Transcranial Alternating Current Stimulation At Theta Frequency To Left Parietal Cortex Impairs Associative, But Not Perceptual, Memory Encoding

For more information:ali.ma@wecistanche.com

ABSTRACT

Neural oscillations in the theta range (4–8 Hz) are thought to underlie associative memory function in the hippocampal-cortical network. While there is ample evidence supporting the role of theta oscillations in animal and human memory, most evidence is correlational. Non-invasive brain stimulation (NIBS) can be employed to modulate cortical oscillatory activity to influence brain activity, and possibly modulate deeper brain regions, such as the hippocampus, through strong and reliable cortico-hippocampal functional connections. We applied focal transcranial alternating current stimulation (tACS) at 6 Hz over the left parietal cortex to modulate brain activity in the putative cortico-hippocampal network to influence associative memory encoding. After encoding and brain stimulation, participants completed an associative memory and a perceptual recognition task. Results showed that theta tACS significantly decreased associative memory performance but did not affect perceptual memory performance. These results show that parietal theta tACS modulates associative processing separately from perceptual processing, and further substantiate the hypothesis that theta oscillations are implicated in the cortico-hippocampal network and associative encoding.

Click to Cistanche herba for improving memory

1 Introduction

Associative memory is commonly defined as a memory system that constructs relationships between initially unrelated items (Suzuki, 2008; Tulving, 1984). The ability to form associations is core to episodic memory function (Kahana, Howard, & Polyn, 2008), and allows us to remember details of specific events and recall abstract information, required for daily tasks and interactions (Herweg, Solomon, & Kahana, 2020). The study of neural mechanisms underlying associative memory has traditionally focused on the hippocampus and its central role in widespread brain networks. Given the rapid and flexible nature of associative memory, a dynamic mechanism for neural communication is required. In this regard, neural oscillatory rhythms are of particular interest because they are thought to contribute to the communication between neuronal populations that are simultaneously engaged in cognitive processes (Buzs´ Aki & Draguhn, 2004; Wang, 2010). In particular, oscillatory activity in the theta range (4–8 Hz) has been extensively researched in animal and human cognition (Buzs´ aki, 2002; Jacobs, 2014; Meador et al., 1991; Lisman & Jensen, 2013; Maurer et al., 2015; Sauseng, Griesmayr, Freunberger, & Klimesch, 2010), as it has been hypothesized to play an integral part in hippocampal function during spatial navigation and associative memory processing (Buzs´ aki, 2002; Skin, Kilianski, & McNaughton, 2019; Wallenstein, Hasselmo, & Eichenbaum, 1998). The success of memory encoding seems to rely on the increase of theta power in hippocampal regions (Burke et al., 2013; Khader, Jost, Ranganath, & R¨ Osler, 2010), although decreased theta power for subsequently successfully remembered items has also been observed (Crespo-García et al., 2016; Fellner et al., 2016; Lega, Jacobs, & Kahana, 2012).

Alyssa Meng a , Max Kaiser a , Tom A. de Graaf a,b , Felix Dücker a,b , Alexander T. Sack a,b , Peter De Weerd a,b , Vincent van de Ven a,b,*

a Department of Cognitive Neuroscience, Faculty of Psychology and Neuroscience, Maastricht University, PO Box 616, 6200 MD Maastricht, the Netherlands b Maastricht Brain Imaging Center (MBIC), Maastricht University, PO Box 616, 6200 MD Maastricht, the Netherlands

The study of hippocampal oscillatory mechanisms in associative memory, particularly in humans, can be challenging, as its location deep within the brain precludes direct non-invasive measurement of neural activity. Moreover, measuring oscillations does not always allow the assessment of causal contributions, which ideally involves direct experimental manipulation of the neural mechanisms of interest. Such manipulation can be achieved with non-invasive brain stimulation (NIBS), by taking advantage of local (i.e., cortical) and network functional connectivity to investigate cortical and subcortical memory effects. Recent transcranial magnetic stimulation (TMS) studies showed that cortical stimulation may alter activity and connectivity within a hippocampal-cortical network, which could, in turn, lead to enhanced memory performance (Hermiller, VanHaerents, Raij, & Voss, 2019; Tambini, Nee, & D’Esposito, 2018; Wang et al., 2014). Here, experimentally induced changes in cortical activity may spread to more remote areas through default functional connectivity (Kim, Ekstrom, & Tandon, 2016). Particularly, the lateral parietal cortices are strongly and reliably functionally connected to the hippocampus (Buckner, Andrews-Hanna, & Schacter, 2008; van de Ven, Wingen, Kuypers, Ramaekers, & Formisano, 2013; Vincent et al., 2006). The ease of stimulation of the parietal cortex, combined with the limited discomfort from transcranial stimulation over these areas, makes this area an important target for network-based neuromodulation of associative memory.

A different NIBS technique, transcranial alternating current stimulation (tACS), allows the modulation of neural networks through a predetermined exogenous frequency (Abellaneda-P´erez et al., 2020; Cabral Calderin et al., 2016; Helfrich et al., 2014; Reed & Kadosh, 2018; Vossen, Gross, & That, 2015; Vosskuhl, Huster, & Herrmann, 2016; Herrmann et al., 2013). Concurrent tACS-fMRI studies have shown that tACS modulates the neural activity and inter-regional connectivity for cortical and subcortical areas distal to the scalp location of stimulation (B¨ Eichinger et al., 2017; Cabral-Calderin et al., 2016; Vosskuhl et al., 2016; Abellaneda-Perez et al., 2020). Findings of a recent study suggest that tACS can be used to modulate activity in the hippocampal-cortical network. Abellaneda-P´erez et al. (2020) administered theta-tACS over the dorsolateral prefrontal cortex while participants underwent fMRI scanning during a resting state or while performing an N-back working memory task. Compared to direct current stimulation (tDCS), tACS resulted in increased activity of multiple default mode network areas, including the hippocampus and inferior parietal cortex, during resting state as well as the N-back task. These findings further show that non-invasive brain stimulation of cortical areas can be used to modulate hippocampal activity through network connectivity.

To date, tACS has been used to specifically study the oscillatory mechanisms in associative memory in only a limited number of studies. In one study (Lang, Gan, Alrazi, & Monchi, 2019), participants encoded the associations between arbitrary face-scene pairs while undergoing high-definition theta-tACS (or anodal tDCS or sham stimulation) over the fusiform area, after which they completed an associative memory test. Results showed that theta tACS during encoding enhanced associative memory performance, compared to tDCS or sham. However, tACS also improved correct rejections of lure items, which suggests a more general cognitive enhancement effect. Previous studies reported theta oscillatory activity in human and animal prefrontal and sensory cortex during working memory procedures (Lee et al., 2005; Raghavachari et al., 2006; Sauseng, Griesmayr, Freunberger, & Klimesch, 2010), as well as during perceptual processing (Helfrich, Fiebelkorn, Szczepanski, Lin, Parvizi, Knight, & Kastner, 2018; Landau, Schreyer, Van Pelt, & Fries, 2015; Somer, Allen, Brooks, Buttrill, & Javadi, 2020), which support the contribution of various cognitive processes to associative memory performance. Indeed, several studies reported enhanced working memory performance after theta tACS was administered over the parietal (Jauˇsovec & Jauˇsovec, 2014) or prefrontal cortex (Alekseichuk, Turi, de Lara, Antal, & Paulus, 2016; Abellaneda-Perez et al., 2020; Reinhart and Nguyen, 2019), while tACS-induced frontoparietal inter-regional desynchronization at 6 Hz resulted in decreased working memory performance (Alekseichuk, Pabel, Antal, & Paulus, 2017). However, another study in which theta tACS was administered over the left prefrontal cortex during associative memory encoding did not alter memory performance in young healthy participants (Klink, Peter, Wyss, & Kl¨ Oppel, 2020).

In sum, findings of non-invasive theta stimulation on associative memory remain inconclusive. Several factors need to be addressed to further our understanding. First, the effect of theta stimulation on associative memory may be strengthened when applied to cortical areas that are strongly connected to the hippocampal-cortical networks. Based on recent TMS studies, the parietal cortex may be a good candidate for tACS administration. However, a direct application of theta tACS over the parietal cortex to selectively alter associative memory processing has not yet been reported. Second, the specificity of theta stimulation on associative memory requires that non-associative cognitive functions were not modulated. This is especially relevant given that the parietal cortex plays a role in associative memory as well as perceptual processing (Cabeza et al., 2011; Culham & Kanwisher, 2001; Gottlieb, 2007). Parietal theta tACS could thus modulate associative processing indirectly by modulating sensory or attentional encoding, which in turn could modulate associative and non-associative memory performance, such as perceptual recognition. Supporting the functional specificity of theta oscillatory activity in associative memory would require a theta tACS effect on associative memory in the absence of an effect on perceptual memory.

In the current study, we administered focal tACS at 6 Hz over the left parietal cortex in a within-subject sham-controlled design while participants encoded a face-scene pairing task. After the encoding phase and tACS, participants completed two memory tests that measured knowledge about the item pairings (i.e., associative memory) and item identity (i.e., perceptual memory). We hypothesized that parietal theta tACS, compared to sham stimulation, would affect associative, but not perceptual memory performance.

2. Methods

2.1. Participants

We recruited twenty healthy volunteers (mean age [SD] =21.68 years [2.84],12 females) from the academic environment of Maastricht University via an online recruitment system. All participants were pre-screened using a non-invasive brain stimulation screening form for eligibility, health risks, and adverse effects. Potential participants who reported any NIBS contraindications during their screening phase were excluded. All participants gave their written informed consent before participation. The study was approved by the ethics review committee of psychology and neuroscience (ERCPN) of the Faculty of Psychology and Neuroscience (FPN)at Maastricht University. Participant safety and comfort were monitored throughout brain stimulation to catch any adverse effects due to stimulation. Participants were compensated with a gift voucher or course credit.

2.2. Task procedure

Participants completed a visual associative memory task, the"Face and Scene Task"(FAST; Lang et al,2019), which was programmed in PsychoPy((Peirce et al.2019)). The task consisted of an encoding and a test phase. During the encoding phase of the task, participants were presented with image pairs that comprised a neutral face and a natural scene. Neutral faces of all ethnicities and ages were gathered from the Chicago Face Database (CFD Version 2.0.3; Ma, Correll,& Wittenbrink, 2015). The faces did not have easily distinguishable features such as piercings or tattoos. All scene images were taken from an online image database.

On each trial, participants saw a pair of images consisting of a neutral face presented on the left side of the screen, and a scene presented on the right side of the screen. They had to indicate how the person in the presented image would appraise the corresponding gene, Participants were instructed to provide their judgment via a rating scale that was visualized as a scaled slider (range-3 to 3)below the image pair to ensure the participants were actively encoding (Fig.1). The image pair was shown until a button was pressed, allowing self-paced encoding, but with an upper limit of 8 s.

Each face-scene pair was followed by a 100 ms inter-stimulus interval during which a fixation cross was presented. The encoding phase consisted of 45 face-scene pairs in total, which were separated into 5 blocks of 9 image pairs. Each block was separated by a 5 s interval.

Fig.1. Encoding and retrieval phase. During the encoding phase(A), participants saw face scene pairs and indicated how positively or negatively they thought that the person appraised the scene. During the memory test phase (B), participants saw face-scene pairs and had to indicate whether the items were previously seen as a pair(~1: seen together), seen as items of different pairs(~2:seen but not together"), or were not seen in any way before ("3: never seen"). Participants were given 8 s to encode and rate each face scene pair. Each face-scene pair was followed by 100 ms interstimulus. In the retrieval phase, participants were given unlimited time to respond. Each face-scene retrieval pair was followed by 100 ms interstimulus

Each participant performed two versions of this task, which were identical in design but used non-overlapping sets of face and scene images. Each encoding task was presented during either 6 Hz or sham tACS stimulation. To familiarise the participants with the task, an abridged version of 15 novel pairs were given as training before the experimental conditions, and standardized instructions were presented at the beginning of each version.

Following the encoding task, participants completed a video game as a distraction task for 5 min to diminish the recency effect of short-term memory and minimize possible stimulation after-effects. The task is believed to be engaging enough to prevent effortful rehearsal.

After the distraction task, participants completed the memory testing phase, which comprised an associative and perceptual recognition task. In the associative recognition task, participants saw 45 presentations of either 1)the original face-scene pairs seen during encoding,2)pairs of faces and scenes that were previously seen but taken from different encoding pairs, or 3) pairs of novel images that were not previously seen during encoding. Participants were asked to indicate whether they had"seen together", "seen but not together", or"never seen" with the keys '1', '2', and '3'respectively. Participants were asked to make keyboard responses with their dominant hand. There was no time limit.

Associative retrieval blocks were intermixed with the perceptual recognition task, in which a series of novel or previously seen images were presented for 200 ms, after which participants were asked if they recognized the images (Yes/No responses; Fig.2). Per block, participants answered 9 associative retrieval image pairs and then answered 9 perceptual recognition images. Blocks alternated in an A-B-A-B format.

The order of the tACS conditions was counterbalanced across participants. Participants rested their heads in a fixed chin rest before testing began. The experiment lasted approximately 90 min per participant, which included the time required to attach the electrodes. Each participant underwent the sham and theta stimulation conditions on the same day, with approximately 30 min between the two stimulation periods.

2.3. tACS parameters

tACS was applied using a neuroConn DC-STIMULATOR TES device, using a concentric electrode setup for focal stimulation, which included an inner circular electrode(3 cm diameter)placed inside an outer electrode(11 cm diameter; Fig.3A). This setup has previously been used for focal stimulation during cognitive tasks (Bortoletto, Rodella, Salva-dor, Miranda,& Miniussi,2016; Gbadeyan, Steinhauser, McMahon, &Meinzer,2016; Schuhmann et al,2019). The center of the concentric setup was placed over scalp location P3, as identified by the 10–20 EEG electrode positioning system, as this site has been shown to have extensive connections with the hippocampal network (Freedberg et al., 2019; Hermiller et al., 2019; Wang et al., 2014). We used the SimNIBS software to simulate the spread of the electric field for the concentric setup when administered over P3. The simulation results showed an electric field with a focal spot of maximum field strength (0.296 V/m at 99.9th percentile) under the central electrode, which decreased towards the outer ring electrode (Fig. 3B). For visual comparison, Fig. 3C shows the T-map of hippocampal-cortical connectivity from a resting-state functional magnetic resonance imaging (rs-fMRI) study with forty-three participants (Tambini et al., 2018; obtained from ), in which a left parietal area of high hippocampal connectivity overlaps with our stimulation location.

For each participant, we used an EEG cap (www.easycap.de; without electrodes attached) to mark the position of P3, by the central and lateral distances between anatomical markers that are defined by the 10–20 EEG electrode positioning system on each participant’s head. We centered the cap on the head according to standard procedures and marked it on the scalp using a cosmetic, skin-friendly pencil. Each participant was examined for abrasions, inflammation, or small wounds on the scalp before electrode attachment. Participants with any skin abnormalities at or near the location of stimulation were excluded from the experiment before electrode placement. The conductive adhesive paste was used to optimize conductivity between the electrodes and scalp to below 10kΩ for each participant. We verified that paste for each of the two electrodes did not connect. Due to the adhesive paste, no headbands or straps were required to hold the electrodes in place throughout the experiment.

In the theta condition. the frequency of stimulation was set to 6 Hz for 5400 cycles (15min× 60 s×6 Hz)at an intensity of 2 mA(peak-to-peak). Participants were familiarised with the sensation of tACS in a short stimulation session to determine if the stimulation was tolerable. One participant indicated discomfort, and stimulation intensity was reduced for this participant to 1.8 mA. During the experiment, stimulation was applied for 15 min, starting 5 min before the encoding task. Participants were given a filler task (visual discrimination task) during the first 5 min of stimulation to occupy participants and ensure they were in a similar mental state before they began the associative encoding task. The ramp-up and ramp-down durations were set to 30 s in the theta condition. In the sham condition, stimulation lasted for 300 cycles (50 s)before ramping down to create an initially indistinguishable state to mask the conditions.

Due to the difference in continuous somatosensory sensations between the two tACS conditions. participants could theoretically discern the theta stimulation from the sham stimulation. To verify whether this was the case, participants were asked to verbally identify the differences in stimulation conditions posthoc at the end of the experiment. None of the participants informed us they were able to confidently distinguish the theta stimulation from the sham.

3. Data analysis

Statistical analysis was performed in JASP(JASPTeam,2018). Of the twenty participants, data of one participant was excluded for having incomplete data files as a result of an error in the experiment program code. Data of another participant were excluded after reporting suffering from fatigue before and during the experiment Of the remaining data-sets, trials with a clear delayed response (>2×SD)were excluded from the analysis, which constituted a loss of 4.03% on average of all trials.

We analyzed memory performance by calculating memory sensitivity (d, Stanislaw & Todorov,1999; Macmillan & Creelman,2005)using hit rates and false alarm rates. For the associative memory test, the hit rate was calculated as the proportion of fac-scene pairs shown during the encoding task, which were correctly identified as"seen as a pair". The false alarm rate was identified as the proportion of face-scene pairs that were wrongly judged to have been seen as a pair during the encoding task. For the perceptual memory test, hit rate was calculated as the proportion of previously shown single items that were correctly judged as seen before, and false alarm rates as the proportion of novel items that were erroneously judged as seen before. Statistical comparisons between tACS conditions were carried out using paired-sample t-tests, and repeated measures ANOVAs were used to test whether the within-subject effects correlated with covariates Age, Sex, and Order of tACS condition. Additionally, we performed paired sample t-tests and repeated measures ANOVA analyses on encoding response times (time from stimulus pair onset to button press)to determine whether the self-paced nature of encoding created a significant difference in performance between conditions and participants. Effects were deemed significant at a p-value of 0.05.

As our hypotheses included null effects (de Graaf & Sack, 2011), we calculated Bayes Factor to quantify the strength of the evidence in favor of no effect due to tACS, that is, BF1o <1(Wagenmakers,2007). Bayes Factor calculation was performed using JASP.

4. Results

4.1. Response time during encoding

During the encoding task, participants responded 2753.1 ms on average (SE= 213.8 ms) after anset of the picture pairs during theta-tACS.During sham-tACS, encoding response time was 2971.5(227.4)ms. Encoding response times did not significantly differ between the two tACS conditions(t(17)=-1.54,p= 0.14).We also tested whether response time variance across encoding trials of each participant differed between tACS conditions (mean [SE] theta-tACS = 978.8 [68.5],sham-tAGS=1060.4[96.2]),and found no significant difference (t(17)=-1.33,p=0.20).

4.2 Theta-based tACS effects on associative memory responses

Associative d'memory performance during theta-tACS (mean [SE]=1.32 [0.13]) was significantly lower than associative d'during sham-tACs(1.61 [0.16]; t(17)=-2.517,p= 0.022,Cohen's d=-0.59;see 互g. 4A）.A repeated-measures ANOVA with tACS （Theta， Sham） as within-subject factor showed that this effect did not interact with between-subject factors Age (p=0.23)or Sex(p=0.36).Wealso tested for carry-over effects of theta stimulation on the encoding during the subsequent sham stimulation session by using tACS Order(theta fol-lowed by sham vs.sham followed by theta)as between-subject factors, which did not significantly interact with the within-subject tACS effect an associative memory (p= 0.64).

To furtherinvestigatetheeffect of tACS on d',weconducted post-hoc comparisons for hit rate and false alarm rate between the two tACS conditions. The hit rate was significantly higher in the sham condition (67.0%[0.05])compared to the theta condition (53.7%[0.05])for associative memory (t(17)=-3.286,p=0.004, Cohen's d =-0.78). False alarm rates for associative memory performance revealed no sig-nificant difference between the sham(16.2%[0.02]) or theta condition (12.1%[0.02]:t(17)=-1.428,p=0.171). Thesefindings corroborated that theta-tACS impaired associative memory performance.

4.3. Theta-based tACS effect on perceptual item recognition

Next, we analyzed scores for the perceptual recognition test, which was performed on the same visual material as for the associative memory test. We found no significant difference in perceptual item recognition between the theta (1.56 [0.16]) and sham conditions(1.48 [0.10];t(17)=0.462,p=0.65;see Fig.4B).As wehypothesizednotACS effect on perceptual recognition, we calculated Bayes Factor(BF) for this comparison. We found that BF1o= 0.44, providing anecdotal evidence in favor of no effect.

4.4. Comparison between associative and perceptual memory performances

To analyze whether tACS affected associative and perceptual recognition memory differently, we conducted four further analyses. First, we correlated associative memory and perceptual recognition performance for the sham condition to determine whether baseline performance in the two tasks was related.This correlation was not significant (r=0.28,p=0.26). Calculation of the Bayes Factor showed that the evidence was in favor of performance of the two tasks not being related (BFy =0.52), This result fits our hypothesis that associative and perceptual memory tasks rely on different underlying neural mechanisms. Second, we compared d' of the sham condition of the two memory tasks, using a paired-samples t-test. and found that they were not significantly different (t(17)= 0.78,p=0.78 BFyo = 0.25).which indicates that the tasks were equally difficult. Third, we correlated the performance difference due to tACS(ie, d(Theta-d(Sham)between associative and perceptual recognition memory and found no significant correlation (r= 0.12,p= 0.63;BF10=0.33).Fourth, we tested whether the tACSinduced impairment of associative memory performance was larger than the (null) effect of tAGS on recognition memory performance. To model this interaction between tACS and task, we calculated a repeated measures model with tACS and Task as within-subject factors(and Age, Sex and Order as between-subject factors)and found a significant tACS× Task interaction effect (F(1,13)=7.68,p = 0.015,= 0.37).In all, these results indicate that tACS affected associative memory differently from recognition memory performance.

4.5. Comparison of encoding response times as an explanation of tACS effects

It could be possible that individual differences in encoding response times affected the tACS effect on memory, despite a lack of significant tACS effect on encoding response times. For example, participants with faster encoding times could have processed the stimuli less well, which could have made them more vulnerable to the effect of tACS. To investigate this issue, we re-calculated the repeated-measures ANOVAs for associative and recognition memory and the associative x recognition interaction.but now using the encoding response times averaged between the two tACS conditions as covariate, We found no significant effect of encoding response time for associative memory (p = 0.31), recognition memory (p = 0.25) or the associative × recognition inter-action (p = 0.97). To substantiate these null effects, we calculated the Bayes factor for the interaction of encoding response times with the repeated measure's effect of tACS on associative memory, recognition memory, and the associative × recognition interaction, and found that they were at or below/in all cases(for each respective model, BF10 was 0.35,0.37 and 0.13). These findings thus refute the suggestion that encoding times influenced the effect of tACS on memory performance.

5. Discussion

Our results demonstrate that theta tACS over parietal cortex during memory encoding impairs associative memory performance but not perceptual recognition memory of the same task materials. These findings are in line with repeated observations that theta oscillatory activity specifically contributes to associative memory encoding (a]sein et al.,1998; Buzsaki, 2002; Lega et al,2012; Burke et al,2013; Herweg et al,2020; Sauseng et al,2010). The finding that associative memory performance did not correlate with perceptual memory performance further corroborates the notion that the neural correlates of the two memory processes are dissociable. This is of further importance given that the site of parietal stimulation (P3)has been found to play a role in both associative memory and perceptual processing(Cabeza et al,201l; Culham & Kanwisher,2001; Gotlieb, 2007). That is, the neural correlates of the two tasks show spatial overlap in the parietal cortex. Yet, the dissociable effect of theta tACS on the two memory tasks suggests that theta oscillatory activity may functionally dissociate the two memory processes in the parietal cortex. Furthermore, this dissociation cannot be explained by a difference in task difficulty, as performance during the sham conditions of both tasks was similar.

Our finding of associative memory impairment after theta tACS contrasts several previous findings of theta tACSinduced enhanced memory performance. However, comparison between studies is hampered by the heterogeneity in task design and location of brain stimulation. For example, tACS-induced memory enhancement has been reported in working memory studies(Abellaned-Prezet al,2020 Alekseichuk et al.2016; Khader et al.2010), but these tasks arguably did not rely on associative processing. Further, the area of stimulation may be of particular importance due to the differential effects of theta power on different neuroanatomical sites (Long. Burle,&Kahana 2014). For example, the enhanced performance in the face-scene association task after tACSwas administered to the fusiform area(Lang et al, 2019)could have resulted from modulation of the encoding of facial features, rather than increased associative memory encoding. While theta oscillatory activity has been observed in these areas in humans and animals(Lee et al,2005; Raghavachari et al,2006; Sauseng et al, 2010), they may play a different role in memory processes. In a more recent study, Klinlk et al.(2020)found no change in associative memory performance in young participants after tACS over the ventrolateral prefrontal cortex, which further exemplifies the inconsistency in results.

Instead, we suggest that our findings are in line with previous studies showing that parietal cortical stimulation modulates neural activity in a hippocampal-cortical network that supports associative memory pro-cessing (Freedberg et al.,2019; Hermiller et al,2019; Tambini et al 2018; Wanget al,2014). More particularly, our findings are in line with the modulation of associative memory, but not recognition memory, after administration of TMS over the parietal cortex (Tambini et al.,2018). We extend these reports by showing that theta tACS can also modulate as asSociative processing within arguably the same hippocampal-cortical network It is possible that our finding of decreased, rather than enhanced, associative memory may indeed be related to the theta oscillatory pattern of brain stimulation.

An important consideration is that we stimulated at a fixed frequency for all participants, rather than at the individual theta frequency. Stimulating at the individual peak frequency would lead to enhanced neural entrainment, which could thereby enhance memory performance (Hanslmayr, Axmacher,& Inman,2019; Thut et al,2011). This notion is supported by a handful of previous tACS studies that stimulated at the individual theta frequency, identified using theta-gamma phase-amplitude-coupling during task execution, and found enhanced working memory performance after stimulation(Reinhart & Nguyen, 2019; Vosskuhl, Huster,& Hermann.,2015). However, other tACS studies showing working memory enhancement used a fixed stimulation frequency(Jausovec& Jausovec,2014; Lang et al.,2019), suggesting that stimulation at individual theta frequency is not necessary to modulate memory performance. More generally, several studies showed that stimulation at the individual alpha frequency did not consistently enhance neural entrainment (Vossen et al,2015; Stecher & Herrmann, 2018), suggesting that the relation between brain stimulation, neural entrainment, and cognitive performance may not be straightforward. Finally, there is evidence that associative encoding in humans is supported by two frequency bands at the extremes of the theta range (3-8 Hz), rather than varying around the center of the theta range(Lega et al, 2012(Wolinski et al.,2018)).In this case, electric oscillatory stimulation at the center of the theta range would lead to theta de-synchronization or shifts in the phase of the endogenous frequencies, which could disrupt memory processing(Alekseichuk et al.201Z; Polania, Nitsche. Korman Batsikadze,& Paulus,2012). We argue that our findings appear best in line with an account of tACS-induced phase-desynchronization that impairs associative memory but does not affect perceptual recognition performance

The design of our study poses a few limitations to our inferences. We did not include functional brain imaging simultaneous to tACS administration, which prevents us from verifying the effect of parietal tACS on the hippocampal-cortical network. Also, we did not measure EEG simultaneous to tACS administration, which means that we could not empirically test our postulation of theta desynchronization in associative memory. Finally, we did not use a control frequency or location to verify the specificity of theta stimulation of the parietal cortex. We are planning to address these open issues in future studies.

In conclusion, our study shows that theta tACS affected the encoding of associative memory while sparing perceptual recognition. in line with the hypothesis that theta oscillations play a critical role in associative memory formation in parietal-hippocampal interactions. These results align with current neurobiological models of associative memory and encourage further study using non-invasive electric brain stimulation to modulate the oscillatory activity of associative memory brain networks for current memory research.