambiguous, but odd in each of three senses. The absurdity can refer either to the content of the belief or to the act of believing; and if the latter, the reason can be either a rational justification or a causal explanation. In either of the first two contingencies, absurdity is hardly a good reason. In the third - if what he meant was that he was so constituted that he tended to believe things when he saw them to be absurd - the contention raises a paradox of a kind we shall be concerned with in a later chapter.
Commitment
49
fictitious characters as committing themselves, retracting, obliging one another to answer their questions, and so forth, and hence to be equipped with fictitious fact stores and commitment stores like the real ones of other people. Or consider the case of a statement made as a joke. Such a statement does not (we must reckon) commit either the joke-teller or its recipient, but it is not fanciful to say that they both indulge in a half-pretence that it does, and could not otherwise understand the joke. Sometimes, of course, the commitments get inextricably mixed, as when an actor makes an aside to the audience; and sometimes we become confused about who is committed to what or whether we care. But at least it is clear that these circumstances are atypical. The present analysis may help us to discern just what is atypical about them. If I were building a machine that should be capable of taking part in a question-and-answer dialogue the first thing I would design into it certainly before worrying about details of grammar and syntax - would be a suitable system of fact and commitment storage, with a program that specified how items were inserted, retrieved, copied, structured, simplified and retracted. Many of the details of this system can easily be imagined; others require further research. Such a machine could answer questions put to it in its language, and do so with a sort of well-schooled simpleton intelligence. It could hardly yet ask questions, unless at random, since we have not given it any kind of organised drive in this direction. Jones: Adam didn't finish his Art of Discourse, the book in which he wrote about questions. But if he had he would have written more about logic. Brown: Why is that? Jones: The book divides into symmetrical parts, and Part Four, the missing bit, would have been a continuation of Part Two. Brown: But if you haven't seen Part Four how do you know it would have been symmetrical? Jones: Well, that's how it seems to me, anyway.
CHAPTER FOUR IMPERATIVES
In some languages there are verbs with a hundred or more inflected forms in the indicative mood, but it is unusual to find more than one or two in the imperative. This has led many writers of grammar books, both old-fashioned and modern, to say little or nothing about imperatives compared with indicatives. There is also a long-standing prejudice to the effect that it is the prime business of language to state facts rather than to prescribe acts or anything else. Logicians, until recently, have been particularly susceptible to it. It is quite questionable. The truth is that imperatives, of one sort or another, are extremely common in our ordinary speech, very diversified in character, and sorely in need of further study. The present chapter and the next, which deals with yet other kinds of locution is intended as a mind-expanding enterprise rather than as an attempt to achieve a grasp of detail, which would take another book. The important difference between imperatives and indicatives and other kinds of locution is not one of inflection or sentence structure but one of a whole different field of use. Imperatives are sometimes equated with commands (or orders): the word imperative is from the Latin for command. But there are many kinds of imperative besides commands. Commands can normally be issued only by people in a position of authority, to those subordinate to them. This means that they are heavily dependent on social structures. Members of a government give commands to those they govern; officers of the armed services give commands to those of lower rank; and there are certain kinds of natural authority like that of parents over their young children, or of animal trainers or owners over their subjects. But in modern times even these relationships are in a state of change, and commands are really only at home in societies that are more authoritarian than our own. The instructions given an employee by his employer are hardly commands. I know of one case in which a professor of philosophy had to stand on his dignity and say That's an order when giving one of his lecturers a duty to perform, but this is rare enough to make the scene comical in retrospect. On the
Imperatives
51
other hand a return to authoritarian habits is an easy possibility at any time. If I were to kidnap you and threaten to deprive you of food I could soon get myself into the position of being able to issue commands to you. What is the authority-relationship between a computer programmer or operator and his machine? The console may be called a "command panel" and the program consists of "orders", or sometimes, more democratically, "instructions". However, computers are becoming more domineering. It is now common for a mere human to have to "request" some facility or other, such as access to a given reel of tape. Of course, this is nonsense both ways. Computers have no social role, since they are not members of our society. But it is interesting to see which social relationships their designers think it worthwhile to recommend us to imitate in our dealings with them. Requests are imperatives issued to superiors or, at least, equals in respect of the matter in question. It is consequently natural that they are usually made rather politely, with please, or Would you be so good as, though this is not essential. Pass the salt, please, and Lend me five dollars till Monday are requests. Circumlocutions such as Would you be able to lend me five dollars? And I would very much appreciate it if you would mind my suitcase for a moment on the face of them a question and a statement, respectively are perhaps more common in the case of requests than they are in the case of commands. The important difference, however, is one of linguistic use and social role. Another kind of imperative that may be used by inferiors to superiors, or between equals is the demand. This is altogether more common than the command in a democratic society where there is little actual authority but a great deal of interplay of small-scale sanctions. Trade unions make demands on employers and employers make counter-demands. If my neighbour makes a noise in the middle of the night, I shall first request him to stop it, then demand that he do so. Demands are requests backed by threat of a sanction, such as withdrawal of labour or calling the police. Commands differ from demands in that there is a sanction that is systematic, socially approved, and attached to a particular person. Naturally the three are not rigidly distinct but shade into one another. Sometimes someone who has authority may choose not to invoke it and will issue a request instead. A commander may say I won't order you not to talk to the press, but I ask you, for the good of the regiment, ... But subordinates may well feel uneasy when faced with a request of this kind. Generally, where one may command one may not request, and vice versa. We have heard of the second lieutenant who says Stand at ease, please! Commands, requests and demands differ in respect of associated
52
Chapter Four
terminology. Commands are given or issued, requests and demands are made or submitted. Commands are obeyed, requests are granted, demands are met. It is difficult to find neutral terminology for them. Yet they have a great deal in common. Most importantly, they have in common that the kernel of each is the specification of an act or action of some kind, normally of the addressee. This is also the case, with qualifications, of other kinds of imperative. It is characteristic of imperatives that they are act-specifications. This is, actually, also a slight oversimplification, but it is the best simple characterisation. In ordinary language it is quite common for what are really imperatives to be expressed in the grammatical indicative, or something like it. You are to see Mr Jones on Thursday morning is one idiom, and another is You should take the voucher to the cashier at the end of the counter. Should, of course, can represent moral obligation, but does not do so in simple cases like this. Requests, demands and commands form the first main family of imperatives. They differ in two very important respects from the second family, which consists of recipes, instructions, advice and suggestions. The thesis that recipes are imperatives may need argument, since we are inclined to think of them as stating facts about methods of manufacture, ingredients, oven temperatures and the like. Yet they typically use the grammatical imperative: Take three ounces of butter, melt, rub lightly into the flour recipe is itself by origin imperative, being Latin for take. Instructions are also usually in the imperative, though the word instruction is more ambiguous, and the instruction book issued with a piece of apparatus may contain much factual material as well as actual specifications of which switches to operate and levers to push. Once again the grammatical imperative is clearly at home: Assemble plug with striped lead attached to long pin. Turn on, allow warm-up period before pressing knob D. Advice should, perhaps, be regarded as the paradigm of the family. The simplest kind of advice such as route-direction, is again clearly imperative: Turn left at the second traffic light, go straight on past the Town Hall. In other cases, as in I think the most promising plan is to invest in nickel exploration, the imperative is more disguised, but it needs only a little thought to see what the principal function is. A suggestion is a light or loose piece of advice. However, the word suggestion has an unfortunate ambiguity in that one may also suggest the truth of an assertion. In this case a suggestion is indicative, not imperative. Discussions of "hypnotic suggestion" often founder in these shoals. It is clear that a hypnotised person will generally receive and act on suggestions that he should do such-and-such, for example, behave as a newly-
Imperatives
53
hatched chicken. But it is not at all so clearly true that he believes indicative suggestions made to him, such as that he really is a chicken. The main trouble here is philosophical, since it is not clear what it means to say that he does believe something. Now what is the source of the feeling that recipes, instructions, advice and suggestions are not genuine imperatives? These locutions (I can hear an objector saying) are factual statements in disguise: recipes state cookery facts, route advice states geographical facts, and so on. Of course there is something in this. But it would be misleading to say that they are not imperatives. It is much more to the point to say that they are imperatives that, unlike commands, demands and requests, share two important dialectical characteristics with indicatives. The first of these characteristics is that they are what I shall call nonwilful: they are not supposed to reflect, or to be pure products of, the speaker's own wishes or will. If I tell you that roses are red, what I am saying is not supposed to be something that I just want to say, but is subject to certain conditions outside me and independent of my will. My statement that roses are red must meet certain tests external to me and not merely dialectical rules of sequence and the like, but tests independent of these too or it would not be a real statement but only a piece of wantonness. Similarly if I tell you to turn left at the next traffic light or to treat your cold with vitamin C, these imperatives will be of the advice variety only if I am not just issuing them to satisfy my own desires but regard them as subject to external tests of rightness. If, having turned at the traffic lights, you find yourself obviously taking us in the wrong direction and I show no concern you could say Were you advising me, or were you just telling me to do it? If I gave it wilfully, it was not really advice. By contrast, I may make what requests I like, where I like. And if I am in a position of authority, subject to the limits of my competence, I may command how I like. It is true that those who command may have to account to higher authorities, but the purpose of a position of authority is to give its holder discretion within a range of possibilities, and that means that within this range his own wishes may dictate. He is not, of course, a good commander if his wishes are flighty. But it is not at all uncommon for commands even to be framed as wishes: Jones, I want you to go down The other respect in which the two families of imperatives differ is in the matter of accountability. Those who initiate recipes, instructions and advice are expected to be able to provide reasons for them not merely reasons for initiating them, but reasons for thinking them to be good recipes, instructions and advice whereas those who issue commands or
54
Chapter Four
demands or make requests are not. And perhaps the point that needs to be emphasised here is that this is a dialectical matter, not merely one of whether there are or are not good reasons. It is possible for me to make a request to you that is really in your own interest and which I could perfectly easily justify as a piece of advice; say, Please stop giving that rasping cough, which could be justified as being for the good of your throat, but which is a request so long as I ask you in my own interests rather than yours. Sometimes a command or demand is explicitly also a piece of advice, as when a ransom note says Bring the money in a small suitcase. You are advised to co-operate or it will be worse for you. But in typical advice the sanction that may attend refusal is independent of the advice-giver. The important distinction between advice and demand turns on whether he does, or would, give a rational justification if challenged. We have now incidentally demonstrated that accountability is not confined to indicatives. This has been known in principle to logicians since Aristotle, who distinguished "practical" from "pure" reason along the lines of the distinction between indicatives and advice-imperatives: both are accountable, or there would be no reasoning associated with them. The distinction runs, in one sense, quite deep in our intellectual life, since there are two sorts of science, pure and applied, and scientists tend to divide into two groups, one interested in the discovery of facts and the other in the discovery of ways of doing things. The two groups usually co-operate rather well but occasionally do battle as when the pure scientists decry projects (such as manned moon rockets) that imperfectly subserve the cause of discovering new scientific facts, and applied scientists retort that it is they who exploit the facts pure scientists discover. What we can do here to help adjudicate in the dispute though it is a complicated one, with political overtones is to point out that pure science should be concerned with acts as well as facts, imperatives as well as indicatives. The theoretical structures erected by science are a reflection of the accountability of the scientists' raw material. But information about how to do this or that is as accountable as information of the true-false variety, and as susceptible to theoretical study. To put the matter a little paradoxically, applied science can be as pure as pure science, and there should be no snobbery about the primacy of the discovery of truth. (I should add that this does not necessarily mean that I approve of huge expenditure on manned moon rockets, which seem to me not to be among mankind's most pressing needs.) One of the reasons for the neglect of the study of reasoning involving imperatives has without doubt been the notion that all imperatives are like commands, and hence wilful and not accountable. It is thought that one
Imperatives
55
need not have reasons for an imperative, and this is also tempts seers and sages to talk in imperatives and feel they need not justify them. In the long run, this unjustly degrades imperatives in the eyes of those whose interest is in rational inquiry. Philosophers have often talked in imperatives, from the seven sages Know thyself and Ockham's Entities are not to be multiplied beyond necessity to Wittgenstein's Don't ask for the meaning, ask for the use. Pronouncements of this sort sometimes tend to be treated as dogma outside the reach of rational investigation, and this in turn encourages a view of philosophy as a dogmatic activity. Philosophers could counter this tendency by making more conscious systematic use of practical reasoning. The linguistic apparatus needed, in a language user, to handle imperative locutions is rather like that needed to handle indicatives. Let us now turn our attention in this direction. On the whole, imperatives of the second family recipes, instructions, advice and suggestions raise more general issues than do the others, and it is better to let our discussion centre around them. Since the issuer of recipes, instructions, advice or suggestions is assumed to know what he is talking about, he must have some store of these imperatives analogous to the fact store discussed earlier. He may give advice (I cannot go on referring to "recipes, instructions, advice and suggestions" all the time, so will let "advice" do duty for them) gratuitously, or he may be asked for it. One asks for advice by means of a special sort of question, whether of yes-no variety, Should I feed my cat lettuce?, or of W-variety, What do I do this year about Christmas cards?, How much shall I tip the hall porter? The fact that the answers to these questions are imperatives, not indicatives, is one that needs to be impressed on all linguists and logicians concerned with the theory of questions: one may look far before finding any acknowledgment that answers to questions may be imperative at all. When someone is asked an imperative question he answers it, it may be supposed, by drawing on the contents of an internal store in the same way as, in the case of an indicative question, he draws on beliefs. For want of a better name, I shall call this store an act store. Perhaps it is more usual, in the case of imperatives, than it is in the case of indicatives, that the addressee of a question finds no stored guidance and needs to make a "decision" on the spot; but it is difficult to be sure that this is so, or that the question is meaningful, and we shall approach the matter from a different direction later. We may also postpone, for the moment, the question of how imperative act-specifications get into anyone's act store in the first place. The kind of imperative item that corresponds with a belief or a piece of
56
Chapter Four
information is a decision, determination or resolution that such-and-such should be done. Such items may be general or particular, specific to particular people, and may have temporal tags. They may also be of various more complex logical forms; conditional, disjunctive, negative. I shall comment on the variety of these forms later. But there is one feature of this variety that calls for comment here, namely, the relation between the subject and addressee of an imperative, and the doer of the projected act. Modern English has virtually only two grammatical forms of imperative, the plain, root or infinitive form Do, Be, etc., which are "speakerexclusive", and the same preceded by Let's, which are speaker-inclusive. (Not, it should be noted, Let us, which means something different. 23) But grammar is not always a good guide to logical form, and we should remember the alternative forms with should, are to, and so on. Most usually, no doubt, imperatives are in respect of proposed acts of the person to whom they are addressed; but there are cases in which it is reasonable to distinguish between the doer of the act and the addressee. And a particularly important part is played in dialectic by locutions with first-person subject, namely, in which the doer of the proposed act is the speaker himself. Whenever a participant agrees to an imperative for example, if he is asked Close the door, would you and says Yes his answer must be regarded as the first-person correlate of the second-person Close the door. The change of pronouns, if there are any explicit ones, is parallel with the change in the indicative case; thus, if I say to someone You are very happy this morning and he says Yes, his Yes can be paraphrased I am very happy this morning; and if I say to someone Come and see me this morning and he says Yes, his Yes must be regarded as equivalent to a first-person correlate of my imperative. On this view a first-person imperative implies the speaker's readiness, determination or decision to perform the act that it specifies. For its direct expression one may choose between a variety of forms with subtly different shades of meaning, Let me do so-and-so, I am to do so-and-so, perhaps most usually I shall do so-and-so: thus, I shall come and see you this morning. (But qualifications are sometimes required.) Now let us think how we may model a dialogue in which, say, P issues advice to Q. For the same reasons as before, we shall need to supplement the participants' putative act stores with imperative commitment stores, in
23
E.A. Levenston, in an article entitled "Imperative Structures in English", Linguistics, 50, (1969) pp. 38-43, makes this point and also a number of parallel ones concerning insistent forms (Do go, Do let's go), negation, and specification of subject and addressee.
Imperatives
57
which a running, effectively public, tally is kept of what advice has been issued and accepted. P, say, goes on record as advising Q to turn left at the first traffic light, or to tip the man the price of a pint of beer. And Q, unless he explicitly rejects P's advice, may be regarded as having accepted it and as officially intending to carry it out; that is, as having committed himself to a first-person imperative of the same content as P's second-person one. It should be noticed that a first-person imperative, seen as a commitment, is in effect a promise. The best simple characterisation of first-person imperatives is that they are promises.24 P may later rescind his advice, in which case his commitment is expunged. On the other hand he may give contradictory or otherwise impossible advice, so that Q will need to make a retraction demand. In all these respects the situation seems to be precisely analogous to the case of indicatives. Let us now turn to the differences. These are substantial, though less so than has sometimes been supposed. In the first place, since advice is accountable but commands and requests are not, there needs to be some recognition of the difference between them. Often enough this difference is clearly shown by otherwise irrelevant features of context. One goes to a doctor or solicitor for advice, not orders, and if the doctor orders me to go to bed he is, in a sense, exceeding his authority. However, the principal difference between the categories will emerge only if the person issuing the command or advice is asked Why?, meaning What reason is there for thinking I should do that? If he says Never mind, do it it was a command and if he says It would be a good idea because ... and gives a reason, it was advice. Somewhat different reactions are appropriate to the questioning of requests, recipes and so on but the distinction between whether the speaker does, or does not, attempt to provide a rational justification divides the two families of imperatives. We did not find any such way of distinguishing two families of indicatives.25
24 It is perhaps necessary to say that I am not suggesting that promises and the like are to be assimilated to "giving oneself a command" any more than a first-person indicative is a case of telling oneself a fact. First-person imperatives are addressed to other people, not oneself. The notion that one can give oneself a command is a strangely schizoid one that must be rejected: what would it be like, for example, to refuse to carry out one's own commands? 25 If there is any comparable indicative phenomenon it would be connected with (a) predictions, which are sometimes not fully rationalised, and may be such that we allow individuals to play their more-or-less wilful "hunches"; and (b) avowals. In both cases, as we have noticed, the failure to meet a request for reasons tends to change the nature of the assertion towards being a personal, subjective or I think
58
Chapter Four
I do not want to suggest that there is always a clear distinction between commands and advice in practice. Perhaps the distinction would be clear in a rigidly stratified society but, at all events, our society is not like that and imperatives often hover between the alternatives. A commander will sometimes provide a rational justification of his commands, if only because he expects readier obedience to them if he does so, having read Clausewitz.26 Similarly if I request you to lend me five dollars you may be reluctant and ask me why you should, and I may give you an explanation of my need. What these exceptions show is that we are not very strict about the distinctions. Some religious sects that stress that the interests of one are the interests of all seem to me to be preaching, rightly or wrongly, that there is no distinction to be made. Another important point in practice is that sometimes the recipient of the command or advice does not ask for a justification, and we may have no means of deciding what the outcome of his asking would have been: the nondescript imperative is accepted (or rejected) without question. Many of our everyday imperatives are of this character. Traffic signs can generally be interpreted as imperatives, and are loosely divided into mandatory and advisory, but their precise legal force is not closely dependent on their rational justification and depends to a large extent on precedents. A slippery kind of imperative which should perhaps be put in a category all its own is the invitation. If I invite a friend to dinner and say Come a little before eight, it is not clear whether I have or have not good reasons for stipulating this time; but he will in any case not ask for them. A second dialectical difference between imperatives and indicatives concerns the kind of justification that may be offered. In the case of imperatives this is a very complicated matter. In the first place it is clear that the justification of an imperative may depend, not merely on the fact that it follows from other, more general imperative principles or is composed of other more elementary ones, but also on facts that are relevant to the situation; so that there are relations between fact store and act store entries, and between indicative and imperative commitments. To justify his advice to Q, P may also have to refer to Q's wishes in the matter, since the advice may depend on what P thinks Q wants to achieve. (I shall have more to say about wishes in the next chapter.) Justification of imperatives hence involves a wider range of locutions than justification of indicatives. We should also notice that, as soon as the question of justification
one, and this suggests a different kind of treatment. But we shall notice some other similarities between predictions and imperatives. 26 "The good soldier is the one who knows what he is fighting for."
Imperatives
59
arises, most of the locutions involved become translated into grammatical indicatives anyway. Q asks P Why should I do that? and P replies You should do it because should, in strict grammar, is in the indicative mood; and so soon as it is used in asking for or giving justification of an imperative it tends to slide about from imperative to indicative meanings. Sometimes even a moral justification will be given. I shall also say more later about moral locutions; but it is at least clear that they are not necessarily involved in the rationalisation of imperatives. A third difference between the dialectic of imperatives and indicatives concerns answers to questions. If I am asked whether the lunch menu has trout on it, it is clear that there is an answer: either it has or it has not. But if I am asked, after we have studied the menu, Would you recommend the trout or the filet mignon? I may feel not merely that I do not know how to advise but that there is nothing to choose that the alternatives are evenly balanced and equally attractive. I shall say Whichever you like. In this case it cannot be said that I have failed to answer the question. Rather, I have given an answer that is just as much in order as Have the trout or Take the mignon, and is a rational alternative answer to them. Sometimes people prefer precise advice and find it disconcerting to be left with a choice, but to resolve their dilemma with an arbitrary answer is not so much to give advice as to go through a ritual parody of it, to decide for them. In other cases the right answer to an imperative question may narrow the range of alternatives but leave more than one of them open. Take any road leading north might be the answer to How do I get to Hatfield? This means that I should take whichever one suits me. More specific directions would have been necessary if I had wanted to go to Barnet. An imperative question may be answered by giving any non-null subclass of the alternatives, including the class of them all, and saying Whichever you like. There is no comparable phenomenon in the case of indicative questions.27 There are as many kinds of imperative question as there are kinds of imperative. What shall I do? may be a command-seeking or requestseeking question, and so on. The above point about Whichever you like answers applies to all of them. Now we may attempt some preliminary summing-up of the dialectic of imperatives. It might be thought that the speaker, whether human or mechanical, of a language containing imperatives would need to be equipped with as many separate stores or sections of store as there are kinds of imperative. But in virtue of what we have said about nondescript
27 Unless, again, it is a question of choosing a prediction from a number of alternative ones, which has some similarity.
60
Chapter Four
imperatives this seems inappropriate. In some sense the imperative is the same whatever kind it is, apart only from an indicator, which may sometimes be inexplicit or missing. At all events imperatives are, for the most part, subject to the same rules: the major exception is in the matter of accountability. When an imperative question like What shall I do now? is asked it may sometimes be so nondescript that any kind of imperative is acceptable as an answer, whether I advise you to go to bed and rest, or Please wait outside a minute, or I suggest a cup of coffee, or You must give me a blood sample, or whatever. We may say that "Silence gives consent" as before, in the sense that someone who does not explicitly reject an imperative addressed to him may be regarded as having accepted it. In this case it is entered (with appropriate alteration of pronouns) in his commitment store as an imperative of his own; that is to say, if the original imperative was a normal second-person one, he has "agreed to do" whatever is specified. In other cases he may want to ask for justification. If he does so, this means that he commits himself as taking the imperative as being of the second family; and if the other speaker offers a justification he may be taken to do so too. Though sometimes, of course, this is made clear in other ways, for example by the asking of the original question in the form Do you advise or a reply in the form I advise In my opinion you should or similarly for imperatives for other kinds than advice. It is, of course, possible that a speaker's behaviour should be inconsistent in some of these respects. He may, for example, make a request in which he says Please in such an icy tone of voice that it is clearly much more like a command or demand; or he may say If I were you I wouldn't go into the living room at the moment and, when asked why, say Please, I advise you not to, so that it appears that the advice is much more like a demand or request: he has prepared a surprise, say, and doesn't want it spoiled. Under pressure he may retract the I advise you and make it an explicit demand or request, Please, you mustn't. But all this means for our theory is that the classification system for imperatives in act or commitment store must be such as to allow for uncertainties. With the exceptions noted, the operation of act and imperative commitment stores will be very like that of fact and indicative commitment stores outlined in the previous chapter. Retraction and retraction demands are the same in principle, and the latter provide an escape (other than pure rejection) for someone issued with contradictory imperatives. There are many other special points to be made about the dialectic of imperatives and their theory. I shall add a few notes on those that strike me
Imperatives
61
as the most important. 1. There may be attached to an imperative an indication of a time at which it is supposed to be carried out. There are all sorts of possibilities here, since acts are sometimes specified as to be carried out "at all times" (or at all "relevant" times), or repeatedly, or when a given contingency arises, and so on. But if we confine ourselves to the case of a specific act to be carried out once, there are still two possibilities, or at least two extreme cases, which I hope it may not be too misleading to label "present tense" and "future tense". Grammatically there is no tense-distinction in imperatives: past tense would make no sense, and we make no distinction in the verb between Do it now and Do it tomorrow. (Now, in an imperative, means forthwith or immediately, not at the present instant, and Do it now is actually incompatible with the act's being, literally now, already done. Similarly always, in an imperative, means now and in the future, not at all times past, present and future.) The important variable is whether the time between "now" and the proposed execution-time is short or long compared with the duration of a locution. If it is long, and what we are discussing is, say, what shall be done tomorrow or next year, acceptance and rejection, accountability and the other dialectical considerations are much as we have already said. But if it is short, and imperatives are supposed to be put into effect as soon as said, things may be different. A pilot being talked down by ground control in a fog has no time to discuss the merits of his instructions, so that accountability is limited by practical considerations; and although it is open to him to accept or reject a given instruction, he does so perhaps not verbally but simply by carrying it out or not carrying it out. There are fewer dialectical dimensions in this situation since various alternatives are conflated. If I call out Look out! to you while you are crossing a road it is hardly even clear that this is an imperative, since Car! or Hey! would do just as well. 2. Connected with the timing of locution and act is the question of lapsing of an imperative, whether because it has been carried out according to plan, or because the time in which it might have been carried out has expired. A command or piece of advice in respect of an act already carried out, or which cannot now be carried out, cannot be regarded as being still in force, and consequently any imperative commitments in respect of a particular time must, in some way or other, cease to be commitments when that time is past. This, again, is a logically very complicated matter, not only because the forms of imperatives are very various but also because the question of our temporal perceptions and knowledge is introduced. But I may be pardoned also for setting this matter aside here. What I have called a "present tense" imperative is one that lapses as
62
Chapter Four
soon as it is said as soon, that is, as the addressee has had a chance to react to it. 3. Imperatives are commonly addressed to one or more specific people, but these people may or may not be the ones who are supposed to carry out the specified act. Marie Antoinette said Let them eat cake to a courtier, not to the starving citizens. But it is generally the case, of course, that the person addressed is expected to do something about it, or the locution would be pointless. If you and I gravely agree that Nixon should pull out of Vietnam but do not tell him, we may be making a moral judgment but we are not advising him. Sometimes imperatives are addressed generally, as "to whom it may concern". Everybody inside is addressed to everybody understood to be present. Somebody come and help is also addressed to everybody, and those concerned are expected in some way to see to it that one of their number comes to help. Nobody leave the hall is not, of course, addressed to nobody, but again to everybody, and those concerned must all refrain from leaving the hall. 4. I have called the contents of imperative locutions acts, but this is too restrictive a word. Not leaving the hall is not an act but abstinence from an act. Imperatives moreover sometimes specify a state, rather than an act, as in Be calm when he addresses you; and they sometimes specify a goal without explicitly specifying the means to it. A very useful fourfold classification of verbs and verb phrases has been suggested by Zeno Vendler: 28 (a) activity verbs like run, jump, laugh, write which refer to processes that can go on continuously for short or long periods of time; (b) accomplishment verbs like run a mile, write a letter, draw a circle, which refer to processes that take time but have a natural beginning or end; (c) achievement verbs like reach the hilltop, win the race, and spot, learn (a fact) and recognise that refer to happenings at particular instants of time; and (d) state verbs, like know, believe, love, have, and usually be (with a predicate), which are associated with periods of time but do not refer to processes going on or happenings taking place in them. Vendler says it is "tempting to think" that the classification may be an exhaustive one. Verbs of categories (a) and (b) are distinguished from those of (c) and (d) by the fact that they have continuous tenses: one can say He is running, I am writing a letter, but not He is learning the fact that He is recognising his mother, He is knowing the multiplication table, He is having ten
28
Vendler, Zeno, "Verbs and Times":, Philosophical Review 66 (1957), pp. 14360; also as chapter 4 of his Linguistics in Philosophy, Ithaca, N.Y., Cornell U.P., 1967.
Imperatives
63
dollars in his wallet or He is being rheumatic. (Except in derived senses, such as if an actor is putting effort into representing a rheumatic character.) I do not know whether it has previously been noticed that imperatives are really only at home with verbs of the first two categories. Basically this is because, when someone can be asked or told to do something, it needs to be meaningful to say in due course that he is doing it, or that he is not doing it. The verb in the imperative, that is to say, needs to be one that has continuous tenses. When it does not, as in state verbs of category (d), the imperative is always rather indirect. Be calm for your exam means Take steps to ensure you are calm for your exam, or Pull yourself together when you come to your exam, and the verbs here are of category (b). In the case of category (c) imperatives are even more indirect: thus Know your lines at next rehearsal means Read and learn your lines before next rehearsal. It hardly makes any other sense to tell someone to know something, in spite of the old idiom (from Arabic) Know, O King, that the armies of your enemies are gathering in the East. One cannot do anything about carrying out such a command or piece of advice; or, certainly, about refusing to carry it out. This means that it is not a genuine imperative at all. Some part of this objection also rubs off on the hypnotist's Believe you are a newly hatched chicken. 5. Negative imperatives have been much discussed by the few logicians that have written on imperatives at all. The ordinary negation of an imperative is still an imperative: the negation of Go is Don't go, and this is equivalent to Stay. But another kind of negation seems to give not an imperative but a permissive, as when Go becomes You don't have to go, which is equivalent to You may stay or Stay if you wish. Generally, the easiest idiomatic way of turning an imperative into a permissive is to add if you wish. Hence in a sense they are conditional imperatives; though this categorisation does not quite capture their force. It has sometimes been said that permissives like Stay if you wish are not really negations of imperatives like Go since on any particular occasion it may be the case that neither is in force. This seems to me to confuse the question of logical form of the locutions with the question of whether anyone is committed to either one at a given time. But this is a detail we need not pursue here. I suppose it must be necessary to distinguish between command permissives, request permissives, advice permissives, recipe permissives and so on. I have no appetite for discussing this distinction, which does not seem to raise any interesting difficulties.
64
Chapter Four
6. There is one further matter we must touch on concerning the dialectic of imperatives, namely, what kinds of presuppositions they may have. Here again we find a richer field than in the case of indicatives. Close the door presupposes not merely that there is an indicated door, but that it is open at the relevant time, that it will not close or be closed independently of the person addressed and that, on the other hand, he is able to close it if he wants to. This means that he may have several grounds for objecting to the imperative, apart from grounds he may have for refusing to carry it out. The same presuppositions may occur in imperative questions, for example in Shall I close the door? Once again we may say that a person who answers such a question directly with Yes or No may be taken to have accepted the presuppositions. They go, of course, into his indicative commitment store, again illustrating that there is interaction between the contents of the two kinds of store. The discussions of the past few pages show that there is much left to be worked out about the dialectic of imperatives and, surely, that the subject has been shockingly neglected. When we say of a computer that it obeys our orders or carries out our requests we are referring to quite mechanical and unsophisticated reactions on the part of the machine to the signals we feed into it. If we wanted to talk to it in English other than the eviscerated English that programmers sometimes use and which is hardly more than coded mathematics we should have to build or program into it a very much more elaborate set of facilities. It is not clear that it would be impossible to do this; and, in fact, it is beginning to be clear how it could be done. But, I repeat, there is much left to be decided. This is only an impressionist sketch. That all the facilities would be necessary in such a machine is less clear. But I would like, in due course, to suggest that even apparently irrelevant tasks such as translating a scientific paper into Russian could well call on many of the imperative facilities from time to time. Not only do imperatives themselves appear in scientific papers, needing to be properly identified for translation, but indirect reference to imperatives will appear also and will need to be "understood". I shall turn in later chapters to the use of the primary linguistic faculties for such apparently unrelated tasks.
CHAPTER FIVE EMOTIONS AND ATTITUDES
The grammar books' treatment of imperatives is bad enough. But no feature of language has been so neglected and mistreated as the expression of emotion. One would think that it hardly needs argument that one of the primary functions of language is to express our emotions. It has its most obvious manifestations in words like Alas and Hurrah, which do little else. But also many - perhaps most - nouns, verbs and adjectives have an emotive element. It is not possible to call something marvellous, or somebody a miscreant, without conveying an emotional attitude. Even where a tone of voice conveys more than the words in print do, it is often possible to pick out and identify particular idioms - the cocky, the snide, the sarcastic, the bland. And there are subtler tricks of juxtaposition and emphasis. The conspiracy of near-silence in the formal grammar books, and the mystical wool-pulling in so many more literary treatments, alike fail to do emotive locutions justice. Let us start at the lowest level. Alas, Hurrah and the others are called "interjections", which means that they are "thrown in", or "ejaculations", which means that they are "thrown out", and hence in both cases no part of true grammar. Or they are called "exclamations", which suggests that they are shouted out aloud and are like reflex actions, momentary, specific, and logicless because unlike true language. Or it will be said that Alas "means that the speaker feels sad", which makes it the same as the statement I feel sad, whose logic is in reality quite different. The word sad it is true, refers to an emotion, and may be used sometimes to express or convey the emotion referred to, but I feel sad, This is a sad state of affairs, I regard this as sad all say different things, and the plain Alas says something different from any of them. The same attitudes are to be found when we consider other less pure emotive words. To call a thing marvellous (I can imagine being told) is to describe it, just as calling it large or green is and even the latter words may convey emotions in some circumstances! Yes, but it would be ridiculous to pursue this line so far as to suggest that it makes sense to speak neutrally and impersonally about a thing's degree of marvellousness,
66
Chapter Five
as we can about its size or colour. I do not want to suggest that we can keep indicative and emotive descriptions distinct from one another, and it will in fact be one of the theses of this chapter that we cannot do so. But the alternative thesis that, since we cannot do so, there is therefore no difference between them is quite misleading and throws out two babies in different directions, retaining only the bath-water. Perhaps the trouble goes deeper than grammatical theory and is characteristic of our culture or civilisation. It is a feature of commercial and scientific life that emotion is supposed to be, at best, an irrelevancy and at worst, injurious to judgment and progress. Emotions, it is thought, should be indulged only during our hours of recreation. When young and in enforced attendance at church I once heard a preacher claim that the only "good emotions" were brotherly love and righteous indignation. This grudging attitude is reflected in the Dewey system of classification of library books, which allots the same classification number to books on Emotion as it does to books on Abnormal Psychology. It is true that extreme emotion often clouds the judgment, and it is also true that there are fields of human knowledge, such as physics and chemistry, into which emotion does not need to enter. But in so much of what we do emotions are, at least, part of our raw material, and the judgments we make are no worse for having an emotive element, kept (like all the other elements) in its right place. It is difficult, for example, to imagine any kind of political or administrative work well performed if no emotional considerations entered. It is also true that pure emotive words like Alas and Hurrah are not very different from natural cries such as the cries of animals. They are, however, conventional words in language and are different in different languages: for Alas a Frenchman says Hélas and a German Ach. Moreover, we do animals an injustice if we think that their natural cries do no more than express emotion, for many of them, such as the warning cries of birds, are more closely analogous to indicatives telling of the approach of a dangerous thing, or imperatives prescribing evasive action. The fact that they are instinctive is beside the point. Hence I am conscious that I may stir up fundamental prejudices if I suggest that emotive words have a logic; and even more so if I suggest that emotive language is mechanisable. What business could a machine have with emotional language? The attempt to build emotions into a machine must surely be a sterile, at best pointless, exercise? The answer is that this is one of the things we have already let ourselves in for in even starting to consider machines as capable of being engaged in linguistic transactions, or language as mechanisable. Emotions, perhaps, machines do not have;
Emotions And Attitudes
67
but then, they do not have the aims or purposes that may give rise to the issuing or acceptance of imperatives either, and they do not, in any humanly relevant sense, give or receive information. Yet we find it profitable to make them copy human linguistic behaviour in respect of indicatives, and perhaps imperatives so why not human linguistic behaviour in its other aspects? Only if we do this will we be able to get them to handle the more sophisticated of our linguistic tasks for us. There is one other argument that ought to be adduced here. It is possible that we want our computers, for most purposes, to be unemotional to act, as it were, as our slaves, and not let their own feelings intrude or influence them in doing what we tell them to. However, it is not always as easy to be unemotional as one might at first think, and one cannot achieve it without an understanding of what is involved. If animals are never unemotional this is because they have never learnt how to be. If I assemble a collection of nouns, verbs, adjectives and other words at random out of the dictionary into a grammatical sequence, they will, if the make sense at all be likely to make emotive sense; and I could not ban all the emotive words from my dictionary without banning many wanted descriptive words. Lack of emotion, that is to say, is achieved by first achieving an understanding of what is emotive and what is not. So if that is what we want, we must still first follow the same road and study the expression of emotion. This we may now proceed to do. We might as well start simple-mindedly by seeing what emotions there are, and trying to correlate them with "interjections". So far as making a list of emotions is concerned it is remarkable that most of the books on emotion are of no use to us. We could find a list in Aristotle,29 by searching their names out of the dictionary or Roget; but modern treatments such as those of Freud or Sartre30 are generally too theorybound and apply only in a restricted range. My favourite book on the subject is that of the naturalist Charles Darwin.31, who goes through the gamut in an empirical survey and has affecting illustrations of human and animal facial expressions and postures. Darwin discusses, in successive chapters: suffering and weeping; low spirits, anxiety, grief, dejection, despair; joy, high spirits, love, tender feelings, devotion; reflection,
29
Not in The Mind which ignores emotion, but in Rhetoric Book II; to some extent in Nicomachean Ethics. 30 Freud's theories are scattered through his voluminous collected works. Sartre, Jean-Paul, Sketch for a Theory of the Emotions, translated by P. Mairet. London, Methuen, 1962. 31 Darwin, Charles, The Expression of Emotion in Men and Animals. London, John Murray, 1872.
68
Chapter Five
meditation, moroseness, obstinacy, ill-temper, sulkiness, determination; hatred, anger, sneering, defiance; disdain, contempt, disgust, guilt, deceit, pride; helplessness, patience, shrugging; surprise, astonishment, admiration, terror, fear horror; shame, shyness, modesty, blushing. As a makeweight he considers signs of affirmation and negation. The grouping seems a little eccentric in places but I invite anyone who is inclined to criticise this list to try to compile a better one. I am not aware that we have advanced much in this respect since Darwin's day. Physiological discoveries, such as the discovery that high adrenalin level is characteristic of both anger and fear, are of doubtful relevance unless they reinforce insights possible without them, and the resemblance of anger to fear is a case in point since it has been recognized from the earliest times. We cannot match all these with simple interjections but I offer the following imperfect list: alas and oh dear for grief and dejection; a smooth ah for pleasurable feelings and a more emphatic one for astonishment and admiration; pooh for sulkiness or disdain; huh for contempt; ow for sharp pain; hurrah for high spirits; ugh or erk for disgust; tut tut for mild disapproval and impatience. It has sometimes been pointed out that pronunciation of some of these involves sounds that are not normal constituents of English words for example, ugh and tut tut, the latter sometimes written tst tst but really unspellable. We should add ha ha for amusement, an emotion which Darwin has solemnly forgotten. Many could be added that are more local or specialised, boo hoo and oo-er of schoolboy stories, my my with an intonation like a wolf-whistle, the humph of unenthusiastic acknowledgment, Mr Punch's cock-a-doodle-doo of triumph. It is remarkable that there is no word here to express anger. As it happens, this gap and several others can be filled by the expedient, frequently resorted to by both believers and atheists, of invoking the Almighty, or his son, or his son's mother, or one of their band of saints, or their opposite numbers in the nether regions. Many apparently secular exclamations such as Gee and Golly are of this origin. A hundred years ago many people said Lawks, and in Shakespeare they say Marry. If you use the apparently contemporary American Yeah man you probably do not realise that you are perpetuating a Nordic oath that is short for something like Yes, by the holy men: Swedes say Ja män. An alternative to invoking the deity is to call for the performance of a sexual act, normal or deviant; though actually bugger means Bulgarian, inhabitants of that country being at one time supposed to be more prone than ourselves to the activity in question. It would be interesting to know how many of alas, hurrah and the
Emotions And Attitudes
69
others are really truncated oaths or sex-words, perhaps from a preChristian era. With the modern decline in religious belief and spread of sexual tolerance it is clear that our descendants are in for a thin and unexpressive emotional life; or would be if it were not that words are so easily detached from their primary meanings. A large proportion of Darwin's list of emotions can be covered in some way or other by means of this expanded list of interjections. However, the question arises whether the list is really as wide as we want it to be. The word emotion may itself be a little narrow here: feeling is broader. I can feel conspicuous, confident, inadequate, friendly and various other things but I would not say that I can have emotions of conspicuousness and so on, though these feelings clearly border on emotions and ought to be considered. I can also feel hungry or thirsty, though hunger and thirst are not emotions in the ordinary sense at all. I also have desires, wants and wishes. And I feel pain, which is like an emotion but different. Finally, let us consider what would rather be called attitudes: the principal attitudes are various kinds and shades of approval or disapproval, but I can also take up a hostile, distant, friendly, menacing or equivocal attitude. Emotions, feelings, desires and attitudes overlap to some extent and we should be interested in the widest field, at least until we discover distinctions. There are some special interjections such as fie that might be captured in this wider net. It is not surprising that there are none for feelings of conspicuousness, inferiority and such like, which are much too specific. But a language could be imagined in which there were. We should now distinguish again between expressing an emotion and describing it. When I say I feel sore I make a statement with which you may agree or disagree: you may say Oh, do you? I'm sorry to hear that or you may say I don't believe you; you are a hypochondriac. In either case it is my feelings you are discussing or, more particularly, the question of whether I have them. But if instead of I feel sore I had simply said Ow!, although this would in some sense have conveyed the same meaning to you, there would have been nothing for you to agree or disagree with. You could still sympathise or accuse me of bluffing but you could not say that I was telling the truth or lying. The difference, in short, is primarily dialectical. If I say I feel sore I go on record with a commitment that is primarily an indicative one, like that of any other statement. If I say Ow! I go on record with a commitment that is of another kind. Now let us consider another case. We are dining together and simultaneously taste the quince pie: I say Delicious! and you agree. What have you agreed to? Not merely that it tastes delicious to me, for you would not agree if you disliked it yourself. Nor that it tastes delicious to us
70
Chapter Five
both, which was not what I said, since in the first place I had no right to prejudge your tastes. That it is delicious? Yes; but we would neither of us be unduly worried if others found it not so, and would not feel that there was an irresoluble conflict. Rather, you simply indicate that your attitude or feeling is the same as my attitude or feeling, which my locution Delicious! conveyed to you. And remember, I could have said the same thing with a pure interjection Ah!. This tastes delicious to me, like I feel sore, describes my feelings, whereas Ah! and Ow!, we might say, express them. But which does This is delicious do? Notice that we can have an analogy to this in the case of pain, since you and I may have identical sore arms following a vaccination and may agree It's painful. One is tempted to say that each of these locutions states an objective fact, and that if Ah! and Ow! have the same meaning as the longer formulations then all four must really be ordinary indicatives. But one could as easily say that their association with Ow! and Ah! suggests that It is delicious and It is painful are not really indicatives but (what I shall call) emotives. So, since we must make a subtle distinction somewhere, let us notice the difference between Delicious! with an exclamation mark and It is delicious without. The first, like Ah!, has a certain immediacy about it, denoting the actual presence of an emotion; the second, which can also be put into past, future and other tenses, may be used to state a fact unemotionally: that is, It was delicious said tomorrow can refer to the same experience as It is delicious said today, and it seems that I ought to be able, if it were not for my salivary glands, to utter It is delicious unemotionally even in the presence of the pie. Of course, I may also say Delicious, that pudding! tomorrow, still in reference to today's pudding that is no longer with us, but if I do so it will be because I am reliving the experience rather than just remembering it. There is always an air of drama about an emotive locution, an impromptu staginess with special use of voice or accompanying gesture, that we denote in writing by the exclamation mark32. To say Delicious in a flat tone of voice would be to misuse the word. The difference between an emotive and a simple indicative is often marked by word order: Shocking, that riot!, Huge, that tree!, Ingenious, that contraption!, Bitter, that wind!, Lovely, those marigolds! This word order is hardly available with less emotive adjectives: Red, that book is simply vapid.
32
I am sometimes surprised to find logicians use an exclamation mark to mark imperatives. This is another confusion. Exclamation marks are for exclamations, of whatever mood; but particularly for emotives.
Emotions And Attitudes
71
So what distinguishes emotives from present tense indicatives stating that a thing or event or circumstance has a certain property, if that property is one that is apt to arouse feelings in us? In the first place, as I have just said, only certain words are emotive. Secondly, I properly utter or agree to emotive locutions only when the feelings are actually aroused in me, not just when I agree that they might be or are apt to be. And, thirdly, emotives are relatively non-accountable. The first point needs no further comment. To underline the second we need only contrast the case of a discussion of whether this cake is delicious, between two people who have not yet tasted it, with that of a discussion of the same question between two people actively engaged in eating it up. In the latter case when either one says Ah! or Delicious! he may be regarded as committed to actually having pleasurable sensations, and it would be paradoxical in the extreme if he simultaneously maintained that he did not. Degree of accountability of emotives varies with the emotion considered. From Delicious! I can rapidly retreat to Well, it tastes that way to me which, as we have seen, is an indicative of a non-accountable variety, a so-called avowal. In other cases the chain may be longer, but we are never far from this termination. If I characterise the view of rooftops from my hotel window as inspiring and you ask me why, I may point out to you the separate wisps of smoke, the happy pigeons and the engaging geometrical rows of terrace houses as evidence for my assertion; but it is stretching logic to call this an argument, and before long I shall have to say Oh well, if you can't see ... This is not to say that emotives have no logical relations. Hurrah! contradicts Boo! and Marvellous! is stronger than Impressive! James Thurber quoted Gertrude Stein's poem "The pigeons on the grass, alas!" and objected "It is neither just nor accurate to connect the word alas with pigeons. Pigeons are definitely not alas. They have nothing to do with alas ... " in an essay filling out this misornithic point of view33. But ultimates, it is clear even here, are quickly reached. What we know as "attitudes" are generally more accountable than feelings. My reasons for my friendly or hostile attitudes, or my attitudes of approval or disapproval, are commonly rather rational. However, hostility may be due to plain dislike, and approval may be justified with I just do., which is more a weakened reaffirmation of the attitude than a genuine reason, and effectively indicates non-accountability. In fact, whether such a locution is accountable or non-accountable is more or less a matter of the speaker's whim. We might note that attitudes of approval and disapproval
33 "There's an Owl in my Room", from Thurber, James The Middle-Aged Man on the Flying Trapeze, New York and London, Harper and Bros., 1935.
72
Chapter Five
of particular acts or goals are often adduced as themselves providing justifications of imperatives; but that their lack of objectivity would normally be evidence that the imperatives in question should be regarded as non-accountable and as belonging to our first imperative category rather than to our second. Thus a dialogue such as Why do you advise me to do that? -Because I approve of it. -But why do you approve of it? -I just do., gives some cause for objection: the advice-giver is not really entitled to be so cavalier. Unrationalised approval might be grounds for a request, but hardly for advice; or perhaps it should not, in the first place, be dignified with the title "approval", but should be regarded as a predilection or quirk. We have an analogue here of the change of meaning of an obstinately held statement. What is the difference between a request (or command or demand) and the expression of a wish? It is not, I think, negligible; though, as so often in this field, idioms merge. I may wish you would come and see me, and say so, yet refuse to request you to do so - and certainly to command or demand that you do so, or even invite you - on grounds such as that I think you have more important preoccupations. Or I may request or command you to do a thing, because it is my official duty to do so while personally hoping, known to you, that you will not. The difference, in short, is a difference of commitment. An emotive commitment is not to be equated with an imperative one. But before going on to emotive commitments and their dialectic let us rehearse another argument for their existence, based on the distinction between emotive locutions and natural cries. Since natural cries can be construed as informative or imperative as well as emotive, the distinction is appropriate to other kinds of locution as well; but it is of especial importance here. Let us suppose that P has a certain emotion or feeling and cries out E! to Q: E may represent pain, delight, dejection, mirth or any of the others of our list. We may suppose, if we wish, that it is attached to some object O in the vicinity: P has just barked his shins on a rock, come across an impressive bed of tulips, or been struck by his publican's resemblance to President de Gaulle. Or he and Q have just discovered that they are going to miss their train. At first sight, one of the differences between a natural cry and a true emotive locution seems to be that the emotive locution is a learnt response and the natural cry unlearnt, however, this is not essential, since a socalled "natural" cry may still be to some extent a conditioned, rather than unconditioned reflex. Pavlov's dogs' mouths watered naturally when he rang bells, at least in the sense that the response was not like a linguistic one. Even to say that a response is "conventional" does not characterise it
Emotions And Attitudes
73
as linguistic. It needs to be a communication. Let us suppose, then, that P says E! to Q because he wants Q to react in a certain way. Is this sufficient for us to call E! a locution? Not quite, for I can make a meaningless noise to someone to make him jump or to attract or distract his attention. In these cases the hearer, although he reacts, does not react in a learnt or conventional manner. But even if he did, in the sense that his response was a conditioned rather than an unconditioned one, it would still not be sufficient. His response must be due to the fact that he correctly takes P's noise to be a communication. Now suppose P gives the sign E! of the existence of his emotion and Q correctly takes it to be this sign. Is this sufficient? Not necessarily, because Q could have correctly interpreted a natural sign in this way, without P's realising that he would or has done so. And even if P had so realised, Q might not have understood that P so realised. And even if Q did so understand, P might not have realised that he did or would. We have embarked on an infinite voyage of distinctions. In the case of a true locution, it seems, P must realise that Q understands that P realises that Q understands that ... without end. We met a chain like this before in discussing indicative commitment. After Adrian and Beatrice tell Celia something, they know that she knows that they know that .... The solution in that case was to postulate that all participants have indicative commitment stores. We have a parallel solution in the case of emotives. A natural cry, emotive or any other, may be the result of emotion in the being that cries, but cannot be said to commit that being to having the emotion. He or it is not regarded as under any dialectical pressures or obligations, cannot contradict himself or itself or be asked to provide a rationalisation of the cry, as distinct from being asked for its motive or cause. Hearers of the cry, though they may be emotionally affected by it, are not regarded as committed by it to any belief or action, and their failure to react will certainly not be regarded as recognition of or agreement with the emotive reaction. But in the case of a true locution all this is different. Your stylised cry of pain Ow! is intended to have me understand that you intend me to understand that you intend me to understand ... that you are in pain. It goes into your emotive commitment store, and something relevant goes into mine. This is, then, the difference between a natural cry and a locution: locutions commit, and natural cries do not. Locutions need to be tallied, and there are dialectical rules that take into account the tally. But in the case of emotives we still have one more, rather difficult distinction to make. It concerns hearer-commitment or, perhaps, what could be called
74
Chapter Five
the distinction between subjective and objective emotions. Let us suppose E! is a genuine emotive locution, not a natural cry. When P says E! to Q, E! goes in P's emotive commitment store. P is dejected, and says Dear, dear! Or Alackaday!, and the tally goes on to his slate. But does it go on to Q's? Is Q expected to feel dejected just because P does? It is possible that Q, whether through a similar present propensity or through suggestibility will agree with P that the circumstance of the dejection is an objective, shared circumstance appropriate to the same reaction in himself, and in others relevantly placed. He may say Yes in such a way as to imply agreement with P's feelings. Or he may accept the objectivity but reject the attitude, saying No in such a way as to indicate that he is contradicting P and thinks P should retract or revise. Or he may react in neither of these ways. It is possible, that is, that Q may go his happy or miserable way having perfectly understood and accepted P's locution but without incorporating in his own commitment store any item other than the indicative one P is dejected. If he dislikes P this may even make him feel happy. P's locution, that is, may be taken as subjective to P carrying no hearer-commitment. Thus the status of emotives regarding hearer-commitment varies very much not only from one kind to another but also from person to person and in different circumstances and societies and according to different styles of life. Some people, whether or not they actually feel or are affected by others' emotions, at least think it necessary in a wide range of particular cases to make clear in principle whether they are or not; others regard it as none of their business, except to note what emotions other people have and take them into account in a rational way. These attitudes are not closely correlated with attitudes of kindness or indifference, since I can very well be sorry for you if you are in extreme pain without regarding your painful experiences as a pain of my own. If you wince I can fetch the aspirin without having to wince myself. But some people would feel that they were called on to wince in sympathy with you, and would be somehow contradicting you or denying the painfulness of your pain if they did not. And again, this is a dialectical matter, affecting the expression of emotions rather than the emotions themselves. Either way of using these linguistic idioms is feasible, and either is consistent with any actual state of feelings. It is a matter of whether people consider emotive utterances to be, on the one hand, things different speakers should agree or disagree about or, on the other, personal to each speaker and hence such that different speakers do not get to grips over them. If we think emotives have no hearer-commitment we shall to this
Emotions And Attitudes
75
extent think their dialectical rules to be different from those for indicatives and imperatives. They may still be accountable, since I can expect you to give reasons for your emotive locutions without making this in any sense a condition of my endorsement of them as mine. However, there will not be the same need for locutions of the retraction variety since these will be required only to permit a speaker to retract an earlier locution and not for a hearer to reject one. The idea that a person should have some kind of store of emotions - we might call it a sentiment store - is at first repellent. Emotions, we think, come and go as psychosomatic processes dictate. But although this may be true for some kinds of emotion, others are both more lasting and less chemical, and are expected to be so. If one day I claim to abominate plastic bathmats and say Erk plastic bathmats! you will be surprised at my fickleness if, the following day, I extol them instead. This does not mean that the cause of any one of my emotions needs to be continuously operative; rather, once produced, the emotion is registered somewhere within me and emerges to order when its object is in question. It is possible that the actual causal mechanism by which it is retriggered involves psychosomatic processes, but this is a matter the linguist need not be concerned with. Let us, then, treat each speaker as equipped with a sentiment store, in parallel with his fact and act stores; and an emotive commitment store in parallel with his indicative and imperative ones. Their contents, we may suppose, consist of emotive particles or purely emotive descriptive words, either alone or attached to objects: to particular things or kinds of thing, actions or states of affairs or kinds of them. Alas, poor Yorick! and Hurrah for Grammar! are examples of attachment to particular things, Erk, plastic bathmats! is an example of attachment to a kind of thing. Fun, water-skiing! is an example of an attachment to a kind of action, Miserable that it is so overcast! an example of attachment to a particular state of affairs, and so on. The last example is of attachment to an actually existing state of affairs, and in so far as this is so it will be accompanied by an indicative commitment. Many emotions seem to be more appropriate as immediate reactions to an actual situation than as reactions to mere past or possible or distant ones, and to this extent there is often, though not as a matter of necessity, a correlation between contents of indicative and emotive commitment stores. It is especially common for a single locution to carry both commitments. Thus in practice many emotive locutions are inseparable from indicative ones, and the precise force of the respective commitments may be variable and controversial. (The same applies to the
76
Chapter Five
locutions that express wants or desires at the same time as requests or commands.) Are there questions that take emotive locutions as answers? Of course there are; but our language is lacking in suitably direct idioms and it is common for them to be phrased as indicative questions in forms such as How do you feel about so-and-so? or What is your reaction to the fact that ...? And in these cases they will usually be answered in the indicative too, with avowals. Instead of Alas! one answers I am sorry about it. The same change of idiom would be normal in the case of retractions; not No, not alas!, which is hardly grammatical, but No, I'm not sorry. This change of idiom tends to obscure the simplicity of the emotive idiom and one may suspect that it does not take place to the same extent in all languages. Retraction demands in connection with inconsistent emotives will be made and answered in the same way, and such rationalisations as occur will almost inevitably be so. Nevertheless it will still often be necessary, if we are to understand what is going on, to think of I feel such-and-such as an emotive rather than as an indicative avowal. Finally let me say something about attitudes to indicative statements: I mean such attitudes as belief and doubt. It has sometimes been thought that since these attitudes can be regarded as feelings and since imperatives are like wishes we are well on the way to a unitary theory of mental processes in which feelings are the fundamental kind of mental event. However, we should distinguish beliefs from belief-feelings, and this can be argued in detail. First of all, of course, belief-feelings are not an essential part of a belief. Neither are feelings of lack of belief - feelings of doubt or of puzzlement - an essential part of genuine doubt or puzzlement. Many of our beliefs concern things of which we are not normally even conscious. The fact that I believe the floor of my living-room to be solid is shown by the fact that I walk on it with confidence, and there is not necessarily any other contemporary mental correlate of my belief. But even if we set aside in a separate category these cases of what would sometimes be singled out as "implicit" beliefs, and concentrate only on the case of a belief actively entertained by person P at time T, I do not think it need be assumed that P has any kind of feelings connected with his belief. It may, of course be usual that some kind of feeling, such as that of being drawn towards an object, or a rosy glow as a sentence rings through the head, is experienced, but I do not think even this is so, and if it is so it is contingently so and hence irrelevant. The fact that we use the verb feel in this connection as an approximate synonym of think or believe, as in I feel that it is so, may be misleading here and belief may be quite different from the kind of thing
Emotions And Attitudes
77
we refer to by means of the noun feeling. On the other hand there can also be feelings of belief without belief. Suppose I am standing on top of a cliff and suddenly have a feeling I could step off and walk through the air without falling, but don't in fact try to do so. Suppose that in spite of my dizzy delusions I walk all round the clifftop sticking to safe paths and doing nothing abnormal or dangerous. In this case there is simply no evidence that I really believe I can walk through the air. My feelings are not relevant to what I really believe. People can have feelings of doubt concerning things they not only genuinely believe to be true, but even know to be true. So-called "philosophical" doubts are an extreme case: one sometimes has insane doubt-feelings about such things as logical consequences or the laws of arithmetic, and can seem to imagine having two apples and two oranges and nothing else and yet having some other number than four pieces of fruit. I remember having a curious metaphysical dream in which it was Tuesday and I had a fear that Wednesday, on which something important (I forget what) was due to happen, would disappear so that the next day would be Thursday. It would be quite wrong to say of such doubt-feelings that they were in any genuine sense doubts. We must, it seems, recognise belief-feelings as a species of feeling; but without identifying belief-feelings with belief. But the most important distinction is that belief proper is an all-or-nothing affair and has no degrees. What we refer to as "half-belief" or "strong" or "weak" belief, is to be analysed in terms of middling or equivocal belief-feelings, and is not true belief at all. Or, at any rate, this is so far as the first-person I believe that so-and-so is concerned. In the third person, it is true, we can say He half-believes so-and-so and justify it by pointing to his vacillating or uncertain behaviour in respect of whatever-it-is; but when I say I halfbelieve so-and-so I cannot be taken to be referring to my own vacillating behaviour but must rather be introspecting and reporting on my feelingtone. I half-believed I could fly, I might say, though prepared to add Of course I really knew I couldn't. To verify whether I half-believe something I do see how I feel about it. Half-belief, then, is not only not belief, but not even the same kind of thing. In fact there is a very simple explanation of the difference between belief and belief-feeling. We have already seen that I believe that so-andso is rather more than a report on an inner state: it is a statement which itself raises commitments in respect of the so-and-so that is the stated object of the belief, though not quite the same ones as the straight So-andso is the case. (They lack hearer-commitment and accountability.) The third-person He believes so-and-so cannot, of course, foist any such
78
Chapter Five
commitments on the person referred to. But I have belief-feelings to the effect that so-and-so does not raise the same commitments either, and is an altogether different statement. Above all, to say that I have weak or strong or middling belief-feelings does not raise the same commitments and could not, since I cannot half-commit myself to a fact or half-realise any of the dialectical consequences of committing myself to it. And I would not want to, though I suppose I could, commit myself to inconsistent or vacillating behaviour or action such as would give others evidence for saying that I half-believed a certain fact. Belief that a statement is true is quite a different thing from a feeling towards it, and this is shown by the difference between indicative and emotive commitments.
THE CENTRAL THESIS
CHAPTER SIX THE PARTS OF THE MIND
The reader who has persevered so far will be beginning to have a suspicion concerning the main thesis of this book, namely, that there is a connection between the grammar of our language and the structure of the minds of language users. Language users, real or mechanical, must have fact, act and sentiment stores, structured, interrelated and operated on in ways we can partly enumerate. At least, they must have some kind of arrangement of parts that serves the same purpose as this arrangement would. We shall later tighten up this thesis and modify it in certain respects. However, at this stage it is interesting to pause and note some historical parallels with the thesis as it is beginning to formulate itself. It seems to have been Plato who first asserted that the mind has parts. He approaches this assertion, oddly enough, from the side of politics. In his Republic,34 in discussing the ideal organisation of a community of people, he decides that there are three classes into which its inhabitants must be divided by profession: government, military and commercial. This is because, he says, there are three different "virtues" in which different people excel, and which fit each for a particular kind of life; namely, love of learning, spirit and love of riches. The easiest way to explain his meaning to his Greek audience is to give examples of different nations known to them who have one or other element to excess to the partial exclusion of the others. Thus the Thracians and Scythians were noted for their spirited character and were generally warlike, rather than learned or commercial peoples. The Athenians (Plato's own community, which had recently been faring badly in both warfare and commerce) were fortunately noted for their love of learning. The Phoenicians and Egyptians, on the other hand, were commercially-minded people without the other virtues. Ideally the three elements should all be present in a proper balance, in every community. In the process of his exposition Plato makes it clear that he is talking about the individual human mind almost as much as about societies of
34
The most relevant sections are Book IV, pp. 435-45.
82
Chapter Six
people. In practice, he thinks, different people excel in different virtues and should be arranged in social classes accordingly - those in whom spirit is prominent in the army, those excelling in love of riches in the commercial classes and those who love learning in the government. Plato, above all, thought that the business of government should be in the hands of philosophers; in this he is unlike most moderns who (rightly or wrongly, I am personally not qualified to judge) regard academics as on the whole best kept harmless in institutions. But Plato also believed, a little inconsistently, in the ideal of the balanced man, who had all three virtues himself and kept each in its proper place. Presumably the balanced man, though governed in the last resort by his love of learning, could turn a hand to soldiering or commerce and excel in these when necessary. It is not clear whether the balanced man is able to do better at commerce than the commercial specialist, or better at soldiering than the truly spirited, but it is clear that he is supposed to be capable of leading an all-round better life as a whole. And he is apparently better at government than those who love learning to excess. Plato has quite a distinct argument to the effect that the mind has parts, namely, that it is sometimes divided against itself. That hunger and thirst are in a different part of the mind from intellect (which is another name for love of learning) can be seen from the fact that sometimes, although hungry and thirsty, we decline to eat and drink when intellect tells us it would be better not to - for example, in treating a fever. Again, the desires are different from spirit, since the soldier in battle must overcome his desires with the spirited part of his mind. Other examples are given to support the division in three. This is a very important argument which has cropped up throughout the history of psychology. However, for many hundreds of years after Plato, psychology was dominated not by Plato's teachings but by those of his pupil Aristotle. It was only in the seventeenth and eighteenth centuries, in the writings of Descartes and the British empiricists (particularly Locke and Hume), that interest in Plato's division revived. In the slightly altered shape of a distinction between intellect, will and emotion, or more imposingly cognition, volition and affection - or more prosaically thinking, willing and feeling - it has since become a kind of official starting-point for the study of mind. Most of the psychology books of the nineteenth and early twentieth centuries have some division along these lines written into their tables of contents, and discuss different parts of the mind in different chapters. Roget's Thesaurus, which divides all concepts whatever into six categories for the purpose of making a classified list of words of the language, has Intellect, Volition and Affection ranged along with Space,
The Parts Of The Mind
83
Matter and Abstract Relations. In these disguises the distinction occupies so elevated a place that its validity can usually be taken for granted without examination. However, in an early Aristotelian Society Symposium we find it debated by G. F. Stout, J. Brough and A. Bain, 35 with the first-named in favour of it. Till the twentieth century the supremacy of intellect or cognition as the ruler of the mind had not been challenged. But in 1911 Samuel Alexander introduced his famous "conational psychology". He still took Plato's threefold distinction as his point of departure, but argued that cognition and emotion, though apparently distinct from the third element, are really "conational" at root.36 (Why "conation" replaced "volition" I do not know, but it should be clear by now that various writers in any case had different ideas about just where the dividing lines should be drawn.) In a paper entitled "Mind as Feeling" the Australian philosopher John Anderson in 193137 took the by now conspicuously untaken step of arguing that the basic mental activity is feeling rather than thinking or willing. Anderson's argument is that a mind, if it exists independently of other things, cannot be defined solely in terms of its relation to them as that which knows, or strives for, some external object, but must have qualities of its own. Feelings, he thinks, though they may have relations with other things, are themselves qualitative and hence real. To discuss this argument fully we should have to consider Anderson's contention that real things (whatever that may mean) must have qualities, and not just relations with other things. But the twentieth century's greatest theory of the mind is that of Sigmund Freud. This shares some part of Alexander's and Anderson's unitary theories. Freud has his own tripartite division of the mind into Ego, Super-Ego and Id, whose separate existence is demonstrated, as with Plato, by the fact of conflict between them. For Freud, all three parts are ultimately of the same kind and consist of forces, or libido. A close reading might, however, suggest that they manifest this character in somewhat different degrees, the Id being the most forceful or libidinous and the Super-Ego the least. There is a curious parallel with Plato here since, if different parts of the mind are to be in conflict, they must be to
35 Stout, G. F., Brough, J. and Bain, A. "Is the Distinction of Feeling, Cognition and Conation Valid as an Ultimate Distinction of the Mental Functions?" Aristotelian Society, Proceedings, old series, Vol. 1 no.3 (1889-90), pp. 142-56. 36 Alexander, Samuel, Space, Time and Deity, New York, Humanities Press, 1920. Book III, ch. 5. 37 Anderson, John, "Mind as Feeling". Australasian Journal of Psychology and Philosophy, 12 (1934), pp. 81-94.
84
Chapter Six
some extent of the same kind; and Plato actually contrasts love of learning and love of riches with spirit, surreptitiously making emotions or drives of all three. It is only when we think of intellect as having some kind of driving force that we can think of it as ruling or governing the other faculties, and it is only when we think of emotions as forceful or active that we conceive them as being in conflict with the will. Plato makes these adjustments by stealth, Freud more consciously. While saying this we should also recognise that Freud not only removes intellectual activity from its place of prominence but also virtually leaves it out of the picture altogether. In this respect he fails to give a satisfactory account of it, which is perhaps strange in a man whose own activities were predominantly intellectual ones. Freud's division of the mind is thus along quite different lines from that of Plato. The problems that interested him were different. Freud, with his theory of unconscious mind, added a whole new dimension to psychology; but he did not necessarily add to our understanding of the operations of the conscious mind. How well can we say we understand the mind today? I think most psychologists would be justifiably a little irritated by this question, because they do not subscribe to the idea that there is an entity called "the mind" that can be explored in the way one can explore a countryside or an organism. They do not think the mind has divisions in the sense in which a garden plot has, or working parts like a motor car. And in this they are undoubtedly right. The threefold division inaugurated by Plato, after all, is a very nebulous one. The categories are very fluid, and their labels have changed very much over the years. Plato's own descriptions are fluid: "spirit" is sometimes "courage" and sometimes "that by which we feel indignant", and in modern writers this is identified not only with "will" or "volition", but (as we saw) with "conation" or "striving", sometimes with "attention", and sometimes even (in the case of Stout in the article referred to) with "motor activity"! Similarly the words "desire" and "emotion" must be mercilessly stretched and squeezed if we are to make sense of the third category; for hunger and thirst are not obviously emotions, nor is pain a desire. It is never clear what sort of things the objects of cognition and emotion are or what their relation is to the objects of sense perception; whether, for example, knowledge is of things or only of facts or events and whether, on the other hand, emotions such as admiration and disgust can attach to facts or events as well as things. But in spite of all this, as a rough and ready categorisation of mental activities, the threefold distinction does have something to recommend it. The long tradition supporting it cannot be dismissed with just a shrug, and
The Parts Of The Mind
85
it is matched by some support from intuition. The question we should ask ourselves is: what are its survival value and intuitive strength really due to. If we can answer this we may be in a better position to appraise the distinction itself. I shall start with two negative answers. First, the distinction between intellect, will and emotion (or whatever they are) is not given to us by introspection; that is, by looking into our own minds with some kind of inward eye. Although I come to know about my own thoughts, feelings and (say) drives in some way that might be described like this, I do not observe in this way that they are the product of separate parts or faculties of my mind. Many of my emotions, in fact, are localised not so much in my mind at all as in various parts of my body, as when I have a sinking feeling in my stomach or blushes in my cheeks, a light head or a heavy heart. Pain is always more or less localised in this way. In these cases our observation of the feelings is clearly not an observation to the effect that they are the workings of a central emotive faculty, and sometimes it is difficult to separate it from straightforwardly sensory observation of the condition of the organs in question. In the case of the other faculties the introspective thesis is even harder to maintain. It scarcely makes sense to say that I observe my own intellect or will; and though I know about my own thoughts and drives (or some of them) it is not at all clear how I come to know about them, or that they originate in separate parts or faculties in my mind, or even that they are really distinct from one another. The idea that we can observe the operations of our own minds derives principally from Descartes; but he supports this mainly by a fallacious logical argument that confuses "thinking" with "knowing that one is thinking".38 But even those who agree with Descartes that we can observe our own minds do not necessarily say that we observe that our minds have three parts to them; and it seems quite implausible that any significantly large body of people could be found who claimed to observe this and were in agreement about it. Our observations in these matters are considerably more nebulous than the claim itself. Second, the threefold distinction is not derived from surgical or physiological study of the brain. If I dissect a brain I find various lobes, fissures and folds and various points of entry and exit of nerves. An attempt has sometimes been made to discover different locations in the brain for the different mental functions: the cerebrum (we are told), and particularly the cerebral cortex, is the seat of higher mental activity; the hypothalamus the control-centre for sympathetic processes and hence
38
Descartes, Rene, Discourse on Method (1637) Part IV.
86
Chapter Six
something to do with emotions; the striatum in charge of efferent nervous currents and hence of the springs of action. Other parts of the brain perform subsidiary functions such as analysing sensations and generating and properly distributing energy. However, the anatomical researches that uncovered all these supposed facts postdate, rather than predate, the theory of the parts of the mind and the account has all the marks of an attempt to fit the experimental results to the theory rather than vice versa. There are inconsistencies and unsatisfactory features in the identifications, and they do not bear close examination and need not be pursued here. This failure, of course, was predictable. Let us return for a moment to Aristotle. He criticised his teacher Plato, in a brief reference in his treatise usually called De Anima, or The Mind, 39 on the grounds that Plato need not have said that the mind has three different parts but only that it has three different functions. This started a controversy between those who preferred to think of the mind as an undivided whole and those who thought of it as having a geography. Aristotle's position was the better one for empirical psychology, if only because, by saving psychologists from the domination of a dogma, it freed them to recognise the facts as they found them. Aristotle does in fact give a much freer and wider-ranging catalogue of the mind's functions, and does not attempt to arrange them in neat categories. This is, however, beside the point if we are looking for the origin of Plato's threefold distinction. It is difficult to see that a claim that the mind has three functions is empirically very different from one that it has three parts, for no one takes the second claim so literally as to suggest that the mind can literally be cut up, or that one may walk around in it from one part to another. Could we discover three functions, as distinct from three parts, in the brain? This looks a more promising question. But when we examine it we come to see that we are all too likely to fit experimental results to theory, rather than vice versa, again. For what is a function? If we already think that thinking, willing and feeling are different functions we shall look for phenomena corresponding with them in the brain In fact we find that there are electrochemical operations going on in the brain all the time, in varying patterns; and if we were to look hard enough we might find a pattern that occurred in everyone's brain every time he stated a fact about red roses, another one that occurred when making a decision to go to bed, another in case of approval of an act of kindness to children. But electrochemical patterns are in themselves unclassifiable, or, at least, there is no
39
4lla23-b31.
The Parts Of The Mind
87
"natural" classification of them. If we were to find a classification in this way it would again be one dictated by our desire to find it. Consideration of computers can help a little here. The earlier computers handled their programs, which governed their calculations, quite separately from the numbers or other data on which their calculations were carried out. But it was discovered that this was unnecessary, and in fact in most computers programs and data are regularly all mixed up together in their store. We could say, if we wished, that the items in the store of a computer have different "functions" depending on whether they are program or data. But they "look" the same in the store, and it is only when we consider the operation of the computer as a whole that the distinction can be made. The electrochemical patterns of activity in the brain may similarly be completely indistinguishable from one another except in so far as they affect the behaviour of the whole organism. So let us come back to our main, and positive, thesis. If we really want to know what the mind is like, the best plan is possibly to stop trying to look inside it and to regard it instead as a Black Box whose internal workings we can specify only in so far as they can be deduced from input and output signals. And this, I conjecture, is just what Plato and Platonic psychologists have done, without realising it. One of the most distinctive characteristics of human beings is that they are language-users. We cannot study the mind directly, but we can study the pieces of language that go into it and come out of it. These do not tell us what it is like inside, and perhaps there is not much sense in asking. But they do at least tell us what it might be like, in the sense that they enable us to see what kind of mechanism could accept and produce the same kind of pieces of language. I do not mean that Plato or the psychologists would have, or could have, justified their picture of the mind in these terms. They did not realise they were studying language. They used, like all of us when we start to theorise about anything at all, an intuitive approach. But they were all language-users themselves, and they had formed the habit of thinking of the people around them as beings who exchanged linguistic messages of certain forms. If the categories in which we have put linguistic utterances in previous chapters are at all natural to us, they would have been natural to Plato and the others, all of whom spoke languages not markedly dissimilar from ours. Their intuitive approaches produced results not unlike our more conscious ones. Not all intuitions correspond exactly and our boundary lines between facts, acts and sentiments correspond neither exactly with Plato's nor with anyone else's. The cut-up is a little artificial at points anyway, and some modifications could be tolerated. The important function of the
88
Chapter Six
distinctions is that they give us a broad contour-map of the terrain. Which came first, the language or the language-user? This is a question that, since the only way I know of investigating the mind is through language, I do not know how to discuss. There are, of course, other kinds of behaviour than linguistic, and some of these may be relevant to the question of what the mind is like. But language is so much a part of human behaviour that we cannot meaningfully imagine language without humans, or humans without language. Perhaps, in the long term, our language changes, but then so do our minds. Some details of our characterisation of language may be contingent on a particular historical circumstance or style of social organisation, but, in general, language and our minds are indistinguishably adapted to one another. Now let us pause and take stock. What we have in fact discovered about the locutions of speakers engaged in dialogue is that they fall into (at least) three largely separate systems. In the first place they state facts, or retract them, or ask questions whose answers are statements. In the second they issue imperatives of various kinds, or retract them, or ask questions whose answers are imperatives. And in the third they give vent to emotive utterances and are capable of finding ways of similarly retracting them and eliciting them with questions. We have postulated three kinds of mental store in each speaker capable of playing his part in dialogue: fact, act and sentiment stores or, that is, stores for indicatives, imperatives, and emotives respectively. And we have said that each participant in a dialogue, as he takes part, may incur commitments of these three kinds which must be regarded as tallied on some kind of public slate; or, at least, that this model is satisfactory for the simplest cases in which each participant clearly knows what each other has said since the start of the dialogue. I have spoken of the commitments of speakers as contained in their indicative, imperative and emotive commitment stores. Some of the contents of stores are "structured" by rationalisation considerations; that is, by the fact that some items are held to be reasons for or justifications of others. How complete is this characterisation of locutions? Shall we discover that there are more and yet more kinds, and that they call for more distinct kinds of store, indicating an incomplete categorisation of the geography of the mind? I want now to argue that there is a sense in which we have more-or-less covered the ground. In subsequent chapters we shall be concerned with elaborating this model in certain directions, but not with introducing any parallel kinds of locution. We have already introduced all the kinds of locution that can be claimed to be dialectically elementary.
The Parts Of The Mind
89
Grammarians have sometimes been interested in classifying sentences into different kinds, and have made a start by considering the "moods" that a main verb can have. Verbs can be indicative, imperative, subjunctive, infinitive, and perhaps a few others - optative, hortative. But "hortative" is hardly different from "imperative", and both "subjunctive" and "infinitive" are only rarely and atypically the moods of main verbs. Hamlet's To be or not to be was an imperative question, and the few surviving subjunctive verb-forms in English, as in So be it, seem to be usable in main verb position only for third-person imperatives. "Optative" is not really an English verb-form, but the archaic Would it were so!, if that counts as optative, expresses a wish. Since we have admitted emotive locutions that have no verbs at all, our list of locutions is already wider than this. Some grammarians, for example Jespersen40, have produced modified classifications but I have not found any that is at all likely to lead us to much modification of our threefold one. This does not mean that there are no stragglers or difficult borderline cases. To start with, there are various utterly special-purpose locutions such as greetings, Hallo, How do you do! and Good-bye!. Whatever their origin, these go in a category of their own; but, after saying that, we can surely cheerfully forget them. Of course, in some cases we may be in doubt how meaningful a ceremonial locution really is. Cheers! (or Prosit! or Skål) could be considered a kind of magical imperative bidding the hearer have health and happiness, and Pleased to meet you could be considered an indicative or an emotive. But we shall have enough to worry about without going into detail over borderline cases, and should be very happy if this were the only kind of question left. A more serious contender for separate treatment is the performative locution, originally identified by J.L. Austin41: Performatives are of the generic form I hereby..., with verbs such as promise, vow, agree, offer, protest, repudiate, appeal, authorise, nominate, insist, denounce, deprecate, salute, ... In the first place it is clear that performatives are not to be identified with indicatives that report the truth or falsity of a fact, for though I can truly or falsely say He promised to pay for my lunch, if I say I hereby promise to pay for your lunch the fact that I say it makes it (under appropriate circumstances) a valid promise, and there is no sense in the
40
Jespersen, Otto, The Philosophy of Grammar, London, George Allen and Unwin, 1924, Ch. 22-3. 41 "Performative Utterances", in Austin, J.L., Philosophical Papers, edited by J. O. Urmson and G. J. Warnock, Oxford, Blackwell, 1961. This paper was originally a broadcast talk (1956). A more general account in Austin, J.L., How to Do Things with Words, edited by J. O. Urmson, Oxford, Blackwell, 1962.
90
Chapter Six
supposition that I could be saying something false. And if I say I denounce you as a traitor to our Party, in saying so I have denounced you. These are cases in which the act of uttering a locution involves the speaker in a set of commitments that is normally quite different from indicative commitments and seems to be characteristic of the particular verb used. Offers, protests, appeals, denunciations all raise commitments of various kinds but none of the commitments seems to be easily identifiable as indicative, imperative or emotive. Are we to have as many kinds of commitment as there are verbs occurring in performatives? A suspicion that this extreme shift is not necessary might be raised by the observation that I promise, one of the leading members of the performative family, has already been approximately accommodated as a first-person imperative. A promise is a promise to do something or other, and the doing is an act which, with a suitable subject, can be registered in an act store or imperative commitment store. A further suspicion might be raised by the fact that I say that (so-and-so) and I command you to do (soand-so) are apparently of performative form; yet the first is a thinlydisguised statement and the second a reasonable normal form of command. In fact the normal way of discriminating commands, requests, advice and the other kinds of imperatives is to say or I request ... or I advise ..., and so on respectively. And I deprecate his behaviour, regarded as a performative rather than an avowal, is a fairly clear example of an emotive, at least in the broad sense we require of that word. In fact it seems possible to provide dialectical analyses of nearly all these performatives, in terms of operations involving only the indicative, imperative and emotive commitment stores; and although these are performatives it is difficult to classify, they are not such as to call for a radical revision of principles of classification. If I say I bid five dollars! at an auction sale, I am making a conditional promise to pay, but though it is clear that I am also doing more than this in that I am playing a role in a complicated social operation it would not be reasonable to postulate the existence of any kind of "offertory commitment store" that registers bids, as a general feature of a linguistic situation. And if I nominate you as a parliamentary candidate by saying I nominate ... , I am instructing the appropriate officer to register your name and perhaps certifying that you are a proper person to stand or commending you to potential voters. These are all already classifiable activities and it is difficult to see that I am doing anything that cannot be characterised in such ways. I shall shortly produce a general argument to this effect. As a matter of fact, we might even consider some reductions in our catalogue. Various people, for example, have suggested ways of
The Parts Of The Mind
91
representing questions as locutions of other kinds. I do not think all the suggestions have been good ones, as we would sometimes very easily see by considering what commitments are involved, respectively, in the question and in the supposed translation. For example, it has been claimed42 that a question is equivalent to the disjunction of its answers, so that Is he in Spain or is he in Italy? would be equivalent to the statement Either he is in Spain or he is in Italy. The latter is what we would call the presupposition of the question, and it is, of course, perfectly true that someone who asks the question could be regarded as committed to this statement, as also someone who attempts to answer it - at least when it is a question in an information-seeking context. But here the resemblance stops. Someone who asks the question Is he in Spain or is he in Italy? having made the statement Either he is in Spain or he is in Italy cannot be regarded as repeating himself, and the person who asks the question of someone who is already committed to the statement cannot be regarded as trying to tell his hearer something his hearer already knows. Moreover, if the hearer is already committed to the statement He is in Italy, it would be unnecessary to ask him the question Is he in Spain or is he in Italy, since it would be known which answer he should give, at least if he is going to be consistent; but there could conceivably be some point in telling him Either he is in Spain or he is in Italy, since, although he would be in a position to deduce this rather simple consequence, he might not have done so and there might be some special point in bringing it to his notice. But above all, asking someone a question does more than making a statement, certainly more than the statement that is the question's presupposition. When I ask Is Mr Smith at home? I do not merely give him the tautologous piece of information Either Mr Smith is at home or he isn't, but put him under a linguistic obligation to tell me whether Mr Smith is at home or not; that is, to make to me either the statement Mr Smith is at home or the statement Mr Smith is not at home. The proposed translation of a question into a statement completely ignores this aspect of the dialectical functions of questions. One other proposed translation of questions works better. It has been suggested that questions are really imperatives, of the form Tell me either that ... or that ... or of the form Bring it about either that I know that ... or that I know that .... By the first of these translations Are you the driver of this car? becomes Tell me either that you are the driver of this car or that you are not the driver of this car, and by the second, Bring it about either
42
Harrah, David, Communication: A Logical Model. Cambridge, Mass., M.I.T. Press, 1963. Also Belnap, Nuel D., Jr. An Analysis of Questions: Preliminary Report, Santa Monica, System Development Corp., 1963.
92
Chapter Six
that I know that you are the driver of this car or that I know that you are not the driver of this car. The two are virtually the same for standard circumstances of question-asking, but the first excludes various "indirect" ways of answering questions and the second fails for non-informationseeking contexts such as examination questions (and perhaps in any case demands too much), In normal or standard circumstances both seem to preserve the principal dialectical features of questions. Either one is appropriate enough when the speaker wants some information and thinks the hearer can supply it, and both create a linguistic obligation in the hearer to supply the information concerned. Questions with presuppositions can be accommodated with a slightly elaborated formulation. The first form has been suggested by Belnap and compared by him with the second, which is due to Åqvist.43 It appears, therefore, that questions might reasonably be regarded as a special sort of imperative; presumably as commands, requests or demands, since the asking of questions is generally wilful and non-accountable. If P asks a question and Q says Why?, P may reply I just want to know, that's all or I just want you to tell me. Similarly retraction demands, as we have defined them, might be reducible in this way: if P is committed to three statements S, T and U, together contradictory, and Q issues a retraction demand, its function is to get P to withdraw at least one of the three statements and it could be formulated as the imperative Withdraw at least one of S, T, U or better Issue a non-commitment locution in respect of at least one of S, T and U. I have already used the word demand in naming these locutions "retraction demands", and this may or may not be quite accurate, but it does capture the general idea. Someone who asks a question, or issues a retraction demand, undertakes the same kind of commitment as someone who issues a wilful, non-accountable imperative and creates the same kind in the addressee. Hence we do not need to assume the existence of any special kind of "interrogative" commitment or commitment store. Similar arguments could be used negatively against attempts to reduce our three kinds of mental content - fact, act and sentiment - and our three kinds of commitments - indicative, imperative and emotive - to one another. A wilful imperative, for instance, might be supposed to be similar to the expression of a wish: Pass the salt is like I would like the salt, with in either case please as a possible additive. No doubt the two are sometimes interchangeable. But I would like the salt could also be uttered
43 Åqvist, Lennart, A New Approach to the Logical Theory of Interrogatives; Part I, Analysis. Uppsala, Almqvist & Wiksells, 1965.
The Parts Of The Mind
93
in circumstances in which it was completely understood that there was no question of anyone's passing it, as for example on a picnic when we had forgotten to bring it, and it would then be clear that it expressed a desire, not a request; and I could utter Pass the salt as accountable advice when I did not want any and merely thought it should be removed so that the dog did not knock it over. Hence there are two kinds of mental content and of commitment here, not one. And in all, as is by now clear, there are at least three. The three kinds are necessary, but have we yet proved that they are sufficient? How do we deal with a claim, if it arises, to have discovered others, or a refusal to reduce questions or performatives as suggested? It could be objected that our method so far has still too large an intuitive element in it. It may not be possible completely to remove this, but something more can, in fact, be done towards it. I shall describe what seems to me to an appropriate method of investigating the completeness of the picture, and give a (somewhat more tentative) suggestion of what I think the result of the application of this method is. Briefly, it seems to me that, with one exception, our picture of the language-using part of the mind is complete. There are no necessary additions of any importance. Let us denote by the phrase "mental state verbs" that family of verbs of which the typical representatives are know, believe, wish, feel, hate, resolve, see, hear subjects. They cannot be delimited exactly but a survey of the dictionary or Roget would soon provide us with a selection that would satisfy us as representative. For completeness we might have to include some verbphrases such as is angry about or feels hungry, but these are not likely to alter the general picture of the mind that emerges. These verbs can be classified according to the kind of object they take. There are four importantly different kinds of object available, namely: (1) indirect statement, generally expressed as that-clause with indicative, as in I know (or believe, or see, or am sorry) that he is wearing a hip-holster, or as accusative-and-infinitive, as in I know Cartwright to be a man of substance. The objects are the statements · He is wearing a hipholster and Cartwright is a man of substance respectively, with only trivial grammatical alterations. (2) indirect questions, introduced by whether or some other relative interrogative word, as in I wonder whether Susan tried the cheesecake or I know who wrote Buddenbrooks The objects are the questions Did Susan
94
Chapter Six
try the cheesecake? and Who wrote Buddenbrooks? Only a few mental state verbs can take indirect questions, and although know is one of them it is in this respect quite unlike believe and feel. I know who wrote Buddenbrooks, incidentally, cannot be regarded, like I know that Mann wrote Buddenbrooks, as fully specifying the fact that it says I know, but only as specifying that I know a fact satisfying certain conditions, namely, a fact which is an answer to a certain question. (3) act, expressed commonly with (agent plus) infinitive, or with thatand-subjunctive, or by means of a gerund. Indirect imperatives fall into this form, as in Someone asked us to dance a rhumba or They suggested that I should go out to Schönbrunn or The Foreign Office advised against appointing him to a section leadership, but so do many mental state verbs such as wish, resolve, promise. (4) simple noun or pronoun denoting a person or thing. Know is again in order, and all perception-verbs; as well as many emotion-verbs such as like and fear. It should be remarked, of course, that nouns and pronouns sometimes denote acts, as in (3), or facts or events or processes, expressible in statements as in (1), but if we are a little critical of what we mean by a "thing" we shall classify these under (3) or (1) respectively, and not here under (4). They do not really, that is, denote "things" but are nominalisations of facts or acts. Further, some mental state verbs take no object at all, so that there is really a fifth category. Into this category go many of the verbs expressing feelings and emotions, if we make due allowance for some special idioms. For example, be in pain seems best classified as an objectless verb. Many subdivisions of all these categories could be distinguished but they are not necessary to our present purpose. Nearly all the mental state verbs in the language take one of these kinds of object and no other: this, at least, can be checked by surveying them. There are, of course, individual exceptions, such as feel, which seems able to take almost any kind of object at all. In some of these cases, however, - including this one - we feel impelled to say that the verb concerned has different meanings in the different cases. Thus a feeling that it will be sunny tomorrow morning or that Mrs Gamp would be a good committee chairman is different in kind from a feeling of hunger, or pain, or stiffness. Again, knowing a person is a different kind of knowing from knowing a fact, and seeing a table is a different kind of seeing from seeing a conclusion.
The Parts Of The Mind
95
If we treat questions as imperatives along the lines suggested above we can dispense with category (2) in favour of (3). We adopt comparable translations in the case of the other verbs taking objects in this category. Thus I wonder whether he will come might be translated I am concerned that I should discover either that he will come or that he will not come, where the underlined clause is a subjunctive indicating an act. It also seems possible to treat objectless occurrences of mental state verbs as abbreviations of sentences in which objects are stated. If someone announces I feel satisfied we can ask him What about?, and if he refuses to be specific and says Oh, not about anything; just satisfied we shall gloss this as meaning that he is satisfied with everything, or at least everything within his immediate ken. When a mystic goes into a trance and announces that he is in a state of supreme knowledge we may be unable to discover precisely what it is that he thinks he knows, but we need not be in any doubt that he is claiming to know something; that is, some fact which, unfortunately, he is unable to communicate to those of us not so enraptured. Or it may be that he has had some vision of an object of the "thing" category invested with especially moving properties; but whatever it is, his "knowing" has an object. Again, if I feel sad it is usually about a fact of some kind such as that I shall never be a great pianist or that Mrs Jones has lost her cat; and if, on a particular occasion, I cannot identify any object, but just feel sad, it will at least make sense to say that it is about something imaginary or something that has somehow failed to register on my other senses. If these moves are accepted we are left with three kinds of object. Now it is clear that the first or "indirect statement" kind of object is characteristic of a set of verbs that refer, primarily, to the existence of items in the fact stores of their subjects - know, believe, doubt, and so on. Also, the second or "act" type of object refers to entries in the subject's act stores - intend, resolve, perhaps will, and the like. Of the mental state verbs not in these categories there are two kinds; (a) emotion or attitude verbs and (b) perception verbs. The former often attach to simple "thing" objects but apparently also often to objects of types (1) or (3), as in I am sorry that he could not come, I am happy that Johnson won the medal, I want Jean to see the ring events, everyone enjoys getting out of town sometimes. It remains to say something about perception verbs. Since our concern has been primarily with dialogue, and since dialogue does not necessarily involve the participants in the use of their senseorgans beyond the minimum extent necessary to permit the transfer of verbal messages from one to another, we can hardly be blamed for not having considered perception sooner. It is possible, of course, that this is a
96
Chapter Six
serious omission in the long run, for the same kind of reasons as were advanced in arguing for the inclusion of a sentiment store even in a machine designed to be unemotional. Briefly, it is possible that we need at least some logical analogue of the perception process in order to be able to handle the language of perception; centring around, for example, a register of real or apparent current perceptions. There is no special verbal form obviously characteristic of perceptions, though one is reminded of the archaic interjections (if that is what they are) Lo and Hark. These take objects of types (1) or (4), like many emotive interjections and emotive mental state verbs, and in general perceptions seem to be linguistically rather like emotions. But with these few remarks let us set the subject aside for a later chapter. That there are no other kinds of object for mental state verbs to take suggests that there are no other kinds of mental content. This conclusion cannot be absolutely definite, for two reasons; first, that there might be mental state verbs that we have overlooked, and second, that we might at any time decide to enlarge our concept of mental state and alter our language to introduce new idioms and concepts. It is possible, perhaps, that there is no sharp line to be drawn between those mental phenomena that are reflected in grammatical constructions in the way those described by mental state verbs are, and those that are not. Certainly there are many kinds of human mental phenomenon - instincts of various kinds, natural morality and aesthetic sense, susceptibility to conditioning, conflict and neurosis, forgetting, dreams, hypnosis, play - that, although of interest to psychologists, seem irrelevant to the project of building a theory of language. It does not seem to me that any of these is such as to involve any kind of elaboration of our model, much less the provision of new kinds of mental content; though I am prepared to keep an open mind on this point. But finally, we must meet an objection. We cannot - I can hear the objector saying - draw any conclusion at all about mental contents from this kind of source. What the latter part of this chapter has demonstrated, perhaps - subject, at any rate, to questions of detail - is that there are three kinds of mental content recognised within the English language; but this says nothing about what that mind is really like. The structure of our language could be completely misleading as a basis for a theory of the mind. There is some substance in this objection. To meet it it is necessary to take the thesis back and reframe it. Let us return to our original question. We saw that Plato, and many writers since Plato, divided the mind into three parts, and we asked: where did they get this theory, and where does it derive its strength? And it is not at all fanciful to suppose that our
The Parts Of The Mind
97
discussion of grammar answers that question. The theory owes its origin and force to the fact that English, and related languages, make just such a threefold distinction in reporting and talking about mental contents. But our thesis goes a little deeper than this too. In taking part in dialogue in English we can, it seems, incur (approximately) three kinds of commitment - indicative, imperative and emotive. And this is a matter that can be established grammatically. The objection may have some force where fact, act and sentiment stores and operations on them are concerned, but it does not touch our analysis of commitment. But now, granted that there are three kinds of commitment, isn't it a tenable thesis that a good or natural model of the mind could be built on the same basis? Whatever the Black Box may really have inside it - assuming that question to have meaning - isn't it the case that, if we succeed in showing that such-andsuch a structure could produce the sorts of output the mind produces, a very considerable advance has been made? Conformity to commitment rules has the effect of "filtering" the mind's output, and this produces, in any case, something of the effect of a set of mental content stores resembling the commitment ones. To speak in Jungian terms, the commitment-stores and associated mechanism function as a kind of persona. For the rest, it is now clear that in the remainder of this book we shall have to be concerned with exploring the relationship between mental contents and commitment more closely.
SOME ELABORATIONS
CHAPTER SEVEN EPIMENIDES THE CRETAN
All Cretans are liars, said Epimenides, an otherwise unknown, perhaps mythical, figure. He was himself a Cretan, so he should have known. Moreover, since racial pride is a common human characteristic, we might expect that his natural bias would be to praise Cretans rather than to denigrate them; and since lying is generally held to be at least a little reprehensible, a Cretan would not be inclined, for no good reason, to admit that Cretans are exceptionally prone to it. This might be held to give the statement a better than usual chance of being a true one, which is the conclusion St. Paul properly drew when he quoted it in a letter to Titus, his man in Crete.44 So far as this particular incident is concerned, that should have been that. But, as is well known, that is just what that has been very far from being. The Epimenides story has been used throughout history as an admonitory example of paradox. Epimenides, it is suggested, really meant that every statement made by any Cretan was a lie, and therefore false; whence, his statement, if true, since made by a Cretan, must be false; whence there can exist statements that are true and false at the same time. This is not a valid conclusion, but it is not nearly so wild as further ones that have been drawn. Opinion is divided about these: some people maintain that it follows that no good can ever come of doing logic, and others, whose faith in logic is unshakeable, hold to the more moderate conclusion that something must have gone wrong. Perhaps (say some of the latter) Epimenides' words do not constitute a genuine statement, at any rate when uttered by Cretans. In the early part of this century Bertrand Russell found a similar paradox threatening the consistency of Frege's class logic, and drew the conclusion that no statement can be genuine if it refers to itself. The paradox of greatest concern to modern logicians is actually not of the Epimenides kind but of the kind that occurs in the statement This
44
"It was a Cretan - one of their own teachers - who said: Cretans are always liars, low criminals and greedy bludgers, And what he said is true." Letter to Titus, 1, 12. (My digest of various translations.)
102
Chapter Seven
statement is false. If the this refers to the statement itself, then if it is true it is false; and if false, true. The conclusion may be a worrying one for the logical theorist but it need be no more than a nine-seconds wonder for the ordinary man, if only because anyone who went round saying This statement is false, meaning it in this way, would be rightly treated as bent on unwarranted mischief. Yet I do not think the same can be said about the Epimenides paradox in versions nearer its original form. People do make statements like All Cretans are liars, and these statements raise an interesting range of dialectical questions. The accent here should be on the word "dialectical". It is only when we put the statement All Cretans are liars in context as a statement made by a known Cretan to give information to a hearer that its true character as a dialectical paradox emerges. But logicians uninterested in dialectic could not have been expected to recognise this, and it is only at this instant of time that it is possible to provide a precise formal analysis. I start by remarking that to tell a lie is not the same thing as to utter a falsehood. When I tell a lie, it is not necessarily the case that what I say is false, but only that I believe it to be false. If I say Professor X isn't in today, because I think Professor X would not want to be disturbed by this particular caller, I tell a lie even if, unknown to me, Professor X has in fact gone out. Of course, there is an element of uncertainty in this categorisation, since we tend to assume knowledge of truth as the norm. If the truth is known by the speaker there is no practical difference between a lie and a falsehood. In general, however, there is a distinction to be made, even if we sometimes fail to make it and sometimes equivocate. In future, therefore, I shall use tell a lie in this special sense, namely, in the sense say what one believes to be false. Or, of course say what one knows to be false, if know is not understood as contained in believe. In the case of telling the truth we are in trouble over terminology since the phrase is clearly ambiguous. If I say Professor X is in believing him to be in, I could be held to have told the truth if he is out. That is, tell the truth sometimes, though not always, means say what one believes to be true. To resolve the doubt we would sometimes say of someone who unwittingly uttered a falsehood that he told the truth as he saw it or told the truth according to his lights, and for the stronger sense we would say that he told the real truth. Once again, there will be no practical distinction to be made if the speaker's beliefs in the matter are true ones. There would be no paradox in our here supposing Epimenides' statement to be true, or in our supposing it to be false. I am not personally sufficiently familiar with Cretans to be competent to offer an opinion, but I would suppose them to be in general no more mendacious than anybody
Epimenides The Cretan
103
else: that is, I think the statement to have the same modicum of truth as All Eskimos are liars, All Englishmen are liars or All Americans are liars. If Epimenides had not been Cretan - if his statement had been made by Pericles of Tyre or Timon of Athens - no objection could have been taken to it on grounds other than that of its falsity. This shows conclusively that the paradox is a dialectical one not a logical one in the usual sense of that word. Actually, in unfolding the dialectical paradox committed by Epimenides, I do not need to rely on interpreting his statement in the strong sense Every statement uttered by any Cretan is a lie. Such a statement, uttered by a known Cretan, will certainly produce a paradox for us, since it implies its own falsity; and, if it is false, some statement made by a Cretan must be true, and hence some other statement made by a Cretan must be true, which is a conclusion that can hardly be regarded as a logically necessary one; and the paradox is that it has been proved by a purely logical argument. But we shall not be interested here in this "tight" paradox, so much as in that raised even by the interpretation of Epimenides' statement as All Cretans have a propensity towards lying or Cretans commonly tell lies or It is unwise to trust a Cretan. To show what is paradoxical and peculiar in these statements I shall rely on an argument of the form traditionally known as a dilemma, namely, an argument that can be developed on each of two assumptions, which are together exhaustive of the possibilities. These are, firstly, that Epimenides, when pressed, acknowledges his statement to be among the statements by Cretans that he states to be suspect as lies; and secondly that, when pressed, he makes an explicit exception of his own statement and seeks to set it above or apart from other statements by Cretans. Either of these assumptions, I claim, is either paradox-producing or, at least, such as to make us regard Epimenides' claim as a very peculiar one. There is an assumption made in this argument, namely, that Epimenides is in fact pressed to declare whether he means his statement to be included among those it stigmatises, or not, and that he does so declare, one way or the other. It is no doubt the case that statements like that of Epimenides are often made and pass unquestioned, so that no paradox is explicitly developed from them. St. Paul, as we saw, let the statement pass. However, this fact does not necessarily set us at ease regarding the potential paradox. Let us, then, take the first horn of the dilemma. Let us assume that when Epimenides says All Cretans are liars he is reproached with But you are a Cretan yourself, aren't you? Do you mean that in what you have just said you aren't to be trusted? And replies with an enigmatic smile Maybe. We shall immediately be at a loss how to take
104
Chapter Seven
him. And it is not difficult to see our uncertainty is not just uncertainty whether to believe him or not, but something more fundamental. We are unsure exactly what we should take him as having said. We do not know how to make sense of the entries in his commitment store. An ordinary lie is a statement that the person who makes it believes to be false; that is, it is a case of his getting an item put in his commitment store that is not in his fact store. Since his hearers do not necessarily know what is in his fact store they do not necessarily know the lie from the truth; and in any case they do not have difficulty in interpreting the statement made and in seeing which item is in the commitment store. But here we seem to be in doubt how to enter up Epimenides' commitment store. He has, on the one hand, made a statement, All Cretans are liars; and then he has, on the other, told us that this statement is not necessarily to be trusted. In the first place, the statement goes into his commitment store; but in the second, shouldn't it be taken out again or modified? A moment's reflection will show us that to consider the second statement as a retraction of the first would not be realistic, since this would be like having Epimenides say, No, I take back that all Cretans are liars; really, that was only an untrustworthy suggestion, and this is not what his Maybe implies. So we are in the position of saying that Epimenides must have both a statement S and the statement S is untrustworthy in his commitment store at the same time. I have been using the word untrustworthy to mean liable to be a lie, that is, to apply to items in someone's commitment store liable to be in contradiction with items in his fact store. We can enlarge on our analysis of the Epimenides paradox - on this first arm of the dilemma - by showing that it is always paradoxical to say or admit that one's commitments are not backed by belief or knowledge. (I shall not here discuss the precise distinction between belief and knowledge: for present purposes both are expressed by fact store entries.) That is, it is always paradoxical to say or admit that one of one's own statements - which one does not retract - is a lie, or anything less than the plain truth. This follows, actually, from what was said earlier about the commitments of I believe that S is the case, which are the same as those of S itself except for hearer-commitment. Now if I state, say, that your train leaves at 3 p.m. and add But I believe it may be 2.45 you will understand the addition as a retraction and modification of the first statement. But if I reiterate No, it leaves at 3; it's just that I have a belief to the effect that it's at 2.45 you will not know how to take me. Don't you believe your own beliefs? you might say. I cannot say what my beliefs are at the same time as I sustain a commitment at variance with them.
Epimenides The Cretan
105
Oddly enough, what cannot be said can sometimes be shown. If I say Your train leaves at 3 uncertainly, or vacillatingly, I may reveal to you that my belief does not quite back my commitment. Or, if you ask me to return a book I have borrowed from you and I say You did not lend it to me but say it vacantly or shiftily - or a little too quickly or loudly - you will know that I am lying. And if I say it with a suspicion of an eyelash-flicker or a smile you may know that I intend you to know. You will perhaps even accept the commitment yourself and connive in the lie. But none of this alters anything that has been said. The distinction between saying something and showing it was drawn in another (but analogous) connection by Wittgenstein in his famous Tractatus Logico-Philosophicus. This book itself contains an interesting example of paradox, in the statement on its last page that the reader who has understood the book will know that all the statements in it are nonsense. This is clearly an Epimenidean or (as we may call it for short) Cretan statement. Is it meant to include itself, or not? I am not aware that this question has ever have been discussed. In fact, most writers on the Tractatus have been content to pass the statement over with at most a comment, filling the rest of their books with "nonsense" of the kind it refers to.45 But let us turn to the second horn of the dilemma. Let us suppose that Epimenides when we say But doesn't that mean that your own statement is to be mistrusted too?, says No, this is a frank insight into Cretans that I am giving you. Let us suppose, that is, that he says that, however unreliable other Cretans may be, and however unreliable even he himself may be on other occasions, his statement on this occasion is the truth. There is clearly nothing inconsistent in this contention; and, in fact, we have shown that this is the only consistent thing he can say under the circumstances - short, at least, of changing his mind and retracting his earlier statement. We cannot contend that this statement is paradoxical on the grounds of any kind of self-contradiction. Yet this statement too, in its different way, is a peculiar one. The paradox is that Epimenides has been
45 In appropriating Wittgenstein's distinction between saying and showing for this purpose I do not want to suggest that the argument from paradox is the only argument for it, or that it fulfills no other function for him. But he could, in fact, have made more of the argument from paradox than he did. There is something paradoxical, and Cretan, about any general theory of language or logic, in so far as it is self-referential; and attempts to express, in language, an attempt to transcend language - to draw a limit to language by seeing both sides of the limit, as Wittgenstein put it in the introduction to the Tractatus - are precisely what is here at issue.
106
Chapter Seven
put in the position of going bail for himself. The criticism of What I said was the truth is not that it is contradictory but that it is pointless. First of all it should be remarked that anyone who makes any ordinary statement such as Grass is green or This is a comfortable chair both puts it as an item in his commitment store and implies that it is an item in his fact store. The sense of the word implies in this statement is a special one that has sometimes been called a "pragmatic" sense: it is the sense in which, when I make a statement S, I may be said to imply that I know or believe that S is the case. To say that I know or believe that S is the case is, of course, just to say in other words that S is an item in what, in this book, is called my fact store. Consequently, I can say, by implication, something about the contents of my fact store, simply by saying S, and what I say may be true or false: that is, my statement S may be the truth (in that S is in both my stores), or a lie (if, though now in my commitment store, it is not in my fact store). Of course, the truth or falsity of the implied statement that S is in my fact store is quite independent of the fact that S becomes an item in my commitment store. (I am not concerned with whether S is itself true or false.) Now suppose you are looking for a book you have lost and, when I say You didn't lend it to me, a doubt arises as to whether I am telling the truth; that is, as to whether the item in my fact store is You did not lend it to me or You did lend it to me. (I ignore the possible alternative that neither item is present.) Can I reassure you? I can say, if I wish, What I told you was the truth, and this actually says in effect The item in my fact store is 'You did not lend it to me', just like the item in my commitment store. But this is something I have said by implication already. I am merely repeating myself. At best I am making explicit something that had already been said implicitly. To put it another way, I do not need to tell you I am telling the truth. True or false, my statement that I am telling the truth does not achieve any object for me. If, as is possible, I do think that I have not borrowed your book, then it is true that this statement says so, but so did the earlier one. And if, as is equally possible, I am lying and think or know that your book is on my shelf or in my drawer, my statement I am telling the truth merely tells the same lie a second time. Then why do people make statements of the form I am telling the truth? This is easy enough to explain. They do it for psychological reasons, to add emphasis and to help create conviction. I can say I did not borrow your book and that's the truth! in such a way as to show you that I am sincere - calmly, firmly, indignantly, belligerently or how you like. But the words and that's the truth! do not play an essential role here, for I could also say I did not borrow your book! calmly, indignantly etc., and
Epimenides The Cretan
107
achieve the same effect. The words and that's the truth! in fact, are a piece of logical superstition based on the very assumption we are now criticising, and could be replaced by various other superstitious phrases such as May the Lord bear witness! or Strike me dead! It should be noticed that an exclamation mark is appropriate to punctuate any of these. So let us sum up. All Cretans are liars, uttered by a Cretan, is paradoxical because there is such a short step from this utterance to a dialectical situation in which the speaker is called on to criticise or defend his own veracity. Both I am lying and I am telling the truth are, for different reasons, impossible for him, since they try to say things that can only be shown, not said. Broadly, they try to make comparisons between the speaker's commitment store and his fact store. Now we should notice that the trouble spreads from I am lying and I am telling the truth to related locutions: second-person locutions, questions, imperatives and so on. These locutions, however unparadoxical in isolation, are scarcely capable of being used in dialogue without raising the same problems as the first-person indicative formulations. At best, they have uses that are different from their ostensible or straightforward ones. Thus You are lying and You are telling the truth, whatever special roles they may have in expressing emphatic distrust or trust, or in raising or allaying objections to what has been said, cannot be regarded as simple statements of fact. This conclusion is straightforward as soon as we examine their hearer-commitment properties. If you say You are telling the truth to me I may be pleased by your confidence or, conceivably, elated by my success in deceiving you but the question of my agreeing or disagreeing with you hardly arises. At most a perfunctory or tacit agreement seems to be in order. You are telling the truth could be held to be a singularly pointless statement for you to make, since you could achieve the same commitment by tacit assent to whatever I had said. You are lying is similarly pointless if all it can provoke is a denial; and in practice it is usually more like a challenge to retract, which is a very different thing from a plain statement. As before, what either of these statements states could be shown without being said. But again, that is beside the point. A question such as Is that the truth?, since the possible answers Yes and No are both paradoxical, is clearly a peculiar question. In practice No would usually be counted not as a simple statement but as a retraction, and consequently the question itself might be considered alternatively as a retraction demand, and the answer Yes as a perfunctory (and non-paradoxical) refusal to retract. Similarly the imperative Tell the truth, since agreeing to carry it out and refusing to do so are alike paradoxical, can only have some kind of emphatic or admonitory force. In all these cases it
108
Chapter Seven
seems that the simple interpretations of the various locutions are barred, whence, to prevent hitches and blocks in our uses of them, we have invented secondary interpretations. Sometimes the presence of these secondary interpretations tends to hide the illogicality of the primary ones. But this should, by now, be sufficiently clear. One interesting area that becomes infected by these difficulties is that of the morality of truth-telling. It is often held to be wrong to tell lies, and a popular argument for this conclusion is Kant's, to the effect that if everybody told lies we should be in a sorry mess when it came to communication. Actually, the concept of a world in which everyone always told lies is hardly a coherent one, since lying presupposes some expectation that the hearer will take the lie for truth and be deceived. But in a slightly modified form in which it might appeal to some modern readers this argument might run: "We should, in any case, tell the truth most of the time. Telling the truth may not be intrinsically important, but it is important to tell the truth sufficiently often to create and preserve the commitments, presumptions and conventions of language. Although it may be occasionally in anyone's interest to tell a lie, it is in everyone's interest that most people should usually tell the truth, because even the occasional lie would be ineffective if they did not. We need to tell the truth at least often enough and on such occasions as to encourage a respect for truth". Kant would have rejected this argument in favour of one that made the obligation towards truth-telling universal. It is not, in fact, a very good argument. We should give it credit, however, for what it achieves. It illustrates that there is, as it were, a good game-theoretical basis for truthtelling, and a good ulterior motive for comparative truthfulness even in those who most wish to deceive. (It may, of course, be small comfort, when I buy an inferior product, to find that it is only, say, 10 percent of the advertising claims that are actually false.) The first objection to the argument revolves round the statistical element in it, since no guidance is given as to how one may choose one's occasions for lying; for if opportunism in these matters is allowable at all, it must be allowable on every individual occasion. Secondly, it rests on the prior, yet unargued, assumption that communication is intrinsically valuable; for which, if we are investigating fundamentals, we must demand a justification. But thirdly and centrally, the question is a Cretan one anyway. Only if, while we discuss the matter, we somehow exempt ourselves and our motives from scrutiny on the grounds of a possible interest in deception on this issue, can what we say be taken at its face value. The charge You say this only because you have a prior prejudice in favour of truth-telling and communication would bring this into the open, since either the reply Of
Epimenides The Cretan
109
course or the reply Not at all, I really think it would raise the same issues as Epimenides' I am lying or I am telling the truth respectively. I could attempt to show the desirability of truth-telling by inspiring others with my example of it; or, say, the undesirability of communication by generally remaining silent. But to specify these things explicitly is another matter. When philosophers debate issues like these, they sometimes seem to be engaged in a conspiracy of silence regarding the possible trouble they could get into if they did not rigorously avoid certain dialectical moves. Neurath spoke of philosophy as the attempt by mariners to rebuild their ship while afloat in it, from the materials it already contains. They must never, it seems, try to replace more than one plank at a time, however many are rotten. I am not personally of the opinion that so coy an approach to the job of rebuilding is necessary, or even that we need be unduly worried about holing the ship. The parts worth saving have their own buoyancy tanks. In drawing the line between what can be said in dialogue and what cannot, it is sometimes helpful to distinguish between what can be said by participants and what can only be said by onlookers. When Smith and Jones are engaged in a dialogue it is peculiar, in the terms described, for Smith to say I am lying or I am telling the truth or for Jones to say to Smith You are lying or You are telling the truth but as an onlooker I can say Smith is lying or Smith is telling the truth without running into any dialectical difficulty whatsoever. Similarly I can ask you Was Smith telling the truth? or argue to you Smith and Jones ought to tell the truth more; it would save them from these crises of mutual distrust. You and I, as onlookers, can have our own, non-paradoxical, dialogue about these matters. What onlookers say, of course, does not have to run the gauntlet of criticism by participants and, to this extent, it might be regarded as facile. If what I say about Smith behind his back is something I cannot reasonably say to his face it should, perhaps, not be given much attention. Nevertheless, onlookers' statements do not generally lead to paradox in the way the same statements may do when made by participants. For short, third-person statements are not paradoxical as first- and second-person ones are. Now let us turn for a moment to truth as distinct from telling the truth. Philosophical problems are beginning to enter this book in strength, and it is appropriate to raise the explicit question What is truth?, and to see what connection it has with the matters we have been discussing. We find, I think, that the connection is very close. Specifically, it is almost as difficult in practice to draw a distinction between "the real truth" and the contents of various people's fact stores as it is to distinguish fact store
110
Chapter Seven
contents from commitments. Traditional theories of truth are of such forms as that it is whatever it is "advantageously leading" for us to believe, or that it is correspondence with external reality, or that it is the coherence of the opinions of the "scientists of our culture circle".46 These theories give us, firstly, a way of interpreting any assertion of the form Statement S is true and, secondly, some guidance as to which statements, when we are in doubt, we should believe and which not. The first, or lexicographical, function of a theory is really a very trivial matter except in so far as it helps to fulfil the second. We shall see that the two functions can, in reality, be profitably separated, and involve quite different issues. First, then, let us consider what S is true means. I take it that this is equivalent to considering how someone who says S is true might alternatively phrase his assertion without using the word true. And this is a very simple matter: he can simply say S. Moreover, in place of Is S true? he can ask the question S?; in place of I believe that S is true he can say I believe that S, and so on.47 Only when the statement to which it refers is specified by description instead of explicitly - for example, in President Nixon's statement to Congress was, as it happens, true - is the word true not eliminable; but even in these contexts we can regard it as a place-holder whose function is to fill out the sentence so that it may obey the rules of grammar: a language can be imagined in which saying just President Nixon's statement to Congress, in a tone of voice that indicated that this was a complete sentence, was regarded as endorsing President Nixon's statement to Congress. In short, the word true is dialectically empty. I do not know whether anyone would want to contend that true should be given some sort of dialogue-free or absolute non-dialectical analysis
46 The first theory is that of William James in Pragmatism, a New Name for some Old Ways of Thinking; New York, Longmans, Green and Co., 1907; ch.6. Dispute between proponents of the other two theories has been going on for over a century in discussions of Hegel, but was revived among logical positivists in the 30s. The appeal to the opinions of the scientists of our culture circle was made by Carnap in a discussion article, "Erwiderung auf die vorstehenden Aufsätze von E. Zilsel und K. Duncker", Erkenntnis 3 (1932), pp. 177-88. 47 Strictly speaking, according to one modern logical doctrine, the "S" in I believe that S is true is the name of a statement whereas in I believe that S it is a placeholder for the statement itself. Thus note the quote-marks in I believe that 'Roses are red' is true but not in I believe that roses are red. The word true then, has the auxiliary function of cancelling quote-marks (or of cancelling the quotationfunction that should strictly stand over "S" in the first of the two sentences). The point deserves more discussion than I can give it here, but I may as well say that I do not think that it alters anything in the long run.
Epimenides The Cretan
111
that is different from this one. If they do, all I can say is that it seems to me that if one has given a translation of a word applicable to all occasions of its use, nothing more is called for by way of explaining its meaning. It is true that we sometimes distinguish between "the real truth" and "what people say", but it seems to me to be implicit in this way of speaking that "the real truth" is what people would say if they believed it to be true. And, of course, if I say The real truth is so-and-so I might just as well say Really, so-and-so. The "real truth", that is, is usually just what the speaker says, as distinct from the person or persons he contrasts himself with. Or at least, since this sounds cynical and is open to misinterpretation, when P says or agrees that S is the real truth, he is simply saying or agreeing to S and contrasting his stand with that of real of hypothetical people who would say something else. I think, therefore, that we need waste no further words on the problem of what true means. But this does not mean that we have a theory of truth, for we also need a set of rules or criteria for how we should or should not apportion our belief. Should we believe statements when they are advantageously leading for us, when they correspond with reality, or when they are affirmed by the scientists of our culture circle? The trouble with these prescriptions is that belief is not, in general, the sort of thing we can control. As doctrines concerning what we should say - that is, what commitments we should be prepared to incur - and what reasons we may properly give for saying it, they may or may not be acceptable But if I find myself unable to believe something that would be advantageously leading for me, or that the scientists of my culture circle all endorse, it does not seem right that I should be counselled to crush my belief. The correct answer to What should I say? is The truth. This is also the correct answer to What should I believe, Both answers, of course, are empty but I do not think any answer that did not have this kind of emptiness could be strictly correct. Not, at least, in all dialectical contexts. If you ask me What should I believe?, and if I am a fairly rational person, the answer I give you will have to be one that, generally speaking, correctly characterises my own beliefs. If I am a deist I shall give you, if I give a general theory at all, one that is consistent with my deism; and if I am a conservative member of the scientific establishment I shall tell you to believe what the scientists of our cultural circle tell you. And so on. I do not suggest that the individual beliefs always come first and the theory later, but, however they arise, in a rational person they more-or-less accommodate themselves to one another. If it happens to be one of my beliefs that the planet Venus is a captive comet, and you ask me Why should I believe Venus is a captive comet?, I cannot answer Because I say
112
Chapter Seven
so but I can answer, emptily, Because it's true. That is, this answer is at least grammatically satisfactory, in spite of the fact that it really merely reasserts Because Venus is a captive comet. I could, of course, give you my particular reasons for thinking Venus to be a captive comet, and these might or might not convince you. But I am unlikely to have any general set of reasons from which all my beliefs follow. Asked in general, What should I believe?, I am unlikely to be able to give you a general prescription that provides reasons for all, or even most, of my beliefs. The truth is the best answer I can give. Now it seems as if this explanation ought to be considered unsatisfactory on the following grounds. It gives, it might be said, an account of the truth in terms of what I say. The point of this chapter, however, is that we are never, in any practical situation, in a position to distinguish these two things. To be sure, when we are onlookers on Smith's and Jones's dialogue, it is easy enough to distinguish "the truth" from "what Smith says" or "what Jones says", I can say Smith says Venus is an original planet but the truth is it's a captive comet. But I cannot say I say so-and-so but the truth is ... When I seem to say something like this, as in speculating whether the truth might be different from what I say, I am really contemplating hypothetical circumstances in which I might withdraw or modify my assertion. Or, if I am not, my speculation is Cretan. Hence a theory of truth, other than an empty one, is impossible. Or, if I happen to have a theory of truth that satisfies me, this cannot be distinguished from a generalisation concerning what I am prepared to assert. Our general conclusion, that participants in a dialogue cannot distinguish uninhibitedly between the contents of their own or one another's fact stores and commitment stores, or between either of these and the real truth, has one rather disturbing general consequence for the investigation being conducted in this book. In discussing fact stores, commitment stores and the real truth we have been setting up a general theory which is supposedly applicable to discourse of all kinds - including the present discourse! It is true that this book, however much the author might welcome the participation of his readers, is for practical reasons a monologue and not a dialogue. But this does not make any difference of principle, and, in any case, anyone who disagrees and who puts up an alternative theory is inevitably involved in the rudiments of a dialogue, with all the relevant concepts and apparatus. But which theory of dialogue are we to use, his or mine? What each of us says about fact and commitment stores is self-referential, in that it applies by implication to the very debate of which our statements are part. I think this is without doubt a highly unsatisfactory state of affairs. But
Epimenides The Cretan
113
I also think it is clear that there is absolutely nothing anyone can do about it. Even to abstain from investigations of this sort altogether, as we pointed out earlier, is not particularly logical if others indulge in them and involve us by such means as asking us questions. All we can do, as it were, is to pretend to be mere onlookers on the dialectical scene, discussing Smith's and Jones's dialogue and not our own. And when something happens, as it occasionally will, to shatter this happy illusion, we must extricate ourselves as best we can and start again. If this means that the results of our investigations have less universal validity than we would hope, we should at least console ourselves with the thought that the same applies to every parallel investigation. And those who read philosophy for comfort and find it disquieting to have no firm ground under their feet should console themselves with their other blessings and reflect that philosophers themselves get used to it.
CHAPTER EIGHT MODELLING OTHER MINDS
We have seen that the mind of any participant in a dialogue may be approximately represented as a tripartite device, consisting of stores of facts, acts and sentiments, with appropriate machinery for the insertion and retrieval of items and a parallel register of commitments. But at several earlier points we have noticed that, in order to carry on an intelligent conversation, a speaker needs to know a lot about his hearers' backgrounds. He needs to have an appreciation of how his hearers think and feel, what he can take for granted that they will know and how he can expect them to react. For the most part this knowledge is implicit in the cultural background and common humanity he shares with them. This knowledge is only partly "psychological" knowledge of the kind that can be learnt from books. I know that most people like eating apple pie, can tell me what country Washington is capital of, and get disturbed or angry if spoken to sharply, without necessarily having items in my fact store of the form "Most people like eating apple pie" and so on. My knowledge of these facts about other people is more in the nature of a tendency to expect them to react in certain ways than explicit theoretical knowledge of facts. Very often, in fact, human beings make mistakes when they try to translate their intuitive knowledge into explicit generalisations. This suggests that there is something missing from our model of the mind, connected with our grasp of the contents of other minds than our own. To some extent this is connected with the grammar of quotation and of indirect speech, as when I say of someone that he knows that such-andsuch is the case or wants me to do so-and-so. I shall not be concerned with the philosophical problem of whether we really know of the existence of other minds than our own; though there is, perhaps, some marginal relevance to that problem in the fact that our assumption that we have such knowledge is deeply embedded in our language. I want to maintain, in short, that our linguistic behaviour indicates a quite particular facility at modelling other people's minds within our own. We can think ourselves into the minds of other people, and imagine the
Modelling Other Minds
115
reactions of real or hypothetical people under wide ranges of circumstances, in a way that is inconsistent with the assumption that our knowledge of these people is theoretical and that the manipulations of the locutions representing their commitments and mental contents all have to be performed, as it were, on statements that represent these commitments and contents between quote-marks. But to make this clear I shall first have to return to some grammatical considerations. Our classification of locutions was into indicative, imperative, emotive and a few stragglers. But besides uttering these locutions ourselves we also sometimes report their utterance by others. The most straightforward way of doing this is by direct quotation, as in He said 'It's breezy in the garden' or The dentist asked 'Does it hurt when you bite?: or 'Come to Monday lunch', he suggested or His only reaction was 'Pity it broke!' We use the same device for hypothetical or proposed locutions as in If he had said 'Good morning' I wouldn't have minded so much or Ask him 'Did you give her her essay?'. In all these cases certain locutions occur not spoken, asked or issued as genuine statements, commands and so on but in a more shadowy way as constituents of dialogues that are reported or imagined or projected. In one sense it would be proper to say that these are not really locutions at all since, when they occur between quote-marks, they do not commit any speaker or hearer in the way that is characteristic of a real locution. But this, of course, under-states the situation. It is better to think of quoted locutions as having the same function in a hypothetical dialogue as ordinary locutions have in a real one. Besides using quote-marks (or their verbal equivalent) we may use a set of idioms that can be lumped together as "indirect speech". They include that-clauses, whether-clauses and other w-clauses, accusative-andinfinitive, gerundives and various kinds of nominalisations. In most cases these do not indicate the precise words of any locution but only its general sense. For our purposes we need not distinguish direct from indirect speech. Sometimes, in fact, the distinction is not clearly made anyway. We need a word to describe the locutions of direct and indirect quoted speech and I shall call them "encapsulated" locutions. There are also other more elaborate ways of encapsulating locutions than these two. For example, you and I might act out a dialogue we have overheard or would like to imagine. A play as performed in a theatre is another example; where the actors' locutions do not commit them personally - still less the audience - and it is only at a second level that the author and actors may be said to be using the situation to "convey" something to those present. To understand the play, we need to imagine its participants as equipped, on behalf of the characters they represent, with a complete mechanism of
116
Chapter Eight
commitments and mental contents separate from the ones they enjoy in real life. This, at least, is the case with orthodox representative drama, whatever analysis may be offered of other kinds. Another kind of encapsulation occurs when something is said "as a joke"; it would generally have no point if it did not represent a hypothetical reality. It is perhaps less obvious that the same sometimes needs to be said about lying. At any rate if a lie is at all elaborate it will be necessary for the liar to imagine for himself a partly new personality or set of mental contents, which he faithfully represents in his lying locutions. That there are cases in which the fiction becomes fused with the reality, I do not doubt; but, at least, we must now do something to see how we may represent the fiction. The apparatus of quotation and of indirect speech gets taken over for a task different from that of representing what people say, namely, for representing what they think. Or feel, or want, or wonder, and so on. This is, in fact, one of our strongest arguments for the thesis, advanced earlier, that our conception of the mind and mental contents is primarily derived from grammar. When I want to say what is in somebody's mind, the only way I usually have of doing so is in terms of how he would express it in words if he were to tell us himself. We usually use indirect speech, as in He thinks that the garden is breezy or She wondered whether the library would be open. But sometimes, of course, especially in novels, we find direct speech also, as in 'The garden will be breezy', he thought or 'Will the library be open?' she wondered. Or we omit the quote-marks. In comic strips in which a character's spoken words appear in a balloon, his thoughts sometimes appear in a similar form in a balloon with a dotted or puffy outline. In the case of emotions, perhaps, we tend to use what might be called a description of the emotion rather than a direct or indirect locution, but even here the latter is not uncommon and is sometimes the only possibility. That the contents of our minds are normally described in terms of how we would express them in words is not very surprising. But we have seen in the previous chapter that there is an important restriction on the description of any such content, namely, that although words can be quoted or described by participant and onlooker alike, my thoughts cannot be described by myself or other participants in dialogue with me, without crossed-wire involvement in commitments or similar dialectical phenomena. If as onlooker I say P thinks that flowers have feeling I commit myself only to a fact about the contents of P's thoughts; but if I say I think that flowers have feeling, I convey no more about my own mind than if I had said Flowers have feeling, but I do commit myself to flowers
Modelling Other Minds
117
having feeling in the same way as this second statement does. And if you say to me You think flowers have feeling you involve me dialectically in the question whether flowers have feeling in challenging me to accept or deny it. Wittgenstein once said that we ought to find it strange that we use the same verb in saying I believe so-and-so as in saying He believes soand-so. His distinction is an important one, but drawn in slightly the wrong place: he should distinguish not just first from third person but rather participants' remarks from onlookers'. Moreover, a large range of similar words - taking not merely indicatives but also imperatives, interrogatives and emotives as objects - share the distinction with believe. It is not necessary to argue in detail the cases of other moods than indicative, but it should be clear that I wonder whether so-and-so is different from He wonders whether so-and-so; that I would like my bags brought in, said to someone whose business it is to see to it (a complication of the imperative case), is different from He wants his bags brought in said to a bystander; and that My foot hurts is different from His foot hurts. The last-mentioned is worth following through (though not here) for its relevance to the debate over the intensionality of painstatements. But although participants in a dialogue are inhibited by dialectical restraints from saying what is in the minds of other participants, it is certain that they are not indifferent to the question of what other participants think and feel. As I talk to you I continually form to myself hypotheses about your mental contents; not necessarily because I do not believe what you say (though that is one possible ground for speculation) but because life is too short for you to say all you think, and similarly for me, and because I shall want to apportion my conversation according to what I think needs to be said to you. It follows that, as P talks to Q, he is forever manipulating in his own thoughts items that are in some sense like Q thinks that S, Does Q want A done?, Pity Q worries about T's being so, Does Q suspect I think he wonders why I said U?, and so on in great complication. But although I can think these things about you when I talk with you, I cannot, or do not need to, think them about myself; and so there is, after all, a difference between the first person and the other cases. It is a symptom of the schizoid person that he seems to see himself "objectively", as he sees other people, and make an exception to this rule; but precisely for this reason the schizoid person is described as "split", as if he were really two minds in the same body. (We all, of course, occasionally debate with ourselves, try to be objective about ourselves, and so on, without necessarily becoming pathological cases.) Now let us think how these thoughts about other people's mental
118
Chapter Eight
contents are represented in our own minds. Specifically - since the enterprise of turning our eyes inwards is an unrewarding one unlikely to result in much apart from strained eye-muscles - let us think of the problem as one of how the relevant operations might be programmed into a machine designed to speak the English language. It is by now clear that knowledge of, questions and feelings about, and imperative drives concerning other people's commitments and mental contents are highly relevant to our ability to carry on intelligent conversation with them. And it should by now be equally clear that conversation with a machine presupposes the same "mental" contents (or analogues of them) in the machine. It would be possible to handle this problem purely as one of understanding the grammar of quote-marks, and their equivalent in indirect speech. In the first place, knowing the rules of dialectic for "zeroorder" locutions - that is, for locutions not containing other locutions in quotes or in indirect speech - already implies knowing the logic of locutions of the form P said quote S unquote, or perhaps P at time T said quote S unquote or P in circumstance C said quote S unquote. (I apologise for telegrammese in the formulation of these locutions but their nature must be made as clear as possible. I dare hope that this way of expressing things makes it clear that I am literally quoting and not naming, and that nevertheless I may use the name S of a statement, between the words quote and unquote, to indicate a quoted statement. And if this displeases purists, so be it.) Now all the indicative statements I make about other people's locutions can be regarded simply as indicative commitments of my own, of first-order forms; that is, containing zero-order locutions between at most one pair of quotes each. And if I make statements about other people's locutions about my own or other people's locutions these will be of second order and will generate commitments containing quotes nested between quotes. Similarly for higher-order locutions, and similarly for my locutions of other moods than indicative. And similarly for my statements about other (non-participant) people's mental contents, and my own thoughts and feelings about anyone's commitments or mental contents other than my own. All this bids fair to generate a sizable slab of logic; and it needs to be assumed that every participant in dialogue is capable of carrying out a lot of logical manipulations on the run. For example, from P said quote S unquote and P said quote T unquote and S is incompatible with T we need to be able to deduce P contradicted himself. From P said that Q said that S and P said that Q said that T and S is incompatible with T we need to be able to deduce not T contradicted himself nor P contradicted himself nor even P said that Q contradicted himself but only From what P said it
Modelling Other Minds
119
follows that Q contradicted himself. (And there are saving clauses about whether P or Q retracted anything in the meantime and whether the occasions were the same.) From P said that Q said that S and S is false we deduce not P uttered a falsehood nor Q uttered a falsehood nor P said that Q uttered a falsehood, but at most From what P said, Q uttered a falsehood (and even this could be misleading). We need to be able to form hypotheses about what P thinks from knowing what he says; to verify them inductively; to deduce what next statements of P and Q will be appropriate, given what they have already said; and so on. And all this at a theoretical level. These various locutions and thoughts concerning other people's locutions and thoughts need to be handled in a way that reflects a theoretical knowledge of the way in which the parts between quote-marks might be handled, as zero-order locutions, in commitment and fact (and other mental content) stores. I do not mean to suggest that we are incapable of carrying out these manipulations. It is clear that, in some shape or form, we do. It is also possible that the easiest way to program a machine to do so would be the direct way, by building into it a set of rules for drawing logical deductions among statements containing quote marks, and performing the requisite other manipulations in the case of locutions of other moods. But I do not think this is so. At the very least, as an aid to logical manipulation we should want a mechanism for setting up models of commitment and mental content stores; mind-models of the various people and hypothetical people being referred to. And, once we do this, there are more obvious ways for the program to be written. Let us consider the mechanism by which we might draw a conclusion from P said quote S unquote and P said quote T unquote, on the assumption that S and T are incompatible. Rather than handle the logic of the statements in just this form, we should, if we are machine-programmers, nominate a section of store as the indicative commitment store of the person P referred to. (It does not matter whether P is real or hypothetical.) Now acceptance of a locution of the form P said quote ... unquote is accompanied in our machine, by insertion in this store of the item represented by the row of dots; and hence items corresponding with S and T get inserted. The recognition that this store contains a contradiction does not require the operation of any special inference involving locutions in quotes, because they are in quotes no longer. All that is needed is that the inputs to this store, together with any conclusions drawn from the state of its contents, are appropriately drawn from or attributed to P rather than regarded as inputs to, or conclusions from, the machine of which the store actually forms part.
120
Chapter Eight
Or consider one of the other activities we assimilated to quotation, the activity of play-acting. I do not want to enter the lists in connection with whether actors do, or should, think "objectively" about the characters they are portraying, or whether they should "feel themselves into" their roles and rely on intuition. Let us consider, if you prefer, the audience. Hamlet's thoughts and feelings, as well as his words, become very much a part of the audience's concern as they listen to him, and it is not at all fanciful to suggest that they would generally know much of what is in his mind at any stage of the play. But this does not mean that they even could give you a logical breakdown of this knowledge, in the form "The actor playing Polonius uttered the words quote ... unquote; Shakespeare, having written these, supposed Hamlet to be a human being of type such-and-such; according to him, human beings of this type would react in manner so-andso; the actor playing Hamlet uttered the words so-and-so; the author's intentions ... etc.; therefore probably Shakespeare intended actors of Hamlet to convey that Hamlet thinks that ...". With the utmost difficulty, a spectator of analytical bent might reconstruct such an account. Most of us, however, enter up Hamlet's words in a special Hamlet-commitment-store, draw the conclusions concerning Hamlet's mental contents as we would those of anyone else, and enter them up in like manner. At Hamlet's first entrance, as it were, we set up a set of stores for him; and after that, Hamlet is just someone we have met; or, perhaps, as the psychologists say, introjected. When we produce outputs from these stores or draw conclusions about their contents we put them straightforwardly back in context with Hamlet said quote ... unquote or According to Hamlet, ...48 This means that there is an operative piece of "mind" program that can be switched from one set of stores to another. The putative contents of other people's minds may be stored in the same form in my mind as the contents I acknowledge as my own. I do not want to draw any extravagant
48 I have deliberately ignored an important logical complication, concerning what Quine called "referential opacity" of some encapsulation contexts: Quine, W.V. From a Logical Point of View, Cambridge, Mass., Harvard UP, 1953. In fact this means that different kinds of encapsulations need to be treated differently when logical conclusions are drawn: Hamlet said the man behind the curtain was eavesdropping does not imply Hamlet said Polonius was eavesdropping, even though Polonius was the man behind the curtain and even if, perhaps, Hamlet knew this. The problem of codifying the logic is not, however, insuperable. A similar, rather related, question is that of indexicals, words like I, you, now, here, there and others whose referential-meaning changes with the occasion of their use: the problem with these, too, is simply that of what logical relations we recognise between locutions involving them, uttered on different occasions.
Modelling Other Minds
121
or romantic conclusions from this suggestion, such as that our minds are in any kind of special sympathy. Rather it is primarily a programming matter, a matter at most of how our minds are adapted to social needs. Our language, our minds and our social relations have developed side-by-side throughout our history, and if a particular way of programming a computer to speak is the easiest, it is more than possible that humans have already adapted themselves to this easiest way. There is one other way our common cultural background enters the picture, namely as an invariant store, When I model Hamlet's mind in my own I do not start from scratch with, as it were, blank sheets of paper but award Hamlet a great deal in the way of stored facts, acts and sentiments a priori. I assume, that is, that Hamlet has something of my own background, And, of course, if I make detailed assumptions of this kind about a clearly historical person such as Charles I or Cleopatra, I may quite easily be wrong; some of the things I assume about them may be untrue But this is the peril of assuming anything at all, and it is much smaller than the peril of assuming nothing, which would make it impossible for us to model them. Even the intuitive application of our best knowledge of people whose backgrounds are remote from our own sometimes requires great mental effort. Commonly we start from some kind of standard model and modify it, so that the items in the special store of a particular character represent deviations from a norm and are specified relatively rather than absolutely. And then, perhaps, the norm requires to be changed from time to time, with updating of items throughout. But these are details. Now let me suggest one possible reduction that results in a conceptual simplification of our whole scheme of stores; namely, the treatment of the mental content stores as stores of hypothetical commitments. In our earlier exposition we started by assuming the existence of mental contents and found that the exigencies of practical dialogue forced us to assume also the existence of stores of commitments One of the arguments for these, for example, was that situations sometimes seem to arise such that, if we had to analyse them in terms of mental contents alone, we would be faced with an infinite regress involving the assumption of an infinite number of mental contents. When you and I converse and fully understand one another, there are things that you know, that I know you know, that you know I know you know, and so on indefinitely. But if I had to have in my head not only a model of what you know, but also a model of what you know I know you know, and a model of what you know I know you know I know you know, and so on without end, I should need an infinite mind in order to be able to converse, and would forever be
122
Chapter Eight
carrying out infinity upon infinity of operations within it. The concept of commitment gets rid of the trouble since, provided each of us has a model of his own and the others' commitments, the information that is the object of the infinite chain of knowing can be stored in this form and rightly regarded as distinct from knowledge properly so-called. But there is still something unsatisfactory about this two-level sixfold classification of locutions. It has already been suggested that our ideas about the mental content stores are parasitic on our ideas about commitments; that is, on our grammar. Let us make this a little more explicit. Mental contents are hypothetical commitments. They are what their proprietors would commit themselves to if asked, and if they answered sincerely. This reduction has an air of tautology about it, and I do not want to present it as if it were anything more than a useful theoretical simplification. As such, however, there is something to be said for it. It amounts to saying that the basic unit in our mind-models is a kind of threefold dialectical persona, and that when we try to put into words what the real contents of anyone's mind are, all we succeed in coming up with is a persona appropriate to a hypothetical context in which the person concerned really and truly said or admitted everything he thought and felt. That this context is always, strictly speaking, a fictitious one is obvious; but we need not draw from this the conclusion that it is utterly pointless or unfounded to imagine it. For it may be a limiting case that can be approached in practice, even if never actually reached. The concept of a commitment-persona is, we should remind ourselves, itself an approximation. If, in line with Noel Coward's song, there were only two people in the world, and if, per impossibile, they had learnt English and conducted an exclusive dialogue in it, it might be in order to speak accurately of their respective commitments. But in the world as we know it, as we meet different people, converse with them and take our leave of them it is not possible to say of anyone at any time that he has a fixed list of commitments, but only that he has just taken on such-and-such commitments towards P or retracted such-and-such in the presence of Q. Yet among the various lists of commitments that might be made in respect of a given person from time to time it is often possible to choose some as more clearly characteristic of him as a person than others. This or that commitment was undertaken hotly, not coolly, and should be regarded as impermanent; this or that other was self-interested, this or that one sly, this or that relatively highly consilient with others or otherwise apparently honest. Jekyll-and-Hyde personalities apart, we can pick out in theory a composite commitment-set that forms the pattern of a person's discourse.
Modelling Other Minds
123
But is this really all that his mind consists in? On a logical point, I suppose, one would have to admit that it is in theory possible for someone systematically to misrepresent his beliefs and other mental contents such that we should be completely wrong in our hypotheses about them. The professional spy, for example, might live a lie for more of his life than he lives the truth. In such a case we should have to admit that the pattern of his commitments did not truly represent his mind. But this is not really the point. I am not saying that a person's mind is an average or composite persona of his actual commitments, but only that it can be conceived as a set of hypothetical ones. Thus even of the spy who lives a lie we can say: If he really told you what he thought ...We imagine, rightly or wrongly, that there is some circumstance in which he could be induced to speak fully and frankly. And that is what we mean - according to my simplifying doctrine - when we talk about his thoughts and feelings. If this account can be assumed to work, we can make use of it at least as a programming expedient. My model of another person's mind is as a set of commitment-personae, valid for different dialectical contexts; and including, if I wish, a kind of composite one that I regard as valid for the context of full and frank objective discussion. And for myself? Do I model my own mind as a hypothetical set of my own commitments? I am afraid this question is not one that I know how to answer. Some philosophers, in the tradition of Descartes, are quite sure that we know the contents of our own minds; others, in the tradition of Freud, are equally sure we do not. I have already pointed out that there is something peculiar about locutions in which we refer to the contents of our own minds. The latter, perhaps, is the answer. The question How do I model my own mind? is a Cretan one.
CHAPTER NINE THE PARADOX OF THE MORALISER
Now let us turn to meeting an obligation we incurred in an earlier chapter when we postponed consideration of moral utterances, the locutions of practical and theoretical ethics. These locutions are basic to a whole branch of philosophy and cannot be ignored. Primarily they are the locutions by the use of which we are told that something is good or bad, right or wrong, or evil, or that we ought or should not do something or other, or that something is someone's duty or responsibility, or that he has an obligation. There are also locutions by which these commitments are retracted, and locutions that invite these commitments by asking questions; and there are various kinds of locution of a higher theoretical order that state principles or analyse the lower-order terms and compare them with others. We should also cast our net a little wider by specifying that, in calling these locutions "moral" or "ethical", we do not intend the terms in their narrowest sense but include locutions of the same kind that are based on principles of etiquette, tact, prudence and convention as well as on principles of morality. The philosopher's jargon word "normative" sums up the field better than "moral". Where, on the logical terrain, are these locutions? Can we accommodate them, and the commitments and mental states of those who utter them, within our existing scheme of grammatical categories? This is approximately equivalent to asking: Could a machine equipped to manipulate facts, acts and sentiments and indicative, imperative and emotive commitments play a rational part in discussions of moral attitudes and issues? Or would it need something else such as a moral sense? Familiarity with philosophical writings about ethics, of course, could easily lead us to believe that our existing categories, far from being insufficient, are more than enough and that the question is rather one of choosing between them. Many philosophers, such as G.E. Moore, take the commonsensical view that since the verbs in these locutions are in the indicative mood the locutions are statements of fact. But a variety of others, from Kant to R.M. Hare, have held that some or all of them are really imperatives; and we have ourselves treated some should-statements
The Paradox Of The Moraliser
125
as imperatives in our earlier discussions. And then, A.J. Ayer and perhaps C.L. Stevenson once held that they were emotives. The dispute over their correct categorisation was not due so much to ignorance of their nature as to lack of a clear criterion for distinguishing the categories. Now that we have such criteria we should be able to make short work of this muchdebated question. There is one important preliminary distinction to be made, the distinction between should and good. Most of the terms in our list are like should in that they have something to do with acts or actions of human beings or, at worst, of personified near-human things or animals; but good, besides being applied to actions, can also be applied to people and things themselves and, when it is, could be better described as evaluative than as normative. The other terms all, broadly, translate into should: right and wrong actions are things that should or should not be done, duties and responsibilities should be honoured. It is true that Kant thought there was something debatable about the principle We ought to do our duty, but even he came to the conclusion that it was true; and most people would regard it as a simple matter of meaning. But to call an action good does not imply that it should be or should have been performed, and to call a person good does not imply anything at all about how he, or we, should behave. Even to call an action bad does not necessarily imply that it should not be done, since it may be the lesser of evils. There is, in short, a range of meanings of good, bad and evil in which they express, if not emotions, at least attitudes of the speaker or hypothetical speaker, and are to be classed with emotives. Whether this is the only range of meanings of these words, or whether they have others in which they are intertranslatable with locutions involving should, I shall not discuss. The other terms in the list are of greater importance to us, and in what follows I shall confine attention to them. Our first question must be why anyone should ever doubt that moral locutions are indicatives, In dialogue they seem to be debatable in exactly the same way as indicatives are, to sustain all the same commitments, to be simply negatable, to have consequences and contraries, and generally to be stated, withdrawn, questioned, supported, inferred and induced in just the same way as locutions more properly called "statements of fact". When I say You ought to give Deirdre a nice birthday present this year, or Faculty was wrong to rescind the motion on inter-university options, or Australia has a duty towards New Guinea, there does not seem to be any good reason to deny that these locutions become entered in my indicative commitment store and, if you agree with them, in yours. Above all, unlike imperatives and emotives, they can be true or false. In fact, when we look
126
Chapter Nine
at the reasons that people like Kant, Hare, Ayer and Stevenson have had for classifying them differently, we find that they are irrelevant to any of the considerations we have advanced so far, deriving from concern over the fact that there seems to be nothing observable in the world corresponding with the facts of ethics as there is corresponding with the facts of, say, the various natural sciences. Since moral statements, the argument runs, are not based on our sense-experience of the world around us, and are not matters of pure logic or reasoning like the statements of mathematics, they cannot really be statements at all; and, if this is so, they must either be meaningless or must be locutions of another kind. In a dialectical context we should find it very easy to deny the implicit premisses of this argument. Statements, we should say, are locutions that are handled in a particular way in dialogue and require a characteristic kind of commitment store, and there is no case for excluding from statementhood any locution that is not distinct from other statements in the way it is manipulated. And we have not yet found the need to bring observation of the world around us into the dialectical picture at all. (Though we shall do so in the next chapter.) Some dissatisfaction with this conclusion might be sparked by the remark that normative words all refer, like imperatives, to acts or actions; or, again like imperatives, to end-states that can only be attained as the result of acts or actions, as in You should be here at 9 in the mornings, or It is his duty to be ready for emergencies. The closeness of moral terminology to that of commands is also cause for speculation; for the armed services, for example, very often phrase orders in forms such as It is the responsibility of personnel to keep barracks tidy and Dress uniform obligatory. It has often been thought that moral obligations are a matter of God's commands. Leaving aside the difficult question of how commands may be supposed to be transmitted to us from heaven, it is worthwhile taking the divine theory seriously at least as a grammatical one; for, if we suppose ourselves to receive and ever to accept God's commands we must put them in our imperative commitment stores rather than our indicative. It is true that commands cannot, like indicatives, be true or false, but we might replace the concept of truth of a moral statement with that of validity or authority of a command; and, in any case, the primary representation of the command would be in an internalised model of God's imperative commitment store. Commands, as we saw, can be negated, withdrawn, questioned, and have consequences and contraries just as indicatives can; so that, although some difficulties might arise in the working out of the details, the theory that moral locutions are commands, whether of God or of some more
The Paradox Of The Moraliser
127
earthly dictator, is not to be thrown out on that score. But there are two properties moral locutions have that commands do not, namely, of being non-wilful and, relatively, accountable. When we earlier (in chapter 4) drew a distinction between two classes of imperatives - on the one hand, commands, demands and requests and, on the other, advice, recipes, instructions and suggestions - we noticed in the first place that although commands, demands and requests, within the limits of the social authority of the speaker, may express the speaker's own wishes concerning what he wants to be done, imperatives of the second category may not; or, if they do, this is incidental to their role. And in the second place, although when an addressee questions a command, or request, or demand the person who issued it may consent to underpin his locution with a rationale (as distinct from a causal explanation), saying why it should be carried out or granted, it is not in the nature of the locution that he should be obliged to do so; whereas, in the case of the "advice"-group, the existence of such a rationale is an essential part of the meaning and readiness to give it is normally understood. Now we must certainly classify normative locutions, if they are imperatives, with the second group; for if I tell you you should be kinder to Doris, or that you have a responsibility to take an interest in local politics, these will hardly be moral locutions - in even the broadest sense - if they can be interpreted as meaning just that these things are my wish. And, if you ask me Why? I shall answer Because she depends on you or There are so many opportunities to improve things and be conscious that, if I had no such ready answer, the status of my locution would be in question. When children ask their elders Why? these elders, it is true, sometimes answer You should, that's all!, but this, though it usually means that the said elders have no reason and are bluffing, is at least in principle supposed to indicate that there does exist a rationale and that the young should take its existence on trust. Consequently it begins to seem that normative locutions can be put in the same category as advice, recipes, instructions and suggestions, namely, as imperatives of the second class. And in fact, although at first sight this suggestion seems to raise difficulties in the interpretation of indicative features of statements concerning right and wrong, duty, obligation and the like, these vanish on a little consideration. Thus indicative negation, for example, is quickly seen here to have the double internal-external feature that is characteristic of imperatives, since there is a distinction between not having a duty to do so-and-so and having a duty not to do it. And patterns of justification, broadly, will be the same as for imperatives. Point by point we can go through the complications inherent in the logic of imperatives
128
Chapter Nine
and check that they are all relevant to normatives. There remains one possible ground of objection to this identification, namely, concerning the ultimate premisses of accountability. Let us suppose that, having questioned my statement It is necessary to be ruthless with ignorant people and obtained an answer say Life is too short, and we must get things done - you go on, childlike or Socrates-like, to continue asking Why? and force me back to ever more primitive moral attitudes: where can we stop? In the case of indicatives it was reasonable to assume that the chain of reasons would come to an end when we reached some kind of substratum of agreed fact; or, if basic facts were subject to disagreement, that admission of simple ignorance would, if not resolve the issue, at least terminate the discussion. In the case of adviceimperatives something similar applies since the goals to be aimed at are largely a matter for the addressee, and the advice-giver is largely concerned with the factual question of what actions will achieve what ends. But in the case of moral locutions it seems that there can be quite fundamental and primitive disagreements - between those, for example, who value the well-being of others and those who regard only themselves or their own small group, those who believe in ambition and those who alternatively value personal modesty; those who tolerate righteous anger and those who regard it as immoderate or a sin, moral theists and atheists, pleasure-approvers and Puritans and so forth. And so soon as a moral argument or reason-chain gets down to fundamental differences of these kinds, it is commonly found that there is very little more that can be said. But it is also found that this is not the kind of situation in which one may shrug off the differences as due to ignorance, since it is precisely in this region that the peculiar nature of moral locutions is clearest and most important. This is not because the same dialectical moves are not available as for indicatives or for advice-imperatives. In particular, it is common for a hard-pressed rationaliser to retreat to I think or I believe: thus Well, I think you should be kind to people when they depend on you, or I believe in doing something with my life, and if you don't I don't know what to say to you. But in the nature of the case this move is somehow less effective than in other cases. When two participants who differ over whether S is or is not the case retreat respectively to I think S the case and I think S is not the case, they have disengaged: they are no longer under any pressure to resolve their disagreement, since their respective statements are not actually in contradiction. But when two people disagree about whether one should, or should not, perform some act A, the retreat to I think one should do and I think one should not do A though apparently similar, is less likely
The Paradox Of The Moraliser
129
to lead to an agreement to differ and less likely to be interpreted as any sort of agreement at all. And it now seems clear that it is this feature that tends to mark moral locutions off from (other) imperatives of the second category. When, as sometimes happens, two people readily and easily agree to differ in respect of what is apparently a moral matter, it loses its moral character and becomes an advice-imperative proper. We shall return to the question of agreement and difference shortly. Let us, however, turn for a moment to the theory that moral locutions are emotives. The older statements of this theory suffer from the fact that they tended to be based on the assumption that emotives were the only alternative to indicative or "descriptive" statements. Now it is all very well to say that moral locutions are emotive if one uses the word emotive so widely that it includes everything that is not a matter of pure description; but it is not at all the same thing to make this claim if one means thereby that they cannot enjoin any kind of action in the way imperatives do. The claim, in fact, only becomes at all plausible if consideration is confined (as in many older books it used to be) to locutions involving the words good and bad rather than any of the others; and we have already noticed that these are different. Even in the case of the superficially similar right and wrong it can hardly be denied that action is involved. To take an older example of A.J. Ayer, Stealing money is wrong can hardly be taken (as he says) just to mean Stealing money!!, with the exclamation marks indicating "a peculiar tone of horror" or "a special form of moral disapproval".49 To steal money is to perform a certain kind of act; and the person who morally disapproves of stealing money, though he is not necessarily actually ordering, requesting or advising anyone not to perform this kind of act, nevertheless indicates opposition to it. It should be added that no peculiar tone of horror is really necessary to indicate moral disapproval, which may be as unemotional as anyone pleases. Moral judgments are just as meaningful if said flatly and unaccompanied by heat or any other special feeling-tone. And sometimes the feeling attached to a moral injunction may be the opposite of what would be required under Ayer's stipulation, as in It would be great fun, but you shouldn't do it or It's horrible, but you must. Moral locutions, we might say, are not a matter of passion, but of action; and this is easy to see as soon as two mood-candidates are allowed beside indicatives, not just one. To call moral locutions emotives is to degrade them; for, generally
49 Ayer, A.J., Language, Truth and Logic, London, Gollancz, 2nd, edn. 1946, p. 107.
130
Chapter Nine
speaking, emotion has had a bad press among academics. I do not mean that this is an attitude I would support, but there is no doubt that this has been the general effect of the emotivists' doctrine, and perhaps their intention. But it is also probably true that they have been driven to it by the excesses of the indicativists. So long as morality may be considered to be debated by moralists without fear of their ever being involved in the tricky business of actually acting on their principles, it becomes academic indeed; and our minds get stuffed with non-operative codes of behaviour to which one is forever reacting with I agree I should do that, considered in the abstract, but ... or arguing that one has a duty to oneself that permits regular moral holidays. This kind of indicative moralising, if it were not so nearly established usage, would lead all reasonable men to rebel. Unfortunately, the easiest line of rebellion is into intuitive, unrationalised ethics of the kind that rejects accountability; and this might just as well be an emotive matter or, at least, a matter (since we need to broaden the word "emotive") of attitudinising The failure of indicative moralists to rationalise our behaviour for us leads to the doctrine that morality is all a matter of passing empty judgment anyway (and aesthetics the same), and we may do as we feel. This is not a position that can even be stated without Cretan overtones. Consequently, we are left with the conclusion that normative locutions are a species of imperative - of the second kind, resembling advice. This kind of identification, of course, can cause trouble for us in philosophy unless we are careful to avoid drawing unwarranted conclusions from it. In short, I am not saying that moral locutions are advice, but only that they should be classified alongside it. And this decision still does very little for us. In our discussion of theories of truth we noticed that there are two jobs to be done by such a theory - first, to say how the word true is used, and second, to give criteria for actually discriminating between true statements and false ones. Dialectical considerations were of great assistance to us with the first of these tasks, but left us with a feeling that we had not got very far with the second. In the same way, the discovery that normative locutions are like advice, though it helps to identify their function in language, does nothing at all to give us any actual moral guidance. This is not unusual in modern ethical writing. Older moralists were interested in putting up and discussing actual principles or codes for the regulation of our lives, but twentieth century moralists - and many would consider it the scandal of modern philosophy - seem more concerned to explore the meanings and logical relations of moral terms than they are with actually making use of these terms to recommend, advise or evaluate conduct. Unfortunately, dialectic, as a study, cannot do much to remedy
The Paradox Of The Moraliser
131
this state of affairs, and I consequently make no claim to try to do so in this book. But there is one further piece of analysis that it does encourage us to undertake. Moral principles, if they are to be set up at all, must not have their foundations in sand. And it so happens that there is a serious paradox implicit in all the older philosophical moralities. Moral imperatives may be low-level or high-level, from You should give the dog a nice pat to such things as the Principle of the Greatest Happiness of the Greatest Number. But there is a sense in which moral philosophers, at least of the old-fashioned type, should be concerned only with the latter kind. In fact, if moralists had their way, low-level moral locutions would be all mere applications of high-level ones to particular circumstances. I should be able to decide whether to pat the dog - or, at least, whether I ought to pat the dog - by asking myself whether the circumstances are such that this action accords with the highest-level principles to which I subscribe, or which have been agreed to by moral philosophers. (It will please him, of course, and please me; but, on the other hand, it is dinner-time and I shall need to wash my hands again before handing food around, ...) The highest-level moral locutions can be used to derive lower-level ones by specialising them with the insertion of details concerning time, persons and circumstances. It follows that the highest-level moral locutions have a certain kind of universality. It would generally be held, in fact, that they should contain no mention of special times, places, persons or circumstances, since the presence of such mention would only indicate that they were not yet of the appropriate level of generality. For example, the principle that one should be kind to animals on Mondays could not be a highest level moral principle, since Mondays is too specific a term; and it could only be a moral principle at all if it were derivable from some principle in which no special reference to Mondays - or, indeed, animals - is made. Such principles are hedonism, the principle that pleasure is the greatest good; utilitarianism, the principle of the greatest happiness of the greatest number; stoicism, the principle of nobility and resignation; conventionalism, the principle of conformity; religious ethics, the principle of obedience to revealed laws of supernatural origin; and others of this kind. A very few people have occasionally thought that there were special principles or obligations applicable to themselves, or their class, or some other special class of people or actions. Men have sometimes been thought to have a special duty to women, or vice versa, as it were independently of their possibly different natures and needs. The white races have sometimes been thought to have a special duty to the black races, or again (more probably) vice versa. But it would almost always be argued in such cases
132
Chapter Nine
that the difference of moral obligation was a consequence of differences among the people concerned, and not to be enshrined in any absolute principle of its own. Family mottoes are, perhaps, a special case: the motto Suffer of a certain English aristocratic family sounds like a moral obligation special to members of the family, and the accident of birth is hardly sufficient differentia for this to have backing from a generally operative moral principle. But I suppose even here it would be possible to argue that the principle is laid down - like stoicism - as one that is supposed ideally to be such as all good men should adopt. One can imagine that a Cartesian, who conceives his own mind as existing on a different plane from others or from the world around him, might draw a fundamental distinction between principles that apply to himself and those that apply to other people. However, if he does so, he almost certainly cannot talk to other people about it, since the difference he experiences between himself and other people is not one that they are capable of appreciating except, as it were, by analogical extension of their own Cartesian experiences of themselves. The accountability of moral locutions is an objective, interpersonal affair and does not tolerate this kind of distinction between persons. This being so, moral principles are supposed to be universal. That is, they apply to all actions at all times. And, this being so, they apply to the actions of the people who put them forward and debate them; including the actual actions of putting forward and debating. Now it might be held that moral locutions, though universal in respect of time and place, are not quite universal in the sense of regulating every choice made by their adherents. That is, it might be said that moral questions - like other imperative questions, as we noticed in introducing them above - are sometimes to be answered not by selecting a precise one of the alternatives offered, but only by selecting a sub-class and saying Do whichever you like of these. And it might be held, therefore, that the actual process of putting forward and debating moral principles might itself escape the operations of morality. Debate, we might think, in itself does nobody either good or harm, and can be carried on in a morally neutral atmosphere. It is only when I have actually selected my moral principles and come to put them into practice that the question of the rightness or wrongness of my behaviour can begin to arise. But the more we begin to reflect on this assumption the more preposterous it begins to appear. Moral locutions are not only to be judged as right or wrong in respect of what they say; they are themselves moral actions, and the saying of them is something that must be regarded as capable of being right or wrong, depending on circumstance. Usually, of course, one should utter correct moral locutions - unless, that is, the
The Paradox Of The Moraliser
133
circumstances are especially inappropriate - and should not utter incorrect ones. And this is not just a tautology. On the whole, telling people to do what is (actually) right helps to get them to do it, and thus helps to get the right thing done. Above all, in discussions of moral philosophy, a part of the object of the exercise is to secure the adoption of whatever principles can be agreed to. A moral philosopher who said that his only aim was to learn the truth about morals, and that he had no interest whatsoever in getting anybody actually to let the discussion influence his behaviour or life, would be regarded as having abdicated his task even more completely than have the logical analysts of the twentieth century. The link between philosophical discussion of moral principles and the content of the principles themselves may be tenuous, but it is not nonexistent. Philosophers do influence generations of colleagues, students and readers. They are not the exclusive influences on them: they must compete with novelists, scientists, psychologists, world wars, natural disasters, climate, topography and clergymen. But they are part of a process, and a rather especially intimate part since they alone concern themselves explicitly with the general principles of morals. This concern would be pointless if they were to claim to stand aside from the process as a whole. Moral discussion, then, is itself especially subject to moral approval and condemnation. This means that moral locutions are relevant to their own rightness or wrongness. We can now develop our paradox. The connection between correctness of a moral principle and the rightness of the act of stating it, though it doubtless exists, is not a universal one. It is, moreover, contingent. There is no necessary moral principle to the effect that we should tell the truth about morals. There are, rather, occasions on which we should not do so, or should utter falsehood instead. I say "tell the truth" and "utter falsehoods" because this is the usual way of speaking about moral principles in the grammatical indicative mood. There is no inconsistency here with the thesis that they are really imperatives. This is implied by the usual assumptions as to the nature of moral principles and discussion. Let us consider a moral dialogue between persons P and Q, which starts with P's saying You should do A, for some action A. Since moral locutions are accountable, Q may (if it is a "future tense" imperative and there is plenty of time for debate) ask Why?, and expect an answer that provides A with a justification. But, as we saw, the locution Why? is ambiguous and can also represent a demand for a justification of a different kind, namely, a justification for having said You should do A. And one is tempted to say that this is not a true case of ambiguity but that the two demands overlap. It may be sufficient justification for having said You
134
Chapter Nine
should do A that the action A itself rather than the act of saying You should do A - has a justification. One tends invalidly to treat We should give correct advice as a kind of tautology. From an alternative point of view let us consider two different kinds of consistency a speaker's locutions may exhibit (or fail to exhibit). We shall regard P as consistent in one sense if, when asked for a justification of You should do A, he provides a set of premisses from which it follows that A is the right thing to do, and, in subsequent discussion unfailingly upholds A and actions of a similar kind subsumable under the principles he adduced in A's support, and condemns actions inconsistent with these. In this sense a speaker is consistent if he maintains a consistent set of commitments in his store; whence we may call it commitment consistency. A speaker can maintain commitment consistency by saying the right things, for which he needs only a good memory and a head for logical calculation. Many people of many different moral persuasions manage to be more or less commitment-consistent. The other kind, which I shall call execution consistency, is rarer. I mean by this the conformity of a person's actions and behaviour to the moral imperatives he utters. Thus it will often be held against someone's moral locutions not that they are inconsistent on their own account but that they do not represent his true morality as revealed in his actions. Either, through carelessness or weakness of the will, he does not live up to his own principles or, for some other reason, he misstates his principles and implicitly conforms with different ones. In either case what he does is different from what he says he ought to do; and I am not sure that we can always distinguish between the two supposed reasons why. Commitment consistency is a precondition of execution consistency, since one cannot consistently execute an inconsistent set of commitments. (Or, perhaps, one cannot fail to execute it: how we describe the case depends on whether the commitments are thought of as obligatory or permissive.) But failure of execution consistency is often urged, in any case, against moral locutions, in a form of argument that is sometimes regarded as a species of argumentum ad hominem. Someone who preaches the virtues of poverty will not be taken seriously if it is pointed out that he has a large private income; and someone who claims to believe in fresh air and exercise will be regarded as inconsistent if, without other reason, he refuses to go for a walk up a hill. Even someone who regularly finds excuses for his apparent inconsistencies of execution will often be regarded as inconsistent on the grounds that the excuses are always ad hoc and that broad patterns of his behaviour and professed principle diverge. What a speaker says, however, is a part of his behaviour just as much as are his other actions. And it now seems that moral locutions have a
The Paradox Of The Moraliser
135
double standard to meet. Not only must they be commitment-consistent one with another, but the fact that they are put forward must be executionconsistent with the kind of behaviour they counsel. And this double standard leads to some very curious consequences. Utilitarianism, for example, is the doctrine that our aim should always be to serve the general happiness. But it is at least possible in principle that we should find that utilitarians are themselves the most unhappy of people. It would follow from this that for commitment-consistency they should hold that no one should ever try to spread utilitarian doctrine. Believing it themselves, they should suppress it, and attempt to spread some more pleasant or less demanding doctrine, such as pure hedonism. Perhaps they should even attempt to believe this alternative doctrine themselves. Such a situation of enforced hypocrisy and muzzling of discussion could hardly please the self-respecting moral philosopher. Alternatively, of course, it might be found that utilitarians are all wildly happy people, whence every utilitarian can reason that he should try to spread his doctrine as widely as possible. But even this happily self-reinforcing situation is not one with which a moral philosopher could long remain content, since he would feel that the proponents of the doctrine were spreading it for the wrong reasons, on account of its social effect rather than its truth. (And this fact might regularly intrude itself into dialogues, since a lack of respect for truth might regularly lead to an impatience, for example, with precise accountability.) Or again, and more plausibly, it would be found that there are some occasions on which utilitarianism is, by utilitarian principles, a good doctrine to push, and others on which it is not. The only occasions on which our convinced utilitarian would be able to indulge in an objective truth-oriented discussion of his doctrine would be those on which the outcome of the discussion could not possibly have any effect, happy or otherwise, on subsequent events. This would, no doubt, make him very unhappy indeed. Or again, let us suppose we meet a moral conventionalist, someone who holds that moral worth attaches only to conformity with the expressed conventions of our society. He may be able to hold this view with complete commitment-consistency. But it seems to follow from his view that he should not hold this view, at least in public, unless it is a conventional view to hold. And this is a conclusion that he might, in fact, find quite repugnant if, for example, his community generally subscribes to some different belief such as that actions should be guided by the will of a prophet as expressed in sacred books. Our atheistic conventionalist can be true to his conventionalism only by always claiming that the will of the prophet transcends that of the community. It is in principle conceivable
136
Chapter Nine
that this "noble lie" is one that he should not reveal as such even to intimate philosophically-minded friends, and even, perhaps, one that he should endeavour to expunge from his own mind. The only difficulty about the latter operation is that, if it were to succeed, it would destroy its own rationale. I do not think the dilemma these considerations create is in any way an academic one of no practical application. On the contrary, it seems to me that some moral philosophers, when they come to debate their doctrines, quite frequently experience a tug-of-war between two standards, a truthstandard and a moral one. Although I have never personally been much torn by this dilemma it is possible that it is because I am not someone who has ever taken very seriously the enterprise of moralising; that is, of setting up a consistent and all-pervasive set of moral principles. And it may be the case that my position is the usual one and that most of us do not take moralising very seriously. However that may be, I think it is clear that the dilemma awaits everyone who tries to do so. I should immediately add that a lack of interest in moralising does not necessarily at all imply lack of moral principles. Many of the most moral people I know - in the sense of being keen workers in this or that morally desirable cause - are disinterested in, and even contemptuous of, moral philosophy. A morality is something we can study in others; and what we study is a pattern of behaviour, including or excluding verbal behaviour in the nature of statements of personal principle or policy or of self-analysis. But the pattern of behaviour shows us a morality, and the statements contribute. And when a morality is conceived in this way, it begins to appear that what the statements say is irrelevant except for what this shows; that is, except for what is shown by the fact that they are made. Let V-ism be some moral viewpoint. A doctrinal V-ist is one who on appropriate occasions claims that we should do what is V: a behavioural V-ist is one whose acts are predominantly V. The latter is clearly the more morally important characterisation. What we have shown is that doctrinal V-ism, far from being an appropriate accompaniment of behavioural V-ism, is sometimes even inconsistent with it. Isn't it possible that it should be a bad thing for mankind ever to know the truth about morality? But then, this observation would itself be dangerous since it might give a clue which would lead others to the bad truth. I should not say it, I should put it out of my mind! ... And where, now, is the academic study known as Ethics? It would be incumbent on the moral philosopher to produce, to his classes and his colleagues, a "proof" of the good falsehood. Voltaire, remember, said that if there were no God it
The Paradox Of The Moraliser
137
would be necessary to invent one. How many have invented one or another God, not cynically but, almost, piously? And admitted the ulterior motive, if at all, to some small hand-picked band of brainwashed proselytes, whose function was to preserve the lie for posterity. These sceptical insinuations conceal, moreover, a Cretan difficulty at the base of moral discussion. There are times when I do not at all suspect the motives of moralists; but, when I do, it does not necessarily help matters to try to bring the suspicions into the open. To ask Is that what you really think, or are your motives ulterior ones? will be effective only against a moralist who is not quite convinced of the merits of commitment consistency. You should do A, and I say that from ulterior motives is a Cretan statement, and the only way any upholder of moral principles can preserve commitment consistency is by professing that it accords with that principle to uphold it; though this is Cretan too. It follows that, if we are to avoid Cretan disputes, we cannot ever question a moralist's motives in debate with him, but only as onlookers. And this in turn implies that moral argument is necessarily always carried on in an arena in which even someone who does not believe himself capable of rationalising his explicit moral prejudices will be forced to go through the moves of doing so. And the stronger his moral feelings, the more - since moral argument may be instrumental in securing his ends - he will be required to rationalise them. The result may sometimes be more like a battle than a discussion, but there is absolutely nothing that can be done about it by the logician. That is why "agreement to differ" has such a different aspect in moral arguments from the one it has in less sensitive areas. When P says I think A should be done and Q says I don't, they cannot be regarded as merely having made a mutual revelation of interesting mental contents, but more likely as having reached a point of fundamental difference. In some circumstances the logical next stage would be for them to start throwing the furniture at one another.
CHAPTER TEN PERCEPTION
Does a book on language need to take account of perception? We have already touched on the subject of perception at several points: first, when we noticed that the mental state verbs see and hear resemble emotives in sometimes taking simple noun-objects; second, when we considered truth; third, when we speculated whether elementary moral locutions could be regarded as having a perceptual basis. This chapter should need no more excuse: it is keeping a promise. But we should also notice that one of the reasons some people have for refusing to take seriously the idea of mechanical dialogue - conversing with a machine - is that they feel that a machine, dependent on its programmer and on verbal inputs for all its information, cannot make an independent contribution, except in empty tautologies, to any conversation in which value is attached to truth. This argument immediately falls if we equip the machine with sense-organs. On the other hand it is not at all clear how one is to go about connecting a machine's sense-organs with its linguistic apparatus. For that matter it is not clear what constitutes a "sense-organ". I shall argue in this chapter that it is the linguistic end of the link that is the important one here, in the sense that it determines what is to count as a sense-organ, and what is to count as a perceptual truth. Computer technologists have sometimes imagined that if they equip a machine with an eye or, say, a television camera they have immediately put it in touch with empirical truth. Given an ability to analyse pictures, an equipped computer should be able on demand to make reliable statements such as I see a laboratory bench. But this assumes that it has been programmed to connect this form of words with such-and-such a visual shape; and it could just as easily be programmed to say, when shown a laboratory bench, I am witnessing the defeat of the English at the Battle of Hastings, and Harold has just been hit by an arrow. That is, it is the programmer, not the computer, who does the interpreting. It is also, in a sense, the programmer, rather than the computer, who does the seeing; for although the stimulus may be a laboratory bench it does not follow that it
Perception
139
looks like a laboratory bench to the computer. If the computer really could see, it would be possible for it to have a disturbance of vision so that the laboratory bench always looked like the Battle of Hastings to it. And then it would, in one sense, be telling the truth when it reported I see the Battle of Hastings. As it is, the computer is merely giving a programmed response to light-stimuli. There is more to seeing than this. The provision of optical apparatus, we might say, though it may be a necessary condition of seeing, does not in itself make possible the gathering of visual information. Similarly, ears are only necessary, not sufficient for the gaining of information by ear. And so on for the other senses, and even then the necessity is only a physical one. In some sense sense-organs are not necessary at all to perception. At best they are external to the process, coding devices that translate incoming stimuli from one medium to another. I could get all my sensory information electrically, through a headplug; or, if the laws of physics were different, by clairvoyance. What goes on once signals get into the brain is a different matter. But now I want to ask a very difficult and puzzling logical question: Do we perceive facts? The verb see (and the other verbs) can take as object a plain noun-phrase such as, a book, the sky, The Sydney Harbour Bridge, Chairman Mao; or it can take a that-clause such as that the cat has finished her dinner or that the Normans are winning. It can also take act or event specifications as when I see John blow his nose or see the running of the Derby; and it can take property-descriptions such as the smoothness of the sea. But which of these is primary? What do we really see, and which idioms are just ways of speaking? There is enough material here to keep a transformational grammarian busy for half a book. Information, we should remind ourselves, is expressed in indicative sentences; so that if we are to be able to say that we get information through our senses it must end up as facts. But it is not quite satisfactory to say that we perceive facts directly. It would, at least, be much more normal to say that we perceive things, or their properties, or at most events. And there is a difference between perceiving a thing and perceiving that it is there, or between perceiving an event and perceiving that it takes place. Or between perceiving some property of a thing and perceiving that it has that property. A computer, of course, cannot do either; but there are occasions on which we want to say that we have the first or thing-event-property sort of perception without having the correlative second sort. The first sort seems more natural in the case of lower organisms. And we can perceive an imaginary thing or event or property, but not - however deceived we may be - an imaginary fact. The truth is that facts are expressible in words, whereas perception
140
Chapter Ten
does not alone or automatically give us anything expressible in words. Perception gives us words only when we see them written or hear them spoken; and even then we see or hear them as things, and do not necessarily or automatically perceive either any facts about them or any facts expressed by them. (Anyway, words may express other things than facts.) That the process of sensory interpretation must intervene even between words and what they express can be underlined by considering another fancy of the science-fiction writer, package insertion of information. The heads of future men, we may suppose, are equipped with directinput channels such that information can be fed in at high speed or inserted in a package. Input is accompanied by no relevant sensations (unlike the case of headplug sensory input mentioned earlier), but afterwards one is able to answer question on the subject of the inserted information. And Jones, let us say, one day has a package on 20th century history inserted. Now there is no doubt that something resembling this is, in principle, possible. However, the task of accurately describing the situation presents us with serious philosophical problems. There is serious doubt whether we should say that Jones knows anything he did not know before the insertion operation. Precisely what we do say depends on details. One possibility is that the inserted package is like a kind of book that he may refer to and read off at will; so that when we ask him What happened to Hitler? and he answers He died in his bunker, we say that he has "looked it up". In this case we shall not want to say that he knows his packaged history any more than (necessarily) does someone who has a real history book on his shelf; though he may, in the course of time, come to learn it by his reading. It would be a necessary condition of this case that he should be able to say reliably whether a piece of information is in his package or not, perhaps giving location, precise wording and so on. Also that he should be capable of wilfully answering wrongly, that he should progressively integrate the inserted information with that already in his possession, and other things of this kind. Another possibility - perhaps the only other - is that we should say that the insertion operation is a conditioning process, and that his subsequent answers to questions are mere reflex actions. The circumstances in which we would say this are obvious enough. But someone who is conditioned to answer in a certain way does not "know" the answers any more than a parrot does; or even perhaps have "information". Consequently, it seems that even verbal information cannot bypass the senses - or rather, cannot bypass the sensory digestion and integration elements of the mind - without losing its character and becoming a mere conditioning stimulus. But what these elements are, of which we are
Perception
141
beginning to feel the presence, we are yet far from specifying. Let us look at the problem, for a moment, from the other end and ask what features in our use of language cause us to look beyond language to perception. The first thing to notice is that language has no sharp limits. When someone asks me Where is the so-and-so? and I point, I assume not only that he will be able to see and understand my gesture and give it its conventional significance but also that he will be able to use his senses to locate the so-and-so. If I ask someone Hand me the newspaper I expect him to treat the the as a weak demonstrative and, even if the newspaper has not previously figured in our conversation or been brought to his attention, locate it by conducting a suitable sensory search. We often enthymematically answer unasked questions, carry out ungiven advice, even withdraw unincurred commitments or take up unissued challenges or points of order. To the extent that a theory of dialogue is to stay on the rails at all, we must recognise - without trying to draw a hard line - that though some of these phenomena involve virtual locutions we cannot incorporate all of them in a neat theory. The concept of a virtual locution, as it happens, can be tidied up quite a lot, since a gesturer can often be required to reframe his locution in real words. Moreover, human beings converse without gesture, or much else in the way of shared perception, when they do so by telephone or letter. It follows that this kind of use of perception is not of central importance in a developed language. It may be important in the development and learning of such a language, but that is another question. Then where do we really need perception? The answer is, of course, that some of our statements are reports of our perceptions, and that, in some sense perceptions underpin these reports. But again we are conscious of the gap between fact and language. When all goes well and language flows smoothly, the presumption that perceived fact underpins our perceptual statements remains untested. But when something goes amiss and perceptual facts are disputed, what happens? Statements of fact are not accounted for by perceptions, but by other statements of fact; until, perhaps, I say Look harder, it's to the left of the mast on the headland and you say I don't believe you; you must be seeing something else. And now I cannot produce my perception to persuade you to accept my statement. We cannot exchange perceptions, but only statements reporting what we see. Or rather, what we seem to see; for the question of whether we really see what we seem to see is not one that can be in turn decided by perception but must be discussed and resolved, if at all, in language. So let us turn to tracing the typical course of a dispute over a perceived
142
Chapter Ten
fact. Don't you like that pink hibiscus! says P, and Q replies You mean the white one over there? P says No, it's pink; it's Apple Blossom and Q says No, it must be that other variety the white one. Attempts to dogmatise having failed, the next move is probably to try a verbal tack: I'd call that pink, says P, and if Q replies I call it white the perceptual dispute is over, to be replaced by one about English usage. But let us suppose Q says instead I don't care what you call it, it's white. The next move is to question the lighting or the aspect: Look, you have the sun in your eyes there, or The pink is really the stamen; it has white petals. But if this fails, there is nothing for it but There's something wrong with your eyes. Have you had them tested recently? or Looks pink to you, does it? Ah, well ...! At this stage of the dispute participants may be preparing for a characteristic shift of ground. Although Q still says It's white, P has started to encourage him to alter his claim to It looks white to me, which is very nearly the same claim but lacks hearer-commitment. And Q may be doing the same in reverse. Subsequently either, or both, may make this shift: if both do, as in It looks pink to me. Oh, it looks white to me, we may regard an armistice as having been signed, with mutual withdrawal of troops. But another possibility is one-sided withdrawal. P may end up saying It looks pink to me and Q saying But it really is white; and P may even have to back down completely and say I realise now there is something funny with my eyes. The role of the locution It looks pink to me can be compared with that of I believe that so-and-so. We noticed in an earlier chapter that when a statement S is challenged the speaker will often retreat to Well, I believe that S, which is a statement that the addressee can accept without being committed to S. The apparent firmness and resolution of a credo is really a front for a dialectical weakness. It now seems that the same is true of a report of a perception. I believe that, it will be remembered, seemed at first sight to be a report of an inner state, but turned out on examination to be quite misleading. We found reason to say that it is impossible, anyway, for a speaker in any straightforward sense to give a report of the contents of his fact store, since the intimate association of the fact store with ordinary fact-stating confuses the truth-falsehood issue for any such report. It is now clear that something very similar is true of reports of perceptions. The two cases are not exactly parallel, since It looks X to me but I know it isn't is possible, whereas I believe that S but I know it's not so is nonsense. In fact, instead of a parallel between two idioms we have an actual relation, since perception may underpin belief. But the dialectical weakness of an appeal to perception is doubly apparent. We frequently, no
Perception
143
doubt, make strong use of facts which could be the subject of appeal to perception, and expect hearers to agree to them without such an appeal; but we do so in plain indicative mood, as in It's pink, or in Don't you like the pink one? But when it becomes necessary to use the actual language of perception, as in I see that it is pink or It looks pink to me not all is well with our perceptions or with the information supposedly issuing from them. I see that ... and It looks to me that ... do not report inner states so much as make dialectically weak claims concerning plain outer ones. Moreover, it is clear that perception itself cannot act as a guarantee of truth of any such claim. Since it has often been thought that some kind of special meaning attaches to the words true and truth when they are used in connection with sensory observation, I may be pardoned for reopening the question of truth. To say of a statement S that it is true is, we said, the same as to make statement S itself. But ... is true, attached to a perceptually verifiable locution such as There is a teacup on my desk, has the same function here as in any other case; namely, no function of any importance. It might be difficult to build a machine capable of verifying the statement There is a teacup on my desk, but if this could be done it would be no harder to build one capable of verifying It is true that there is a teacup on my desk; or, for that matter, It is true that it is true that there is a teacup on my desk. And this is not, in any case, our main preoccupation. In so far as there are characteristic proper moves for the conduct of dialogues into which locutions reporting perceptions may enter, there are moves of the same kind in respect of locutions of the form S is true as there are for plain S. And that is all, at this level, that needs to be said about the supposed special kind of truth that attaches to perceptual locutions. What would we say of a conversational machine that made perceptual claims very unlike our own? We might, of course, write it off as wild and unpredictable. But I do not think it is necessarily the case that we should do so. There are even imaginable circumstances in which we would treat the machine as being more perceptive or as having greater acuity or perceptual discrimination, or as being less open to illusion, than we are. If its designer were to attempt to suppress all tendencies to make judgments diverging from our own, he would, after all, need to build into the machine a susceptibility to certain kinds of illusion to which we are prone. Whether this would be reasonable would depend on the purpose of the project. But who is right? We seem to be arguing that there is no truth obtainable from perception. A so-called "theory of truth", we said, is more than a rule for the use of ...is true it must give a set of criteria for the acceptance or rejection of statements, independently of whether the words true and truth are involved in them. And we now seem to be denying that percep-
144
Chapter Ten
tion has any special role in connection with disposing us to accept or reject perceptual statements; or, at least, declaring our inability to say what this role is. Isn't it preposterous to suppose that a machine without senseorgans, or with arbitrary ones, could really be worth conversing with on the subject of perceptual fact? Before we tackle this question we must consider what the factors are that dispose us to reach decisions - in one direction rather than another - in the case of perceptual disputes. Claims can be pressed strongly or weakly. And a claim that a certain immediately perceptible fact is so may be one of importance, or one of no importance whatever. That we do not often get at one another's throats over questions of perceptual fact is, no doubt, partly a reflection of the fact that our disagreements are generally unimportant. Under these circumstances it is a matter of good sense to take one of the easy escape routes. The move to That's how it looks to me is, in any case, no great retreat. It is very different from a denial, and it does not necessarily resolve the issue. At least, there is a stronger sense in which the issue may be "resolved", namely, that agreement should be achieved concerning the perceived fact and not merely concerning the appearance to respective participants. I do not want to be accused of suggesting that the question of what is really the case is a matter of majority vote. However, it is a simple fact that claims are strongest when pressed corporately by a large number of people; and if Q marshals R and others to support him against P, they will usually soon raise morale to the point of claiming S is the case rather than It seems to us that S; and then it is likely that P starting from S is not the case, will retreat at least to It seems to me that not-S, and eventually to conceding in spite of how things seem to him. (Further retreats would be possible even to It seems to me after all that S.) P, it may be universally agreed (even by P), is suffering from a disturbance of his perceptual faculties - for example, from colour-blindness - and must make allowances for this in future judgments. I say it is likely that this will be the outcome; though it is, conversely, possible that P by superior determination or arguing power, will convince Q and everyone else that they are all wrong and that he alone has made the correct observation. Possibly P will succeed in maintaining the simple indicative S is not the case against all the weight of opposition, and will persuade Q, R and all others to concede S looks to us to be the case, but we must be under an illusion. What pressures or arguments P may employ in securing this end is not really to the point here. The fact is that this is a possible, if unlikely, outcome. And the relevance of majority rule to such cases is expressed as follows: counting heads is one, among others, of the ways in which we measure the
Perception
145
relative merits of claims. This is generally true even when we are ourselves claimants. Illusions are treated - in this cart-before-horse fashion - as being deviants from the norm, rather than the norm; and we like, when possible, to have causal explanations of them. Physiological malfunction, distortion of perspective, inattention, mental illness, hypnotism - all these are causal candidates. Lying and verbal misdescription are on the sidelines. But explanations are secondary, except in so far as they may themselves influence the argument and aid the pressures. No one point is decisive. Consequently it is not possible that there should be a general theory of empirical truth, a theory of the validation of perceptual claims. We should like to think that P ought to be able to determine his attitude to any dispute with Q and R by applying criteria obtained from a theory; then having a reasonable chance of convincing Q and R similarly. And it is true that if P is armed with a convincing theory he may be able to make use of it either in bolstering his own morale or in persuading others of his superior grasp of the truth. But it is also the case that Q and R could make use of a theory - perhaps the same theory! - in their own representations. If this is so, the issue is as open or as shut as before. And an onlooker, who may apply a theory of perception in adjudicating the dispute for his own edification, can hardly be regarded as standing above the participants themselves, however much he may pretend to do so. One kind of theory of the validation of perception, for example, claims that certain perceptions, at least, which stand out from others as especially vivid or distinct, can be relied on. Descartes' "clear and distinct ideas", though they are not necessarily drawn directly from sensory perception, are an example of this; and Hume sometimes speaks of "particularly lively impressions" as being self-validating. But someone who says My perception is particularly distinct, clear and lively and consequently it must be correct cannot always expect this to be accepted as an argument by others with whom he is in contradiction. If he does succeed in persuading them by making this statement especially solemnly or forcefully, the credit should go to his arguing powers, not to his theory; for it would be a simple matter for opponents to claim that his perceptions only seemed to him clear, distinct and lively. An onlooker who says I can see clearly that P is right and Q wrong may be on the soundest of grounds in saying so but, remaining uncontradicted in the absence of a counter-view (from Q or from another onlooker), has to this extent failed to meet the pressures that might be applied against him. If he did, the fact that he claimed that his perception was a clear one would add little or no weight. Another way in which people sometimes try to support perceptions is
146
Chapter Ten
by appeal to past successes. Justification of perceptions, it might be said, proceeds a posteriori, and if Jones has usually been right in making observations of this kind in the past we ought to trust him in the present. This is all very well if you and I are not Jones, but it is less satisfactory for me to try to justify my own perceptual claims by saying I am usually right for it is necessary to suppose that there are independent means of checking up on my record. That a barrister has won a large proportion of his cases, though it might be a good reason for supposing him to be likely to win future ones, is hardly an argument he should be allowed to urge in court in support of any of these future ones themselves. And if a particularly hypnotic artist, say, were to claim that reality is all cuboid, or predominantly purple, the fact that he had succeeded in making us believe many objects to have this form would be no argument for the case of the next object if we found it looked different to us. Again, it might be supposed that engineers will succeed in designing an extremely reliable perception-machine which, attached to a conversational computer, will be able to make sensory judgments that are more acute and perspicuous, and less liable to illusion, than our own. The theory underlying its design would represent a theory of the validation of perception. And I do not deny the possibility of such a machine. But let us suppose that one day the machine, functioning perfectly, makes a perceptual judgment that differs from what everybody else declares to be the case. Its designer, perhaps, will trust the machine against the perceptions of everyone else, even himself; but lesser mortals would not give in without argument. In short, when it comes to the crunch, a theory does not function as a criterion but is thrown into the dialectical melting-pot like everything else. We can now apply what we have been saying to the question of whether locutions of other kinds than indicative can be underpinned by perception. Do moral imperatives come to us in this way? And is there a perceptual basis for our emotive reactions and other attitudes? In what we have said about perception so far we have been concerned only with the senses of sight, hearing and the like; but it has sometimes been suggested that we have a "moral sense" that operates independently of these, and it could as easily be suggested that, if emotions come to us from elsewhere rather than being generated within us, they do so through one or more "emotive senses". No one, of course, has any notion how such additional senses might operate. In most of the uses that have been made of "moral sense" by philosophers the basic information, such as that there is a blind beggar holding out his hat, has come through the ordinary senses anyway, and the moral sense gets to work on the products of these in generating its own products, such as that I should (or should not) contrib-
Perception
147
ute a coin. It seems to follow that a moral sense is an unnecessary passenger in our conceptual scheme, and presumably the same applies to emotive ones. Extra eyes or ears are no help in perceiving right and wrong, but neither are they necessary. This does not, however, settle the question of whether moral and emotive "truths" can be perceived and it is clear, from what we have been saying, that it cannot. If someone claims that, in various everyday situations, moral imperatives come to him directly through his senses, he makes a claim that can easily be dismissed with You only seem to or You only think you do - but which could as easily be supported by like-minded people, with possible persuasion and conversion of their opponents. And if we object to Jones's characterising the sight of a starving child as pitiful, on the grounds that the emotion of pity is in ourselves and not objectively in the appearance of the child, we are likely to be accused, at the least, of quibbling. Some sights, we shall be told, just are pitiable. And although it is the right of philosophers to quibble, it is not clear that they have any knock-down reasons in cases like this, of the kind that determine that these non-indicative locutions are necessarily non-empirical. When all is said, the strongest - and perhaps only - argument that can be adduced against the existence of imperative and emotive perceptions is that there is relative disagreement about the supposed perceived entities. If we see these things we see them differently; sometimes radically differently, to the extent that the hypothesis of systematic sensory malfunction becomes implausible and needs to be replaced by the alternative one of differential upbringing, conditioning processes and psychosomatic constitution. That is to say, moral imperatives and emotives need to be regarded as produced primarily from within, merely triggered from without. It follows that the qualitative difference between imperatives and emotives on the one hand, and indicatives on the other, is smaller than it would seem. If there were less disagreement about moral locutions we might start regarding them as the products of sight and hearing; and if there were less agreement about indicative ones we might start regarding them as not so. In either case, there would be a consequent readjustment of idiom and of available dialectical moves.
THE NATURE OF THE THEORY
CHAPTER ELEVEN WHO MAKES THE RULES?
The theory of language outlined in this book contains contentions at two different levels. In the first place I have suggested that there are three more or less interrelated kinds of basic utterance or locution, the indicative, imperative and emotive, and that the "logic" of linguistic interchange between people - or machines - is the logic of these three kinds with their associated paraphernalia such as questions, retractions and retraction demands. The three-fold classification is primarily a classification of commitments; and, as this word is properly understood, this implies a trio of commitment stores, with some interconnections, in the brain or the machine memory. And in a more shadowy way this implies a threefold classification of mental contents - facts, acts, and sentiments - behind the commitment stores. To what extent this is really a psychological thesis I leave open. At this level, however, the theory is not one towards which I am at all inclined to declare a passionate philosophical attachment. It seems to me not a bad theory as these things go. It makes a number of important distinctions, and it is at least closer to the truth than I have seen any comparably detailed theory come. But I would be the first to admit that it has many shortcomings, and that I have had to fudge a number of idioms to make them fit. The only claim I am really disposed to make about the threefold cut-up and the plethora of stores is a rather weak one, that this seems as good a way as any of summing up the linguistic phenomena. This part of the problem, in any case, is one for linguists. I am not a linguist, and tackle these matters only in quixotic spirit, because it seems that something needs to be done. But I would be delighted if some betterqualified researcher would take the field in my place, and I would not be perturbed if he should find that my cut-up of it is wrong. But that brings me to the second level of the theory. It seems to me utterly indisputable - and this is my main point that language is primarily "dialectical", in the sense in which I have used this word, namely, that it is oriented towards the exchange of information, requests, attitudes, and such like, in dialogue. The rules of dialogue are the
152
Chapter Eleven
most fundamental rules of language. Beside these rules, the ones that spell out the fine detail - such as those that give us the grammatical structure of individual sentences - are secondary and unimportant, and may for that matter be definable dialectically too. But this at once puts language into contact with its surroundings, and reveals it as having features dependent on them. The apparently arbitrary character of grammatical rules vanishes and is replaced by a set of adjustments to contextual needs and pressures. And this thesis, now, is a philosophical one which has consequences for the whole enterprise of language and for the study of philosophy itself. We have become involved in a great deal of uncomfortable scepticism in the effort to adjust ourselves to this thesis. I have argued that this is unavoidable, but it is important in any case to lay it at the right door. The theory that language consists basically of indicatives, imperatives and emotives - or that this is a promising basis on which to start building a linguistic theory - is not itself an encouragement to scepticism. At its first level the theory is not only too weak but also too unphilosophical to have any such consequences. It would also be possible to imagine various excursions at the second level that could be conducted without these complications. For example, it is almost a truism to say that languages are for communication and reflect the fact in their structure. But it is all too frequently the case that those who theorise at the second level start with presumptions of the nature of reality which, on examination, are found to be cases of the carriages pushing the engine in that they foist on reality features that have really been derived from language. This fact, again, is well recognised by linguists, but the conclusions are seldom drawn. Our doubts arise, rather, from the realisation that the rules of language are not arbitrary, as those of elementary grammar seem to be, but really bind us. So long as we think they are arbitrary or artificial we fondly imagine that it is possible for us, at will, to transcend them, and to communicate "directly" with those around us, otherwise than through the medium of language. We think we can, as it were, change the rules to suit the occasion. And so far as elementary grammar is concerned, this is certainly the case. But to vary these detailed features of a language is one thing, and to vary its more fundamental ones another. So far as these more fundamental features are concerned we have really no alternative, on pain of sheer failure to communicate, to simply going on using our language as we have it. Questions of the justification of the shape of the existing rules all too frequently cannot be meaningfully asked. Can we alter our language at all? Since language has grown up as the product of social forces can't we, if we come to learn about those forces and their operation, turn them to account and mould the future of our
Who Makes The Rules?
153
language consciously? Don't our more inventive writers and speakers in fact continually do this? There is no doubt that language steadily changes, even in its fundamentals. But it is very doubtful whether important changes can ever be regarded as the work of individuals. The invention of an adjective, the popularisation of a particular apposition, may be, but these are minor matters. Of changes affecting our classification of locutions we have noticed two in earlier pages: varying fashions in the use of command, demand and request corresponding with changes in the authoritarian nature of society; and (though this is harder to document) a movement away from pure emotives in favour of elliptical ones that are grammatically indicative, perhaps corresponding with increased literacy and the greater use of print. Even these are far from being really fundamental although, given a few hundred years, they could superannuate any classification that was too rigidly conceived. Much more important would be changes, if they were possible, that would overcome the limitations of language as they are conceived at the second level. Can we come to say things that at present cannot be said? things that will dissipate our philosophical scepticism? Some people claim to do this. But they are of disparate kinds. In the first place, there are the mystics, who tell us there is truth in falsehood, paradox and nonsense, and that we may in this or that way liberate ourselves from the limitations of language; that words are a barrier we build between ourselves and things, and that we should penetrate or knock down the barrier and relate ourselves to things directly. This kind of doctrine is a familiar enough one but I do not at all know what to say by way of appraising it. Some people, no doubt, achieve pleasurable states or useful insights by convincing themselves of it. I am not of their number. But there is another kind of liberation that is often proposed, not by mystics but by what I might call "activists", who see a preoccupation with verbal trivia as a barrier in the way of their pursuit of this or that social or political goal. With some of these people - depending on what their actual goal is - one must have the greatest sympathy. They remind us, like Marx, that "philosophers have explained the world; our task is to change it", and turn our attentions from theory to practice. But again, once this is said, it should be clear that the actual words in which their contention is put do not have much to recommend them. It is an irony of history that Marx has himself become elevated to the ranks of the philosophers, with an explanatory theory of his own; and there must always, in any case, be something odd about saying that words cannot give us truth or that we must transcend language. The question what the world is like and the
154
Chapter Eleven
question what we should do about it are two different, if related, questions and there is no need to neglect one in the process of trying to answer the other. In fact, one of the limitations that activists see in language has always been illusory, and this fact should have been completely exposed by this book. So long as we think of language as an instrument for fact-stating, it may be asked with some reason what its connections are with the conduct of life; but so soon as it is pointed out that to urge, recommend, or exhort people to particular kinds of action or life-styles are themselves linguistic acts, with their own grammar and logic, and that even the emotions have their characteristic forms of linguistic expression, it must be clear that language is too useful a tool for anyone seriously to suggest dispensing with it. It is the activist's best friend, and a refuge in company with mystics ill becomes him. Consequently I do not think that any contention that we can enrich life by transcending language need be taken seriously. Richness of life, if attainable at all, is attainable without deliberate attempts to talk nonsense. Actually, a far more likely modern dénouement of such attempts is the impoverishment of life through failure of communication between the various sectors of society, brought about precisely by lack of appreciation of linguistic niceties. But I am not sure that the shape of our language really has much to do with these matters. If it had, social problems would be both less serious and more permanent. The particular limitations of language that we feel we should like to remove are the Cretan ones, as revealed, for example, in our inability to say whether what we are saying is really true. P, Q and R make statement S; but, says T, what is the use of my hearing from Q that what P says is true, or from R that the same is true of Q, when all of them are trading in mere words? Why can't they say something relevant about the real truth of S, as distinct from merely repeating S and related statements in different words which is all P, Q and R ever do? This is obviously an impossible request. Yet people say But we do break the barriers sometimes. P might overcome our doubts by saying simply You know I'm telling the truth. But the point here is that, if this is effective, it is for reasons that are outside language and reason. P, for example, makes his statement firmly and with a steady gaze, that inspires trust and would be difficult to counterfeit. We do not need to deny that this may be effective. The firm voice and steady eye show what cannot be said. And sometimes, of course, even "mere" words are effective. None of these things is necessarily effective, of course; and we sometimes need to be reminded of how easily confidence-inspiring externals can be
Who Makes The Rules?
155
counterfeited. Nevertheless, we would not be doing these phenomena justice if we did not admit that there is something to be said for regarding them as ultimately essential to the operation of language. If no one trusted anyone else, linguistic communication would not be possible. Before developing this point, let us consider a feature of our own theory of language about which similar things might be said. What is the nature of the rules of dialogue that we have been putting forward? Who makes them and where do they derive their force? Are they empirically discovered, by seeing how people actually converse? What is it to recognise a rule as a rule, and to conform with it - or break it? Is the existence of rules a matter of intuition? It can be imagined that someone might become impatient with the whole scheme of rules and commitments in something like the same terms as before: P, Q and R converse in accordance with the rules of dialogue and expect others to do so; but, says T, what is the use of it all? Now that I am told that dialogue, as it were, is a kind of rule-governed trading in commitments, isn't it degraded to the status of a sterile game? If the rules are essential to discourse, show me directly that they are. (Don't just claim it, continuing to play the game within the rules.) And if they are not, let us have nothing to do with them. Like the demand for "real" truth, this is a demand that cannot be satisfied within language. Both are cases of loss of linguistic morale. And they call for similar therapies: the firm assurance, steady eye, perhaps words which have the same effect. And once again it seems that general willingness to continue to use the regular mechanisms of discourse is a precondition of the existence of language. This consideration does not justify any particular theory of language, but it refutes any suggestion that there should be no theory. There is nothing unchangeable about the rules of dialogue; they do change. But not only is it not possible for one person to change them all by himself; even a group of people cannot impose a change on others without their agreement. By means of a suitable campaign one might persuade people to change, or condition them; but this becomes a political question. It is particularly remarkable, moreover, that much is apparently constant in all natural languages. Thus the questions Where do the rules come from? and Who enforces them? are to some extent misplaced. They tempt us to rationalise the rules, and this would be like committing Moore's naturalistic fallacy. Rules are rules. If there is any answer to the question how they are enforced, it is that we enforce them on one another. We may now turn to a short examination of the usual enforcement mechanism. In the formal proceedings of boards, parliaments and committees there
156
Chapter Eleven
is commonly a chairman who is concerned with procedural matters such as who speaks when, whether what is said is relevant to the proceedings generally, and so forth. In ordinary dialogues there is no such person, but there is often some parallel or a supplementary feature of committee procedure, the objection or point of order. Irregular locutions are greeted with one or these, such as "What do you mean by that?" or "You haven't answered my question", "That's not relevant" or "Wait a moment, let me finish". If we call these in turn points of order we must have something to contrast them with; and I suggest that the locutions that are, unlike these, directly concerned with the subject or topic of the dialogue (assuming it to have one) be called topic points or, for short, topical. Points of order are concerned with the occurrence and propriety of particular topic points. It is no objection to this characterisation that there are circumstances in which it cannot be applied. Thus a dialogue may wander and have no clear topic. A point of order may be debated at length, and then become the topic, subject to yet further points of order, and so on. The interests of participants ensure that these cases, though possible, are not usual. The point is that points of order - or at least some of them - are simultaneously sanctions, exerting pressure to conform with this or that relevant rule. Or else they are sanction-warnings, or sanction-surrogates; for to the extent they can be misused they might be said to be no true sanctions, as are acts which would directly harm someone's interests. On the whole, points of order are accompanied by the suggestion or contention that there is some linguistic impropriety, and by the demand that it be put right. Commonly, the only way for the dialogue to proceed is for some retraction and reframing to be made. And if points of order need to be raised continually, or if they are not suitably answered, communication may be said to have broken down, at least as regards the topic under discussion. There may be cases in which it is difficult to tell a point of order precisely from a topic point. For example if P asks Did she wear the red hat, or the white, or the blue? and Q replies None of these, we may be uncertain whether to regard this as a rejection of the question and a demand that it be reframed or simply to regard it as an answer. It is not, of course, very important which we say. Something will depend, however, on Q's attitude and tone of voice. If he sounds as if he disapproves of this non-exhaustive enumeration, we may well decide that he is really raising a point of order; and if P takes it as such he may go on to say Sorry. Which was it, then? Thus there may or may not be a rule that question-askers should enumerate all possible alternatives. What are points of order like when they are explicit? The first thing to
Who Makes The Rules?
157
notice is that they may be of any mood. That is, the addition of points of order to our locution-armoury does not invalidate our previous classification or, actually, augment it. It cuts across it. A point of order may consist of saying that a rule has been broken, or asking whether it has or what a given locution means, or of telling or advising someone to avoid or remedy a breach, or of stating or asking whether such-and-such should be allowed, or of expressing an attitude towards a speaker or locution; or of withdrawing a procedural commitment. There are other possibilities, and it is difficult to be sure of having exhausted the kinds. Points of order, in brief, may be debated in all the forms of locution available for topic points. The second thing to notice is that points of order inevitably involve reference by the speaker to his own and his addressees' locutions, and their dialectical contexts. This can be effected either (a) by quotation, direct or indirect, as in Your statements 'Adam was a Frenchman' and 'Adam was born near Cambridge' are in contradiction or What did you mean by saying that there was nobody at the lecture?, or (b) by some system of naming, numbering or descriptive reference. For example, I can say Your last statement but one has already answered that question, or What I meant by what I said in answer to Robinson was ... We could build a simple model of a dialectical system containing points of order by supposing that every locution is numbered and that speakers may refer to their own and one another's locutions by their numbers. But there can also be points of order involving whole classes of locutions, as in Everything you have said so far has been obscure, or open universal ones such as Stop contradicting yourself, that is Let no future locutions of yours be in contradiction to your commitments; and these require reference by description. All in all, the task of enumerating the possible kinds of point of order does not seem an easy one. With self-reference comes the possibility of paradox of the orthodox Epimenidean variety; that is, of the variety normally discussed by logicians. Since there will need to be reference to truth and falsehood, it must be possible to say of statements referred to by description that they are true or false, and it therefore seems impossible to avoid permitting statements such as What I am now saying is false; or, say, Locution no. n is false, where this is itself locution no. n. Or, if direct self-reference of this kind were somehow banned, there could be more roundabout varieties, perhaps involving an element of cooperation, as when P says Your next statement will be true and Q replies That's not so; or of generalisation, as when someone says There are no true logical generalisations. But beyond calling attention to these possibilities, I have nothing to say about them.
158
Chapter Eleven
They seem to me to be unavoidable features of language. What happens when one speaker objects to another speaker's point of order? Can he raise a point of order at second remove? In standard committee procedure there are safeguards against this, since points of order must generally be debated only briefly and put to vote without interruption. We might feel that in an ideal dialogue outside the committee room the same should apply. But this feeling is hardly more than an extension of our possible feeling that points of order should never be necessary in the first place. If there is a rule that points of order are not themselves subject to points of order - or, in fact, if there is any rule at all about how points of order should be raised and discussed - it is subject to breaches, and all that anyone can do about breaches of a rule is object to them - by raising yet further points of order! When a point of order is settled it is, perhaps, usual for speakers to resume their original topic; but sometimes, no doubt, they fail to do so, and the point of order becomes the topic, subject to yet further points of order. It is possible to imagine someone objecting to this, with We are getting further and further away from the original point, but it is also possible to imagine this objection contested. Some points of order involve reference not so much to particular locutions of the dialogue or to classes of them but to what we might call "conditions of rational discourse". If I say You're not listening to me I call attention to the fact that it is a condition of the satisfactory conduct of any dialogue that the participants pay attention to one another. This does not mean that we are necessarily better off if we listen to one another, for there might very well be cases in which not listening to someone is more in my interests than doing so. Listening, rather, is a linguistic necessity, and someone who does not listen to this extent breaks a linguistic rule. It is a condition of communication by language that the parties concerned be conscious, be (reasonably) sober, know the language, be within earshot (or within reach by whatever other sense), pay attention, and be capable in principle of replying; and perhaps other things. And if they are not, and if circumstances permit an objection, an objection in appropriate terms is likely to be the outcome. Of course, if your addressee is really unconscious there will be no point in objecting to him, and any objections anyone raises will be better addressed to yourself: an onlooker may say It's no use talking to him or he himself may wake up and say What was that? I've been asleep. But these practical points do not affect the point of principle. An especially odd kind of locution is the kind addressed to a deaf, drunk or sleeping man, or to someone who cannot hear or does not know your language, in order to see whether (or to bring out that) the relevant
Who Makes The Rules?
159
condition of discourse is satisfied I may vainly say Can you hear me? into a dead telephone, or Parlez-vous français to an uncomprehending Arab. Or I may say Hey, wake up! to a sleeper, knowing, if I think about it, that it will be the loudness or suddenness of my locution that will wake him rather than any understanding of, and willingness to comply with, the order itself. There is a sense in which all such locutions are misuses. When I want to know whether an Arab knows French, asking him in French is an odd way of going about it: if I cannot ask him in a language I know he knows, I must try him out on a passage of French to see how he reacts, but any passage at all would do, and I merely waste words (and get my logical wires crossed) by using one which raises the point of order rather than one which might perform a topical function as well. The same applies to the other examples. But let us return to the main point about points of order. Whatever their logical form, their principal function - which distinguishes them from topical locutions - is that, taken together, they control dialogues rather than contribute to their content. They show the flag. They apply pressure on other participants to conduct the dialogue - and other dialogues - in one way rather than another. Or they resist such pressure, or raise questions concerning it, or remove or relax it, and so on. But just for this reason it is a little odd to inquire closely into their form. If points of order are what gives substance to the rules of dialogue, they must also be what gives substance to themselves. If the point of my raising a point of order is to get you to say such-and-such rather than soand-so, and you think I should raise my point of order in such-and-such a way rather than in such-and-such other, your own consequential point of order must be seen not only as debating with mine but also as being engaged in some kind of trial of strength with it. This does not mean that we shall necessarily both abandon reasonable language, for there is no easy alternative to continuing to debate matters on whatever common ground we have, which is presumably considerable. In saying that points of order are sanction-surrogates I do not mean that the ultimate sanctions of throwing chairs, social ostracism, or pistols at dawn are in any sense in the mind of anyone who uses points of order. These things are only the extreme and ultimate form of what appears here at its gentlest. There are all sorts and levels of sanction, and the threat of failure to communicate is usually enough. But this does not alter the fact that points of order have a force. In saying that the main function of points of order is to be low-level sanctions I do not want to suggest that rational justifications of points of order are irrelevant. Rather, rational justifications themselves - bringing
160
Chapter Eleven
out the extent to which communication would be prejudiced if the points of order were not heeded - are part and parcel of the sanction-machinery. I seem to be saying Rules don't really bind. It's all right to break them if you can get away with it. But this is a confusion, since it seems to say (a) that there are rules, and (b) there are other considerations that determine whether the rules may be kept; that is, that there are other rules governing the first ones. This would be a clear case of both eating one's cake and keeping it for the next meal. For if the rules are the expression of some kind of consensus of feelings for language, someone who breaks the existing ones though he is not necessarily answerable to the old consensus, is perpetually answerable to whatever consensus eventuates in the new situation. To say It is all right if you can get away with it is like claiming that there is no argument for conformity; but this is hardly something the majority, who create the consensus, are likely to agree to. To this extent it is a Cretan paradox of a kind we are beginning to be familiar with. In so far as all procedural locutions help to mould language it would be circular to try to justify them in the name of language, and viciously circular to condemn them. In fact one might as well turn the contention round. If there is a case for questioning the rules of language - of a kind that might create a consensus that would stand against the rules - then one could as easily argue that it would be wrong to keep the rules. But right and wrong, though they enter our language in a casual way as an expression of linguistic conformity or non-conformity, hardly enter the picture as a whole. Language as a whole, in its social setting, does not obey super rules. If there is a justification of any idiom in this enlarged picture, it is not a linguistic justification, but a moral, practical or political one. And precisely then we are outside the scope of appeal of linguistic rules. The paradox involved in the discussion of the force of points of order resembles a political one I might call the "Bismarck-Lincoln" paradox. Bismarck, in a speech to the Reichstag, toyed with the slogan Might is Right; and he subsequently became tagged with it by his political opponents, who said it represented his policy. Abraham Lincoln thereupon committed his government to the supposed converse policy Right is Might. Since to say that A is B is the same thing as to say that B is A, it is not entirely clear what the difference between the two is. To say that the rules of language are the expression of pressures exerted by a linguistically powerful majority sounds very much like saying that, in language, Might is Right. It is a species of logical Bismarckism. But if this thesis offends people I am quite willing to substitute for it the corresponding Lincolnian thesis, to the effect that conformity with
Who Makes The Rules?
161
linguistic rules commends itself, and that bringing to someone's attention the existence of a linguistic rule is itself a means of applying pressure to conform. I do not see that these theses are significantly different. Both are rather Cretan. It will incidentally be the case that, since much - I will not say all philosophy turns on linguistic issues, a theory of language is a philosophical doctrine. And it further follows from what I have been saying that, in many areas of philosophy, Might is Right; or, if you prefer, Right is Might. But it also follows that philosophers often tread Cretan ground and deal in paradox. As I have said before, I do not think there is anything at all that can be done about it.
BIBLIOGRAPHY
Alexander, Samuel. 1920. In Space, Time and Deity. Book III, ch. 5, New York: Humanities Press. 5,-89653)5#699)5,-35)7 ;-38 B%0-!;8-)3+;3;9 6.5:)134-5:CJournal of Symbolic Logic, 27 pp. 19-52. 5,-8965605 B 15,)9--315/CAustralasian Journal of Psychology and Philosophy, 12 pp. 81-94. Åqvist, Lennart. 1965. B!)8:C5A New Approach to the Logical Theory of Interrogatives. Analysis. Uppsala, Almqvist & Wiksells. Aristotle. n.d. Nicomachean Ethics. A. n.d. Rhetoric Book II. A. n.d. The Mind. Austin, J.L. 1962. How to Do Things with Words. edited by J. O. Urmson. Oxford: Blackwell. A. 1961. Philosophical Papers. edited by J. O. Urmson and G. J. Warnock, Oxford: Blackwell. Ayer, A.J. 1946. In Language, Truth and Logic, 107. London: Gollancz, 2nd, edn. Belnap, Nuel D., Jr. 1963. An Analysis of Questions: Preliminary Report. Santa Monica: System Development Corp. 86=515/#5,B8)44)81)59;5-8)3C
164
Bibliography
;90&)55- B9'- )>%0152CAtlantic Monthly pp. 101-8. Carnap, Rudolf. 1934. The Logical Syntax of Language. translated by Amethe Smeaton. London: Routledge & Kegan Paul, 1937. 06492> 6)45,B$>5:)+:1+$:8;+:;8-9C7 Dakin, Robert. n.d. University of Papua and New Guinea. Darwin, Charles. 1872. The Expression of Emotion in Men and Animals. London: John Murray. Descartes, Rene. 1637. Discourse on Method. Part IV. Fries, C. C. 1952. The Structure of English. New York: Harcourt Brace and Company. )4*315 ;/;9:&63 B ):0-4):1+)3 6,-396. 1)36/;-CTheoria pp. 130-155. A. discussion pap-8)5 B";-9:16598-5:$:):-4-5:9CPhilosophy of Science, 30, pp 62-3. A. -+-4*-8&63
B%0-..-+:6.'0-5:9$)1,CTheoria pp.249-263. Harrah, David. 1963. Communication: A Logical Model. Cambridge, Mass: M.I.T. Press. Healey, Al)5;/;9: B5/3190,1649C5Kivung, vol 1, No2. James, William. 1907. In Pragmatism, A New Name for some Old ways of Thinking, ch. 6. New York,: Longmans, Green and Co. Jespersen, Otto. 1924. In The Philosophy of Grammar, Ch. 22-3. London: George Allen and Unwin.
Bibliography
165
):?-8863,B5:-878-:):1