Neurobiology of Learning and Memory 145 (2017) 165–171

Contents lists available at ScienceDirect

Neurobiology of Learning and Memory

journal homepage: www.elsevier.com/locate/ynlme

Enhancing early consolidation of human episodic memory by theta EEG MARK

neurofeedback
⁎
Roman Rozengurta, Limor Shtootsb, Aviv Sheriffb, Ofir Sadkab, Daniel A. Levyb,
a
Sagol Department of Neurobiology, University of Haifa, Haifa 3498838, Israel
b
Baruch Ivcher School of Psychology, Interdisciplinary Center Herzliya, Herzliya 4610101, Israel

A R T I C L E I N F O A B S T R A C T

Keywords: Consolidation of newly formed memories is readily disrupted, but can it be enhanced? Given the prominent role
Memory of hippocampal theta oscillations in memory formation and retrieval, we hypothesized that upregulating theta
Consolidation power during early stages of consolidation might benefit memory stability and persistence. We used EEG neu-
EEG rofeedback to enable participants to selectively increase theta power in their EEG spectra following episodic
Neurofeedback
memory encoding, while other participants engaged in low beta-focused neurofeedback or passively viewed a
Theta
neutral nature movie. Free recall assessments immediately following the interventions, 24 h later and 7 d later
Recall
all indicated benefit to memory of theta neurofeedback, relative to low beta neurofeedback or passive movie-
viewing control conditions. The degree of benefit to memory was correlated with the extent of theta power
modulation, but not with other spectral changes. Theta enhancement may provide optimal conditions for sta-
bilization of new hippocampus-dependent memories.

1. Introduction compared to participants who selectively increased low beta power, or

passive controls. In addition, as in sleep consolidation studies
The initial formation of a memory trace is followed by a time in- (Rasch & Born, 2013), theta NFB led not only to protection of learning
terval during which that trace may become consolidated (Dudai, 2004, from decay, but to offline improvement relative to best prior perfor-
2012). Such consolidation has been demonstrated to be vital for the mance. Since the initial stages of procedural memory involve hippo-
persistence of all types of long-term memory (McGaugh, 2000; campal mechanisms (Albouy, King, Maquet, & Doyon, 2013), we set out
Rasch & Born, 2013). Early stages of consolidation involve the stabili- to investigate whether similar theta EEG NFB enhancement of early
zation of structural and functional synaptic and other neural changes consolidation would be found for episodic memory.
engendered by a learning event (Bailey, Kandel, & Harris, 2015; Healthy young adult participants viewed 30 object pictures, each
Redondo & Morris, 2011). Pharmacological interventions such as post- presented for 3 s, and performed an immediate free recall test, re-
learning administration of protein synthesis inhibitors block synaptic peating the process twice more to establish a learning curve.
consolidation, while other substances can potentiate synaptic con- Participants then engaged in NFB modulation of theta power, or in one
solidation (Rosenberg et al., 2014). Finding non-invasive interventions of two control conditions: low beta NFB (an active control condition
that might benefit consolidation in humans, especially of declarative matching the experience of engaging in NFB procedures), or movie
memory for facts and events, seems to be a desideratum. A novel viewing (a passive control condition modelling how memory would be
method of optimizing consolidation is suggested by findings regarding affected under ecological activity conditions), all for 30 min. All par-
the important role of brain oscillations, especially theta rhythms, in ticipants then took the fourth recall test. Two additional follow-up re-
mnemonic processes (Burke et al., 2014; Fell & Axmacher, 2011; call tests were administered: 24 h later, to determine the interaction of
Hsieh & Ranganath, 2014). This raises the possibility that post-learning NFB training and sleep on consolidation, and 1 w later, to assess the
theta rhythm modulation might be a method of promoting consolida- stability of consolidation effects engendered by NFB. We found notable
tion. In recent research (Reiner, Rozengurt, & Barnea, 2014; Rozengurt, subsequent memory benefits for the Theta NFB group at all three as-
Barnea, Uchida, & Levy, 2016), we found significantly greater post- sessment points relative to the active and passive control conditions.
training procedural memory performance following neurofeedback Such theta NFB effects on this episodic memory task suggest that theta
(NFB) in a group of participants who selectively increased theta power, enhancement may provide a method for potentiating hippocampus-

⁎
Corresponding author.
E-mail address: daniel.levy@idc.ac.il (D.A. Levy).

http://dx.doi.org/10.1016/j.nlm.2017.10.005
Received 30 July 2017; Received in revised form 4 October 2017; Accepted 7 October 2017
Available online 10 October 2017
1074-7427/ © 2017 Elsevier Inc. All rights reserved.
R. Rozengurt et al. Neurobiology of Learning and Memory 145 (2017) 165–171

dependent memory. interest: theta (4–8 Hz), low beta (15–18 Hz) and high beta (18–22 Hz).
For the recording of resting stage baseline, all participants sat quietly
2. Materials and methods with their eyes open for 4 min. The Fz electrode was employed to
provide real-time NFB, as in our previous studies (Reiner et al., 2014;
2.1. Participants Rozengurt et al., 2016), and in accordance with additional studies
(Egner & Gruzelier, 2003, 2004). Notably, as frontal midline theta has
75 volunteers (45 females; mean age 29.6 y, SD 7.1 y) took part in been implicated in episodic memory encoding and retrieval processes
the study in return for payment and/or academic requirement credit. (Hsieh & Ranganath, 2014), this was seemingly the optimal location for
They reported no history of psychiatric or neurological disorders, nor both conditions in this task. The NFB program was set to provide real-
chronic use of medication. All participants reported having a minimum time positive feedback using a visual signal displayed on the computer
of 6 h of nocturnal sleep in the night before and during the week of the screen. This took the form of a bar display, in which the height of a
experiment, and no physiological sleep disruptions. Informed consent vertical green bar was determined by EEG target-band power ratio (i.e.,
was obtained from all participants for a protocol approved by the theta/low beta or low beta/theta). A horizontal criterion line was
Institutional Review Board of the Interdisciplinary Center Herzliya. The presented overlying the bar, representing the goal band power ratio
75 participants were randomly assigned to Theta group (17 F, 8 M, level. Participants were instructed to keep the bar above the criterion
mean age 30.9 y, SD 7.7 y), (low) Beta active control group (14 F, 11 M, line as much as possible. It was explained to them that bar height is
mean age 28.6 y, SD 7.0 y), or Movie passive control group (14 F, 11 M, determined by the character of their EEG, and that they must learn to
mean age 28.8 y, SD 6.5 y). control it by maintaining whatever mental state provides them with
positive feedback.
2.2. Experimental design To directly contrast the two active NFB conditions, we provided
participants with real-time feedback based on their theta/low beta
After random assignment to one of the three experimental groups, ratio, as is common in clinical application of NFB in putative treatment
participants were prepared for EEG recording. Resting baseline EEG of attention deficits (e.g., Bluschke, Broschwitz, Kohl,
was recorded for 4 min. Participants were given task instructions and Roessner, & Beste, 2016; Van Doren et al., 2017). Theta group partici-
viewed 30 object pictures (e.g., spoon, door, leaf, hammer, bus) each pants received positive feedback for increasing theta/low beta ratio;
presented for 3 s, followed by a free recall test which took approxi- Beta group participants received positive feedback for increasing their
mately 5 min. Next, participants viewed the 30 objects again in a dif- low beta/theta ratio. Positive bar rise feedback was only provided if in
ferent random order, followed by a second recall test, which was fol- addition to increasing their target band-power ratio (theta/low beta or
lowed by a third study-test cycle. Participants then engaged in NFB for a low beta/theta), participants did not increase high beta power (to
period of 30 min (three 10-min sessions with short rest breaks) in the minimize motion artifacts). Initially, the WINEEG software auto-
two NFB groups, or 30 min of neutral nature movie viewing in the matically adjusted the threshold to be 90% of the participant’s mean
passive control group. Following the NFB/movie viewing session, par- target band-power ratio during the first 2 min of the NFB session.
ticipants took the fourth recall test. 24 h after the initial session, par- Participants generally improve in their ability to increase target band
ticipants took the fifth recall text, via a phone call (∼50% of the par- power ratio, so this adjustment continued dynamically until the end of
ticipants) or by returning to the lab. This follow-up recall test was the NFB session (Ros et al., 2013). Band power measures for the elec-
intended to determine the interaction of NFB training and sleep on trophysiological and psychophysiological analyses were averaged
consolidation of episodic memory. One week after the initial session, across the entire 30 min NFB session for the Fz electrode. For offline
participants took the sixth and final recall test (via phone call or in lab, analysis, independent component analysis instantiated in WINEEG
as before), intended to assess the stability of episodic memory con- software was used for removing blink artifacts. In order to remove other
solidation effects engendered by NFB. Six participants did not complete artifacts from the EEG recording, a comparison between the signal
the post-1-week free recall test (3 participants from the Beta experi- parameters and the threshold values was used, based on several criteria:
mental group, and 3 participants from the Theta experimental group). deviation of the potentials from the isoline exceeding 75 μV, deviation
Participants were unaware of the goals of the experiment. of the low frequency (0–1 Hz) signal component exceeding 50 μV, and
deviation of the high frequency (20–35 Hz) signal component ex-
2.3. Neurofeedback protocol ceeding 35 μV. Following artifact removal, mean EEG band power for
each relevant frequency for each participant was derived from the
NFB was performed and recorded using Mitsar-202 EEG system WINEEG software (using fast Fourier analysis), to be subjected to
(Mitsar, St. Petersburg, Russia). Nineteen silver-chloride electrodes analyses of variance (ANOVA) to characterize group differences as well
were placed on the scalp using an elastic cap, according the standard as relationships between frequency change and episodic memory per-
10–20 system, at the following sites: Fp1, Fp2, F7, F3, Fz, F4, F8, T3, formance. The non-NFB control group (“Movie group”) had their EEG
C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1, and O2. The input signals were recorded passively at baseline and during movie viewing using a Bio-
recorded in a monopolar montage, with linked earlobe reference elec- Semi Active Two system (BioSemi, Amsterdam, The Netherlands) from
trodes. The ground electrode was placed on the forehead. Impedance 64 electrodes mounted in an elastic cap according to the extended
was kept below 5 kΩ. EEG was amplified, band-pass filtered at 10–20 system. The on-line filter settings of the EEG amplifiers were
0.5–30 Hz, and sampled at the rate of 250 Hz. During online NFB, a 0.16–100 Hz. EEG was continuously sampled at 2048 Hz. The BioSemi
100 μV artifact-rejection threshold was used to interrupt NFB during system enables very high precision EEG recording, and was used in this
eye and body movements that produced gross EEG fluctuations. The group since the proprietary interface with NFB software was not re-
spectral distribution of the ongoing oscillatory brain activity was de- quired. For offline analysis, 19 electrodes corresponding to those
rived from EEG in real-time, using WINEEG software (Mitsar), and comprising the Mitsar montage were selected, and the data was
stored for further offline analysis. This program, running on a portable downsampled to 256 Hz and converted to the WinEEG format, which
computer, received digitized EEG data from all channels. Mean absolute was used for all further offline pre-processing and analysis as described
and relative EEG spectral power for each bandwidth was calculated above.
using fast Fourier analysis, with the following parameters: epoch
duration of 4096 ms, epoch overlapping of 50%, time smoothing with 2.4. Statistical analysis
the Hann window. The spectral characteristics were computed for fre-
quencies ranging from 1–22 Hz, and analyzed for frequency bands of NFB effects were examined by repeated measures ANOVA (Figs. 1,

166
R. Rozengurt et al. Neurobiology of Learning and Memory 145 (2017) 165–171

Fig. 1. NFB modulation of (A) EEG theta/low beta ratio, (B) theta power, and (C) low beta power. P-values are provided for group differences in these metrics between baseline and NFB/
movie means. Error bars indicate ± SEM; *, P < .05; ***, P < .001; NS, non-significant.

Fig. 4. Percent performance improvement immediately following NFB compared to im-

Fig. 2. Free recall scores during each stage of the experiment, for Theta NFB, low Beta mediate pre-NFB training measures, as a function of the degree of change in theta/low
NFB, and Movie groups. Error bars indicate ± SEM; NS, non-significant; **, P < .01; ***, beta ratio for all NFB participants across Theta and Beta groups (n = 50).
P < .001.
3. Results
2, 5, 6) and regression analyses (Figs. 3 and 4). Data are presented as
percent band-power values or, when mentioned, as band-power ratios. 3.1. NFB modulation of EEG power ratio
All statistically significant analyses were set at p < .05, error bars
represent SEM, and n represent number of participants. Statistical cal- As described in the Methods, feedback was given to both Theta and
culations were performed with SPSS software. Beta groups using theta/low beta ratio, which therefore provides a
synoptic index of modulatory success across NFB groups, and a parallel
indication of incidental frequency band change during movie viewing

Fig. 3. Percent performance immediately improvement following NFB compared to immediate pre-NFB training measures, as a function of the degree of change in (A) theta power during
NFB compared to baseline, for n = 25 participants of the Theta group; and in (B) low beta power during NFB compared to baseline, for n = 25 participants of the Beta group.

167
R. Rozengurt et al. Neurobiology of Learning and Memory 145 (2017) 165–171

in the passive control condition. For theta/low beta ratio (Fig. 1a), Movie group: −14.7%); these group differences were significant,
repeated measures ANOVAs, with factors of group (Theta, Beta, Movie) F(2,72) = 47.15, p < .001, with the Theta group differing from both
and stage (resting baseline, NFB/movie), indicated no main effect of other groups, p < .001, and the Beta group differing from the Movie
stage, F(1,72) = 1.57, p = .21, no main effect of group, F(2,72) = 1.43, group, p < .01. Similarly, at post-1 w test, the Theta group exhibited
p = .25, but a significant group X stage interaction, F(2,72) = 10.86, further improvement (to +14.0%), while the other groups exhibited
p < .001. We examined the interaction by conducting separate re- further performance decline (Beta group: to −8.9%, Movie group: to
peated measures ANOVAs for each group, with theta/low beta ratio −24.1%), these group differences were significant, F(2,66) = 73.83,
change as the dependent variable. This revealed that theta/low beta p < .001, with the Theta group differing from both other groups,
ratio increased (+27.9%) significantly in the Theta group, p < .001, and the Beta group differing from the Movie group,
F(1,24) = 18.83, p < .001, but decreased (−4.3%) non-significantly in p < .001. Thus, theta-upregulating NFB had a noticeable impact on
the Beta group, F(1,24) = 2.59, p = 0.12, and in the Movie group episodic memory performance, both immediately and in follow-up
(−5.7%), F(1,24) = 0.67. These group differences were not a function of testing.
initial variance in EEG band power, as baseline theta/low beta ratio did Another perspective on the impact of theta upregulation by NFB on
not significantly differ between groups, F(2,72) = 0.07. episodic memory is provided by calculating the correlation between the
We further examined the effects of the intervention on each band percent target-band power change during NFB and immediate post-NFB
separately in the three experimental groups. For theta power (Fig. 1b), memory performance improvement. For the Theta group (Fig. 3a),
group differences in baseline power were not significant, F(2,72) < 1.0. there was a significant correlation between percent theta change during
Regarding the changes during NFB, repeated-measures ANOVA with NFB and the improvement in immediate post-NFB successful recall,
factors of stage (repeated) and group (between-Ss) yielded a significant Pearson’s r(25) = 0.609, p = .001 (2-tailed), while for the Beta group
main effect of stage, F(1, 72) = 10.64, p < 0.01, no main effect of the relationship was not significant, r(25) = 0.240, p = .249 (2-tailed),
group, F(2, 72) < 1.0, and a significant group X stage interaction, F(2, nor was it for the Movie group r(25) = −0.279, p = .177 (2-tailed). In
72) = 4.47, p < 0.02. We examined the interaction by conducting re- contrast, the correlation between percent low beta change during NFB
peated measures ANOVA for each group separately, which revealed and the improvement in immediate post-NFB successful recall for the
that the difference between baseline and NFB theta power was sig- Beta group (Fig. 3b) was not significant, r(25) = 0.116, p = .581 (2-
nificant for the Theta group (+16.2%), F(1, 24) = 19.64, p < 0.001, tailed); nor was it for the Theta group, r(25) = −0.052, p = .805 (2-
but not for the Beta group (−1.8%), F < 1.0, nor for the Movie group tailed), nor for the Movie group r(25) = −0.344, p = .092 (2-tailed).
(+9.5%), F(1, 24) = 3.58, p = 0.07. This indication of specificity of theta upregulation on off-line im-
For low beta power (Fig. 1c), group differences in baseline power provement in recall success may also be seen in examination of the
were similarly not significant, F(2,72) < 1.0. Regarding the changes correlation between theta/low beta ratio change and improvement in
during NFB, repeated-measures ANOVA indicated that the main effect immediate post-NFB successful recall, which for the Theta group was
of stage was not significant, F(1, 72) = 1.63, p = .20, nor was the main significant, r(25) = 0.439, p = .028 (2-tailed), while not for the Beta
effect of group, F(2, 72) < 1.0, but there was a significant group X stage group, r(25) = 0.130, p = .434 (2-tailed), or the Movie control group
interaction, F(2, 72) = 4.07, p < .05. We examined the interaction by r(25) = 0.064, p = .761 (2-tailed). This relationship is portrayed sy-
conducting repeated measures ANOVA for each group separately, noptically in Fig. 4 for the Theta and Beta active intervention groups.
which revealed that for the Theta group, the difference between base- Taken together, these correlations clarify that this effect was driven
line and NFB low beta power (−8.0%) was not significant, F(1, solely by theta upregulation in the Theta group, and that it cannot be
24) = 1.93, p = .18, nor was it significant for the Beta group (+4.3%), attributed to task success or to the experience of positive feedback.
F(1, 24) < 1.0. In contrast, in the Movie group low beta power increase
(+13.7%) was significant, F(1, 24) = 6.50, p < .02. The Beta group 3.3. Analysis of NFB-effective participants’ performance
might have experienced difficulty in upregulating low beta power since,
as noted above, we implemented control of movement-related activity While Theta group participants showed significant NFB modulation
by withholding positive feedback for target low beta when high beta of their relative theta band activity and theta/low beta ratio, Beta group
was increased along with it. In contrast, the movie group was un- participants’ low beta band power and theta/low beta ratio during NFB
constrained, and attention paid to the movie could yield increased low were not significantly different than baseline. This raises the possibility
beta as observed. We addressed the Beta group’s modulation difficulties that it was band power regulation success, with the concomitant more
in additional analyses reported below. positive feedback enjoyed by successful regulators, rather than speci-
fically theta band power that was responsible for the Theta group’s
3.2. NFB effects on episodic memory improvement superiority in post-intervention free recall. To rule out that possibility,
we selected from each NFB group only those participants who increased
NFB/movie effects on free recall performance for the three experi- target band power ratio by at least 5% relative to baseline (n = 19 for
mental groups are presented in Fig. 2. We first determined by one-way the Theta group and n = 13 for the Beta group; Fig. 5). These Theta
ANOVAs that performance (expressed as number of correctly recalled group participants exhibited a 29.4% mean increase in target theta
items) in each of the initial, pre-NFB/movie study-test trials did not power between baseline and NFB, F(1,18) = 38.95, p < .001, and these
differ significantly between groups, all Fs < 1.0. This indicates that Beta group participants exhibited a 14.7% increase in target low beta
later effects of NFB cannot be attributed to prior group differences in power, F(1,12) = 20.69, p = .001. Furthermore, these effects were ex-
baseline memory ability or initial learning. pressed in NFB changes relative to baseline in theta/low beta ratio, with
We then examined the impact of NFB/movie intervention on sub- the Theta group exhibiting an increase of 29.3%, F(1,18) = 10.49,
sequent free recall trials. For the immediate post-NFB/movie test, the p < .01, and the Beta group exhibiting a decrease of −14.5% (i.e., a
Theta group exhibited improved free recall (+8.8%), while decline was relatively greater increase in low beta than in theta), F(1,12) = 5.85, p
observed in the Beta group (−2.9%) and Movie group (−4.2%); these = .05.
group differences were significant, F(2,72) = 19.51, p < .001, with We then conducted the same analyses of free recall performance
Bonferroni-corrected comparisons indicating that the Theta group dif- comparing these groups as was done for the entire sample (Fig. 6).
fered from both other groups, p < .001, but that they did not differ These analyses revealed that the two groups did not differ in any of the
from each other, p = .59. For the post-24 h test, the Theta group pre-NFB study-test free recall trials, all Fs < 1.0. However, these
showed further improvement (to +11.2%), in contrast to the other groups differed in percent post-NFB free recall performance change
groups that exhibited further performance decline (Beta group: −4.5%, relative to final pre-NFB score at all three test points, all Fs > 49.5, all

168
R. Rozengurt et al. Neurobiology of Learning and Memory 145 (2017) 165–171

Fig. 5. Group differences in mean theta power, low beta power, and theta/low beta power ratio changes between baseline and NFB, for the sub-group of participants effective in
modulating their target band power (n = 19 participants in Theta group, n = 13 participants in Beta group). Error bars indicate ± SEM; **, P < .01; ***, P < .001; NS, non-significant.

As mentioned in the Materials and Methods section, because of lo-

gistical reasons not all participants were able to return to the lab for the
24 h post-NFB/Movie and the 1 w post-NFB/Movie recall tests. We
therefore arranged that half of the participants in each group return to
the lab, and the other half of each group was tested by telephone. There
were no significant differences in performance between participants in
these two conditions, both within each NFB/Movie group and across
groups, all Ps > .15. Data from all participants were therefore col-
lapsed across follow-up test method for all reported analyses.

4. Discussion

In the current study, we demonstrate enhancement of free recall

performance by theta NFB implemented between encoding and re-
trieval stages of an episodic memory task. This benefit seemingly re-
presents a consolidation-related process. This finding resonates with
recent reports regarding the importance of early post-encoding activa-
tion, asserted to reflect consolidation-related activity, for the persis-
tence of episodic memories (Dudai, Karni, & Born, 2015). Given the
immediacy of theta NFB impact on episodic memory, its effects are
Fig. 6. Free recall scores during each stage of the experiment, for Theta NFB and Beta
NFB effective modulator subgroups (n = 19 participants in Theta group, n = 13 parti- likely attributable to synaptic consolidation (cellular-molecular pro-
cipants in Beta group). Error bars indicate ± SEM; NS, non-significant; ***, P < .001. cesses transpiring shortly after new memory formation) rather than
systems consolidation (shifts in the locus of mnemonic indexing from
hippocampal to neocortical loci; Dudai et al., 2015), though both pro-
ps < .001. Furthermore, there was a significant correlation between
cesses might have been affected by the intervention.
percent target-band power upregulation during NFB and percent post-
This study employed an NFB intervention, in which participants
NFB free recall improvement for the Theta effective regulators group,
engendered the upregulation of target-band power themselves, using
r(19) = 0.780, p < .001, but not for the Beta effective regulators group,
on-line feedback. It might therefore be argued that the improvement in
r(13) = 0.367, p = .218. These analyses indicate that the Theta group
free recall performance seen in the Theta group resulted not from
advantage cannot be attributed to NFB success and its concomitant
physiological processes associated with increased theta power, but to
greater amount of positive feedback in itself, but rather reflects a pro-
the fact that they were successful in their appointed task, providing
cess specific to theta upregulation.
them with strong positive feedback that might boost their memory
We employed the Fz electrode to provide real-time feedback for
performance. By this account, though, we would expect to find a par-
theoretical reasons noted in the Methods section. It might be wondered
allel benefit in the Beta group; in fact, low beta regulation success in the
whether using a different electrode might have yielded different find-
Beta group overall, and even in the more selective sample of successful
ings. However, in our prior research (Rozengurt et al., 2016), we found
regulators, was not correlated with subsequent memory benefits. In
a high degree of correlation between theta/beta ratio change yielded by
contrast, there was a very strong correlation between the degree of
NFB at Fz and the degree of NFB change at other electrodes. We re-
success in theta enhancement and benefit to memory, specific to the
peated that analysis for the present dataset (Fig. 7), for n = 44 parti-
Theta group. Similarly, the Theta group’s memory benefit might theo-
cipants for whom we had the raw data available. This revealed a
retically be attributed solely to the group differences in the initial
comparable degree of correlation across electrodes (and conceivably,
suggestions given to participants regarding the strategies they might
across cortical areas as well, although volume conduction could be
use to succeed in band power regulation (relaxation for theta and
responsible for the scalp effects). This indicates that while Fz appears to
concentration for low beta). However, this would not account for the
be a relevant reporting electrode for memory consolidation by NFB, the
strong correlation between the degree of theta upregulation and sub-
effects of NFB are global, and other reporting electrodes might function
sequent memory benefit within the Theta group.
similarly well.
The movie group was included in this study primarily to provide an

169
R. Rozengurt et al. Neurobiology of Learning and Memory 145 (2017) 165–171

**
Fig. 7. Pearson’s R correlations between NFB theta/beta ratio change at Fz and NFB change at other recording electrodes, for n = 44 participants. , P < .01.

indication of what would happen to the memory traces formed in the competition filtering has been associated with sleep consolidation
initial learning trials during a controlled period of ongoing cognitive- (Albouy et al., 2013), NFB theta potentiation during waking could also
perceptual activity, modelling ecological conditions of ongoing ex- enhance such consolidation processes. It is noteworthy that, as in our
perience. The data regarding the Movie group indicates that under the prior studies of NFB effects on motor sequence learning (Reiner et al.,
particular conditions of the experiment, without the benefits to memory 2014; Rozengurt et al., 2016), theta NFB led not only to preservation
afforded by theta NFB, recall performance declines over time. While the but to off-line improvement in recall, which this mechanism might have
Beta group did not differ from the Movie group in recall success in the effected.
immediate post-intervention test, it is the case that they showed better It should also be noted that an alternative interpretation of these
memory at 24 h and 1 w delay, which is likely attributable to the lesser findings is that rather than reflecting consolidation processes, the
amount of specific retroactive interference for both NFB groups relative benefit exhibited by the Theta group resulted from theta oscillatory
to the movie viewers, which seemingly impacted on later stability of the activity and/or the mental states engendering it having facilitated re-
initial learning. There does not seem to be evidence from these data that trieval in the immediately following test. That initial boost to retrieval
low beta NFB provides any specific memory benefits. might then in turn have facilitated the subsequent free recall tests. The
We have considered several mechanisms by which theta upregula- present data do not allow adjudicating between these interpretations,
tion might benefit synaptic consolidation (Rozengurt et al., 2016). One which might be assessed in future studies in which the initial post-NFB
possibility is that theta rhythms differentially favor waking replay of recall test is conducted following a delay of several hours, during which
firing patterns in neuronal ensembles activated in encoding (Carr, time the acute effects of theta activity state might dissipate.
Jadhav, & Frank, 2011; Miller et al., 2013). Relevant to scalp-based NFB Relatedly, we have framed these findings in terms of consolidation
as implemented in the current study, cortico-hippocampal theta co- processes, on the most basic level since the intervention occurs after
herence has been linked to replay strength and subsequent consolida- encoding but before retrieval. The consolidation period in a broad sense
tion (Benchenane et al., 2010). Another candidate process in which is the locus of many memory-related processes: decay, non-specific
theta might be facilitatory is the resetting of synapses that have not interference, and forgetting, as well as stabilization and sharpening of
crossed a threshold for Hebbian plasticity back to baseline activity le- representations. Future research should examine to which of these
vels, thus sharpening memory traces by increasing the representational processes theta rhythm modulation might be relevant. For example,
signal-to-noise ratio. Norman and colleagues have proposed a biased combining EEG NFB with real-time fMRI could allow using re-
competition model, in which cycling between high and low inhibition presentational similarity analysis to examine whether the representa-
states associated with different theta phases leads to the strengthening tional patterns of the stimuli at encoding are more strongly represented
of effective memories by long-term potentiation, and weakening of the during neurofeedback in the theta than in the low beta condition.
competitors by long-term depression (Norman, Newman, & Detre, A number of limitations of this study should be mentioned. NFB
2007). The net effect would be reduction of a memory signal-to-noise success was seemingly challenging for some Beta group participants, as
ratio and better overall memory for target items. Although such biased we implemented control of movement-related activity by withholding

170
R. Rozengurt et al. Neurobiology of Learning and Memory 145 (2017) 165–171

positive feedback for target low beta when high beta was increased Dudai, Y. (2012). The restless engram: Consolidations never end. Annual Review of
Neuroscience, 35, 227–247.
along with it (Fig. 1c). However, theta NFB effects remained robust Dudai, Y., Karni, A., & Born, J. (2015). The consolidation and transformation of memory.
when examining effective modulators (Figs. 5 and 6). Additionally, Neuron, 88, 20–32.
recent research indicates that human hippocampal memory-related Egner, T., & Gruzelier, J. H. (2003). EEG biofeedback of low beta band components:
Frequency-specific effects on variables of attention and event-related brain poten-
theta activity may be focused at lower frequencies (∼3 Hz) than the tials. NeuroReport, 14, 1221–1224.
classic theta band (4–8 Hz) we employed (Jacobs, 2013; Lega, Burke, Egner, T., & Gruzelier, J. H. (2004). Ecological validity of neurofeedback: Modulation of
Jacobs, & Kahana, 2016; Pastötter & Bäuml, 2014). Individual band- slow wave EEG enhances musical performance. Clinical Neurophysiology, 115,
131–139.
frequency tailoring (as done for alpha power by Escolano et al., 2014) Escolano, C., Navarro-Gil, M., Garcia-Campayo, J., Congedo, M., De Ridder, D., &
may enable better specificity. Minguez, J. (2014). A controlled study on the cognitive effect of alpha neurofeedback
Our study is the first to indicate that theta rhythm modulation may training in patients with major depressive disorder. Frontiers in Behavioral
Neuroscience, 8, 296.
serve as a method of enhancing early consolidation of episodic memory.
Fell, J., & Axmacher, N. (2011). The role of phase synchronization in memory processes.
Follow-up studies are required to examine the generalizability of the Nature Reviews Neuroscience, 12, 105–118.
effect to additional aspects of episodic memory and semantic learning, Hsieh, L. T., & Ranganath, C. (2014). Frontal midline theta oscillations during working
and to explore whether such interventions may provide effective pro- memory maintenance and episodic encoding and retrieval. Neuroimage, 85, 721–729.
Jacobs, J. (2013). Hippocampal theta oscillations are slower in humans than in rodents:
tection from retroactive interference, as would be expected from con- Implications for models of spatial navigation and memory. Philosophical Transactions
solidation enhancement. of the Royal Society of London. Series B, Biological Sciences, 369, 1635.
Lega, B., Burke, J., Jacobs, J., & Kahana, M. J. (2016). Slow-theta-to-gamma phase-am-
plitude coupling in human hippocampus supports the formation of new episodic
Acknowledgements memories. Cerebral Cortex, 26, 268–278.
McGaugh, J. L. (2000). Memory - a century of consolidation. Science, 287, 248–251.
The authors declare no competing financial interests. Miller, J. F., Neufang, M., Solway, A., Brandt, A., Trippel, M., Mader, I., ... Schulze-
Bonhage, A. (2013). Neural activity in human hippocampal formation reveals the
spatial context of retrieved memories. Science, 342, 1111–1114.
References Norman, K. A., Newman, E. L., & Detre, G. (2007). A neural network model of retrieval-
induced forgetting. Psychological Review, 114, 887.
Pastötter, B., & Bäuml, K. H. (2014). Distinct slow and fast cortical theta dynamics in
Albouy, G., King, B. R., Maquet, P., & Doyon, J. (2013). Hippocampus and striatum:
episodic memory retrieval. Neuroimage, 94, 155–161.
Dynamics and interaction during acquisition and sleep-related motor sequence
Rasch, B., & Born, J. (2013). About sleep's role in memory. Physiological Reviews, 93,
memory consolidation. Hippocampus, 23, 985–1004.
681–766.
Bailey, C. H., Kandel, E. R., & Harris, K. M. (2015). Structural components of synaptic
Redondo, R. L., & Morris, R. G. (2011). Making memories last: The synaptic tagging and
plasticity and memory consolidation. Cold Spring Harbor Perspectives in Biology, 7,
capture hypothesis. Nature Reviews Neuroscience, 12, 17–30.
a021758.
Reiner, M., Rozengurt, R., & Barnea, A. (2014). Better than sleep: Theta neurofeedback
Benchenane, K., Peyrache, A., Khamassi, M., Tierney, P. L., Gioanni, Y., Battaglia, F. P., &
training accelerates memory consolidation. Biological Psychology, 95, 45–53.
Wiener, S. I. (2010). Coherent theta oscillations and reorganization of spike timing in
Ros, T., Théberge, J., Frewen, P. A., Kluetsch, R., Densmore, M., Calhoun, V. D., & Lanius,
the hippocampal-prefrontal network upon learning. Neuron, 66, 921–936.
R. A. (2013). Mind over chatter: Plastic up-regulation of the fMRI salience network
Bluschke, A., Broschwitz, F., Kohl, S., Roessner, V., & Beste, C. (2016). The neuronal
directly after EEG neurofeedback. NeuroImage, 65, 324–335.
mechanisms underlying improvement of impulsivity in ADHD by theta/beta neuro-
Rosenberg, T., Gal-Ben-Ari, S., Dieterich, D. C., Kreutz, M. R., Ziv, N. E., Gundelfinger, E.
feedback. Scientific Reports, 6 31178–31178.
D., & Rosenblum, K. (2014). The roles of protein expression in synaptic plasticity and
Burke, J. F., Sharan, A. D., Sperling, M. R., Ramayya, A. G., Evans, J. J., Healey, M. K., ...
memory consolidation. Frontiers in Molecular Neuroscience, 7, 86.
Kahana, M. J. (2014). Theta and high-frequency activity mark spontaneous recall of
Rozengurt, R., Barnea, A., Uchida, S., & Levy, D. A. (2016). Theta EEG neurofeedback
episodic memories. Journal of Neuroscience, 34, 11355–11365.
benefits early consolidation of motor sequence learning. Psychophysiology, 53,
Carr, M. F., Jadhav, S. P., & Frank, L. M. (2011). Hippocampal replay in the awake state:
965–973.
A potential substrate for memory consolidation and retrieval. Nature Neuroscience, 14,
Van Doren, J., Heinrich, H., Bezold, M., Reuter, N., Kratz, O., Horndasch, S., ... Studer, P.
147–153.
(2017). Theta/beta neurofeedback in children with ADHD: Feasibility of a short-term
Dudai, Y. (2004). The neurobiology of consolidations, or, how stable is the engram?
setting and plasticity effects. International Journal of Psychophysiology, 112, 80–88.
Annual Review of Psychology, 55, 51–86.

171