WORKING MEMORY AND ELECTROENCEPHALOGRAPHIC (EEG) COHERENCE

Brežan Simon, Štukovnik Vita
Faculty of Medicine, University of Ljubljana, Ljubljana, Slovenia Institute of Clinical Neurophysiology, University Medical Centre Ljubljana
ABSTRACT
Cognitive neuroscience searches for neuronal correlates of higher mental functions, including working memory. Electrophysiological methods explore brain functions on the level of electrical activity of the brain with high time resolution. This allows the study of basic mechanisms of information processing and operational system of the brain. EEG coherence, as a newer method of EEG signal analysis, is suitable for studying integrative brain function. It measures interregional synchronized oscillatory activity of neurons as one possible mechanism of the functional integration – binding of different communicating brain areas. The binding could serve as a mechanism of functional coupling, present with different distributed brain processes and behavioural modes, for example working memory.

WORKING MEMORY
Memory includes the following main processes: encoding (cognitive events during presentation of information), storage (maintenance) and retrieval (recall) of information. One of the best studied models of working memory is Baddeley’s model (Baddeley and Hitch, 1974). The model consists of the attentional control system – the central executive (CE), and three subsystems (slave systems): the phonological loop, the visuospatial sketchpad and the episodic buffer. CE is responsible for controlling subsystems, strategy selection, planning and manipulation of information, providing link between subsystems and long-term memory. Information is stored and actively rehearsed separately in the modal specific subsystems – visuospatial sketchpad and phonological loop. The role of episodic buffer is binding of information from the slave systems and long-term memory and temporary storing them in a multi-dimensional code.

NEUROANATOMICAL AND NEUROPHYSIOLOGICAL BASIS OF WORKING MEMORY
Various components of working memory are mediated by different anatomically separated neuronal networks. Cell electrophysiological studies on animal models and functional brain imaging studies on human subjects have shown that working memory is mainly based on the activity in prefrontal, premotor, limbic, and posterior association parietal areas of the brain (Fletcher and Henson, 2001). The basic neurophysiological mechanism of working memory is supposed to be based on repeated reverberations of electrical impulses in reverberational (feedback) loops (Štrucl, 1999). A repeated excitation of a synapse in such an excitational loop leads to postsynaptic facilitation, and this can be a basis for the maintenance of specific information, coded by activated neural network pattern. The exact mechanisms for coordination and dynamic interactions of anatomically separated brain regions with neural mechanisms for information processing itself are not yet known. The functional integration (‘binding’) of different brain areas, responsible for specific (working memory) functions, is one of the key problems in understanding the brain function – a question not yet resolved. Binding is perhaps mediated by a synchronized oscillatory activity of neuronal networks, which can be determined by the electroencephalographic (EEG) coherence analysis.

Figure 3. Illustration of EEG scalp electrodes and functional coupling between different brain areas (the colour of inter-electrode connections shows different coherence values).

Figure 1. Basic structure of memory with three memory stores (Atkinson and Schiffrin, 1968).

The short-term memory is also termed working memory. This emphasizes its active role in cognitive processes – besides temporal storage of information, the information is also available for active manipulation. Working memory can therefore be defined as a complex of cognitive processes for maintenance, manipulation and utilization of mental representations.
Figure 2. Baddeley’s working memory model (Baddeley, 2000).

COGNITIVE PARADIGMS FOR WORKING MEMORY ASSESSMENT
Different cognitive paradigms are used for assessment of specific working memory processes: At the Institute of Clinical Neurophysiology our research group plans to carry out the EEG study of working memory in the context of synchronized brain oscillations, using the EEG coherence method. The study design will be based on the modified Sternberg paradigm, which in comparison to other current EEG studies allows independent assessment of both central executive component and maintenance processes of working memory. This modified Sternberg paradigm includes: – maintenance of sequentially presented verbal stimuli (letters) and their serial position (maintenance processes of working memory) – alphabetically reordering of presented letters and maintenance of reordered sequence of letters (executive component and maintenance processes of working memory) – control task (different dilemmas and problems exist what should be the appropriate control task which would ideally exclude working memory and other uncontrolled intrinsic mental processes)

EEG coherence and power spectra analyses are newer methods for analyzing the EEG signal. Power spectra are a measure for a degree of the representation of a specific frequency band in the signal. The power spectra changes reflect different levels of regional cortical activity or different levels of their synchronization. Synchronization usually reflects cortical inactivity, while desynchronization reflects activity. EEG coherence reflects different degrees of synchronization between separate cortical regions and is independent measure of synchronized oscillatory activity, which could mediate the functional coupling – binding and communication between distributed brain centers involved in specific tasks. Power spectra and coherence changes therefore reflect two different operational systems of the brain. Many studies have found that synchronous oscillations correlate with specific behavioural contexts and cognitive tasks (Hallet, 2000). Most important is the difference between coherence values in the test task compared to basal values in the control tasks (∆Cxy).

– STERNBERG PARADIGM: maintenance of information (activity of phonological loop, visuospatial sketch pad) – GO/NO-GO TEST: central executive component of working memory – STROOP COLOUR-WORD TEST: central executive component of working memory – TOWER OF LONDON: central executive component of working memory, maintenance of information (activity of phonological loop, visuospatial sketch pad). – N-BACK TASK: maintenance of information (activity of phonological loop, visuospatial sketchpad), central executive component of working memory

ELECTROENCEPHALOGRAPHY (EEG), SIGNAL ANALYSIS: EEG COHERENCE AND POWER SPECTRA
Electroencephalography (EEG) is a method, which measures repeated, periodic electrical activity of cortical neurons. The summed activity of many neurons results in field potentials, many of them constituting macropotential (EEG signal). Macropotential is a result of changing pattern of the synchronization and desynchronization of regional brain cells; this results in amplitude changes of specific frequency bands. EEG has great time resolution and shows distinct patterns of activity (brain rhythms, oscillations). Brain rhythms can be divided in many frequency bands (delta: 0,5-4 Hz, theta: 4-7 Hz, alpha: 8-13 Hz, beta: 13-30 Hz, gamma: over 30 Hz) with their specific functional and behavioural correlates, activating contexts and different spatial scales. Their possible role is switching neural networks between different functional states with activating or inhibiting proper neural systems.

Cxy(ω): coherence value between signals x and y
Φxy(ω)- value of cross-correlation power spectrum of signals x, y Φxx(ω)- value of auto-correlation power spectrum of signal x Φyy(ω)- value of auto-correlation power spectrum of signal y

Figure 4. Power spectrum increase (red colour) in theta frequency band for working memory retention task compared to control task (increases in frontal, central and parietal brain areas); computer brain model; /personal archive material/.

Figure 5. Coherence increases (red colour) and decreases (blue colour) in theta frequency band for working memory retention task compared to control task (increases between frontal, central and parietal brain areas); computer brain model; /personal archive material/

WORKING MEMORY AND SYNCHRONIZED BRAIN OSCILLATIONS: POWER SPECTRA AND EEG COHERENCE STUDIES
The neurophysiological theory of (working) memory:

The majority of EEG studies have shown that working memory tasks are often related to: • regional neuronal desynchronization and consequent power spectrum decrease in lower alpha band (showing a non-specific effect of attention, mental effort); • regional desynchronization leading to power spectrum decrease in upper alpha band (correlate of semantic processing); • paradoxical alpha synchronization, power spectrum increases (active inhibition of disturbing neural networks not needed for the memory task as optimization of processes; Klimesch et al., 1998, etc.); • increased interregional synchronization resulting in coherence increases in theta frequency band – working memory processes (Serrien et al., 2003; Sauseng et al., 2004; Sarnthein et al., 1998; Jensen et al., 2002); • increased regional synchronization with increases of power spectrum in theta frequency band: frontal midline theta rhythm – memory maintenance, attention, mental effort (Klimesch et al., 1997, Gevins et al., 1997, model Lisman, Idiart and Jensen, 1998, etc.). The source of theta rhythm? – anterior cingulate cortex, cortico-hippocampal feedback loops; • increased power spectra and coherence in gamma band (sensory, perceptional and attentional processes, working memory processes); • changes of power spectra and coherence with different memory load (Gevins et al., 1997; Jensen, 2000, etc.). RESEARCH DILEMMAS: different non-consistent definitions of specific frequency bands for different brain rhythms by authors, possible interindividual differences in oscillatory frequencies, usage of different cognitive paradigms, methodological and technical problems, artefacts, signal to noise ratio…

Spatial scales of brain oscillations involved in working memory Mainly pre(frontal) and posterior association brain areas were found oscillating synchronously (theta, gamma, alpha band; Sarnthein et al., 1998, etc.): • prefrontal cortex itself rehearses and maintains the information transferred from posterior storage areas and manipulates with it; • top-down control by prefrontal cortex and its central executive, gating of irrelevant sensory signals coming from posterior regions and memory process optimization. Lisman/Idiart/Jensen (LIJ) model – example of neurophysiological model of working memory Results of studies showing the effect of different memory load on theta oscillations confirm the LIJ model (1998). Model predicts that different memory representations (7 items +-2) are encoded and activated separately and sequentially in subsequent phases of theta cycle. Theta input serves as a signal carrier and induces repeating of scanning and working memory maintenance processes. This means that model is based on serial information processing principles. Therefore it can explain empirical data (Sternberg, 1964) on increase of reaction times with respect to working memory load (number of items). The memory code itself is mediated by synchronous gamma oscillations, induced by pattern of alternating excitation and inhibition of different memory codes. The source of theta oscillations could be the hippocampo-cortical reverberating loops. Working memory processes are thereby potentially mediated by posterior (sensory)-hippocampal-prefrontal neural network, which is integrated and functionally coupled by the synchronization of oscillations in theta frequency band.
Figure 6. Diagrams indicating the concept of LIJ (Lisman, Idiart, Jensen, 1998) working memory model. Three memory items (A, B, C) are loaded in memory buffer, the theta period increases by one gamma period with each item added. In retention interval (delay period), items are maintained by activity-dependent intrinsic properties of the neurons coding these items. After probe presentation the items can be scanned – compared with the probe as they are activated. After scanning, motor response and answer can be initiated. RT – reaction time

This is in accordance with Baddeley’s model of working memory which predicts constant interplay and functional coupling between central executive control system (probably mediated by prefrontal cortex, Collette et al., 2002) and slave subsystems for storage of verbal and visuospatial information, located in posterior brain regions (Gathercole, 1999).

• Brain oscillations in different frequency bands subserve specific (memory) functions and operate over different spatial scales. • Multiple superimposed synchronized (coherent) oscillations in different frequency bands with different spatial patterns and functional correlates govern specific mental functions.

CONCLUSION WITH CLINICAL IMPLICATIONS OF EEG COHERENCE
Synchronization in gamma frequency band is believed to be correlated with sensory processing and the very content of information processing, but could also reflect increased attentiveness. The increased coherence for interactions between posterior and frontal areas in theta frequency band (also with probable involvement of hippocampus) on the other hand, possibly mediates working memory processes (storage, rehearsal and scanning). In this way, slower rhythms could act as signal carriers constituting the context whereas fast waves refer to the entire content of the representation. The neuronal synchronization – functional coupling plays a role in interaction of posterior association cortex, where sensory information is stored, and (pre)frontal cortex, where relevant current information is held, rehearsed and updated. Verbal memory tasks seem to activate primarily the left brain hemisphere, while visuospatial memory tasks activate predominantly the right brain hemisphere. The value of EEG coherence method for brain research is greatest when combined with other brain research approaches which show different aspects of the brain function. With choosing appropriate cognitive paradigms and neuropsychological tests, it is possible to study physiological and pathophysiological aspects of cognitive, motor and sensory brain function. Recently, some studies (Hogan et al., 2003; Spencer et al., 2003, etc.) also showed important differences in coherence values when comparing working memory in neurodegenerative or psychiatric patients (Alzheimer disease, Parkinson disease, schizophrenia) to healthy controls. This opens new future perspectives for possible search of pathophysiological mechanisms and etiological factors contributing to many different neurological diseases.
REFERENCES: 1. Babiloni, C., Carducci, F., Vecchio, F., Rossi, S., Babiloni, F., Cincotti, F., Cola, B., Miniussi, C., Rossini, P.M. (2004). Functional frontoparietal connectivity during short-term memory as revealed by high resolution EEG coherence analysis. Behavioral Neurosciencies, 118(4), 687-697. 2. Baddeley, A. (2000). The episodic buffer: a new component of working memory? Trends in Cognitive Science, 4(11), 417-423. 3. Gevins, A., Smith, M.E., McEvoy, L., Yu, D. (1997). High-resolution EEG mapping of cortical activation related to working memory: effects of task difficulty, type of processing and practice. Cerebral Cortex, 7, 374-385. 4. Jensen, O., Lisman, J.E. (1998). An oscillatory short-term memory buffer model can account for data on the Sternberg task. The Journal of Neuroscience, 18(24), 10688-10699. 5. Klimesch W. (1996). Memory processes, brain oscillations and EEG synchronization. International Journal of Psychophysiology, 24, 6-100. 6. Sarnthein, J., Petsche, H., Rappelsberger, P., Shaw, G.L., von Stein, A. (1998). Synchronization between prefrontal and posterior association cortex during human working memory. Neurobiology, 95, 7092-7096. 7. Serrien, D.J., Pogosyan A.H., Brown, P. (2003). Influence of working memory on patterns of motor related cortico-cortical coupling. Experimental Brain Research.
Figure 7. LIJ model of a network functioning as a multi-item short-term memory buffer. Theta and gamma oscillations play an important role in the concept. An afterdepolarization (ADP) is triggered after a cell fires (sensory input) and it causes depolarizing ramp that serves to trigger the same cell to fire again after delay. These ramps are temporarily offset for different memories, an offset that causes different memories to fire in different gamma cycles. The key function of this buffer is to perpetuate the firing of cells in a way that retains serial order. The repeat time is determined by theta oscillations due to external input. Gamma oscillations arise from alternating global feedback inhibition and excitation (the cell with most depolarized ramp will fire again) due to separate firing of different memory codes.