3. Conditioning and Associative Learning N. J. Mackintosh
4. Visual Masking B. G. Breitmeyer
5. The Musical Mind J. A Sloboda
6. Elements of Psychophysical Theory J.-C. Falmagne
7. Animal Intelligence L. Weiskrantz
8. Response Times R. D. Luce
9. Mental Representations A. Paivio
10. Memory, Imprinting, and the Brain G. Horn
11. Working Memory A. Baddeley
12. Blindsight L. Weiskrantz
13. Profile Analysis D. M. Green
14. Spatial Vision R. L DeValois and K. K DeValois
15. The Neural and Behavioural Organization of Goal-Directed Movements M. Jeannerod
16. Visual Pattern Analyzers N. V. S. Graham
17. Cognitive Foundations of Musical Pitch C. L. Krumhansl
18. Perceptual and Associative Learning G. Hall
19. Implicit Learning and Tacit Knowledge A. S. Reber
20. Neuromotor Mechanisms in Human Communication D. Kimura
21. The Frontal Lobes and Voluntary Action R. Passingham
22. Classification and Cognition W. K. Estes
23. Vowel Perception and Production B. S Rosner and J. B. Pickering
24. Visual Stress A. Wilkins
25. Electrophysiology of Mind Edited by M. D. Rugg and M. G. H. Coles
26. Attention and Memory N. Cowan
27. The Visual Brain in Action A. D. Milner and M. A. Goodale
28. Perceptual Consequences of Cochlear Damage B. C. J. Moore
29. Binocular Vision and Stereopsis I. P. Howard and B. J. Rogers
30. The Measurement of Sensation D. Laming
31. Conditioned Taste Aversion J. Bures, F. Bermúdez–Rattoni, and T. Yamamoto
32. The Developing Visual Brain J. Atkinson
33. The Neuropsychology of Anxiety, 2e J. A. Gray and N. McNaughton
34. Looking Down on Human Intelligence I. J. Deary
35. From Conditioning to Conscious Recollection H. Eichenbaum and N. J. Cohen
36. Understanding Figurative Language S. Glucksberg
37. Active Vision J. M. Findlay and I. D. Gilchrist
38. The Science of False Memory C. J. Brainerd and V. F. Reyna
39. The Case for Mental Imagery S. M. Kosslyn, W. L Thompson, and G. Ganis
40. Seeing Black and White A. Gilchrist
41. Visual Masking, 2e B. Breitmeyer and H. Öğmen
42. Motor Cognition M. Jeannerod
43. The Visual Brain in Action A. D. Milner and M. A. Goodale
44. The Continuity of Mind M. Spivey
45. Working Memory, Thought, and Action A. Baddeley
46. What Is Special about the Human Brain? R. Passingham
47. Visual Reflections M. McCloskey
48. Principles of Visual Attention C. Bundesen and T. Habekost
49. Major Issues in Cognitive Aging T. A. Salthouse
50. Perceiving in Depth Ian P. Howard
51. The Neurobiology of the Prefrontal Cortex: Anatomy, Evolution, and the Origin of Insight Richard E. Passingham and Steven P. Wise
52. The Evolution of Memory Systems: Ancestors, Anatomy, and Adaptations
Elisabeth A. Murray, Steven P. Wise, and Kim S. Graham
53. Understanding the Prefrontal Cortex: Selective advantage, connectivity, and neural operations Richard E. Passingham
54. Remembering: An Activity of Mind and Brain Fergus I. M. Craik
Remembering
An Activity of Mind and Brain
FERGUS I. M. CRAIK
3
Great Clarendon Street, Oxford, OX2 6DP, United Kingdom
Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries
All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above
You must not circulate this work in any other form and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2021930831
ISBN 978–0–19–289522–6
DOI: 10.1093/oso/9780192895226.001.0001
Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.
To
Anne
Preface
This book has had a long gestation period. I first had the idea of writing a book on memory when I applied for a Killam Research Fellowship in 1981. These excellent Fellowships are awarded by the Canada Council for the Arts to Canadian scholars from universities and research institutes “to pursue groundbreaking research” for two years, untrammeled by normal teaching and administrative duties. In my application I stated my intention to write a “groundbreaking book on memory” while spending a year at the Center for Advanced Studies in the Behavioral Sciences at Stanford University, and this aspiration seemed to impress the Killam selection committee. So I spent an interesting year at the Stanford Center, writing up completed studies, lunching in the Californian sunshine, discussing aspects of memory theory with a congenial group of researchers (see Chapters 2 and 7), but unfortunately – no book! The intention then lay dormant for several years, until I began thinking about it seriously again in early 2017. By this time the idea had morphed from a rather vague hypothetical construct to a more definite plan. My career as an experimental psychologist was clearly drawing to a close, so it was appealing to write up many of the studies I had done over the years with a view to highlighting the theoretical ideas that lay behind the experiments, how the ideas had evolved, and how they made contact with current concepts in cognition and cognitive neuroscience. Progress over the next couple of years was somewhat sporadic, but in March 2020 the Covid-19 pandemic arrived in Canada. Like everyone else I was confined to quarters – nothing to do but write! – and so the 40-year project was finally completed.
The book is partly a record of my own experimental work in memory, attention and aging research over 50-plus years, augmented by descriptions and discussions of relevant studies by colleagues and collaborators, and partly a discussion of how theoretical ideas have changed over the years with the advent of new technologies and new concepts from neighboring disciplines. So it is not a comprehensive textbook of memory; rather, the emphasis is on empirical issues and theoretical puzzles that have caught my attention from the mid-1960s to the present day. The main theme is the idea that human memory should be thought of as a mental activity rather than as a ‘thing in the head’ – as the verb ‘remembering’ rather than the noun ‘memory.’ The book’s title Remembering: An activity of mind and brain intends to capture this central point, and I have traced my own thinking along these lines from the cognitive work on levels of processing with Robert Lockhart and Endel Tulving to current collaborative work with colleagues at the Rotman Research Institute involving neuroimaging and neural networks. I should emphasize that
although the word ‘brain’ is in the title, the contents focus very largely on cognitive, behavioral studies. However, one cheering aspect of current work in cognitive neuroscience is the extent to which constructs emerging from cognitive studies –attention, working memory, perception/memory interactions etc. – are guiding and organizing current work on brain processes.
With its focus on my own work (and its implications for the rest of the field) the book is therefore modeled on such previous classics as Endel Tulving’s 1983 book on episodic memory and Alan Baddeley’s books on working memory, in the sense that it is a personal view of attention, memory and learning. In greater detail, the book starts with a description of the ideas and experiments that led to the 1972 article on levels of processing by Craik and Lockhart and to the 1975 empirical article by Craik and Tulving. From my perspective these ideas arose directly from work on attention by the English psychologists Donald Broadbent and Anne Treisman, so their seminal ideas are sketched briefly. Chapter 2 discusses some historical and more recent ideas on remembering as an activity of mind. One less familiar source in this context is the Russian work stemming from the activity theories of Vygotsky and his colleagues. The work on memory from this perspective by such researchers as P. I. Zinchenko and A. A. Smirnov in the 1940s and 1950s is really quite similar to the ideas proposed some decades later by Craik and Lockhart (as emphasized by some trenchant Russian critics!), although the starting points for the two sets of ideas were quite different. Chapter 3 is an attempt to flesh out my ideas on remembering as a processing activity, and so sets the scene for work discussed in the following chapters.
Chapters 4 and 5 deal respectively with the activities of encoding and retrieval, although it quickly becomes apparent that the two sets of processes are intimately related and simply cannot be described in isolation; retrieval processes are in many ways the mirror image of encoding processes. Chapter 6 covers experiments and theoretical ideas on short-term memory, and how the focus in this area turned progressively to the more current notion of working memory. Chapters 7 and 8 describe some experiments and concepts in the area of cognitive aging, again with a focus on age-related differences in remembering. This topic has been a strong second interest of mine throughout my career, so a lot of ground is covered. I therefore split the descriptions into two chapters, in order not to tax the reader’s patience too severely. Chapter 7 mainly sets out a number of theoretical ideas, while Chapter 8 focuses more on empirical studies. After a good deal of Scottish ‘swithering’ I finally decided to include a chapter on the brain (Chapter 9). Early reviewers were divided in their advice on this point, with some suggesting that “Craik should stick to what he actually knows about!” and others pointing to the strong emphasis on neuroscience in current work on memory. I have been involved as a collaborator on many neuroimaging studies since the early 1990s, so I eventually decided to include the chapter. Finally, Chapter 10 draws several research lines
together, summarizes the main arguments, and hopefully makes contact with a number of current issues in human memory research.
In overview, the book is essentially a research monograph focusing on work in my lab over 50 years or so, and how that work relates to other ideas and findings in the field of human memory research. It does not attempt to cover studies of human memory comprehensively, but will hopefully involve readers in a wide range of interesting topics. Although a number of the experiments described are ‘historical’ at this point, I believe that the main ideas are still very current. The book is aimed principally at research colleagues, from graduate students to seasoned professionals, but I have also tried to keep the descriptions accessible, so I am hopeful that more general readers may catch some of the excitement associated with the findings and ideas. The tone is informal and somewhat ‘chatty’ – my normal writing style, for better or for worse!
The work described in this book could not have been carried out without the support from many individuals and institutions. My lab work on attention and memory has been generously supported over many years by the Natural Sciences and Engineering Council of Canada (NSERC), augmented by funds from the Killam Trusts and from the University of Toronto. I have greatly benefited over the years from the stimulation provided by the rather grandly entitled Ebbinghaus Empire; a research seminar on topics in memory and cognition, meeting weekly in the Psychology Department of U of T. When I first arrived in Toronto the stern coEmperors were Ben Murdock and Endel Tulving, presiding over a respectful and mostly silent group of graduate students and postdocs. The silence (although not of course the respect) was finally broken in the early 1970s with the addition of Bob Lockhart, Morris Moscovitch and others to the group, and the ‘EE’ remains a terrific source of new ideas. I also profited from similar groups in London – a seminar organized with Tim Shallice from UCL – and at Erindale College; a joint ErindaleMcMaster group run by Larry Jacoby and myself.
With regard to the many individuals who have helped to formulate my ideas and to guide my career I can pick out several who have been particularly influential. This group includes Ellen Bialystok, Larry Jacoby, Betty Ann Levy, Robert Lockhart, Morris Moscovitch, Moshe Naveh-Benjamin, Donald Stuss, Endel Tulving and Boris Velichkovsky; I am truly grateful for their support, wise guidance and inspiration over the years. At the present time I am fortunate to have outstanding colleagues at the Rotman Research Institute at Baycrest in Toronto. It is a very interactive group of scientists, and I have collaborated on published papers with many of them. I have enjoyed working with them all and learned a lot from our interactions.
I have also been extremely fortunate in the students, postdoctoral fellows, research assistants and visitors who have made up my labs in London and Toronto. I have particularly fond memories of my lab at Erindale College in the late 1970s and early 1980s. We called the lab ‘The LMR’ for obscure reasons. It was a very
productive group, with many ideas for experiments thrashed out in the local pub! Other lab members who have greatly influenced my thinking over the years include my excellent graduate students at Birkbeck College in London, whose work is described in the book; also the graduate students, postdoctoral fellows and international visitors who have populated my labs at Erindale College, the main campus at U of T, and the Rotman Institute. I am greatly indebted to them all.
The book’s final form has been greatly influenced by the heroic efforts of three good friends; Ellen Bialystok, Larry Jacoby and Bob Lockhart, who each read the complete draft, chapter by painful chapter. I am immensely grateful for their wise counsel at every level from commas to concepts; the finished product is hugely improved by their suggestions. I also received some very helpful guidance from several anonymous reviewers appointed by potential publishers; I very much appreciated their (mostly!) kind comments. Two reviewers who revealed their names are Roddy Roediger and Morris Goldsmith. I am grateful to them both. I can add that Morris Goldsmith went way beyond the call of duty by producing 20 singlespaced pages of detailed comments and suggestions. A jaw-dropping effort, and very much appreciated.
I am very grateful to Kevin Tang, a research assistant in my lab, who did a splendid job of compiling the long list of references. And I must say a special word of thanks to Jennie Sawula who typed up the great bulk of the manuscript. I have to confess to a medieval style of scientific writing – I still write everything in longhand, barely a step beyond scratching it out laboriously with a quill pen on vellum. However, I then ingeniously harness modern technology by FAXing each completed sheet to Jennie, who deciphers my scrawl, types it up on Word and sends it back by email for final editing. Bizarre but effective! Thank you Jennie!
I am also very grateful to the team at Oxford University Press who steered me through the publication process with commendable efficiency and graciousness. My first helpful contact was with Martin Baum and James Oates; production was handled by Karen Moore, with copyediting by Jayne MacArthur and indexing by Sue Leech. I am greatly indebted to all of them.
Above all, I have to thank my family for putting up with my endless scribbling at times when I should have been out throwing footballs with my son, taking my daughter to ballet lessons, driving my grandchildren to the beach, or whatever it is that better fathers and grandfathers do. When my children were teenagers, a friend asked them “Is your father an actual workaholic?” Now things will be different! Finally, this book is dedicated to my wife Anne, who graduated from being a subject in one of my early experiments to being a wife, mother and loving companion on trips to the opera and walks in the Alps. Thank you for everything.
The major purpose of this book is to examine the proposition that human memory should be regarded as an activity of mind and brain. This perspective stands in contrast to accounts in terms of structures; as items and associations for example, or as memory stores. The present approach is thus a processing account—remembering as a set of processes—with encoding and retrieval processes viewed as activities that can be described in both behavioral and neural terms. A secondary theme is to explore how far it is possible to stretch the idea that remembering is basically a form of perceiving and thinking, with encoding processes being nothing more than the normal activities of perceiving, comprehending, thinking, and deciding, and with retrieval processes being essentially an attempt on the organism’s part to recapitulate the same pattern of processing that took place during encoding. I would like to believe, in fact, that the processing view can be extended to cognition generally, although the present account will focus more narrowly on attending and remembering.
The view of human memory that I will set out suggests that (following Bartlett, 1932) we should talk of remembering as a mental activity rather than memory as a stored record. Along with many other researchers I will emphasize the similarity between encoding and retrieval processes—basically that retrieval is an attempt to recapitulate encoding—and that both sets of processes are, in turn, similar to the processes of perception and comprehension. I will argue, in fact, that “memory” consists essentially of modifications to these perceptual and conceptual systems, especially at higher levels of analysis. Whereas perceiving obviously reflects an interaction between incoming sense data and existing analytic and interpretive processes, my proposal is that remembering does also—although with a focus on the evoked internal representations rather than on the interpretation of external sense data. According to this view retrieval does not consist of a search through many stored records (the library analogy), but is the product of an interaction between current input in the form of sensory information plus questions, cues, and context, and the previously modified perceptual/conceptual analytic mechanisms (my version of “the memory trace”). My colleague Bob Lockhart suggested that a more appropriate analogy is the example of a tree falling in the forest—does it
make a sound if nobody is present? At first it seems obvious that the answer is “yes”—surely the sound waves occur regardless of the audience. But the waves by themselves make no sound, so the correct answer is “no”—the experience of a sound requires an interaction between the physical sound waves and the auditory sensory system of a listener, just as remembering requires an interaction between a question or other input and the existing analytic mechanisms. Hopefully these rather arcane ideas will become clearer as the chapters unfold!
I should also say a few words about my use of the word “remembering,” which is prominent in the book’s title and throughout the chapters. Remembering refers to the activities involved in retrieving information of different types and for different purposes; it does not have to be deliberate, however, as “involuntary remembering” clearly occurs (Berntsen, 2010; see Chapter 5). It does typically refer to the conscious awareness of retrieved information, and this is the sense in which I mostly use the word. However, this usage apparently excludes such important categories as implicit and procedural memory, which I certainly intend to include in the general framework of remembering as an activity of mind. So, for the most part, I will use the word to connote the conscious experience associated with the retrieval of facts and events, but I also intend “remembering” in a looser sense to refer to the activities associated with the retrieval and use of all types of encoded information.
In general terms the book will set out what I believe about memory after studying it for many years. It is thus partly an account of some of my own studies as they relate to the central theme of “memory as activity” and partly a critical assessment of other studies—selected on the basis of their relevance to the overall theme. I will embed these descriptions of experiments, critical accounts, and theoretical speculations in a first-person autobiographical framework in order to provide context in terms of the people, places, and events that have influenced and shaped my thinking over the years.
The book starts with an account of the levels of processing framework proposed by Robert Lockhart and myself in 1972, followed by an assessment of how these ideas have fared in the light of subsequent theoretical and empirical advances. A second chapter describes a number of precursors to the levels of processing ideas, and examines how they fit the central theme of remembering as an activity. Further chapters deal with encoding and retrieval processes in greater detail, with the transition of the concept of short-term store to that of working memory, with agerelated changes in memory, and with its neural correlates. A final chapter attempts to integrate the various findings and ideas into a coherent big picture account.
Levels of processing
In the late 1960s and early 1970s I was a faculty member in the Psychology Department of Birkbeck College, which is part of the University of London. British
psychology at that time was considerably different from psychology in North America—largely because the behaviorist movement and its influence on theories of animal and human learning had never fully caught on. In Britain, the emphasis in experimental psychology was more on perception, motor skills, and higherlevel cognition. The neuropsychological study of clinical patients was also prominent. Models of attention were popular—largely influenced by the seminal work of Donald Broadbent, whose ideas were set out in his 1958 book Perception and Communication. Broadbent’s “filter theory” of attention was formulated to address the practical problem of errors made by machine operators working under conditions of information overload, specifically errors made by pilots of World War II aircraft. The errors were first attributed to carelessness or to inattention on the part of pilots, but Broadbent’s view was that the errors reflected an inability of the human brain to deal simultaneously with a number of competing sources of information, regardless of whether the inputs were all visual or were visual and auditory. Broadbent argued that the brain has a limited capacity to process incoming streams of information and that an attentional mechanism selects one input channel and necessarily ignores others. As shown in Figure 1.1, Broadbent’s filter theory proposed that incoming sensory information undergoes initial processing and is then held in a preconscious s-system. I assume that the s-system is more precisely a set
System for varying output until some input is secured E ectors
Store of conditional probabilities of past events
Fig. 1.1 Broadbent’s filter theory (Broadbent, 1958, p. 299). All stimuli impinging on the senses enter the short-term store (s-system); the attentional filter selects one channel to be processed further in the limited-capacity p-system, which may be equated with primary memory and conscious awareness. Information in the p-system may be recirculated to the s-system; information also makes contact with knowledge in long-term memory (store of conditional probabilities of past events), and with potential actions back into the environment via appropriate effectors.
Reproduced from Broadbent, D. E., Perception and Communication, Pergamon Press Ltd. 1958 with permission from Elsevier.
of modality-specific sensory stores, but the main point is that all incoming information is partially analyzed and held briefly at that level. So the s-system is preattentive and has a large capacity but a rapid forgetting rate.
Broadbent’s proposal was that an attentional mechanism then selects one incoming “channel” of information and passes it on for further processing and conscious awareness in the p-system (“p” for perception, I assume). This selective filter was assumed to act in an all-or-none manner—that is, it is “tuned” to select one channel and ignore all others. The basis for selection could be top-down (e.g., “listen to the right ear and ignore the left ear”) or bottom-up (e.g., the selection switch might be attracted by a loud, bright, or otherwise salient stimulus stream). Once in the p-system the sensory information has access to previously acquired knowledge (long-term memory) and acquires meaning through rapid transactions with that knowledge. Broadbent also suggested that the information in the p-system (and thus “in mind”) could also be rehearsed before being used to formulate a response and that rehearsal involves a recirculation of material through the s-system and back to the p-system.
The filter theory gave a good account of experiments involving two ears (dichotic listening) and also eye and ear (Broadbent, 1958), but was questioned by an observation made by the English psychologist Neville Moray. He found that when participants were fully engaged in “shadowing” (repeating back) auditory information presented to one ear, they had little or no knowledge of messages presented to the other ear—in line with Broadbent’s model. If the unattended message contained the participant’s own name, however, it was perceived on roughly one-third of occasions. This should not happen if the filter acts to block all unattended messages (Moray, 1959). The same problem was illustrated by a simple experiment reported by two undergraduates (Gray and Wedderburn, 1960). They showed that if two meaningful word triplets, such as “Who goes there?” and “My Aunt Jane” were presented dichotically so that the meaning switched from ear to ear, e.g., “Who Aunt there” in the right ear and “My goes Jane simultaneously in the left ear, participants reported hearing either “Who goes there?” or “My Aunt Jane” without realizing that they had switched ears during the message. Somehow meaningfulness switched the filter before the messages entered the p-system, and this just could not happen with an all-ornone switch located before meaning was assigned. An everyday parallel to this ingenious experiment is the experience of attending closely to one speaker (for example, at a “cocktail party”—much frequented by experimental psychologists in the 1950s it seems!), yet plainly hearing a highly significant fragment of conversation (the listener’s own name, for instance) from a different speaker, whose conversation had otherwise not been heard or understood. Again, meaningfulness appears to act in a bottom-up manner, which gives problems to many models of attention viewed as a series of stages, with meaning allocated at a relatively late stage.
One solution is the suggestion that all inputs are analyzed to the level of meaning, and that selection takes place only after this full analysis. Such socalled “late-selection models” (e.g., Deutsch and Deutsch, 1963) seem biologically wasteful, however—given that most deep analyses of meaning will simply be discarded. A middle way was proposed by Anne Treisman in a series of papers (Treisman, 1964a,b). She suggested that selection takes place throughout the process of analysis, with only attended, salient, or highly meaningful messages surviving a progressive set of analytical stages to reach conscious awareness. The essence of Treisman’s model is sketched in Figure 1.2. It comprises a series of “levels of analysis” running from shallow sensory analyses through intermediate analyses (e.g., of phonemes or simple visual forms) to deeper meaningful analyses of words, sentences, objects, and scenes. Each level of analysis functions as a pass–fail test, specific to that particular input. In Treisman’s model conscious perception is
Fig. 1.2 My interpretation of Treisman’s (1964a) model of attention. Incoming stimuli enter the processing system with various degrees of physical strength; they are then processed to various “depths of analysis” depending both on their initial strength and on their probability of occurrence in the current context. In the diagram, stimuli a, b, c, d, and e have registered in sense organs with various degrees of strength (shown here as line thickness). At the lexical level, the probability of a word’s occurrence before stimulus onset is shown by length of line (e.g., A has a very low probability, E is highly probable). Stimulus a is processed deeply and activates A due to its incoming strength. Stimulus e is weak but is “drawn in” by the high probability of its corresponding word’s occurrence. Stimuli b, c, and d are partially analyzed to various depths; the level attained determines which aspects of the stimulus are consciously perceived.
a correlate of the analyses being performed; that is, partial (typically early) analyses will be perceived even if the stimulus is not fully analyzed. So a listener may be aware that input from a second speaker is present on an unattended channel (e.g., the left, unattended ear in a dichotic listening experiment), and may even be aware that the speaker is a woman, yet have no awareness of the meaning of the unattended speaker’s conversation.
Treisman further suggested that two sets of factors determine the fate of incoming stimuli. One set works bottom-up, and includes signal strength variables such as loudness and brightness. Paying attention to a particular input channel also has the effect of increasing the effective signal strength of stimuli on that channel. The second set works top-down and essentially reflects the probabilities of occurrence at that time of various possible stimulus inputs. Some stimuli are treated as permanently important to the person (e.g., the person’s own name), others as important under specific circumstances (e.g., expecting a telephone call or being aware of a fretful baby), and yet others are given a high probability rating by a temporary context (e.g., after hearing “the boy leaned out the _ ?”, the word “window” becomes salient). Treisman proposed that the pass–fail tests at each level of analysis act like tests in signal detection theory (Swets et al., 1961) in which the criterion (ß) is set by temporary or long-lasting probabilities of occurrence, and signal strength (d’) is set by the physical qualities of the input stimulus and by the deployment of attention to that channel. An incoming stimulus item will thus “pass the test” and proceed to deeper tests either if it is a strong signal or if the criterion “pass mark” is set to a lenient level.
By this scheme, incoming stimuli will be processed to a shallow or deep level according to the interplay of these two factors. Strong signals will “bulldoze” their way through to full conscious awareness, regardless of how probable or improbable they are at that moment, and highly probable stimuli will also be pulled through successive levels of analysis by virtue of top-down influences setting appropriate criteria at favorably low levels. Figure 1.2 depicts a situation in which several sources of stimulation are impinging on the system at one time. All inputs are registered at the earliest sensory levels of analysis, but they then proceed to deeper levels depending on their signal strength and probability of occurrence at that moment. Strong signals pass successive tests despite the fact that appropriate analytic criteria are not set in their favor but may not get through to a full analysis of meaning. Weak signals may fail early in the analytic sequence or proceed to a full analysis if their contents are probable in the current context. The perceptual end result is thus typically one in which messages from an attended source are fully analyzed and comprehended, while others are only partially analyzed but are nonetheless perceived in terms of their sensory characteristics.
I found one of Treisman’s (1964c) experiments particularly intriguing. She played identical speech messages to the left and right ears in a dichotic listening paradigm but staggered in time so that the messages arrived at the two ears several
seconds apart. She had participants “shadow” the speech coming in one ear; that is, the participant repeated back out loud the words coming in the attended ear. Treisman then brought the unattended message closer in time to the shadowed message and noted how far apart the messages were when the participant realized that the same message was being played to both ears. The result depended on whether the shadowed (attended) message was leading or lagging in time; if it was leading, participants recognized that the unattended message was the same when it occurred 5 seconds later. But when the shadowed message followed the unattended message, the two messages had to be brought within 1.5 seconds before participants realized that they were the same. Apparently, the full analysis of the attended message had the effect of tripling its survival time in short-term memory to enable a match to be made.
This result suggested to me that memory and attention are very closely interlinked. Not only the everyday observation that to remember something well you must pay attention to it, but the more precise idea that the durability of the encoded memory record might depend on how deeply analyzed the stimulus had been. Perhaps memory processing was essentially the same thing as processing for attention and perception.
This thought was followed by some back-of-the envelope sketches, some shared with my graduate students whose reactions were somewhat less than wildly enthusiastic. My student Michael Watkins, who went on to have an illustrious career at Princeton and Rice Universities, remembered one such occasion in the following terms (Watkins, 2001):
A little more than three decades ago, a singular young researcher joined his students at lunch and mentioned that he had been thinking. His students harkened at once, for Fergus Craik is a modest man, not given to bluster. Craik reached for a scrap of paper and drew a horizontal line, from which dropped two vertical lines, one short and one long.
“What’s that, Gus?”
“It seems to me,” said he, “that memory for something depends on how deeply the something is processed or analyzed.”
“That’s it, Gus??”
“Well, that’s the essence of it . . .”
To be fair, my London students were interested in this somewhat different perspective, but I moved to Toronto shortly after this time, so the initial empirical explorations of the levels-of-processing (LOP) ideas were carried out with other colleagues. I also wrote about these nascent ideas to Endel Tulving, who was then at Yale University. He was “cautiously encouraging,” I would say, acknowledging that the LOP framework fitted some recent data quite well (e.g., Hyde and Jenkins, 1969), and also pointing out some shortcomings. These latter aspects included the
point that the notion emphasizes storage with no clear link to retrieval processes, and also that the LOP idea appears to focus on the encoding of single items and does not address interdependencies among items, such as organization and mutual inhibition. He was quite correct on both points, I believe. However, as we talked and corresponded further, Tulving became progressively more enthusiastic about the ideas and contributed two crucial components to their development. The first, as related further in the next section, was his invitation to write up the ideas as a theoretical article for the Journal of Verbal Learning and Verbal Behavior (JVLVB), and the second was his agreement to collaborate on the empirical studies that were suggested by the LOP framework.
The Craik and Lockhart article
When I arrived in Toronto to take up a faculty position in the fall of 1971 I was delighted to find that my friend and colleague Bob Lockhart had been thinking of memory in much the same way as I had. We shared the view that remembering should be thought of as a set of encoding and retrieval processes, rather than as a series of stores. Clearly there were qualitatively different ways in which information could be represented in the cognitive system; for example, words could be represented as visual images, in terms of their phonological, articulatory, or lexical features, and also in terms of the meaning they conveyed, either as single items or as components of a sentence. But these different encoding and representational dimensions are not stages or stores; they are simply different aspects of representation. Lockhart and I also shared the view that “deeper” processing, in the sense of greater involvement of meaning and implication, was likely associated with longerlasting memory records.
In early 1972 Endel Tulving was finishing his stint as Editor of JVLVB in those days the top journal in the areas of memory and verbal learning. His original plan at that time was to have the last issue under his editorship as an all-invited issue; as it happened, this did not work out, but Bob Lockhart and I were the beneficiaries of Tulving’s scheme as he invited us to write an article on the “levels” notions for the December issue of the journal. Tulving was working at Yale University at that time, and he wrote to me from there in January, 1972, with the following proposition:
I am writing to you to ask you whether you would not wish to write a paper, not too short and not too long, in which you would extol the virtues of whatever grand system of thought, theory, or metaphysical speculations relevant to human memory, or to some other manifestation of Mind, you hold at the moment of writing. I had especially in mind your notions about levels of analysis in shortterm memory, although you yourself may wish to do something else. I would publish the paper essentially as you would submit it, so you would not have to
worry about fighting and haggling with the establishment who might normally resist your innovative ideas. The nature of the article would be primarily theoretical, and integrative, rather than a vehicle for presenting new experiment results. Just about the only catch here is that I would have to have a reasonably tidy and completely finished manuscript no later than June 1.
Lockhart and I accepted with alacrity of course, and got down to work pretty well immediately. Endel Tulving took a paternal interest in our progress and indeed, was immensely helpful in shaping the final published version. Bob Lockhart and I work well together, so writing was not too much of a chore. Our styles are somewhat different, however, with Lockhart taking a rather philosophical approach relative to Craik’s more concrete and empirical approach. It is possible that these differences in emphasis yielded a rounder, and in some sense a “deeper,” final version. One incident occurred during a summer evening writing session as we labored to meet the June 1 deadline (we failed as I remember!). Our colleague Ben Murdock came into our room and tossed a book on the table with some phrase like “You fellows may be interested in this!” before walking out grinning broadly. The book was Laird Cermak’s recently published Human Memory, Research and Theory (1972) in which he lays out a very similar set of ideas to those that Lockhart and I were working on. So, levels were in the air in 1972 it seems. I did not know Cermak at the time, but I later met him and we became good friends. We jointly organized a conference on LOP, held in 1977, and this yielded a book of the proceedings (Cermak and Craik, 1979). Meanwhile, back in Toronto in 1972 Lockhart and I finally finished a draft that we and Tulving were happy with, so off it went to the publishers, with Bob Lockhart and I going through manic-depressive cycles— alternating between thinking we had come up with something quite useful, and thinking that the main message was trite and obvious—"meaningful things are remembered better.” Put that way it hardly seemed like a major breakthrough!
In fact, the article by Craik and Lockhart (1972) made a number of suggestions on how to conceptualize human memory in cognitive terms, and, naturally enough, some have stood the test of time better than others. At the time of writing our main aim was to question the usefulness of the dominant “stores” model (the idea that memory could be understood in terms of various memory stores: sensory memories plus short-term and long-term memory) and instead suggest that the persistence in memory of a perceived event depended on its depth of analysis, where “deep” analysis referred to processing of meaning and implication. Following some perceptual theorists, principally Treisman (1964a, 1969), we endorsed the notion of a hierarchy of processing stages running from shallow sensory analyses through intermediate stages such as phonology to deeper semantic analysis. We did also suggest (p. 675) that after identification the processed event may undergo further processing by enrichment or elaboration by integration with past experience. One major point was our emphasis on the notion that the purpose
of virtually all day-to-day processing is perceptual; at shallow levels to successfully navigate our environment, and at deeper levels to identify events and process their implications. Further, that these various sensory, perceptual, and conceptual analyses automatically yield a record of the analyses, and we may think of this record as the memory trace. Thus, importantly, deep perceptual and conceptual analyses are by themselves sufficient to form a long-lasting memory trace; intention to learn or memorize is not important; trace persistence is entirely a matter of the qualitative nature of the processing carried out—for whatever reason.
This last point was borne out strongly by studies involving incidental learning, in which participants are asked to process a series of words in a given way (e.g., judge the number of syllables each word contains or rate each word for pleasantness) and are then unexpectedly asked to recall or recognize the words. At the time of writing the Craik and Lockhart article, the literature (e.g., Hyde and Jenkins, 1969; Postman, 1964) confirmed the notions that later retention depended on the initial “orienting task,” tasks involving greater degrees of semantic processing were associated with better memory performance, and intentional learning differed from incidental learning only to the extent that intentional learning induced participants to carry out different operations from those induced by the incidental task. One case in point is that intentional learning is typically associated with better free recall in young adults than the level following a semantic orienting task (Challis et al., 1996; Craik, 1977a), probably owing to the fact that intentional learning induces participants to carry out inter-item organizational processing in addition to processing each item; such organization is beneficial to free recall, in particular.
Perhaps the main contribution of the “levels” paper, however, was its emphasis on memory as a set of dynamic processes rather than as a set of structures. In line with the main theme of this book, encoding was conceptualized as the same processes underlying perception and comprehension. In agreement with the suggestions of a number of previous theorists described in Chapter 2 (notably Bartlett, 1932), we should think of remembering as an activity of mind and brain. Assuming that retrieval is also a matter of active processing, the question immediately arises as to the relation between the processes associated with encoding and retrieval. In later articles (e.g., Craik, 1983) I endorsed the notion that retrieval is essentially a matter of recapitulating the same processes (as nearly as possible) as those carried out at the time of encoding, and this point of view is explored further and substantiated in Chapter 5. However, in the Craik and Lockhart article we were not so thoroughgoing, and suggested simply that encoding processes gives rise to an encoded record of these processes—the memory trace. Clearly something must change in the brain in order for remembering to occur at a later time, but the notion of a set of “memory traces” each representing a specific encoded event does not seem to be the way to go. However, merely repeating the same encoding operations at the time of retrieval, as suggested by Paul Kolers (1973), brings its own problems—for example, how do we know we are remembering an event as opposed
to simply perceiving and understanding it for the first time? These thorny questions are taken up again in Chapter 4.
The Craik and Lockhart article is sometimes cited as the article that questioned the validity of the distinction between short-term memory (STM) and long-term memory (LTM). Indeed, we questioned the validity of the “stores” concept but strongly maintained the distinction between “primary memory” and LTM, although in different terms. In our view primary memory was equivalent to recirculating information at a constant level of processing—typically a relatively shallow level. We commented that “in our view, such descriptions as ‘continued attention to certain aspects of the stimulus,’ ‘keeping the items in consciousness,’ ‘holding the items in the rehearsal buffer,’ and ‘retention of the items in primary memory’ all refer to the same concept of maintaining information at one level of processing” (Craik and Lockhart, 1972, p. 476). This notion of rehearsal being equivalent to the recirculation of information at a given level was also linked to the overall view that retention is a function of the depth of processing achieved, by the suggestion that maintenance of information at a constant level would not enhance subsequent memory performance. In order for rehearsal to improve memory it is necessary for the rehearsal processes to engage deeper levels. This contrast was given the rather unimaginative label of “Type I and Type II processing.” Better terms—maintenance and elaborative processing—were later suggested (Craik and Watkins, 1973; Woodward et al., 1973), and these terms are now in general use. So Craik and Lockhart maintained the distinction between STM and LTM, but with STM thought of as primary memory (PM)—those items held by attentional processing in conscious awareness. PM, in our terms, had a limited capacity, but the limit was in terms of the limited amount of attentional resources available to process aspect of words (typically phonology or articulation) rather than some structural limit. We also suggested that the “coding characteristic” of PM was not fixed but was flexible—depending on the aspect of held information being attended to—phonology, articulation, imagery, semantic, etc. That is, “the processor itself is neutral with regard to coding characteristics; the observed PM code will depend on the processing modality within which the processor is operating” (Craik and Lockhart, 1972, p. 676). Further thoughts on PM, and on how primary memory and working memory are related, are deferred until Chapter 6.
A troubling absence from the Craik and Lockhart article is the lack of an objective index of depth or processing. In the absence of such an independent index it is all too easy to claim that well-remembered items must therefore have been deeply processed. That is, the scientific logic is “circular” as our critics (e.g., Baddeley, 1978; Eysenck, 1978; Nelson, 1977) were quick to point out! We have countered this criticism in various ways. First, although the dimension of depth has been described in intuitive terms it seems that researchers’ intuitions do converge (Seamon and Virostek, 1978; Anderson and Reder, 1979). For example, Anderson and Reder conducted an informal study (n = 2) in which the authors independently rated
