ArticlePDF Available

Excitatory TMS Modulates Memory Representations

Authors:
  • Stars Behavioral Health Group

Abstract and Figures

Brain stimulation technologies have seen increasing application in basic science investigations, specifically toward the goal of improving memory function. However, proposals concerning the neural mechanisms underlying cognitive enhancement often rely on simplified notions of excitation. As a result, most applications examining the effects of transcranial magnetic stimulation (TMS) on functional neuroimaging measures have been limited to univariate analyses of brain activity. We present here analyses using representational similarity analysis (RSA) and encoding-retrieval similarity (ERS) analysis to quantify the effect of TMS on memory representations. To test whether an increase in local excitability in PFC can have measurable influences on upstream representations in earlier temporal memory regions, we compared 1Hz and 5Hz stimulation to the left dorsolateral PFC (DLPFC). We found that 5Hz rTMS, relative to 1Hz, had multiple effects on neural representations: 1) greater representational similarity during both encoding and retrieval in ventral stream regions, 2) greater ERS in the hippocampus, and, critically, 3) increasing ERS in MTL was correlated with increasing univariate activity in DLPFC, and greater functional connectivity for hits than misses between these regions. These results provide the first evidence of rTMS modulating semantic representations and strengthen the idea that rTMS may affect the reinstatement of previously experienced events in upstream regions.
Content may be subject to copyright.
Excitatory TMS Modulates Memory Representations
Wei-Chun Wang1, Erik A. Wing1, David L.K. Murphy1, Bruce M. Luber3,5, Sarah H.
Lisanby3,4,5,*, Roberto Cabeza1,4, and Simon W. Davis1,2
1Center for Cognitive Neuroscience, Duke University, Durham, NC, USA
2Neurology, Duke University School of Medicine, Durham, NC, USA
3Psychiatry and Behavioral Neuroscience, Duke University School of Medicine, Durham, NC,
USA
4Psychology & Neuroscience, Duke University, Durham, NC, USA
5National Institute of Mental Health, Bethesda, MD, USA
Abstract
Brain stimulation technologies have seen increasing application in basic science investigations,
specifically toward the goal of improving memory function. However, proposals concerning the
neural mechanisms underlying cognitive enhancement often rely on simplified notions of
excitation. As a result, most applications examining the effects of transcranial magnetic
stimulation (TMS) on functional neuroimaging measures have been limited to univariate analyses
of brain activity. We present here analyses using representational similarity analysis (RSA) and
encoding-retrieval similarity (ERS) analysis to quantify the effect of TMS on memory
representations. To test whether an increase in local excitability in PFC can have measurable
influences on upstream representations in earlier temporal memory regions, we compared 1Hz and
5Hz stimulation to the left dorsolateral PFC (DLPFC). We found that 5Hz rTMS, relative to 1Hz,
had multiple effects on neural representations: 1) greater representational similarity during both
encoding and retrieval in ventral stream regions, 2) greater ERS in the hippocampus, and,
critically, 3) increasing ERS in MTL was correlated with increasing univariate activity in DLPFC,
and greater functional connectivity for hits than misses between these regions. These results
provide the first evidence of rTMS modulating semantic representations and strengthen the idea
that rTMS may affect the reinstatement of previously experienced events in upstream regions.
Keywords
connectivity; representational similarity analysis; TMS; episodic memory
*Now at the National Institute of Mental Health. Drs. Sarah H. Lisanby and Bruce Luber contributed to this article while at Duke
University, prior to joining NIMH. The views expressed are their own and do not necessarily represent the views of the National
Institutes of Health or the United States Government.
HHS Public Access
Author manuscript
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Published in final edited form as:
Cogn Neurosci
. 2018 ; 9(3-4): 151–166. doi:10.1080/17588928.2018.1512482.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Introduction
Over the past few decades, repetitive transcranial magnetic stimulation (rTMS) has
developed into a powerful tool to establish causal brain-behavior relationships and improve
cognition. Brain stimulation has been coupled with neuroimaging techniques in order to
identify neural activity associated with enhancement of working (Mottaghy et al., 2002;
Luber et al., 2007), episodic (Kohler et al., 2004), and semantic (Binney et al., 2010)
memory. However, most investigations on the effects of rTMS on neural substrates of
memory functioning have been limited to univariate analyses of brain activity (Bollinger et
al., 2010; Lee and D’Esposito, 2012; Vidal-Pineiro et al., 2014). Here we use a combination
of rTMS, representational similarity analyses (RSA), and encoding-retrieval similarity
(ERS) in order to test the efficacy of this technique in eliciting changes in neural
representations that are a fundamental component of memory functioning.
Many rTMS studies have sought to relate various rTMS stimulation parameters to memory-
related brain processes, focusing on the underlying neural correlates, rather than a
measurable difference in memory. Because the relationship between any individual measure
of memory and corresponding brain function is highly complex and is often mediated by
multiple brain regions and psychological processes, this approach is advantageous for
targeting specific components of neural function that contribute to memory encoding and
retrieval. It is generally accepted that rTMS at higher frequencies (≥ 5Hz) induces local
increases
in BOLD activity during a host of different cognitive operations including motor
planning (Schneider et al., 2010), working memory (Esslinger et al., 2014), and episodic
memory (Vidal-Pineiro et al., 2014),
boosting
ongoing activity associated with successful
performance. Conversely, 1Hz rTMS has been shown to reliably
depress
local hemodynamic
activity (Muellbacher et al., 2000; de Vries et al., 2012; Plow et al., 2014). While the use of
rTMS as a tool for behavioral neuroenhancement is undoubtedly informed by understanding
these local effects, it remains unclear how local changes in neural activity affect large-scale
system dynamics. Research combining rTMS with functional neuroimaging techniques has
shown that rTMS can also have effects on distal brain regions that are functionally or
structurally connected with the area targeted by stimulation (Esslinger et al., 2014; Wang
and Voss, 2015). However, while the focus on “excitability” (a tenuous inference given the
generality of the hemodynamic BOLD signal) in more distal regions of a circuit is becoming
widespread (Fox et al., 2014), the effect of brain stimulation on the underlying
representations upon which various cognitive processes operate has been almost completely
unexplored.
To date, most applications examining the effects of rTMS on neuroimaging data have been
limited to univariate analyses of brain activity. In contrast to standard univariate models,
RSA instead asks how distributed patterns of brain activity evoked by different stimuli are
related to one another, and thus provides the means to directly address questions of how
mental representations are implemented in the brain as unique patterns of information. RSA
methods have been used to explore how the relationship between stimuli of various
dimensions—such as the visual or conceptual similarity of objects, words, or sentences—is
reflected in the pattern of evoked neural responses associated with those stimuli. By
comparing the similarity between each pair of stimuli in an fMRI dataset along a given
Wang et al. Page 2
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
dimension, we obtain a representational dissimilarity matrix (RDM), which serves to
characterize a certain computational model of stimulus properties. These model RDMs can
be compared to brain RDMs reflecting the similarity in activity patterns across the same
pairs of stimuli. This approach has helped identify brain regions where representational
structure captures higher-level semantic meaning that is not strongly contingent on stimulus
modality (Devereux et al., 2013). Regions associated with semantic representations in RSA
analyses include left anterior temporal lobe (ATL), fusiform gyrus, and left angular gyrus
(AG) (Clarke and Tyler, 2014; Martin et al., 2018).
In the domain of episodic memory, representational approaches have been used to test a
fundamental hypothesis that retrieving information from memory produces a partial
recapitulation or reactivation of brain states and representations present when information
was initially experienced (Danker and Anderson, 2010). One way of testing for this
reactivation has involved comparing the similarity of brain patterns between the encoding
and retrieval of items, known as encoding-retrieval similarity (ERS; Ritchey et al., 2013;
Wing et al., 2015). Episodic memory reactivation has been reported in posterior cortical
regions, as well as in the medial temporal lobes (MTL). PFC reactivation effects in
prefrontal regions (ERS and RSA) are generally less common, possibly owing to the
population- and process-level heterogeneity of this region; in fact, a recent meta-analysis
suggests decoding accuracy is generally lower in PFC (Bhandari et al., 2018). Nonetheless,
dorsolateral PFC (DLPFC) plays an important role in controlling relational memory binding
(Blumenfeld and Ranganath, 2007) and the encoding of robust representations (Clarke et al.,
2014), and likely operates on such encoding operations through functionally connected
pathways to subcortical memory systems (Summerfield et al., 2006; Schott et al., 2013).
Furthermore, various frequencies of noninvasive brain stimulation to DLPFC has been
shown to reliably affect episodic memory performance (Sandrini et al., 2003; Sole-Padulles
et al., 2006; Meinzer et al., 2013), and from a practical perspective, this site provides one of
the most reasonable cortical targets to access subcortical memory systems via TMS (though
see Wang and Voss, 2015 for the use of a parietal target).
Whereas previous studies combining TMS with fMRI focused on univariate activity, we
used multivariate RSA and ERS methods to investigate how TMS might alter the underlying
stimulus representations at the time of initial learning, and the implications potential
alteration might have for reactivated mnemonic information at retrieval. We turned to an
existing dataset on which we originally focused on the network-level dynamics of 5Hz vs.
1Hz rTMS (Davis et al., 2017). We applied TMS to DLPFC, a region which plays an
important role in controlling relational memory binding (Blumenfeld and Ranganath, 2007),
before a relational memory encoding-retrieval session in which fMRI data was collected. If
DLPFC-focused stimulation is to have an effect in upstream encoding locations, we
expected greater connectivity from the initial stimulation site to upstream sites managing
these representations. Thus, we made three predictions in the current analysis: 5Hz relative
to 1Hz TMS of DLPFC would enhance (1) the quality of semantic representations (as
measured by RSA), (2) the quality of episodic memory representation reinstatement (as
measured by ERS), and (3) DLPFC activity and functional connectivity accounting for the
effect of rTMS on representations. Answers to these questions would help to clarify the
Wang et al. Page 3
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
variables that may influence the neural response to stimulation during a wide array of
episodic memory paradigms.
Materials & Methods
Participants
Fourteen healthy older adults were recruited for this study (all native English speakers; 8
females; age mean +/− SD, 67.2 +/− 4.4 years; range 61–74 years); one subject was
excluded due to tolerability during rTMS; two additional participants did not complete all
retrieval runs and thus were included in Encoding RSA and univariate analyses, but excluded
from Retrieval RSA and ERS analyses. Each older adult was screened for exclusion criteria
for TMS (history of seizure, brain/head injuries) as well as psychiatric condition (MINI
International Neuropsychiatric Interview, English Version 5.0.0 DSM-IV, (Sheehan et al.,
1998)). None of the older participants reported subjective memory complaints in everyday
life or had MMSE score below 27 (mean +/− SD = 29.1 +/− 0.8).
Stimuli & Procedure
The associative memory task was comprised of a set of 360 sentences, each of which
included a concrete subject and direct object. In each sentence, both subject and direct object
were capitalized to indicate to the subject which specific nouns were to be remembered (
“A
SURFBOARD was on top of the TRUCK.”
). Associative strength between nouns in a
sentence (as determined by the USF word association norms (Nelson et al., 2004), were
normally distributed (mean = 0.025, SD = 0.038), and both the mean imageability (mean =
561.3, SD = 42.3), mean familiarity (mean = 564.4, SD = 37.9), mean concreteness (mean =
528.8, SD v = 38.7) and total length of each set of sentences (mean = 6.38, range = 5–9
words) was counterbalanced across all sentences used therein.
One run of the associative task comprised an encoding block of 90 sentences presented
visually for 3 seconds, separated by a jittered inter-trial interval (mean 7 seconds), with a
total duration of 9 minutes for the entire encoding block. After a brief reminder of the
instructions for the task (8 seconds) a subsequent retrieval block began, comprising 68 word
pairs from the same previously studied sentence, and 22 new pairs composed of two words
recombined from two different sentences, with a duration of 6 minutes for the entire retrieval
block. Participants were asked to judge whether each word pair was intact or recombined
and indicate how confident they were in their decision on a 4-point rating scale. Given the
low proportion of low-confidence responses in the current data (average of 13.6% of
responses to old trials, across participants), we collapsed low and high-confidence responses,
and excluded encoding trials in which participants failed to indicate a response at retrieval.
Image Acquisition
An outline of all data acquisition events is depicted in Figure 1. Scanning was divided
between two days, 1–4 days apart, with the first day comprised of an individualized
functional DLPFC memory-success localizer for stimulation on Day 2 (Figure 2A), which
comprised the same sentence encoding task described above. All procedures were completed
on a GE MR 750 3-Tesla scanner (General Electric 3.0 tesla Signa Excite HD short-bore
Wang et al. Page 4
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
scanner, equipped with an 8-channel head coil). Coplanar functional images were acquired
with an 8-channel head coil using an inverse spiral sequence with the following imaging
parameters: 64 × 64 matrix, flip angle = 77o, TR = 2000ms, TE = 31ms, FOV = 24.0 mm2,
and a slice thickness of 3.8mm, for 37 slices. The anatomical MRI was acquired using a 3D
T1-weighted echo-planar sequence (256 × 256 matrix, TR = 12 ms, TE = 5 ms, FOV = 24
cm, 68 slices, 1.9 mm slice thickness). Scanner noise was reduced with earplugs and head
motion was minimized with foam pads. Behavioral responses were recorded with a four-key
fiber optic response box (Resonance Technology), and when necessary, vision was corrected
using MRI-compatible lenses that matched the distance prescription used by the participant.
TMS Procedure
Prior to the TMS-fMRI session, the scalp location for TMS coil position over the left
DLPFC location had been found by using an infrared neuronavigation system (Brainsight:
Rogue Research, Montreal, Canada). Specifically, the point of greatest activation in the left
DLPFC in the fMRI memory contrast (i.e., encoding trials which were subsequently
remembered versus forgotten) from the first day of scanning was chosen from the fMRI
overlay on the subject’s structural MRI, both of which had been uploaded into BrainSight.
After co-registration of the subject’s head with his MRI, the DLPFC location was marked on
a tight fitting acrylic swim cap that stayed on the subject’s head until TMS-fMRI
interleaving was completed on the same day. At that time, participants were acclimated to
the sort of TMS pulses to be delivered later in the scanner with a series of single pulses at
the target site, as well as a short burst of 5Hz stimulation. The motor threshold (MT) for
each subject was determined using a MagVenture R30M device located outside the scanner
room, part of an MRI compatible TMS system which included a non-ferrous figure-8 coil
with 12m long cable and artifact reducing counter-current charging system (MagVenture,
Farum, Denmark). MTs were determined using electromyography of the right first dorsal
interosseous (FDI) muscle and defined as the lowest setting of TMS device intensity at
which ≥ 5 out of 10 motor evoked potentials of at least 50μV peak to peak amplitude could
be elicited.
Before each functional scan, two 10-minute trains of either 1Hz or 5Hz stimulation were
delivered at 120% MT immediately prior to an fMRI acquisition. The position of the TMS
coil was reset to the same target site before the beginning of each rTMS session, and
monitored continuously while the subject lay supine in the bed of the MR scanner. 1Hz
rTMS was delivered in a continuous train of 10 minutes, while 5Hz rTMS was delivered in
intermittent 6 sec trains with a 24 sec inter-train interval, also for 10 minutes. Dosage was
equivalent between 1Hz and 5Hz rTMS conditions (600 total pulses), and the order of
stimulation frequency was counterbalanced across participants. Immediately after the 10
minutes of TMS, participants were positioned in the scanner, and performed the encoding
and subsequent retrieval portion of the sentence task while fMRI was acquired. The amount
of time elapsed between the end of the rTMS train to the beginning of the functional scan
was 9.4 minutes (SD = 1.7 minutes).
Wang et al. Page 5
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Data Analyses
Functional MRI preprocessing
Functional images were preprocessed using image processing tools, including FLIRT and
FEAT also from FSL, in a publicly available analysis pipeline developed by the Duke Brain
Imaging and Analysis Center (https://wiki.biac.duke.edu/biac:analysis:resting_pipeline).
Images were corrected for slice acquisition timing, motion, and linear trend; motion
correction was performed using FSL’s MCFLIRT, and 6 motion parameters estimated from
the step were they regressed out of each functional voxel using standard linear regression.
Images were then temporally smoothed with a high-pass filter using a 190s cutoff, and
normalized to the Montreal Neurological Institute (MNI) stereotaxic space. White matter
and CSF signals were also removed from the data, using WM/CSF masks generated by
FAST and regressed from the functional data using the same method as the motion
parameters. Spatial filtering with a Gaussian kernel of full-width half-maximum (FWHM) of
6mm was applied.
To examine multivariate activity patterns, each sentence was modeled in a separate GLM
using the least squares single (LSS) approach on the unsmoothed data, yielding first-level
single-trial beta images for each trial in each participant (Mumford et al., 2012). This
method estimates a first-level model for every trial such that the trial of interest serves as one
regressor and all other trials serve as another regressor. Trials were modeled as a single stick
function at the onset of a sentence (Encoding) or word pair (Retrieval) stimulus, which is
then convolved hemodynamic response function (HRF); models also included six motion
regressors (x-dim, y-dim, z-dim, roll, pitch, and yaw) in addition to block-wise run
regressors. Group contrasts masked out white matter/CSF (SPM gray matter template values
<0.1), all contrast images display effects at p <0.005, and cluster-level corrections for
multiple comparisons to p <.05 with Monte Carlo Simulations (Slotnick et al., 2003;
Slotnick, 2017).
RDM Construction and RSA Analysis
Semantic similarity (RDMs) for the sentences (at encoding, Figure 3A) and the word pairs
(at retrieval, Figure 3B) were constructed using the website cortical.io (Webber, 2015).
Cortical.io utilizes an unsupervised machine learning algorithm to encapsulate the entirety
of semantic space (trained on Wikipedia) into a vector of 16,384 co-occurring concepts, or
‘semantic contexts.’ To calculate semantic similarity, it extracts the semantic contexts (in the
vector of 16,384) associated with the words in each sentence at encoding and each word pair
at retrieval (i.e., a binary ‘semantic fingerprint’ that is visualized with a 1282 matrix). To
then create the RDMs, the dissimilarity (1-cosine similarity) between the semantic
fingerprint for each pair of sentences or word pairs was then calculated. This ‘semantic
fingerprint’ method is relatively new, but has been successfully used to group similar firms
based on their business descriptions (Ibriyamova et al., 2017) and group academic authors
based on the content of their publications (Han et al., 2017). Thus, it provides a useful tool
for extracting semantic similarity beyond just single words.
Wang et al. Page 6
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
For each subject, second-order correlations were computed between our semantic RDM and
single-trial beta images for each participant using an in-house searchlight script (https://
github.com/brg015) with a 5-voxel searchlight sphere, separately for the three stimulation
conditions. For group analyses, these second-order correlation maps were then spatially
smoothed (6-mm isotropic FWHM Gaussian filter).
ERS Analysis
ERS was calculated using an in-house searchlight script with a 5-voxel searchlight sphere
(https://github.com/brg015) at both the item level, where encoding and retrieval trials
involved the same item, and the set level, where encoding and retrieval trials belonged to the
same set (i.e., stimulation condition) but involved different items. Item-ERS and set-ERS
were calculated by correlating encoding and retrieval activity patterns within each
searchlight sphere. For item-ERS, the encoding and retrieval activity patterns corresponding
to the same trial were correlated, whereas for set-ERS, retrieval activity patterns for each
trial was correlated with the encoding activity patterns for all trials in that stimulation
condition (e.g., 5Hz) and then averaged to create the whole-brain similarity volume for that
retrieval trial.
After ERS volumes had been calculated for each retrieval trial, fixed-effect contrasts were
generated separately for item-ERS and set-ERS by averaging together all ERS volumes,
yielding six mean ERS images per participant. For group analyses, these mean ERS images
were then spatially smoothed (6-mm isotropic FWHM Gaussian filter). Because Baseline
information was collected on a separate day, both RSA and ERS analyses are focused on the
difference between RSA/ERS observed after two rTMS frequencies (5Hz, 1Hz);
nonetheless, we include Baseline information here, as well as significant clusters in the
Baseline condition, for descriptive purposes.
Lastly, to control for univariate activity, we also conducted confirmatory within-subject
binary logistic regressions in clusters showing ERS differences as a function of stimulation
condition, while still considering the influence of univariate activity at both encoding and
retrieval. Such a result would help to confirm the hypothesis that TMS input specifically
modulates the pattern of stimulus representations, and not merely the mean BOLD activity
associated with processing the stimulus. In this regression, stimulation condition (1Hz vs
5Hz) was the dependent variable, and there were three independent variables: (1) ERS, (2)
univariate encoding activity, and (3) univariate retrieval activity. A test on the parameter
estimates corresponding to the ERS regressor (see Table 3, rightmost column) indicated
whether similarity measures uniquely predicted stimulation condition while accounting for
the effects of univariate activity. Thus, the logistic regression model represents a
confirmatory test of extant results, and therefore one-tailed significance testing was used
given that we expected to find an effect in the direction exhibited by the ERS.
Relationships between ERS and the DLPFC Stimulation Site
Lastly, we conducted two additional tests to examine the direct influence of left DLPFC
stimulation on upstream representations. While the analyses above are focused on
multivariate pattern information, estimates of univariate activity serve as useful markers of
Wang et al. Page 7
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
excitation or inhibition after 5Hz or 1Hz rTMS, respectively. Thus, we examined whether
the magnitude of ERS correlated with this univariate activity in the region of stimulation in
DLPFC. To do so, we extracted the mean univariate activity (beta values) in the DLPFC
stimulation site, and ERS (Pearson’s
r
values) in clusters exhibiting significant effects and
correlated them separately for each stimulation condition.
Second, we used task-based functional connectivity estimates to determine the degree to
which these regions were coupled with DLPFC across encoding, i.e., immediately after 5Hz
or 1Hz rTMS. Functional connection matrices representing task-related connection strengths
were estimated using a correlational psychophysical interaction (cPPI) analysis (Fornito et
al., 2012). Briefly, the model relies on the calculation of a PPI regressor for each region,
based on the product of that region’s timecourse and a task regressor of interest, in order to
generate a term reflecting the psychophysical interaction between the seed region’s activity
and the specified experimental manipulation. In the current study the task regressors based
on the convolved task regressors from the univariate model described above were used as the
psychological regressor, which coded subsequently remembered and subsequently forgotten
word pairs with positive and negative weights, respectively, of equal value. This
psychological regressor was multiplied with two network timecourses for region
i
and j. We
then computed the partial correlation ρPPI i,PPI jz, removing the variance z associated with
the psychological regressor, the timecourses for regions
i
and j, and constituent noise
regressors. We accounted for the potential effects of head motion and other confounds by
assessing the 6 motion parameters and including these parameters in our partial correlation
between regions.
Results
Behavioral results
As explained in the Introduction, the goal of the current analysis was to investigate the
effects of TMS on mnemonic representations and their fidelity across encoding and retrieval
conditions. Behavioral results for our associative memory task are summarized in Table 1.
We found no significant differences between 5Hz vs. 1Hz rTMS conditions, either in Hit
Rate (t(12) = 1.18, p = 0.27), False Alarm Rate (t(12) = 1.44, p = 0.18), or d’ (t(12) = 0.08, p
= 0.94); response times for correct trials were similarly not differentiated across stimulation
conditions (all p > 0.05), and significance levels were unchanged when removing two
participants who failed to complete retrieval runs. While rTMS order was evenly
counterbalanced across participants, we nonetheless failed to see a difference in stimulation
frequency order in either the post-1Hz (t(12) = 0.76, p = 0.45) or post-5Hz performance
(t(12) = 1.42, p = 0.18). In order to address the possibility that 1Hz or 5Hz rTMS exhibited
differential rates of decay in their effect on behavior, we examined the slope of a regression
line fit to responses to all non-null-response trials, for each stimulation condition, where the
x-axis described successive trials, and the y-axis represented individual trial response times.
Slope values across participants (means across participants shown in Table 1) were then
compared between Baseline, 5Hz, or 1Hz rTMS conditions; here we found no difference in
slopes between any pairwise comparison between conditions (t(12) < 1.0). These null
findings for behavioral differences between rTMS conditions suggest that any brain-related
Wang et al. Page 8
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
differences in multivoxel pattern information (RSA/ERS) are not readily attributable to
differences in task strategy or strength of encoding, and do not readily suggest evidence for
differential decay rates between stimulation conditions.
Univariate Results
In order to help validate our premise that 1Hz rTMS is inhibitory and 5Hz is excitatory, we
first extracted univariate activity at the stimulation site during encoding (Figure 2C). As
expected, rTMS significantly modulated memory-related activity, as isolated by the
subsequent memory difference in univariate activity between remembered and forgotten
trials. As reported in our previous analysis of these data (Davis et al., 2017), the univariate
activity for successfully remembered trials in the stimulated left DLPFC region was reduced
by 1 Hz rTMS compared with Baseline (t(12) = −2.43, p = 0.03), but was increased by 5 Hz
rTMS (t(12) = 2.83, p = 0.015) compared with Baseline activation for subsequently
remembered encoding trials. Next, in order to test whether these univariate effects were
specific to the stimulation site, we tested the contralateral right DLPFC site for each subject
—we observed no univariate modulation due to subsequent memory (Figure 2C). Lastly, in
order to address the possibility that 1Hz or 5Hz rTMS exhibited differential rates of decay in
their effect on the BOLD response during our task, we examined the slope of a regression
line fit to individual trial BOLD responses to all non-null-response trials, for each
stimulation condition. Similar to the RT-slope analysis above, the x-axis represented
successive trials, and the y-axis represented univariate BOLD activity during encoding trials,
and slope values for all participants were compared pairwise between conditions. We found
no difference in slopes for baseline (slope = 0.0014), 5Hz (slope = 0.0015), or 1Hz (slope =
−0.0028) conditions (t(12) < 1.0), suggesting both that in our study, rTMS effects were
consistent across the length of the 9 minute encoding run, and that the decay did not differ
between different forms of rTMS.
RSA results
Our main goal was to examine evidence that rTMS has upstream effects on the quality of
information processed in a standard memory paradigm. To test our first prediction on the
effects of TMS on semantic representations, we examined the influence of rTMS on the
semantic quality of representations (2nd order correlation with the semantic RDM) during
the Encoding period. The searchlight RSA revealed greater second-order correlations
between activity patterns and our semantic RDM for excitatory (5Hz) than inhibitory (1Hz)
rTMS (inclusively masked by 5Hz effect at p < .05) in the left fusiform cortex (Figure 3A;
Table 2, top panel). No significant clusters emerged in the reverse contrast (i.e., 1Hz > 5Hz,
inclusively masked by the 1Hz effect at p < .05). Next, we examined the fidelity of
representations during the Retrieval period, when participants were completing an
associative recognition test for word pairs from the sentences read at encoding. The
searchlight RSA revealed greater second-order correlations between activity patterns and our
semantic RDM for 5Hz than 1Hz rTMS (inclusively masked by 5Hz effect at p < .05) in
regions including the PHC and ATL (Figure 3B; Table 2, bottom panel). Again, no clusters
emerged in the reverse contrast (1Hz > 5Hz, inclusively masked by the 1Hz effect at p < .
05). In sum, consistent with our first prediction, we found that TMS enhanced the quality of
Wang et al. Page 9
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
semantic representations (greater 2nd order correlation with semantic RDM) in typical
semantic regions such as left fusiform and left ATL.
Lastly, we examine the effects of TMS stimulation regions that show RSA in the Baseline
condition; we include the z-statistic for the 5Hz > 1Hz comparison within a mask derived
from the cluster showing RSA effects in the Baseline condition. This value helps to provide
context for the upregulation (or not) in RSA associated with the 5Hz > 1Hz difference.
However, in both Encoding and Retrieval, Baseline regions (i.e., ITG) show no significant
modulation associated with stimulation condition in our sample.
ERS Results
To test our second prediction, we then examined the influence of rTMS on the fidelity of
memory representations measured using ERS. ERS allows us to identify the reactivation of
episodic representations between encoding and retrieval by comparing activity patterns for
trials belonging to the same item (i.e., item-ERS) and trials belonging to different items (i.e.,
set-ERS) at encoding and retrieval. We first examined whether any regions exhibited
Stimulation by ERS level interactions. This contrast revealed greater item-ERS than set-ERS
for 5Hz than 1Hz stimulation (inclusively masked by 5Hz effect at p < .05) in MTL and
temporal regions (Table 3), including the left hippocampus (Figure 4). These effects were
significant even when using a logistic regression model to test whether ERS predicted
stimulation condition while simultaneously controlling for univariate activity at both
encoding and retrieval (Table 3, rightmost column). Furthermore, they also remained
significant when restricting ERS analysis to remembered trials. Memory Success did not
appear to have a strong influence on these regions, such that we observed no significant
Success x ERS effects in either 5Hz or 1Hz stimulation conditions (nor any Success x
Stimulation interaction). Critically, we did not observe greater item-ERS than set-ERS for
misses during either 5Hz or 1Hz stimulation (p’s > .70). Nonetheless, it is difficult to draw
any lasting conclusions from Success-related analyses of these data given the lower
proportion of misses in the current study compared to previous studies; future work focused
on Success-related modulation should be mindful of these concerns in their study design.
Relationships between ERS and the DLPFC Stimulation Site
Lastly, to test our third prediction that the representational would be associated with
increases in DLPFC activity and connectivity. We first extracted univariate activity at the
stimulation site during encoding (see above, and Figure 2) to compare with our observed
ERS values. A pairwise comparison between 1Hz and 5Hz fMRI activity for successfully
remembered trials was also significant (t(12) = 3.43, p = 0.006). This result confirms that
rTMS modulated memory-related activity at our site of stimulation and enables our two
subsequent analysis of how this upregulation of univariate activity in left PFC is related to
upstream effects in MTL.
We then examined whether univariate activity at the stimulation site (DLPFC) correlated,
across trials, with the magnitude of our observed stimulation by ERS level interaction
effects. This analysis found a significantly greater correlation (t(10) = 2.28, p = .042)
between left DLPFC activity at encoding and the left MTL cluster for 5Hz > 1Hz
Wang et al. Page 10
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
stimulation (Figure 5A). There were no significant correlations found in any other ERS
regions.
Lastly, in order to link local stimulation effects in the DLPFC to representational changes as
a function of memory success, we examined the functional connectivity between the DLPFC
stimulation site and regions that were found to exhibit significant stimulation by ERS level
interactions. We found a significant Success x Frequency interaction (F(1,10) = 3.76, p =
0.004), wherein connectivity at encoding between DLPFC and the left MTL cluster was
greater for subsequently remembered than forgotten trials following 5Hz compared to 1Hz
rTMS (Figure 5B). There were no significant correlations found in any other ERS regions.
These results help to clarify the role of DLPFC in mediating upstream semantic
representations that are important for successful memory functioning.
Discussion
To our knowledge, this is the first study to demonstrate that rTMS can modulate memory
representations as measured by RSA and ERS. Thus, this work links brain stimulation
techniques with semantic representations and episodic memory functioning using a novel
multimodal paradigm. The study yielded three principal findings. First, 5Hz, relative to 1Hz,
rTMS boosted the quality of semantic representation (i.e., 2nd order correlation with
semantic RDM) in left fusiform gyrus. Second, 5Hz, but not 1Hz, rTMS enhanced quality of
memory representations in MTL, as assessed by ERS analyses. Lastly, linking the ERS
effect to the DLPFC stimulation, we found that increases in univariate activity at the
stimulation site were correlated with increases in item-level ERS in MTL, and that these two
sites were more functionally connected during subsequently remembered trials. We discuss
each of these results and their implications below.
rTMS effects on Semantic Representations
Our result demonstrates that external manipulations of local excitability in left DLPFC
impact neural pattern similarity in distinct temporal and parietal regions known to code for
semantic and episodic features. Foundational to this finding is the idea that the neural and
semantic similarity between individual items represents meaningful information, and that
this representation in a given brain region or computational model can be characterized by
the matrix of dissimilarities between the stimulus representations (Kriegeskorte and Kievit,
2013). The information encoded in our task represented complex sentences with conceptual
memoranda common to everyday life. Despite its fundamental role in cognition, the
informational content that is represented during reading remains elusive. To understand how
semantic information is represented in the brain, we used a comprehensive database of
relationships between words capturing a broad semantic space (Cortical.io). The similarity
between the semantic contexts in each sentence (encoding) or word pair (retrieval) across
our task can be summarized with a single representational dissimilarity matrix (or RDM),
and thus presents a powerful means of describing relatively complex information patterns
across multiple trials.
While no study has used TMS to examine changes in pattern similarity, a number of studies
have targeted activation patterns associated with more upstream components of conceptual
Wang et al. Page 11
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
encoding. Recent studies suggest that TMS, when applied early in the ventral processing
stream, may strongly impact univariate estimates related to visual and mnemonic processing.
20Hz TMS applied to early visual cortex during short-term retention of visual stimuli results
in a reduction in behavioral measures of working memory precision (van de Ven et al., 2012;
van de Ven and Sack, 2013), supporting the notion that visual cortex stores precise
representations of visual working memory contents. While most of these studies are focused
on disrupting posterior regions that are thought to represent stimulus-specific information,
there is also evidence that disruption of semantic processing can also occur as a consequence
of 1Hz rTMS to the anterior temporal lobes (Lambon Ralph et al., 2009; Binney et al.,
2010).
While our study represents an early application of work combining RSA with
neuromodulation to examine semantic and episodic memories, it should be noted that a
number of high-profile working memory studies have used multivariate techniques to assess
how neuromodulation affects item-level representations in working memory (Rose et al.,
2016; Wolff et al., 2017). While the focus of these approaches tends to focus on tagging a
specific memoranda or transient “memory states”, such techniques do not speak to the
modulation of an existing, model-driven assessment of a semantic system, e.g., via the
semantic associations between items as assessed in our sample by Cortical.io, or feature-
normed techniques as used by Tyler and colleagues (Devereux et al., 2013; Tyler et al.,
2013; Clarke and Tyler, 2014). As such, these working memory memoranda serve as
representations only in that they are unique neural representations, but still have no
underlying psychological model of how different stimuli relate to each other.
Nonetheless, our results provide a clear step forward for the investigation of the encoding
and retrieval of semantic concepts. Critically, both RSA and ERS were greater for trials after
excitatory stimulation in regions typically associated with more abstract semantic
representations like the fusiform gyrus and ATL (Figure 3); while the former it typically
associated with representing object hierarchies or categories (Clarke and Tyler, 2014), the
latter is known to play a role in stimulus-independent semantic representations (Olson et al.,
2007; Patterson et al., 2007). These results suggest that conceptual information represented
in temporal regions is modulated by TMS. For meaningful word stimuli, recent work has
shown that representations in ATL relate to measures of true and false memory, suggesting
that exogenously induced representational changes in this region might underpin aspects of
memory function for similar stimuli (Chadwick et al., 2016). These convergent results
therefore suggest a distinct mechanism whereby specific mnemonic representations linked to
specific item information become consolidated and available for later access through the
reactivation of specific neural patterns. While the frequency-specific effects observed in our
study did not dissociate remembered and forgotten items (as observed by a lack of a 3-way
interaction or 2-way Success x Stimulation interaction), they nonetheless demonstrate the
ability of rTMS to boost the specificity of mnemonic representations, as evidenced by a
significant Stimulation x ERS interaction. To date there is no evidence that brain stimulation
can modulate these effects, but the current results help to shed some light on this potential
mnemonic process.
Wang et al. Page 12
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Top-down influences on memory
As noted in the Introduction, one needs to consider that the region-specific effect of rTMS
on memory performance might have been boosted by indirect manipulation of regions
connected with the stimulation site. The current results provide preliminary evidence that
excitatory influences in DLPFC can have meaningful influences on the representational
information processed in more posterior cortices. We found two complimentary pieces of
evidence to support the idea that PFC-mediated increases in representational similarity can
be induced with rTMS. First, and most critically, the local activity increases at the site of
stimulation were more strongly correlated with item-level ERS in the left MTL during
excitatory relative to inhibitory stimulation (Figure 5A). Additionally, this hippocampal ERS
effect was significant for remembered and but not forgotten items in our sample. Although
we cannot discount this null finding due to fewer forgotten trials than previous ERS studies
(Ritchey et al., 2012; Wing et al., 2015), the influence of this region in differentiating
between Item > Set-level representations was clearly increased by 5Hz rTMS, and was
significant even when restricted to remembered trials. Second, functional connectivity
between the DLPFC stimulation site and the left MTL during encoding demonstrated a
significant Stimulation x Memory Success interaction (Figure 5B), such that connectivity
was greater during successfully encoded trials than subsequently forgotten trials after 5Hz
(but not 1Hz) stimulation. These results suggest that representations in MTL were boosted
by indirect manipulation.
While this is, to our knowledge, the first time TMS has been shown to modulate pattern
similarity, it is not the first to suggest that PFC stimulation may boost the top-down
activation (or controlled retrieval) of semantic knowledge (Badre and Wagner, 2007; Bunge
et al., 2009). Causal links between PFC activity and the properties of visual cortical neurons
has been established through intervention with TMS. Disruption of PFC activity using 1Hz
rTMS causes a significant reduction in the selectivity of fMRI responses in visual cortex,
and appears to inhibit top-down selective attentional modulation to specific category types in
the occipitotemporal cortex, during both online perception (Higo et al., 2011), and during
working memory delay phases, (Lee and D’Esposito, 2012). The principle implied here is
that DLPFC inputs enhance selectivity in visual cortex; in the current study we failed to
observe any significant modulation of ERS in such early visual regions (Table 3), which
did
show Baseline ERS. While it is difficulty to draw lasting conclusions from such a null effect,
we did find stimulation-related modulation in more downstream regions associated with
amodal memory representation (Martin et al., 2018). However, given that our stimulation
region was more functionally connected to MTL regions exhibiting significant ERS effects,
our results are more in agreement with findings describing the importance of frontotemporal
connectivity in accessing semantic and episodic memory representations (Clarke et al.,
2013; Backus et al., 2016). These results therefore suggest that the amodal representational
content evoked during encoding and reactivated during retrieval is fundamentally affected by
stimulation in frontal control regions.
The importance of frequency
Lastly, it is key to note that both our RSA/ERS pattern similarity findings, and that success-
related functional connectivity between left DLPFC and left MTL during encoding (Figure
Wang et al. Page 13
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
5B) were uniquely associated with fMRI data collected after 5Hz rTMS. This result is
reasonable given the importance of frontotemporal synchrony within the theta band (4–
10Hz) during episodic encoding in regions functionally connected to the hippocampus,
including LPFC and VPC (Osipova et al., 2006; Hsieh and Ranganath, 2014). EEG signals
generated by the hippocampus are typically dominated by regular theta waves, often
continuing for many seconds after being elicited by a focused event (Buzsaki, 2002, 2005).
Increased rhythmic synchrony across distant regions within a cortical network is thought to
improve information processing by increasing network efficiency, an effect particularly
important during demanding cognitive processing (Fries, 2009; Deco et al., 2011), such as
episodic encoding and retrieval. For example, a number of scalp EEG studies have reported
that frontotemporal theta rhythms are enhanced for items that are subsequently remembered
more than items that are forgotten (Summerfield and Mangels, 2005; Hanslmayr et al., 2009;
Nyhus and Curran, 2010). Even more recent intracranial work from Kahana and colleagues
has helped to validate this observation, which suggest that widespread synchronization of
theta networks predict successful encoding (Ezzyat et al., 2017; Solomon et al., 2017), and
that selective modulations of theta- and high-gamma-band activity in the temporal lobes can
be used to decode successful memory states and inform targeted, closed-loop brain
stimulation paradigms (Ezzyat et al., 2018). Such direct applications of frequency-specific
information and brain stimulation represent powerful tools for future research and clinical
application.
Stimulation variables such as intensity, frequency and duration clearly have an strong
influence on the psychological and neurobiological effects of rTMS. One critical aspect of
offline rTMS applications like ours is the notion of decay; rTMS was delivered for 10 min,
ostensibly yielding an effect that has been shown to last ~9 min (Thut et al., 2006) and hence
persist during the entire encoding run (6-min). Nonetheless, some authors have suggested
differential decay rates in the effect of stimulation may differ for rTMS of different
frequencies. Previous studies have provided some information about the duration of
behavioral (Robertson et al., 2001; Mottaghy et al., 2002) and neural (Eisenegger et al.,
2008) effects after low-frequency “off-line” long-train rTMS applied to the dorsolateral
prefrontal cortex (DLPFC). While the general consensus is that 1–10Hz rTMS has a sharper
decay than higher-frequency theta-burst stimulation (Huang et al., 2005; Nyffeler et al.,
2006), there is nonetheless still considerable debate on the duration of low- and high-
frequency rTMS (Gamboa et al., 2010). Though our study was not adequately designed to
elucidate these factors affecting differential decay between 1Hz and 5Hz, we cannot
discount the possibility that the greater ERS after 5Hz rTMS was driven by greater decay in
the behavioral and neural effects after 1Hz rTMS. Nonetheless, the lack of any difference in
slopes for either RT (Table 1) or mean univariate BOLD in the stimulation site suggests no
evidence for such a differential decay in our sample.
Caveats
Our study has a number of limitations. First, there are unique subjective differences between
the experience of 1Hz and 5Hz rTMS. Thus, it is possible that greater RSA/ERS effects after
5Hz TMS were due to participants being cued by the greater intensity of sound produced by
Wang et al. Page 14
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
a 5Hz rTMS and may have been in a state of heightened awareness. While we cannot
discount this possibility, we note that 1) the memory task was not concurrent with the rTMS,
so the influence of such heightened attention would have to persist through the ~10 minutes
necessary to initiate the fMRI scan, 2) we observed no significant univariate effects of 1Hz
stimulation compared to baseline, suggesting that the application of any TMS was not
sufficient to elicit this effect, and 3) performance on the memory task did not differ as a
function of stimulation frequency.
Second, our study suffers from a low sample size; a recent meta-analysis of clinical rTMS
studies found that the average sample size for such studies is low (mean sample size = 18;
Martin et al., 2003). Clearly this standard is not viable for the field to progress beyond
qualitative observations of rTMS impact. However, our findings nonetheless provide some
limited support for the conclusion that the modulatory influence of rTMS may be limited to
specific frequency parameters, offer an exploratory foundation in characterizing
representational changes associated with rTMS, and suggest an analytical format that may
reveal important information as yet unexplored in more well-powered brain stimulation
studies.
Conclusion
Overall, our findings indicate a direct link between 5Hz rTMS applied to DLPFC, and
memory representations in posterior parietal and temporal cortices. Taken together, these
results provide the first evidence of excitatory TMS enhancing memory representations
using two different multivariate methods and suggests that rTMS may affect the
reinstatement of previously experienced events in regions distal to the site of activation. By
showing that stimulation impacts the neural basis of episodic memory at specific
frequencies, our data provide a foundation for future work to apply stimulation when it is
most likely to improve memory function.
Acknowledgements:
This work was supported by National Institutes of Health Grant R01 AG19731 and F32 AG049574.
References
Backus AR, Schoffelen JM, Szebenyi S, Hanslmayr S, Doeller CF (2016) Hippocampal-Prefrontal
Theta Oscillations Support Memory Integration. Curr Biol 26:450–457. [PubMed: 26832442]
Badre D, Wagner AD (2007) Left ventrolateral prefrontal cortex and the cognitive control of memory.
Neuropsychologia 45:2883–2901. [PubMed: 17675110]
Bhandari A, Gagne C, Badre D (2018) Just above Chance: Is It Harder to Decode Information from
Human Prefrontal Cortex Blood Oxygenation Level-dependent Signals? Journal of cognitive
neuroscience:1–26.
Binney RJ, Embleton KV, Jefferies E, Parker GJ, Ralph MA (2010) The ventral and inferolateral
aspects of the anterior temporal lobe are crucial in semantic memory: evidence from a novel direct
comparison of distortion-corrected fMRI, rTMS, and semantic dementia. Cereb Cortex 20:2728–
2738. [PubMed: 20190005]
Blumenfeld RS, Ranganath C (2007) Prefrontal cortex and long-term memory encoding: an integrative
review of findings from neuropsychology and neuroimaging. The Neuroscientist 13:280–291.
[PubMed: 17519370]
Wang et al. Page 15
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Bollinger J, Rubens MT, Zanto TP, Gazzaley A (2010) Expectation-driven changes in cortical
functional connectivity influence working memory and long-term memory performance. The
Journal of neuroscience : the official journal of the Society for Neuroscience 30:14399–14410.
[PubMed: 20980597]
Bunge SA, Helskog EH, Wendelken C (2009) Left, but not right, rostrolateral prefrontal cortex meets a
stringent test of the relational integration hypothesis. Neuroimage 46:338–342. [PubMed:
19457362]
Buzsaki G (2002) Theta oscillations in the hippocampus. Neuron 33:325–340. [PubMed: 11832222]
Buzsaki G (2005) Theta rhythm of navigation: link between path integration and landmark navigation,
episodic and semantic memory. Hippocampus 15:827–840. [PubMed: 16149082]
Chadwick MJ, Anjum RS, Kumaran D, Schacter DL, Spiers HJ, Hassabis D (2016) Semantic
representations in the temporal pole predict false memories. Proc Natl Acad Sci U S A 113:10180–
10185. [PubMed: 27551087]
Clarke A, Tyler LK (2014) Object-Specific Semantic Coding in Human Perirhinal Cortex. J Neurosci
34:4766–4775. [PubMed: 24695697]
Clarke A, Devereux BJ, Randall B, Tyler LK (2014) Predicting the Time Course of Individual Objects
with MEG. Cereb Cortex.
Clarke A, Taylor KI, Devereux B, Randall B, Tyler LK (2013) From perception to conception: how
meaningful objects are processed over time. Cereb Cortex 23:187–197. [PubMed: 22275484]
Danker JF, Anderson JR (2010) The Ghosts of Brain States Past: Remembering Reactivates the Brain
Regions Engaged During Encoding. Psychological bulletin 136:87–102. [PubMed: 20063927]
Davis SW, Luber B, Murphy DLK, Lisanby SH, Cabeza R (2017) Frequency-specific neuromodulation
of local and distant connectivity in aging and episodic memory function. Hum Brain Mapp.
de Vries PM, de Jong BM, Bohning DE, Hinson VK, George MS, Leenders KL (2012) Reduced
parietal activation in cervical dystonia after parietal TMS interleaved with fMRI. Clin Neurol
Neurosurg 114:914–921. [PubMed: 22386903]
Deco G, Buehlmann A, Masquelier T, Hugues E (2011) The role of rhythmic neural synchronization in
rest and task conditions. Frontiers in human neuroscience 5:4. [PubMed: 21326617]
Devereux BJ, Clarke A, Marouchos A, Tyler LK (2013) Representational similarity analysis reveals
commonalities and differences in the semantic processing of words and objects. The Journal of
neuroscience : the official journal of the Society for Neuroscience 33:18906–18916. [PubMed:
24285896]
Eisenegger C, Treyer V, Fehr E, Knoch D (2008) Time-course of “off-line” prefrontal rTMS effects--a
PET study. Neuroimage 42:379–384. [PubMed: 18511301]
Esslinger C, Schuler N, Sauer C, Gass D, Mier D, Braun U, Ochs E, Schulze TG, Rietschel M, Kirsch
P, Meyer-Lindenberg A (2014) Induction and quantification of prefrontal cortical network
plasticity using 5 Hz rTMS and fMRI. Hum Brain Mapp 35:140–151. [PubMed: 22965696]
Ezzyat Y et al. (2018) Closed-loop stimulation of temporal cortex rescues functional networks and
improves memory. Nature communications 9:365.
Ezzyat Y et al. (2017) Direct Brain Stimulation Modulates Encoding States and Memory Performance
in Humans. Curr Biol 27:1251–1258. [PubMed: 28434860]
Fornito A, Harrison BJ, Zalesky A, Simons JS (2012) Competitive and cooperative dynamics of large-
scale brain functional networks supporting recollection. Proc Natl Acad Sci U S A 109:12788–
12793. [PubMed: 22807481]
Fox MD, Buckner RL, Liu H, Chakravarty MM, Lozano AM, Pascual-Leone A (2014) Resting-state
networks link invasive and noninvasive brain stimulation across diverse psychiatric and
neurological diseases. Proc Natl Acad Sci U S A 111:E4367–4375. [PubMed: 25267639]
Fries P (2009) Neuronal gamma-band synchronization as a fundamental process in cortical
computation. Annu Rev Neurosci 32:209–224. [PubMed: 19400723]
Gamboa OL, Antal A, Moliadze V, Paulus W (2010) Simply longer is not better: reversal of theta burst
after-effect with prolonged stimulation. Experimental brain research 204:181–187. [PubMed:
20567808]
Han H, Yao C, Fu Y, Yu Y, Zhang Y, Xu S (2017) Semantic fingerprints-based author name
disambiguation in Chinese documents. Scientometrics 111:1879–1896.
Wang et al. Page 16
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Hanslmayr S, Spitzer B, Bauml KH (2009) Brain oscillations dissociate between semantic and
nonsemantic encoding of episodic memories. Cereb Cortex 19:1631–1640. [PubMed: 19001457]
Higo T, Mars RB, Boorman ED, Buch ER, Rushworth MF (2011) Distributed and causal influence of
frontal operculum in task control. Proc Natl Acad Sci U S A 108:4230–4235. [PubMed:
21368109]
Hsieh LT, Ranganath C (2014) Frontal midline theta oscillations during working memory maintenance
and episodic encoding and retrieval. Neuroimage 85:721–729. [PubMed: 23933041]
Huang YZ, Edwards MJ, Rounis E, Bhatia KP, Rothwell JC (2005) Theta burst stimulation of the
human motor cortex. Neuron 45:201–206. [PubMed: 15664172]
Ibriyamova F, Kogan S, Salganik-Shoshan G, Stolin D (2017) Using semantic fingerprinting in
finance. Applied Economics 49:2719–2735.
Kohler S, Paus T, Buckner RL, Milner B (2004) Effects of left inferior prefrontal stimulation on
episodic memory formation: a two-stage fMRI-rTMS study. Journal of cognitive neuroscience
16:178–188. [PubMed: 15068590]
Kriegeskorte N, Kievit RA (2013) Representational geometry: integrating cognition, computation, and
the brain. Trends Cogn Sci 17:401–412. [PubMed: 23876494]
Lambon Ralph MA, Pobric G, Jefferies E (2009) Conceptual knowledge is underpinned by the
temporal pole bilaterally: convergent evidence from rTMS. Cereb Cortex 19:832–838. [PubMed:
18678765]
Lee TG, D’Esposito M (2012) The dynamic nature of top-down signals originating from prefrontal
cortex: a combined fMRI-TMS study. The Journal of neuroscience : the official journal of the
Society for Neuroscience 32:15458–15466. [PubMed: 23115183]
Luber B, Kinnunen LH, Rakitin BC, Ellsasser R, Stern Y, Lisanby SH (2007) Facilitation of
performance in a working memory task with rTMS stimulation of the precuneus: frequency- and
time-dependent effects. Brain Res 1128:120–129. [PubMed: 17113573]
Martin CB, Douglas D, Newsome RN, Man LL, Barense MD (2018) Integrative and distinctive coding
of visual and conceptual object features in the ventral visual stream. Elife 7.
Martin JL, Barbanoj MJ, Schlaepfer TE, Thompson E, Perez V, Kulisevsky J (2003) Repetitive
transcranial magnetic stimulation for the treatment of depression. Systematic review and meta-
analysis. The British journal of psychiatry : the journal of mental science 182:480–491. [PubMed:
12777338]
Meinzer M, Lindenberg R, Antonenko D, Flaisch T, Floel A (2013) Anodal transcranial direct current
stimulation temporarily reverses age-associated cognitive decline and functional brain activity
changes. The Journal of neuroscience : the official journal of the Society for Neuroscience
33:12470–12478. [PubMed: 23884951]
Mottaghy FM, Gangitano M, Sparing R, Krause BJ, Pascual-Leone A (2002) Segregation of areas
related to visual working memory in the prefrontal cortex revealed by rTMS. Cereb Cortex
12:369–375. [PubMed: 11884352]
Muellbacher W, Ziemann U, Boroojerdi B, Hallett M (2000) Effects of low-frequency transcranial
magnetic stimulation on motor excitability and basic motor behavior. Clinical neurophysiology :
official journal of the International Federation of Clinical Neurophysiology 111:1002–1007.
[PubMed: 10825706]
Mumford JA, Turner BO, Ashby FG, Poldrack RA (2012) Deconvolving BOLD activation in event-
related designs for multivoxel pattern classification analyses. Neuroimage 59:2636–2643.
[PubMed: 21924359]
Nelson DL, McEvoy CL, Schreiber TA (2004) The University of South Florida free association,
rhyme, and word fragment norms. Behav Res Methods Instrum Comput 36:402–407. [PubMed:
15641430]
Nyffeler T, Wurtz P, Luscher HR, Hess CW, Senn W, Pflugshaupt T, von Wartburg R, Luthi M, Muri
RM (2006) Repetitive TMS over the human oculomotor cortex: comparison of 1-Hz and theta
burst stimulation. Neuroscience letters 409:57–60. [PubMed: 17049743]
Nyhus E, Curran T (2010) Functional role of gamma and theta oscillations in episodic memory.
Neurosci Biobehav Rev 34:1023–1035. [PubMed: 20060015]
Wang et al. Page 17
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Olson IR, Ploaker A, Ezzyat Y (2007) The Enigmatic temporal pole: a review of findings on social and
emotional processing. Brain 130:1718–1731. [PubMed: 17392317]
Osipova D, Takashima A, Oostenveld R, Fernandez G, Maris E, Jensen O (2006) Theta and gamma
oscillations predict encoding and retrieval of declarative memory. The Journal of neuroscience :
the official journal of the Society for Neuroscience 26:7523–7531. [PubMed: 16837600]
Patterson K, Nestor PJ, Rogers TT (2007) Where do you know what you know? The representation of
semantic knowledge in the human brain. Nat Rev Neurosci 8:976–987. [PubMed: 18026167]
Plow EB, Cattaneo Z, Carlson TA, Alvarez GA, Pascual-Leone A, Battelli L (2014) The compensatory
dynamic of inter-hemispheric interactions in visuospatial attention revealed using rTMS and fMRI.
Frontiers in human neuroscience 8:226. [PubMed: 24860462]
Ritchey M, Wing EA, LaBar KS, Cabeza R (2013) Neural similarity between encoding and retrieval is
related to memory via hippocampal interactions. Cereb Cortex 23:2818–2828. [PubMed:
22967731]
Robertson EM, Tormos JM, Maeda F, Pascual-Leone A (2001) The role of the dorsolateral prefrontal
cortex during sequence learning is specific for spatial information. Cereb Cortex 11:628–635.
[PubMed: 11415965]
Rose NS, LaRocque JJ, Riggall AC, Gosseries O, Starrett MJ, Meyering EE, Postle BR (2016)
Reactivation of latent working memories with transcranial magnetic stimulation. Science
354:1136–1139. [PubMed: 27934762]
Sandrini M, Cappa SF, Rossi S, Rossini PM, Miniussi C (2003) The role of prefrontal cortex in verbal
episodic memory: rTMS evidence. Journal of cognitive neuroscience 15:855–861. [PubMed:
14511538]
Schneider SA, Pleger B, Draganski B, Cordivari C, Rothwell JC, Bhatia KP, Dolan RJ (2010)
Modulatory effects of 5Hz rTMS over the primary somatosensory cortex in focal dystonia--an
fMRI-TMS study. Mov Disord 25:76–83. [PubMed: 20058321]
Schott BH, Wustenberg T, Wimber M, Fenker DB, Zierhut KC, Seidenbecher CI, Heinze HJ, Walter H,
Duzel E, Richardson-Klavehn A (2013) The relationship between level of processing and
hippocampal-cortical functional connectivity during episodic memory formation in humans. Hum
Brain Mapp 34:407–424. [PubMed: 22042493]
Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weiller E, Hergueta T, Baker R, Dunbar
GC (1998) The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and
validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin
Psychiatry 59 Suppl 20:22–33;quiz 34–57.
Slotnick SD (2017) Cluster success: fMRI inferences for spatial extent have acceptable false-positive
rates. Cogn Neurosci 8:150–155. [PubMed: 28403749]
Slotnick SD, Moo LR, Segal JB, Hart J, Jr (2003) Distinct prefrontal cortex activity associated with
item memory and source memory for visual shapes. Cognitive Brain Research 17:75–82.
[PubMed: 12763194]
Sole-Padulles C, Bartres-Faz D, Junque C, Clemente IC, Molinuevo JL, Bargallo N, Sanchez-Aldeguer
J, Bosch B, Falcon C, Valls-Sole J (2006) Repetitive transcranial magnetic stimulation effects on
brain function and cognition among elders with memory dysfunction. A randomized sham-
controlled study. Cereb Cortex 16:1487–1493. [PubMed: 16339086]
Solomon EA, Kragel JE, Sperling MR, Sharan A, Worrell G, Kucewicz M, Inman CS, Lega B, Davis
KA, Stein JM, Jobst BC, Zaghloul KA, Sheth SA, Rizzuto DS, Kahana MJ (2017) Widespread
theta synchrony and high-frequency desynchronization underlies enhanced cognition. Nature
communications 8:1704.
Summerfield C, Mangels JA (2005) Coherent theta-band EEG activity predicts item-context binding
during encoding. Neuroimage 24:692–703. [PubMed: 15652304]
Summerfield C, Greene M, Wager T, Egner T, Hirsch J, Mangels J (2006) Neocortical connectivity
during episodic memory formation. PLoS Biol 4:e128. [PubMed: 16605307]
Thut G, Nietzel A, Brandt SA, Pascual-Leone A (2006) alpha-Band electroencephalographic activity
over occipital cortex indexes visuospatial attention bias and predicts visual target detection. J
Neurosci 26:9494–9502. [PubMed: 16971533]
Wang et al. Page 18
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Tyler LK, Chiu S, Zhuang J, Randall B, Devereux BJ, Wright P, Clarke A, Taylor KI (2013) Objects
and categories: feature statistics and object processing in the ventral stream. Journal of cognitive
neuroscience 25:1723–1735. [PubMed: 23662861]
van de Ven V, Sack AT (2013) Transcranial magnetic stimulation of visual cortex in memory: cortical
state, interference and reactivation of visual content in memory. Behav Brain Res 236:67–77.
[PubMed: 22921373]
van de Ven V, Jacobs C, Sack AT (2012) Topographic contribution of early visual cortex to short-term
memory consolidation: a transcranial magnetic stimulation study. The Journal of neuroscience :
the official journal of the Society for Neuroscience 32:4–11. [PubMed: 22219265]
Vidal-Pineiro D, Martin-Trias P, Arenaza-Urquijo EM, Sala-Llonch R, Clemente IC, Mena-Sanchez I,
Bargallo N, Falcon C, Pascual-Leone A, Bartres-Faz D (2014) Task-dependent activity and
connectivity predict episodic memory network-based responses to brain stimulation in healthy
aging. Brain Stimul 7:287–296. [PubMed: 24485466]
Wang JX, Voss JL (2015) Long-lasting enhancements of memory and hippocampal-cortical functional
connectivity following multiple-day targeted noninvasive stimulation. Hippocampus 25:877–883.
[PubMed: 25639205]
Webber FDS (2015) Semantic Folding Theory And its Application in Semantic Fingerprinting. arXiv
preprint arXiv:151108855.
Wing EA, Ritchey M, Cabeza R (2015) Reinstatement of individual past events revealed by the
similarity of distributed activation patterns during encoding and retrieval. Journal of cognitive
neuroscience 27:679–691. [PubMed: 25313659]
Wolff MJ, Jochim J, Akyurek EG, Stokes MG (2017) Dynamic hidden states underlying working-
memory-guided behavior. Nat Neurosci 20:864–871. [PubMed: 28414333]
Wang et al. Page 19
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Figure 1. Timeline of imaging protocol.
Wang et al. Page 20
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Figure 2. Localization of memory-specific rTMS targets, and univariate effects of stimulation.
(A) A functional memory-success localizer collected on Day 1 in individual participants; (B)
these subject specific peaks demonstrated a reasonable overlap across participants. (C)
Univariate BOLD effects for subsequently Remembered and Forgotten trials during
Baseline, 1Hz, and 5Hz conditions reveal a reliable excitation/suppression of the subsequent
memory effect after 5Hz/1Hz rTMS. No effect was observed in the contralateral cortex.
Wang et al. Page 21
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Figure 3.
Regions exhibiting significant RSA effects during (A) encoding and (B) retrieval. Model
RDMs based on sentential (A) or word pair similarity (B) are shown to the left, while at
right are graphs representing mean 2nd order correlations between model RDMs and brain
RDMs.
Wang et al. Page 22
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Figure 4. Encoding-Retrieval Similarity after 5Hz or 1Hz rTMS.
The schematic on the left describes encoding-retrieval items to be compared (yellow cells)
versus set-level matches between unmatched pairs (light purple cells). These set-level
matches serve as a reasonable comparison for ERS effects. At right, mean pattern-to-pattern
similarity across stimulation conditions for item- and set-level matches.
Wang et al. Page 23
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Figure 5. Relationships between Left PFC stimulation site and MTL ERS effects.
(A) The increase in univariate activity associated with 5Hz rTMS (see Figure 2) served to
boost ERS in left hippocampus after 5Hz rTMS. (B) Functional connectivity estimates based
on the cPPI analysis during the encoding period, which demonstrate a selective increase in
PFC-MTL connectivity during successfully remembered > forgotten trials, and effect present
only in the 5Hz rTMS condition.
Wang et al. Page 24
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Wang et al. Page 25
Table 1.
Behavioral performance
Accuracy H (SE) FA (SE) d' (SE)
Baseline 0.73 0.04 0.20 0.04 1.78 0.23
1Hz rTMS 0.80 0.05 0.10 0.03 2.36 0.20
5Hz rTMS 0.85 0.03 0.15 0.03 2.29 0.22
Response Time H (SE) FA (SE) Slope/trial* (SE)
Baseline 1723 101 2496 121 −0.0006 0.004
1Hz rTMS 1645 105 2424 124 −0.0018 0.002
5Hz rTMS 1576 108 2237 110 0.0031 0.006
*
Note: Slope was calculated as the slope of a regression line fit to all responses during encoding for each stimulation condition. Values shown here
represent a mean across participants.
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Wang et al. Page 26
Table 2.
Significant RSA clusters.
Encoding RSA
Contrast Region k Z-score x y z
Baseline Inferior Temporal Gyrus 27 3.09 −42 −8 −34
5Hz > 1Hz Fusiform Gyrus 68 3.22 −36 −58 −4
Inferior Temporal Gyrus*
1.03
Retrieval RSA
Condition Region k Z-score x y z
Baseline Inferior Temporal Gyrus 22 2.99 48 −10 −40
5Hz > 1Hz Left Precentral gyrus 107 3.73 −16 −24 66
Left Anterior PHC 92 3.59 32 −22 −16
Left ATL 137 3.30 −38 4 −44
Posterior cingulate 24 3.01 12 −16 38
Inferior Temporal Gyrus*
1.59
Note: A mask (p < 0.005) for the Inferior Temporal Gyrus is included in both Encoding and Retrieval RSA contrasts for 5Hz > 1Hz in order to
determine the extent of stimulation-related modulation in Baseline regions.
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
Author Manuscript Author Manuscript Author Manuscript Author Manuscript
Wang et al. Page 27
Table 3.
Significant ERS clusters.
ERS
Contrast Region k Z-score x y z Reg. t
Baseline Lateral Occipital Cortex 20 3.03 42 −74 12
5Hz > 1Hz Right Superior temporal gyrus 33 3.49 48 0 −18 2.71
Left Hippocampus/Amygdala 42 2.94 −18 0 −24 1.91
Lateral Occipital Cortex*
0.94
Notes: Regression sig. (see methods):
p = .01 one tailed
p = .04 one tailed.
A mask (p < 0.005) for the Lateral Occipital Cortex is included in contrasts for 5Hz > 1Hz in order to determine the extent of stimulation-related
modulation in Baseline regions.
Cogn Neurosci
. Author manuscript; available in PMC 2019 September 05.
... rTMS of left dlPFC has produced moderately consistent effects on RSFC but less consistent cognitive outcomes. Regarding cognitive changes associated with left dlPFC rTMS, eight of twelve studies reviewed here reported significant cognitive improvements associated with high-frequency stimulation [14,33,34,37,38,40,41,[43][44][45]70,71]. Further, one study using low-frequency rTMS reported acute cognitive impairment [42]. ...
... In these two studies [40,41], rTMS was associated with changes in RSFC or EEG variables. Further, Davis and colleagues observed that RSFC changes were associated with better cognitive performance, including increased representational similarity during encoding and retrieval during a memory task [41,43]. This association is consistent with a mechanistic explanation for cognitive effects of rTMS: rTMS affects the activity or RSFC of brain tissue (for some period of time after stimulation) with consequences for associated cognitive abilities [72]. ...
... Very few published studies have measured whether different rTMS frequencies with the same expected activation valence (e.g., high-frequency, 10 Hz vs 20 Hz) produce different effects. Instead, published work has more often contrasted high and low frequencies or the same stimulation frequency at one stimulation location versus another [41,43,49]. This gap in the literature may be important because the few publications on the topic suggest that varying stimulation frequency can affect cognitive outcomes. ...
Article
Full-text available
Human memory systems are imperfect recording devices that are affected by age and disease, but recent findings suggest that the performance of these systems may be modifiable through interventions using non-invasive brain stimulation such as repetitive transcranial magnetic stimulation (rTMS). The translational potential of these rTMS interventions is clear: memory problems are the most common cognitive complaint associated with healthy aging, while pathological conditions such as Alzheimer’s disease are often associated with severe deficits in memory. Therapies to improve memory or treat memory loss could enhance independence while reducing costs for public health systems. Despite this promise, several important factors limit the generalizability and translational potential of rTMS interventions for memory. Heterogeneity of protocol design, rTMS parameters, and outcome measures present significant challenges to interpretation and reproducibility. However, recent advances in cognitive neuroscience, including rTMS approaches in addition to a new understanding of functional brain networks and related insights, may offer methodological tools necessary to design new interventional studies with enhanced experimental rigor, improved reproducibility, and greater likelihood of successful translation to clinical settings. In this review, we first discuss the current state of the literature on memory modulation with rTMS, then offer a commentary on developments in cognitive neuroscience that are relevant to rTMS interventions, and finally close by offering several recommendations for the design of future investigations using rTMS to modulate human memory performance.
... These results suggested that rTMS over the PFC directly induced the neural plasticity in the PFC, and representations in the hippocampus were boosted by indirect manipulation. Consistent with our study, Wang et al. (2014Wang et al. ( , 2018 found that the targeted enhancement of PFC-hippocampal brain networks improved associative semantic and episodic memory performance, which involved localized long-term plasticity. The roles of the hippocampus and PFC in memory processing, individually or in concert, are a major topic of interest in memory research, and PFC and hippocampus are strongly connected by direct and indirect pathways (Eichenbaum, 2017). ...
... Repetitive transcranial magnetic stimulation-induced upregulation of NR1 in the prefrontal cortex, hippocampus, and primary motor cortex correlates with cognitive enhancement A number of studies have shown that rTMS-induced neural plasticity in the PFC or hippocampus correlated with cognition enhancement (Wang et al., , 2018. Recently, Li et al. (2019) reported that 6 weeks of HF-rTMS treatment over the left DLPFC improved long-term potentiation-like plasticity in M1 in AD patients, and the long-term potentiationlike plasticity improvement in M1 correlated to the observed cognition change. ...
Article
Full-text available
Background Repetitive transcranial magnetic stimulation (rTMS) is an effective way to stimulate changes in structural and functional plasticity, which is a part of learning and memory. However, to our knowledge, rTMS-induced specific activity and neural plasticity in different brain regions that affect cognition are not fully understood; nor are its mechanisms. Therefore, we aimed to investigate rTMS-induced cognition-related neural plasticity changes and their mechanisms in different brain regions. Methods A total of 30 healthy adult rats were randomly divided into the control group and the rTMS group ( n = 15 rats per group). The rats in the control and the rTMS group received either 4 weeks of sham or high-frequency rTMS (HF-rTMS) over the prefrontal cortex (PFC). Cognitive function was detected by Morris water maze. Functional imaging was acquired by resting-state functional magnetic resonance imaging (rs-fMRI) before and after rTMS. The protein expressions of BDNF, TrkB, p-Akt, Akt, NR1, NR2A, and NR2B in the PFC, hippocampus, and primary motor cortex (M1) were detected by Western blot following rTMS. Results After 4 weeks of rTMS, the cognitive ability of healthy rats who underwent rTMS showed a small but significant behavioral improvement in spatial episodic learning and memory performance. Compared with the pre-rTMS or the control group, rats in the rTMS group showed increased regional homogeneity (ReHo) in multiple brain regions in the interoceptive/default mode network (DMN) and cortico-striatal-thalamic network, specifically the bilateral PFC, bilateral hippocampus, and the left M1. Western blot analyses showed that rTMS led to a significant increase in the expressions of N -methyl- D -aspartic acid (NMDA) receptors, including NR1, NR2A, and NR2B in the PFC, hippocampus, and M1, as well as an upregulation of BDNF, TrkB, and p-Akt in these three brain regions. In addition, the expression of NR1 in these three brain regions correlated with rTMS-induced cognitive improvement. Conclusion Overall, these data suggested that HF-rTMS can enhance cognitive performance through modulation of NMDA receptor-dependent brain plasticity.
... Evidence supports transsynaptic action of superficial TMS into deep brain regions. TMS studies utilizing concurrent neuroimaging with fMRI ( Baudewig et al., 2001 ;Bestmann et al., 2003 ;Roberts et al., 1997 ;Bohning et al., 1998 ;Davis et al., 2017 ;Tik et al., 2017 ), PET ( Fox et al., 1997 ;Ferrarelli et al., 2004 ), and EEG ( Ilmoniemi et al., 1997 ), have demonstrated that TMS stimulation of superficial cortical targets can re-sult in measurable effects at remote sites including those that are deep relative to the surface of the brain such as the hippocampus and other limbic regions ( Denslow et al., 2005 ;Li et al., 2004 ;Bestmann et al., 2004 ;Wang et al., 2018 ). For example, TMS of the motor cortex induced measurable dopamine release in the putamen while rTMS of the DLPFC induced dopamine release in the head of the caudate ( Strafella et al., 2003( Strafella et al., , 2005Cho and Strafella, 2009 ). ...
Article
TMS has become a powerful tool to explore cortical function, and in parallel has proven promising in the development of therapies for various psychiatric and neurological disorders. Unfortunately, much of the inference of the direct effects of TMS has been assumed to be limited to the area a few centimeters beneath the scalp, though clearly more distant regions are likely to be influenced by structurally connected stimulation sites. In this study, we sought to develop a novel paradigm to individualize TMS coil placement to non-invasively achieve activation of specific deep brain targets of relevance to the treatment of psychiatric disorders. In ten subjects, structural diffusion imaging tractography data were used to identify an accessible cortical target in the right frontal pole that demonstrated both anatomic and functional connectivity to right Brodmann area 25 (BA25). Concurrent TMS-fMRI interleaving was used with a series of single, interleaved TMS pulses applied to the right frontal pole at four intensity levels ranging from 80% to 140% of motor threshold. In nine of ten subjects, TMS to the individualized frontal pole sites resulted in significant linear increase in BOLD activation of BA25 with increasing TMS intensity. The reliable activation of BA25 in a dosage-dependent manner suggests the possibility that the careful combination of imaging with TMS can make use of network properties to help overcome depth limitations and allow noninvasive brain stimulation to influence deep brain structures.
... The DLPFC plays an important role in circuits underlying working memory [103] and relational memory binding for longterm memory (LTM) [104]. Wang et al. applied 5-Hz rTMS for 10 min over the left DLPFC, which is important in controlling relational memory binding, prior to an associative LTM task and observed changes in memory circuit excitability and connectivity with fMRI in healthy older adults [105]. They showed that increased local PFC excitability can have measurable influences on upstream semantic representations in earlier temporal memory regions of a memory circuit. ...
Article
More than any other brain region, the prefrontal cortex (PFC) gives rise to the singularity of human experience. It is therefore frequently implicated in the most distinctly human of all disorders, those of mental health. Noninvasive neuromodulation, including electroconvulsive therapy (ECT), repetitive transcranial magnetic stimulation (rTMS), and transcranial direct current stimulation (tDCS) among others, can—unlike pharmacotherapy—directly target the PFC and its neural circuits. Direct targeting enables significantly greater on-target therapeutic effects compared with off-target adverse effects. In contrast to invasive neuromodulation approaches, such as deep-brain stimulation (DBS), noninvasive neuromodulation can reversibly modulate neural activity from outside the scalp. This combination of direct targeting and reversibility enables noninvasive neuromodulation to iteratively change activity in the PFC and its neural circuits to reveal causal mechanisms of both disease processes and healthy function. When coupled with neuronavigation and neurophysiological readouts, noninvasive neuromodulation holds promise for personalizing PFC neuromodulation to relieve symptoms of mental health disorders by optimizing the function of the PFC and its neural circuits. ClinicalTrials.gov Identifier: NCT03191058.
... However, we contend that global ERS is an appropriate and theoretically meaningful baseline assessment of recapitulation between younger and older adults when assessing single-item ERS. Regarding future directions, recent work suggests that applying excitatory transcranial magnetic stimulation to dorsolateral prefrontal cortex increases pattern similarity in visual cortex during encoding and retrieval, as well as single-item ERS in the hippocampus in younger adults during an episodic memory task (Wang et al., 2018). An exciting avenue for future research is therefore to . ...
Preprint
Full-text available
Typical aging is associated with increases in false memory rates among older adults. Such errors are frequently associated with differential neural activity during encoding and retrieval in older compared to younger adults within visual cortices, hippocampus, and front-parietal regions. It remains unknown how pattern similarity reductions relate to false memories in healthy aging. Using encoding-retrieval similarity (ERS) analyses in a sample of younger and older adults, we examined how the similarity of neural patterns between memory phases associated with target and lure objects was impacted by age and contributed to false memory rates. Single-item ERS for targets and lures was reduced by age throughout much of the ventral visual stream and the posterior hippocampus. Furthermore, ERS associated with perceptual lures within the visual stream maintained differential relationships with false memory. Finally, a global ERS metric accounted for age deficits in single-item ERS, but did not contribute to false memory rates. These findings highlight the contribution of age-related reductions in ERS across multiple representational levels to false memories in healthy aging.
Article
Effectiveness of cognitive test “n-back” performing was assessed in 32 subjects, right-handed men, 18—21 years old. EGG was recorded monopolarly in “10—20” system during initial state, in the period after the first learning stage and during the second learning stage while the subjects performed the test on their own. Two groups of subjects were marked, basing on differences between efficiency of their cognitive activity. Spectral analysis revealed none differences in theta-band spectral power between subjects of the two groups at the initial state. Successful subjects demonstrated increase of the theta-band spectral power in occipital, parietal, central and left frontal zones of cortex in the period after the first learning stage comparing to unsuccessful subjects. On the contrary, unsuccessful subjects demonstrated significant increase of the theta-band spectral power in the second learning stage during the test self-performing comparing to the initial state and the period after the first learning stage. Increase of spectral power of the theta-band can be considered as a regression stage of systemogenesis, accompanied by increase of the representation of low-differentiated systems in new behavior realization and causing high efficiency of cognitive activity.
Article
Typical aging is associated with increases in false memory rates among older adults. Such errors are frequently associated with differential neural activity during encoding and retrieval in older compared to younger adults within visual cortices and the hippocampus. It remains unknown how pattern similarity reductions relate to false memories in healthy aging. Using encoding-retrieval similarity (ERS) analyses in a sample of younger and older adults, we examined how the similarity of neural patterns between memory phases associated with target and lure objects was impacted by age and contributed to false memory rates. Single-item ERS for targets and lures was reduced by age throughout much of the ventral visual stream and the posterior hippocampus. Furthermore, ERS associated with perceptual lures within the visual stream maintained differential relationships with false memory. Finally, a global ERS metric accounted for age deficits in single-item ERS, but did not contribute to false memory rates. These findings highlight the contribution of age-related reductions in ERS across multiple representational levels to false memories in healthy aging.
Article
Although a large proportion of the lexicon consists of abstract concepts, little is known about how they are represented by the brain. Here, we investigated how the mind represents relations shared between sets of mental representations that are superficially unrelated, such as car–engine and dog–tongue, but that nonetheless share a more general, abstract relation, such as whole–part. Participants saw a pair of words on each trial and were asked to indicate whether they could think of a relation between them. Importantly, they were not explicitly asked whether different word pairs shared the same relation, as in analogical reasoning tasks. We observed representational similarity for abstract relations in regions in the “conceptual hub” network, even when controlling for semantic relatedness between word pairs. By contrast, we did not observe representational similarity in regions previously implicated in explicit analogical reasoning. A given relation was sometimes repeated across sequential word pairs, allowing us to test for behavioral and neural priming of abstract relations. Indeed, we observed faster RTs and greater representational similarity for primed than unprimed trials, suggesting that mental representations of abstract relations are transiently activated on this incidental analogy task. Finally, we found a significant correlation between behavioral and neural priming across participants. To our knowledge, this is the first study to investigate relational priming using functional neuroimaging and to show that neural representations are strengthened by relational priming. This research shows how abstract concepts can be brought to mind momentarily, even when not required for task performance.
Article
Prevailing theories of hippocampal function argue that memories are rapidly encoded by non-overlapping memory traces. Concurrently, the hippocampus has been argued to integrate across related experiences, enabling generalization. The cognitive neuroscience of memory has been transformed by the recent proliferation of studies using pattern similarity analyses to investigate the neural substrates of memory in humans, marking an exciting and significant advance in our understanding of population-level neural representations. We provide an overview of hippocampal pattern similarity studies published to date. By considering the effects of stimulus type, time-scale, and hippocampal subregions, we account for both increases and decreases in representational similarity. We argue that hippocampal representations for related memories are not fixed. Instead, the evoked representations are flexibly modulated, depending on whether the current goal is to extract generalities or to reinstate specific experiences. In the first comprehensive review of hippocampal pattern similarity analyses, we provide insight into the mechanisms of memory representation and implications for the interpretation of pattern similarity more generally.
Article
Full-text available
The prefrontal cortex (pFC) is central to flexible, goal-directed cognition, and understanding its representational code is an important problem in cognitive neuroscience. In humans, multivariate pattern analysis (MVPA) of fMRI blood oxygenation level-dependent (BOLD) measurements has emerged as an important approach for studying neural representations. Many previous studies have implicitly assumed that MVPA of fMRI BOLD is just as effective in decoding information encoded in pFC neural activity as it is in visual cortex. However, MVPA studies of pFC have had mixed success. Here we estimate the base rate of decoding information from pFC BOLD activity patterns from a meta-analysis of published MVPA studies. We show that pFC has a significantly lower base rate (55.4%) than visual areas in occipital (66.6%) and temporal (71.0%) cortices and one that is close to chance levels. Our results have implications for the design and interpretation of MVPA studies of pFC and raise important questions about its functional organization.
Article
Full-text available
Memory failures are frustrating and often the result of ineffective encoding. One approach to improving memory outcomes is through direct modulation of brain activity with electrical stimulation. Previous efforts, however, have reported inconsistent effects when using open-loop stimulation and often target the hippocampus and medial temporal lobes. Here we use a closed-loop system to monitor and decode neural activity from direct brain recordings in humans. We apply targeted stimulation to lateral temporal cortex and report that this stimulation rescues periods of poor memory encoding. This system also improves later recall, revealing that the lateral temporal cortex is a reliable target for memory enhancement. Taken together, our results suggest that such systems may provide a therapeutic approach for treating memory dysfunction.
Article
Full-text available
A significant body of research in cognitive neuroscience is aimed at understanding how object concepts are represented in the human brain. However, it remains unknown whether and where the visual and abstract conceptual features that define an object concept are integrated. We addressed this issue by comparing the neural pattern similarities among object-evoked fMRI responses with behavior-based models that independently captured the visual and conceptual similarities among these stimuli. Our results revealed evidence for distinctive coding of visual features in lateral occipital cortex, and conceptual features in the temporal pole and parahippocampal cortex. By contrast, we found evidence for integrative coding of visual and conceptual object features in perirhinal cortex. The neuroanatomical specificity of this effect was highlighted by results from a searchlight analysis. Taken together, our findings suggest that perirhinal cortex uniquely supports the representation of fully specified object concepts through the integration of their visual and conceptual features.
Article
Full-text available
The idea that synchronous neural activity underlies cognition has driven an extensive body of research in human and animal neuroscience. Yet, insufficient data on intracranial electrical connectivity has precluded a direct test of this hypothesis in a whole-brain setting. Through the lens of memory encoding and retrieval processes, we construct whole-brain connectivity maps of fast gamma (30–100 Hz) and slow theta (3–8 Hz) spectral neural activity, based on data from 294 neurosurgical patients fitted with indwelling electrodes. Here we report that gamma networks desynchronize and theta networks synchronize during encoding and retrieval. Furthermore, for nearly all brain regions we studied, gamma power rises as that region desynchronizes with gamma activity elsewhere in the brain, establishing gamma as a largely asynchronous phenomenon. The abundant phenomenon of theta synchrony is positively correlated with a brain region’s gamma power, suggesting a predominant low-frequency mechanism for inter-regional communication.
Article
Full-text available
In an editorial (this issue), I argued that Eklund, Nichols, and Knutsson's 'null data' reflected resting-state/default network activity that inflated their false-positive rates. Commentaries on that paper were received by Nichols, Eklund, and Knutsson (this issue), Hopfinger (this issue), and Cunningham and Koscik (this issue). In this author response, I consider these commentaries. Many issues stemming from Nichols et al. are identified including: (1) Nichols et al. did not provide convincing arguments that resting-state fMRI data reflect null data. (2) Eklund et al. presented one-sample t-test results in the main body of their paper showing that their permutation method was acceptable, while their supplementary results showed that this method produced false-positive rates that were similar to other methods. (3) Eklund et al. used the same event protocol for all the participants, which artifactually inflated the one-sample t-test false-positive rates. (4) At p < .001, using two-sample t-tests (which corrected for the flawed analysis), all the methods employed to correct for multiple comparisons had acceptable false-positive rates. (5) Eklund et al. contrasted resting-state periods, which produced many significant clusters of activity, while null data should arguably be associated with few, if any, significant activations. Eklund et al.'s entire set of results show that commonly employed methods to correct for multiple comparisons have acceptable false-positive rates. Following Hopfinger along with Cunningham and Koscik, it is also highlighted that rather than focusing on only type I error, type I error and type II error should be balanced in fMRI analysis.
Article
Full-text available
Recent theoretical models propose that working memory is mediated by rapid transitions in 'activity-silent' neural states (for example, short-term synaptic plasticity). According to the dynamic coding framework, such hidden state transitions flexibly configure memory networks for memory-guided behavior and dissolve them equally fast to allow forgetting. We developed a perturbation approach to measure mnemonic hidden states in an electroencephalogram. By 'pinging' the brain during maintenance, we show that memory-item-specific information is decodable from the impulse response, even in the absence of attention and lingering delay activity. Moreover, hidden memories are remarkably flexible: an instruction cue that directs people to forget one item is sufficient to wipe the corresponding trace from the hidden state. In contrast, temporarily unattended items remain robustly coded in the hidden state, decoupling attentional focus from cue-directed forgetting. Finally, the strength of hidden-state coding predicts the accuracy of working-memory-guided behavior, including memory precision.
Article
A growing literature has focused on the brain's ability to augment processing in local regions by recruiting distant communities of neurons in response to neural decline or insult. In particular, both younger and older adult populations recruit bilateral prefrontal cortex (PFC) as a means of compensating for increasing neural effort to maintain successful cognitive function. However, it remains unclear how local changes in neural activity affect the recruitment of this adaptive mechanism. To address this problem, we combined graph theoretical measures from functional MRI with diffusion weighted imaging and repetitive transcranial magnetic stimulation (rTMS) to resolve a central hypothesis: how do aged brains flexibly adapt to local changes in cortical activity? Specifically, we applied neuromodulation to increase or decrease local activity in a cortical region supporting successful memory encoding (left dorsolateral PFC or DLPFC) using 5 or 1 Hz rTMS, respectively. We then assessed a region's local within-module degree, or the distributed between-module degree (BMD) between distant cortical communities. We predicted that (1) local stimulation-related deficits may be counteracted by boosting BMD between bilateral PFC, and that this effect should be (2) positively correlated with structural connectivity. Both predictions were confirmed; 5 Hz rTMS increased local success-related activity and local increases in PFC connectivity, while 1 Hz rTMS decreases local activity and triggered a more distributed pattern of bilateral PFC connectivity to compensate for this local inhibitory effect. These results provide an integrated, causal explanation for the network interactions associated with successful memory encoding in older adults. Hum Brain Mapp, 2017. © 2017 Wiley Periodicals, Inc.
Article
Author name disambiguation is an important problem that needs to be resolved in bibliometric analysis or tech mining. Many techniques have been presented; however, most of them require a long run time or additional information. A new method based on semantic fingerprints was presented to disambiguate author names without external data. A manually annotated dataset was built to testify on the efficiency of the presented method. Experiments using co-author features, institution features, and text fingerprints were conducted respectively. We found that the first two methods had higher precision, but their recall was low, and the text fingerprint method had higher recall and satisfied precision. Based on these results, we integrated co-author features, institution features, and text fingerprints to provide semantic fingerprints for disambiguating author names and achieving better performance on the F-measure.
Article
People often forget information because they fail to effectively encode it. Here, we test the hypothesis that targeted electrical stimulation can modulate neural encoding states and subsequent memory outcomes. Using recordings from neurosurgical epilepsy patients with intracranially implanted electrodes, we trained multivariate classifiers to discriminate spectral activity during learning that predicted remembering from forgetting, then decoded neural activity in later sessions in which we applied stimulation during learning. Stimulation increased encoding-state estimates and recall if delivered when the classifier indicated low encoding efficiency but had the reverse effect if stimulation was delivered when the classifier indicated high encoding efficiency. Higher encoding-state estimates from stimulation were associated with greater evidence of neural activity linked to contextual memory encoding. In identifying the conditions under which stimulation modulates memory, the data suggest strategies for therapeutically treating memory dysfunction.