Memory and Brain Dynamics: Oscillations Integrating Attention, Perception, Learning, and Memory

Memory and Brain Dynamics Oscillations Integrating Attention, Perception, Learning, and Memory © 2004 by CRC Press, LLC...

Author: Erol Basar

102 downloads 1318 Views 11MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

Memory and Brain Dynamics Oscillations Integrating Attention, Perception, Learning, and Memory

© 2004 by CRC Press, LLC

Conceptual Advances in Brain Research A series of books focusing on brain dynamics and information processing systems of the brain. Edited by Robert Miller, Otago Centre for Theoretical Studies in Psychiatry and Neuroscience, New Zealand (Editor-in-Chief); Günther Palm, University of Ulm, Germany; and Gordon Shaw, University of California at Irvine, USA.

Brain Dynamics and the Striatal Complex edited by R. Miller and J.R. Wickens Complex Brain Functions: Conceptual Advances in Russian Neuroscience edited by R. Miller, A.M. Ivanitsky and P.M. Balaban Time and the Brain edited by R. Miller Sex Differences in Lateralization in the Animal Brain by V.L. Bianki and E.B. Filippova Cortical Areas: Unity and Diversity edited by A. Schuz and R. Miller The Female Brain by Cynthia L. Darlington


Memory and Brain Dynamics Oscillations Integrating Attention, Perception, Learning, and Memory

Erol Basar ç Brain Dynamics Research Center Dokuz Eylül University, Izmir Brain Dynamics Multidisciplinary Research Network of Turkish Scientific and Technical Research Council TÜBITAK, Ankara The International and Multidisciplinary Research Network: Brain Dynamics and Cognition Affiliated with the IDP/United Nations, New York

CRC PR E S S Boca Raton London New York Washington, D.C. © 2004 by CRC Press, LLC

TF1663_C00.fm Page 4 Friday, May 21, 2004 1:15 PM

Dr. Erol Basar ¸ Director, Brain Dynamics Research Center and Chairman, Department of Biophysics, Faculty of Medicine Dokuz Eylül University Izmir, Turkey Chairman, The International and Multidisciplinary Research Network: Brain Dynamics and Cognition Affiliated with the I.O.P./United Nations, New York Brain Dynamics Multidisciplinary Research Network of Turkish Scientific and Technical Research Council TÜBITAK Ankara, Turkey E-mail: [email protected] Web page: http://braindynamics.deu.edu.tr/basar.htm

Library of Congress Cataloging-in-Publication Data Basar, ¸ Erol. Memory and brain dynamics : oscillations integrating attention, perception, learning, and memory / Erol Basar ¸ . p. ; cm. — (Conceptual advances in brain research ; v. 7) Includes bibliographical references and index. ISBN 0-415-30836-4 1. Memory. 2. Electroencephalography. 3. Brain. 4. Oscillations. I. Title. II. Series. [DNLM: 1. Brain—physiology. 2. Memory—physiology. WL 300 B297m 2004] QP406.B366 2004 612.8'23312—dc22 2003069760

This book contains information obtained from authentic and highly regarded sources. Reprinted material is quoted with permission, and sources are indicated. A wide variety of references are listed. Reasonable efforts have been made to publish reliable data and information, but the author and the publisher cannot assume responsibility for the validity of all materials or for the consequences of their use. Neither this book nor any part may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, microfilming, and recording, or by any information storage or retrieval system, without prior permission in writing from the publisher. The consent of CRC Press LLC does not extend to copying for general distribution, for promotion, for creating new works, or for resale. Specific permission must be obtained in writing from CRC Press LLC for such copying. Direct all inquiries to CRC Press LLC, 2000 N.W. Corporate Blvd., Boca Raton, Florida 33431. Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation, without intent to infringe.

Visit the CRC Press Web site at www.crcpress.com © 2004 by CRC Press LLC No claim to original U.S. Government works International Standard Book Number 0-415-30836-4 Library of Congress Card Number 2003069760 Printed in the United States of America 1 2 3 4 5 6 7 8 9 0 Printed on acid-free paper



Series Preface For over 30 years Erol Basar ¸ has been pursuing a distinctive line of research into brain function. In his approach, signals from spontaneous EEGs and those produced in response to stimuli (evoked or event-related potentials) are analyzed into their frequency components and these components are then taken as the elements for correlation with psychological variables. His work is distinctive not only in scientific terms, but is also guided by a distinctive philosophy. This book, the seventh in the Conceptual Advances in Brain Research (CABR) series, is his most complete exposition to date of this philosophical and scientific perspective on brain function. Basar ¸ ’s approach has its origins in part in physics rather than biology. This is hinted at in the early chapters of his book, and is made explicit in Chapter 11 where he refers to the strategy adopted by Isaac Newton in understanding planetary motion. This approach is very apt. A wellknown aphorism of Newton came to mind as I read this final chapter: “Hypothesis non fingo” (usually translated as “I do not feign explanations”). Newton formulated the concept of gravity and showed how this concept allowed one to understand planetary motion and various other things, but he did not try to explain gravity in terms of something more fundamental. Likewise, Basar ¸ shows how frequency-specific patterns of electrical activity in the brain help one understand psychological processes, but does not try to explain those frequency-specific patterns in terms of lower level phenomena (i.e., neurones). This approach is far from the focus of researchers schooled in single-unit electrophysiology. However, many of the oscillatory phenomena described by Basar ¸ can already be explained in terms of neuronal biophysics and related interactions between networks of neurones. In principle, there is no reason to doubt that they can all be so explained. Basar ¸ occasionally refers to single-unit studies, and there is no doubt that he does in fact accept that these rhythmic patterns of activity are derived from patterns of synaptic activation in single neurones. However, that is not the conceptual language he prefers. Instead he sees the different frequency-specific components of massed neuronal activity as the real units for understanding brain functions. He presents a great deal of evidence (especially his own) showing that frequency-specific electrographic activity, selectively distributed in various brain sites, correlates with the psychological aspects of the tasks his subjects are performing. If we accept the conceptual language used by Basar ¸ , he ably demonstrates that what can be described using this language is very substantial. Here are some of the many examples. Regularly occurring, accurately timed sequences of stimuli lead to phase locking of EEG rhythms that develop as the stimulus pattern becomes familiar. Such regularization of frequency-filtered EEG components is related to the difficulty of the task and factors such as task fatigue. Well-known, event-related potential (ERP) components such as the P300 can be analyzed as various frequency components that have different psychological correlates. The frequency components induced by stimuli depend on the frequency composition of the prestimulus EEG. Coherence of oscillations between different parts of the brain is increased by stimulation and the entropy (scatter of frequency components) of the EEG is decreased by stimuli. Different EEG frequencies appear to be of differential importance in different parts of the hemispheres. Familiar stimuli (like a photo of the subject’s own grandmother) induce EEG rhythms across the whole of the hemispheres, with different frequencies dominant in different regions. Unfamiliar faces produce patterns of resonance different from familiar ones like a subject’s grandmother. All these findings depend on the use of frequency filtering of the EEG or ERP signals.



One of the great strengths of this book is what can be called “psychobiological holism.” On the psychological side, Basar ¸ refers continually to an alliance of attention, perception, learning, and remembering, emphasizing that they are not sharply separate functions. This is a far more realistic view of cognitive function than the traditional one consisting of a series of independent “black boxes,” for which one is always tempted to search, but in vain, for their strict anatomical localizations. On the biological side, the electrographic correlates of psychological function are oscillations capable of interacting over the whole of the brain. As a result, the sort of phenomena that become important in Basar ¸ ’s view are rhythms distributed somewhat selectively over the whole brain, correlations at one region between oscillations at different frequencies, and coherences between oscillations in widely separate locations at the same frequency. The psychological and biological sides of this holism fit together naturally and convincingly. In the main, the empirical evidence is dealt with in separate chapters or sections from the theory development. These two parts of the book are welded together using a carefully developed didactic style. The evidence will be a rich source for future researchers, both empirical and theoretical. The theory development comes at various stages of the book, the most substantial of which is Chapter 9. Overall this is a bold and forward-looking essay that explains the brain as a whole rather than artificially separating it into components. At times the author admits realistically that his formulations are somewhat tentative and in need of future revision. Books like this are exactly what the CABR series was set up to promote. Robert Miller



Preface … a major task for neuroscience is to devise ways to study and to analyze the activity of distributed systems in waking brains, including particularly human brains, and to seek direct correlations and explanations of the relevant behavior in terms of those patterns of neural activity. V.B. Mountcastle, 1998

DYNAMIC AND SELECTIVELY DISTRIBUTED MEMORY IN THE WHOLE BRAIN This book aims to bridge the disciplines of neurophysiology, cognitive psychology and EEG–brain dynamics with the aim of describing how the brain represents mental events that are interwoven with memory. Memory is inseparable from all other brain functions and involves distributed dynamic neural processes. The analysis of concerted action of multiple oscillatory processes is a major key to understanding distributed memory. The role of the memory in human behavior cannot be overemphasized because no high level nervous function can operate successfully without memory contributions. Perception, cognition, problem solving, and decision making all rely on memory. Thus, a major task for neuroscience is to choose strategies to analyze the activated memory in the awake brain. Based on the explosion of neuroscience literature, the concerted actions of distributed multiple oscillatory processes (EEG oscillations) play a major role in brain functioning. New important strategies are introduced in this book, one of which treats the alliance of attention, perception, learning, and remembering (APLR alliance) by means of EEG oscillations. According to Baddeley (1996), working memory provides a crucial interface of perception, attention, memory, and action. During experiments involving learning and working memory processing, EEG oscillations manifest continuously evolving dynamics. Empirical results led to development of a model of the hierarchy of memories as a continuum and a theory covering the concerted actions of function and memory in the whole brain. Unique leitmotifs and strategies for this book include use of the expression dynamic memory to describe memory processes that evoke relevant changes in alpha, beta, gamma, theta, and delta activities. The concerted actions of distributed multiple oscillatory processes constitute major keys for revealing distributed memory. The notion of physiological or fundamental memory is introduced. This type of memory includes phyletic memory and reflexes. The evolving memory incorporating reciprocal actions or reverberations in the APLR alliance and during working memory processes is emphasized. A new model related to the hierarchy of memories as a continuum is introduced. The notions of longer-activated memory and persistent memory are proposed as replacements for long-term memory. A new strategy related to recognition of faces emphasizes the importance of EEG oscillations in neurophysiology and gestalt analysis. According to Damasio, memory depends on several brain structures working in concert across many levels of neural organization. Memory is a constant work in progress. The proposition of a brain theory based on supersynergy in neural populations is most pertinent for understanding this constant work in progress. The proposed basic framework called memory in whole-brain work



emphasizes that memory and all brain functions are inseparable; they act as a whole in the whole brain. The model and EEG strategies introduced in this book may be relevant for analyzing pathological changes in Alzheimer’s disease, patients and psychiatric patients with attention, learning, and memory deficits. The results presented in this book are based on the application of frequency analysis to EEG records from human and animal brains. Emphasis is placed on event-related oscillations and/or function-related oscillations. Because neuroscientists have concluded that different brain regions must cooperate to accomplish all brain functions, the analysis of the relationships of different regions of the brain is becoming more important. Lashley (1929) proposed that memories are in fact scattered across the entire brain rather than concentrated in specific regions. The results described in this book demonstrate that the whole brain is involved in these processes and that the memory function is selectively distributed in the brain. Lashley did not indicate the selective distribution because adequate experimental techniques to reveal it were not available in the 1920s. Hebb's fundamental concept of cooperativity concepts opened the area of interactive and growing networks in cognition research (1949). This book raises more questions than it claims to answer. It opens many new windows although some remain closed. As I finished the writing of my 1980 book on EEG–brain dynamics, I raised many questions. Answers came from a large number of neuroscientists, and research in many areas continues to expand results. I hope that the new pathways described in this book will gain importance and that many unsolved problems will be solved by young scientists working on the dynamics of memory function.

COPERNICAN CHANGES IN MEMORY RESEARCH According to Fuster (1997), our thinking about the cortical organization of primate memory is undergoing a Copernican change, from a neurophysiology that localizes different memories in different areas to one that views memory as a distributed property of cortical systems. According to Steven Rose (1997), memory is not, as previously thought, a vast cerebral warehouse filled with rows and rows of neatly ordered filing cabinets. It is impossible to know where in the brain a particular memory is located. Memory is a dynamic property of the brain as a whole rather than of any specific region. Memory resides simultaneously everywhere and nowhere in the brain. Our long-standing experiments with the Brain Dynamics Research Program that started in the 1970s led us to conclude that memory function is selectively distributed in the whole brain because oscillatory processes evoked by memory load occur in a selectively distributed manner in the whole brain as concerted (coherent) actions.

WHAT JUSTIFIES WRITING ANOTHER BOOK ON MEMORY? Several treatises have covered the neural presentation of memory. Fuster (1995) asked, “Who needs yet another?” The time has come to build a framework surrounding EEG-related memory processes. As early as 1985, I used the dynamic memory to describe memory processes that evoked relevant changes in alpha, theta, and delta activity (Chapter 3). My 1980 monograph titled EEG–Brain Dynamics: Relation between EEG and Brain Evoked Potentials was not in the main line of brain research when it was reviewed in Trends in Neuroscience in 1981. Since then, analysis of functionrelated brain oscillations is one of the most important areas of neuroscience research. My motivation to write this book was triggered by increasing numbers of publications in this area and also my experience in several other areas. I belong to a small group of scientists working on oscillation phenomena in the brains of a broad variety of species ranging from Aplysia ganglia to humans. © 2004 by CRC Press, LLC


The brain dynamics concept is most relevant for determining how memory is distributed because measurements on a timescale from 1 to 1000 MS (impossible to perform with functional magnetic resonance imaging [fMRI]) can now be achieved. Single sweep analysis of oscillations enormously contributes to analyzing the interactions of attentive states, learning, and evolution of memory, i.e., how EEGs are altered during the development of working memory (Basar ¸ and Stampfer, 1985; Chapter 3, this volume). Experiments on oscillatory dynamics provide the only possibility of elucidating the process of memory evolving over a short time interval. Our group started such studies very early. Our work also included research on chaos, entropy, and comparative studies of EEGs and MEGs. Our experiments with implanted cats allowed us to analyze distributed processes of the brain stem and cerebellum. As a result, memory traces in the whole brain can be analyzed by using our experimental data. In order to establish experimental strategies to reveal cognitive processes and integrative brain function, the neurons–brain theory and the notion of superbinding instead the concept of the cardinal pontine cell may play a major role. The goal of this short book is not to be the most comprehensive discussion of experiments and the dynamics of electrophysiology. The descriptions of biochemical and electrophysiological micromechanisms that serve to store information in the brain are not within the scope of this book. The models presented cannot be perfect. They are intended as examples to help build a new frame for the so-called dynamic memory. Accordingly, I hope that this book fulfills its purpose of proposing a new framework in the new domain of EEG-related memory research.

ABOUT THE CONTENTS AND ORGANIZATION OF THIS BOOK Beginning in the 1970s, a series of experiments examined the oscillatory character of event-related potentials in animal and human brains. At that time, the understanding of memory correlates of the measured oscillations was not the main goal. Instead, we aimed to attack basic physiological components of event-related oscillations. The situation changed dramatically in the mid-1980s when event-related oscillations provided an important window for revealing descriptions of cognitive functions and memory. Although it may provide guidance for design and interpretation of experiments, this book is not a text. Apart from the general chapters in the first part, it cannot be read in the usual manner. It must be tackled piece by piece, step-by-step, and will hopefully advance your knowledge of the fertile terrain of brain dynamics. When the preliminary experiments related to working memory and memory-related oscillations were published in the 1980s, we could not have predicted that gamma, alpha, theta, and delta oscillations would provide core material for scientists working on memory-related research. Accordingly, we use in this book a strategy that should orient readers to assimilate the experimental and theoretical material in parallel by going back and forth between experimental and theoretical chapters. The book is divided in three parts. Part I covers foundations; it presents the introductory core material and the rationale for the book. The material is essential for understanding how the oscillatory approach reveals information about brain functions and memory. Part II details core experiments and their interpretations. Part III includes theoretical and modeling-oriented chapters. Although Chapter 7 has a more theoretical character, it contains experimental results that serve as a theoretical framework. It is included in Part II to explain why the grandmother experiments were initiated. Part I — Chapter 1 describes preliminary concepts and some frameworks initiated since the 1920s. Chapters 3, 6, and 8 provide empirical data obtained by application of these concepts. The new data, in turn, led to new theories or principles. The paradigm change in cognitive sciences put more emphasis on analyses of macrodynamics instead of microdynamics. Accordingly, the © 2004 by CRC Press, LLC


conjecture is now open to establish a new conceptual framework or theory of neural populations to extend or replace Sherrington’s neuron doctrine and also to include memory in its framework. Chapter 2 explains definitions and concepts related to different types of memories. For readers starting to learn the essentials of memory function, a reading of this chapter is obligatory before attacking more difficult descriptions in coming chapters. This chapter is kept short. Readers who want to acquire more knowledge from the literature are referred to other readings cited at the end of this section. Part II — Chapter 3 is a key chapter that describes experiments on dynamics of memory by means of EEG- and event-related oscillations during cognitive processes based on performance of working memory tasks. Analysis of single epochs prior to and following target signals led to the concept of dynamic memory at the EEG level. The preliminary experiments performed more than 20 years ago gave us the first hints of the dynamics of evolving memory and reciprocal activation of attention, perception, and memory during working memory tasks. The results of the experiments support the theory of reentrant or recurrent networks. After reading Chapter 9, readers will possibly return to Chapter 3 to review the experimental grounds of the theoretical treatise on transitions, evolving memory, matching processes, and the new model presented in Chapter 9. Chapter 4 has a similar character to Chapter 3. It deals with similar experiments with attention and working memory paradigms performed on freely moving and behaving cats. The results with behaving cats made it possible to find correlates to memory function, perception, and attention in the whole brain including the brain stem. These key results in the whole brain allow us to state that memory function is manifested with selectively distributed oscillations in the whole brain and that results of investigators working with limited locations of electrodes in the human brain should be interpreted with great caution. Memory functions cannot be localized as Lashley (1929) pointed out. Prestimulus EEG activity and its roles in brain responsiveness and short-term memory were already explained in Chapter 3. In Chapter 5, the relation between prestimulus EEG and brain responses is analyzed with detailed analytical and systematic steps, thus allowing the interpretation of prestimulus EEGs as important factors in the causality of brain responses. This causality is strongly related to endogenous brain activity and, in turn, related to its cognitive states. It is also an important controlling factor for the reciprocal activation of functions of the APLR alliance, as will be discussed in Chapters 7, 8, 9, and 11. The causality principle of Newton’s dynamics and quantum dynamics play an essential constructive and interpreting role in memory-related brain dynamics. In Chapter 6, function- and memory-related oscillations are treated systematically by starting with a chronological survey. Readers who have less knowledge about brain oscillations may jump to Chapter 6 after reading Chapter 1. Chapter 6 contains representative examples of gamma, alpha, theta, and delta frequency channels. It shows that integrative brain functions are manifested by multiple oscillations; about 50 functional correlates of oscillatory responses are discussed. The principle of superposition and its functional role are explained and accompanied by samples, The selectively distributed alpha, gamma, theta, and delta systems are described by showing that frequency responses are modality- and topology-dependent. Another important feature of this chapter is the analysis of long-distance coherences in the brain. This opens the issue of action in concert of selectively distributed frequency systems and superbinding in Chapter 7. The chapter serves as an intermediary one, orienting readers not yet informed about conceptual developments of the last 5 years. It aims to bridge the various theoretical steps in somewhat chronological order. The essential idea is to train experimenters in neuroscience to develop new frameworks for treating the electrophysiology of cognitive functions. EEG research scientists oriented to functional analyses in neuroscience suffer from a lack of rules and principles similar to those used in conventional neurophysiology, and they seek theoretical frameworks and new rules for proper understanding of EEG recordings. Table 7.1 is self-explanatory. It explains activities of distributed oscillatory systems and their relation to integrative functions and memory. © 2004 by CRC Press, LLC


Chapter 8 presents advanced steps in the analysis of cognitive processes outlined in this book by treating the enigma of the grandmother neuron — a prominent problem in the neurophysiology literature of the last century. A crucial point is the fact that both anonymous and known faces evoke oscillations that are clearly different from those produced by anonymous faces. This differentiation is absolutely impossible to make by means of conventional ERP analysis techniques. Accordingly, readers oriented to electrophysiological methods may find useful tools in this chapter. These experiments were the first involving recognition of pictures of the subjects’ grandmothers. Preliminary but tenable consequences of the so-called grandmother experiments are explained in Chapter 8. The most important issue is that the whole brain and all oscillations are activated during recognition or remembering of the faces of the subjects’ grandmothers and faces that were unknown at the beginnings of the experiments. The responses behave as a three-dimensional construct consisting of temporal, spatial, and frequency spaces. The responses to the faces were not represented solely by one location or unique frequency. The selectively distributed nature of multiple oscillations of the whole cortex clearly denies the possibility of a new version of the neuron doctrine of Barlow (1995) as an extension to Sherrington’s initial doctrine. The grandmother experiments and their implications resemble the tip of an iceberg and may be expanded into new versions including pictures of known episodes or pictures that induce emotions, thus enabling electrophysiological differentiation and transition between semantic and episodic memories. Part III — Chapter 9 is the heart of the book and discusses the essential model derived from the study of memory by means of EEG oscillations. Readers who are curious about the results emerging from studies of EEG oscillations may start with this chapter, then review earlier material based on cross-references throughout the book. Chapter 9 begins with schematic descriptions of selectively distributed alpha, theta, and delta response systems. One important reference is the work by Fuster (1995 and 1997) in which the notion of distributed memory in the cortex and the hierarchy of memories were anchored by relevant physiological findings. A new model presenting various levels of memory function and hierarchies of various types of memories (memory states) is proposed; it constitutes the core of Chapter 9. The role of physiological processes, their contributions to memory function, and the transitions between memory states are emphasized. The chapter also covers a new hypothesis based on frequency tuning and resonance between brain neural populations (multiple frequency matching). Interwoven with the proposed new model are questions related to equipotentiality by Lashley (1929), the reverberation hypothesis by Hebb (1949), and the reentry hypothesis of Edelman (1977). This chapter can be considered a real workshop. Readers may page back and forth to other chapters in order to assimilate and/or criticize the ideas or notions of the new model based on findings with EEG oscillations. We are open to interactions with readers and welcome their emails. Our homepage will include a presentation of this model ([email protected]; http://braindynamics.deu.edu.tr). Chapters 10 and 11 contain important information about new trends and emerging ideas discussed throughout all chapters of the book, but both chapters have different aims. Chapter 10 provides a type of concluding synthesis of the new trends in analysis of memory function by means of EEG oscillations. Since it combines results and ideas presented throughout this book, it is useful for gaining a general comprehension of the subject of brain dynamics. After reading Chapter 10, readers may return to previous chapters, possibly after acquiring a general orientation after the reading of this chapter. Chapter 11 focuses on the future. Readers who are interested in theories related to brain function may find in this chapter a theoretical framework to orient them to designing new experiments, devising new theoretical proposals, or possibly modifying this proposal by using some of its empirical foundations or basic principles. The epilogue points out the hope that the draft of the theory on whole-brain work will provide a new groundwork for understanding dynamic memory. The glossary contains some of the © 2004 by CRC Press, LLC


abbreviations, nomenclatures, definitions, and descriptions of terms used in this book. The appendix explains relevant mathematical procedures frequently cited in this book in order to provide continuity of the text for readers who are familiar with common mathematical tools.

ACKNOWLEDGMENTS The Scientific and Technical Research Council of Turkey TÜBITAK has provided the major contribution for the preparation of this book and the achievement of joint experiments with German and Turkish scientists in the last 10 years or so. In 2000 I moved from the Medical University Lübeck in Germany, where I was leading the Neurophysiology Research Group, to Dokuz Eylül University in Izmir. Before and during this transition period, The German Research Council DFG and The German Ministry of Education BMBF offered major support for the realization of joint experiments in Bremen, Lübeck, Ankara, Izmir, and joint publications with the groups in Moscow, Sofia, Buenos Aires, and earlier in Perth. The McDonnell Foundation in the United States further supported the cooperation between the Lübeck and Sofia groups. The interaction between Lübeck and Helsinki was supported also by DFG and the European research organization BIRCH. The fruitful cooperation of scientists from three continents could be realized due to the generous support of these foundations. An essential contribution for establishing the International Multidisciplinary Network “Brain Dynamics and Cognition,” which operates under the official legacy of the International Organization of Psychophysiology (I.O.P.) associated with the United Nations in New York, was achieved by I.O.P. President Professor Dr. C.A. Mangina. I express my appreciation to Prof. Dr. Mangina for his efforts to motivate scientists for joint cooperation and thus to enrich worldwide understanding and peace. Dr. Murat Özgören, M.D., Ph.D., coauthor of Chapter 8, and Dr. Adile Öniz made excellent contributions to the preparation of the difficult manuscript. Moreover, they prepared a number of illustrations and contributed to the organization and preparation of references and the glossary. Dipl. Psychol. Christina Schmiedt and Cand. Psychol. Ingo Fründ in Bremen read and corrected ˇ the manuscript. Mrs. Ahrens, secretary of the Bremen Institute, and Mrs. C. Yegin, my secretary in Izmir, greatly helped in the organization of international joint research programs. A special note of thanks is due my spouse and most important colleague, Professor Dr. Canan ˘ at the University Bremen and mother of our children Eren and Pelin. Since the Basar ¸ -Eroglu 1980s, she has performed in Germany the most important experiments that constitute the core of this book. In the last years we have been able to perform the intriguing experiments of Chapter 8 in her laboratory in Bremen. Accordingly, her work has been invaluable throughout my entire career and also in the development of the present book. I also express my deepest appreciation to Professor Dr. Sirel Karakas¸, my former graduate student in Ankara. She has had a major role in all my monographs for more than 30 years as well as in this book as coauthor of the last chapter. Her incessant questions, constructive suggestions and ability to predict the new emerging hypotheses were extremely helpful. Therefore, she has been my most important companion in the new avenue of memory and brain dynamics. I greatly appreciate the contributions of all these persons and foundations.

SUGGESTED READINGS In order to achieve maximum gain from reading this book, readers must be somewhat familiar with the principles of neurophysiology, psychophysiology, and the psychology of memory. Excellent references include: Baddeley, A., Wilson, B.A., and Watts, F.N. (1995), Handbook of Memory Disorders, John Wiley & Sons, New York. © 2004 by CRC Press, LLC


Basar, ¸ E. (1998), Brain Function and Oscillations, Vol. I, Brain Oscillations: Principles and Approaches, Springer, Berlin. Basar, ¸ E. (1999), Brain Function and Oscillations, Vol. II, Integrative Brain Function: Neurophysiology and Cognitive Processes, Springer, Berlin. Damasio, A.R. (1994), Descartes’ Error: Emotion, Reason, and the Human Brain, Grosset/Putnam, New York. Eichenbaum, H. (1999), The hippocampus and mechanisms of declarative memory, Behavioral Brain Research, 103: 123–133. Eichenbaum, H. (2000), A cortical–hippocampal system for declarative memory, Nature: Reviews in Neuroscience (U.S.), 1: 41–50. Fuster, J.M. (1995), Memory in the Cerebral Cortex, MIT Press, Cambridge, MA. Goldman-Rakic, P.S. (1988), Topography of cognition: parallel distributed networks in primate association cortex, Annual Review of Neuroscience, 11: 137–156. Goldman-Rakic, P.S. (1996), Regional and cellular fractionation of working memory, Proceedings of the National Academy of Sciences of the U.S.A., 93: 13473–13480. Goldman-Rakic, P.S. (1997), Space and time in the mental universe, Nature, 386: 559–560. Kandel, E.R., Schwartz, J.H., and Jessel, T.M. (1991), Principles of Neural Science, Elsevier, New York. Miller, E.K. (2000), The prefrontal cortex and cognitive control, Nature: Reviews in Neuroscience (U.S.), 1: 59–65. Miller, R. (1991), Cortico-Hippocampal Interplay and the Representation of Contexts in the Brain, Springer, Berlin.



Author Professor Erol Basar ¸ was born in Istanbul, Turkey and studied at the Universities of Munich, Hamburg, and Hanover in Germany. He was educated as a physicist and physiologist and earned a Ph.D. in biophysics. He joined the staff of the Physiology Institute in Hamburg in 1965 and was appointed a postdoctoral fellow at the Nathan Kline Brain Research Institute in New York in 1968. In 1971, he was appointed associate professor and founding director of the Institute of Biophysics at Hacettepe University in Ankara, Turkey, where he performed basic research on brain oscillations and integrative brain function. In 1978, Professor Basar ¸ was appointed the Richard Merton Professor of the German Research Council at the University Kiel. He served as head of the Neurophysiology Research Group at the Physiology Institute of The Medical University in Lübeck from 1980 through 2000. During that period, he worked on several international projects with scientists from San Diego, California; Perth, Australia; Moscow, Russia; Sofia, Bulgaria; Istanbul, Turkey; Helsinki, Finland; Buenos Aires, Argentina; Copenhagen, Denmark; Shanghai, China; and Vancouver, Canada that merited considerable attention. His most important collaboration was the study of invertebrate ganglia with Professor T.H. Bullock, as a result of which both scientists have organized conferences and edited books. Since 1993 Professor Basar ¸ has served as president of the Brain Dynamics Research Network of TÜBITAK (the Research Council of Turkey). He was named a professor at Dokuz Eylül University in Izmir, Turkey in 2000 and currently serves as director of the Brain Dynamics Multidisciplinary Research Center and the Department of the Biophysics at the University’s Medical School. Professor Basar ¸ is strongly involved with the founding of a premier international research center in Izmir with the support of DPT, the governmental planning agency in Ankara. Professor Basar ¸ is currently the Vice President for Academic Affairs of the International Organization of Psychophysiology (I.O.P.) associated with the United Nations (New York). He is also the chairman of The International Research Network on “Brain Dynamics and Cognition” affiliated with I.O.P./U.N. (New York). Professor Basar ¸ has published 12 books (five of which are monographs) and approximately 200 other publications and has organized six international conferences. Since the 1970s, he has been one of the pioneers who noted the importance of oscillatory brain dynamics for integrative brain function and memory. His monograph titled EEG–Brain Dynamics: Relation between EEG and Brain Evoked Potentials published by Elsevier in 1980 is known as a milestone in the field of brain dynamics. ˘ Professor Basar ¸ is married to Professor Canan Basar ¸ -Eroglu, a staff member at the Institute of Cognition Research in Bremen, Germany. They have written several publications together.



Contents PART I Foundations Chapter 1

Introduction and Core Philosophy

1.1

Landmarks: Memory Is Distributed, Memory is a Dynamic Property 1.1.1 Lashley’s Equipotentiality 1.1.2 Hebb’s Rules of Cooperativity 1.1.3 Hayek: Perceiving is Classification of Objects by Activation of Associative Nets 1.2 New Trends in Neuroscience 1.3 Copernican Changes in Memory Research 1.3.1 Distributed Networks 1.3.1.1 Distributed Memory: Findings with Functional Magnetic Resonance Imaging 1.3.2 Parallel Distributed Processing 1.4 EEG-Brain Dynamics 1.4.1 Importance of EEG Studies 1.5 Pioneering Studies of Brain Macrodynamics and Whole Brain Approach 1.5.1 Griffith: Statistical Mechanics in Biology and Physics 1.5.2 Rosen: Global Neurodynamics 1.5.3 Fessard: General Transfer Functions of the Brain 1.5.4 Edelman: Reentrant Signalling Theory of Higher Brain Function 1.6 Freeman, Katschalsky, and Haken: Preliminary Steps in Introducing Macrodynamics of Electrical Activity 1.7 Application of Principles of Biological System Analysis to Brain Research 1.7.1 Reasons for Establishing Programs for Brain Research 1.7.1.1 Program Steps 1.7.1.2 Mathematical Methods of Program 1.8 New Approaches to Brain Functioning at Macroscopic Level 1.8.1 Sherrington’s Neuron Doctrine Revisited 1.8.2 Renaissance of EEG Use in Search of Integrative Brain Functions 1.9 Neurons-Brain Theory: An Approach that Includes Whole Brain Organization 1.9.1 Topography of Cognition and Elements of Neurons–Brain Theory 1.10 Significance of EEG Brain Dynamics in Memory States and Integrative Brain Functions Chapter 2 2.1

2.2

Concepts and Theories

Memory Machineries 2.1.1 Dynamic Memory and APLR Alliance 2.1.2 Steps of Memory Processing 2.1.3 Encoding and Retrieval Fractionation of Memory



2.3 2.4

2.5

2.6

2.7

2.8

2.2.1 Long-Term Memory versus Short-Term Memory 2.2.2 Working Memory Distinction between Implicit and Explicit Memory States Nondeclarative Memory 2.4.1 Phyletic Memory 2.4.2 Perceptual Memory 2.4.3 Procedural Learning 2.4.4 Priming 2.4.5 Evolving Memory Declarative Memory 2.5.1 Episodic Memory 2.5.2 Semantic Memory 2.5.3 Relationship of Episodic and Semantic Memories Neurobiology of Memory 2.6.1 Molecular and Cellular Bases of Memory 2.6.1.1 Hebb’s Proposal 2.6.1.2 Kandel’s Fundamental Results 2.6.1.3 EEG Oscillations in Aplysia and Helix pomatia New Scheme Based on EEG Studies for Categorization of Memory Levels 2.7.1 Physiological (Fundamental-Functional) Memory 2.7.2 Transition and Combination of Memory Stages (Evolving Memory) Longer-Acting Memory and Transition to Persistent Memory in Whole Brain

PART II Experiments and Their Interpretation Chapter 3 3.1 3.2

3.3

Shaping Dynamic and Evolving Memories by Reciprocal Activation of Attention, Perception, Learning, and Remembering

Essential Experiments Involving Dynamic Memory and Top-Down Activity Dynamic Memory Manifested by Induced Alpha Activity 3.2.1 Selective Attention 3.2.2 APLR Alliance 3.2.3 Importance of Internal Event-Related Oscillations 3.2.4 Coherent and Ordered States of EEG due to Cognitive Tasks 3.2.4.1 Preliminary Experiments 3.2.4.2 Preliminary Results 3.4.3 Global Trends of Pretarget Event-Related Rhythms: Subject Variability 3.2.5 Paradigms with Increasing Occurrence Probability 3.2.5.1 3.5- to 8-Hz Range 3.2.5.2 8- to 13-Hz Range 3.2.5.3 40-Hz Range 3.2.6 Experiments with Light Stimulation 3.2.6.1 Experiments with Varied Probabilities of Stimulus Occurrence 3.2.7 Experiments with Subject A.F. 3.2.8 Quasideterministic EEGs, Cognitive States, and Dynamic Memories 3.2.8.1 Dynamics of Time-Locked EEG Patterns Relations between Memory States and P300 Responses: EROs 3.3.1 Experimental Set-Up and Paradigms



3.3.2

3.4

Frequency Analysis of ERPs: Preliminary Results 3.3.2.1 Comparison of EPs and ERPs 3.3.2.2 Comparative Analysis of Poststimulus Frequency Changes under Different Conditions and Their Contributions to Different Latency Peaks 3.3.2.3 Formation of Peaks 3.3.2.4 Comparison of ERP Responses to Regular and Random Infrequent Target Stimuli 3.3.3 Orientation Reaction and Learning during Repetitive Stimulation Requirement of Preparation Rhythms for Activation of Working Memory: Analysis of Pre- and Poststimulus Activity in Single Sweeps 3.4.1 Event-Related Theta Oscillations 3.4.2 Event-Related 10-Hz Oscillations 3.4.2.1 Interim Summary 3.4.3 Modulation of P300 Activity by Preparation Rhythms 3.4.4 Control of Learnable Sequences by Prestimulus EEG Activity or Building of Memory Templates 3.4.5 Varied Degrees of Augmentation and Prolongation: Gamma Oscillations in Memory Tasks 3.4.6 Action of APLR Alliance and Hyphothesis Concerning Reentrant Circuits 3.4.7 Habituation 3.4.8 Augmentation of Knowledge or Learned Material Is Reflected by Regular and Increased Alpha Activities

Chapter 4

Perception and Memory-Related Oscillations in Whole Brain

Canan Basar ¸ -Eroglu ¸ ˘ and Erol Basar 4.1 4.2

4.3 4.4

Relevance of Chapter Theta and Alpha Responses in Cat Brains during Cognitive and Memory-Related Tasks 4.2.1 Introduction 4.2.2 Methods and Paradigms Utilized for Obtaining P300 Responses from Freely Moving Cats 4.2.3 Systematic Analysis of Effects of Repetition Rate of Omitted Tones on ERPs Recorded from Cat Hippocampi 4.2.4 Utility of Analysis in Frequency Domain 4.2.5 Multiple Electrodes in Hippocampus 4.2.6 Hippocampal P300 and Cognitive Correlates: Theta Components in CA3 Layer Compound P300–40-Hz Response of Cat Hippocampus 4.3.1 P300–40-Hz Compound Potential Event-Related Oscillations in Cat Hippocampus, Cortex, and Reticular Formation during States of High Expectancy: Comparison with Human Data 4.4.1 Unit Activity and Behavior 4.4.2 Event-Related Potentials in Cortex and Hippocampus in a P300-Like Paradigm 4.4.3 Selectively Distributed Theta System: Involvement of Limbic, Frontal, and Parietal Areas 4.4.3.1 Integrative Analysis of Increased Theta Response 4.4.4 Interpretation of Changes in ERPs 4.4.4.1 Comparison with Human Responses



4.4.5 4.4.6

4.4.7 4.4.8 Chapter 5 5.1 5.2 5.3

5.4

5.5 5.6 5.7 5.8

6.2

6.3

Causal Factors Controlling Brain Responsiveness and Memory: Prestimulus EEG Activity, Entropy, and Genetics

Introduction Relationship of EEG and ERP Algorithm for Selective Averaging 5.3.1 Dependence of EP Amplitudes and Waveforms on Prestimulus EEG: Vertex Recordings 5.3.1.1 Auditory-Evoked Potentials 5.3.1.2 Visual-Evoked Potentials 5.3.1.3 Topographic Aspects 5.3.2 Frontal Visual-Evoked Potentials 5.3.3 Inverse Relations of EEGs and Visual Responses Frequency Content of EROs from Different Locations: Major Operating Rhythms 5.4.1 Major Operating Rhythm (MOR) of Frontal Lobe: Theta? 5.4.2 MORs of Occipital and Central Region (Vertex) 5.4.3 Functional Significance of EEG–EP Interrelations Barry: Preferred States in Brain Activity 5.5.1 Creation of Preferred Brain States by APLR Alliance Causality of Brain Responses According to Changes in Oscillatory Networks Entropy as Causal Factor in Responses and Mechanisms of Super-Synergy Genetics as a Causal Factor in Delta and Theta Responses and Beta Rhythms

Chapter 6 6.1

Why Compare EP Results with Conventional Experiments? Structures Involved in States of APLR Alliance 4.4.6.1 Hippocampus as Supramodal Structure 4.4.6.2 Frontal Cortex 4.4.6.3 Global Function of Reticular Formation 4.4.6.4 Cognitive Functions of Cerebellum Secondary Alpha Response and Alpha Response with Delay Comparison with Human Brain Results

Correlation of Multiple Oscillations with Integrative Functions and Memory

Introduction 6.1.1 Aim of Chapter 6.1.1.1 Emphasis on Multiple Oscillations in Brain Research 6.1.1.2 Role of Oscillations in Memory Processing 6.1.1.3 Steps for New Synthesis and Binding Problem Survey of EEG Oscillations 6.2.1 Alpha Activity 6.2.1.1 Survey by Andersen and Andersson (1968) and Basar ¸ (1999) 6.2.1.2 Toward a Renaissance of Alphas 6.2.2 Earlier Experiments on Induced or Evoked Theta Oscillations 6.2.3 Gamma Frequency Range Selectively Distributed Oscillatory Systems: Distributed Multiple Oscillations 6.3.1 Concept, Definitions, and Methods 6.3.2 Oscillatory Responses in Invertebrate Ganglia 6.3.3 Gamma Oscillations in Sensory, Cognitive, and Motor Processes 6.3.3.1 Multiple Functions in Gamma Band 6.3.3.2 Important Causality Factor for Human Gamma Response



6.3.4

6.4 6.5 6.6 6.7

6.8

Alpha Oscillations in Perception and Cognition: The Alphas 6.3.4.1 Sensory Components 6.3.4.2 Cognitive Components 6.3.4.3 Resonance in Brain Responses 6.3.4.4 Multiple Functions in Alpha Frequency Window 6.3.5 Theta Oscillations in Perception and Cognition 6.3.6 Delta Oscillations in Cognition 6.3.7 Klimesch: Multiple Oscillatory Activities in Alpha Band 6.3.8 Oscillations in Highest Frequency Window Superposition Principle and Superimposed Multiple Oscillations in Theta and Delta Frequency Windows in Cognitive Processes: Examples Selectively Distributed and Selectively Coherent Oscillatory Networks Interim Conclusions Distributed Oscillatory Systems and Distributed Memory 6.7.1 Event Processing in Distributed Systems 6.7.2 Multiple Functions of EROs and Multiple Functions of Memory: Convergence of Concepts 6.7.3 Human Memory Performance and Time-Locked Theta Responses EEGs and EROs as Information Codes 6.8.1 Frequency Coding at Different Levels 6.8.2 Most General Transfer Functions and Multiple Oscillations

Chapter 7 7.1 7.2 7.3 7.4 7.5 7.6 7.7

Are Integrative Brain Functions Shaped by Superbinding and Selectively Distributed Oscillations?

Rationale and Usefulness of this Chapter Binding Problem in Memory Processing and Gestalt Neurons–Brain Theory and Oscillatory Codes Description of Function–Memory Table Super-Synergy: A Spatio-Temporal and Functional Organization of Multiple and Distributed Oscillations Gedanken Model: Involvement of Selectively Distributed and Coherent Activities of Neural Populations in Grandmother Percept Neural Populations and “Feature” Cells 7.7.1 Sokolov: Feature Detectors

Chapter 8

Grandmother Experiments in Perception of Memory: Recognition of Gestalts

Erol Basar ¸ and Murat Özgören 8.1 8.2 8.3

Introductory Remarks Klimesch: Role of Theta and Alpha Oscillations in Memory and Attention Functions Grandmother Paradigm and Gestalt Experiments 8.3.1 Experimental Strategy 8.3.1.1 Electrophysiological Recording 8.3.1.2 First Data Recording (Random) Set 8.3.1.3 Second Data Recording (Regular) Set 8.3.2 Event-Related Oscillations Arising from Light, Anonymous Face, and Grandmother Face Stimulations 8.3.2.1 Topologies of Delta Responses 8.3.2.2 Topologies of Theta Responses 8.3.2.3 Topologies of Alpha Responses



8.3.2.4 Distributed Beta and Gamma Responses Differentiation of Responses in Delta, Theta, and Lower and Upper Alpha Frequency Bands: Preliminary Statistics Recognition Memory and Gamma Oscillations What Does the Grandmother Paradigm Mean? Are Oscillations Distributed Templates in Memory Activation? 8.5.1 Selectively Distributed Enhancements in Whole Cortex 8.5.2 Efficiency of Grandmother Paradigm for Differentiation of Memory Components or States 8.5.3 Does Activation of Larger Neural Populations Indicate Reactivation of Episodic Memory? 8.5.4 Transition from Semantic to Episodic Memory: Distinctions between Semantic and Episodic Memories 8.5.5 Importance of Frontal Lobes and Other Brain Areas for Memory Processing and Perception 8.5.5.1 fMRI Experiments Related to Distributed Memory in Cortex 8.5.5.2 Critique of Experiments of Fernandez and Fell 8.5.5.3 Major Activation Areas of Semantic and Episodic Memories 8.5.5.4 Superbinding and Stryker’s Question about Oscillations 8.5.6 Do Grandmother Experiments Favor Hebb’s Hypothesis? Are the Descriptions of Gestalts and Emotions Related to More Complex Percepts Possible? 8.3.3

8.4 8.5

8.6

PART III Memory Function: Models and Theories Chapter 9 9.1 9.2

9.3

EEG-Related Models of Memory States and Hierarchies

Introduction of a New Construct on Memory Categorization Physiology of Selectively Distributed Oscillatory Processes 9.2.1 Connections of Sensory–Cognitive Systems 9.2.2 Activation of Alpha System with Light 9.2.3 Activation of Alpha System with Auditory Stimulation 9.2.4 Activation of Theta and Delta Systems Following Cognitive Inputs 9.2.5 Nonspecific Interactions Hierarchical Categorization of Different Levels of Memory 9.3.1 Fuster’s View of Memory Networks: A Milestone in Neuroscience 9.3.2 Tentative Model Related to EEG Activation 9.3.3 Inborn (Built-In) Networks (Level I) 9.3.3.1 Reflexes 9.3.3.2 Stereotypic Fixed Action Patterns 9.3.3.3 Phyletic Memory and Oscillatory Response Codes 9.3.3.4 Feature Detectors 9.3.3.5 Living System Settings 9.3.4 Physiological or Fundamental Memory 9.3.4.1 Changes of Sensory Memory: Spontaneous and Evoked Alpha Activity at Occipital Sites in Three Age Groups 9.3.5 Working Memory (Level II) 9.3.5.1 Perceptual Memory



9.3.6

9.4

9.5

9.6

Incorporation of Oscillatory Codes in Physiological Memory Consisting of Phyletic, Sensory and Perceptual Memories 9.3.7 What is Motor Memory? 9.3.8 Development of Procedural Memory throughout Life Dynamic Memory in Whole Brain: Memory States instead of Memories 9.4.1 Alpha, Theta, and Delta Oscillatory Processes during APLR 9.4.2 Are Dynamic EEG Templates Created during Processing of the APLR Alliance? Do They Build (Virtual) Short-Term Storage of Newly Learned Material? 9.4.3 Are All Brain Functions Linked with Memory? Complex Memory or Multiple Matching: Evolving Memory and APLR Alliance 9.5.1 Memory Activation: Working Memory and Hierarchical Organization of Memory States 9.5.2 Multiple and Complex Matching Processes: Reciprocal Activation of Alpha, Delta, Theta, and Gamma Circuits in Whole Brain 9.5.2.1 Reentry? 9.5.3 Prolonged Oscillations, Delays, and Coherent States during Complex Matching 9.5.4 Complex Matching 9.5.4.1 Matching of Multiple Oscillations in Whole Brain Longer-Acting Memory and Transition to Persistent Memory in Whole Brain 9.6.1 Evolving Memory: Multiple Level Functioning in CNS 9.6.2 Level III Activities Portrayed in Figure 9.7

Chapter 10 New Trends in Memory Dynamics: Concluding Remarks 10.1 The Emphasis of this Book: From a Research Program to a Theory on Whole-Brain Work 10.2 Distributed Memory in the Whole Brain 10.3 Correlation of Brain Oscillations with Multiple Brain Functions 10.3.1 Are All Memory States Tuned with Frequencies of EEG Oscillations? 10.4 Gestalts and the Grandmother Percept 10.5 Activated Memory Manifested by EEG Oscillations 10.5.1 Plausibility of Hebb’s Reverberating Activity Based on EEG Experiments 10.6 Model Related to Memory States 10.6.1 Active Memory and Reverberation Hypothesis 10.6.2 Memory State as Continuum 10.6.3 Multiple Matching with EEG Frequency Codes as an Essential of Recognition 10.6.4 Longer Acting Memory and Persistent Memory 10.7 Importance of EEG Analysis Chapter 11 Memory and Whole-Brain Work: Draft of a Theory Based on EEG Oscillations Erol Basar ¸ and Sirel Karakas¸ 11.1 Integration of Proposals Related to Whole-Brain Work 11.2 Whole-Brain Work Theory: How to Approach Brain Functions by Means of EEG Oscillations 11.2.1 Level A: Transition from Single Neurons to Oscillatory Dynamics 11.2.2 Level B: Superbinding of Neural Assemblies (Supersynergy) © 2004 by CRC Press, LLC


11.2.3 Level C: Integration, Alliance and Interplay in Memory 11.2.4 Level D: Causality and Brain Responsiveness 11.3 Newtonian Causality, Chaotic Dynamics, and Brain Language Epilogue:

From EEG–Brain Dynamics to Memory–Brain Dynamics

References Abbreviations and Glossary Appendix: Relevant Mathematical Methods


TF1663_book.fm Page 1 Friday, May 7, 2004 8:47 AM

Part I Foundations


TF1663_C01.fm Page 3 Thursday, May 13, 2004 2:24 PM

and Core 1 Introduction Philosophy 1.1 LANDMARKS: MEMORY IS DISTRIBUTED, MEMORY IS A DYNAMIC PROPERTY Memory is a functional property. Since brain functioning is based on dynamic processes, memory is also a dynamic process. Seeing even the simplest light signal is a memory process related to a fundamental inborn retrieval response. A baby perceives and shows reflex responses to light before he or she is exposed to more complicated learning processes. The response to light is probably a basic decoding process. Fuster (1997) stated that memory reflects a distributed property of a cortical system. Important components of higher nervous system functioning such as perception, recognition, language, planning, problem solving, and decision making are interwoven with memory. This author considers memory a property of the neurobiological systems it serves; it is inseparable from their other functions. Memory is a dynamic property of the brain as a whole rather than a characteristic of any single specific region; it resides simultaneously everywhere and nowhere in the brain (Rose, 1997). According to Antonio Damasio (1997), “Memory depends on several brain systems working in concert across many levels of neural organization. Memory is a constant work in progress.” How did neuroscientists arrive at such conclusions? The conclusions are based on a long evolution of thoughts and concepts originating with Karl Lashley, Donald Hebb, and F.A. Hayek in the first half of the 20th century. The following sections briefly explain the works of these pioneers.

1.1.1 LASHLEY’S EQUIPOTENTIALITY To study learning and memory concepts in mammals, Karl Lashley (1929) taught rats to successfully negotiate complex mazes. He then began incrementally removing thin slices of each rat’s cerebral cortex in an effort to pinpoint the memory locus for this task. No matter which sections of brain Lashley removed, the rats were still able to run the maze. Their performances diminished progressively as more brain tissue was excised, but Lashley found no single region whose ablation completely erased memory. In a landmark paper, Lashley proposed the theory of equipotentiality: memory is in fact scattered across the entire brain and is not concentrated in specific regions.

1.1.2 HEBB’S RULES

OF

COOPERATIVITY

Hebb`s rule (1949) implies that information processing requires functional cooperation by distributed neurons. More precisely, this rule postulates that groups of synapses that have a tendency to fire together and converge on a single neuron become strengthened as a group. This is known as the principle of cooperativity. Does some kind of modification of neurons or modification of connections between neurons occur as a result of learning? For example, when we learn to associate two stimuli (e.g., an unconditioned stimulus and conditioned stimulus as in classical conditioning), what happens in the brain to support the learning process?



Early attempts to answer this question can be traced back to Donald Hebb who in 1949 proposed that the coactivation of connected cells would result in a modification of weights and when a presynaptic cell fired, the probability of firing by a postsynaptic cell firing was increased. Hebb said, “When an axon of cell A is near enough to excite cell B or repeatedly or persistently takes part in firing it, some growth or metabolic change takes place in both cells such that A’s efficiency as one of the cells firing B is increased.” This learning principle did not specify exactly what was meant by growth or metabolic change, but it served as a useful starting point and has become the widely cited heuristic for neurobiological investigations of learning and memory. The distributed nature of activations in cognitive tasks described in this chapter may explain why Lashley thought that the brain operated as a whole. The cooperation among distributed structures of the brain is also a factor because the coherences are selectively distributed. Analysis of oscillations in several neural populations of the brain in parallel and in various frequency windows brought a new refinement to descriptions of the whole brain and cooperativity: The whole brain is activated in all perceptual and memory-related mechanisms. The intensity of electrical oscillatory responses is selective in neural populations. The links or cooperativity, measured by means of coherences and phase differences, also show varied degrees of intensities. Accordingly, we may explore new interpretations of the statements of Lashley and Hebb by using new tools to analyze the electrical activities of the brain during sensory−cognitive activities. Hebb rejected the notion that stimulus−response relationships could be explained by simple reflex arcs connecting sensory neurons to motor neurons. It was necessary to postulate “a central neural mechanism to account for the delay between stimulation and response.” Hebb believed that sensory stimulation could initiate patterns of neural activity that were centrally maintained by circulation in synaptic feedback loops. Such reverberatory activities made it possible for response to follow stimulus after a delay. Seung (2000) claimed that the validity of Hebb’s theory remained uncertain. Although the existence of the Hebbian synapse is not in doubt, whether delay activity is thoroughly reverberatory is still unclear. (See also Section 2.6.1 in Chapter 2.) Electroencephalogram (EEG) studies recorded several delays and prolongations of responses (Chapter 3 and Chapter 4). Are the delays and prolongations candidates for reflecting Hebbian reverberatory mechanisms? Although no concrete answer can be provided, the possibilities will be discussed in Chapter 9. In the author’s opinion, the delays and prolongations of oscillatory responses reflect prolonged work of neural populations following difficult cognitive or memory tasks and their analysis can provide important hints for establishing learning and remembering models.

1.1.3 HAYEK: PERCEIVING ASSOCIATIVE NETS

IS

CLASSIFICATION

OF

OBJECTS

BY

ACTIVATION

OF

Associative networks play important roles in complex dynamics. Such networks are also considered essential building blocks in modern memory research. Hayek’s work (1952) was described in a very comprehensive and useful manner by Fuster (1995) who found Hayek’s work more important than Hebb’s related to describing memory function. Perceiving is the classification of objects by activation of the associative nets that represent them in memory. According to Fuster (1997), our thinking about the cortical organization of primate memory is undergoing a Copernican change — from a neurophysiology that localizes different memories in different areas to one that views memory as a distributed property of cortical systems. According to Fuster’s empirically founded hypothesis, the same cortical systems that serve us in perceiving the world also serve us in remembering it. Perceiving is the classification of objects by activation of the associative nets that represent them in memory. It is reasonable to assume, as Hayek did, that memory and perception share the same cortical networks, neurons, and connections to a large extent. To understand the formation and topography of memory, it is useful to think of the primary and sensory motor areas of the cortex that we may call the phyletic memory or the memory of the species. The primary sensory © 2004 by CRC Press, LLC


and motor cortices may be considered funds of memory acquired by a species through evolution. We can use memory as part of the term because, like personal memory, the phyletic memory consists of information that has been acquired and stored and can be retrieved (recalled) by sensory stimuli or the need to act.

1.2 NEW TRENDS IN NEUROSCIENCE Between 1980 and 2000, seven important steps in neuroscience research advanced our understanding of brain dynamics and function: 1. The discovery of oscillatory phenomena at the cellular level based on the 40-Hz studies by Singer (1989) and Eckhorn (1988), measurements of 10- and 5-Hz oscillatory behavior at the membrane level, and extracellular single recordings (Dinse et al., 1997, Llinás, 1988). 2. The application of chaos theory to electroencephalogram (EEG) signals, demonstrating that the EEG is not only a noise signal (for reviews see Basar ¸ , 1990; Duke and Pritchard, 1991; Molnár, 1999). 3. Developments based on the acceptance of cognitive function analysis by the use of the EEG and event-related potentials (ERPs). 4. The use of the magnetoencephalography (MEG) and functional magnetic resonance imaging (fMRI) as complementary tools. 5. The development of fast laboratory computers and availability of sophisticated neurocomputing software that accelerated progress in all fields of research. 6. The binding hypothesis occupied an important place in conceptual discussions, although we strongly emphasized that it is not sufficient to explain the mechanisms of complex percept building. 7. Copernican changes in memory research, particularly as discussed in the publications of Fuster (1995 and 1997); Goldman-Rakic (1997); Mesulam (1990 and 1994); and Kandel (1982). (See also Section 2.6.1.3.)

1.3 COPERNICAN CHANGES IN MEMORY RESEARCH 1.3.1 DISTRIBUTED NETWORKS According to Fuster (1997), the classic terms (representation, retrieval, recall, recognition, shortterm memory, and long-term memory) remain valid, but need to be neurobiologically redefined. Arguably, the smallest memory network (netlet) is a cortical cell group or module representing a simple sensory response; a memory reflects a distributed property of a cortical system. It can be hypothesized that selectively distributed oscillatory systems (or networks) may provide a general communication framework and be useful for functional mapping of the brain (Mesulam, 1990 and 1994). Communications in these networks may contribute to the formation of specific templates belonging to objects and memories. According to a model of cognition, this formation occurs as selectively distributed processing with considerable specialization and in anatomically differentiated localizations (Mesulam, 1990 and 1994; for details about memory as a distributed property of a cortical system, see also Fuster, 1997). In particular, analysis of hypothetical distributed oscillatory systems may lead to fundamental functional mapping of the brain, complementary to morphological studies. Perceptual memory is acquired through the senses. It comprises all that is commonly understood as personal memory and knowledge, i.e., representation of events, objects, persons, animals, facts, names, and concepts. From a hierarchical view, at the bottom level are memories of elementary © 2004 by CRC Press, LLC


sensations; at the top are abstract concepts that, although originally acquired by sensory experience, have become independent as a result of cognitive operations. Single neuron recordings in monkeys trained to perform working memory tasks have identified components of a working memory circuit in the prefrontal cortex. The neuronal processes related to task performance can be dissociated on the scale of milliseconds to seconds. During a working memory task, as the stimulus is sequentially registered and stored over a period of seconds and then translated into a motor response, specific neural populations respond in characteristic ways. One class of prefrontal neuron responds to a visual stimulus as long as the stimulus is in view. In contrast, other prefrontal neurons are activated at the offset of the stimulus and remain active as long as the monkey must remember the location or features of an object (Fuster, 1995; GoldmanRakic 1988 and 1997). As one can deduce from the work of Mesulam and Fuster, common codes for perpetual signal transfers between neural networks for parallel and serial processing and also for possible reverberation circuits and loops between neural network must exist. Oscillations in the brain may serve as adequate codes for this general communication by inciting networks to resonate. A more general view is that functional or oscillatory network modules are distributed in both the cortex and throughout the whole brain (Basar ¸ , 1999). We will now discuss an electrophysiological (EP) parallel between Fuster’s memory network and the distributed oscillatory systems mentioned earlier. When analyzing field potentials, it is difficult to define boundaries of brain nuclei and their electrical activities. Nevertheless, this approach is useful because great amounts of data can be collected and interpreted from several electrodes distributed in the brain. Furthermore, it is possible to perform measurements during continuously changing cognitive states. EP studies and EEG segments from the cortex, limbic system, thalamus, and cerebellum can be recorded and compared in waking and freely behaving animals. This type of recording during behavioral states cannot possibly be managed with single cell electrodes. Studies of functional correlates of structures like sensory cortices, hippocampi, and thalamic relay nuclei are based mostly on experiments using unit recordings. A major difficulty with interpretation of experiments made by single unit recordings (for example, experiments on corticothalamic information transfer) is that the results are limited to a few neurons. Accordingly, the author assumes that every hypothesis on localization of the thalamocortical circuit as a 10-Hz generator is restricted and not acceptable with regard to the results of experiments described in this book: the alpha, theta, and gamma generators are selectively distributed in the brain. 1.3.1.1 Distributed Memory: Findings with Functional Magnetic Resonance Imaging Cohen et al. (1997), and Courtney et al. (1997) used fMRI studies of humans to find parallels to the knowledge gained from single-cell recordings of animals. Courtney presented subjects with pictures of human faces and asked them to recall whether each picture was the same or different, from a picture presented 8 s earlier. Activations in the prefrontal areas correlated most strongly with delay periods, compared with activations in the visual areas that more strongly correlated with sensory stimulation. Cohen et al. presented subjects with single written consonants every 10 s and asked the subjects to judge whether each consonant was the same as a consonant presented one, two, or three trials back in the sequence. This task required subjects to remember the identities of the consonants and the order in which they were presented. The farther back in the sequence the consonant to be recalled appeared, the greater the load on the working memory. The authors showed that activations in the prefrontal cortex were maintained throughout the 10-s interstimulus intervals. The degree of prefrontal activation was higher for conditions with the greatest memory loads. By contrast, activations by the primary visual, somatosensory, and motor cortices and in several secondary regions were not sustained across the 10-s interval and were not related to memory demand. They © 2004 by CRC Press, LLC


were probably responsive to sensory or perceptual stimuli, but did not aid the working memory in performing tasks. Based on the fMRI results of Courtney et al. (1997), early extrastriate visual areas demonstrated transient, relatively nonselective responses to complex visual stimuli and later extrastriate visual areas demonstrated transient, selective responses to faces. This indicated a more specialized role in the processing of meaningful images. Both extrastriate visual and prefrontal cortical areas demonstrated sustained activity during memory delays, indicating a role in maintaining an active representation of the face in working memory.

1.3.2 PARALLEL DISTRIBUTED PROCESSING According to parallel distributed processing (PDP) or the connectionist model (McClelland, Rumelhart, and PDP Research Group, 1986; Rumelhart, McClelland, and PDP Research Group, 1986) of cognitive psychology, information processing takes place through the interactions of a large number of simple processing elements. The connections between elements of information can be active at the same time and this enables the system to manipulate a large number of cognitive operations simultaneously. The PDP model conjectures that parallel distributed processing occurs through a network distributed across incalculable numbers of locations in the brain. The Goldman-Rakic (1988) hypothesis on parallel sensory−cognitive processing, Mesulam’s (1990) distributed processing in large-scale neurocognitive networks, Fuster’s (1995) cortical memory, and Basar ¸ ’s (1998 and 1999) theory of oscillatory neural assemblies are psychophysiological counterparts of the PDP model. Karakas¸ et al. (2000) commented on parallel processing: In the formulation of our research group, parallel distributed processing is based on the oscillatory activity, the EEG and event-related oscillations (EROs) of various frequencies; each oscillation represents multiple functions and, conversely, a given function is represented by multiple oscillations. The cognitive functions are represented by integrative activity of neuroelectric oscillations that occur in parallel (Basar ¸ , 1998 and 1999; Basar ¸ and Karakas¸, 1998; Quiroga and Schürmann, 1998).

1.4 EEG−BRAIN DYNAMICS Electroencephalogram−brain dynamics can be defined simply as measuring electrical activity (or magnetic fields) of the brain recorded from large numbers of neural populations selectively distributed in the whole brain by using large scalp or intracranial electrodes approximately 100 µm in diameter. It seems clear that this method is one of the fundamental approaches to understanding integrative brain functions. However, scientists working in the field of brain macrodynamics had a long way to go before the relevance of EEG and MEG studies for elucidating brain functioning became clear. Although scalp EEG activity was measured by Hans Berger in the 1920s and Lord Adrian (1942) initiated basic research with EEG oscillations, functional EEG research remained in the shadows of the single cellular level until the 1980s.

1.4.1 IMPORTANCE

OF

EEG STUDIES

Figure 1.1 illustrates new approaches and strategies in functional neuroscience. The utility of the ensemble of methods is emphasized because the application of a single method has severe shortcomings for elucidating integrative brain functions. The methods range from indirect means of measuring changes in cerebral blood flow in local regions of the human cortex with fMRI to measuring changes in electrical activity via EEG recordings of the human brain with multiple electrodes and surgically implanted multiple electrodes in primates. According to Mountcastle (1998), measurement of large populations of neurons is presently the most useful experimental paradigm used in perception experiments. However, fMRI has the © 2004 by CRC Press, LLC


APPLIED DOMAINS _________ EVOLUTION OF BRAINS

Single Unit Studies ?

AGING PATHOLOGY (e.g., Alzheimer’s disease, Parkinson’s disease, epilepsy, multiple sclerosis) + Biochemistry Pharmacology

Mathematical and Technical strategies Chaos, Neural Networks

PET

EEG/ERP MEG/MEF [Event-Related Oscillations] EROs

BASIC FUNCTIONAL LEVEL __________ SENSORY DETECTION

fMRI

PERCEPTION COGNITION MEMORY MOVEMENT

Psychophysiology Attention, Perception Learning and Memory Paradigms and Tests

ATTENTION

FIGURE 1.1 New approaches and strategies in functional neuroscience.

disadvantage of low temporal resolution and long distance measurements with multiple microelectrodes cannot be yet performed. Therefore, measurements of macro-activities (EEG, ERP, and MEG) seem to be the most adequate methods of measuring the dynamic properties of memory and integrative brain function. Since neuroscientists have concluded generally that several different brain regions must cooperate to accomplish any brain function, the analysis of the relationships of different regions of the brain is becoming more important. We will now discuss the methods and strategies cited in Figure 1.1. Strategy is defined as combined (parallel or sequential) applications of several methods. Studies at single-cell level have been of great importance in eludicating the basic physiological mechanisms of communications between cells (Mountcastle, 1998; Eccles, 1973). However, the importance of these studies for understanding of integrative brain functions is questionable because the whole brain is involved during integrative processes, as Ross Adey (1960, 1966, and 1989) noted and the new trends in neuroscience clearly emphasize (see Freeman, 1999). Positron emission tomography (PET) is an invasive method that allows large temporal resolution within 30 min, but offers no possibility of dynamic measurements within microseconds. The methods incorporating analyses of EEG, ERP, and EROs with fMRI provide additional strategies to illuminate brain functions since they cover dynamic changes in the brain and morphological structures. MEG and studies of event-related magnetic fields (MEFs) greatly increase spatial resolution in comparison to EEG and ERP. Accordingly these methods show great promise in future applications. The new strategies are interwoven with relevant use of mathematical and psychophysiological strategies including: 1. Theoretical mathematical and systems approaches such as (a) chaos, entropy, (b) modelling with neural networks, (c) a frequency domain approach combined with wavelet analysis and spatial and temporal coherence (Bullock, 1989; Petsche, 1998; von Stein, 2000). 2. Psychological strategies involving behavioral paradigms and application of neuropsychological tests (Karakas¸ et al. 2002 and 2003). One important strategy not cited in Figure 1.1 is recording of data via surgically implanted intracranial electrodes in animal brains. © 2004 by CRC Press, LLC


To achieve relevant progress in functional neuroscience, it is essential to combine several methods (Freeman, 1999) although application of all strategies in every laboratory is impossible. Figure 1.1 illustrates levels of basic central nervous system (CNS) functions (right side) and applied domains (left side). Sensory detection, movement, and memory functions can be analyzed successfully by using individual methods or strategies in research domains that study evolutionary developments, aging, pathology, and pharmacology (use of drugs or pharmacological agents to treat pathologies). Applications of combined strategies at basic scientific levels and in all these specialized fields may reveal new horizons for understanding integrative functions of the brain and especially memory function. The importance of memory and its influence on behavior cannot be overly emphasized because few aspects of higher nervous functioning could operate successfully without some memory contribution. Perception, recognition, language, planning, problem solving, and decision-making abilities all rely on memory (Damasio and Damasio, 1994).

1.5 PIONEERING STUDIES OF BRAIN MACRODYNAMICS AND WHOLE BRAIN APPROACH The domain of mechanics that involves the motions of bodies without reference to the causes of motion is called kinematics; the domain that studies the resulting motions is called kinetics. These two domains constitute the field of dynamics, also known as Newtonian dynamics: The dynamic activities of the brain involve mutual influences of bodies as reflected by kinetics. The analysis of trajectories reflecting the activities of neuronal populations is somewhat similar to the analysis of motion. Accordingly, we will use brain dynamics to describe the causes or mechanisms that give rise to trajectories manifested as electrical signals from the brain. The early studies of Lashley (1929) and Hebb (1949) established important principles of memory and integrative brain function. Between 1960 and 1971, several biophysicists and neuroscientists including Fessard (1961), Griffith (1971), and Rosen (1969) published relevant studies indicating the transition from single neuron dynamics to the dynamics of neural populations. The finding of this transition by a group of scientists opened a new era of behavioral neuroscience based on the oscillatory dynamics of neural populations. The important issue of oscillatory brain dynamics at the functional level is explained in several books and papers (Basar ¸ et al., 2003).

1.5.1 GRIFFITH: STATISTICAL MECHANICS

IN

BIOLOGY

AND

PHYSICS

Griffith (1971) discussed the concepts of statistical neuron dynamics and tried to formulate a similarity between statistical mechanics and neurodynamics as follows: The situation is superficially very similar to that which is obtained in statistical mechanics, as it applies to the relation between macroscopic thermodynamic quantities and the underlying microscopic description in terms of the complete specification of the states of all the individual atoms or molecules …. These are, first, that we could not, even if we knew all the necessary parameters, actually solve in detail the 1010 or more coupled neuronal “equations of motion” necessary to follow the state of the system in detail as a function of time. Second, that there exists a simpler “macroscopic” level of description which is really our main ultimate object of interest so that we do not wish, even if we could, to follow the “microscopic” state in detail but merely wish to use it to understand the time development of the macroscopic state. One most important aspect of this is that we only wish to specify, at the macroscopic level, the initial conditions of any calculation we may make. This leads immediately to the problem of whether the fundamental assumptions of equal a priori probabilities and random a priori phases hold for nerve cell aggregates, and, if not, whether we can find anything to replace them.

Griffith’s remarks are more important now than they were 30 years ago because new trends in brain research clearly indicate the need to introduce new frameworks to analyze integrative brain functions by studying cell aggregates rather than single cells. © 2004 by CRC Press, LLC


1.5.2 ROSEN: GLOBAL NEURODYNAMICS In 1969, Rosen questioned the role of statistical mechanics in gas dynamics? The physics laws describing gas dynamics are based on an ensemble of molecules in an isolated system. One does not describe gas dynamics based on single molecules in an isolated system. After the laws were determined experimentally, attempts were made to correlate the macrosystem laws with dynamics at the microlevel, i.e., with gas molecules. In other words, the laws of gas dynamics were determined before they were correlated with molecular properties. This principle is complementary to Griffith’s statements cited above. Basar ¸ (1980 and 1998) commented on the works of Rosen and Griffith: In the analysis of brain waves, we are certainly interested to discover the particular properties of individual neurons and their relation to the gross activity. To further examine the problem of the correlation of single unit activity (microactivity) and gross activity (macroactivity), Rosen (1969) explained the concepts of statistical mechanics and physics and their relation to neurobiology: What is the micro-description? We know that here the fundamental state variables are the displacements and momenta of the individual particles which make up our system. According to Newtonian dynamics, the kinetic properties of the system are given by the equations of motion of the system, which express the momenta as functions of the state variables. The basic postulate of Newtonian dynamics is the following point: knowing the state variables at one instant and the equations of motion, we are supposed to be able to answer any meaningful question that can be asked about the system at any level. Statistical mechanics however identifies a macrostate with a class of underlying microstates, and then expresses the global state variables as averages of appropriately chosen micro-observables over the corresponding class of microstates.

1.5.3 FESSARD: GENERAL TRANSFER FUNCTIONS

OF THE

BRAIN

Fessard (1961) emphasized that the brain must not be considered simply as a juxtaposition of individual lines, leading to a mosaic of independent cortical territories, one for each sense modality, with internal subdivisions corresponding to topical differentiations. What are the principles that dominate the operations of hetero-sensory communications in the brain? The answer requires extensive use of multiple microelectrode recordings along with a systematic treatment of data by computers (Gray and Singer, 1989; Elkhorn et al. 1988). Fessard indicated the necessity of discovering principles that govern the most general — or transfer — functions of multiunit homogeneous messages through neuronal networks. The transfer function describes the ability of a network to increase or impede transmission of signals in given frequency channels. The transfer function represented mathematically by frequency ˘ et al., 1992) constitutes the main framework characteristics or wavelets (Basar ¸ , 1980; Basar ¸ -Eroglu for signal processing and communication. The existence of general transfer functions could then be related to a series of networks having similar frequency characteristics to facilitate or optimize signal transmission in resonant frequency channels in the brain (Basar ¸ , 1998). In an electric system, optimal transmission of signals is reached when subsystems are tuned to the same frequency range. Does the brain have such subsystems tuned to similar frequency ranges or do common frequency modes exist in the brain? The empirical results reviewed here imply a positive answer and provide a satisfactory framework to Fessard’s question. Frequency selectivities in all brain tissues containing selectively distributed oscillatory networks (delta, theta, alpha, beta, and gamma) constitute and govern mathematically the general transfer functions of the brain. To fulfill Fessard’s hypothesis, all brain tissues of mammalians and invertebrates would have to react to sensitive and cognitive inputs with oscillatory activities or similar transfer functions. The synchronies, amplitudes, locations, and © 2004 by CRC Press, LLC


durations or phase lags vary continuously, but similar oscillations are most often present in activated brain tissues (Basar ¸ , 1999). The general transfer functions of the brain manifested in oscillations strongly indicate that frequency coding is one of the major candidates of brain functioning, as noted earlier. We will further discuss Fessard’s work in Section 6.8.2.

1.5.4 EDELMAN: REENTRANT SIGNALLING THEORY

OF

HIGHER BRAIN FUNCTION

G.M. Edelman (1978) posed important questions about brain functioning. Does the brain operate according to a single principle carrying out its high-order cognitive functions? That is, despite manifold differences in brain subsystems and the particularities of their connections, can one discern a general mechanism or principle required for the realization of cognitive facilities? If so, at what level — cellular, molecular, or circuit — does the mechanism operate? By means of EEG analysis we can try to determine general mechanisms or principles at the levels of neural populations or circuits of cells. We may also open other important avenues and find additional laws, as Rosen and Griffith did. More important is Fessard’s question about general transfer functions of signal communications in the brain. Mountcastle (1976) noted that the central problem of brain physiology was how to understand the actions of large populations of neurons, actions that may not be wholly predictable from properties of subsets. He also noted that the central problem of the intrinsic neurophysiology of the cerebral cortex was to discover the nature of neuronal processing with the translaminar chains of interconnected cells (in columns). Edelman (1978) transformed these statements by noting that the main problem of brain physiology was “to understand the nature of repertoire building by populations of cell groups.” As noted earlier, EEG oscillations represent primitive study methods and should be considered as building blocks for further techniques. Edelman (1987) developed a theory of neuronal group selection. This theory assumes a genetic endowment of neuronal groups such as the columnar modules of the cortex with inherent degrees of variability and plasticity in their connections. They constitute the units of selection of the primary repertoire. By exposure to external stimulation and a Hebbian mechanism, certain groups of cells that tend to fire together will be selected by stimuli insofar as groups respond to them, and thus their connections will be strengthened (Figure 1.2). Some of those connections will make recurrent or re-entrant circuits that are essential features of the model and of its theoretical and computational elaborations (Tononi et al., 1992). Groups not selected will be crowded out by the competition. According to Edelman, re-entry is dynamic and can occur via multiple parallel and reciprocal connections. It takes place between populations of neurons rather than between single units. Neurons within a group tend to be strongly connected. At higher levels, the integration of perceptual and conceptual components is required to categorize objects. See also Chapter 9 and Chapter 10.

1.6 FREEMAN, KATSCHALSKY, AND HAKEN: PRELIMINARY STEPS IN INTRODUCING MACRODYNAMICS OF ELECTRICAL ACTIVITY The mechanisms of self-organization via oscillations through various kinds of interactions in physical, chemical, biological, psychological, and social systems have been most deeply explored by Aharaon Katzir-Katchalsky (1974) and Ilya Prigogine (1980) in studies of dissipative structures and chaotic state transitions. According to Prigogine’s theory, no system is structurally stable; fluctuations lead to instabilities and two new types of functions and structures. The evolution of a dissipative structure is a self-determining sequence. Scheme 1.1 shows the relationships of function, structure, and fluctuation. This approach combines both deterministic and probabilistic elements in the time evolution of the macroscopic system. © 2004 by CRC Press, LLC


S1

S2

S3

R1

R2

R3

Si

Rk

Rj

FIGURE 1.2 Edelman’s principle of group selection. At birth, a primary repertoire of responses (R1, R2, R3, …) by cortical neuron groups can be potentially elicited by any of a series of stimuli (S1, S2, S3…). After learning or repeated experience, a given stimulus, Si, will elicit many or only one of those responses RK. (Modified from Edelman, G.M., Neural Darwinism, Basic Books, New York, 1987.)

Freeman’s viewpoint (1999) is that complex biochemical feedback pathways within cells support the emergence of oscillations at cycle durations of minutes, hours, and days and underline the recurrence patterns of normal cyclical behavior as well as epilepsies, mood disorders, and other pathologies. A large number of neurons form macroscopic populations under the influence of external and internal stimuli and endogenous neurohormones. Freeman’s opinion is that these populations are more closely related to the nerve cell assemblies conceived by Hebb (1949). The relationships of the neurons to the mass are explained by Haken’s synergetic theory (1977) whereby microscopic neurons contribute to the macroscopic order and then are “enslaved” in a manner similar to the containment of particles in lasers and soap bubbles. Freeman (1975) achieved an important step in understanding the dynamics of populations of neurons (macrosystems) and also finding correlations between the activities of single neurons and population responses by starting with induced gamma activity in the olfactory bulb of the rabbit. Our group tried to determine the dynamics of brain responses in an abstract way, then tried to show, based on existing neurophysiological data, what particular neural responses could give rise to the general transfer functions cited by Rosen and Fessard (see Basar ¸ , 1980 and 1999). Oscillatory responses and resonance phenomena in the alpha, beta, theta, delta, and gamma frequency ranges govern the brain dynamics as revealed by macroscopic brain activity. Resonance is the response that may be expected of underdamped systems when a periodic signal of a characteristic frequency is applied to the system. The response is characterized by surprisingly large output amplitude for relatively small input amplitude. We were looking for codes related to general dynamical rules and links between macrodynamics and microdynamics and between brain oscillations and functions. Our research and experimental foundations are described by Basar ¸ et al. (2004). The alliance of perception and memory based on concepts of Hayek (1952) is described with electrophysiological measurements and systems theory tools. The building of a general framework of macroscopic brain dynamics led to useful categorization of integrative brain functions. In order to find codes and general dynamic rules in the sense of Rosen (1969) and Fessard (1961), the biological systems analysis program was applied. This will be explained in the next section.



1.7 APPLICATION OF PRINCIPLES OF BIOLOGICAL SYSTEM ANALYSIS TO BRAIN RESEARCH 1.7.1 REASONS

FOR

ESTABLISHING PROGRAMS

FOR

BRAIN RESEARCH

In the 1970s, Basar ¸ and coworkers tried to determine the dynamics of brain responses in an abstract way and named their approach a Program For Biological System Analysis. They tried to show, based on existing neurophysiological data, what particular neural responses could give rise to general transfer functions. The program was extended and modified in 1998 and designated the Brain Dynamics Research Program (Basar ¸ , 1976, 1980, and 1998). In the meantime, a number of other research groups applied some of the steps or the global concept of the program. In addition to the classical analysis tools of general systems theory, some supplementary experimental methods and methods based on the special natures of living systems are parts of this program. The program has three main classes of methods: (1) abstract methods of general systems theory, (2) specific methods for living systems, and (3) methods of thoughts and research principles. Figure 1.3 illustrates a more advanced version of a biological systems analysis and brain dynamics research program. The rationale for developing a research program was based on elucidating the black box (brain). Three basic quantities involved in biological investigations are the input (stimulus), the system, and the output (response). If the stimulus and response are known or are measured variables, it should be possible to estimate the properties of the system (the whole brain). The determination of the abstract frequency characteristics or transfer function of the system under study usually causes experimental and sometimes conceptual difficulties. This is partly due to rapid changes of the parameters measured. Mathematical representations merely help identify the frequency positions of all components without determining the exact natures of the components. At this stage, a researcher must elucidate the black box. Since the determination of mathematical characteristics alone did not allow concluBRAIN DYNAMICS RESEARCH PROGRAM I. ABSTRACT METHODS FOR SYSTEM ANALYSIS 1. a) b) c) d)

Methods to analyze brain states Power spectral density Cross correlation Cross spectrum Coherence

2. a) b) c) d) e)

Methods to analyze evoked brain activity Transient response analysis Frequency analysis Response adaptive filtering Combined EEG-EP analysis Evoked coherence

II. SPECIFIC METHODS FOR ANALYSIS OF THE BRAIN FUNCTION 1. Application of pharmocological agents

III. METHODS OF THOUGHT OR RESEARCH PRINCIPLES 1. Going into the system 2. Going out of the system

2. Selective blocking ot the system

3. New emerging methods to analyze eventrelated oscillations a) Wavelet analysis b) Wavelet entropy c) Single sweep wave identification d) Event related oscillations e) Study of nonlinearities and chaos approach

3. Reduction of the system into its passive response

3. Consideration ot the system as a whole

4. Application of various paradigms to influence the state of consciousness and alertness (attention, learning etc.) 5. Paradigms with complex gestalts, as “grandmother face”

A new framework to extend the Neuron Doctrine, considering the brain as a whole: “Neurons Brain Theory” related to whole brain work

FIGURE 1.3 Brain dynamics research program. (Modified and extended from Basar, ¸ E., Brain Function and Oscillations, Vol. 1, Brain Oscillations: Principles and Approaches, Springer, Berlin, 1998, p. 153.) © 2004 by CRC Press, LLC


sions about the biophysical nature of the phenomenon, the difficult problem was to establish the biological systems analysis theory. The application of this program has followed the lines of thinking of Fessard, Griffith, and Rosen to develop transfer functions and the thinking of Hebb related to long distance coherence of macroscopic electrical activity of the brain. Investigators in the field of brain studies usually deal with gray boxes (partially elucidated black boxes). An apparatus or system is designated a gray box when it performs a defined operation and provides information about the structure or processes making possible (realizing) the defined operation. A gray box generates partial information concerning the structures and processes that realize input−output relations (Basar ¸ , 1998). In the context of the general framework of the brain dynamics research program, we developed certain research principles or strategies that allowed us to add to our knowledge about brain functioning. In fact, every neuroscientist has his own surroundings and develops his own definitions and classifications of signals studied. This approach has helped to expand our knowledge of global brain dynamics and global brain functions as reflected by EEG and oscillatory brain responses. 1.7.1.1 Program Steps The ensemble of abstract methods of brain state analysis shown in Figure 1.3 includes (1) power spectral density, (2) cross-correlation, (3) cross-spectrum, and (4) coherence. Combined EEG and EP analysis and wavelet analysis methods are also used to analyze brain activity. Basar ¸ ’s group first used conventional methods to apply abstract techniques to brain wave analysis. The group later performed studies on event-related oscillations (EROs) using abstract methods including long distance coherence and new methods such as wavelet entropy (Rosso et al., 2001; Quiroga et al., 1999). The third group of abstract methods shown in the figure includes emerging methods for analyzing EROs. The study of nonlinearities and chaos approach aims at understanding additional properties of the system. Specific methods for analysis of brain function included application of pharmacological agents and blocking of the system. Most importantly, the applications of different paradigms produced very interesting results and formed the frameworks of studies for complex gestalts such as the grandmother cell and similar techniques (Basar ¸ , 2003; also see Chapter 8, this volume). Treating the brain as a system means that the brain consists of a collection of components or subsystems arranged and interconnected in a definite way. One possible approach to understanding the brain system as an entity is to isolate the subsystems and study their specific properties. As a next step, one should understand how the subsystems are interconnected and which specific relations determine their integrative functioning. After determining subsystems and their interrelations, the next step is to try to model and reconstruct the whole entity. Abstract methods and their analogues used to analyze living systems aim at isolating distinct components. This approach is informative and defines a strategy generally called going into the system.* The conceptual framework provides us another, far more important, research strategy that cannot be realized by any of the analysis methods available or by their combined application. This strategy is called going out of the system and is defined as a method of thought. It is well known, for example, that a word is an abstract representation extracting the most essential attributive features from an enormous group of single concrete objects. In a similar manner, by using the method of thought, one can approach the essential principles of brain functioning by removing specific concrete representations and simultaneously extracting common building units. This can be achieved by going out of the system.** The principle of going out of the system is important for comparing the anatomy and physiology of the brains of humans and invertebrates, for example. This was the essential step undertaken by * This strategy is the search of the microstructure. ** This is the comparison of the properties of the analyzed system with those of other systems. Example: comparison of the circulatory system of the kidneys with the circulatory system of the brain.



Darwin in establishing comparative biology. Going out of the system involves another important comparison. Analyzing the frequency responses of the cortex, hippocampus, and other structures within the same brain can provide important information regarding parallel processing, thus contributing to our knowledge of fundamental building blocks. We should also consider interpretation of results obtained from investigation of cellular and structural systems. Several research groups derived some of the most accepted consequences through application of the brain dynamics research program. The functional significance of oscillatory neural activity began to emerge from the analysis of responses to well-defined events (EROs phase- or time-locked to a sensory or cognitive event). It is possible to investigate such oscillations by frequency domain analysis of ERPs based on the following hypothesis. An EEG analyzes the activities of an ensemble of generators producing rhythmic activities in several frequency ranges. These oscillators are active usually in a random way. However, the application of sensory stimulation to these generators enables them to couple and act together coherently. This synchronization and enhancement of EEG activity gives rise to evoked or induced rhythms. Evoked potentials representing ensembles of neural population responses were considered the results of a transition from a disordered to an ordered state. A compound ERP manifests a superposition of evoked oscillations in EEG frequencies ranging from delta to gamma. Natural frequencies of the brain range include alpha (8 to 13 Hz), theta (3.5 to 7 Hz), delta (0.5 to 3.5 Hz), and gamma (30 to 70 Hz). (See Yordanova and Kolev, 1998 and Chapter 6, this volume.) These statements clearly indicate that the macrodynamics of the brain are governed by oscillatory EEG dynamics that provide important keys to understanding brain function. The concerted application of all three steps of the brain dynamics research program led to a new framework, namely the neurons−brain theory that will be explained in the next section. The development and achievements of the theory and the grandmother gestalt experiments discussed in Chapter 8 are based mainly on the implications of this program and extend the neuron doctrine to consideration of the brain as a whole. The brain dynamics research program methods provided conventional tools, continuously developing principles, and new applications. The program also provided a wide spectrum approach. It did not limit the research field to a single frequency (e.g., 40 Hz) window and allowed us to pursue the super-synergy concept. New methods such as wavelet entropy studies are also applicable within this framework (Quiroga et al., 1999 and 2001; Rosso et al., 2001). 1.7.1.2 Mathematical Methods of Program Mathematical methods are covered by Basar ¸ (1998) and Basar ¸ et al. (2001d). A vast amount of literature discusses the analysis of chaos, wavelets, and wavelet entropy. We introduce amplitude frequency characteristics in the Appendix at the end of this book because the method is not yet covered in the literature.

1.8 NEW APPROACHES TO BRAIN FUNCTIONING AT MACROSCOPIC LEVEL 1.8.1 SHERRINGTON’S NEURON DOCTRINE REVISITED Studies of Ramon Cajal (1911) related to neuron morphology and the physiological approach of Sherrington (1948) led the way to the single-neuron doctrine with the notion of one ultimate pontifical nerve cell that integrated CNS function. In this concept, the integration was related to motor activity; the functional mapping was a type of movement mapping. Memory and cognitive functions were not interwoven in the physiological descriptions. Horace Barlow (1972 and 1995) transformed and replaced the first interpretation by specifying a feature of an object, such as a line, color, or tone represented by the firing of a neuron. © 2004 by CRC Press, LLC


In the first half of the 20th century, the invention of the EEG was followed by an explosion of publications related to brain function. The invention renewed the hope of tapping a physical correlate of mental performance (psychic energy described by Berger, 1929). The principles introduced by Berger and experimentally supported by Adrian (1941) remained in the shadows of neurophysiology research limited to the single-neuron approach. Mountcastle (1992) said: Suddenly a paradigm change is upon us, considering slow oscillations as active agents for signal transmission … stimulusinduced slow wave oscillations are related to/are signs of/generate such complex brain functions as perception, execution of movement patterns, or storage in memory — in short what is called cognitive neuroscience. The new developments demonstrate that it is not possible to interpret the functional contributions of alpha, theta, and delta, and gamma responses with only the neuron doctrine originally proposed by Sherrington (1948). The generators giving rise to these frequency responses are extremely sensitive to the modalities of sensory and cognitive inputs. Tracking properties of functionally related distant single neurons is not yet possible because of technical limitations. Goldman-Rakic (1988 and 1997), in search of a topography of cognition, concluded: If subdivisions of limbic, motor, sensory, and associative cortex exist in developmentally linked and functionally unified networks, as the anatomical, physiological, and behavioural evidence suggests, it may be more useful to study the cortex in terms of information processing functions and systems rather than traditional but artificially segregated sensory, motor, or limbic components and individual neurons within only one of these components. These new developments followed the proposals of Griffith (1971) and Rosen (1969) and the concepts of unification of functional networks of Hayek (1952), and supported the renaissance of use of the EEG in functional neuroscience.

1.8.2 RENAISSANCE

OF

EEG USE

IN

SEARCH

OF INTEGRATIVE

BRAIN FUNCTIONS

A special issue of the International Journal of Psychophysiology (Basar ¸ , Hari, Lopes da Silva, and Schürmann, 1997) and a new book dedicated to Hans Berger (Basar ¸ , 1999) described a quiet revolution in neuroscience. In the previous decade, increasing numbers of brain scientists employed approaches using oscillatory components of event-related potentials, EEG, and MEG. When we recall the important remark made by Mountcastle (1998) about a possible turnaround in the analysis of the brain field potentials and their importance, we can imagine that this field will grow remarkably and the new millennium will be witness to a new era in brain research in which EEG oscillations together with complementary brain imaging techniques will become focal points for new discoveries. An important step in this revolution is the fact that the neuroscience experimenters started to consider the brain as an integrative system and no longer limited their analyses to results from a special structure or application of a single paradigm only. This change may have arisen from the availability of powerful computers and algorithms in the analysis of EEG events and new systems theory tools. Another important achievement is the parallel use of whole-cortex MEG and fMRI. The experimenter who chooses these algorithms to explore higher level nervous activities should essentially accept the EEG components as the most important functional building blocks of the brain both at the cellular and neural assemblies levels. We have probably witnessed more than a paradigm change in neuroscience as predicted by Mountcastle (1992). Neuroscientists are now able to attack a core problem: understanding brain functioning by means of its natural frequencies or EEG oscillations. The sudden paradigm change indicated by Mountcastle (1992) is interwoven with the discoveries of the activities of neural populations via new brain imaging techniques. These new techniques opened the way toward new ideas about renewing or extending the single neuron doctrine. Freeman (1975) designated the theory of using dynamics of neural masses as the new Sherrington doctrine, in which neural populations play a significant functional role. Roy John (1988) described a © 2004 by CRC Press, LLC


hyperneuron consisting of neural populations as a functionally important entity of the brain. Barlow (1972) and Mountcastle (1992 and 1998) proposed modern views of Sherrington’s doctrine. Szenthágothai’s well-known illustration of a 300-µm diameter cortical module (1983) is one of the important examples of neural modules and local nerve circuits — an ensemble that plays a significant functional role. Mountcastle (1976) defined the basic function unit as a minicolumn approximately 30 µm in diameter and containing 100 to 300 neurons. Larger processing units called macrocolumns contained up to several hundred minicolumns. Mountcastle said: Prominent among them is the concept that the brain is a complex of widely and reciprocally interconnected systems and that the dynamic interplay of neural activity within and between these systems is the very essence of brain function. The large entities of the brain are composed of replicated modules. The linked sets of modules of the various brain entities comprise a distributed system (see Chapter 6). Another important trend in describing integrative brain activity and memory involving functionally and selectively distributed neural networks started with the publications of Goldman-Rakic (1997), Fuster (1995), and Mesulam (1990 and 1994). In line with their proposals, Basar ¸ et al. (2004) surveyed results of functional oscillatory activities compiled by more than 100 laboratories in the last 20 years at the cellular, field potential, and EEG−MEG levels. Depending on the regimes or states of the brain, the limbic system, brainstem, thalamus, and cortex are all involved with 2-Hz, 4-Hz, 10-Hz, and 40-Hz firing or with all of them. The new trends imply that the following are involved in integrative brain function: 1. Single neurons and also neural assemblies 2. Spikes of single neurons and also oscillatory activity of neurons and neural assemblies 3. Movements and also cognitive and memory processes Sherrington’s (1948) description of integrative brain activity preceded the empirical results that emerged in the past 20 or 30 years.

−BRAIN THEORY: AN APPROACH THAT INCLUDES 1.9 NEURONS− WHOLE BRAIN ORGANIZATION 1.9.1 TOPOGRAPHY

OF

COGNITION

AND

ELEMENTS

OF

NEURONS–BRAIN THEORY

In Mesulam’s model of cognition, the formation of specific templates belonging to objects and memories occurs as selectively distributed processing with considerable specialization. This functional selectivity exists in anatomically differentiated localizations. The view of Fuster (1995 and 1997) is that memory reflects a distributed property of cortical systems. Accordingly, it can be hypothesized that selectively distributed oscillatory systems may provide a general communication framework for morphology and may be very useful for functional mapping of the brain (Basar ¸ et al., 1999). The neurons−brain theory was based on the concepts mentioned above and empirical findings reviewed by Basar ¸ et al. (1999 and 2001), Bressler (1990), and Gruzelier (1996). It aims to partially replace and extend Sherrington’s neuron doctrine for exploring integrative brain functions manifested with activities of neural populations. New rules to describe brain functions by means of neural populations instead of single neurons were developed: 1. The neuron is the basic signaling element of the brain. Oscillatory activities of the brain (gamma, alpha, beta, theta, and delta) reflect natural frequencies and/or real responses (Basar ¸ et al., 2001). 2. Neural assemblies replace neurons in descriptions of integrative brain functions; this view diverges from Sherrington’s neuron doctrine. © 2004 by CRC Press, LLC


3. The EEG is a quasi-deterministic or chaotic signal; it is not always noise. Electroencephalogram oscillatory activities govern most general transfer functions in the brain. 4. Selectively distributed oscillatory neural populations are activated upon sensory stimulation or event-related tasks by manifestation of: a. Enhancements or resonances b. Delay of oscillations c. Blocking or desynchronization of oscillations d. Prolongation of oscillations e. Increases or decreases of coherences f. Increases or decreases of entropies 5. According to published data, parallel processing also functions selectively. Oscillatory systems showed varied degrees of coherence (Basar ¸ , 1980; Kocsis et al., 2001; Schürmann et al., 2000). 6. Types of neurons do not play a major role for frequency tuning of oscillatory networks since morphological different neural networks are excitable and communicate with the frequency codes of EEG oscillations. 7. Functions in the brain are manifested by varying degrees of superpositions of oscillations in EEG frequency ranges. Accordingly, neuron assemblies do not obey the all-or-none rule of the single neuron doctrine. 8. Although the existence of feature detectors has been demonstrated, integrative brain functioning needs the synergy of selectively distributed and selectively coherent neural populations. The role of the feature detectors is well described (Sokolov, 2001; Chapter 7, this volume). 9. Integrative functions in the brain are manifested by varied degrees of coherences. 10. A strong inverse relation exists between prestimulus oscillations and brain responses; spontaneous activity of the brain is effective as a control parameter. Do the super-synergy and superbinding concepts allow us to build a bridge to interpret manifestations of integrative brain functions? As a corollary to the brain−neuron theory, we recently (Basar ¸ et al., 2001 and 2003) we introduced the concept of supersynergy in brain electrical activity as an ensemble of at least six processes that act in synergy upon sensory−cognitive input. According to our hypothesis based on results of human and animal experiments, the electrical manifestations of integrative brain functions are shaped by: 1. The superposition of oscillations including the alpha, beta, gamma, theta, and delta bands. 2. Activation of two or more selectively distributed oscillations in gamma, alpha, theta, and delta bands upon exogenous or endogenous input; these activities are manifested with parameters such as enhancement, delay, blocking (desynchronization), and prolongation. 3. Temporal and spatial changes of entropy in the brain. 4. Temporal coherence between cells in cortical columns for simple binding. 5. Varying degrees of spatial coherence that occur as parallel processing over long distances. 6. Inverse relations of EEGs and EROs; prestimulus EEGs serve as control parameters. Recent experiments performed with the faces of a known grandmother and an anonymous person support the concept of EEG superbinding. We return to this concept in Chapter 7 and Chapter 8. This chapter mentions the neurons−brain theory and the supersynergy concept because they are almost prerequisites for a global reading of this book. However, the evolution of this framework can be better understood after carefully reading the experiments discussed in Chapters 3 through 6. Accordingly the new framework will be discussed analytically in Chapter 7. Chapter 11 aims to combine this chain of ideas into a theory. © 2004 by CRC Press, LLC


Suggestions and ideas by Fessard, Griffith, and Rosen; the principles of Hebb, Lashley, and Hayek; and the theory of Haken provided important tracks that converge in the supersynergy and superbinding concepts; the method developed by Hans Berger was the most important tool. Certainly, the cooperation between neural populations cannot elucidate phenomena at the synaptic level as described by Hebb. However the concepts presented in this book may bridge or possibly unify all previous views as a new proposal for scientists working at macroscopic levels of brain dynamics.

1.10 SIGNIFICANCE OF EEG BRAIN DYNAMICS IN MEMORY STATES AND INTEGRATIVE BRAIN FUNCTIONS The ERP is a compound neuroelectric signal that is rich in functional information (Bullock, 1993) and related to a large spectrum of activities ranging from single percepts to complicated memory processes. In the analysis of integrative brain functions, one must consider not only one specific ERP in a given brain structure, but must also take into account the interrelations of distributed ERPs due to strong parallel processing in the whole brain. Accordingly, it is necessary to analyze the entire brain in order to understand a specific function manifested by neurolectric activity of a given structure. For example, when we consider or analyze cognitive processes, the most marked ERPs are recorded usually in frontoparietal areas or in various association cortices. However, it is necessary to take into account recordings from other areas as well, e.g., from sensory cortices that may indicate parallel processing (Basar ¸ and Schürmann, 1994; Basar ¸ , 1998a and b). Remembering and memory are manifestations of various and multiple functional processes, depending on the complexity of the input to the CNS. The electrical response to a simple light flash is based on simple memory processes at the lowest hierarchical order. When we talk about a memory process — a short- or long-term one — we perceive a sensory input that is matched with information already stored in neural tissue. If a simple light evokes alpha and gamma responses, it is almost obligatory to assume that elementary oscillatory responses are also manifestations of several memory processes at different hierarchical levels. The topology of the memories, depending on the modality of the input, must be different (see examples provided by Basar ¸ and Schurmann (1994) and Basar ¸ (1998a,b) related to cross-modality experiments and measurements of cortical and subcortical structures). Such studies have rarely been performed and the results and their interpretations must be considered preliminary. Accordingly, multiple distributed memories cannot be treated in detail and classifications of all levels of distributed memories cannot be yet provided. In performing many complex tasks, it is necessary to retain information in temporary storage until it is needed to complete the task. The system used for this is called working memory (Baddeley 1996). Working memory is the temporary ad hoc activation of an extensive network of short-or long-term perceptual components that are, like any other perceptual memories, retrievable and expandable by new stimuli or experiences. Fuster states that working memory has the same cortical substrate as the kind of short-term memory traditionally considered the gateway to long-term memory. According to the functional descriptions above, a simple or complex light stimulation or a light stimulation or a light stimulation involving some task or event should evoke oscillatory responses with different time hierarchies. Our view is that functional or oscillatory network modules are distributed not only in the cortex but also throughout other parts of the brain. Several types of analyses are crucial to functional interpretation of ERPs: 1. The analysis of the stimulus. What can a stimulus evoke in the brain? It can evoke simple sensory percepts, complex sensory percepts, bimodal percepts, memory related functions, etc. © 2004 by CRC Press, LLC


2. The analysis of ERPs should be performed in related or unrelated function-dependent areas. For example, if a complex semantic event or memory-demanding task is presented as stimulation to the brain, frontal and/or parietal recordings are considered to carry the most important information. In this case, it is very important to analyze ERPs recorded in the occipital cortex (an area thought to be less involved in high level cognitive processing). This shows what is missing in occipital ERPs in comparison to association areas or what is recorded additionally. These steps are analogous to the fMRI analysis mentioned earlier in this chapter. 3. Component analysis by means of EROs provides an advantage over conventional ERP analysis as, for example, the results of cross-modality measurements demonstrate. In occipital areas, auditory stimulation does not evoke 10-Hz responses, although ERPs are measured upon visual stimulation. This demonstrates the dependence of 10-Hz responses on visual perception. Accordingly, the spatial resolution of ERPs is highly increased. 4. Studies with single-cell recordings and fMRI indicate that memory networks are distributed. Although the ERPs and EROs do not provide the excellent spatial resolution of fMRI or the exact one-to-one locations of single-cell recordings, they have several outstanding advantages in memory research. 5. When compared with fMRI, the time resolution levels of ERPs and EROs are excellent, since it is possible to measure function-related neuroelectric changes within a few milliseconds. 6. In ERP studies, the neuroelectric or neuromagnetic recordings can be obtained in humans; this is almost impossible with single-cell recordings. Moreover, it is possible to apply simultaneous measurements with several recording electrodes in distant locations. This allows dynamic comparisons of various structures of the human cortex and diverse subcortical structures in animal brains. For example, immediate comparison of frontal theta or alpha activity with occipital activity is possible. 7. As Fuster (1997) noted, the brain has as many memory types as the number of percepts. The application of event-related oscillations for the analysis of working memory and for simultaneous analysis of perceptual memory is very useful as a complement to fMRI and single-cell studies. These remarks clearly show that the analysis of EROs fills an important gap in the analysis of selectively distributed percepts and memories. 8. Dynamic changes in the attention−perception−learning−remembering (APLR) alliance can be studied only with strategies involving EROs. This chapter describes concepts and frameworks developed since the 1920s, whereas Chapter 3, Chapter 4, Chapter 6, and Chapter 8 provide empirical data obtained by application of these concepts. The new data, of course, led to new theories and principles. The paradigm change in cognitive sciences emphasizes analyses of macrodynamics instead of microdynamics. Accordingly, the way is now open to establish a new conceptual framework or theory of neural populations to extend or replace Sherrington’s neuron doctrine and to include memory in the framework developed.


TF1663_C02.fm Page 21 Wednesday, May 12, 2004 7:51 PM

2 Concepts and Theories 2.1 MEMORY MACHINERIES 2.1.1 DYNAMIC MEMORY

AND

APLR ALLIANCE

This chapter will describe relevant research and categorizations that explain the theoretical frameworks and models by Baddeley, Damasio, Squire, Tulving, and especially Fuster. Definitions, classifications, and categorizations of memory are described in several books and reviews often with divergent opinions that usually reflect the experiences of the authors. In the present book, a model of dynamic memory is proposed. The dynamic changes arising during memory processes will be demonstrated by evaluation of EEG oscillations. The experimental onset of dynamic changes by means of reciprocal activation will be described in Chapters 3, 6, 7, and 8 and may lead to new perspectives for future research. Chapter 9 discusses the essential model derived from memory studies involving EEG oscillations. With this model, we aim to associate the evolving memory with reciprocal activations of the processes of attention, perception, learning and remembering that we call the APLR alliance.

2.1.2 STEPS

OF

MEMORY PROCESSING

According to Tranel and Damasio (1995), “The process of forming memory involves three basic steps: (1) acquisition, (2) consolidation, and (3) storage.” • • •

Acquisition is the process of bringing knowledge into the brain and into a first-stage memory buffer via sensory organs and primary sensory cortices. Consolidation is the process of rehearsing knowledge and building a robust representation of it in the brain. Storage is the creation of a relatively stable memory trace or record of knowledge in the brain.

In learning to recognize a new face, for example, an individual would consolidate information concerning its visual pattern and create a relatively permanent record of the pattern that would then be connected to other pertinent knowledge (the person’s name, the situation in which the individual met the new person, etc. (Tranel and Damasio, 1995)

2.1.3 ENCODING

AND

RETRIEVAL

Any system for storing information, whether biological or artificial, must be able to (1) encode or register information, (2) store it, preferably without much loss, and (3) subsequently access or retrieve that information. Baddeley noted that because the three stages are closely linked, it is difficult to isolate any phenomenon as exclusively occurring at a single stage. Nevertheless, this division into processing stages continues to be useful in helping explain the processing or operation of memory systems. Chapter 9 will show that the transitions between processing stages are dynamic processes. As experiments in Chapter 3 indicate, the attention, perception, learning, and remembering processes



are interwoven. Interplay among them is perpetual and they are not separable on a time axis (see also Baddeley, 1996). Human brains use dynamic records rather than static, immutable memory traces. For example, the record of a face an individual recognizes is a set of neuron circuit changes that can be reactivated rather than a “picture” stored somewhere in the brain. Dynamic records can be modified and in this way reflect evolving experience (Damasio, 1989 and 1994). Encoding is the initial processing of information to be learned or memorized. Immediate memory for arbitrary sequences of verbal material, such as sequences used in a digit span test, typically relies on encoding based on phonological or sound characteristics of the material. Retrieval is the process of reactivating knowledge in a way that will allow it to become an image in consciousness (as in recall and recognition) or translated into a motor output (movement of a limb, activation of vocal apparatus, autonomic activity). See Tranel and Damasio, 1995. New learning models using parallel distributed processing (PDP) or connection architectures that are assumed to closely simulate the parallel processing of the neural networks of the brain have once again raised the issues of interference effects and how the brain deals with interference (Ratcliffe, 1990; Rumelhart and McClelland, 1986). Parallel processing of distributed oscillatory systems with multiple frequency windows was first described by Basar ¸ in the 1980s by means of ERP experiments with animals. The experiments were later extended to humans (Basar ¸ , 1980 and 1999).

2.2 FRACTIONATION OF MEMORY Atkinson and Shiffrin (1968) developed the well-known scheme of fractionation of memory illustrated in Figure 2.1. According to their model, information from the environment enters a series of brief sensory registers that then pass on information to a short-term store. This temporary storage system plays a crucial role. Without it, information cannot be transferred into or from the third final component, the long-term store. Long-term storage is assumed to occur when information is transferred as described, with the probability that the transfer is a direct function of the duration of time an item resides in the short-term store.

2.2.1 LONG-TERM MEMORY

VERSUS

SHORT-TERM MEMORY

The classic explanation of encoding reflects the current view originally stated by Shiffrin and Geisler (1973): “The process of encoding is essentially one of recognition; the appropriate image or feature is contacted in long-term memory (LTM) and then placed (i.e., copied) in short-term memory (STM).” Complex cognitive processes such as speaking and thinking may also be described in terms of a close interaction between the working memory and LTM systems. The basic difference of the Shiffrin and Geisler model is that a sensory code is lacking and a code generated in STM, for example, during speaking, plans what the individual will say. The codes generated in STM trigger search processes in LTM to retrieve the relevant knowledge about the appropriate semantic, syntactic, and articulatory information. This idea is similar to Baddeley's (1986, 1992, 1997) concept of working memory comprising an attentional controller and central executive and subsidiary slave systems (Klimesch, 1999). Baddeley pointed out that despite its advantages, the described model rapidly encountered problems. The assumption that merely holding an item in a short-term system would guarantee learning proved difficult to sustain (Craik and Watkins, 1973). The addition of levels of processing seemed to give a better account of learning than the modal model. Even more problematic was the evidence for normal learning in patients with short-term store deficits (Shallice and Warrington, 1970). Such patients appeared to have remarkably few problems in coping with everyday life and this was interpreted as an argument against the importance of short-term memory as a crucial © 2004 by CRC Press, LLC


Environmental input

Sensory registers Visual Auditory Haptic

Short-term storage (STS) temporary working memory Control processes: Rehearsal Decision Coding Retrieval strategies

Response output

Long-term storage (LTS) permanent memory storage

FIGURE 2.1 Structure of memory. (Modified from Atkinson, R.C. and Shiffrin, R.M. (1968), in The Psychology of Learning and Motivation: Advances in Research and Theory, Spence, K.W., Ed., Academic Press, New York, p. 195.)

control center for cognition. Baddeley and Hitch (1974) therefore suggested abandoning the assumption of a unitary short-term store and suggested a multicomponent working memory system.

2.2.2 WORKING MEMORY Baddeley and Hitch proposed a division of the working memory model into at least three subsystems, as illustrated in Figure 2.2. An important part of the system is an attentional controller or central executive that forms an interface between long-term memory and two or possibly more slave systems. These subsystems constitute the capacity for the temporary storage of information with an active set of control processes that allow information to be registered intentionally and maintained within the subsystem. The visuo-spatial scratchpad or sketchpad specializes in maintaining visuo-spatial information; verbal information is held by the phonological or articulatory loop. The central executive is assumed to be responsible for the selection and operation of strategies and for maintaining and switching attention as the needs arise. It is assumed to be associated with the functional activities of the frontal lobes and thus sensitive to frontal atrophy and lesions (Baddeley, 1986; Baddeley and Wilson, 1988). According to Tranel and Damasio (1995), working memory is a transient type of memory processing on a time scale of seconds during which an individual can maintain “online” the relevant stimuli, rules, and mental representations required to execute a particular task (Baddeley, 1992; © 2004 by CRC Press, LLC


Phonological loop

Central executive

Visuo-spatial sketch pad

FIGURE 2.2 Working memory model. (Modified from Baddeley, A. and Hitch, G. (1974), in The Psychology of Learning and Motivation, Bower, G.A., Ed., Academic Press, New York, p. 47.)

Goldman-Rakic, 1987; Chapter 1, this volume). Working memory is used to bridge temporal gaps, that is, to hold representations in a mental workspace long enough to formulate appropriate responses to stimulus configurations or contingencies for which some or even all the basic ingredients no longer exist in perceptual space (Fuster, 1989; Goldman-Rakic, 1987). The concept of working memory in a way overlaps with assumptions about STM. Both are considered relatively transient and are thought to have limited capacities. Fuster noted (1995, 1997, and 2000) that working memory, also known as operant memory, is an operant concept of active memory and postulates (1995) that active memory is a state rather than a system of memory. Single neuron recordings in monkeys trained to perform working memory tasks identified components of a working memory circuit in the prefrontal cortex. In these studies, the neuronal processes related to task performance can be dissociated on a scale of milliseconds to seconds. During performance of a working memory task, as the stimulus is sequentially registered, stored for seconds, and then translated into a motor response, specific neural populations respond in characteristic ways. One class of prefrontal neurons responds to a visual stimulus as long as the stimulus is in view. In contrast, other prefrontal neurons are activated at the onset of the stimulus and remain active during the time the monkey must remember the location or features of an object (Fuster, 1995; Goldman-Rakic, 1988 and 1997).

2.3 DISTINCTION BETWEEN IMPLICIT AND EXPLICIT MEMORY STATES Another important categorization of memories (or memory states) is the distinction between implicit and explicit types. Squire (1992) distinguishes between declarative and nondeclarative memory — terms that more or less map onto the explicit and implicit terms (Figure 2.3). An early distinction was made between procedural and declarative learning. Procedural learning represented the acquisition of skills or “learning how.” Declarative learning involved the acquisition of facts or “learning that ….” (Squire, 1992; Baddeley, 1985). While many preserved learning capacities may be regarded as skills, regarding conditioning or stem completion as genuinely procedural seems to be stretching the term. According to Baddeley (1995), it has become increasingly clear that memory comprises not a single system, but consists of an alliance of interrelated subsystems. Empirical evidence for the distinction between long- and short-term memories began to emerge in the 1960s. Baddeley’s view (1995) bears repeating: One source came from the previously described evidence that immediate memory for verbal material appears to rely on phonological coding, while LTM appears to be semantically based. © 2004 by CRC Press, LLC


MEMORY

NONDECLARATIVE (IMPLICIT)

DECLARATIVE (EXPLICIT)

Facts (semantic)

Events (episodic)

Skills and habits

Priming

Nonassociative learning

Simple classical conditioning

FIGURE 2.3 Components of long-term memory. (Modified from Squire, L.R. (1992), J. Cognitive Neurosci., 4, 232.)

A second source of evidence came from the observation that certain tasks appear to have two components. If a subject is presented with a list of words for immediate free recall, there will typically be extremely good recall of the last few items presented (Glanzer and Cunitz, 1966). One interpretation of this result is to suggest that the last few items are held in a labile short-term store, whereas earlier items reside in LTM.

2.4 NONDECLARATIVE MEMORY Baddeley (1995) proposes a cluster of learning systems that have in common the fact that they are independent of episodic memory. This means that they are capable of accumulating information, but not of pulling out and identifying specific episodes. The episodic memory system has a remarkable capacity to associate previously unrelated events in a single trial. It can associate an event with a context, and hence locate it in time and place. In contrast, nondeclarative systems are specialized for accumulating information from the world, but incapable of keeping separate the individual episodes (Baddeley, 1994b). A number of different kinds of nondeclarative phenomena have been identified.

2.4.1 PHYLETIC MEMORY Perceiving is the classification of objects by activation of the associative nets that represent them in memory. It is reasonable to assume, as Hayek (1952) did, that to a large extent, memory and perception share the same cortical networks, neurons, and connections. To understand the formation and topography of memory, it is useful to think that the reaction ability of the primary sensory and motor areas of the cortex is called phyletic memory or memory of the species. The primary sensory and motor cortices may be considered a fund of memories that the species acquired during evolution. We use the memory term because, like personal memory, phyletic memory constitutes information that has been acquired and stored and can be retrieved (recalled) by sensory stimuli or the need to act.

2.4.2 PERCEPTUAL MEMORY Perceptual memory is acquired through the senses. It comprises all that is commonly understood as personal memory and knowledge, i.e., representations of events, objects, persons, animals, facts, names, and concepts. From a hierarchical view, memories of elementary sensations are at the



bottom; at the top, abstract concepts originally acquired by sensory experience have become independent via cognitive operations.

2.4.3 PROCEDURAL LEARNING Procedural memory is the representation of a series of actions or perceptual processing functions that occur unconsciously, and repetitions typically result in increased speed or accuracy. This relates to the acquisition of skills, whether perceptual motor skills such as riding a bicycle or driving a car or cognitive skills such as reading or problem solving. Skills clearly represent an important area of learning and serve as archetypal examples of procedural learning — learning how rather than learning that (Eichenbaum, 2000). See Figure 9.7 for a categorization of perceptual memory in the hierarchy of memory states. Skills are continuous (each component of the skill serves as a cue to the next as in cycling or steering a car) and discontinuous (a series of discrete stimulusresponse links are involved as in typing).

2.4.4 PRIMING Priming is the facilitation of recognition, reproduction, or bases in the selection of recently perceived stimuli (Eichenbaum, 2000). If a word has been presented, subjects are subsequently more likely to identify a noisy representation of the word or produce the word when faced with its stem or a fragment. As Tulving and Schacter (1990) pointed out, priming effects occur across a wide range of modalities and are typically dependent upon the repetition of the physical characteristics of the original stimulus; priming is typically much less sensitive to semantic or conceptual aspects of the primed material. It is assumed that priming is some form of neural residue that either enhances its subsequent speed of use (positive priming) or has an inhibitory effect (negative priming).

2.4.5 EVOLVING MEMORY The processes of evolving memory constitute the most important core of the presented empirical knowledge and serve as an important framework of this book (see Chapter 9). The process of formation of memory, which we denote also as evolving memory, probably constitutes the most important process during the transition from one memory state to another (see Chapter 8 for discussion of the transition between semantic and episodic memories).

2.5 DECLARATIVE MEMORY Declarative (or explicit) memory is the recall of events and facts; it is commonly known as personal memory. It constitutes two subsystems originally defined by Tulving (1972) as episodic and semantic memories.

2.5.1 EPISODIC MEMORY Episodic memory is a system that collects temporarily and spatially encoded events in a subject’s life, i.e., recalls of particular experiences or episodes. Remembering what was received as a birthday present last year and what was eaten for breakfast are examples of episodic memories. These types of memories are strongly influenced by the degrees of attention and organization that reflect the importance of the events in order to set up memory structures that are accessible to retrieval.

2.5.2 SEMANTIC MEMORY Semantic knowledge is an organization of factual information independent of specific episodes during which that information was acquired. Semantic memory, on the other hand, is knowledge © 2004 by CRC Press, LLC


of the world. Knowing the chemical formula for salt, the name of the French capital, and the number of inches in a foot are all examples of semantic memory (Baddeley, 1995). They are facts that through single or repeated mentions have come to categorize concepts, abstractions, and evidences of reality, although the subject may not necessarily remember when and where he or she acquired the information. Education can be regarded as the gradual growing and/or enriching of semantic memory, starting with perceptual knowledge of the physical world and progressing to language use, knowledge of society, and acquisition of detailed specialized information acquired via a trade or profession.

2.5.3 RELATIONSHIP

OF

EPISODIC

AND

SEMANTIC MEMORIES

Tulving's initial conceptualization (1972) proposed that semantic and episodic memories are based on separate memory systems but evidence indicates that they reflect the same system operating under different circumstances. Semantic memory stores information that may have originated from many separate experiences that are no longer individually retrievable (Fuster, 1995). The view of Baddeley is that semantic memory consists of the accumulation of many episodes. According to this author, a useful analogy is to think of a series of individual episodes piled on one another; episodic memory represents the capacity to pull one episode from the pile, whereas semantic memory reflects the capacity to look at the pile from above and draw out features that are common to many of the constituent episodes. In Chapter 8, we will analyze dynamic changes of theta responses after the transition of semantic knowledge to episodic knowledge. From the view of electrophysiological records, semantic and episodic memory systems probably share similar oscillatory activated neural populations or both systems are neurally unseparable. Section 8.5.4 is an attempt to describe the transition between semantic and episodic memory research and the electrophysiological manifestations of this crucial transition.

2.6 NEUROBIOLOGY OF MEMORY 2.6.1 MOLECULAR

AND

CELLULAR BASES

OF

MEMORY

2.6.1.1 Hebb’s Proposal Does some type of modification of neurons or of connections between neurons take place as a result of learning? For example, when we learn to associate two stimuli (e.g., an unconditioned stimulus and conditioned stimulus, as in classical conditioning), what happens in the brain to support this process (Tranel and Damasio, 1995)? Donald Hebb (1949) proposed that the coactivation of connected cells would result in a modification of weights so that when a presynaptic cell fires, the probability of postsynaptic cell firing is increased. Hebb stated: When an axon of a cell A is near enough to excite cell B or repeatedly or persistently takes part in firing it, some growth or metabolic change takes place in both cells such that A’s efficiency as one of the cells firing B is increased (p. 62). This learning principle did not describe what was meant by growth or metabolic change. However, this principle served as a useful pioneering idea, and has become one of the widely cited concepts for neurobiological investigations of learning and memory (see Chapter 1, this volume). Another important step forward arose from the work of Bliss and Lomo (1973). When the excitability of a postsynaptic cell was increased for hours, or even days or weeks, by stimulation with a high-frequency volley of pulses known as a tetanus (specifically, when the primary afferents of dentate granule cells in the hippocampus were exposed to a tetanic stimulus), the depolarization potential of the postsynaptic cell was enhanced, and this potentiation lasted for a long period. The effect is known as long-term potentiation (LTP) and it has become a very important model in modern conceptualizations of the cellular bases of learning and memory. Morris et al. (1982) © 2004 by CRC Press, LLC


demonstrated that the retardation of the behavioral learning curve in the performance of the water maze task was directly congruent with the extent to which LTP was blocked in the hippocampus. In other words, less LTP correlated with poorer learning and more LTP correlated with better learning. These results provided strong behavioral evidence supporting the role of LTP in the cellular basis of learning. 2.6.1.2 Kandel’s Fundamental Results Important advances in the understanding of learning and memory at the molecular level have come from the work of Eric Kandel and his colleagues (Kandel and Schwartz, 1982; Hawkins et al., 1983). Much of this work has been done with the Aplysia californica marine mollusk, which has a simple nervous system composed of approximately 10,000 neurons. The neurons are unusually large and easily identifiable, making Aplysia far more convenient for cellular level studies than vertebrates with far more complex nervous systems. Research by Kandel and colleagues provided the first direct evidence that alterations of synaptic efficacy play a causal role in learning. They discovered that behavioral habituation of the gill and siphon withdrawal reflex, a staple behavioral activity of Aplysia, was mediated by a reduction in transmitter release at a defined synaptic locus (Pinsker et al., 1970; Castellucci and Kandel, 1974). These results supported Hebb’s principle. Bailey and Chen (1983) later showed that habituation was accompanied by alterations in the morphologies of electrophysiologically identified synapses. These investigations provided direct evidence for forms of synaptic plasticity that may provide cellular and molecular bases for at least some forms of learning and memory. 2.6.1.3 EEG Oscillations in Aplysia and Helix pomatia Isolated invertebrate ganglia also show types of EEG oscillations in delta, theta, alpha, and gamma frequency windows. These results published by Schütt et al. (1992 and 1999) and Basar ¸ (1999) will be partly illustrated in Section 6.3.2. Do EEG oscillations manifest universal functional codes during the evolution of species? Does the phyletic memory or memory of species process EEG codes similar to those observed in the human brain? These questions cannot be answered in this book, but it is worth mentioning that correlations of oscillations with changes at the molecular level may provide essential material for tracking the molecular basis through possible electrical associations. The model by Kandel (1982) and his associates will play an important role in future memory research.

2.7 NEW SCHEME BASED ON EEG STUDIES FOR CATEGORIZATION OF MEMORY LEVELS This chapter discussing established theories and definitions briefly describes the new categorizations that emerged from experiments related to EROs. The model is new and must undergo a maturation process. However, based on experimental results cited in Chapters 3, 5, 6, and 8 describing the relevance of memory function-related oscillations, a new model is needed. Dynamic and reciprocal activations of integrative brain and memory functions can be more adequately described with oscillations because of their dynamic nature.

2.7.1 PHYSIOLOGICAL (FUNDAMENTAL-FUNCTIONAL) MEMORY We would like to introduce physiological memory as an alliance or collection of phyletic, perceptual, and procedural memories partially built in LTM storage or persistent memory. Throughout our lives, we acquire new physiological and cognitive facts and strategies. The ability to perceive red, for example, already exists in the phyletic memory. Knowledge about a flag, face, or the uniform © 2004 by CRC Press, LLC


of an enemy is acquired during life, and such images are stored easily in the LTM system, similar to skills like diving or driving. Accordingly, phyletic memory may be altered, extended, or formed into a physiological memory vital for functioning. Physiological functions that are vital for survival are genetically coded to a degree but are also partially acquired throughout life. Chapter 3 and Chapter 6 will show that all CNS functions are accompanied by or associated with some type of memory or components of whole brain memory. As noted earlier, physiological and cognitive functions are inseparable (Basar ¸ , 1999). According to our EEG-related physiology and cognition-related functions, we go a step further than Hayek (1952) and adapt the scope of Fuster (1995). We hypothesize that all functions of the CNS and memory are inseparable. Since the physiology expression is almost synonymous with function and acting, we also use the functional memory phrase. Although we have here a type of redundancy, this idea constitutes the leitmotif of this book. Accordingly, in Chapter 6 and Chapter 9, we explain the concept and functioning of physiological memory that serves as part of every memory action. A main point is that physiological memory also includes phyletic memory and (partially) procedural memory. It combines a collection of inborn (built-in) memories and newly developed and stabilized memory traces accumulated via everyday brain functioning. To see something, even the simplest light signal, is already a memory process related to a fundamental inborn or built-in retrieval process. A baby perceives a light and shows reflex responses to the light before going through learning processes.

2.7.2 TRANSITION AND COMBINATION OF MEMORY STAGES (EVOLVING MEMORY) According to Damasio (1997), memory depends on concerted work by several brain systems across many levels of neural organization. Memory is a constant work in progress and EEG oscillations evolve parallel to this constant work. Recording of function-related or memory-related EEG or MEG oscillations serves as an excellent measuring tool for transitions and other dynamic processes in human and animal brains. This is the only method that can analyze distributed dynamic processes during long experiments involving behaving and conscious brains. Sections 8.5.2 through 8.5.4 discuss an important example.

2.8 LONGER-ACTING MEMORY AND TRANSITION TO PERSISTENT MEMORY IN WHOLE BRAIN According to Section 9.5.5, event-related changes in EROs lead to substantial changes in the electrical manifestations of evolving memory. It will be shown that new learned material is transferred to LTM for longer time intervals in comparison to working memory. Astonishingly, the durations of time spent in working memory and in long-term memory storage are not defined clearly in the literature. In our opinion, longer-acting memory is a better description than long-term memory because it distinguishes between working memory and persistent memory. As a new proposition in memory categorization (Figures 9.7 and 9.12), fresh memory traces acquired in everyday experiences are temporarily stored in longer-acting memory, before reaching the persistent memory level. According to the description of memory levels introduced in Chapter 9, persistent memory combines built-in memory with physiological memory (an ensemble of submemories such as echoic memory, iconic memory, olfactory memory, etc.) and stabilized parts of longer-acting memory acquired throughout life (see Figure 9.7 and Figure 9.12). The answer to the question of how new information acquired during processes of memory evolution or memory building (manifested by multiple oscillations and enhanced coherence in the whole brain) is transferred and stored in persistent memory surpasses the scope of this book. However, it is important to note that the networks of persistent memory operate with the same oscillatory dynamics of evolving memory, i.e., they use the same basic oscillatory codes (alpha, beta, etc.). This indicates that frequency codes may be transferred to persistent memory or may play an essential role during the transition. © 2004 by CRC Press, LLC


Part II Experiments and Their Interpretation



Dynamic and Evolving 3 Shaping Memories by Reciprocal Activation of Attention, Perception, Learning, and Remembering “Attention” is the taking possession of by the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or train of thought. It implies withdrawal from some things in order to deal effectively with others. William James (1890)

3.1 ESSENTIAL EXPERIMENTS INVOLVING DYNAMIC MEMORY AND TOP-DOWN ACTIVITY Chapter 2 introduced the combination of attention, perception, learning, and remembering we called the APLR alliance. These processes are reciprocally active and overlapping constructs that are difficult to separate. Each term has a variety of meanings and implications. Attention, like memory, is a function of a system. Fuster (1995) noted that it made no more sense to speak of a special neural system for attention than it did to speak of one memory. Attention and memory are intimately interrelated. What we remember and how long we remember it depend on the selective function known as attention, as do the dynamics of active memory. Results of EEG oscillation studies clearly demonstrate that all integrative functions and memory are interrelated (Chapter 6). As we explained in discussing the tentative scheme related to levels of memory states, reflexes are also inborn memory components (Figure 9.7). This chapter will present experiments related to the dynamics of electroencephalogram (EEG) oscillations. Both prestimulus EEG and interstimulus activity EEG merit considerable attention because EEG activity prior to a sensory or cognitive input greatly influences brain responsiveness (Barry et al., 2003; Chapter 5, this volume). In order to understand the evolution of memory components in ERPs we must analyze the dynamic changes of prestimulus and poststimulus oscillations during cognitive paradigms that use sensory stimulations. In Chapter 1, we mentioned the importance of the hypotheses of Hebb (1949) and Edelman (1977) related to dynamic properties of the brain. Hebb assumed that brain morphology should be considerably changed after stimulations (excitations of cells creating new activated states in neural populations). Edelman mentioned the possibility of reentry, i.e., meaning that stimulation of the brain influences its responsiveness. Good research models based on these hypotheses are provided by combining prestimulus and poststimulus EEG segments; both are considered active or activated brain states. Combined EEG–event-related potential (ERP) experiments currently provide the only possibilities of studying dynamic changes for durations shorter than 1 s. The evidence of a minimal activation period of



200 to 500 ms for awareness of a near-threshold stimulus (Libet, 1991) makes it possible to describe cognitive interactions occurring in this time interval only with EEG oscillations in behaving subjects. Top-down brain signals convey knowledge derived by prior experience rather than sensory stimulation. The analysis of changes in cognitive activation of the human brain, which is influenced by prior experience, i.e., processing of top-down signals from the brain, provides the core material of this chapter.

3.2 DYNAMIC MEMORY MANIFESTED BY INDUCED ALPHA ACTIVITY 3.2.1 SELECTIVE ATTENTION Selective attention, or simply attention, is a construct that has a rather broad but circumscribed set of meanings; selective attention is clearly distinct from nonselective central nervous system (CNS) processes such as arousal or alertness. Attentional processes are CNS functions that enable perceptual or motor responses to be made selectively to one stimulus category or dimension in preference to others. Irrelevant stimuli that are not required are partially or completely rejected from perceptual experience, entry into long-term memory (LTM), and control over behavior. Since attention refers to selective aspects of sensory processing, it follows that all experimental demonstrations of attention must measure the responsiveness of the organism to more than one category of stimulus. The differential response to attended versus unattended stimuli provides the operational basis for this construct. If an animal is tested with only one stimulus, one cannot be sure whether improvements in processing are the results of paying attention selectively or of an increased level of arousal or alertness that by definition influences a broader spectrum of sensory inputs or response propensities nonselectively (Hillyard and Picton, 1979).

3.2.2 APLR ALLIANCE Attention, perception, learning, and memory are interwoven processes. During experiments with working memory paradigms, short-term memory (STM) presents a continuously evolving interaction and reciprocal activation of each of these functions. During such experiments, memory processes are altered, enhanced, and evolved. Accordingly, dynamic experiments (single-trial oscillations) provide excellent opportunities for elucidating the processes of evolving memory during short time intervals, as we will describe in the following sections.

3.2.3 IMPORTANCE

OF INTERNAL

EVENT-RELATED OSCILLATIONS

Various neural populations in the brain can generate coherent states in which oscillatory 10-Hz activity and theta (3.5 to 7 Hz) activities are recorded. A light flash can elicit a 10-Hz enhancement in the brain if the brain shows disordered activity prior to stimulation. We can evoke a 40-Hz response with sharp onset light, acoustical stimulation, and other techniques. At this point an important question is whether we can find a way to put the brain in such coherent states of EEG activity without external sensory stimulation. Can we find a sensory-cognitive task to produce coherent internal evoked potentials, or better, internal event-related oscillations (EROs)? Petsche (1998) noted that in the past two decades inquiries involving spontaneous brain oscillations revealed new impulses and eventually led to a renaissance of alpha research (Basar ¸ , 1997). Petsche further stated: In 1994 Basar ¸ organized an influential symposium on this topic in which the multiplicity of phenomena in the alpha band was demonstrated. One of the agreements of this symposium was that alpha rhythms are not unitary phenomena but represent a large ensemble of integrative brain functions, the probable roles of which were observed (Basar ¸ et al., 1997a and b). It is for these reasons that research into possible reflections of cognition in the spontaneous EEG © 2004 by CRC Press, LLC


has attracted psychologists. The rationale of planning the 1994 symposium was mainly based on the fact that memory-related alpha oscillations were considered as the most important signaling of integrative brain function. We would also like to mention the important work of Klimesch and his group (1996, 1997a, 1997b, 2000a and b, 2001a and b) on the memory functions of alpha activities (see also Petsche and Etlinger, 1998; Chapter 8, this volume).

3.2.4 COHERENT

AND

ORDERED STATES

OF

EEG

DUE TO

COGNITIVE TASKS

3.2.4.1 Preliminary Experiments The experiments were carried out with 16 healthy volunteers, mostly students 19 to 21 years of age. The EEGs were recorded in vertex, parietal, and occipital locations against references of ear lobes (Cz, P3, P4, and O1 in the 10–20 system). The EEG signals were amplified by using a Schwarzer machine. The subjects sat in a soundproof and echo-free room that was dimly illuminated. For stimulus preparation, evaluation of selective averaging procedure, and digital filtering, a Hewlett Packard 1000F computer was used. The filtering of EEGs and ERPs were carried out. The digital filters did not create any phase shifts. Auditory stimulation of 2000 Hz and 80 dB tones of 800-ms duration were applied at regular intervals of 2600 ms. Every third or fourth tone was omitted. The subjects were asked to predict and mark mentally the times of occurrences of the omitted signals. The EEG 1 s prior to the omitted stimulation was also recorded with the ERP. The light stimulator was a 20-W fluorescent bulb that was electrically triggered. The duration of the light step was also 800 ms. 3.2.4.2 Preliminary Results After the subjects learned and successfully followed the rhythmicity contained in the paradigm, they were usually able to increase their attention and rhythmic prestimulus EEG patterns could be observed. Most subjects reported that at the beginning of an experimental session with repetitive signals, they had difficulty predicting the time of occurrence of the stimulus omission. During the second half of the experiment, they were usually able to predict the time of the omitted signal. Accordingly, in our signal analysis we applied a selective averaging by grouping approximately the first 10 prestimulus sweeps at the beginning of the experiment and the last 10. Figure 3.1 illustrates comparatively the averages of the first 10 and last 10 prestimulus EEG epochs (digitally filtered between 1 and 25 Hz) recorded at the vertex of a subject who reported

–

10.00 µV +

–500

–400

–300

–200

–100

0.0

ms

FIGURE 3.1 Averages of the first (broken line) and last (solid line) 10 prestimulus EEG epochs of the experiment, filtered in the 1- to 25-Hz frequency band. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)



A

Filtered: 7–13 Hz

– 7.5 µV +

–500

B

– 400

–300

–200

–100

0.0 ms

–300 –200 –100 Omission of stimulus

0.0 ms

Filtered: 7–13 Hz

– 7.5 µV + –500

– 400

FIGURE 3.2 (A) Prestimulus EEG sweeps at end of experiment. (B) Prestimulus EEG sweeps at beginning of experiment. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)

that he felt unsure and diffuse at the beginning of the experiment. Toward the end of the experimental session, he was more focused and performed his task better. The first 10 sweeps tended to follow the same rhythmicity, but the average was less regular and showed lower amplitudes. The average of the 10 sweeps at the end of the experiment depicted regular rhythmic behavior with large amplitudes. Rhythms similar to those illustrated in Figure 3.1 were observed with all subjects. The alignment and phase reordering were not the same in all the subjects. The exact times of regularity and phase reordering showed fluctuations from 0 to 700 ms prior to the event. Figure 3.2 shows 10 prestimulus EEG epochs from Subject C at the end (A) and at the beginning (B) of an experiment. Single sweeps were digitally filtered in a frequency range between 7 and 13 Hz according to the rhythmicity revealed in the wideband curve. It is easy to see the repeatable patterns at the end of the experiment in contrast to the lack of such patterns at the beginning. Are they recurrent networks or are the observed changes in alpha activity due to reentry following learning? Although the question cannot be answered with a clear yes, the descriptions of these results greatly favor the Hebb (1949) and Edelmann (1977) hypotheses. We will return to this question in Chapter 9.

3.4.3 GLOBAL TRENDS OF PRETARGET EVENT-RELATED RHYTHMS: SUBJECT VARIABILITY The subjects often decided responses in relation to their own set targets or mentally predicted targets. Some subjects could better predict the omitted signals after the second tones, others after the third tones. Most subjects showed better performance at the end of the experiment, but some were able to recognize the time of the occurrences early in the experiments. The reliability of the results was based on a comparison of phase-ordered EEG states with the subjects’ statements of whether they were able to mark the target signals mentally. It is well known that during long © 2004 by CRC Press, LLC


TABLE 3.1 Comparison of Paradigms Type of Paradigm

Probability of Occurrence of Target

Every fourth or seventh stimulation randomly omitted (most difficult) Every third to fourth stimulation randomly omitted (intermediary, less difficult) Every fourth stimulation omitted (easiest; no randomness)

25% after third tone 50% after second tone 100% after third tone

Source: Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.

recording sessions, EEGs can show highly stationary behavior, random synchronization, and alignments. Our findings showed, however, that the phase-ordered patterns correlated fairly well with the subjects’ reports and, as will be shown below, fluctuations occurred with a number of subjects.

3.2.5 PARADIGMS

WITH INCREASING

OCCURRENCE PROBABILITY

To reduce the possibilities of recording randomly occurring coherent EEG signals, we extended our paradigm. In addition to paradigm 1, randomly omitting the third or fourth stimulation, we extended our measurements to include a more difficult paradigm 2 and thus decrease the probability of the occurrences of the target signals. In paradigm 2, the omitted signal could be changed from the fourth to the seventh randomly (the occurrence of a target signal was 25% when the subject already heard the second tone; see Table 3.1). During the same recording session, the subjects had to perform paradigm 3, which offered an easier way to mark the target signal mentally — every fourth signal was omitted. Paradigm 3 was the easiest because the probability of the target occurrence was 100%, after the subject heard the third signal. The comparison of the experimental results showed that the subjects could emit coherent and phase-ordered pretarget EEG signals almost in all cases with the easiest paradigm. The same subjects did not show the same good coherent and phase-ordered pretarget EEG responses with the most difficult paradigm. This section describes global results with mean value curves from experiments with 16 subjects. We will later present descriptions of experiments on single subjects. Figure 3.3 illustrates the comparison between the easiest and most difficult paradigms. The curves were wideband filtered (between 1 and 100 Hz); results were as follows: Every fourth to seventh signal omitted (the most difficult paradigm) — We observed no regular rhythmicity prior to target (or omitted) stimulation. We observed enhancement of the unfiltered EEG following the omitted stimulation (target signal). Target occurrence was rare; the surprise effect should be greater when the stimulation is omitted. We did not not analyze the variability of the P300 wave among subjects who showed relevant individual fluctuations and variabilities at the beginnings and the ends of the experimental sessions. The mean value curves from 16 subjects showed slight EEG enhancements with peaks around 300 to 400 ms. The latency changes of waves of the P300 family were large (see Galambos and Hillyard, 1981). Every fourth signal omitted (the easiest paradigm) — The pretarget EEG showed a rhythmicity around 9.5 Hz. No EEG enhancement was observed after the omitted stimulation. The mean value EEG signal looked like a continuation with a slightly slower rhythm. 3.2.5.1 3.5- to 8-Hz Range During the most difficult paradigm (every fourth to seventh signal omitted), enhancement following the omitted stimulation was observed in the theta frequency range, whereas no regular rhythmicity was noted in the pretarget EEG. In other words, the reaction was due to surprise (Figure 3.4). (Compare results with enhanced theta responses in the hippocampus cited in Chapter 4.) © 2004 by CRC Press, LLC


FIGURE 3.3 Comparison of the most difficult (random omission of every fourth to seventh stimulation, top) and easiest (omission of every fourth stimulation, bottom) paradigms as mean value curve from 16 subjects. Filter limits: broad band, 1 to 100 Hz. Target signals at time 0. Derivation: vertex. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)

FIGURE 3.4 Comparison of most difficult (random omission of every fourth to seventh stimulation, top) and easiest (omission of every fourth stimulation, bottom) paradigms as mean value curve from 16 subjects. Filter limits: 8 to 13 Hz (alpha frequency range). Target signals at time 0. Derivation: vertex. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)



FIGURE 3.5 Comparison of most difficult (random omission of every fourth to seventh stimulation, top) and easiest (omission of every fourth stimulation, bottom) paradigms as mean value curve from 16 subjects. Filter limits: 3.5 to 8 Hz (theta frequency range). Target signals at time 0. Derivation: vertex. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)

3.2.5.2 8- to 13-Hz Range No coherent and ordered 10-Hz activity was observed preceding the target signal (Figure 3.5) during the most difficult paradigm. However, following the omitted stimulation, a 10-Hz enhancement was observed. During the easiest paradigm (every fourth signal omitted) a coherent 9-Hz rhythmicity preceding the target was observed. Conversely, a blocking of regular 10-Hz activity was observed after the omitted stimulus. 3.2.5.3 40-Hz Range Figure 3.6 shows the mean value curves of 16 subjects in the 40-Hz frequency range. During the easiest paradigm, we observed increased regular rhythmicity of 40-Hz activity just prior to stimulation (50 ms prior to omitted stimulation) in the mean value curve. The 40-Hz activity (or blocking) decreased following the omitted stimulation. During the most difficult paradigm (every third to seventh signal omitted), we noted no increased regular rhythmicity prior to target amd enhancement after the omitted stimulation (approximately 250 ms after stimulation). Results of enhancement or blocking of 40 Hz in this global analysis followed the same trend as the 10-Hz activity. Our analysis is not sufficient to describe whether the 10- and 40-Hz enhancements (or blockings) occurred simultaneously. Readers can compare these results with results in the cat hippocampus discussed in Chapter 4.

3.2.6 EXPERIMENTS

WITH

LIGHT STIMULATION

We will describe experiments during which repetitive light stimuli were used. Although the experiments involved small numbers of subjects, large numbers of experiments were carried out for each



FIGURE 3.6 Comparison of most difficult (random omission of every fourth to seventh stimulation, top) and easiest (omission of every fourth stimulation, bottom) paradigms as mean value curve from 16 subjects. Filter limits, 30 to 50 Hz (40-Hz frequency range). Target signals at time 0. Derivation: vertex. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)

subject. We discuss only studies carried out in the 8- to 13-Hz frequency range; detailed accounts of all the experiments at various frequency ranges are not feasible. We used the same experimental set-up and procedure described earlier in this chapter. The stimulation consisted of light steps of 800-ms duration. The light source was a 20-W fluorescent bulb that could be triggered with a short time constant. The intervals between stimuli were 2600 ms in duration. 3.2.6.1 Experiments with Varied Probabilities of Stimulus Occurrence J.K. is a medical student who quickly learned the goal of the experiments and was very cooperative during the experiments. Figure 3.7 illustrates samples of the filtered resting EEG as a control before an experiment with the cognitive task. There are three plots of the filtered EEG segments, with 10 sweeps in each plot. The three plots present samples from the same recording session. The mean correlation coefficient ( C ) of each ensemble of sweeps in a time range from -500 to 0 ms is also shown. The subject was instructed to be attentive to repetitive light stimuli. Every fourth light stimulation was omitted (the easiest paradigm). He reported at the beginning of the experiment that he could easily mark the target signal; however, after approximately 10 omitted signals or the first 40 sensory stimulations he could not concentrate as well; toward the end of the measurement, he had enormous difficulties in concentrating. Figure 3.8 shows the first 10 filtered sweeps together with the filtered mean values and wideband mean curves (1 to 30 Hz). Clear rhythmicity and good congruency are observed for most sweeps. In the following sessions of the experiments (B and C), the rhythms were less regular and the congruency among sweeps almost disappeared. Also at this stage, 10-Hz EEGs with larger amplitudes were observed in comparison with the resting EEG shown in Figure 3.7. At the beginning, when the subjects reported good performance, the correlation coefficient was high (0.38). It later diminished (0.13 and 0.01) and decreased drastically by the end of the experiment. © 2004 by CRC Press, LLC


FIGURE 3.7 Resting EEG of subject J.K. Top: mean value results on ten sweeps. Bottom: 10 sweeps of EEG segments that were digitally filtered in the frequency range of 8 to 13 Hz. Time 0 was chosen arbitrarily. EEG samples were recorded at the beginning (A), middle (B), and end (C) of session. Correlation coefficients were evaluated from 3 ensembles of 10 sweeps. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)

Figure 3.9 illustrates a similar experiment with J.K. a few months later. He again reported at the beginning of the experiment that he was able to mark the target mentally with ease; during the experiment, he lost his ability to follow the target. Near the end he again had better control in marking the target. Figure 3.9(B) shows the decrease in congruency and diminishing of the correlation coefficient. In Figure 3.9(C), the congruency is better ( C = 0.28). On the following day, we started the first experiment with the most difficult paradigm (every third to seventh stimulation omitted) and proceeded next to the easiest paradigm. During the most difficult paradigm, J.K. said he felt unsure whether he could follow the rhythmicity of the light signals at the beginning of the experiment. During the last two thirds of the experimental period, he was able to mark a larger number of target signals. Figure 3.10(A) shows the beginning and Figure 3.10(B) shows the middle stage. The amplitudes of the EEG increased during the experiments but the correlation coefficient did not. During another experiment with the easiest paradigm, J.K. reported that he had not performed well at the beginning (Figure 3.11). However, toward the end of the experiment, he definitely had better control in marking the target. Comparison of Figure 3.8 and Figure 3.11 shows that an opposite effect occurred. In the experiment shown in Figure 3.11, the congruency between single curves was better toward the end of the experiment and C increased from 0.00 to 0.16. In five subjects, the EEG measurements during the easiest paradigm using light signals were taken after application of the most difficult paradigm. During a session with the most difficult paradigm, congruency of single rhythms like the epochs of Figure 3.8 and Figure 3.11 were not observed. Further, the correlation coefficients calculated during the four stages of the experiment remained in all cases around 0.05; they never reached values around 0.4. Comparison of easiest and most difficult paradigms — We want to mention again why the comparison of results using the easiest and the most difficult paradigms for the same subject is important to formulate a judgment about event-related pretarget rhythms. It is possible for the same subject to increase the probability of the occurrence of the target by up to 100%. The increase in the EEG amplitude and the tendency to regularity and phase ordering are reflected in correlation coefficients. If the probability of the occurrence of a target were then decreased, one would expect a less good or even bad performance. In the latter case, it might be expected that the phase ordering © 2004 by CRC Press, LLC


FIGURE 3.8 Pretarget EEG of subject J.K. (experiment 3) during the easiest paradigm (every fourth signal omitted). EEG segments were filtered in the frequency range of 8 to 13 Hz. Time scale from -1000 ms to 0 indicates 1 s recording time prior to target (omitted tone). (A) Ten single EEG samples at the beginning of experimental session (bottom). Mean value curves of 10 sweeps (middle). Broadband mean value curve from 10 sweeps (top). Filter range: 1 to 30 Hz. (B) Ten EEG samples in middle of session (bottom). Mean value curve from 10 sweeps (top). (C) Ten EEG samples at end of the session (bottom) Mean value curve from 10 sweeps (top). The correlation coefficients evaluated from 3 ensembles of 10 sweeps are shown at the top of each ensemble. (C) covers only the period from –500 to 0 ms., i.e., 500 ms prior to target. Subject's report: (A) = good performance; (B) and (C) = bad performance. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)

of the EEG and the tendency to a repeatable pattern would finish. On the basis of this reasoning, we applied both paradigms to five subjects on the same day and always obtained comparable results that were similar to the results from subject J.K. The increase in correlation coefficient means an increase in similarity of single epochs. The fact that subjects who reported good performance produced mean correlation coefficients up to 0.4 shows that an EEG can attain good phase-ordered patterns; this is contrary to the cases of recordings with less probability of occurrence. We must also emphasize that the recording of almost repeatable EEG patterns during defined experiments with cognitive targets required a large number of experiments and good cooperation of the subjects. Different time windows — In Figure 3.7 and Figure 3.11, we consistently considered the time window between 500 and 0 ms prior to target signals. Although the EEGs of most subjects depicted phase orderings starting 1000 to 700 ms prior to target signals, the time scale of -500 to 0 ms is the most common one for a rough preliminary evaluation. To avoid errors of visual inspection, we started each analysis with some moving time windows prior to target. This means that we chose six time windows at various points along the time axis of -1000 to 0 ms. The narrowest window © 2004 by CRC Press, LLC


FIGURE 3.9 Pretarget EEG of subject J.K. (experiment 15) during the same (easiest) paradigm. EEG segments were filtered in the frequency range of 8 to 13 Hz. Time scale from -1000 ms to 0 indicates 1 s recording time prior to target (omitted tone). (A) Ten single EEG samples at the beginning of experimental session (bottom). Mean value curves of 10 sweeps (middle). Broadband mean value curve from 10 sweeps (top). Filter range: 1 to 30 Hz. (B) Ten EEG samples in middle of session (bottom). Mean value curve from 10 sweeps (top). (C) Ten EEG samples at end of the session (bottom) Mean value curve from 10 sweeps (top). The correlation coefficients evaluated from 3 ensembles of 10 sweeps are shown at the top of each ensemble. (C) covers only the period from –500 to 0 ms., i.e., 500 ms prior to target. Subject's report: (A) = good performance; (B) and (C) = bad performance. Results show repetition after a few months. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)

had a duration of 300 ms. Larger correlation coefficients were to be expected for the smaller windows. Let us consider the 10 EEG sweeps illustrated in Figure 3.8. For the 500 ms before stimulation, C = 0.40. As Table 3.2 shows, C takes different values depending on the length and position of the time window. For a time window of -300 to 0 ms before stimulation, C has the highest value; the window from -700 to -300 ms has a much lower value. Table 3.2 also shows correlation coefficients of control EEG sweeps from Figure 3.7. During the recording of EEG sweeps where the subject did not report good performance, the correlation coefficients were not much higher even by choosing narrow time windows (mean value of -0.05). The control EEG of the same subject (sweeps from Figure 3.7) did not show significant values of C even with narrow time windows. For all performed experiments, searches with different time windows were carried out; the results are similar to those in Table 3.2. They indicated highly increased mean values of correlation coefficients during good performance sessions compared with resting EEGs or bad performance sessions. Evaluation of all the subjects under study gave similar results, showing that with analysis of time, the correlation coefficient is always highest during the easiest paradigm.



FIGURE 3.10 Pretarget EEG of subject J.K. (experiment 16) during the most difficult paradigm (every fourth to seventh signal omitted). EEG segments were filtered in the frequency range of 8 to 13 Hz. The time scale from –1000 ms to 0 indicates 1 s recording time prior to target (omitted light). (A) Ten single EEG samples at beginning of the experimental session (bottom). Mean value curve (top). (B) Ten EEG samples in middle of session (bottom). Mean value curve from 10 sweeps (top). (C) Ten EEG samples at end of session. Correlation coefficients cover only the period from –500 to 0 ms. Subject's report: tried to do well. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)

FIGURE 3.11 Pretarget EEG (8 to 13 Hz) of subject J.K. (experiment 19) during the easiest paradigm (every fourth signal omitted). (A) EEG samples at beginning of experimental session (bottom). Mean value curve (top). (B) EEG samples in middle of session (bottom). Mean value curve from 10 sweeps (top). (C) EEG samples at end of session. Subject's report: performance “bad” at beginning (A); increasingly good toward end of experiment [(B) and (C)]. (Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.)



TABLE 3.2 Correlation Coefficients of EEG of Left Occipital Recording of Subject J.K. during Several Time Windows Experiment 3 Time Window (ms)

Experiment 15

Experiment 19

Bad Performance C

Good Performance C

Bad Performance C

Good Performance C

Bad Performance C

Control EEG C

–0.07 –0.04 –0.1 0.1 –0.07 0.03 –0.04

0.15 0.3 0.4 –0.07 0.14 0.46 0.23

–0.04 –0.04 –0.06 –0.07 –0.06 –0.06 –0.05

0.16 0.15 0.12 0.18 0.14 0.11 0.14

0.01 0.03 0 0.02 0.04 –0.01 0.01

–0.01 –0.05 –0.04 –0.01 –0.01 –0.02 0

–1000 to 0 –700 to 0 –500 to 0 –1000 to –500 –700 to –300 –300 to 0 Mean value (6 time windows)

Source: Modified from Basar, ¸ E. et al. (1989a), in Brain Dynamics, Springer, Berlin, p. 43.

3.2.7 EXPERIMENTS

WITH

SUBJECT A.F.

Subject A.F. was a technical assistant in our research group. From the beginning of the experiments, he had a great interest in serving as a subject and provided detailed reports after every measurement session. The analysis of the single sweeps by visual inspection correlated highly with his reports in most cases. When he reported that he was able to mark the target signal mentally during a given measurement session, the single pretarget EEG curves usually showed good congruence. The agreement with his report was about 80%. Taking this degree of reliability into account, we performed seven experiments with A.F. over a period of about 3 months. The measurements and reports were as follows: The resting EEG prior to application of a paradigm was measured. The easiest paradigm with repetitive light stimuli was applied. After application of 30 light stimuli (and 10 omitted signals), A.F. wrote a report and describe the sessions as good or bad performances. The single sweeps of the pretarget EEG were plotted and the reliability of the subject's report was checked by means of analysis with correlation coefficients that revealed the degree of single sweep congruence. In six of the seven experiments A.F. reported that he had measurement periods with good and bad performance. For one experiment, he reported only bad performance. Table 3.3 shows C values in the frequency band of 8 to 13 Hz for resting EEGs and good and bad performance periods for the easiest paradigm. The subject of Experiment 32 could not achieve good performance in any measurement period. The correlation coefficient for the resting EEG had mean values no higher than 0.05 and they averaged

Memory and Brain Dynamics: Oscillations Integrating Attention, Perception, Learning, and Memory

Memory and Brain

Learning and memory: the brain in action

Learning And Memory. The Brain in Action

Learning and Memory: From Brain to Behavior

Learning and Memory

Human Learning and Memory

Learning and Memory

Learning And Memory (Gray Matter)

Neurobiology of Learning and Memory,

Learning Maps and Memory Skills

Learning And Memory (Gray Matter)

Neurobiology of Learning and Memory

Memory Rehabilitation: Integrating Theory and Practice

Brain, Vision, Memory

Knowledge, Perception and Memory (Philosophical Studies Series)

Brain Vision Memory

Memory (Handbook of Perception and Cognition)

Novel Trends in Brain Science: Brain Imaging, Learning and Memory, Stress and Fear, and Pain

Attention and Implicit Learning

Molecular Mechanisms of Learning and Memory

Concise Learning and Memory: The Editor's Selection

Molecular Mechanisms of Learning and Memory

Learning and Memory (Psychology Reference Series, 2)

Learning and Memory. A Comprehensive Reference

Memory

Memory

Memory

Memory

Memory

Memory

Memory and Brain Dynamics: Oscillations Integrating Attention, Perception, Learning, and Memory

Memory and Brain

Learning and memory: the brain in action

Learning And Memory. The Brain in Action

Learning and Memory: From Brain to Behavior

Learning and Memory

Human Learning and Memory

Learning and Memory

Learning And Memory (Gray Matter)

Neurobiology of Learning and Memory,

Learning Maps and Memory Skills

Learning And Memory (Gray Matter)

Neurobiology of Learning and Memory

Memory Rehabilitation: Integrating Theory and Practice

Brain, Vision, Memory

Knowledge, Perception and Memory (Philosophical Studies Series)

Brain Vision Memory

Memory (Handbook of Perception and Cognition)

Novel Trends in Brain Science: Brain Imaging, Learning and Memory, Stress and Fear, and Pain

Attention and Implicit Learning

Molecular Mechanisms of Learning and Memory

Concise Learning and Memory: The Editor's Selection

Molecular Mechanisms of Learning and Memory

Learning and Memory (Psychology Reference Series, 2)

Learning and Memory. A Comprehensive Reference

Memory

Memory

Memory

Memory

Memory

Memory

Recommend Documents