Cognitive science now and then
DOI:10.1093/acprof:oso/9780199228768.003.0006
Abstract and Keywords
This chapter analyses developments in cognitive science during the past fifty years from a personal perspective. It describes psychology issues and topics that were of personal interest to the author of this chapter. These include a paper titled ‘A logical calculus of the ideas immanent in nervous activity’ and Donald Broadbent's Perception and Communication, which not only used flowcharts to express models of cognition but also made a clear and explicit statement of cognitive science as abstracting computational procedures from the physical systems that implement them.
Keywords: cognitive science, logical calculus, neural activity, Donald Broadbent, Perception and Communication, cognition models
A science that is no better now then it was fifty years ago—that offers no greater illumination of its subject matter now than it did then—hardly excites. But that does not mean that we would like currently exciting science to have been worthless half a century ago. On the contrary: if some of the ideas favoured currently are ideas that were also favoured then, these have stood the test of time—which must testify to their worth, surely?
Are there any such ideas in cognitive science?
I was a feckless youth. Memory tells me that the reason I went to university from the small country town (population 4999) where I attended high school was because I had been offered a government scholarship that would pay me a living allowance sufficient to keep me alive for the three or four years in which my University studies were to be pursued. Which meant that I didn't need to think about having to get a job straight after I'd finished school.
As my best school subject was English, I chose without further consideration to study English at university; I thought also that studying English would have the advantage that it would allow me to become a novelist and to live the kind of life that my then favourite novelists lived, in the South of France, Paris, London, or Capri.
University regulations prescribed that I would have to choose two other subjects for first-year study. I chose philosophy (because to be studying philosophy seemed an impressive thing for a boy from the bush to be doing) and psychology (because that seemed something that a novelist ought be able to make use of).
English I at the University of Sydney in 1957 was a shock. The subject was taught in a desiccated and disdainful way. Never expressed was the idea that books were there to be read for enjoyment; books were for criticizing, not for enjoying. I found this dull and distasteful, and limped through to a Pass in the subject.
Psychology I and Philosophy I, on the other hand, were glorious. The Professor of Philosophy at that time, John Anderson,1 was in his last year (p.62) before retirement. He did all the wrong things as a lecturer: his Glaswegian accent was near to impenetrable, and he read his lectures straight from his lecture notes, looking up only occasionally. Yet his lecture course on the pre-Socratic philosophers was enthralling, and his last lecture received a lengthy round of applause from the students. I was astonished to learn that there were people thinking about thought two and a half millennia ago—and that we could today still know something about the views they had. No less astonishing was that it was possible to do scientific experiments on thought, as I learned in the practical classes in Psychology I. The study of English, and all ideas of being a novelist, were discarded, and the rest of my undergraduate career was devoted to a double major in psychology and philosophy.
Though there was a slight interruption: three months of National Service before returning to university. I don't need to describe what that was like, because Clive James has done so, unforgettably, in Chapters 14 and 15 of Unreliable Memoirs (James 1980). He and I shared that experience, as well as sharing Psychology I (which he thoroughly disliked) and English I (which was more to his taste than to mine).
John Anderson was not the only exciting philosophy teacher in the University of Sydney at that time. I learned a great deal from David Armstrong,2 John Mackie3 (themselves both much influenced by Anderson), and Alan Stout, son of G. F. Stout4 (Monro 1983). But the one person whose intellectual influence on me was enormous was a temporary lecturer and PhD student in the Department of Psychology, a man called Peter Kenny, whom I met in my second year at University.
Kenny was a little older than most PhD students, having previously spent some time training for the priesthood, although by the time I met him he had become an ex-Catholic (as had I, Anderson having done an effective job of corrupting this particular youth5). His speciality was psychometrics, and he and I published a paper on factor analysis while I was an undergraduate (Kenny and Coltheart 1960). As part of this work, I even did a little programming on (p.63) the first digital computer in an Australian University, SILLIAC.6 It was gigantic (2.5 m high, 3 m wide, and 0.6 m deep), primitive (paper tape was the input and output medium), and slow; but it was an electronic computer, and I was being allowed to program it. Eighteen months before I had been shooting rabbits in the bush and skinning them for dinner.
Kenny's interests turned out to be much wider than just psychometrics; crucially, he was particularly interested in what today would be called cognitive science.
Cognition was not part of the undergraduate curriculum at Sydney in those days, of course, since the Cognitive Revolution was only just beginning then; Broadbent's Perception and Communication appeared in the same year I met Kenny. But Kenny knew of and told me about much significant work in cognitive science that had already been published. He was studiedly insolent to his academic seniors, and a notorious libertine, and so (to my regret) soon vanished from the academic scene; but I read all of the material he recommended, and it fixed permanently the kind of work I subsequently did and still do.
I will mention here some of this work that I still particularly remember being deeply attracted to. The earliest was the paper by McCulloch and Pitts (1943) with the gorgeous title ‘A logical calculus of the ideas immanent in nervous activity’. Pitts was a philosopher of logic and McCulloch a neuroscientist. This was probably the first paper on artificial neural nets (and hence on computational modelling, which is something I am still doing). The possibility set out in that paper of abstracting computational procedures from the physical systems that implement them, and the implication that the same procedure can be run on many different kinds of physical system and so is independent of particular physical realizations, seems to me still to be one of the most important ideas in cognitive science—and I was exposed to this idea fifty years ago.
Then there was the Hixon Symposium (Jeffress 1951), which was held at Caltech in September 1948; the symposiasts came from psychology, philosophy, psychiatry, neurology, mathematics, biology, and zoology. Every paper in this volume is a classic, but one might single out Lashley's ‘The problem of serial order in behaviour’, McCulloch's ‘Why the mind is in the head’, and von Neumann's ‘The general and logical theory of automata’ (it is believed that the first occurrence of the term ‘black box’ was in von Neumann's paper: ‘I shall, therefore, view a neuron as a “black box” with a certain number of inputs that receive stimuli and an output that emits stimuli’ (von Neumann 1951, p. 22).
(p.64) Across the Atlantic ten years later, at Teddington on the Thames near London, a symposium on ‘Mechanisation of Thought Processes’ was held (National Physical Laboratory 1959). This book, which I read as avidly as I had the Hixon Symposium volume, also contains some classics in cognitive science, including Selfridge's paper on Pandemonium, Minsky's ‘Some methods of artificial intelligence and heuristic programming’, McCarthy's ‘Programs with common sense’ and Rosenblatt's paper on Perceptrons. Well over 200 people attended this meeting, including Sir Frederick Bartlett, John Morton, Stuart Sutherland, Horace Barlow, Richard Gregory, Fergus Campbell, Donald MacKay, Peter Ladefoged, and J. Z. Young.
And then there was Perception and Communication (Broadbent 1958). Not only did it use flowcharts to express models of cognition (see pp. 216 and 219 of the book), but it made a clear and explicit statement of cognitive science as abstracting computational procedures from the physical systems that implement them:
The proper relation between the physiologist and the psychologist may be regarded as analogous to that between the automobile mechanic and the test driver. Each of these men has his own domain of knowledge…for many purposes a knowledge of the mechanism is not essential to the driver; no more so than a knowledge of the problems of driving is essential to the mechanic. (In the scientific field, one may note, the psychologist usually knows much more of physiology than the physiologist does about behaviour.)Broadbent 1958, p. 306
Broadbent could have pursued his example further by pointing out that the driver does not care, nor need to know, what metal carburettors are made from, or even whether they are made from metal at all, whereas these are matters of intense interest to the mechanic—which illustrates the implication that the same processing procedure (in this case, aerating the petrol) can be run on many different kinds of physical system and so is independent of any particular physical realization.
I was as an undergraduate spared ultra-behaviourism. Learning was taught in the third year of the psychology course at the University of Sydney by a person who had obtained his MA from the University of Iowa under the supervision of Kenneth Spence, and could be described as an ultra-Hullian. But he was away that year on sabbatical leave, and the psychology of learning was taught instead by a visitor from the American University of Beirut, J. D. ‘Peter’ Keehn, a man of wide interests who introduced me to some thoroughly cognitive theorizing about animal behaviour—the work of Tolman, whose research interests would today be referred to as ‘Animal Cognition’, and whose paper ‘Cognitive maps in rats and men’ (Tolman 1948) counts as another early classic in cognitive science.
(p.65) Having pursued my undergraduate studies with mixed success (finishing top in the Psychology class in each of my first three years, but collapsing into Second Class honours in my fourth), I went on from being an undergraduate to being a PhD student. Why? Because I had been offered a postgraduate scholarship that paid me a living allowance sufficient to keep me alive for the three years in which my PhD studies were to be pursued. Which meant that I didn't need to think about having to get a job straight after finishing my first degree.
In the years when I was a PhD student (1962–1964) my thesis topic was the Perceptual Moment Hypothesis—the idea that psychological time is quantized, the quantum often argued to be something like 100 milliseconds. My interest in this came from reading John M. Stroud's 1956 chapter ‘The fine structure of psychological time’ (Stroud 1956).7 There must have been some serious morphic resonance going on here, because at this time both Alan Allport and Tim Shallice were working on exactly this topic for their PhDs.
Both Tim and Alan went on to publish their work on this (Allport 1968; Shallice 1964). I didn't. My three years as a full-time PhD student on a scholarship were spent mostly playing tournament and rubber bridge in clubs in downtown Sydney. I did find time to read a lot of cognitive science, and to do a desultory experiment or two, but I never saw my PhD supervisor and wrote nothing of my PhD thesis. In the last few weeks before my scholarship was due to expire—at the end of 1964—I therefore had nothing to show for it and no plans for employment. One day during those last few weeks, though, the head of my department, as he was passing me in the corridor, remembered that one of his staff members was about to go on sabbatical leave for the whole of the next year, and that he had forgotten to arrange a replacement. I was therefore asked on the spot whether I would like a temporary lectureship in the department for the year 1965. It was easy to accept, and that meant that I didn't need to think about having to get a job straight after finishing my PhD scholarship.
This lectureship was renewed for 1966. My main responsibility in those two years was lecturing to Psychology 1, and I seem to recall being required to deliver to these first-year students 63 hours of lectures covering every aspect of psychology. During this period I must have decided that I did need to have a PhD, because I embarked on a new PhD topic—the interaction of size and distance information in visual space perception—and during the two years I was a temporary lecturer I completed enough experiments (mainly at night) (p.66) to be able to write up later a (short) PhD thesis. This work was stimulated by Ross Day, then a Reader in the Sydney Psychology Department. He had taught me perception during my second undergraduate year, with a valuable historical perspective (it was from him that I learned about the Hemholtzian concept of unconscious inference in perception, an idea still important in current cognitive science). He was remarkably tolerant of youthful intellectual over-enthusiasm (my lab report after our practical class on Wohlgemuth and reversible perspective was 103 pages long, rather than the requested 10. It was returned to me imperturbably, with comments on most pages, but no comment on its length).
Although the work for my real PhD was on perception, in retrospect I can see that its flavour was highly cognitive—even in some ways computational. My subjects viewed a disc of light, monocularly, through a pinhole, in a completely dark room, without any visual cues as to how far away it was, and were asked to judge its distance from them. To ensure that the subjects could not even use knowledge of the size of the room to establish a maximum possible distance for the stimulus, I made certain that they did not know what room they were in by meeting them elsewhere, blindfolding them, leading them around the building until they were completely disoriented, and then into the room where the experiment was conducted. Hence the only information the subject had that was relevant to judging the distance of the disc was the size of its retinal image. That information is completely ambiguous: a big far object and a small near object would produce retinal images of the same size, despite being at different distances.
So I provided disambiguating information that was non-visual—it was either verbal size information (‘That disc you are seeing is three inches across’) or haptic size information (‘Here's a cardboard disc to feel: it is the same size as the disc you are looking at’). Combining information about the size of the retinal image with the second source of information about disc size allows distance to be unambiguously determined—in principle. But can people perform such computations? At that time, it was argued by many that people are not sensitive to the absolute size of the retinal image of an object and hence that absolute size can play no role in visual perception. If that were so, then varying the size of the retinal image while holding constant the non-visual size information would have no effect on the subject's judgement of distance: if the subject was told that the disc was six inches across, a 3-inch disc 10 feet away would be judged to be at the same distance as a 12-inch disc 10 feet away (as the only difference between these two conditions is the absolute size of the retinal image). If, on the other hand, subjects can use the absolute size of the retinal image to perform such computations, they will (p.67) judge the smaller disc to be more distant (it must be more distant since its retinal size is small relative to its objective size as specified in the non-visual size information). I did find such effects. This yielded a paper in Nature (Coltheart 1969).
In 1965, Ross Day was offered the Foundation Chair of Psychology at Monash University in Melbourne, and I accepted an offer of a lectureship in that department in 1967. Upon arrival I found the atmosphere electric. People were there at all hours of the day and night. Everybody was doing research. Everybody was publishing. Everybody was enthusiastic about their subject. I met Ken Forster. I wrote up and submitted my PhD thesis, the degree being awarded in June 1969.
Two months after that date I had left Monash. The siren songs from the South of France, Paris, London, and Capri had eventually proven too seductive. I didn't manage to move to Europe, but at least Waterloo, Ontario, was a lot closer to those places than Melbourne, Australia. I arrived at the Waterloo Psychology Department as a cognitive scientist of theoretical persuasion with a background in visual perception, and this generated a paper in Psychological Review, written while I was there (Coltheart 1971); but what I mostly did at Waterloo was work on a tachistoscopic report of unpronounceable random letter strings. I worked on that simply because there was an enthusiastic group already there working on it: Phil Merikle, Phil Bryden, and their graduate students. What they were doing looked like fun, and so I joined in. It was this experience that led to my later research interest in letter strings that are words, and letter strings that are not words but which are nevertheless pronounceable.
The next year, at last, I got to Europe: an extended European summer holiday trip including some time in (still slightly) swinging London. Europe was all I had hoped for and more, but it took me another two years to find a job there, a Readership at the University of Reading. John Morton and John Marshall had been at Reading but were gone by then; however, I found Alan Allport in post. Over the next three years I spent many absorbing hours in conversation with him about the foundations of cognitive science, his characteristically slightly jaundiced attitude helping to curb my characteristic over-enthusiasm. One product of our extensive interactions was one of the most embarrassing footnotes imaginable. This was to do with experimental work I did with my PhD students at Reading (notable amongst whom was Derek Besner) and published as Coltheart, Davelaar, Jonasson, and Besner (1977). This paper has had a large number of citations—about 585 at present—largely because of the experiment in it which investigated the effects of orthographic neighbourhood size on lexical decision. This variable is now often called ‘Coltheart's N’. The footnote says: ‘We thank Alan Allport for suggesting this (p.68) experiment and Nick Davison for running the subjects’. No one seems to have noticed this footnote. At least, no one has ever said to me, ‘Tell me, Max, if the experiment was Alan's idea, and the subjects were run by Nick, then your own contribution was…?’.
An offer of a Chair in the Psychology Department of Birkbeck College at the University of London meant that, at last, I would actually be able to live in London. A house was therefore bought in Tufnell Park, one street away from where Clive James had inhabited a squalid bedsit fifteen years before—he got to London long before I did—and I took up my new post in 1975. By then I had become immersed in work on the psychology of reading and in developing a dual-route model of reading based on ideas first put forward by Forster and Chambers (1973) and Marshall and Newcombe (1973).
As well as proposing some dual-route ideas, the Marshall and Newcombe paper made a claim that was completely new to me. It was that one might be able to learn something about how reading is normally accomplished by studying people in whom the processes of reading had been perturbed by brain damage. Their demonstrations seemed to me completely persuasive. So I wanted to start doing research immediately with people with acquired dyslexia. Or, as I soon appreciated, research with people with any acquired disorder of cognition. I realized this when, as I was interviewing Jane Riddoch for entry to the undergraduate programme at Birkbeck, she told me about dressing apraxia (she was a physiotherapist at the time).
I had no idea how to go about finding patients to work with, but when I mentioned this to Maria Wyke, a neuropsychologist at the Institute of Psychiatry whom I happened to know, she invited me to do some work with her and a patient of hers, KC. He was a charming elderly man of letters who had been at Oriel College in his youth, whose language was cultured and fluent, but who had an almost complete inability to repeat single words: a severe conduction aphasia. I thought it would be interesting to investigate the role of working memory in sentence comprehension with this man, and we did. Having completed an extensive set of experiments with him, I went to read the aphasiological literature on this topic, to find that everything I had done with him had already been done before, and thoroughly. I wasn't too downhearted: at least this showed that I knew how to do cognitive neuropsychology, a field of work to which I was by then deeply committed.
I was also fortunate at Birkbeck to encounter Morag Stuart. She had been a schoolteacher for many years, and eventually found that she needed a degree in psychology to advance further in her profession, so she enrolled at Birkbeck for that purpose. However, she never returned to teaching, having acquired a taste for research from her undergraduate studies at Birkbeck, and gone on to (p.69) do a PhD there. She is now a Professor in the University of London. Through being her supervisor I acquired some understanding of learning to read and developmental disorders of reading, and so extended my research interests in reading into these new areas.
I had left Sydney in 1967. Twenty years later it had become a different place—in 1987 it had become as glittering a destination as London had seemed to me in 1967. So I returned, to a Chair in the Psychology Department at Macquarie University, a position from which I retired last month. In fact, I had been on leave from that position since January 1 2000, when I became Director of the Macquarie Centre for Cognitive Science, a position in which I am continuing.
My work on reading continued apace at Macquarie, particularly in collaboration with my PhD students, Kathy Rastle (with whom I worked on computational modelling of reading; she is now a Professor at the University of London) and Anne Castles (with whom I worked on learning to read and developmental dyslexia; she is now a Professor at Macquarie University). But I also acquired a completely new research interest, the study of delusional belief, which was the subject of my recent Bartlett Lecture to the Experimental Psychology Society (Coltheart, 2007).
This happened because a student in the department who was working on schizophrenia, Robyn Langdon, lost her supervisor to another university, and asked me to take on this role. I demurred, explaining to her that I knew nothing about schizophrenia,8 but she said that she would teach me about that, which she has done, and continues to do. Shortly afterwards, the same thing happened again: a clinical neuropsychologist of my acquaintance, Nora Breen, told me about two remarkable men she had come across in her clinic who each had the fixed belief that the person he saw when he looked into the mirror was not him but some complete stranger who looked like him. She wanted to do a PhD on delusional belief including these men as part of her project, and wanted me to supervise her, assuring me that she would take responsibility for filling in all the gaps in my flimsy knowledge of neuropsychiatry and clinical neuropsychology. This she did with great success. This work resulted in the formulation of a two-factor theory of delusional belief (see e.g. Coltheart 2007; Coltheart Langdon and McKay, 2007), according to which delusional beliefs arise when two cognitive abnormalities are present simultaneously. The first of these is what causes the delusional thought to occur to the (p.70) person in the first place, and the second is an impairment to the normal processes of belief evaluation which, if intact, would have allowed the person to reject the thought that had occurred to them rather than accepting it as a belief. Each of these two impairments is, of course, associated with some particular form of brain damage, but the theory has been described, and tested, at a level that abstracts away from such facts about the neural implementation of the relevant cognitive systems.
Cognitive science then and now
I'm happy to see that certain ideas about cognitive science that I liked so much almost fifty years ago are still popular The idea that computational procedures can be abstracted from the physical systems that implement them, the implication that the same procedures can be run on many different kinds of physical system and so are independent of particular physical realizations, and the consequence that the operation of any such procedure can be described at (at least) two different levels, are still very strong in cognitive science.
That is not to say that there are not barbarians at the gate; there are, and they belong to several different tribes.
There are the eliminative materialists (e.g. Churchland 1981), whose view is that there is no such thing as cognitive science; there is only cognitive neuro-science. They identify cognitive psychology with folk psychology, and argue that the more we learn about the brain the less we will need folk psychology to discuss cognition, until eventually folk psychology will have withered away and the folk will use only brainspeak to discuss cognition.
Then there are the connectionists, currently too numerous to mention. They reject the idea that computational procedures should be studied by abstracting them away from the physical systems that implement them. The models of cognition they offer are described as ‘brain-like’ or ‘neurally plausible’, and elements of their models are often described as ‘neurone-like’. So their models are not attempts at describing cognition in a way that abstracts away from the physical system upon which cognition depends.
And then there are contemporary neo-behaviourists of a Wittgensteinian persuasion such as Bennett and Hacker (2006), who repudiate the very concept of mental representation; these authors indeed assert that any attempt at describing mental activities in the form of flowcharts or box-and-arrow diagrams is ‘misleading’, ‘confused’, ‘a mythological redescription’, ‘incoherent’, ‘a mythical tale’, and ‘risible’ (this is just a selection from the numerous derogatory epithets sprinkled throughout their paper). And this paper doesn't (p.71)hesitate to name names: amongst those whom the paper accuses of risible myth-making are Willem Levelt, John Morton, and Anne Treisman.
But these gates, I am delighted to see, are still being defended fervently by such luminaries of cognitive science as Ned Block, Saul Sternberg, and Jerry Fodor.
Cognitive scientists often say that the mind is the software of the brainBlock 1995, p. 21During the second half of the century there was a change in the kinds of questions that psychologists asked and in the acceptable answers. This change was influenced by the growth of computer science, which persuaded psychologists that programming concepts might be acceptable as precise descriptions of information processing by people as well as by machines. Also the software–hardware distinction supported the legitimacy of theories couched in terms of abstract information processing operations in the mind rather than only neurophysiological processes in the brain. In the “human information processing” approach, complex activities of perception, decision, and thought, whether conscious or unconscious, came to be conceptualized in terms of functionally distinct and relatively independent (“modular”) sub-processes responsible for separate operations such as input, transformation, storage, retrieval, and comparison applied to internal representations—modules whose arrangement was expressed in systematic flow charts.Sternberg (submitted)It isn't, after all, seriously in doubt that talking (or riding a bicycle, or building a bridge) depends on things that go on in the brain somewhere or other. If the mind happens in space at all, it happens somewhere north of the neck. What exactly turns on knowing how far north? It belongs to understanding how the engine in your auto works that the functioning of its carburetor is to aerate the petrol; that's part of the story about how the engine's parts contribute to its running right. But why (unless you're thinking of having it taken out) does it matter where in the engine the carburettor is? What part of how your engine works have you failed to understand if you don't know that?Fodor 1999, p. 69
It would have been interesting to have been present when the engineer Broadbent and the philosopher Fodor opened the bonnet of Fodor's car and began a discussion of carburettors, carburetion, and cognitive science.
No comments:
Post a Comment