Investigation of Neuronal Oscillations of the Working Memory System by Transcranial Rhythmic Stimulation

Размещено на http://www.Allbest.ru/

Размещено на http://www.Allbest.ru/

The Government of the Russian federation

Federal State Autonomous Educational Institution of Higher Professional Education

National Research University - Higher School of Economics

Faculty of Social Sciences, School of Psychology, Master's program

Cognitive sciences and technologies: from neuron to cognition

FINAL QUALIFYING WORK - MASTER THESIS

Investigation of Neuronal Oscillations of the Working Memory System by Transcranial Rhythmic Stimulation

Student Ermolova Maria

Scientific adviser

Professor, PhD Feurra M.

Moscow, 2018



Table of Contents

• Introduction
• Chapter 1. Literature review
• 1.1Models of working memory
• 1.2 Working memory and cortical activity
• 1.3 Working memory and rhythmic brain activity
• 1.4 Non-invasive brain stimulation in the study of working memory
• Chapter 2. Experimental study
• 2.1 Methods
• 2.1.1 Task
• 2.1.2 Stimuli
• 2.1.3 Randomizations and Сounterbalancing
• 2.1.4 Procedure
• 2.1.5 Participants
• 2.1.6 Behavioral experiment
• 2.1.7 TACS experiment
• 2.2 Results
• 2.2.1 The behavioral experiment
• 2.2.2 The tACS experiment
• 2.2.3 Training effect
• Chapter 3. Discussion
• 3.1 Stimuli
• 3.2 The behavioral experiment
• 3.3 The tACS experiment
• References



Introduction

Working memory as a cognitive and neural phenomenon is characterized by a short-term dynamic storage of information (sensory images, motor plans, expected rewards, etc.) that is actively maintained and used in structuring goal-directed actions, decision-making and reasoning. Functionally working memory is one of the key cognitive processes underlying human cognitive ability and goal-directed behavior. In fact, multiple psychiatric and neurological troubles of cognition, affect, motivation (e.g. Alzheimers disease, Schizophrenia, morbid drug addiction) are linked to specific impairments of the working memory function.

Working memory studies are conducted in leading laboratories that use modern behavioral paradigms, neuroimaging, neurostimulation, multi-electrode recording of neuronal activity, as well as numerical and mathematical modeling.?Since the late 1970s significant amount of electrophysiological data has been obtained in animal studies, using delayed response behavioral tasks. Existing data suggest that the retention of information in working memory is implemented by self-sustaining neural activity in various areas of the cortex, including prefrontal (Fuster&Alexander, 1971), parietal (Chafee&Goldman-Rakic, 1998) and intro-temporal (Miyashita&Chang, 1988) cortices.

From the physiological point of view, working memory is a complex heterogeneous phenomenon that is constituted by a number of various neuronal processes related to the sustaining of local activity, competition and cooperation between neural ensembles and synthesis of activity with afferent and efferent information flows; this activity is characterized by memory-dependent modulations of spiking activity and population oscillations.

This project uses non-invasive brain stimulation method (tACS) to study and identify causal relationships between oscillatory neural activity and the mechanisms of working memory, more notably, its flexible and dynamic control and gating. The over-arching hypothesis for the project is that specific modulations in different oscillation bands in neural networks of higher associative cortices participate in dynamical gating of afferent information into WM, functional binding of distributed neural representations of WM content, and inhibition of processing irrelevant information that could interfere with memory trace, thus ensuring proper and efficient WM function.



Chapter 1. Literature review

A wide practical importance of resolving research questions associated with working memory stems from the fact that working memory plays a key role in the implementation of human cognitive functions that determine the quality of human life, effectiveness and success in their everyday activities. In particular, working memory is closely and causally related with such higher cognitive functions as sustained attention, motor control, speech generation, and the ability to make rational decisions.

1.1 Models of working memory

Throughout the history of working memory studies there have been more than then one theoretical model of which basic constituent working memory is comprised of and how it functions. One of the major models, called multicomponent model, has been around since 1970s (Baddley, 2003) and was proposed by Alan Baddley and Graham Hitch (D'Esposito&Postle, 2015). It is comprised of three functional components, each having its' own role: a central executive, a verbal storage system called the phonological loop, and a visual storage system called the visuospatial sketchpad. According to Baddley (2003), the phonological loop allows to hold information in memory via an articulatory rehearsal process. It applies mostly to symbolic information, eg. verbal, or information that generally has semantic representation in the long term memory. Storage of information is maintained by constant re-articulation of it. The span of working memory is limited by the amount of items which can be repeated before the first item fades from the storage. The visuospatial sketchpad, which can work in parallel with the phonological loop, provides means to temporarily store visual and spatial information, as follows from its' name. It is often studied using perceptual characteristics of stimuli. Finally, the central executive rules over the activity of other functions, providing top-down control, eg. connection of the network's nodes, allocation of attention or inhibition of activity. This is a general description of a multicomponent model, which has been developed into several variations by other scientists (D'Esposito&Postle, 2015).

Another class of theoretical models was developed rather recently and is generally called state-based models. “As a class, these models assume that the allocation of attention to internal representations - be they semantic long-term memory (eg. letters, digits, words), sensory, or motoric - underlies the short-term retention of information in working memory” (D'Esposito&Postle, 2015). These models are grounded in the idea that the content of working memory is defined by perceptual and long term memory representations being in a particular state of accessibility, largely controlled by attentional processes (Eriksson, Vogel, Lansner, Bergstrom,&Nyberg, 2015). Moreover, according to these models, there are also memory-influencing elements outside of attentional span. This is information which has not received any allocated attention but is nonetheless in a state of “activated long term memory” - a state of higher accessibility as compared to everything that is stored in a long term memory in general. A good example of such situation would be information previously held in and removed from working memory which can nonetheless cause interference (eg. slower response times) if probed within a working memory task (Eriksson et al., 2015).

Aside from functional psychological perspective, cognitive science employs perspectives and methodologies from neurobiology, biochemistry and other areas in order to study working memory. Working memory manifests itself not only in behavior but also in electrochemical activity within cortical and sub-cortical structures, which always accompanies behavior.

1.2 Working memory and cortical activity

As a complex phenomenon, working memory is associated with a distributed net of basic functions and cortical regions associated with them. According to human and animal imaging studies, working memory engages regions in frontal, parietal, occipital and temporal areas. D'Esposito and Postle (2015) go further and argue that any population of neurons in any region, from primary cortex to multimodal association cortex, can serve as storage of information until it is used in the service of goal-directed behavior.

Brain areas for encoding of the information can vary depending on the type of input (eg. in the case of visual input visual cortical areas are presumably responsible for encoding of information, formation of representations and their retention), while executive control functions, such as selective attention and inhibition, are generally associated with prefrontal cortex in general, and fronto-parietal network in particular (Palva, Kulashekhar, Hдmдlдinen,&Palva, 2010; Wager&Smith, 2003; Constantinidis&Klingberg, 2016).

Working memory has been numerously shown to have a limit to how much information it can hold simultaneously (Constantinidis&Klingberg, 2016; Gazzaley&Nobre, 2012; Alvarez&Cavanagh, 2004). For that reason, it is essential to be able to attend to stimuli selectively and inhibit the encoding of irrelevant information. Inhibition of sensory input works via top-down mechanisms (Gazzaley&Nobre, 2012; D'Esposito&Postle, 2015). D'Esposito and Postle (2015) give an example of a situation where one has to search for a friend in a crowd. Top-down mechanisms give rules of what to search for (the friend) as well as rules of ignoring irrelevant information (the rest of the faces in the crowd), i.e. not storing it in the working memory.

Extensive evidence has been given of a major role that prefrontal cortex plays in the working memory in general, and in top-down control in particular (Lara&Wallis, 2015; D'Esposito&Postle, 2015; Gazzaley&Nobre, 2012; Baddeley, 2003). Both human and animal studies have demonstrated persistent activation of prefrontal cortex during the tasks that require holding information in memory during a delay period (Lara&Wallis, 2015). A meta-analysis by Wager and Smith (2003) of 60 neuroimaging studies has demonstrated that tasks requiring executive processing generally produce more dorsal frontal activations than do storage-only tasks. Moreover, prefrontal cortex has been shown to exhibit differential preferences for various abstract higher-order variables: task rules, goals, abstract representations of object categories, and mixtures of multiple task-related variables (D'Esposito&Postle, 2015). As it is, prefrontal cortex poses as a perfect medium for top-down control, or as a “central executive”, connected to the rest of the brain regions engaged in the working memory activity.

Parietal cortex is associated with top-down control in general, and selective attention in particular (Wager&Smith, 2003; Lara&Wallis, 2015; D'Esposito&Postle, 2015). Together with prefrontal cortex it constitutes the fronto-parietal network which produces a bidirectional persistent activity during working memory tasks (D'Esposito&Postle, 2015; Siegel, Buschman,&Miller, 2015). The possible explanation to this activity is the support of top-down and bottom-up functions other than information storage per se, e.g. selective attention and filtering.

At the systems level, working memory is characterized not only by spatial allocation of neural activity but also by the frequency-domain attributes of this activity.

1.3 Working memory and rhythmic brain activity

There are indications that link individual differences in the working memory function and individual differences in brain oscillatory activity (Kwon et al., 2015; Zakrzewska&Brzezicka, 2014; Maurer et al., 2015). At the same time, a number of psychiatric and neurodegenerative diseases are associated with both pathologies of working memory functions and changes in the structure of rhythmic activity in the brain. Clarifying the role of oscillations in working memory functioning should contribute to the understanding of both the causes of individual differences in cognitive abilities as well as mechanisms of disorders in pathology.

Our theoretical understanding of how brain oscillatory activity implements and controls cognitive functions has been notably lagging behind the large body of experimental data on this phenomenon. Over the past decade a wide-ranging body of experimental data has shown coupling and interactions between oscillatory brain activity in different frequency bands that are seen during cognitive tasks. In particular, evidence points towards strong correlations between oscillatory activity characteristics and performance of working memory tasks (Bonnefond&Jensen, 2012; Wimmer, Ramon, Pasternak,&Compte, 2016; Liebe, Hoerzer, Logothetis,&Rainer, 2012; Gevins, Smith, McEvoy,&Yu, 1997; Axmacher et al., 2010). At the same time theoretical explanations of such correlations are still in their relative naissance. So far, theories are often descriptive as opposed to mechanistic, thereby lacking a causal link between oscillations and control of working memory function.

Recently, in addition to the firing rate modulation, a body of experimental data has pointed towards working memory-dependent modulation of the brain oscillatory activity. In humans and primates performing working memory tasks, increased oscillations in the gamma range (Pesaran, Pezaris, Sahani, Mitra,&Andersen, 2002), the beta range (Wimmer et al., 2016) were observed along with lower frequency ranges such as theta (Liebe et al, 2012.); while for oscillation in the alpha-range, amplification (Leiberg, Lutzenberger,&Kaiser, 2006; Bonnefond&Jensen, 2012) or suppression (Gevins et al, 1997.) were seen depending on the task contingencies. Task-dependent increases of the coherence between the various brain circuits involved in working memory in the beta range (Salazar, Dotson, Bressler,&Gray, 2012) and theta range (Liebe et al., 2012) were also demonstrated. Moreover, increased coherence was detected between theta and gamma rhythms in the delay memory maintenance period (Axmacher et al., 2010).

Increased gamma rhythm is seen as a correlate of the activation of neuronal populations involved in information processing (Schack, Vath, Petsche, Geissler,&Mцller, 2002). This is partly consistent with the experimental data showing selective strengthening of the gamma rhythm in the neural populations encoding the information that is held in the working memory (Pesaran et al, 2002; Lundqvist et al, 2016.). However, other data did not show such working memory-specific selectivity gamma rhythm (Wimmer et al., 2016) suggesting a more general role in stimulus encoding. Another hypothesized function of the gamma rhythm is to ensure functionally appropriate competition between the local neural ensembles encoding specific items and attributes in working memory whereby this gamma activity results in a selective inhibition of neural populations that are selective for task-irrelevant information. This function of gamma rhythm has been mostly advanced in theoretical studies (Ainsworth et al., 2011; Lundqvist, Herman,&Lansner, 2011). In addition, it is believed that the gamma rhythm is a mechanism for the establishment of functional connections between populations of neurons (i.e. the “communication through coherence” hypothesis of Fries (2005)), that allows us to consider it as a potential mechanism of interaction between distinct neuronal populations implementing memory trace maintenance, and the afferent sensory and efferent motor circuits of the cortex and subcortical structures (Ardid, Wang, Gomez-Cabrero,&Compte, 2010; Fujisawa&Buzsaki, 2011).

Beta-range oscillations have been on the other hand proposed to underlie functional interneuronal coupling that is distinct from the links dependent on fluctuations in the gamma range. Firstly, the most pronounced gamma modulation is observed in situations where the pattern of neuronal population activity is non-stationary (for example, in the sensory regions of the cortex in the perception of the new information and in the motor systems during movement planning). On the other hand, the most pronounced Beta-rhythm is seen during relatively stationary neuronal activity (e.g. in the primary sensory areas during stimulus expectation and motor sections while maintaining a given posture). Secondly, gamma modulation is more typical for ascending sensory information flows (from the primary to higher integrative cortex) and beta rhythm - for descending modulatory signal flow and interactions between populations located in the higher association areas of the cortex (Bressler&Richter, 2015). Third, unlike gamma rhythm providing local competition, beta rhythm plays a prominent functional role associative binding of local representations into hierarchically organized patterns of neural activity (Donner&Siegel, 2011). Fourth, gamma creates favorable conditions for the transmission of information between the populations of neurons, but not self-sustaining spike activity within a population. In contrast, beta rhythm can potentially facilitate the maintenance of spike activity (Kopell et al., 2011). These features of the beta rhythm served as a basis for hypotheses proposing beta rhythm as a mechanism for maintaining stationary conditions in cortical networks, of which information maintenance in working memory is a particular example (Engel&Fries, 2010).

The functional hypotheses about the role of gamma and beta rhythm map well onto recent experimental studies that show that the greatest increase in gamma power in the working memory tasks is seen in the stimulus perception stage, as well as at the beginning and end of the maintenance period (i.e., when the information is loaded into the memory and when it is used (retrieved) to implement behavior). The maximum beta-rhythm power, in contrast, is seen in the middle of the maintenance period, while in the perceptual stage one observes a depression of the beta rhythm (Wimmer et al, 2016; Lundqvist et al., 2016). Interestingly, evidence showed that the gamma and beta activity during working memory information maintenance are not the periodic rhythms in the classical sense, but are formed of stochastically generated by short bursts of oscillations (Lundqvist et al., 2016). Notably, the probability of generating gamma and beta bursts is negatively correlated.

At the moment the hypothesized role of the alpha rhythm in working memory function falls along two theories. The first idea purports that synchronization of neural networks in the alpha range is directly involved in the retention of the memory trace (Palva, Kulashekhar, Hдmдlдinen,&Palva, 2011). Meanwhile, ample experimental evidence points to a second competing theory according to which the oscillations in the alpha range stabilize memory trace by inhibiting irrelevant processes and impeding interference between working memory and irrelevant information. Thus, strengthening of alpha rhythm in areas of the brain, irrelevant to the working memory maintained data was found (Haegens, Osipova, Oostenveld,&Jensen, 2009; Jokisch&Jensen, 2007); It was also shown that the alpha rhythm plays a key role in the protection of working memory maintained information from distractors (Bonnefond&Jensen, 2012; Sauseng et al, 2009). Same studies also determined a correlation between alpha-power in circuits encoding distractors and success in working memory tasks, as measured by reaction times and working memory span/capacity.

To date there have been number of theoretical studies on modelling the role of oscillations in working memory functions. Recently, Dipoppa and Gutkin (2013) showed how an afferent or efferent alpha input to the attractor spiking neural network can prevent the loading of information into the working memory store and at the same time erase the current memory trace. The work explored the role of multiple oscillatory rhythms on working memory-associated sustained activity and its control. They suggested that changing spectral content of on-going brain activity can act as a rapid and flexible mechanism controlling working memory networks, between gating information into working memory, retention and clearance. In that work, an important role was assigned to the lower oscillatory bands, notably alpha. The ability to rapidly clear the memory trace and prevent the entry of new information in memory using the alpha rhythm was successfully demonstrated in the work.

In regard of the studies and ideas described above, the focus of this project is to identify the role of neuronal oscillations in the control of information encoding in the working memory and oscillatory-induced stabilization of a priori formed memory trace in the face of distracting irrelevant stimuli.

For the aforementioned reasons we hypothesized that artificial entrainment of alpha and beta rhythms will lead to improved behavioral results in working memory tasks. Pronounced effect of alpha band is expected during distractor inhibition. However, during memorization of stimuli effect of alpha will be the opposite, impairing the performance. Expected effects are the results of the following theoretical and experimental premises:

· Amplification of endogenous beta-oscillations strengthens connections within distributed representation of the information held in the working memory, thus stabilizing memory trace;

· The effect of tACS delivered in the beta band on working memory retention is likely to be observed because corresponding effects have been observed previously in the similar experimental setup (Feurra, Galli, Pavone, Rossi,&Rossi, 2016) and stands in agreement with contemporary theory of functional role of beta-rhythm (Engel&Fries, 2010; Kopell et al., 2011);

· Amplification of endogenous alpha-oscillations inhibits distribution of information about distractor within cortical networks, thus protecting memory trace from interfering with this information;

· The effect of alpha-band tACS on suppression of distractor processing is expected due to the following facts: (1) alpha-rhythm is increased during distractor presentation in the paradigm (Bonnefond&Jensen, 2012) that is used as a basis by the present project; (2) alpha-rhythm can be potentially modulated by tACS (Herrmann, Rach, Neuling,&Strьber, 2013), (3) distractor suppression can be enhanced by alpha-band rTMS (Sauseng et al., 2009).

1.4 Non-invasive brain stimulation in the study of working memory

Techniques such as repetitive transcranial magnetic stimulation (rTMS) and transcranial alternating current stimulation (tACS) give us possibilities to alter oscillatory cortical activity directly, non-invasively and painlessly. They allow us to test causal links between cognitive functions and particular oscillation frequencies. These methods, despite being rather novel to research domain, have been extensively used to study various neurophysiological and cognitive phenomena, from motor excitability (Moliadze, Atalay, Antal,&Paulus, 2012) to decision making (Ch. Herrmann et al., 2013).

TACS is a technique that allows production of sinusoidal electrical current on a desired frequency which, when applied to the skull, interferes with endogenous oscillatory cortical activity and results in the effect of an entrainment, i.e. it superimposes frequency of endogenous oscillations. The advantages of tACS include relative absence of somatosensory sensations during stimulation thereby allowing for better control conditions. Effects of tACS span from electrophysiological (Ozen et al., 2010; Moliadze et al., 2012) to behavioral (Engel, Fries,&Singer, 2001; Baєar, Baєar-Eroglu, Karakaє,&Schьrmann, 2001), allowing a wide range of experiments.

Systematic working memory studies using tACS began only relatively recently, and at the moment the results are relatively few. The most well-studied is the influence of tACS on working memory when stimulation is in the theta and gamma ranges. The possibility of influencing working memory using tACS was notably demonstrated in the works of Polanнa, Nitsche, Korman, Batsikadze,&Paulus (2012) and Feurra et al. (2016). In this study it was shown that the tACS-established gain in synchronization between the parietal and frontal cortex departments in the theta range leads to reduced reaction time in a visual working memory task. Further, Vosskuhl, Huster,&Herrmann (2015) demonstrated increased number of stored numbers using tACS to suppress the theta rhythm at frequencies identified in individual subjects. Chander et al. (2016), in contrast, showed a reduction in performance of a working memory task when endogenous frontal theta rhythm is perturbed using tACS. In recent study Hoy et al. (2015) showed that working memory can differentially improve after stimulation in the prefrontal cortex by gamma-range compared to non-oscillatory DC stimulation. Finally, Alekseichuk et al. (2016) showed improvement of spatial working memory under current stimulation, composed of theta fluctuations combined with bursts of oscillations in the upper gamma band, and the effect was dependent on the phase relationship between these two oscillations. These works thus show that tACS is indeed capable of modulating working memory depending on the location and frequency of the stimulation.

There have not been many studies on the effect of tACS in the beta-range on working memory. In the study of Feurra et al. (2016) increasing the span of stored numerals was implemented by beta-stimulation of the parietal cortex as compared to the stimulation in other frequency ranges or the control (non-stimulated) situation. However, unlike most of the works in which the effect was observed in the theta and gamma range, this study did not stimulate prefrontal but parietal cortices, and the experimental task used placed relatively light demands on executive cognitive functions (for example, did not require the conscious manipulation stored information). Thus, we can assume that the beta rhythm in this case is related directly to the retention of information in working memory, and not to the executive processes that depend on the prefrontal cortex and associated with fluctuations in the gamma/theta range. This assumption is in agreement with theoretical views on the functional role of beta-rhythm (Engel&Fries, 2010; Kopell et al., 2011).

As far as we know, at the moment there are no studies demonstrating the ability to control selective inhibition of irrelevant information (related to the processing of distractors) during the execution of tasks on working memory using the tACS. At the same time, Bonnefond&Jensen (2012) have shown an increase in the alpha rhythm on frontoparietal network in the time of presentation of distractors, and found that the subjects with a more pronounced strengthening of the alpha rhythm gave an answer more quickly, indicating a functional value of alpha in these conditions. Furthermore, Sauseng et al. (2009) investigated the effect of repeated transcranial magnetic stimulation (rTMS) with a pulse repetition frequency lying in the alpha range, during a task involving storing working memory targets from one optic hemifield and ignoring objects from another hemifield. Stimulation of the occipital-parietal regions contralateral to the ignored-item hemifield increased memory capacity. This confirmed the role of alpha rhythm in the inhibition of distractors and the ability to manage this process with the help of external perturbations. It should also be noted that the possibility to influence cortical alpha rhythm by tACS in the respective frequency was demonstrated in Helfrich et al. (2014). In addition, in this study it was shown that such manipulation impacts visual perception and processing processes.

In summary, the project is based on a large number of theoretical and experimental studies on the causal role of neuronal oscillations in the realization of working memory functions, and is at the forefront of research in this area. Identifying the mechanisms by which oscillatory activity routes and gates information in working memory will likely lead to a paradigmatic shift in our understanding of cellular and neural circuit mechanisms of higher cognitive functions.



Chapter 2. Experimental study

Experimental hypotheses of the study stem from the following theoretical premises:

· Endogenous beta-oscillations strengthen connections within distributed representation of the information held in the working memory, thus stabilizing memory trace;

· Endogenous alpha-oscillations inhibit distribution of information about distractor within cortical networks, thus protecting memory trace from interfering with this information.

We expected cortical stimulations in both alpha and beta band to improve performance in a working memory task which requires to retain memorized information for a delay period and to ignore external distractors in the meanwhile.

2.1 Methods

2.1.1 Task

The experimental paradigm of the study was based on the Sternberg working memory task. The general task involves memorization of a set of several symbols (numbers, letters, sounds etc.) showed sequentially one after another and identification whether a following probe was a part of the memorized set.

The task that was used in the present study is a modification of a Sternberg task version proposed by Bonnefond and Jensen (2012). In the main condition, five stimuli are presented sequentially at the center of the screen, for 0.033 sec each with 1.1 sec intervals. Subjects are asked to memorize the first four stimuli (“memory set”) and to ignore the fifth one (“distractor”). A distractor is taken from the same category as the other stimuli and is referred to as strong distractor. Upon demonstration of the sixth stimulus (“probe”), the subject has to reply, whether this stimulus was a member of the memorized set or not. The probe stays on the screen until participant gives an answer by pressing on one of the two buttons (labeled “Old” and “New”). Accuracy and reaction time are recorded after each answer as behavioral measures of success. Besides, there are two control conditions. If the distractor falls into a different category than other stimuli, it is called weak distractor. In the no-distractor condition, participants were asked to memorize all the five stimuli presented.

Figure 1. Experimental paradigm - depiction of a trial

In accordance with our hypothesis and results presented by Bonnefond and Jensen (2012), the no-distractor condition is the hardest for participants as they have to memorize more stimuli. Difficulty of the strong-distractor condition is expected to be higher than in the case of weak distractors. In the latter condition, interference between distractor and memorized stimuli is weaker as they fall into different categories, thereby reducing the task difficulty.

2.1.2 Stimuli

In the original study conducted by Bonnefond and Jensen (2012) letters of English alphabet were used as stimuli. Such setup did not seem to be optimal for our study for two reasons: (1) participants' performance is extremely high; (2) participants tend to use chunking strategies for memorization (i.e. grouping letters into pseudo-words) which allows them to memorize the set as a single object. For these reasons, we decided to change the stimuli. Our plan entailed creation of new stimuli which would be abstract and contain no symbolic or direct pre-established meaning for the participants. In order to avoid pre-established semantic associations, new symbols for memorization were taken from several ancient alphabets (Brahmi, Phoenician, Orkhon). All the stimuli were modified within Paint.Net program and adjusted to each other in terms of size, thickness of lines, luminance and occupied space.

Figure 2. Stimuli - symbols from ancient Brahmi, Phoenician, Orkhon alphabets, modified and adjusted to each other. They also worked as strong distractors

For the weak distractor condition, the pictures visually different from the main stimuli were needed. A new set of stimuli was created within Matlab R2015a environment. It consisted of images of graphs with thin lines as edges and thick dots as nodes.

Figure 3. Weak distractors - graphs with unilaterally connected 6 nodes

The graphs always had 6 nodes that were unilaterally connected. By appearance, they strongly differed from other symbols used in the experiment. However, size, line width, luminance, and amount of space on the screen which graphs occupied were the same as for the other stimuli.

All stimuli and distractors were reviewed regarding similarities among them. After consideration, 56 symbols and 28 graphs ended up in the final pull of stimuli.

2.1.3 Randomizations and Сounterbalancing

The appearance of all stimuli and distractors in trials was counterbalanced in the following ways:

· The probe is “old” (i.e. matches one of the stimuli in the set) in 50% of trials within one experimental block (20 trials)

· All symbols are randomized and randomizations are controlled

· Distractor never matches neither the symbols of the memory set nor the probe within the trial

· Symbols are never repeated within one set

· In case of matching trials, the probe has equal distribution of matching with either of the set elements within one experimental block (20 trials)

· Probes are never repeated within one experimental block (20 trials)

· Distractors are never repeated within one experimental block (20 trials)

· Frequencies of each symbol appearing within one block are balanced

· Sequencing of the same-condition trials is balanced (i.e. no more than 5 trials with the same condition in a sequence)

Counterbalancing of the experimental conditions (strong distractor, weak distractor, no distractors) was designed with requirements of a tACS protocol in mind (which also has three stimulation conditions: 10 hz, 20 hz, sham). The trials are combined into blocks of 60 trials. Each experiment consists of 3 sessions, each including all three experimental conditions. Sequence of conditions in each session is counterbalanced.

2.1.4 Procedure

Stimuli were presented in white on a black screen (size - 1920 x 1080, refresh rate - 60 Hz) positioned 60 cm from participant. Screen luminosity was reduced to 50% to preserve the eyes from fatigue. The experiment, which was built and run within E-Prime 2.0 Professional program, was tested in terms of presentation timings. Time lags appeared to be within the norm (under 0.015 sec) and therefore negligible.

An experiment started with both oral and written explanation of the procedure and the task. A participant began the experiment with a 10-trials training for each of the three experimental conditions.

The experiment was divided into 9 blocks 60 trials each (adding up to 180 trials for each experimental condition and 540 trials overall). Averaged accuracy and reaction time were presented three times per each block (i.e. every 20 trials). Blocks were combined into 3 sessions with 5-minute rest between sessions and 3-minute rest between blocks within sessions. The stimuli were randomized and counterbalanced within trials, blocks, and sessions. The controlled randomization was conducted within Matlab R2015a environment.

Every trial started with a fixation dot at the center of the screen, staying there for a random period of time from 0.5 to 1.5 sec.

Both hands of the participant were always on the keyboard over the buttons which had words “Old” and “New” on them (positioned over “K” and “D” buttons on a standard keyboard layout). The positions of the answer buttons in respect to each other (i.e. right-left) were counterbalanced among participants.

An experiment ended with a questionnaire about memorization strategies, in particular, about changes in the strategies. The answers were written down verbatim.

2.1.5 Participants

Participants were included in the study if they met several requirements: (1) no neurologic or psychiatric medical history; (2) normal (or corrected to normal) vision; (3) sufficient sleeping time the night prior to an experiment; (4) no psychoactive substances intake the day before an experiment. All participants received instructions about conditions of the experiment and possible unpleasant sensations (skin prickling or phosphenes) and signed a written consent form validated by the HSE ethic committee. Participants were explained the goals of the study after an experiment and received monetary compensation (counted from a rate 250 roubles per hour).

The experimental research that we performed was divided into two stages: the behavioral experiment and the experiment with transcranial alternating current stimulation (tACS).

2.1.6 Behavioral experiment

On the first stage, influence of weak-, strong- and no-distractor conditions on accuracy and response time were tested. This study gave us an opportunity to compare results of the modified experiment (which included new stimuli and a new paradigm) with the data presented in the paper of Bonnefond and Jensen (2012).

Our data was collected from 30 participants from 18 to 32 years old (22 women, mean age - 24). All of them had no medical neurological or psychiatric history, had normal or corrected to normal vision, and reported being right-handed. 5 participants were excluded for technical reasons (eg. incorrect understanding of the task, software malfunction), leaving N = 25 for analysis.

2.1.7 TACS experiment

On the second stage, subjects performed the same task, while three types of electrical stimulation were applied: 10 Hz (alpha), 20 (beta) and sham. Alpha- and beta-band oscillations presumably play a role in such working memory functions as stabilization of memory trace and inhibition of distractors. Sham condition entails stimulation going only for the first 30 seconds and works as a placebo.

Alternating current stimulation was delivered by a Brainstim stimulator through surfaces of saline-soaked sponge electrodes (size 5 x 7 cm). Stimulating electrode was positioned over P3 area according to the International 10-20 EEG System. This area was chosen on the basis of the evidences of the top-down role that parietal cortex plays in working memory. Stimulation of parietal cortex has been shown to have an effect on visual short-term memory and attention (A. Birba et al., 2017; A. Kiyonaga et al., 2014; Feurra et al., 2016),

The “return” electrode was placed over the shoulder in accordance with safety guidelines on electrical stimulation usage. tACS was delivered at an intensity of 1,000 mA (500 mA peak-to-peak). The waveform of the stimulation was sinusoidal, and there was no direct current offset. The low intensity of stimulation was used in order to avoid perception of flickering lights usually reported with higher stimulation intensities (Paulus, 2010). To minimize skin sensations, the electrodes were placed inside sponges and were constantly saturated with saline solution. Impedances wre kept below 10 kЩ throughout stimulation sessions.

Размещено на http://www.Allbest.ru/

Размещено на http://www.Allbest.ru/

Figure 4. The area P3 of parietal cortex was stimulated during tACS experiments

Along this line, the influence of stimulation on behavioral results was investigated. 21 subjects from 18 to 32 took part in the experiment (14 females, mean age - 22.2).

working memory cortical rhythmic brain



2.2 Results

2.2.1 The behavioral experiment

The goal of the behavioral experiment was to evaluate complexity of the task as well as to verify the effects presented in the original paper of Bonnefond and Jensen (2012) with regard to new symbols and a modified procedure.

The primary interest was focused on the comparison between three experimental conditions: with distractors (strong and weak) and with no distractors. According to our hypothesis and the results of Bonnefond and Jensen's study (2012):

· No-distractor condition should be the hardest to complete due to the biggest number of elements in a memory set;

· Strong distractor condition should be harder to complete than weak distractor condition because distractors from the same category are harder to ignore than those from a visually different category.

The analysis included ANOVA with repeated measures. Conditions (strong, weak, no distractor) and sessions (three consecutive sessions, each including all three conditions) were taken as independent variables. Accuracy and reaction time were taken as dependent variables.

Condition had a significant influence on a proportion of right answers (F(2,48) = 5.73, p = 0.006). Weak distractor condition turned out to be the easiest for participants. It differed significantly from both strong distractor (p = 0.04) and no-distractor (p = 0.002) tasks. Reaction time was also significantly influenced by condition (F(2,48) = 4.067, p = 0.02). Weak distractor trials were solved faster than both strong distractor (p = 0.01) and no-distractor (p = 0.03) trials. Strong distractor and no-distractor condition did not differ from each other significantly neither by accuracy nor by reaction time.



Figure 5. The difference of accuracy percentages between experimental conditions. Most of the difference can be observed as a result of the weak distractor condition differing from other two conditions

Figure 6. Differences in reaction times between experimental conditions. Most of the differences can be observed as a result of the weak distractor condition effect from the other two conditions

Overall, acquired results conform to our hypothesis: condition had a significant influence over both accuracy and reaction time.

As was discussed above, new paradigm and new symbols were introduced to the study in order to complicate the task while keeping the effects of distractors. Results of both reaction time and accuracy show the partial success of our modifications. Results from weak distractor as compared to other conditions show the effect of distractor inhibition as expected from the original study. However, the difference between strong distractor and no-distractor conditions turned out to be insignificant, which differs from the results acquired by Bonnefond and Jensen (2012).

Session had a significant influence on both accuracy (F(2,48) = 14.23, p < 0.01) and reaction time (F(2,48) = 26.4, p < 0.01). Considering accuracy, training effect was prominent for the first session but subsided for the two remaining sessions. Reaction time, on the other hand, showed prominent training effect throughout the whole experiment, from the first till the third session.

Figure 7. The difference of accuracy percentages between experimental sessions. Training effect is apparent for the first session

Figure 8. The difference of reaction time between experimental sessions. Training effect is apparent throughout all three sessions

The post-experimental questionnaire allowed us to analyze memorization strategies of the participants, and changes in these strategies in particular. Two dominant strategies came forward: associative (64%) and visual (4%) memorization. Associative memory often included phonological loop - participants reported naming symbols and continuously repeating the names until appearance of the probe. This pool of participants also reported ignoring the distractor by not naming it. 32% of participants reported changing the strategy during the experiment or using a mixed strategy.

2.2.2 The tACS experiment

The goal of the experiment was to test general effect of stimulation on working memory. Two factors were analyzed during this experimental stage - stimulation protocol and experimental condition.

The effect of experimental condition was tested in order to verify whether the task works as intended. Differences in both reaction time (p = 0.002) and accuracy (p = 0.01) between conditions were found to be significant. Post-hoc analysis of accuracy and reaction time showed that, same as in the behavioral experiment, these results are present mainly due to differences between weak condition and both other conditions. The outcome conformed to the results of the behavioral experiment, therefore the conclusion was made that the task works as expected.

Figure 9. The difference of accuracy percentages between experimental conditions in tACS study. The trend conforms to the one in the behavioral study

Figure 10. The difference of reaction time between experimental conditions in tACS study. The trend conforms to the one in the behavioral study

Overall effect of both stimulation conditions (10 hz and 20 hz) appeared to be insignificant: p = 0.23 for accuracy and p = 0.18 for reaction time. The only significant difference was found in post-hoc analysis in accuracy between 20 hz stimulation and sham stimulation (p = 0.04). In reaction time difference between 10 hz and 20 hz conditions was near to be significant (p = 0.08), however, both conditions didn't reach significant difference from the sham stimulation.

Figure 11. The difference of accuracy percentages between stimulation conditions in the tACS study. The only significant difference was found between 20 hz and sham stimulation.



Figure 12. The difference of reaction times between stimulation conditions in the tACS study. No significant differences were found

2.2.3 Training effect

An important issue that arose during both behavioral and tACS experiments was the training effect which manifested itself in the fact that, irrelevant of condition sequence, both behavioral results constantly improved along the sessions - from the start to the end of the experiment (p < 0.01 for both reaction time and accuracy in the behavioral study and for reaction time in the tACS study).

Figure 13. The difference of reaction times between experimental sessions in the tACS study

Learning effect per se is not a problem for a well-counterbalanced experiment. Moreover, learning effect, as well as fatigue effect, is present in most of the experiments in one way or another. However, the problem lies in the individual differences in the learning curves between subjects. While within-subject counterbalancing of experimental conditions in the behavioral experiment can make up for this effect, within-subject counterbalancing of stimulation conditions in the tACS experiment is undesirable (due to a great length of such experiment). Thus, learning effect cannot be accounted for during the analysis of stimulation effects.



Chapter 3. Discussion

3.1 Stimuli

In the original study by Bonnefond and Jensen (2012), which our study was based on, the consonant letters of English alphabet were used as stimuli. In order to complicate the task and to avoid the phonological loop we decided to change the stimuli into symbols unknown to our participants. As a result, we have met the first goal of making the task more difficult. However, the second goal of escaping the phonological loop was not met because most of our participants used the phonological loop strategy: they gave names to the symbols (based on their associations) and held them in working memory by continuously repeating these names, in their mind or even aloud. Still, the difference between strong-distractor and weak-distractor conditions stayed. Thus, we can say that the task works as we intended it to.

3.2 The behavioral experiment

The goal of the first study was to test humans' ability to retain information in the working memory while ignoring external distractors irrelevant to the task.

The study conformed to our hypothesis: the results showed that the weak distractors from a different category, which were introduced to the study, work as planned - they are easier to ignore than the distractors from the same category.

Load-5 condition, however, did not work as we expected. In the original study by Bonnefond and Jensen (2012) load-5 condition resulted in being the most challenging out of all three conditions. It is important to note that in their study the load-5 condition was not a part of the main experiment but was included in a supplementary behavioral experiment on a separate day. In our case, all conditions were included in the experiment. Although in the load-5 condition participants had to hold more items in working memory (5 instead of 4), it did not show any significant difference from a strong distractor condition. The reason for that might be the fact that conscious ignoring of distractors is not functionally identical to encoding less input. Conscious ignoring of a distractor might require additional activity which feeds on resulting accuracy and reaction time.

3.3 The tACS experiment

The goal of the second study was to test the effects of artificially imposed alpha and beta oscillations on the ability to complete a working memory task with distractors and delayed response.

In the original study by Bonnefond and Jensen (2012) the data revealed a robust adjustment of the phase of alpha oscillations in anticipation of the distracter. Our experiment failed to demonstrate overall effect of either stimulation on the performance. The only significant influence of stimulation was exhibited by accuracy rise under 20 hz stimulation as compared to the sham. Additionally, the difference between reaction time under 10 hz and 20 hz stimulation showed a trend, i.e. near-to-significant effect. Apart from the conclusion that investigated frequencies may not play the assumed role in working memory, there are two other possible reasons for this outcome.

The first reason lies in the learning effect discussed above. In the original study by Bonnefond and Jensen (2012) ceiling effect was reached after approximately 10 trials. In our experiment each participant has a unique learning curve, which on average spanned for 7 blocks out of 9 before reaching the ceiling performance. Such outcome may influence the resulting differences in stimulation effects. The possible solution to this challenge is modification of experimental protocol: introduction of one more experimental day only for training before the day of an actual experiment. This can give enough time for most of participants to reach their ceiling performance before the start of an experiment.