186 90 949KB
English Pages 179 [187] Year 2008
Michael Frauchiger Wilhelm K. Essler (Eds.) Representation, Evidence, and Justification
LAUENER LIBRARY OF ANALYTICAL PHILOSOPHY Edited by Wilhelm K. Essler and Michael Frauchiger
VOLUME 1
Michael Frauchiger Wilhelm K. Essler (Eds.)
Representation, Evidence, and Justification Themes from Suppes
ontos verlag Frankfurt I Paris I Ebikon I Lancaster I New Brunswick
Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de.
North and South America by Transaction Books Rutgers University Piscataway, NJ 08854-8042 [email protected]
United Kingdom, Eire, Iceland, Turkey, Malta, Portugal by Gazelle Books Services Limited White Cross Mills Hightown LANCASTER, LA1 4XS [email protected]
Livraison pour la France et la Belgique: Librairie Philosophique J.Vrin 6, place de la Sorbonne; F-75005 PARIS Tel. +33 (0)1 43 54 03 47; Fax +33 (0)1 43 54 48 18 www.vrin.fr
2008 ontos verlag P.O. Box 15 41, D-63133 Heusenstamm www.ontosverlag.com ISBN 13: 978-3-938793-82-4
2008 No part of this book may be reproduced, stored in retrieval systems or transmitted in any form or by any means, electronic, mechanical, photocopying, microfilming, recording or otherwise without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed on a computer system, for exclusive use of the purchaser of the work
Printed on acid-free paper FSC-certified (Forest Stewardship Council) This hardcover binding meets the International Library standard Printed in Germany by buch bücher dd ag
Contents
General Foreword Henri Lauener and the Lauener-Stiftung Wilhelm K. Essler
1
Preface Michael Frauchiger
5
Laudatio for Prof. Dr. Patrick Suppes Dagfinn Føllesdal
9
A Revised Agenda for Philosophy of Mind (and Brain) Patrick Suppes
19
The Emergence of Justification in Ethics Dagfinn Føllesdal
51
Comment on Føllesdal’s “The Emergence of Justification in Ethics” Patrick Suppes 67 The World and the Worlds Wilhelm K. Essler
71
Comment on Essler’s “The World and the Worlds” Patrick Suppes
79
In Praise of The Representation Theorem Nancy Cartwright
83
Comment on Cartwright’s “In Praise of The Representation Theorem” Patrick Suppes 91 Modeling in Philosophy of Science Stephan Hartmann
95
Comment on Hartmann’s “Modeling in Philosophy of Science” Patrick Suppes 123 Symmetry, Invariance and Reference Steven French
127
Comment on French’s “Symmetry, Invariance and Reference” Patrick Suppes 157 Interview with Patrick Suppes by Michael Frauchiger About the Editors
163
General Foreword Henri Lauener and the Lauener-Stiftung∗
Mr. President of the University of Bern, dear colleagues, ladies and gentlemen: Thank you very much for taking part in this great event. For this is the first presentation of the Lauener Prize by the Lauener-Stiftung. Well, what is the Lauener-Stiftung, and first of all, who is Lauener? Of course, this question is somehow wrong; it should be reformulated either as: * “Who was Henri Lauener? ”, or as: * “Who is Henri Lauener to me? ”. Well, I do not intend to answer the question “Who was Henri Lauener? ” right here and now; for that would be somehow like a belated obituary. And, in any case, you know him: You are familiar with what he did for developing philosophy in Switzerland, and for opening this beautiful country’s philosophical activities to the philosophical world of the 20th century in several respects: by the results of his own research, presented in impressive publications, by his teachings esp. at this central university of Switzerland, and – last but not least – by the numerous conferences he organized in Switzerland. He invited important colleagues from over its borders to these conferences. Thus he gave students and colleagues of this country the opportunity to become aware of the way these guests philosophized as well as the opportunity to establish contact with them and remain in contact with them. By the way: I am sure, you are aware of the fact that Lauener and his contributions to contemporary philosophy are considered absent – from California to India – at least [to be repeated: at least] as much as in his native country. ∗
Address of the President of the Lauener-Stiftung at the Lauener Prize Award Ceremony 2004.
But let me try to give a short answer to the second question: “Who is Henri Lauener to me? ” Well, when he died, I lost a friend, a real close friend, a comrade! Of course, he was not my only friend, by far not. But, by far, he was my best academic friend, or, to be more precise: He still is my best academic friend. Of course, with regard to his body, he is dead. But with regard to his mind, in some sense he is still accompanying me. Maybe, in some respects it seemed very hard to become his friend as well as discussing with him. But, as I see it, it was very easy to become his friend and also to discuss with him. For his behaviour was not unpredictable, by far not. To Henri Lauener, philosophy was not a mere job, a Beruf, but a Berufung, an inner call that he had to obey. And to him, therefore, presenting philosophy was much more than holding some position at the university. For him, philosophizing did not stop when he left his office; on the contrary: he insisted that a philosopher never ever must be thinking A, speaking B, and doing C. In this way, by continuously practising this basic attitude, his way of life was straightforward. And therefore, it was very easy for me to become his Freund, since in this respect we were on the same wavelength. In this way – and, of course only in this way – it was very easy to be his Freund. How to be kind and firm at the same time is what I learned from him in being a Freund. And especially in this sense and in this aspect, Henri will still be with me as long as I am alive. Along with that basic attitude, it was very easy to discuss with him: starting during early evening in some small restaurant at the Lac Léman, and ending that discussion perhaps sometime late after midnight. But, while it was easy to discuss with him, it was by far not easy to in the end win such a discussion. Of course, he did not use any rhetoric just for the sake of winning; nevertheless, his voice was somehow powerful. But things of that kind do not impress me. What was indeed impressive to me was the following: During these discussions regarding the relations between philosophizing and developing one’s way of life he presented the sounder arguments almost every time – well, not exactly every time; in the few cases when my arguments turned 2
out to be more profound he eventually accepted them. Also, it was easy to discuss with him detailed problems of philosophy, despite occupying slightly different standpoints. For his epistemological point of view was something like an undogmatic holism, whereas my position is a strictly non-holistic one, accepting of course the accurate results of holism as being helpful. Differences of that kind never ever became obstacles within our discussions or conversations. Living and philosophizing, these two things were indivisible for him, like the two sides of the same coin; and he maintained both at least up to that time when he lost his beloved small daughter, his beloved wife, his beloved mother, and his beloved dog in rapid succession. Up to that time he regarded accurate philosophizing as being most valuable to life in general and to his own life in particular, and vice versa. Of course, after the above mentioned bereavements he still regarded philosophizing and living, both being firmly interrelated, as most valuable. But with regard to his own life, an unfortunate pessimism began to arise and settle within his mind. Contrary to all that he had done and effected, he increasingly clung to the misconception that his work and therefore his life were without any value. This was the first and, according to my experiences, the only time where he became inaccessible to well-founded rational arguments and attempts to encourage him. Nevertheless, the value of accurate philosophizing, this he continued to promote and to support with all his heart. Therefore, in his last will, he decreed that all his mobile and immobile properties should be sold after his death, and that a Stiftung – a foundation – for the purpose of promoting accurate philosophizing was to be his legacy. While visiting him at Alexandra hospice, where he spent the last period of his life, as well as during telephone conversations, he explained the objective and procedure of his will: The objective of this Stiftung is twofold: on the one hand the objective to present deserving representatives of accurate philosophizing to the younger talents so as to encourage them; on the other hand the objective to support these young talents themselves. This is to be done in the following manner: 3
In the 1st, 3rd, 5th year etc. the prize winner will be a deserving philosopher honoured for his life’s work. In the 2nd, 4th, 6th year etc. the prize winner will be a young talent in accurate philosophizing. In exactly this sense the Stiftungsrat of the Lauener-Stiftung decided unanimously that the prize winner of this year – and therefore also the first prize winner – is to be Professor Patrick Suppes. Wilhelm K. Essler
4
Preface Most contributions to this book were originally talks given at the 1st International Lauener Symposium in Berne, Switzerland, on 09 and 10 September 2004. The Symposium, organized by the Lauener Foundation, was held in honour of Patrick Suppes (Lucie Stern Professor of Philosophy, Emeritus, at Stanford University), who was awarded the first Lauener Prize for an Outstanding Oeuvre in Analytical Philosophy on the same occasion. The present volume, though, is not designed as the proceedings of that symposium nor as a Festschrift for Suppes. It is meant, instead, as a series of extended academic conversations with Patrick Suppes, which mostly started long before the Symposium in 2004 and have been carried on since. The authors have been encouraged to implicitly respond, in their articles, to what had been debated in the substantial discussions which followed their lectures. In addition, Patrick Suppes has shown himself ready to write detailed commentaries on each of the articles. His comments do not only put a different critical perspective on each paper’s topic and argumentation, they also illuminate some crucial aspects of Suppes’ own work which are not easily available in the existing primary and secondary literature. The interview with Suppes at the end of the book was originally made during the Symposium and has subsequently been expanded by correspondence, thereby fitting well in the frame of a prolonged academic conversation. The book’s purpose is to give the reader ample scope for finding out about the present development of a wide range of philosophical and methodological themes on which Suppes has set out seminal ideas. This collection opens a new series by ontos verlag - the Lauener Library of Analytical Philosophy - which will present further constructive dialogues with distinguished philosophers. The new series provides an opportunity for leading authors in the field to connect their own problems and projects with the themes and perspectives of some classical exponents of analytical philosophy – those who have been awarded the Lauener Prize for an Outstanding Oeuvre in Analytical Philosophy.
But, what does it mean to deal with analytic philosophy today? Is ‘analytic philosophy’ not becoming an increasingly inclusive label, which gets constantly stretched, growing less and less significant and eventually pointless? It appears, indeed, that there is a trend towards an inflationary use of the term ‘analytic philosophy’, but that’s exactly what makes it appropriate to assemble influential analytical philosophers of exceptional merit, making them to stand out. For their exemplary oeuvres may illustrate those elements of the analytic tradition which are clearly worth retaining and developing further. It is true though that the analytic tradition in philosophy has always been very heterogeneous, as regards the quite different metaphilosophical, epistemological, ontological, ethical, political, etc. positions which have been advocated within it. There has of course never been any analytic school of thought, apart from some variable movements (such as logicism, neopositivism, and naturalism) that never became universal within the tradition, so that a certain openness, pluralism, and pragmatism have always prevailed at last. On the other hand, there has always been a common methodological denominator, since philosophers fully representative of the analytical tradition have always shown a strong concern for the clarity of their concepts, the neatness of their methods, and the conclusiveness of their explanations, that is - as Suppes puts it in a nutshell in the interview - a “concern with justification, evidence, and argument”. In particular, analytical philosophers have been making a serious effort to benefit, philosophically, from the manifold developments within mathematical logics and semantics, so as to establish standards of truly intersubjective communicability, soundness, and reliability (standards which need not necessarily be universal but contextually appropriate). There are undoubtedly some significant philosophers in the 20th century, such as Husserl and Cassirer, who are not usually classed to belong to the analytical tradition, but whose works clearly have the above-mentioned characteristics of analytical, or rather accurate, philosophizing. Surely all these qualities of an analytical philosopher are in an exemplary way possessed by Patrick Suppes. In spite of their partly opposing philosophical views, the “undogmatic naturalist” Patrick Suppes and the “open transcendentalist” Henri Lauener are both truly analytical philoso-
6
phers who have critically and selectively tried out, improved, and sometimes exhausted the intellectual possibilities accessible to them, without really getting confined to them. At this point, Wilhelm K. Essler and I would like to thank the authors assembled in this collection (Nancy Cartwright, Dagfinn Føllesdal, Steven French, and Stephan Hartmann) for their exceptional commitment to the project of this book. Many thanks also to our fellow members of the Foundation Council of the Lauener-Stiftung (Alex Burri, Stephan Hottinger, Dieter Jordi, and Daniel Schulthess) for their friendly consideration and encouragement. Furthermore we’d like to sincerely thank Rafael Hüntelmann at ontos verlag for his support and willing cooperation in setting up this new series, which is being edited on behalf of the Lauener-Stiftung. Finally, we’d like to express our very special thanks to Patrick Suppes for his extremely generous, open and active way of collaborating with us on this book. Michael Frauchiger
7
Laudatio for Prof. Dr. Patrick Suppes Lauener Prize winner 2004 Dagfinn Føllesdal∗
I am grateful to the Lauener Foundation for inviting me to give the Laudatio for Patrick Suppes today. It is one of the pleasant tasks of my life, both because of Henri Lauener, whom I appreciated greatly, and because of Pat Suppes. I shall now try to present Pat's contributions in 30 minutes. This will not be easy. I have known Pat for exactly forty years, from the beginning of September 1964. Since I first met him I have greatly appreciated him as a friend and colleague, and over the years we have taught numerous seminars together at Stanford on a large variety of topics. However, I have to make a confession: Although Pat has been such a good friend and colleague over so many years, I have never come around to reading more than a fraction of his work. This does not mean that I have read very little, on the contrary, I have read and enjoyed a very large number of his articles and books. However, I must make a new confession: not even Pat's bibliography I have read in full. In preparation for this laudatio I have, however, made some counting, and have I found that so far Pat has published more than 300 articles, written 30-40 books and edited a similar number of volumes. I am here not counting his numerous mathematics text books and popular works. These articles and books fall within a variety of fields, and within each of them Pat has made important contributions. When Pat was appointed to Stanford in 1950, still a very young man, in his twenties, he came to the Philosophy Department. But he was very quickly also made a Professor of Statistics and of Education. However, when we look through his bibliography and vitae, a curious oddity about chronology emerges: in 1972 Pat won the Distinguished Scientific Contribution Award of the American Psychological Association. This award is given for the best research in psychology that year and it would be a pride for every psychol∗
Stanford University
ogy department to have a winner of this coveted award on their faculty. However, at that time Pat was not a member of the Psychology Department, so the award went to the Philosophy Department. This may explain why the next year the Psychology Department belatedly came around to making him a member of their department. As impressive as the number of his contributions and their quality is the broad variety of fields in which Pat has played a decisive role. When we look at Pat's work within each of the many fields in which he has worked, and also when we survey his work as a whole, there are two things that strike us about Pat. One is his openness, he is always very eager to hear about new developments, to learn something new, and he is in fact particularly interested in perspectives different from his own. This is, of course, a good thing, but it can be overdone. One can become superficial and uncritical. Pat fortunately compensates for this openness with a second feature which is equally characteristic of him. As soon as he recognizes what is maintained from that very different perspective, he starts asking: "What is the evidence?" This combination of enthusiastic openness and critical questions about evidence is typical of Pat. It is also a character trait he shares with Henri Lauener, who devoted his life to promulgate the use of arguments and evidence in philosophy. This emphasis on evidence brings me to an early episode in Pat's life which I will mention. Pat's mother died when he was four and a half, and he was raised by his stepmother, who married his father before he was six. This stepmother encouraged him in a variety of ways to pursue his intellectual interests. One of these ways was, however, unintended by her. I will quote from Pat's autobiography: She was devoted to the Christian Science of Mary Baker Eddy. From about the age of eight to fourteen years I attended Sunday school and church regularly and studied the works of Eddy as well as the Bible. The naive epistemological idealism of Eddy’s writings stirred my interest, which turned to skepticism by the age of thirteen or so. I can remember rather intense discussions with fellow Sunday-school students about how we were supposed to reconcile, for example, the bacterial theory of disease with the purely mentalist views of Eddy. No doubt our arguments were not at all sophisticated, but our instinct to distrust the flagrant conflicts with common sense and elementary science was sound.
10
Before I turn to Pat's work in philosophy I will mention a couple of other noteworthy features of his early years: he attended public high schools in Tulsa, Oklahoma, where he was born, but had the great luck of participating in a six-year educational experiment. He writes about this: My public school education was more influential on my development than is often the case mainly because I was a participant in what is known as the Tyler eight-year study of the Progressive Education Association. On the basis of examinations given in the sixth grade, able students were selected to participate in a six-year experiment of accelerated education. In many respects the most competitive and ablest classes I ever attended were those in high school. One of the important aspects of this special educational venture was the extended attempt to get us as young students to talk about a wide range of current events and everything else that interested us.
A second important period in Pat's pre-philosophical life was his participation in the Second World War. After having majored in physics at the University of Tulsa he was called up in the Army Reserves in 1942 and then had a short break in which he received a BS degree in meteorology from the University of Chicago in 1943. He writes about this: Knowledge of meteorology has stood me in good stead throughout the years in refuting arguments that attempt to draw some sharp distinction between the precision and perfection of the physical sciences and the vagueness and imprecision of the social sciences. Meteorology is in theory a part of physics, but in practice more like economics, especially in the handling of a vast flow of nonexperimental data.
Pat then went on to three years of service in the Army Air Force in the South Pacific. We Europeans are immensely grateful to the United States and its soldiers, like Pat, whose participation in the Second World War was decisive for saving us from barbarism. This gratitude is, of course, fully compatible with our despair over USA's present involvement in Iraq and in the Israel-Palestine conflict. Particularly for us philosophers the kind of "evidence" that was presented as a reason for going into Iraq was utterly disturbing. However, now back to Pat and his career in Philosophy. After his discharge from the Army Air Force in 1946 he entered Columbia University as a graduate student in philosophy in January 1947 and received a PhD in 1950. He there was influenced by Ernest Nagel more than by any-
11
one else, but he also continued to take courses in areas outside philosophy, such as topology and group theory and also relativity theory, and together with some other students he organized a seminar on von Neumann and Morgenstern's theory of games, which at that time was not offered at Columbia. Pat writes about these excursions outside of philosophy: I thus developed early the habits of absorbing a wide variety of information and feeling at home in the problem of learning a subject in which I had not had much prior training or guidance.
Because of his background and interest in physics, Pat wanted to write his dissertation about the philosophy of physics, and he decided to write on the concept of action at a distance. A good deal of the dissertation was devoted to an analytical study of these concepts in the works of Descartes, Newton, Boscovich, and Kant. The part on Descartes was published in 1954, but some of the historical scholarship that went into the dissertation has later come out in other studies, notably in a 1967 article on Kant and in a long article on Aristotle's theory of matter, from 1974, in which Pat also reviews the theories of matter of Descartes, Boscovich and Kant. In the many seminars I have taught together with Pat I have again and again been struck by his broad and thorough knowledge of the history of philosophy as well as the history of science. I remember in particular a seminar we gave a year ago on Aristotle's De Anima and the medieval discussion of this topic. Pat there brought in highly pertinent material from sources that even the specialists on the Middle Ages who participated in the seminar had not heard of. These were sources that Pat had studied in his early years in college and which still stayed fresh in his mind. However, Pat is not primarily known as a historian of philosophy, so let us now turn to the core of his work. In this laudatio I will divide his work into six fields, with various subfields: 1. Methodology, Probability and Measurement 2. Psychology 3. Physics 4. Language and Logic 5. Computers and Education 6. Mind and Brain -----
12
1. Methodology, Probability and Measurement This has in many ways been Pat's main field, one to which he has contributed during the whole of his long career, from his first article, "A set of independent axioms for extensive quantities" in 1951 to his latest book Representation and Invariance of Scientific Structures for which he won the Lakatos award at the London School of Economics last year. This latest book is the culmination of several decades of work on set-theoretical structures in science and Pat there argues that these structures supply the right framework within which to investigate problems of representation and invariance in any systematic part of science. Also the problems mentioned by the President of the Swiss Science and Technology Council, Professor Susanne Suter, in her Address that we just heard, concerning measurement and evaluation in science policy and education, have been dealt with by Pat. 2. Psychology Psychology is the second of Pat's main fields of research, scientifically and philosophically. His contributions there cover such a vast range of problems that I can mention only the four main subfields to which he has contributed: learning theory, mathematical concept formation in children, psycholinguistics and behaviorism. Within each of these fields he has made important contributions, several of which have revolutionized the field. I will here only say a few words about the last of these four subfields, behaviorism, since here there are many misunderstandings both among philosophers and in the general public. "Behaviorism" is used as a label for many different views; one of these is ontological behaviorism, the view that there is behavior, but nothing mental. There are some philosophers who have this kind of reductionist position, but they are few and are getting fewer. Pat has never had this view. He is and has always been an evidential behaviorist. That is, he holds that we have no direct access to the mental states of others, for example through telepathy or mind-reading, but that the evidence against which we must test our theories of the mental is behavior, in addition to whatever evidence we can get from the biological sciences, through study of the brain and our neural networks. All this behavioral and biological evidence provides boundary conditions, which our philosophical theories have to satisfy.
13
3. Physics As mentioned, Pat's dissertation was on the concept of action at a distance. This was his very first contribution to the philosophy of physics, and many, many more have come, on a variety of issues. As could be expected, several of them relate to the two main developments in 20th century physics, relativity theory and quantum mechanics. Particularly the latter area has engaged Pat, mainly because the conceptual problems have been more challenging in that area. Pat regards conceptual clarification as one of the main contributions philosophers can make to the philosophy of science, and his solid grasp of probability theory and its foundations have been of great importance for his contributions, in particular to quantum mechanics and the entanglement of spatially separate particles. However, Pat, who has made many scientific contributions to the first two fields I surveyed, the theory of measurement and psychology as a science, is a little frustrated that he has not in the same way contributed to physics as a science. In his autobiography he tells us: I find that my own work here is less satisfying to me than other areas I discuss later, just because I do not anticipate making a scientific contribution to the subject. In the case of the theory of measurement or psychology, for example, my contributions have as much a scientific as a philosophical flavor, and I find that this suits my temperament better -- independent of whether the scientific contribution is of greater or less significance than the philosophical one.
This may be Pat's judgment. However, most philosophers of physics would be more than happy and immensely proud for having brought about as much clarity and insight as Pat has done in physics. 4. Language and Logic Pat's book Introduction to Logic in 1957 was his first work in this field. It was followed by another well known text book, Axiomatic Set Theory in 1960. In that year came also his first little article on language: "Problem analysis and ordinary language." Many more have followed. Fortunately, his main articles on language have been collected in a volume, Language for Humans and Robots in 1992. Some more articles have followed. As in the case of psychology, it is natural to group Pat's work on language into
14
four areas, the first of which overlaps with his work in psychology: First the psychology of language, particularly children's acquisition of language and children's use of language. Secondly, work of a formal linguistic kind, partly contributions to linguistics as a science. Thirdly, there is work on language and robots, and fourthly, Pat's ideas on congruence of meaning. In the first of these fields, children's language, Pat has carried out empirical studies which I think many of you will find fascinating, where he gives close analyses of actual verbal communication between children in non-experimental setting. It is characteristic of Pat that he does not merely present data but analyzes them in such a way as to bring out their philosophical aspects. In this case, he shows how semantics seems to be much more important for the acquisition of language than the syntactical features of language. I will briefly mention the fourth of the subthemes, congruence of meaning. This is Pat's contribution to one of the main issues in 20th century philosophy: What is meaning? Bolzano, Frege and most of the major philosophers of the twentieth century have addressed this problem, and this was also one of Henri Lauener's main concerns, particularly in his book on Quine and in various articles. (One of Lauener's articles on this topic has been reprinted in a recent collection of the best articles on Quine published during the last 50 years.) Pat proposes that we should compare the notion of meaning in language with the notion of congruence between geometrical figures. As is his wont, he works this idea out with great precision and definitely makes a good case for his conclusion, that there is a large number of congruence relations that reflect the regularities that enable us to establish, learn and use language. 5. Computers and Education In the summer of 1983 an innovative software product called "Dial a Drill" hit the stores. The product was released by a company called Computer Curriculum Corporation, or short CCC. "Dial-A-Drill" was an automatic recording, over the phone, that the students (or adults) could pick up at appointed times. They would then hear a computerized voice that would put them through reading, spelling and arithmetic drills, which they would an-
15
swer by pushing buttons on the phone's keypad. As students recorded their answers, the phone-voice responded with hints and words of praise: "Excellent work!" The program also adjusted the drill as the phone call progressed, delivering harder problems for skilled children or easier one for those who were struggling. That made Dial-A-Drill "computer adaptive," as testing experts call it, a feature that should become a rage many years later. Computer Curriculum Corporation, or CCC, was basically Pat's firm. He founded it, with several partners, in 1967, and it was based on his ideas and research in education and psychological learning theory. CCC has brought out lots of other programs, mostly computer based, and mostly directed at those parts of learning that require drill, that is, repeated exercises with explanations and feedback. Pat never intended his computerized education to replace teachers. Nothing can replace good teachers, he said, it should replace bad teachers, or help students who had no teachers to get feedback on what they were doing. Pat's educational tools have been, and are, of special value in areas where there are no, or very few, teachers, and hardly any well educated teachers, such as in the developing countries and in the slums. Pat has done much educational work, using these tools, in Ghana, Brazil and other countries in Africa and South America. And I find it significant that "Dial-ADrill", was sold for 15 to 18 dollars with its full implement of three to five telephone calls a week and monthly reports in the mail to the students and their parents. 6. Mind and Brain This is the latest field which Pat has taken up for serious research. Of course, the problems of mind have always been on his mind, in his work in psychology, and also in his work on language and on education, and we have given several seminars together on the mind and on its relationship to the brain. This fall, we will be giving a seminar on Mind and Brain, together with prominent biologists working on the brain. However, Pat himself can take care of much of the biology. Ten years ago, he started to do experimental work on the brain to find out exactly what happens in the brain when we use language. Some of the early experiments he carried out at the Scripps Institution in LaJolla, California, because that was the only
16
place where they had instruments with the kind of spatial and temporal resolution that Pat needed in order to test his hypotheses. One of these hypotheses that was quickly confirmed was that the activities in the brain happen so noisily and quickly that they had not been spotted by the earlier efforts. Pat's experiments are now being carried out at Stanford, and Pat has been able to test and confirm several hypotheses about what happens in the brain when we use language, such as brain-wave recognition of words, sentences and simple visual images and their names. Much of his work on this topic is published in the Proceedings of the National Academy of Sciences, of which he has been a member since 1978. He is one of only two philosophers who are members of this prestigious academy. -----------------This brings me, finally, to the many honors that have been bestowed on Pat over the years, of which today's Lauener Award is the latest. I have already mentioned the Distinguished Scientific Contribution Award of the American Psychological Association, but Pat has received several similarly noteworthy awards within the other fields in which he has been working, including several awards in education. The most distinguished of all his awards, the National Medal of Science, he received in 1990. This prize is restricted to natural and social science and even among Nobel Prize winners only a few have received it. Pat was the third social scientist to get this medal. Not bad for a humanist. Pat is a member of numerous scientific academies, including the illustrious American Philosophical Society, America's oldest academy, founded by Benjamin Franklin in 1743. This is not, as the name might suggest, an academy for philosophers, but an academy for researchers from all fields, with only a handful of philosopher among its selected half hundred members. Pat is one of them. As could be expected, Pat has honorary doctorates from universities in many different countries. What could not be expected by a person who has been so productive in so many fields, is that Pat has also served in a large variety of time-consuming positions: he was chair of philosophy at Stanford in the sixties (and I am grateful that he thereby was the one who brought me to Stanford). He has also served as Dean at Stanford – but only briefly and reluctantly, since he would rather concentrate on his research. He has been president of the American Philosophical Association, the International Union of History and Philosophy of Science and many other
17
national and international organizations. In the midst of all this and a magnificent life of research Pat has also, with his left hand, founded and directed his very successful firm for computer-assisted instruction, which I mentioned earlier, Computer Curriculum Corporation. Like Henri Lauener, Pat has wanted to use his assets to support good research, and he has already given Stanford several endowed chairs, in addition to establishing a Center for the Interdisciplinary Study of Science and Technology, which will be opened on November 1. Pat therefore does not need the Lauener prize, which he is receiving today. But he very highly deserves it. I regard him as the best choice the Prize Committee could make. In establishing a new prize, it is important to find prize winners who can give glory to the prize. Gradually, the prize can thereby more and more give glory to those who received it. I will therefore congratulate the Lauener Foundation with its choice of prize winner. And I will congratulate Pat with the prize!
18
A Revised Agenda for Philosophy of Mind (and Brain) Patrick Suppes∗
It is my intention in this article to set forth the case for revision of the topics considered central in the philosophy of mind and to reorient that agenda toward concepts that are more present in current research in psychology and neuroscience. In my view the philosophy of mind should be a subject like the philosophy of physics, dominated in many ways by scientific findings, in this case, in psychology and neuroscience about the nature of the mental. This does not mean there is no room for philosophy left. Rather, its role is changed to the workman-like job of building a conceptual foundation for the scientific results. The building of such a foundation is itself as much scientific as philosophical. What marks it as philosophical is the emphasis on a certain range of concepts, some of which may remain controversial and will not be clarified for some decades by proper theoretical and empirical scientific findings. I have divided this revised agenda into three parts. The first one deals with computation, perhaps the subject most missing from current philosophical theories of mind. Part two focuses on representation. For reasons that are set forth later, the concentration is on brain rather than mental representations. Finally, the ever present and continually controversial topic of consciousness is looked at anew in the third part, along with habits and their automaticity. I. Computation Preliminary remarks on language. A major area of conflict in thinking about mental computation are ideas about language. From the standpoint of much empirical research, which I believe is broadly correct, the computational processes of comprehension or production are almost entirely unconscious in nature. A particularly telling and important example is that of the prosodic features used by a speaker. In so far as these features express anger, fear, contempt, and so forth, they are often more evident to listeners ∗
Stanford University
than the speaker, even after they occur. This is a familiar observation, especially about anger. For many features of speech, we are aware of the results, but not the processes, by which they are produced or comprehended. This will be a general thesis about consciousness discussed later in Part III—awareness of the results, but seldom detailed awareness of the processes themselves. A radical thesis is that detailed linguistic theories of parsing and generating grammars are mostly wrong as descriptions of brain processing of language. The mental view is that processing is sequential and often conscious in character. In actual fact, most philosophical treatments of the philosophy of mind do not really have a great deal of detailed analysis of this processing. The brain-computation view, in any case, is that processing is massively parallel and mainly unconscious, entirely so in the details. This general thesis I will not defend carefully at this point, but will amplify as I turn to more concrete questions. What I said about parsing applies also to determining meaning or, what I prefer, truth. The folklore account is that we parse the sentence and then determine its meaning. The brain-computation view is that the processing happens nearly simultaneously and in parallel. We have no awareness at all of how we compute the truth value, just as we do not have awareness of how we compute the grammatical correctness of an utterance. So, if I give persons sentences they have not heard before about the geography of Europe, they will be able to tell me rather quickly, within about a 150 milliseconds after the end of each sentence, whether it is true or false. Typical examples are Paris is not the capital of Poland and Rome is north of London. I return to these broad remarks later in talking in a more detailed way about computation, but first I want to turn to some philosophical views of the mental representation of language and their relation to computation. Mental representations. Here is a quote from Jonathan Lear about Quine and Wittgenstein with which to begin, Quine, like Wittgenstein, categorically rejects the notion that meaning can essentially involve anything private to an individual, such as a hidden mental image. This is the myth of the museum—that words name specimen mental objects—which Quine rightly urges us to reject. If we are to explain languagemastery, and thus the meaning our words and sentences have, we must do it on the basis of our experience: the sensory evidence of all types to which we have over time been exposed and our sensitivity to it. Positing interior mental objects that are named by words only gets in the way of an explanation, for it
20
merely papers over the gaps in our understanding of how language-mastery is acquired. (Lear, Going Native, pp. 177–78)
Now let us turn to the contrasting view of Noam Chomsky. As I am using the term, knowledge may be unconscious and not accessible to consciousness. It may be “implicit” or “tacit.” No amount of introspection could tell us that we know, or cognize, or use certain rules or principles of grammar, or that use of language involves mental representations formed by these rules and principles. We have no privileged access to such rules and representations. This conclusion appears to be inconsistent with requirements that have been widely enunciated. Kant, for example, insisted that “All representations have a necessary relation to a possible empirical consciousness. For if they did not have this, and if it were altogether impossible to become conscious of them, this would practically amount to the admission of their nonexistence.” Similar doctrines are familiar in contemporary philosophy. John Searle writes that “It is in general characteristic of attributions of unconscious mental states that the attribution presupposes that the state can become conscious, and without this presupposition the attributions lose much of their explanatory power.” (Chomsky, Rules and Representations, p. 128)
My summary view is that, in historical order, Kant, Wittgenstein, Quine, Searle, and Lear are wrong. Chomsky is right. Our conscious mental representations of language are poorly developed, especially almost all aspects of processing, i.e., computation. As in other areas of the mind, we are conscious of results not mental processes, a point I have stressed earlier and will do so later. On both sides that I have quoted, what is missing, of course, are detailed answers. The views of Lear are particularly mysterious. To turn back on him his own concept of mystery, what in the world is his theory of how language is processed? Surely, there is something going on in the brain. It is not just a matter of external sensory experience. The absence of any conjectures whatsoever about how syntactic and semantic computations are made by the brain is the most striking feature, in fact, of what Wittgenstein and Lear have to say. Quine, who is more sensible than either of them, hints in various accounts of stimulus meaning and the like at the psychology, if not the neuroscience, of language learning. Chomsky also has detailed things to say about what the rules must be like and he is in favor, generally, of a biological view toward language. It is just that he has not
21
ventured into the psychology or neuroscience of such matters in much depth. Problem of the computation of truth. I turn now to an important specific case of computation. One of the scandals of both philosophy and linguistics, as well as psychology, is the absence of any detailed theory of how the most elementary empirical sentences are judged true or false, or to put it more directly, how their truth value is computed. Consider atomic sentences. Tarski's semantic theory of truth offers no help in determining their truth. He was not concerned to give a theory of how we compute or know that individual atomic sentences such as Paris is north of Rome are true or false. Computation of their truth is not generally familiar; I have introduced the subject for deliberate, and indeed necessary, purposes here. There are only one or two views practically possible. The truth or falsity of such sentences is computed from the knowledge and beliefs an individual is able to construct—and I say ‘construct’ not ‘store’—to emphasize processing, or the truth value is simply a mysterious act of the mind beyond comprehension. Obviously, the latter view is absurd. But how are such elementary sentences computed? There are the outlines of a theory in parts of psychology. I shall develop these ideas here without entering into extensive technical details. It is possible that some, perhaps even many, philosophers who read what I have just written will think they remember that the problem of the computation of truth actually was solved in a nearly satisfactory way by Tarski in his famous article of 1936 “The concept of truth in formalized languages”. Everyone who may have forgotten the details will still remember the famous criterion of truth illustrated by: ‘it is snowing’ is a true sentence if and only if it is snowing, which, as Tarski (1936/1983, p. 156) remarks, we are assuming is the case. On the preceding page Tarski states a familiar criterion that, as he says, is more or less that of the classical view of the truth, “a true sentence is one which says that the state of affairs is so and so, and the state of affairs indeed is so and so”. In this connection, he quotes also the earlier source of a very similar remark, well known in philosophy, of Aristotle in the Metaphysics (1011b27), “to say of what is that it is or of what is not that it is not, is true”. Intuitively put, Tarski's task is not at all to investigate what is the state of affairs. His concern is with the recursive definition of truth, when the truth of atomic sentences is given. So, if the truth of atomic sentences is known, we can then compute the truth of any complicated sentence recursively from that of the atomic sentences. What I am saying is not meant to be a technical characterization of
22
Tarski's recursive definition, which explicitly holds only for certain formal languages, but rather to give a feeling for it. It is the truth value of the empirical atomic sentences that is in no sense touched by his characterization, and he understood this perfectly well. It is also not my task here to actually investigate the empirical truth of some indefinite number of atomic sentences. It is rather different. It is how we use our beliefs, our memories, and our associations of other kinds to compute, from this basis, the truth of many different empirical atomic sentences about familiar phenomena. Certain ones we are expected to compute easily, even if we have never heard them uttered before. Such simple geographic examples are what I consider here. We could just as well pick familiar political events of the last twentyfive years. We would also expect to get very quick answers, without any reference to outside sources of knowledge. So, as you can see, what I want to turn to is the psychological theory of how the mind, i.e., brain, of a person computes the truth of new sentences, when the truth should be evident from things already known by that person. Preliminary remarks on association. The investigation is simply of the brain computations, or if you wish, mental computations, needed to compute the truth value of a sentence that has probably not been previously heard or read. A more detailed formulation of ideas is given in Suppes and Béziau (2004), but the supporting neuroscientific literature is not cited there either, and many gaps remain in our scientific understanding of the neural processing. The general ideas I use are not new. They derive from Aristotle and Hume, the Godfathers of what I consider a sound philosophy of the mind. In the direct matter of computation I will refer to Hume, and delay consideration of Aristotle until the next section on representation. Psychological and neural methods of computation are the focus. I will use both kinds of description, because for me they are one and the same. What is psychological is embodied in the brain. What is claimed to be psychological and not embodied in the brain is only fantasy. But often we have not yet discovered exactly how a given psychological process is realized in the brain, and so we have to leave open the neural details. Hume was the first to state so absolutely clearly and unequivocally that there are really just three basic mechanisms of computation in the mind— of course, he did not use the word ‘computation’. The classic formulation is to be found in Section 4 of Book I of the Treatise of Human Nature (1739/1951). He says at the beginning that the faculty of imagination, must
23
be guided by some universal principles. These are the mechanisms of association. The qualities, from which this association arises, and by which the mind is after this manner convey'd from one idea to another, are three, viz. RESEMBLANCE, CONTIGUITY in time or place, and CAUSE and EFFECT. (Hume, Treatise, p. 11)
What is remarkable about this passage is that not only does Hume put his finger on the principal role of association, he also has an excellent hypothesis of what are the main associative mechanisms. One perception of an apple seeming so much like another gives rise to forming the idea of an apple; constant contiguity between fire and smoke, leads to a causal association. Hume goes on to say that of the three, the recognition of cause and effect, that is, that from a certain occurrence there follows another event, which we recognize as that of effect, is the most extensive, and in many ways the most important. I emphasize this remark about cause and effect, for it is sometimes mistakenly thought that Hume was a skeptic about causal notions, because he denied, in contradistinction to Locke and Newton, their necessary character. He was also rightly constrained in recognizing that it is not possible to go beyond a certain depth without additional kinds of knowledge that he did not have. Here is the quotation on this topic, that I like the most. He properly characterizes it as part of the nature of a true philosopher to restrain the intemperate desire to continue searching for causes when he is only led into “obscure and uncertain speculations”. These are therefore the principles of union or cohesion among our simple ideas, and in the imagination supply the place of that inseparable connexion, by which they are united in our memory. Here is a kind of ATTRACTION, which in the mental world will be found to have as extraordinary effects as in the natural, and to shew itself in as many and various forms. Its effects are every where conspicuous; but as to its causes, they are mostly unknown, and must be resolv'd into original qualities of human nature, which I pretend not to explain. Nothing is more requisite for a true philosopher, than to restrain the intemperate desire of searching into causes, and having establish'd any doctrine upon a sufficient number of experiments, rest contented with that, when he sees a farther examination would lead him into obscure and uncertain speculations. (Hume, Treatise, p. 12–13)
24
It is also important to catch the reference to attraction, because Hume felt, in many ways properly, that the role of association in the life of the mind was as dominate and as significant as that of gravitation in the motion of physical objects. The first edition of Hume's Treatise was published in 1739. So now more than 250 years later, we have more elaborate things to say, but in spite of qualms about association in various intervening periods, it has now returned to its proper place in the theory of the brain. Well noticed is the large association area of the human cortex, in comparison even to other primates. Associative networks are being developed by scientists from many disciplines, and for a variety of purposes. Mistakenly, some philosophers and psychologists, only a few decades ago, would have scoffed at the idea that any serious computations of any kind could be made by principles of association. But this is simply a conceptual mistake, well recognized in the general theory of computation. There are now a number of papers showing the computational power of associative networks and it is easy to prove that we can simulate with an associative network a universal Turing machine capable of computing any computable, i.e., any partial recursive, function, in the standard theory of computation. This universal Turing machine need in principle not be large. One of the best examples, even after a good many years, is the one introduced by Marvin Minsky (1967) consisting of four output symbols and seven internal states. So that using the commonly applied measure of the product of the number of symbols and the number of states, Minsky’s universal Turing machine has the number 4 x 7 = 28. Not quite the best, but almost as good as any that can be reached according to present arguments and examples. I will not go through the details of how such a Turing machine can be simulated by an associative network. It is a familiar fact of modern computational theory that rather simple devices, including Minsky’s machine, are quite adequate for computing anything in principle. This does not mean they would serve as practical computers. In like fashion, it is quite another matter to have a deep and detailed understanding of the computations actually made by the brain. My next task here is much simpler. I propose, schematically, how we can, using psychological and neural concepts, compute from associations the truth of obvious empirical sentences. Associative networks. The remarks thus far are a prolegomena to giving a sketch of the theory of how such ordinary computations of truth are made. The basic idea is that the computations are made by an associative network, with brain representations of words being the nodes and the links
25
between the nodes being the associations. More generally, auditory, visual, and other kinds of brain images can also be nodes. There is a reasonable body of evidence to support the hypothesis that the nodes of the network are collections of synchronized neurons. In the initial state, not all nodes are linked, and there are, in this simple formulation, just two states, quiescent and active. No learning or forgetting is considered. It is assumed that, after a given sentence is responded to as being either true or false, all the activated states return to quiescent. The axioms, which are not stated here, are formulated just for the evaluation of a single sentence, not for giving an account of how the process works over a longer slice of discourse. The way to think about the networks introduced is that a person is asked to say whether a sentence about familiar phenomena is true or false. It is very natural to ask, and not to have a quibble about ‘Do you believe this, even though you don't know whether it is true?’ The sentence input comes from outside the associative network in the brain. I will consider only spoken words forming a sentence, although what is said also applies to visual presentation as well. So, as the sentence is spoken, the sound-pressure image of each word that comes to the ear is drastically transformed by a sequence of auditory computations leading to the auditory nerve fibers, which send electromagnetic signals to the cortex. Such signals are examples of those mentioned earlier. In previous work, I have been much concerned with seeing if we can identify such brain signals as brain representations of words. Some references are Suppes, Lu, and Han (1997) and Suppes, Han, Epelboim, and Lu (1999a, 1999b). The brain activates quiescent states by using the signal brought into the cortex as the brain representation of the verbal stimulus input. With the activation of the brain representation of words by external stimuli, the associations between activated brain representations are also activated by using this same signal. Moreover, it is assumed in the theory that activation can be passed along from one associated node to another by a phenomenon characterized some decades ago in psychological research as spreading activation. For example, in a sentence about a city like Rome or Paris, some familiar properties are closely associated with these cities and the brain representation of these properties may well be activated shortly after the activation of the brain representations of the words Rome or Paris, even though the names of these properties, or verbal descriptions of them, did not occur in any current utterance. This is what goes under the heading of spreading activation. Some form of it is essential to activate the nodes and links needed
26
in judging truth, for, often, we must depend upon a search for properties, which means, in terms of processing, a search for brain representations of properties, to settle a question of truth or falsity. A good instance of this, to be seen in the one example considered here, is the one-one property, characteristic of being a capital: x is capital of y, where x is ordinarily a city and y a country. There are some exceptions to this being one-one, but they are quite rare and, in ordinary discourse, the one-one property is automatically assumed. But this is only one of many examples, easily given, that arise in ordinary conversation. (For computer-science applications of spreading activation to information retrieval, see Crestani (1997) for a critical survey.) One other notion introduced in the axioms of Suppes and Béziau (2004) for computing truth is the concept of the associative core of a sentence, in our notation, c(S) of a sentence S. For example, in the kinds of geography sentences given in the experiments referenced above, where similar syntactic forms are given and the sentences are given about every four seconds, persons apparently quickly learn to focus only on the key reference words, which vary in an otherwise fixed sentential context, or occur in a small number of such contexts. So, for example, the associative core of the sentence Berlin is the capital of Germany is a string of brain representations of the three words Berlin, capital and Germany, for which I use the notation BERLIN/CAPITAL/GERMANY, with, obviously, the capitalized words being used to denote the brain representations. A more complicated concept is needed for more general use. In the initial state of the network, associations are all quiescent, e.g., PARIS ~ CAPITAL, and, after activation, we use the notation PARIS ≈ CAPITAL. In the example itself, we show only the activated associations and the activated nodes of the network, which are brain representations of words, visual or auditory images, and so forth. The steps of the associative computation are numbered in temporal steps t1, etc., which are meant to include some parallel processing.
27
Example. Rome is the capital of France. t1. ROME, CAPITAL, FRANCE t2. PARIS, 1–1 Property t3. ROME ≈ CAPITAL, CAPITAL ≈ 1–1 Property
Activation Spreading activation Activation
CAPITAL ≈ FRANCE, PARIS ≈ CAPITAL, PARIS ≈ FRANCE t4. ITALY t5. PARIS/CAPITAL/FRANCE ROME/CAPITAL/ITALY t6. TRUE ≈ PARIS/CAPITAL/ FRANCE
Spreading activation Activation Spreading activation
TRUE ≈ ROME/CAPITAL/ITALY t7. FALSE ≈ ROME/CAPITAL/FRANCE
Spreading activation
This sketch of an example, without stating the axioms and providing other technical details, is meant only to provide a limited intuitive sense of how the theory can be developed for simple empirical sentences. Most important, there is no account of learning associations. Only an idealized performance setup is considered. II. Representation A central topic in the philosophy of science is the analysis of the structure of scientific theories. Much of my own work has been concerned with this topic, but in a particular guise. The fundamental approach I have advocated for a good many years is the analysis of the structure of a theory in terms of the models of the theory. In a general way, the best insight into the structure of a complex theory is by seeking representation theorems for its models, for the syntactic structure of a complex theory ordinarily offers little insight into its nature. In attempting to characterize the models of a theory, the notion of isomorphism enters in a central way. Perhaps the best and strongest characterization of such models is expressed in terms of a significant representation theorem. By such a theorem for a theory the following is meant. A certain class of models of the theory, distinguished for some intuitively clear
28
conceptual reason, is shown to exemplify within isomorphism every model of the theory. More precisely, let M be the set of all models of a theory, and let B be some distinguished subset of M. A representation theorem for M with respect to B would consist of the assertion that given any model M in M there exists a model in B isomorphic to M. In other words, from the standpoint of the theory every possible variation of model is exemplified within the restricted set B. It should be apparent that a trivial theorem can always be proved by taking B = M. A representation theorem is just as interesting as the intuitive significance of the class B of models, and no more so. An example of a simple and beautiful representation theorem is Cayley's theorem that every group is isomorphic to a group of transformations. One source of the concept of a group, as it arose in the nineteenth century, comes from consideration of the one-one functions which map a set onto itself. Such functions are usually called transformations. It is interesting and surprising that the elementary axioms for groups are sufficient to characterize transformations in this abstract sense, namely, in the sense that any model of the axioms, i.e., any group, can be shown to be isomorphic to a group of transformations. Philosophical views of mental representations. Without attempting anything like a detailed and accurate account of the long and complicated history of the concept of mental representation in philosophy, it can be enlightening and relevant to review some of the standard conceptions and issues from Aristotle onward. In fact, here I mainly restrict myself to the Aristotelian and Humean traditions. (For more details, see Suppes (2002, Ch. 3).) The main point will not be to assess the correctness or to criticize the adequacy of the analysis given, but to reflect on whether or not there is a notion of isomorphism in the background, as reflected in such concepts as likeness or resemblance. Comment will also be made as to how such notions are tied to representations. Aristotle. That the defining feature of sense perception is receiving the form of a physical object is, in general terms, the view of perception for which Aristotle is most famous. As he works out the details, it is not an unreasonable view at all, even though it is quite obvious from a modern viewpoint that it cannot be entirely correct. The background of Aristotle's discussion of perception and implicitly, therefore, of mental representation, is his distinction between form and matter, which applies to objects and phenomena of all kinds. For example,
29
the form of an axe is different from the matter that receives that form. This distinction is also relative in the sense that, for example, the matter of a house can be made up of bricks, which, in turn, have their own form and matter, of which they are made up. I defended the continued viability of this general concept of matter in Suppes (1974). Aristotle makes important use of the distinction between form and matter in his theory of perception. It is the role of the sense organ to have the potential to receive the form, but not the matter, of a physical object, as described in this passage from the second book of the De Anima: We must understand as true generally of every sense (1) that sense is that which is receptive of the form of sensible objects without the matter, just as the wax receives the impression of the signet-ring without the iron or the gold, and receives the impression of the gold or bronze, but not as gold or bronze; so in every case sense is affected by that which has colour, or flavour, or sound, but by it, not qua having a particular identity, but qua having a certain quality, and in virtue of its formula; (2) the sense organ in its primary meaning is that in which this potentiality lies. (Aristotle, De Anima, 424a17–424a25)
So, if in perceiving the candle we receive exactly its form, but not its matter, then we have a representation of the physical candle that is isomorphic to it in the mind, just because of the sameness of form. The relevant point here is that this notion of the sameness of form corresponds very closely to the concept of isomorphism that I have been arguing for as a central concept of representation. The concept of form catches nicely that of isomorphism, in the sense that the form does not thereby refer to all the properties of the candle, but only to the properties perceived by the various senses of sight, touch, etc. This kind of restriction of the properties considered is, as already noted in earlier discussions, characteristic of any workable notion of isomorphism. Upon reading the above passage and my comments, someone might reflect that this is not a very rich theory of the mental, but only of perception. Aristotle, however, goes on to extend the same ideas about forms to the intellect. It is not possible here to discuss all the subtleties involved in his worked-out theory, but the following passage makes clear how the part of the soul that thinks and judges operates in the same way with forms and, thus, the characteristic notion of isomorphism is again applicable: Concerning that part of the soul (whether it is separable in extended space, or only in thought) with which the soul knows and thinks, we have to consider
30
what is its distinguishing characteristic, and how thinking comes about. If it is analogous to perceiving, it must be either a process in which the soul is acted upon by what is thinkable, or something else of a similar kind. This part, then, must (although impassive) be receptive of the form of an object, i.e., must be potentially the same as its object, although not identical with it: as the sensitive is to the sensible, so must mind be to the thinkable. …Hence the mind, too, can have no characteristic except its capacity to receive. That part of the soul, then, which we call mind (by mind I mean that part by which the soul thinks and forms judgments) has no actual existence until it thinks. So it is unreasonable to suppose that it is mixed with the body; for in that case it would become somehow qualitative, e.g., hot or cold, or would even have some organ, as the sensitive faculty has; but in fact it has none. It has been well said that the soul is the place of forms, except that this does not apply to the soul as a whole, but only in its thinking capacity, and the forms occupy it not actually but only potentially. (Aristotle, De Anima, 429a10-18, a23-30)
It is a definite part of Aristotelian thought that the forms do not exist separate from individual bodies. There is no separate Platonic universe of forms. A very clear statement on this Aristotelian view is made by Aquinas in the following passage. I say this, because some held that the form alone belongs to the species, while matter is part of the individual, and not of the species. This cannot be true, for to the nature of the species belongs what the definition signifies, and in natural things the definition does not signify the form only, but the form and the matter. Hence, in natural things the matter is part of the species; not, indeed, signate matter, which is the principle of individuation, but common matter. For just as it belongs to the nature of this particular man to be composed of this soul, of this flesh, and of these bones, so it belongs to the nature of man to be composed of soul, flesh, and bones; for whatever belongs in common to the substance of all the individuals contained under a given species must belong also to the substance of the species. (Aquinas, Summa Theologica, I, Q.75. Art. 4)
The discussion and citations given do not adequately portray the rich tradition of Aristotelian psychology. Aristotle’s De Anima and his many commentators, including important Arabic ones, but especially Aquinas, established a long intellectual tradition still vigorously reflected in such seventeenth-century works as Descartes’ Passions of the Soul (1649/1927). It has not been sufficiently remarked that this tradition of Aristotelian psychology, to label it in modern terms, is in its own way comparable to the
31
glorious Ptolemaic tradition in astronomy, which only ended with the Astronomia Nova (1609) of Kepler. Such clear and detailed expositions of the De Anima as Themistius’ “Paraphrase” (1996) written in late antiquity (fourth century) were still widely read, in the Renaissance. Indeed, Aquinas’ commentary formed the basis of Thomistic psychology, still actively studied as a systematic subject in the twentieth century (Brennan, 1941). Hume. On various grounds, in the Treatise of Human Nature, Hume argues vigorously against any direct concept of mental representation between objects in the external world and what is in the mind. He has many things to say about these matters. The first point to emphasize has already been remarked on in the preceding section. The mechanism of the mind for Hume is association. The important point relevant to the discussion here is the pride of place that Hume gives to resemblance, since only it is a quality, as he puts it, that holds between ideas. And, of course, resemblance is exactly Hume's notion close to that of the modern notion of isomorphism as a basis of representation. …'Tis plain, that in the course of our thinking, and in the constant revolution of our ideas, our imagination runs easily from one idea to any other that resembles it, and that this quality alone is to the fancy a sufficient bond and association. (Hume, Treatise, p. 11)
I said earlier that Hume denies a direct resemblance between fragmentary perceptions and the real objects posited to generate those perceptions, but he does defend a realistic notion of physical objects and their continued identity in time. His long and complicated argument about these matters is one of the most substantial parts of Book I of his Treatise. I want to point out various ways in which he uses the concept of resemblance or isomorphism. The following passage is particularly striking, because he argues for the constancy of our perception of something like the sun or other very stable objects by using the fact that our successive perceptions, though interrupted, resemble each other and justify, therefore, as he argues in detail, the inference to identity across time of the object. Here is the key passage about this use of resemblance. When we have been accustom'd to observe a constancy in certain impressions, and have found, that the perception of the sun or ocean, for instance, returns upon us
32
after an absence or annihilation with like parts and in a like order, as at its first appearance, we are not apt to regard these interrupted perceptions as different, (which they really are) but on the contrary consider them as individually the same, upon account of their resemblance. But as this interruption of their existence is contrary to their perfect identity, and makes us regard the first impression as annihilated, and the second as newly created, we find ourselves somewhat at a loss, and are involv'd in a kind of contradiction. In order to free ourselves from this difficulty, we disguise, as much as possible, the interruption, or rather remove it entirely, by supposing that these interrupted perceptions are connected by a real existence, of which we are insensible. This supposition, or idea of continu'd existence, acquires a force and vivacity from the memory of these broken impressions, and from that propensity, which they give us, to suppose them the same; and according to the precedent reasoning, the very essence of belief consists in the force and vivacity of the conception. (Hume, Treatise, p. 199)
Hume does not stop here. Over several pages he explains four different aspects of this setup in the inference from interrupted perceptions to identity of objects through time. It is one of the most sustained arguments in the Treatise and a wonderful piece of philosophical analysis. He first explains the basis for the principle of the individuation of objects, and he introduces here something that is very much in the spirit of isomorphism and a relevant notion of invariance. Hume’s ideas dominated much of the psychology of the nineteenth century, to which even William James agreed in his influential treatise Principles of Psychology (1890). James was quite critical of Hume’s use of simple ideas, out of which to build complex ones. He objected to such as atomistic concept as central to the nature of thought. Brain representations of language: general remarks. When a person hears the word Paris or reads the word Italy, usually in a sentential context, the encoded word token reaching the cortex must be recognized, in some way or another, as isomorphic to a word already encoded in longterm memory. Of course, none of this process of recognition is conscious in the usual settings in which we listen or read. Given that this description is roughly correct, a first problem of research is to find, if possible, brainwave representations of the initial processing of verbal stimuli. There is a substantial tradition of studying components of these waves, usually called evoked response or event-related potential (ERP) components, generated by various verbal stimuli, e.g., visually presented sentences with semantic anomalies. For reviews of this literature, see Brown and Hagoort (1999), and Rugg and Coles (1995). The task considered here is different, namely,
33
identifying particular waves as the token representations in the brain of auditorily or visually presented words or sentences. From a theoretical or statistical standpoint, this is a problem of correct classification or recognition of brain waves, which we observe at the surface of the skull by electroencephalographic (EEG) recordings of the electric field. The concept of isomorphism of two models of a theory is one of the most basic notions used in many parts of mathematical psychology. It has been prominent in the extensive development of the measurement theory of many different psychological phenomena. But the notion has a much wider sweep throughout mathematics and many parts of physics. The reason for this widespread application is easy to explain. An isomorphism captures the concept of two different models, or complicated sets of data, having the same structure with respect to a given set of concepts. The first point to be emphasized is that the sameness of structure is not meant to be the relation of identity for the entities in question, which is trivial, both from a mathematical and scientific standpoint. What is important is that the notion of same structure captures that part or, better, the properties or features of the objects or phenomena under consideration, that are dealt with theoretically or experimentally. In experiments, features that are not considered relevant to the analysis of same structure are, for example, the names of the experimenters or the dates of birth of the experimenters, or even the exact day on which an experiment took place. At a more fundamental level, in classical particle mechanics, to give an example from physics, the color of particles is not part of the mechanical analysis of structure, even though the concept of color is important in other parts of physics. A relation of isomorphism must be an equivalence relation, that is, reflexive, symmetric and transitive, but for ordinary mathematical purposes, a good deal more is necessary. If we just required an equivalence relation, then any one-one mapping of one set onto another would provide an isomorphism. What is required for most structures is not only the mapping of one domain onto another, but also the preservation in the obvious sense, familiar from many examples, of the operations, relations and fixed objects that are part of the structure. Sometimes the mapping from one domain to another is not a one-one function, but a many-one function, giving rise to what is called, in the theory of measurement, a homomorphism. Familiar examples are mappings of empirical structures into numerical ones, where distinct empirical objects or phenomena are assigned the same number. Thus, we have a homo-
34
morphism from the empirical phenomena to the numbers representing measurements. The concern here is the existence of an isomorphism between constituents of language and the processes in the brain that are postulated to represent them. So, for example, we can start at the word level and look for the representation of each word in the brain. Already an ambiguity is present, one that is also present in the representation of words outside the brain. When we say a word, we have a process representation of that word. When we record it on a CD, we have a nonprocess representation that has the potential of giving us such a representation when we play the CD. We make this distinction in the brain as well, even if we do not have a settled view of what corresponds to the CD bit representation of words in the brain. There can even be disagreement as to how words are initially processed in the brain. One view is that words are processed electrically in the axons and synapses of neurons, whereas another is that they are processed in the electromagnetic field generated by the neurons. I will not get into this controversy here. In the experiments mentioned later, we used recordings of the electric field obtained from EEG sensors. I do emphasize that we are observing the electric field, not in any sense directly observing electric currents in individual axons or synapses. For many kinds of concrete objects or phenomena, we must use a quantitative measure of fit between two objects—for example, in the fit between a prototype and a test sample. This leads to a notion of similarity, as a generalization of the concept of isomorphism. This relation is, in general, not transitive. Similarity is not a new concept in psychology. It has been widely studied since the nineteenth century. Even so, it continues to be a source of formal problems. Transitivity is only the beginning of the troubles. In the ordinary concept of isomorphism and the preservation of structure, we have substitution relations for parts of a given object or phenomenon. But these straightforward substitutions, under a standard concept of congruence or isomorphism, are much more complicated when the familiar thresholds of psychological phenomena disturb the transitivity of the equivalence and the isomorphism as well. Isomorphism of language structures. We first characterize contextfree elementary language structures for the restricted purpose of formulating our hypothesis of structural isomorphism for constituents of language and their brain representations. Weaker, context-dependent structures are introduced later.
35
We define an elementary segment as a pair (T,f), where T is a halfopen, half-closed interval [b,e) and f is a real-valued function defined on T. The intended interpretation is that T is an interval of time, with b being the beginning of a phoneme, word or sentence and e being its temporal end. Then f(t) is interpreted as the amplitude at time t of a spoken word, for example, or, also as the amplitude of the word's brain representation for some listener who heard it. The intuitive idea is that a word segment consists of a concatenation of phoneme segments, and a sentence, a concatenation of word segments. In the context-free version of the axioms, the congruence relation ≈ has the expected standard properties. For example, two spoken tokens of the word Paris would be congruent. Definition 1 A structure E = (E, P, W, S, P(P), P(W), P(S), ≈) is an elementary language structure if and only if the following axioms are satisfied: 1. E is a nonempty finite set of elementary segments (T, f) as defined above. 2. E = P ∪ W ∪ S. 3. P ∩ W = W ∩ S = P ∩ S = 0. 4. Each word w in W is a concatenation of phonemes in P, i.e., (i) w = (T,f), (ii) w = p1"pm, (iii) pi = (Ti, fi), (iv) T = [b,e) = [b1, e1) ∪"∪ [bm,em), (v) b = b1"ei = bi + 1"em = e, (vi) f = f1 ∪"∪ fm. 5. Each sentence s in S is a concatenation of words in W. 6. The relation ≈ is an equivalence relation on (P×P) ∪ (W×W) ∪ (S×S). 7. If w = p1"pm, w´ = p´1 "p´m , and pi ≈ p´i , 1 ≤ i ≤ m, then w ≈ w´.
36
8. If s = w1"wn, s´ = w´1 "w´n , and wi ≈ w´i , 1 ≤ i ≤ n, then s ≈ s´. 9. Any two elements of E that belong to the same member of any one of the three partitions P (P), P (W) or P (S) are congruent, independent of the context in which they occur. Although the axioms are stated in formal mathematical language, each one has a natural interpretation. Axiom 1 is just a finiteness restriction on E, characteristic of any sample of spoken or written language, however large it may be. The infinite structures characteristic of the formal theory of grammars do not, of course, impose such a restriction. Axiom 2 guarantees that every elementary segment in E is either a phoneme (p), word (w) or sentence (s). This is sufficient for our purposes. In other contexts, it would be natural to add syllables and, for English at least, many familiar short phrases that would be better to treat as units rather than as being composed of words. Axiom 3 permits no overlap between phonemes, words and sentences. One-word sentences do not occur in the work reported here. Accommodating them in other circumstances can be handled in several different ways. Axioms 4 and 5 characterize how words are concatenated from phonemes, and sentences from words. The temporal-process construction used here is meant to be realistic for verbal stimuli that are auditory or visual, matched in display timing to the corresponding auditory stimulus; and also for the recording and analysis of their brain representations. Axioms 4 and 5 also represent a divergence from the axiomatic formulation of formal grammars where sentences are described as finite sequences of words, and no formulation of the real-time process of production or comprehension is given. Axiom 6 is the requirement that the congruence relation ≈ be an equivalence relation, i.e., that it be reflexive, symmetric and transitive. Axioms 7 and 8 state the conditions that must be satisfied for a binary relation of equivalence to be a congruence relation. The congruence axioms have been written, for simplicity, in a geometric style. They can be converted to a standard algebraic form by explicitly introducing a binary operation of concatenation. Axiom 9 is conceptually important because we think of the partitions as having the property that a set which is a member of a partition can be identified as the type of a phoneme, word or sentence, as the case may be. Such a distinction, within a
37
formal set of axioms for linguistic processes or objects, is unusual, but desirable here, because of the emphasis on tokens—the elementary segments—as the objects of analysis. We now turn to the definition of isomorphism of two elementary language structures. Another concept we formalize along with isomorphism is that of an approximate isometry between corresponding elementary segments. The intuitive idea back of this constraint is that the real-time processing of the constituents of language and their brain representations must take about the same amount of time. For example, the processing of a spoken word ordinarily takes place within a few hundred milliseconds of its being heard, with a small set of troublesome words taking longer. In our experiments, all of the words used were meant to be familiar and easily processed. In a more elaborate statement of the theory, allowance needs to be made for difficult words. For example, in systematic psychological theories of eye movement during reading, if processing of a word or phrase is not complete and non-stimulus supported memory has decayed, backtracking to the immediately preceding word or phrase is likely to occur. In so doing, a tight approximate isometry would be disturbed (for detailed axioms on eye movements in reading, see Suppes, 1990). Definition 2 Let E1 = (E1, P1, W1, S1, P(P1), P(W1), P(S1), ≈1) and E2 = (E2, P2, W2, S2, P(P2), P(W2), P(S2), ≈2) be two elementary language structures. E1 and E2 are isomorphic if and only if there exists a one-one mapping ϕ from E1 onto E2 such that (i) ϕ(P1) = P2, ϕ(W1) = W2, ϕ(S1) = S2 and the partitions of P1, W1 and S1 are also preserved under the mapping ϕ; (ii) For e and e´ in E1, ϕ (e) ≈2 ϕ(e´) iff e ≈1 e´; (iii) If w = p1"pm, then ϕ(w) = ϕ(p1)"ϕ(pm); (iv) If s = w1"wn, then ϕ(s) = ϕ(w1)"ϕ(wn).
38
Moreover, the mapping ϕ is an ε-approximate isometry if for any element (T,f) of E1 with T = [b,e) |(e - b) - (ϕ(e) - ϕ(b)) | < ε. In some occurrences of spoken words and sentences, the context matters. A phoneme changes in the context of different phonemes, and similarly for words. To accommodate and study these phenomena, we weaken the axioms for context-free structures to context-dependent ones. But these modifications will not be considered here. So far in this section I have not said how to characterize congruence for elementary segments, i.e., phonemes, words and sentences, of brain representations of language. Keep in mind that this characterization is critical for data analysis, and does always refer to tokens, not types, of linguistic units, contrary to abstract versions of phonology and syntax. Some form of least-squares fit, summed over the discrete amplitude observations of the two electromagnetic waves, is mainly used in the models we tested. This means that congruence between test samples and prototypes is judged by which prototype fits the test sample best. This is not the occasion to enter into the experiments in which data were collected and analyzed to test the hypothesis of structural isomorphism of language constituents and the brain representations of them, i.e., phonemes, words, and sentences in the present context. I can display some graphic data. Figure 1 shows the close similarity of the brain representations of the word east based on data from different sentences. The close similarity supports, of course, the existence of a structural isomorphism, even though these data make nothing like a complete case. Figure 2 shows the same comparison of brain representations for the four consonants b, g, p, and t, when they occurred as initial consonants. To avoid any misunderstanding, I stress that structural isomorphism between language constituents and their brain representations does not mean that a brain representation of a spoken word, for example, perceptually resembles it. The most vivid example to make this same point is the great physical and perceptual distance between actual empirical procedures of measurement and their isomorphic representation by numerical structures.
39
Figure 1. The prototype for east is the solid curved line, and the dashedand-dotted line the test sample from an experiment. The x axis is measured in milliseconds from the onset of the visual word as stimulus. The y axis is measured in microvolts, with the numerical scale not shown. The vertical line on the left shows the beginning b and the one on the right the end e, thus marking off the temporal segment used for the leastsquares test of fit.
If the hypothesis of structural isomorphism were blatantly wrong, it would seem to make the brain's methods of computation in recognizing external processes and objects much more complex than they probably are. Structural-isomorphism claims are restricted, but still supportive of the general idea that the brain representations of processes and objects are similar, in this strong structural sense, to what they represent. Detailed empirical findings on such an isomorphism are given in (Suppes et al., in preparation).
40
Figure 2. Comparison of prototype and typical test sample after data censoring for each consonant b, g, p, and t. The thick solid line is the prototype and the thin solid line a typical test sample. The measurements on the x and y axes are the same as for Figure 1.
41
III. Habits, Automaticity and Consciousness Our mental concept of ourselves is that of self-aware human beings able to think deliberately, carefully, and thoroughly about all kinds of problems. But contrary to this folklore ideal, and contrary also to much folklore psychology, we are almost entirely unaware or unconscious of our thinking processes. What we often have is excellent knowledge of the results. We know, for example, that we have decided to go to the movies, or we have now made the decision to take a trip to the Arctic the next time we see a special fare. These results are salient, often privately known only to the individual having them, but all the same they reflect the results, not the patterns of thinking that led to them. It is easy to say, “Well, this is just another philosophical opinion.” In fact, the data are quite substantial in supporting the empirical conclusion that we have little consciousness of process, but much of results. Two useful articles providing many references to supporting empirical studies are Nisbett and Wilson (1977) and Wilson (1985). I have also surveyed these studies myself in an earlier article, Suppes (2003). Most of our walking and talking proceeds a pace without any need for prior deliberation or conscious reflection. If we stumble, we notice it, and are momentarily conscious of the steps we are taking. If we mumble when we talk, we are in like manner suddenly aware that we have made a mistake, and need to correct it. But these are exceptions. The normal course of things is wonderfully unselfconscious and smoothly running, without any awkward interference of efforts at deliberate and conscious thought. But clearly, there must be a place for deliberation and conscious reflection. Of course there is. Partly, it is in the contemplations, happily or sadly, of the results of these unconscious processes. Another, and more important, occasion for such consciousness is in the learning of something new and the deliberate effort to pay attention, to concentrate, to control movements, thoughts, and so forth. Such initial learning of habits is in many ways the most significant case of conscious attention. From Aristotle to James. The ideas that I am expressing about habits and about consciousness are not presumed to be new. In an excellent recent article Burnyeat (1999) points out how fundamental for Aristotle is the importance of habit in learning how to be good. Those who have stressed the great importance of the practical syllogism for Aristotle's analysis of the good have too often missed this point. Only the properly prepared person, i.e., someone raised from early youth in the proper manner, is going to be
42
able to achieve the desired state of flourishing described so well by Aristotle. Burnyeat's emphasis on the importance of the psychological development of the young as a central and essential feature of Aristotle's theory of virtue is a welcome and fresh addition to many of the standard readings of the Nichomachean Ethics. Equally important is Burnyeat's emphasis that, in Aristotle's view, learning what is noble and just, that is, learning virtue, does not consist simply of learning neat formulations of rules or traditional maxims. It takes an educated perception, the habitual capacity of going beyond the application of general rules, to know what is needed for the practice of the virtues in specific circumstances (Nichomachean Ethics, 1109b23, 1126b2-4). A similar thesis about Aristotle is developed by Sherman (1999), who also quotes relevant passages from the Politics, De Anima, Poetics, and Rhetoric, to support the importance of the developmental role of learning good habits early. That the Greek teaching of the young by tutors or in elementary schools followed the precepts of Aristotle is, of course, only partly true. A rather realistic account of Greek education in Hellenistic and Roman times may be found in Cribiore (2001). The one thing that is reinforced by her complex account of what the practice was like is the emphasis on the formation of habits. To a very large extent, the teaching in the early years emphasized mental gymnastics, as an analogue of the recognition of the need for physical gymnastics in the training of the body. Much more detailed than Aristotle’s own treatment of habits is that of Aquinas. I shall not try to survey it here. Some attention to it is given in Drolet and Suppes (2008). Hume does not much use the word “habit”, but writes about custom and here he is close to the earlier tradition. His definition is very much in the spirit of what I have been saying here about habits. “We call everything custom which proceeds from a past repetition without any new reasoning or conclusion” (Treatise, p. 104). In passages on the same page and others close by, Hume very much reinforces the kind of automaticity of habits that I want to stress more explicitly a little later. He makes the important point that in very obvious uniform conjunctions of events, what we then infer are causes and their effects; the mind does not reflect at all or reason upon the phenomena. It passes from one to the other, driven by the constant association. Here is the passage that is clear on this point: In general we may observe, that in all the most establish'd and uniform conjunctions of causes and effects, such as those of gravity, impulse, solidity, &c.,
43
the mind never carries its view expressly to consider any past experience: Tho' in other associations of objects, which are more rare and unusual, it may assist the custom and transition of ideas by this reflexion. (Hume, Treatise, p. 104)
But the locus classicus on this topic is Chapter 4 of James’ Principles of Psychology (1890). Here is the important passage about the move from conscious attention to automaticity, James’ point being that “habit diminishes the conscious attention with which our acts are performed”. This is what he has to say about this process: One may state this abstractly thus: If an act require for its execution a chain, A, B, C, D, E, F, G, etc., of successive nervous events, then in the first performances of the action the conscious will must choose each of these events from a number of wrong alternatives that tend to present themselves; but habit soon brings it about that each event calls up its own appropriate successor without any alternative offering itself, and without any reference to the conscious will, until at last the whole chain, A, B, C, D, E, F, G, rattles itself off as soon as A occurs, just as if A and the rest of the chain were fused into a continuous stream. When we are learning to walk, to ride, to swim, skate, fence, write, play, or sing, we interrupt ourselves at every step by unnecessary movements and false notes. When we are proficients, on the contrary, the results not only follow with the very minimum of muscular action requisite to bring them forth, they also follow from a single instantaneous ‘cue’. The marksman sees the bird, and, before he knows it, he has aimed and shot. A gleam in his adversary's eye, a momentary pressure his rapier, and the fencer finds that he has instantly made the right parry and return. A glance at the musical hieroglyphics, and the pianist's fingers have rippled through a cataract of notes. And not only is it the right thing at the right time that we thus involuntarily do, but the wrong thing also, if it be an habitual thing. Who is there that has never wound up his watch on taking off his waistcoat in the daytime, or taken his latchkey out on arriving at the door-step of a friend? Very absent-minded persons in going to their bedroom to dress for dinner have been known to take off one garment after another and finally to get into bed, merely because that was the habitual issue of the first few movements when performed at a later hour. The writer well remembers how, on revisiting Paris after ten years’ absence, and, finding himself in the street in which for one winter he had attended school, he lost himself in a brown study, from which he was awakened by finding himself upon the stairs which led to the apartment in a house many streets away in which he had lived during that earlier time, and to which his steps from the school had then habitually led. (James, Principles of Psychology, p. 114–115)
44
We do not have to be explicit about the conscious will, as James is in this passage, to appreciate fully the empirical correctness of what he says about forming habits. He does not use the term ‘automaticity’ that is now popular in the psychological literature, but it is clear that he has in mind exactly this phenomenon. A habit becomes automatic when it is performed without conscious attention. So in one sense consciousness and automaticity are opposites. Well, so what? The answer is what James stresses and other psychologists of the present time stress. Automaticity is the road to perfection. The conscious performance of acts that we think of as habitual is awkward, badly timed, and often inappropriate. The point of this remark is to emphasize how much daily experience, the amount of which is not fully appreciated by almost any philosophers of mind, is based on automatic, unconscious and habitual responses. Not on conscious deliberations or reflections of any kind. Moreover, this is true of our most cognitive activities, talking and listening. We automatically process the speech we produce and the speech we listen to. In almost no cases are we really conscious of either process. It is only after we have turned over in our minds what we have said, or what we have heard, that some point will come sharply to conscious attention. Again, this is a matter of process being mainly unconscious, but partial results being available for conscious inspection. Very much in the Aristotelian spirit of developing good habits is the following passage from James urging exactly that at a young age: The great thing, then, in all education, is to make our nervous system our ally instead of our enemy. It is to fund and capitalize our acquisitions, and live at ease upon the interest of the fund. For this we must make automatic and habitual, as early as possible, as many useful actions as we can, and guard against the growing into ways that are likely to be disadvantageous to us, as we should guard against the plague. The more of the details of our daily life we can hand over to them effortless custody of automatism, the more our higher powers of mind will be set free for their own proper work. There is no more miserable human being than one in whom nothing is habitual but indecision, and for whom the lighting of every cigar, the drinking of every cup, the time of rising and going to bed, every day, and the beginning of every bit of work, are subjects of express volitional deliberation. Full half the time of such a man goes to deciding, or regretting, of matters which ought to be so ingrained in him as practically not to exist for his consciousness at all. If there be such daily duties not yet ingrained in any one of my readers, let him begin this very hour to set the matter right. (James, Principles of Psychology, p. 122)
45
James goes on to strengthen this passage by emphasizing that in deciding to launch ourselves into a new habit, we must be as decisive and determined as possible. So, the place for consciousness in the initial change of habits is a salient and important one. But salience and importance recede, as should be the case, once the habit becomes automatic. What is good about this discussion of the learning of habits on James’ part is his full recognition that habits are not purely mental in some dualistic Cartesian sense, but always intimately related to both past and present actual performance. This is what James says: Down among his nerve-cells and fibres the molecules are counting it, registering and storing it up to be used against him when the next temptation comes. Nothing we ever do is, in strict scientific literalness, wiped out. Of course, this has its good side as well as its bad one. As we become permanent drunkards by so many separate drinks, so we become saints in the moral, and authorities and experts in the practical and scientific spheres, by so many separate acts and hours of work. (James, Principles of Psychology, p. 127)
What he also fully recognizes is that the basic physiological act of learning habits has remarkable similarity whether habits are good or bad. In other respects, as is emphasized in Drolet and Suppes (2008), the differences between good and bad habits are well accepted and much agreed upon in broad communities of persons ranging from the young to the old. Consciousness. In the later part of the chapter on habits, James examines several ways of thinking about the function of consciousness. I will not try to go through his various arguments. He does have, however, at the end a very suggestive idea. It is a functional one, not one identified by any particular physical or neural manifestation of consciousness. The idea is that the role of consciousness is a selective one. When we are confronted with a number of alternatives it is the function of consciousness to help us think about the choice we will make. Notice the difference here. When automaticity governs, the choice is made smoothly and unconsciously. When the choice is not one governed by habit, then consciousness often, even if not always, comes in to play. It is useful to compare James’ notion of selection with the notion that became dominant a few years later, namely, attention. By the first decade of the twentieth century it had become a central concept of experimental psychology. Some standard references are Wundt (1897), Pillsbury, (1908), and Titchener (1908). One leading idea of this functionalist ap-
46
proach to attention was that the initial condition for onset is stimulus change, and the recognition that the evidence of attention was also manifested in various physiological responses. However, the tight connection of attention to consciousness was subsequently lost in the pursuit of a strict behaviorism, especially in American psychology. It was regarded as unscientific in most of the decades of the first half of the twentieth century for experimental psychologists to speak of consciousness. But the study of attention prospered not only in the United States but also in Europe, especially in Russia, then the Soviet Union, for example, in the work of Luria and Vinogradova (1959) and Sokolov (1960, 1963). The Russian psychologists tended to talk about the orienting reflex rather than attention, but the functional conception was very similar. Such studies were also made in the United States. A good example is the study of the relation of the orienting reflex conditioning in Maltzman and Raskin (1965). Two good, more recent references on automaticity are Shiffrin and Schneider (1977) and Barge and Chartrand (1999). In spite of the new neural and psychological studies of consciousness, many questions do not yet have good answers. What remains constant is the psychological insistence that phenomenological awareness is the central feature of consciousness. This is true wherever awareness occurs and whether it is viewed as essential or accidental on any particular occasion of feeling, perceiving or thinking. Although not without controversy, the experimentally supported stress on awareness of results, not processes, will come to be recognized as sound, and will help narrow the phenomenological range of consciousness. I am skeptical that the physical source of consciousness will be found in any one location in the brain. Much more promising in my view is the conjecture that consciousness reflects a physical process of activation, as briefly described in Part I, and it may occur in many places in the brain, at least in the cortex. For an extensive review of the experimental literature on the relation between activation and consciousness, see Dehaene and Changeux (2005). They also support the hypothesis of a conscious neuronal workspace that distinguishes automatic subsystems from more autonomous, spontaneous supervisory systems. Their ideas build on the earlier cognitive theory of consciousness of Baars (1989). These ideas are very suggestive but still speculative. My own, also speculative, idea is that the physical embodiment of consciousness will be more dynamic and electromagnetic, with exact spatial location being of less importance.
47
References Aquinas, T. (1944). Summa theologica. In A. C. Regis, Ed., Basic writings of Saint Thomas Aquinas, Vol. 1. New York: Random House. Aristotle (1941). Metaphysics. In R. McKeon, Ed., Basic works of Aristotle. New York: Random House. Aristotle (1941). Nichomachean ethics. In R. McKeon, Ed., Basic works of Aristotle. New York: Random House, pp. 1109b23, 1126b2–4. Aristotle (1975). De Anima (On the soul). Cambridge, MA: Harvard University Press, 4th edn. English translation by W. S. Hett. First Published 1936. Baars, B. J. (1989) A cognitive theory of consciousness. Cambridge: Cambridge University Press. Barge, J. and T. Chartrand (1999). The unbearable automaticity of being. American Psychologist, 54: 462–479. Brennan, R. E. (1941). Thomistic psychology: A philosophical analysis of the nature of man. New York, NY: Macmillan. Brown, C. M. and Hagoort, P. (1999). The neurocognition of language. Oxford: Oxford University Press. Burnyeat, M. F. (1999). Aristotle on Learning to be good. In N. Sherman, Ed., Aristotle's ethics. New York: Rowman and Littlefield. Chomsky, N. (1980). Rules and representations. New York: Columbia University Press. Crestani, F. (1997). Application of spreading activation techniques in information retrieval. Artificial Intelligence Review, 11, 453–482. Cribiore, R. (2001). Gymnastics of the mind, Greek education in Hellenistic and Roman Egypt. Princeton: Princeton University Press. Dehaene, S. and J. P. Changeux (2005). Ongoing spontaneous activity controls access to consciousness: a neuronal model for inattentional blindness. PLoS Biology, 3: 0910–0927. Descartes, R. (1649/1927). Passions of the soul. In R. M. Eaton, Ed., Descartes selections, pages 361-403. New York, NY: Charles Scribner’s sons. First published 1649. Drolet, A. and P. Suppes (2008). The good and the bad, the true and the false. In M. C. Galavotti, R. Scazzieri, and P. Suppes, Eds., Reasoning, rationality, and probability. Stanford, CA: CSLI Publications. Hume, D. (1739/1951). A treatise of human nature. London: John Noon. Quotations taken from L. A. Selby-Bigge’s edition, Oxford University Press, London. James, W. (1890/1918). The principles of psychology. New York: Henry Holt and Company. Lear, J. (1978). Going native. Daedalus, 107, 177–78. Luria, A. R. and Olga S. Vinogradova (1959). An objective investigation of the dynamics of semantic systems. British Journal of Psychology, 50, 89–105.
48
Maltzman, I. and D. C. Raskin (1965). Effects of individual differences in the orienting reflex on conditioning and complex processes. Journal of Experimental Research in Personality, 1, 1–16. Minsky, M. L. (1967) Computation: finite and infinite machines. Englewood Cliffs, New Jersey: Prentice Hall. Nisbett, R. E., and T. D. Wilson, (1977). Telling more than we can know: verbal reports on mental processes. Psychological Review, 84, 231–259. Pillsbury, W. B. (1908). Attention. New York: The Macmillan Company. Rugg, M. D., and Coles, M. G. H. (Eds.). (1995). Electrophysiology of mind: Eventrelated brain potentials and cognition. Oxford: Oxford University Press. Shiffrin, R. M. and W. Schneider (1977). Controlled and automatic human information processing: II. Perceptual learning, automatic attending, and a general theory. Psychological Review, 84, 127–190. Sherman, N. (1999). The fabric of character. In N. Sherman, Ed., Aristotle's ethics. New York: Rowman and Littlefield. Sokolov, E. N. (1963). Perception and the conditioned reflex. New York: The Macmillan Company. Sokolov, E. N. (1960). Neuronal models and the orienting reflex. In Mary A. B. Brazier, Ed., The central nervous system and behavior. New York: Josiah Macy, Jr. Foundation, pp. 187–276. Suppes, P. (1974). Aristotle's concept of matter and its relation to modern concepts of matter. Synthese 28, 27–50. Suppes, P. (1990). Eye-movement models for arithmetic and reading performance. In E. Kowler (Ed.), Reviews of oculomotor research, (Vol. IV), Eye movements and their role in visual and cognitive processes (pp. 455–477). New York: Elsevier. Suppes, P. (2002). Representation and invariance of scientific structures. Stanford, CA: CSLI Publications. Suppes, P. (2003). Rationality, habits and freedom. In N. Dimitri, M. Basili, and I. Gilboa, Eds., Cognitive processes and economic behavior. Routledge Siena Studies in Political Economy. New York: Routledge, pp. 137–167. Suppes, P. and J-Y Béziau (2004). Semantic computations of truth, based on associations already learned. Journal of Applied Logic, 2, 457–467. Suppes, P., M. P. Guimaraes, D. K. Wong, and E. T. Uy (in preparation). Testing the hypothesis of structural isomorphism between constituents of language and their brain representations. Suppes, P., B. Han, J. Epelboim, and Z.-L. Lu (1999a). Invariance between subjects of brain wave representations of language. Proceedings National Academy of Sciences USA, 96, 12953–12958. Suppes, P., B. Han, J. Epelboim, and Z.-L. Lu (1999b). Invariance of brain-wave representations of simple visual images and their names. Proceedings National Academy of Sciences USA, 96, 14658–14663. Suppes, P., Z.-L. Lu, and B. Han (1997). Brain wave recognition of words. Proceedings National Academy of Sciences USA, 94, 14965–14969.
49
Tarski, A. (1936). The concept of truth in formalized languages. In A. Tarski, Logic, semantics, metamathematics. Translated by J. H. Woodger. Second edition, 1983, edited by John Corcoran. Indiana: Hackett Publishing Company, Inc. Themistius (350 A.D./1996). On Aristotle’s On the Soul. Translated by Robert B. Todd. New York: Cornell University Press. Titchener, E. B. (1908.) Lectures on the elementary psychology of feeling and attention. New York: The Macmillan Company. Wilson, T. D. (1985). Strangers to ourselves: The origins and accuracy of beliefs about one's own mental states. In J. H. Harvey and G. Weary, Eds., Attribution: Basic issues and applications. Orlando, FL: Academic Press. Wundt, W. (1897/1902). Outlines of psychology. Leipzig, W. Engelmann; New York, G. E. Stechert, 2nd (revised) English edn., from the 4th (revised) German edn., translated by C. H. Judd.
50
The Emergence of Justification in Ethics1 Dagfinn Føllesdal∗
During the first half part of the twentieth century it was quite common among philosophers – ranging from the logical empiricists to existentialists like Sartre and Heidegger – to regard ethics as beyond rational justification. These tendencies are found even today, mostly among scientists who have not followed the developments in philosophy, but also among postmodernists and some other philosophers who are unaware of what is going on in philosophy today. Many of those philosophers who most vehemently reject the possibility of argument and objectivity in ethics and in other fields have learned from Edmund Husserl to emphasize the subjective perspective. It is astonishing that they do not seem to have noticed that Husserl combined his attention to the subjective perspective with extensive arguments against subjectivism and relativism, first in the Prolegomena, the first volume of the Logical Investigations (1900), and later in other works. At least, none of the postmodernists addresses these arguments or indicates any other kind of awareness of their bearing on their own view. In this paper I shall argue that Husserl not only argued against relativism, but that he in his later work anticipated important developments concerning justification that took place in the second half of the 20th century. Husserl's contribution is particularly interesting because he put forward an interesting solution to the following fundamental problem: It is easy to see that the method that we shall now describe can help us to resolve disagreements. But why does it provide justification? Justification in science The 20th century development of justification in ethics was preceded by a similar development in justification in science which I will now briefly describe before I turn to ethics. 1
This article overlaps with my article with this same title in European Review, Vol. 13 (2005), No. 2, 169-182. I am grateful to European Review and its editor, Sir Arnold Burgen, for permission to use this material in this article. ∗ Stanford University
The traditional view on justification stems from Aristotle, who, as you remember, discussed three possible approaches to justification: (1) One could justify something by showing that it follows from something else, which in its turn is shown to follow from something else and so on ad infinitum. (2) Instead of going on forever like this one might eventually end up where one started, one moves in a circle. (3) At some point in the backward movement one stops at some fundamental principles that are not in need of further justification. Aristotle rejected the first two approaches, neither retreating backwards into an infinite regress nor moving around in a circle will give us justification, he maintained. Only the third approach he found satisfactory. The great challenge for Aristotle was therefore to find first principles that can serve as a basis for the justification of the rest of what we claim to know. His assertions about the natural places of the various elements were claimed to be such first principles. In this essay I shall argue for the second of these approaches, the circular one. This may seem preposterous, as it did to Aristotle. In the centuries after Aristotle many philosophers proposed alternative candidates for first principles. However, some of Aristotle's contemporaries, most of them followers of Plato, claimed that at least in the natural sciences one has to work with tentative hypotheses that may be confirmed or disconfirmed on the basis of how well they "save the phenomena." This tradition continued up through the Middle Ages. Thomas Aquinas regarded this hypothetical method as the appropriate method in the natural sciences and contrasted it with Aristotle's method, which he regarded as more suitable for metaphysics. Thomas gives as an example the view that the earth is round. One can argue for this in two ways, he said. One can either, as Aristotle, start from a first principle, in this case the principle that all particles of earth have a natural location, viz. the earth's center, towards which they are all striving. They therefore group into a shape that maximizes their closeness to this point, viz. a sphere. Or, one can, as in astronomy, confirm it from the shape of the lunar eclipse, or from the fact that the same stars are not seen from every part of the earth."2 Thomas pointed out that the astronomer's hypotheses are not necessarily true:
2
Thomas Aquinas, Commentary on Aristotle's Physics, II, the end of Lecture 3. Thomas makes similar observations in other works, for example, in Summa theol. I.32,1 ad 2, and in his Commentary to Aristotle's De caelo et mundo (see next note).
52
The astronomers have tried in different ways to explain the movements of the planets. But the hypotheses they have put forth are not necessarily true; even if these hypotheses seem to save the phenomena, it is not necessary that they are true, since one might perhaps explain the apparent movements of the stars by help of another assumption, which nobody has ever thought of. Aristotle used, however, such hypotheses about the nature of movements as if they were true.3
The search for first principles continued the longest in mathematics and ethics, which seemed not to be based on empirical observations and where, especially in mathematics, various first principles or axioms were so intuitively convincing that they seemed to be beyond any doubt. However, in 1843 John Stuart Mill claimed that mathematics, like the natural sciences, was ultimately founded on observation.4 According to Mill, what makes mathematics so reliable are two features: (1) it is tested again and again when it is used in the marketplace or in surveying and other practical activities; (2) mathematics is systematized in an extremely simple and wide-ranging way, so that if one were to change part of it, most other parts would have to be changed as well. Given that these other parts have been repeatedly tested, we would look upon any observation that would seem to go against mathematics with great suspicion. Rather than starting on the immense enterprise of revising all of mathematics, we would blame the disturbing observation on something else: inattention, somebody trying to fool us or the like. Twentieth-century developments One hundred years later ideas similar to those of Mill were put forth by several philosophers who went even further than he did. Three of these philosophers deserve special mention: W.V. Quine (1908-2000), Nelson Goodman (1906-1998) and Morton White (1917- ). In the following their views will be presented and discussed, and we will, in particular, see how the basic idea in their views on justification was taken over and developed in ethics.
3
Thomas Aquinas, Commentary to Aristotle's De caelo et mundo, Lecture 17, n. 2. John Stuart Mill, A system of logic: ratiocinative and inductive : being a connected view of the principles of evidence, and the methods of scientific investigation. 2 vols. London: Parker, 1843. 4
53
Quine's oft-quoted characterization of his holism at the end of "Two dogmas of empiricism" is a nutshell description of this method of justification applied to a broad field, comprising empirical science, mathematics and logic, but with nothing said about ethics: ... total science is like a field of force whose boundary conditions are experience. A conflict with experience at the periphery occasions readjustments in the interior of the field. ... Having reëvaluated one statement we must reëvaluate some others, which may be statements logically connected with the first or may be statements of logical connections themselves. ... No particular experiences are linked with any particular statements in the interior of the field, except indirectly through considerations of equilibrium affecting the field as a whole.5
Before proceeding to the main features of the method it may be instructive to consider an example of an application of this method of justification to the problem of justification in logic. Let me quote Nelson Goodman's classical description of the method, in his Fact, Fiction and Forecast (1955): How do we justify a deduction? Plainly by showing that it conforms to the general rules of deductive inference. ..... But how is the validity of rules to be determined? Here again we encounter philosophers who insist that these rules follow from some self-evident axiom, and others who try to show that the rules are grounded in the very nature of the human mind. I think the answer lies much nearer the surface. Principles of deductive inference are justified by their conformity with accepted deductive practice. Their validity depends upon accordance with the particular deductive inferences we actually make and sanction. If a rule yields inacceptable inferences, we drop it as invalid. Justification of general rules thus derives from judgments rejecting or accepting particular deductive inferences. This looks flagrantly circular. I have said that deductive inferences are justified by their conformity to valid general rules, and that the general rules are justified by their conformity to valid inferences. But the circle is a virtuous one. The point is that rules and particular inferences alike are justified by being brought into agreement with each other. A rule is amended if it yields an inference we are unwilling to accept; an inference is rejected if it violates a rule we are unwilling to amend. The process of justification is the delicate one of mak5
W.V. Quine, "Two dogmas of empiricism," Philosophical Review 60 (1951), 20-43. Reprinted in Quine, From a Logical Point of View, Harvard University Press, Cambridge, Mass., 1953, 2. ed. 1961. The passage quoted occurs on pp. 42-43 of From a Logical Point of View.
54
ing mutual adjustments between rules and accepted inferences; and in the agreement achieved lies the only justification needed for either.6
Justification in ethics These ideas on justification in science, mathematics and logic were carried over to ethics during the second half of the 20th century. Two of the central figures in this development of justification in ethics were John Rawls (1921-2002) and Israel Scheffler (1923-). Rawls first presented this view on justification in ethics in "An outline of a decision procedure for ethics"7 in 1951 and developed it further in A Theory of Justice (1971).8 Scheffler set forth his view in "On justification and commitment" in 1954.9 Scheffler gives credit to Goodman, but apart from this there are no cross references between these authors to tell us where this approach to justification in ethics originated first. Reflective equilibrium This emerging view on justification was dubbed by Rawls "reflective equilibrium," and this fitting appellation will be used in the following. Using the label "reflective equilibrium" for a method of justification may be misleading, since when Rawls uses this label in A Theory of Justice and his later writings he does not have in mind a theory of justification, but a method for obtaining agreement. There is some vacillation in Rawls on this point, in "Outline of a decision procedure for ethics" Rawls regards the method as providing justification (p. 186); he talks about the method helping us to attain moral knowledge (p. 177) and about moral rules being validated through the use of this method (p. 177) and he states that he is not concerned "with the problem of how to make it psycho-logically effective in the settling of disputes" (p. 177). However, in his 6
Nelson Goodman, Fact, Fiction and Forecast, Harvard University Press, Cambridge, Mass., 1955 and later editions. Here quoted from the second edition, Bobbs-Merrill Company, New York, 1965, pp. 62-64. The italics are Goodman's 7 John Rawls, "An outline of a decision procedure for ethics." The Philosophical Review 60 (1951), 177-197. 8 John Rawls, A Theory of Justice. Cambridge, Mass.: Harvard University Press, 1971. 9 Israel Scheffler, "On justification and commitment." Journal of Philosophy 51 (1954), 180-190
55
later writings Rawls is more cautious, and he distinguishes more clearly between achieving agreement and providing justification. He no longer contends that the method provides justification, but settles for the more modest claim that it is a practical way of achieving agreement. This is just the issue where Husserl made a highly interesting contribution, to which we will come later. We will use "reflective equilibrium" for any method that has certain general characteristics that we will now list, whether it be conceived of as a method of justification or as a method of settling disputes. 'The hypothetico-deductive method' is less appropriate, since that method contains only some of the ingredients characteristic of the method of reflective equilibrium. The method of reflective equilibrium, as I will use the term, is characterized by the following six features: (i) Justification The method is usually a method of justification. As we noted, Rawls in his later writings regards the method as merely a practical way of reaching agreement, and there are some who regard the method as yielding only explanation or prediction but not justification. This is a main issue to which we shall return. However, all versions of the method have the remaining five features in common: (ii) Coherence The method emphasizes the coherence of one's views. The coherence is of the kind that we typically strive for in scientific theories, deductive logical inference plays an important role, so do simplicity and other considerations: some might, for example, want to make use of what is often called 'inference to the best explanation'. Typically, particular statements are justified by being deduced from more general statements, but on the other hand, the general statements in their turn are justified by the fact that the desired particular statements follow from them. So far, the method is nothing over and beyond the hypotheticodeductive method. However, now we come to its distinctive features,
56
which set the method of reflective equilibrium apart from the hypotheticodeductive method: (iii) Total corrigibility. No statement in one's "theory" is immune to revision (I use the word 'theory' in a broad sense, which does not require a theory to be fully worked out into a deductive system, but only requires the statements to be sufficiently related to permit transfer of evidence between them). Any statement may be given up when we find that giving it up brings about simplifications and greater coherence in our over-all theory. The views of some of the logical empiricists on "protocol statements" as non-revisable and unaffected by theory, are incompatible with this, and their methods are therefore not examples of reflective equilibrium, although they are examples of the hypothetico-deductive method. The method of reflective equilibrium is also incompatible with a theory of "sense data," where statements about sense data are supposed to be incorrigible. Adherents of reflective equilibrium are fallibilists not only with respect to some or most statements, such as the hypotheses of the theory, as in the case of the hypothetico-deductive method, but with respect to all, including reports of observations and other "data". (iv) Different fields of application The method of reflective equilibrium can be applied in a number of different fields, four prominent ones being empirical science, mathematics, logic and ethics. Philosophers can regard the method as appropriate for one, two, three, or all four of these fields. Philosophers can also be distinguished according to whether they regard these four fields as separate fields, where evidence from one field does not transfer to the other fields, or whether they are what we could call "unbounded" holists, and regard all four fields as part of one whole, where coherence considerations involve all four of them and where evidence accordingly is transferred from one area to the other. Evidence from the empirical sciences will thus be relevant for questions of values and norms, and more remarkable, evidence from ethics may have a bearing on questions in mathematics, logic or empirical science. The foremost representative of such unbounded holism is Morton White. First in Toward Reunion in Philosophy (1956), pp. 254-58 and 263,
57
and later in What Is and What Ought To Be Done (1981)10 White argued that all the four areas mentioned are interrelated in such a way that statements from all four areas can be put to test together, and that sometimes "we may reject or revise a descriptive statement in response to a recalcitrant moral feeling." (What Is and What Ought To Be Done, p. 122). By including statements from all four areas in the body of statements to be tested White includes more than Quine, who did not discuss ethics, and much more than Duhem, in The Aim and Structure of Physical Theory11, who was an important "holist" but did not include mathematics and logic and did not discuss ethics. However, on the other hand White includes less than Quine in each separate test. While Quine holds that "every one of our beliefs is on trial in any experiment or test" (What Is and What Ought To Be Done, pp. 22-23), White, like Duhem, thinks that only part of our "web of belief" is involved in each test. To distinguish his view from Quine's White therefore calls his view 'limited corporatism'. He uses "corporatism" the way I am using the word "holism", for the view that "we do not test isolated individual statements but bodies, or conjunctions, of statements" (p. 15). The qualification 'limited' indicates that White, like Duhem and unlike Quine, holds that the bodies of statements that are tested in any one test are less comprehensive than they are thought to be by Quine. Avoiding the word "corporatism" that is also used in political theory, I will continue to use "holism" for what White calls 'corporatism'. One might, if one wanted to, introduce different labels for the different variants of holism, for example "piecemeal holism" for Duhem and White's view, where the bodies that are tested in any single test make up only limited pieces of our whole web of belief, and "bounded holism" for a view like Quine', where not all four fields are included. Duhem's view would hence be a "bounded piecemeal holism," while White's would be an "unbounded piecemeal holism." However, rather than burdening my reader's memory, I will spell out what kind of holism I am discussing whenever that is pertinent.
10
Morton White, Toward Reunion in Philosophy, Harvard University Press, Cambridge, Mass., 1956, and What Is and What Ought to Be Done, Oxford University Press, New York, 1981. 11 Pierre Duhem, La Théorie Physique: Son Objet, Sa Structure, Rivière, Paris, 1906, English translation, as The Aim and Structure of Physical Theory, Princeton University Press, Princeton, 1954.
58
In Science and Sentiment in America,12 White shows that William James vacillated between a "trialistic" view and a holistic view. The former predominates in the Psychology and in The Will to Believe, where natural science, logic/mathematics, and ethics are three separate fields, each subject to a method like that of a reflective equilibrium, but with no transfer of evidence between the fields. The latter, holistic view, White finds advanced in some parts of Pragmatism and in A Pluralistic Universe, where all three fields are regarded as part of one unified whole, one stock of beliefs, in a broad sense, with evidence being transferred between the fields: strains in one field may be increased or reduced by what is happening in the other fields. According to White, "an unsatisfied desire may challenge the stock as much as the discovery of a logical contradiction or a recalcitrant fact, and it is James' belief in the parity of unsatisfied desire with the two other creators of strain that distinguishes his later position." (White, op.cit, p. 205) White's latest book, A Philosophy of Culture: The Scope of Holistic Pragmatism13 gives a thorough historical and systematic examination of the development of this corporatist view. One important theme that is more thoroughly discussed in A Philosophy of Culture than in the earlier works is the status of the holism itself: "I have maintained that thinkers who seek knowledge do and should use the method of holistic pragmatism in testing their views" (pp. 184-185). Can pragmatic holism itself be tested and given up? White responds that we should regard it as a rule of good scientific methodology. It reflects the idea that epistemology is a normative discipline, but it should not be regarded as a priori, necessary or immutable. Like some other rules in ethics and science "they are entrenched, but they may be removed from their trench for good reason" (p. 182). Note here the three features of the method of reflective equilibrium that we have discussed so far: justification, coherence and total corrigibility. The passage I quoted from Goodman brings into focus a further, highly important feature, to which we shall now turn: pre-reflective, intuitive acceptance as the basic source of evidence. 12
Morton White, Science and Sentiment in America: Philosophical Thought from Jonathan Edwards to John Dewey, Oxford University Press, New York, 1972 13 Morton White, A Philosophy of Culture: The Scope of Holistic Pragmatism. Princeton University Press, Princeton, 2002.
59
(v) Pre-reflective, intuitive acceptance The method of reflective equilibrium makes crucial use of our prereflective, intuitive acceptance of various statements. Through reflection, systematization and observation it seeks to gradually modify our acceptances, strengthen some of them and weaken others, but it does not attempt a whole-sale rejection of all of them in order to replace them with something radically new. There is no source of evidence upon which such a new edifice could be built, all the evidence there is, is imparted through these intuitive acceptances. (vi) Perception and other sources of evidence. Our intuitive acceptances come in various strengths and are influenced by various factors, some of which we consider more reliable than others. Perception influences many of our acceptances of what the world that affects our senses is like. Perception would, at least by empiricists, be looked upon as one privileged source of evidence, which, although not infallible, provides whatever evidence there is, in addition to the coherence considerations. While most philosophers would attribute to perception and observation such a privileged role in the sciences, the situation is not so clear in ethics. There have been philosophers who have treated moral feelings, or sentiments, as an ethical counterpart to perception. As we shall see, Husserl had this view. Rawls, in "Outline of a decision procedure for ethics," seems to hold that some particular moral judgments have such a privileged status, and that the general ethical principles have whatever acceptability they have in virtue of how well they systematize our particular moral judgments. Our acceptance of some of the particular judgments may be modified through this systematization, but the particular judgments remain the ultimate source of evidence for the ethical principles, much as in science the particular observation statements are the ultimate source of evidence for the general hypotheses of the theory. However, Rawls in his later writings no longer gives particular moral judgments a privileged status when compared with the general judgments, but gives judgments of both kinds the same status.
60
Our intuitive acceptances of ethical statements are often unreliable, depending on egoistic considerations, cultural influences, etc.. Reflecting upon them we come to regard some as less reliable than others. These considerations of reliability are part of the reflections we perform in order to arrive at a reflective equilibrium and these considerations themselves have to fit into the reflective equilibrium. Only a careful study of how various observations, experiences, and changes in our system affect our acceptance can tell us whether, in addition to coherence considerations, which are crucial to the method of reflective equilibrium, there is any source of evidence that is of particularly great importance. Reflective equilibrium in Husserl In an earlier article I have argued that Husserl held a reflective equilibrium view on evidence.14 He accepted all the features of this view that we have discussed: coherence, total corrigibility, pre-reflective acceptances, etc. He does so separately within all the four areas that we have described, natural science, mathematics, logic and ethics, and he gives a special twist to the method, based on how the lifeworld plays a crucial role in justification. I shall not repeat this examination of Husserl's writings here, but I will use the last minutes of my presentation to mention some main points in his view on justification in ethics and sketch the special role that the lifeworld plays in justification. Husserl on justification in ethics Husserl wrote relatively little on ethics compared to what he wrote on epistemology and the philosophy of logic. However, from what he wrote one may gather that he had a reflective equilibrium view on justification in ethics and that our sentiments are the basic evidence against which we try ethical theories.
14
”Husserl on evidence and justification.” In Robert Sokolowski (ed.), Edmund Husserl and the Phenomenological Tradition: Essays in Phenomenology (Proceedings of a lecture series in the Fall of 1985.) (Studies in Philosophy and the History of Philosophy, Vol. 18) Washington: The Catholic University of America Press, 1988, pp. 107-129.
61
As an introduction to his discussion Husserl sets forth the question of what kind of evidence one has in ethics: And thus also in ethics we have to ask: where is the source of the primitive ethical concepts, where are the experiences [Erlebnisse], on the basis of which I can grant these concepts the evidence of conceptual validity?15
For Husserl, acts of sentiment [Gemüt] are the source of evidence for values.16 For: How can the unconditional validity of 'ought' be recognized, if not by some relations or peculiarities of sentiments, and indeed acts of sentiment, lying at the bottom, which the one who is judging is looking to?17
Further, Husserl writes: The English moral sense philosophy [Gefühlsmoral] has after all established beyond doubt: If we imagine a being, who is sentiment-blind in the same way as we know beings who are color-blind, then everything moral loses its content, the moral concepts become words without sense.18 Thus it is obvious that there can be no talk of 'good' and 'bad' when one abstracts from sentiments.19 Not as if we can accept the content of his [Hume's] argumentation in its entirety, but one thing it makes certain, one thing is thereby completely evident: sentiment is essentially involved in the coming about of ethical distinctions.20
Like Rawls, Husserl regards the method of justification in ethics as not axiomatic and fundamentalistic, but clarifying and reflecting. Also, like Rawls, Husserl finds his method anticipated by Socrates: Socrates ... recognized the fundamental sense of this method, expressed in a 15
Husserl-manuscript F I 20, p. 106. Quoted by Alwin Diemer, in Edmund Husserl. Versuch einer systematischen Darstellung seiner Phänomenologie (Monographien zur philosophischen Forschung, Band XV), Anton Hain, Meisenheim am Glan, 2. ed., 1965, 316. 16 Husserl-manuscript F I 23, p.77 (Diemer, op.cit., p. 316, n. 3). 17 Husserl-manuscript F I 20, p.227 (Diemer, p. 317, n. 6, and also p. 48, n. 48). 18 Husserl-manuscript F I 20, p. 227 (Diemer, p. 317 and also p. 48, n. 106). 19 Husserl-manuscript F I 20, p. 107 (Diemer, 317, n. 6). 20 Husserl-manuscript F I 20, p. 99 (Diemer, 317, n. 7).
62
modern way, as intuitive and a priori critique of reason. Or, more accurately, he recognized its fundamental sense as a method of clarifying stocktaking of oneself...21
One final observation on Husserl's view on justification in ethics: Since our concrete particular intuitions are the last source of validity in ethics, Husserl avoids, like Rawls, a danger that often comes up in connection with fanaticism and totalitarianism: one acts against common sense morality in the name of some higher moral principle. The role of the lifeworld in justification Husserl held that the process of justification is ultimately based on intuitive acceptances and that there is no deeper source of justification. The set of all these intuitive acceptances makes up what Husserl called the "lifeworld." I shall not give any presentation of the lifeworld here, but assume that its main features are known.22 We shall now look at the crucial role Husserl assigns to the lifeworld in the process of justification. In his last work, Crisis of the European Sciences, Husserl writes: There has never been a scientific inquiry into the way in which the lifeworld constantly functions as a subsoil, into how its manifold prelogical validities act as ground for the logical ones, for theoretical truths. And perhaps the scientific discipline which this lifeworld as such, in its universality, requires is a peculiar one, one which is precisely not objective and logical but which, as the ultimate grounding one, is not inferior but superior in value.23
Note how Husserl here expresses a view very similar to that of Goodman: the pre-logical validities act as ground for the logical ones, the lifeworld functions as a subsoil. Remember how, according to Goodman, "principles of deductive inference are justified by their conformity with accepted deductive practice. Their validity depends upon accordance with the particular deductive inferences we actually make and sanction." (Goodman, op.cit, p. 63). Rawls, too, appeals ultimately to "what seem to be in21
Erste Philosophie, 2. Vorl., Husserliana. VII, 11.10-14. A presentation and discussion of the lifeworld may be found in my ”The Lebenswelt in Husserl.” In Leila Haaparanta, Martin Kusch, and Ilkka Niiniluoto, eds., Language, Knowledge, and Intentionality: Perspectives on the Philosophy of Jaakko Hintikka (Acta Philosophica Fennica, Vol. 49). Helsinki, 1990, pp. 123-143. 23 Krisis, § 34, Husserliana. VI, 127.13-20 = page 124 of David Carr's translation (Evanston, Ill: Northwestern University Press, 1970). The italics are mine. 22
63
tuitively acceptable and reasonable moral decisions". ("Outline of a decision procedure for ethics," p. 194) Also in Experience and Judgment there are several similar passages, among them the following: ... the retrogression to prepredicative experience and the insight into what is the deepest and ultimately original level of prepredicative experiences signifies a justification of doxa, which is the realm of ultimately original self-evidence, not yet exact and physicomathematically idealized. Thereby, it is also shown that this realm of doxa is not a domain of self-evidence of lesser rank than that of episteme, of judicative knowledge and its sedimentations [Niederschläge], but precisely the domain of ultimate originality to which exact cognition resorts for its sense, such cognition (it must be recognized) having the characteristic of being a mere method and not a way leading to knowledge by itself.24
I would like to end the discussion of this point by quoting a long passage from the Crisis, where Husserl expresses ideas very similar to those of Goodman: What is actually first is the "merely subjective-relative" intuition of prescientific world-life. For us, to be sure, this "merely" has, as an old inheritance, the disdainful coloring of the doxa. In prescientific life itself, of course, it has nothing of this; there it is a realm of good verification and, based upon this, of wellverified predicative cognitions and of truths which are just as secure as is necessary for the practical projects of life that determine their sense. The disdain with which everything "merely subjective and relative" is treated by those scientists who pursue the modern ideal of objectivity changes nothing of its own manner of being, just as it does not change the fact that the scientist himself must be satisfied with this realm whenever he has recourse, as he unavoidably must have recourse, to it.25
The life-world as the ultimate court of appeal Finally, I come to the important point that the life-world for Husserl is an ultimate court of appeal, behind which there is no point in asking for further justification. The main reason Husserl gives for this, is that most of 24
Erfahrung und Urteil, § 10, p. 44 = page 46 of James S. Churchill & Karl Ameriks' translation (Evanston, Ill: Northwestern University Press, 1973). 25 Krisis, §34a, Husserliana. VI, 127.31 - 128.10 = Carr's translation, p. 125.
64
the life-world consists of acceptances that we have never made thematic to ourselves and which have therefore never been the subject of any explicit judicative decision: ...where such completely self-giving intuition of the judicative substrates takes place, there is absolutely no possible doubt with regard to the "so" or "otherwise" and hence no occasion for an explicit judicative decision.26
Every claim to validity and truth rests upon this "iceberg" of largely unthematized prejudgmental acceptances. Every request for justification ultimately has to lead back to this same sort of acceptances. There is nothing more ultimate to turn to, and there is nothing more that can be asked for: there is nothing to "postulate" or to "interpret suitably", but only something to bring to light. Thus alone can that ultimate understanding of the world be attained, behind which, since it is ultimate, there is nothing more that can be sensefully inquired for, nothing more to understand.27
26
Erfahrung und Urteil, § 67, p. 330 = Churchill & Ameriks' translation, p. 275. Formale und transzendentale Logik, § 96b, Husserliana. XVII, 249.17-20 = page 242 in Dorion Cairns' translation (The Hague: Nijhoff, 1969). 27
65
Comment on Føllesdal’s “The Emergence of Justification in Ethics” Patrick Suppes∗
I’m in basic agreement with Husserl’s emphasis on experience and practice necessarily being the basis of the justification in almost every subject. I think that Dagfinn brings out nicely the importance of this approach, and he properly emphasizes that there is nothing subjective in any strong sense about Husserl’s position on the nature of ethics compared, say, to the nature of mathematics or science. I just want to make two points of further detail that I think are of interest and not covered by Dagfinn in his paper. The first concerns an explanation of why we have much more successful formal developments, first of all, in mathematics, second in science, and third in ethics. I think it is hard to disagree with my claim that the ranking for detailed, thoroughly agreed upon, developments should be made in the order I have given. So the existence of this difference between the major areas that Dagfinn compares is of great importance. I emphasize at the same time that I do not find these differences surprising. In all kinds of practices in ordinary affairs, differences arise on the degree of agreement on the given normative standards. Some of the intensity of these differences surely arises from the fact that, for example, the practices of experimental physics are much more special and restricted than the practices of ordinary cooking. So, it is much easier to find disagreement about how to best prepare a given dish, perhaps just because many more persons are more exposed to, more knowledgeable, and therefore more definite in their opinions about cooking than they are about physics experiments. I will not try to explore whether it is reasonable to think that this line of argument is sufficient to explain at a more detailed level the differences between the agreement on normative standards in experimental physics of what it means to conduct an experiment in a proper manner, and agreements on, let us say, major issues of social or economic policy. But it is my belief the arguments about the relative extent of com-
∗
Stanford University
mon knowledge and practice will not alone explain the difference, yet I will not explore this matter in more detail here. My second point concerns the too often overlooked presence of conflict of norms in every area of practice that Dagfinn mentions. Outsiders might be surprised to find such conflict of norms in mathematics, but it is easy to do so. I mention just one outstanding example in the twentieth century, the normative decision as to whether to accept mathematical proofs that necessarily use the axiom of choice or one of its equivalents. Classical mathematical analysts find it almost irresistible to use this axiom to prove many results that seem conceptually important; on the other hand, mathematical constructivists are unsatisfied with such a blatantly nonconstructive principle. Another example of recent times, within mathematics, is the normative attitude toward the use of computers as aids in proofs. One can already see a changing attitude. Not too long ago, many mathematicians scorned such use. As it has become more prominent and proofs using such computers have become increasingly complicated, it is difficult to see how the proofs might be given without them. So normative attitudes are changing, but a conflict is still clearly present. Well, compared to mathematics the conflicts in ethics are everywhere, and I think it is of great importance in the teaching of ethics to emphasize their fundamental character. It is also, I would like to point out, a common experience in ordinary life to be faced with such a problem. Any parent that has raised a number of teenagers, as I have, knows that having some fixed sets of norms for their behavior that are going to be held onto come hell or high water is not going to work very well in many kinds of settings. The desirability of norm conflicts on the behavior of the young may be something of which one can be critical, but its existence, unfortunately, can hardly be denied for large parts of society in many parts of the world. More generally, there is too much optimism about the resolution of such conflicts of norms in ethical discussions of many kinds. Even in some of my favorite books on the subject, I am surprised to find missing, attention to such conflicts, and the importance they hold for the subject. A nice example of such an irresolvable conflict is given in the following imagined case cited by Hume and apparently much discussed in earlier times about the right of possession of property. The fable is about a conflict between two Grecian colonies over ownership of a deserted city. From all these circumstances, ‘tis easy to see how perplex’d many questions may become concerning the acquisition of property by occupation; and the least effort of thought may present us with instances, which are not susceptible
68
of any reasonable decision. If we prefer examples, which are real, to such as are feign’d, we may consider the following one, which is to be met with in almost every writer, that has treated of the laws of nature. Two Grecian colonies, leaving their native country, in search of new seats, were inform’d that a city near them was deserted by its inhabitants. To know the truth of this report, they dispatch’d at once two messengers, one from each colony; who finding on their approach, that their information was true, began a race together with an intention to take possession of the city, each of them for his countrymen One of these messengers, finding that he was not an equal match for the other, launch’d his spear at the gates of the city, and was so fortunate as to fix it there before the arrival of his companion. This produc’d a dispute betwixt the two colonies, which of them was the proprietor of the empty city; and this dispute still subsits among philosophers. For my part I find the dispute impossible to be decided, and that because the whole question hangs upon the fancy, which in this case is not possess’d of any precise or determinate standard, upon which it can give sentence. To make this evident, let us consider, that if these two persons has been simply members of the colonies, and not messengers or deputies, their actions wou’d not have been of any consequence; since in that case their relation to the colonies wou’d have been but feeble and imperfect. Add to this, that nothing determin’d them to run to the gates rather than the walls, or any other part of the city, but that the gates, being the most obvious and remarkable part, satisfy the fancy best in taking them for the whole; as we find by the poets, who frequently draw their images and metaphors from them. Besides we may consider, that the touch or contact of the one messenger is not properly possession, no more than the piercing the gates with a spear; but only forms a relation; and there is a relation, in the other case, equally obvious, tho’ not, perhaps, of equal force. Which of these relations, then, conveys a right and property, or whether any of them be sufficient for that effect, I leave to the decision of such as are wiser than myself. Hume, 1739/1951, pp. 507-508
Reference Hume, D. (1739/1951). A treatise of human nature. London: John Noon. Quotations taken from L. A. Selby-Bigge’s edition, Oxford University Press, London, 1951.
69
The World and the Worlds Wilhelm K. Essler∗
Abstract By using levels of languages with increasing strength instead of one ungraded language of first order logic, the danger of paradoxes is avoidable and, moreover, the gain of categoricity is available. In this paper such a conception of language-levels is accepted in its semantic as well as epistemological aspects. Therefore, to deal with some world means to use a particular language out of these language-levels. In this language, the frame of the world concerned is established. Within that frame, the actual world is determined by empirical means.
Unfortunately, I never had the opportunity to participate in one of the lectures given by Patrick Suppes at Stanford. But luckily I did pay attention to the results of his philosophical investigations published in his books and articles, or at least to some of them. Especially with regard to the philosophy of science, his analysis of definition and definability as well as of fundamental and derived measurement were mind opening and illuminating. They guided me in my further philosophical development. It is significant for great philosophers that students as well as readers are able to learn important things from them, even if they do not follow the fundamental lines of their philosophy. In this sense, I learnt so much from him without following exactly the path of his investigations. Thus, e. g. by remembering results of Skolem and Kanger, I mistrusted and I still am mistrusting typeless set theories as far as their philosophical values are concerned. The main reasons for choosing languages which presuppose poorer ontologies are: • I still agree with Wilhelm von Ockham’s principle “Entia non sint multiplicanda praeter necessitatem”. Therefore, presented with some ∗
Johann-Wolfgang-Goethe-Universität Frankfurt a. M.
theory to be philosophically investigated and logically analysed, I am still searching for the most simple language w. r. t. syntactical expressability and ontology. • To be able to present the ontology of a particular given language, this ontology needs to be uniquely describable; i. e.: the axioms which describe the structure of those entities have to be categorical in order to recognize the structure of the different models of the set of axioms as being identical. • In each case, according to such a respectively minimal ontology, I will be able to go on and receive a maximum of epistemological and métaphysical results, naturally by avoiding all the metaphýsical diversions. • According to Tarski’s levels of semantics and therefore of ontology I then continue looking for levels of epistemologies and métaphysics related to them. Thinking is the mind’s discussion with itself. This was known already by ancient Indian philosophers, e. g. by Buddha Shakyamuni; and it was known later on also by Platon. Of course, this identity is valid only for those movements of the mind that work conceptually; therefore, no other movements of the mind shall be taken into consideration here. And since an inner speaking is absolutely unapproachable to all others, it is useful and appropriate to translate the respective language of inner judgements into some language of outer statements. Observing something means to act in some particular way. The result of performing such an action is observation, i. e. an inner judgement or an outer statement regarding to the object of observation. For purposes of logical reconstruction, it is useful to regard an appropriate language of observations and measuring values, i. e. to investigate the smallest logically and semantically completed language which enables classifications and comparisons of the objects needing to be observed resp. to be measured. Choosing languages which are too powerful very often causes unintended problems which in turn complicate and impede the analysing of the logical form of observation. For epistemological reasons, it is therefore advisable to choose a limited language of first order logic
72
right from the start, e. g. an object-language which contains a sufficiently large class of unary and binary predicates in addition to the individual expressions, but no further kinds of descriptive resp. cognitive constants. In order to establish the semantics of such an object-language M 0S in a simple way, Tarski’s procedure of homogenizing the presupposed ontology is applied, i. e.: the individuals are mapped onto the unit sets of those individuals; and the unary attributes are mapped into the binary ones; i. e. onto the ordered pairs of those of them, where the second member is without any significance. The universe 0 U of such an interpretation of M 0S consists of the set of individuals to be observed – in past, present and future – according to the intensions of the predicates of M 0S . Therefore, by means of the interpretation 0 J upon 0 U all the individual expressions – may they either be constant or free variables – denote individuals of 0 U only, resp. their unit sets, resp. their correlating binary attributes; and furthermore, all the individuals of 0 U may be denoted by some individual expression. Such an interpretation 0 J of M 0S on 0 U then presents an outcome, i. e. a completed result of the performances of the intensions of all the predicates of M 0S on the individuals of 0 U . The final result derived from correcting and completing all the single factual results of those performances as they were finally accepted, will be the factual outcome. A logically possible world w.r.t. M 0S –, or, according to lattice theory: an atom, or, according to the probability theory, an atomic event – is therefore to be regarded as the unit set {< 0 J, 0 U >}; and a state of affairs – an event, excluding hereby the impossibility, the impossible event, the zero–element of the lattice – is a suitable union of possible worlds of that kind. A métaphysically possible world w.r.t. M 0S is a logically possible one which satisfies the métaphysical principles of M 0S , e. g. stating the disjunctivity of measuring intervalls as well as the reflexivity of the relations of equality. These principles of M 0S are, of course, not provable in M 0S in any strict sense; for they are synthetic ones in that particular language. But they are used in it – i. e. for the purpose of using M 0S – a priori in a pragmatic manner, i. e. as principles of M 0S which in M 0S are not es-
73
tablished by the means of this language via empirical procedures. These principles determine the a priori form of the observations, but of course not the empirical content of them. They do not even describe or even indicate how such an observation is to be performed, be it with the help of natural instruments like one’s own body or be it with the help of artificial instruments created according to already established theories. Therefore, in reconstructing the procedure of establishing worlds in empirical sciences after1 having developed that object language M 0S of observations, we have to look back onto all those intellectual actions we performed in creating M 0S and in determining with the aid of the means of M 0S its contents, its worlds in general and its métaphysically established factual world in particular; i. e.: we have to use the metalanguage M1S of M 0S and its contents, esp. its means for M 0S . In order to develop the syntax of M 0S within M1S , this metalanguage needs 4-ary relations on the first three levels at least and at most. Therefore, M1S is sufficiently rich to establish axiom systems for arithmetics and physical geometry as well as for empirical theories, as far as they are not overdeveloped in epistemological aspects. The syntactic as well as the semantic concepts and the concepts of classical logic may then be established in the usual way. These intellectual means enable us to investigate the operational meaning of each set of observation concepts, i. e. of sets of 1-ary predicates like {“F1i ”,...,“Fni ”} of M 0S , each of them denoting some division of the universe 0 U of M 0 S . By looking for the scheme of rules which scientists usually use to receive firm observations, we discover that these are rules according to the test-result-scheme2: The object to be observed has to be transformed into some preparatory state, be it either in waiting up to the time when this will happen, or be it by changing the circumstances to ensure that this will happen now; immediately afterwards, an instrument c which is associated to that preparatory procedure, will show the result, de1
The expression “after” is the literal translation of Greek “meta”! Using more poetical terms, scientists often tell or write “question to the nature – answer of the nature” instead of “test-result”.
2
74
signed as “ Fji ”. In rearranging arguments of Carnap and Suppes, it turns out that, for this scheme to be sound, two conditions are unavoidable: (1) The test has to be performed sometime in the past, present or future. (2) The testsituation must guarantee complete uniformity w.r.t. observer c ; i. e.: it is guaranteed that any other competent observer will perceive the same result. If that competent observer c is a person, he is supposed to have the usual physical abilities and a clear mind w.r.t. this test; whether or not he is in such a state may otherwise be tested. If this competent observer c is a machine, it is supposed to function like its prototype. These two conditions – and esp. condition (2) – are required of each performance of every rule according to that scheme, at least as an idealisation. It may happen that the testing situation differs to a certain amount from that idealisation, be it in sociology or be it in quantum mechanics. But in order to receive intersubjectively valid results – i. e. observations out of perceptions, resp. measuring values out of measuring results – the dependence of the result on the observer has to be eliminated by recurring to principle (2), i. e. to the principle of uniformity according to the observer. Then – presupposing the conditions (1) and (2) to be satisfied – the factual interpretation 0 J on 0 U w.r.t. “ Fji ” is the class of those individuals of 0 U that were, regarding the i-th kind of testing (=: Q i , i. e.: questioning) and the j-th result of that i-th kind of testing (=: R ij ), to be perceived by every competent d (=: Cp ) observer, i. e.:3 λ\ x Λ\ d e Cp : < x, d > e Q i ⇒ < x, d > e R ij
(the class of all the objects x such that every competent observer d who starts testing the i-th kind of attributes w.r.t. x will get the j-th result of that i-th kind w.r.t. this x ).
3
For the reason of simplicity the factor time is left out. If time is now added here, it has to be added already in advance to the predicates of 0 M S . In this case, M 0 S has to be established as a two-sorted language, where the last argument refers to time point and the (one or two) previous arguments refer to the individuals of 0 U .
75
If these expressions “ Q i ” and “ R ij ” of M1S would be used without any appropriate intensions, they would not lead to determined extensions; and then 0 J would not associate any determined extension to “ Fji ”. Therefore, the intensions of “ Q i ” and “ R ij ” have to be established in M1S , which is presupposed to be sufficiently rich for establishing syntax and semantics of M 0S . For this purpose, precise – i.e. categorical - theories of space and time as well as suitable parts of precise theories concerning those concepts “ Q i ” and “ R ij ” are needed here and are to be applied. Since the instruments which are involved here are composed of material parts, these empirical theories have to contain – or rather have to be completed – by theories of composition and partition4 in accordance with the theory of positive rational numbers. Within M 0S the concept “universe” – i. e. “ 0 U ” – was defined by the statement “ λ x : x = x ”. Within the metalanguage M1S of M 0S , the concept “universe” – i. e. “ 1 U ” – now got defined by “ λ\ x : x = x ”, too. “ 1 U ” denotes some universe of discourse 1 U which consists of two parts, i. e. the a priori part A1 U which contains the natural, rational and real numbers, and the empirical part E1 U which contains those entities which are determined by measuring theories belonging to empirical theories. These empirical entities are investigated via applying empirical means, mainly by conditioned perception according to the test-result-scheme, furthermore by objective probability as well as by epistemological probability, i. e. by subjective probability used for epistemological purposes. The strength of this metalanguage M1S is determined by the outcome of these logical and epistemological means. In starting to reflect about M1S , we stop using it and begin to look back on it by now using some suitably rich language M 2S : Using this eye of metametalanguage M 2S , the metalanguage M1S of the object-language M 0S is now viewed and analysed. The semantic part of M1S is discerned from its syntactic part; and the pragmatic factors are distinguished. Again, 4
By the way, even this theory which determines the syntax of M 0 S contains such a theory of composition and partition.
76
a possible world of M1S is determined as {} , i. e. as the unit class of an interpretation 1 J on the universe 1 U where “possible” is to be regarded in ontological and métaphysical aspects; and the factual world is to be found by applying the intensions of the predicates of M1S to the elements of 1 U . But the factual world of M 0S differs largely from the factual world of M1S ; that is what we now become aware of. Of course, these two worlds are not totally different from one another; but the former one now turns out to be a small and very gross part of the latter one, which in different aspects is more subtle and elaborated. Supposing that someone should now ask me: “Which world is the real one, the factual one of M 0S or the factual one of M1S ? ”, I then would have to answer: “None of them.5 For the frames of these worlds are manmade; they are therefore not real. And also the associations of the resp. frames and contents are manmade and are therefore not real.” But real is – that is what I am experiencing by using M 2S – the Lebenswelt, i. e. the world in which I – in using M 2S – am acting now.6 This Lebenswelt is the real world now; that is what I am completely sure of now. That is what I retain completely sure of as long as I keep using this level M 2S of reflection, as long as I do not start to reflect this level M 2S by using the language of a still more subtle, sophisticated and differentiated Lebenswelt.
5
Of course, by speaking precisely, I then should reply instead: “Real w.r.t what language? As long as I do not know the language which you are referring to, this question is as incomplete as the question: ‘Do you like Mary more than?’. Therefore, please, describe to me the language you are using whilst thinking and speaking; and afterwards I will try to describe the world – or at least its ontological and epistemological frame – to which you are referring in using that language! ” 6 In a very strict sense, this statement “The language M i S which I am using now” is neither true nor false but wrong and senseless; for in mentioning it I do not use it. Therefore, it has to be reformulated as such: “The language M i S which I was using up to now and which I will continue to use later on again”.
77
References Carnap, Rudolf “Testability and Meaning”, New Haven, Conn. 21954 Essler, Wilhelm K. “Analytische Philosophie I”, Stuttgart 1972 Gödel, Kurt „Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme“, in: „Monatshefte für Mathematik und Physik“ 38, 1931, 173-198 Suppes, Patrick „Introduction to Logic“, Princeton, NJ 61963 Suppes, Patrick “Studies in the Methodology and Foundations of Science: Selected Papers from 1951 to 1969”, Dordrecht 1969 Tarski, Alfred “Der Wahrheitsbegriff in den formalisierten Sprachen”, in: “Studia Philosophica” 1, 1935, 261-405
78
Comment on Essler’s “The World and the Worlds” Patrick Suppes∗
A rich set of ideas is introduced in Willy Essler’s short paper. There are many things to be commented on, but I will limit myself to three important points. First, I very much agree with his emphasis on the importance of measurement procedures in setting up our fundamental view of the world. The one concept that I found missing is the way in which we formally express how this is done. I prefer the approach that introduces measurement via isomorphism between empirical and numerical structures. This explicit isomorphism makes it clear that it is the structural isomorphism between the results of empirical measurements and various numerical structures. Of course, this isomorphism is only partial. I mean by this that the many excruciating empirical details of how measurement procedures are conducted will not be part of the isomorphism, but really only the results. It is important to recognize this distinction, because the proper detailed characterization, of such procedures goes well beyond anything that can be covered by this isomorphism. The second remark is also related to measurement but of a different nature. This concerns the important principle that Essler states as his second one, namely, that the results of measurement must have “complete uniformity with respect to observers.” The issue here is the interpretation of sameness of result. I think that a slight modification of this principle would fit the practice in physics better, namely, what is required is not sameness of results but an invariance of important physical quantities for different observers. This sounds too abstract. Let me give a particular example. In either classical physics or the physics of special relativity, inertial observers that are moving with respect to each other, that is those whose frames of reference have uniform velocities with respect to each other, certainly do not obtain the same measurement results in terms of the numbers they observe and record; for example, the velocity of particles. But what is certainly the case is that for any given theory there will be some invariant ∗
Stanford University
79
quantities, that is, some quantities that have exactly the same value for both observers. In the case of special relativity, it is well known that the proper time between any two space-time events such that one is after the other in some light cone, that is in other language, that the interval is time-like and not space-like, is a complete invariant. The technical meaning of “complete invariant” here is that such frames of reference must be related by Lorentz transformations. The detailed proof of this and the exact technical formulation is given in an old article of mine (Suppes 1959). A list of such invariance theorems that have some philosophical interest are given in my book (Suppes 2002). It is easy to find additional examples. I do think it is important to stress the difference between the literal meaning of same structure and the concept of invariance. So, to reformulate just to make the point clearly, invariance means that some mathematical combinations of the measurement results have the sameness required, which is usually not possessed by the original measurements themselves. My third point is the most general one and in some ways the most fundamental. Essler properly emphasizes how the notion of factual world is tied to the language we have available for describing this world. This is an important restriction on the way we talk about reality and the way we formulate our theories, but it does suggest a puzzle as to how we are to treat an extraordinarily important separate domain of human activity, namely, how we are to think about nonverbal actions. Like other animals, we have a wide repertoire of actions that we cannot describe with any accuracy verbally—the way we hit a tennis ball, the way we walk up stairs, the way we shake hands, the way we move our heads in response to an observation we agree with—none of these have detailed verbal descriptions in ordinary language and are in fact in many cases extraordinarily difficult to describe in completely accurate ways scientifically because, as we would put it technically, they often involve nonlinear equations in their description. But we feel comfortable with these actions. They are part of us and we do not think of them as being relative to any language and certainly they are not. So one of the puzzles is how we should think about this entire domain of action so important to us physically and mentally. It seems to me that in some sense we must avoid any strong relativity with respect to the language we use. The deeper point is my conviction that philosophical theories of action are much too concerned with the verbal descriptions that are possible, and not with the rich indescribable underlying life of the body and the mind. Our actions are driven by physical parameters, emotional feelings,
80
and mental computations of which we are only fleetingly aware. This is a story that would require considerable further amplification to make the case, that I do think can be made in a convincing way, for the priority of the nonverbal over the verbal in much of our daily lives.
References Suppes, P. (1959). Axioms for relativistic kinematics with or without parity. In L. Henkin, P. Suppes, & A. Tarski (Eds.), The Axiomatic Method with Special Reference to Geometry and Physics. Amsterdam: North-Holland, 291-307. Suppes, P. (2002). Representation and Invariance of Scientific Structures. Stanford, CA: CSLI Publications.
81
In Praise of The Representation Theorem Nancy Cartwright∗
1. Introduction This paper will take up three of Patrick Suppes’s favourite topics: representation, invariance and causality. I begin not immediately with Suppes’s own work but with that of his Stanford colleague, Michael Friedman. Friedman argues that various high level claims of physics theories are not empirical laws at all but rather constitutive principles, principles without which the concepts of the theory would lack empirical content. I do not disagree about the need for constitutive principles. Rather I think Friedman has mislocated them, and entirely at the wrong end of the scale of abstraction. It is representation theorems, as Suppes pictures them, that are the true constitutive principles, and that is true for theories far beyond physics. My disagreement with Friedman has two prongs. First, the high-level principles he calls ‘constitutive’ are not (at least in many cases). Second, representations theorems are. The first half of my claim depends on a quite different view of theory that I have from Friedman, a view I believe I share at least in part with Suppes. But it is not the central part of Suppes’s work I want to connect with today. Here I shall focus on the second half, in favour of representation theorems. I begin by explaining Friedman’s view and describing why I think representation theorems are better candidates than his own for constitutive principles. Then I shall illustrate by looking at a simple representation theorem in a subject in which Suppes has occasioned an important revolution, causality. The example I choose is a representation theorem that links causality with a second of Suppes’s favourite subjects, invariance. I want to illustrate the role of representation theorems as constitutive principles by showing that the theorem on invariance is indeed a presupposition for (one kind of) empirical meaningfulness of the abstract concept of a causal law. ∗
LSE and UCSD
2. Constitutive principles For the Kantian, part of the job of ensuring that our empirical knowledge fits the world of experience is done by the synthetic a priori. This provides the rational framework within which we experience the world. As Michael Friedman puts it, “synthetic a priori knowledge (typified by geometry and mechanics) …functions as the presupposition or condition of possibility of all properly empirical knowledge.”1 Friedman is keen to resurrect the role of the synthetic a priori, but not as a once-and-for-all framework necessary for empirical experience. Rather each proper theory in modern physics has its own framework that is held, relative to it, as a priori and that makes possible the genuinely empirical knowledge within that theory. Friedman’s principal examples are the three laws of Newtonian mechanics, which are a priori in his sense in the Newtonian scheme as currently understood. The law of universal gravitation – “that there is a force of attraction or approach, directly proportional to the two masses and inversely proportional to the square of the distance between them, between any two pieces of matter in the universe”2 – is the one empirical law in the scheme. This, he points out, talks about acceleration. Newton defined acceleration relative to absolute space. Since we do not believe in absolute space, we cannot do this. We say rather that the law of universal gravitation holds in any inertial frame, “where an inertial frame of reference is simply one in which the Newtonian laws hold (the centre of mass frame of the solar system, for example, is a very close approximation to such a frame).”3 This is why in our current rendering of Newtonian theory Newton’s three laws must be taken as a priori: It follows that without the Newtonian laws of mechanics the law of universal gravitation would not even make empirical sense, let alone give a correct account of the empirical phenomena. For the concept of universal acceleration that figures essentially in this law would then have no empirical meaning or application: we would simply have no idea what the relevant frame of reference might be in relation to which such accelerations are defined.4
1
Friedman (2001), 26 Friedman (2001) 36 3 Ibid. 4 Ibid. 2
84
A priori here can have a very local meaning: prior to the empirical laws, though perhaps posterior to other theories, claims, operations and definitions. Friedman argues for something more general. He maintains that these principles cannot be warranted in the same way that genuine empirical laws can. Whether that is so for the claims he counts as constitutive, those I want to count under this label – representation theorems – certainly are a priori in a straightforward sense: they are meant to be provable. According to Friedman constitutive principles, like those defining the frame of reference of the concept of acceleration, make the “empirical application of the theories in question first possible”.5 I think this is a misdescription. The principles that Friedman calls ‘constitutive’ make the concepts intelligible, not empirically applicable. They provide univocal and precise definitions that fit the concepts into the relevant high theory but they are not the principles that make possible the empirical application of these concepts. When it comes to the presuppositions for empirical knowledge, as opposed to presuppositions for fitting these concepts into a highly abstract theory, representation theorems are a far better candidate.
3. Representation theorems We represent features of the empirical world with specific mathematical forms that have specific properties. These forms tend to be far more universal across applications than is any (univocal) interpretation or definition of the related concept. Acceleration for instance: no matter what frame of reference we define it relative to, we almost always represent it as d2x/dt2. So length itself must be represented as a quantity twicedifferentiable with respect to time. But it also has a number of other builtin features as well. Probably the simplest is that length is represented by an additive measure. Can a mathematical representation with these characteristics adequately represent the phenomena to be associated with “length”?
5
Ibid. 49 This of course, as he points out, does not guarantee that the empirical principles that we formulate using them will be true, just that they are candidates for truth or falsity.
85
The answer depends on the structure of the phenomena to be represented. In the case of length, the phenomena might include what happens to sets of measuring rods. For instance, as Suppes puts it, the collection A of rods is longer than B “if and only if the set A of rods, when laid end-to-end in a straight line, is judged longer than the set B of rods also so laid out.”6 Formalizing that fact, along with a couple of other obvious features we ascribe to the empirical concept of length (for instance, that any collection of rods is at least as long as the empty set), we can characterize the structure consisting of the set of rods and the longer-than relation as a finite equally-spaced extensive structure. Now we are in a position to show that an additive measure is an appropriate representation for length by proving a representation theorem. In this case the theorem tells us that for any finite equally-spaced extensive structure, there is an additive measure µ such that for every pair of sets of rods, A and B, µ (A)≥ µ(B) iff A is longer than B. That is only a start of course. In order to guarantee the empirical applicability of the concept of length as we represent it in mathematical physics, we need a representation theorem relating all the qualitative features we assign to length to its mathematical representation. And similarly for all the quantities of empirical reality and their features for which we provide mathematical representations. It is, then, I urge, in the representation theorems we offer for the concepts in use in modern science that we find our best candidates for “constitutive principles”. These are the preconditions for the application of our concepts to empirical reality. Our representations are consistent with the features we ascribe to empirical reality only if the appropriate representation theorems are true.
4. Causality and invariance: a representation theorem I consider here only linear causal systems, where {V,≤,L,C} is a linear causal system iff
6
Suppes (2002), 64
86
V is a finite set of variables, x1,…,xn ≤ is a complete, antisymmetric, transitive ordering on V L is a finite set of linear equations over V of form xi = Σajxj, j≠i C⊂L such that the following axioms hold: 1. L-Consistency: Members of L are mutually consistent. 2. L-completeness: All linear combinations of members of L are in L. 3. Causal precedence and antireflexivity: xi = Σajxj ε C Æ xj