Memory and Brain Dynamics

Oscillations Integrating Attention, Perception, Learning, and Memory

Erol Basar

• Publisher: CRC Press
• Year: 2004
• ISBN:
  9781134394838
• Edition: 1
• Topic: Biological Sciences
• Subtopic: Neuroscience
• Index: Biological Sciences

Memory itself is inseparable from all other brain functions and involves distributed dynamic neural processes. A wealth of publications in neuroscience literature report that the concerted action of distributed multiple oscillatory processes (EEG oscillations) play a major role in brain functioning. The analysis of function-related brain oscillatio

Part I
Foundations

1
Introduction and Core 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 IS CLASSIFICATION OF OBJECTS BY ACTIVATION OF ASSOCIATIVE NETS

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 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 Başar, 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, short-term 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 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; Goldman-Rakic 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 (Başar, 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 act...

Table of contents

• COVER PAGE
• TITLE PAGE
• COPYRIGHT PAGE
• SERIES PREFACE
• PREFACE
• AUTHOR
• PART I: FOUNDATIONS
• PART II: EXPERIMENTS AND THEIR INTERPRETATION
• PART III: MEMORY FUNCTION: MODELS AND THEORIES
• EPILOGUE: FROM EEG-BRAIN DYNAMICS TO MEMORY-BRAIN DYNAMICS
• REFERENCES
• ABBREVIATIONS AND GLOSSARY
• APPENDIX: RELEVANT MATHEMATICAL METHODS