- Email: [email protected]
Successful memory encoding is associated with increased cross-frequency coupling between frontal theta and posterior gamma oscillations in human scalp-recorded EEG
NeuroImage 66 (2013) 642–647
Contents lists available at SciVerse ScienceDirect
NeuroImage journal homepage: www.elsevier.com/locate/ynimg
Successful memory encoding is associated with increased cross-frequency coupling between frontal theta and posterior gamma oscillations in human scalp-recorded EEG Uwe Friese a, b,⁎, 1, Moritz Köster b, c, 1, Uwe Hassler b, Ulla Martens b, Nelson Trujillo-Barreto d, Thomas Gruber b a
University Medical Center Hamburg-Eppendorf, Department of Neurophysiology and Pathophysiology, Hamburg, Germany Institute of Psychology, University of Osnabrueck, Germany c Institute of Cognitive Science, University of Osnabrueck, Germany d Cuban Neuroscience Center, Havana, Cuba b
a r t i c l e
i n f o
Article history: Accepted 1 November 2012 Available online 8 November 2012 Keywords: EEG Oscillations Cross-frequency-coupling Gamma-band Theta-band Memory encoding
a b s t r a c t Although previous studies have established that successful memory encoding is associated with increased synchronization of theta-band and gamma-band oscillations, it is unclear if there is a functional relationship between oscillations in these frequency bands. Using scalp-recorded EEG in healthy human participants, we demonstrate that cross-frequency coupling between frontal theta phase and posterior gamma power is enhanced during the encoding of visual stimuli which participants later on remember versus items which participants subsequently forget (“subsequent memory effect,” SME). Conventional wavelet analyses and source localizations revealed SMEs in spectral power of theta-, alpha-, and gamma-band. Successful compared to unsuccessful encoding was reflected in increased theta-band activity in right frontal cortex as well as increased gamma-band activity in parietal-occipital regions. Moreover, decreased alpha-band activity in prefrontal and occipital cortex was also related to successful encoding. Overall, these findings support the idea that during the formation of new memories frontal cortex regions interact with cortical representations in posterior areas. © 2012 Elsevier Inc. All rights reserved.
Introduction Impressive evidence has been accumulated that synchronized neuronal activity serves the integration of spatially distributed processing in the brain (Engel et al., 2001; Varela et al., 2001). In particular, oscillations in the gamma band range ( > 30 Hz) are thought to reflect processes related to the activation and maintenance of neuronal object representations (Jensen et al., 2007). Low frequency oscillations in the theta band,~ 5 Hz, and in the alpha band, 8–12 Hz, have been related to top-down control modulating the processing of neuronal object representations (Palva et al., 2005). Mostly based on animal research or studies with invasive recordings in epilepsy patients, several authors have suggested that the coordinated interplay of low- and high-frequency oscillations is likely to be crucial for various memory processes (Canolty and Knight, 2010; Fell and Axmacher, 2011; Freunberger et al., 2011; Nyhus and Curran, 2010; Palva and
⁎ Corresponding author at: University Medical Center Hamburg-Eppendorf, Department of Neurophysiology and Pathophysiology, Martinistraße 52, D-20246 Hamburg, Germany. Fax: +49 40 7410 57752. E-mail address: [email protected] (U. Friese). 1 These authors contributed equally to this work. 1053-8119/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.neuroimage.2012.11.002
Palva, 2007). So far, only a few studies using electroencephalography (EEG) or magnetencephalography (MEG) in humans have demonstrated cross-frequency coupling between low- and high-frequency oscillations in working memory and attention tasks (Demiralp et al., 2007; Palva et al., 2005; Sauseng et al., 2008, 2009). In the present study, we employed a subsequent memory paradigm to investigate whether low- and high-frequency oscillations in scalp-recorded EEG data interact during successful encoding of new memories. In a subsequent memory paradigm, neuronal activity during the encoding of stimuli which are remembered in a subsequent memory test, is contrasted versus activity during the encoding of stimuli which are later forgotten (Paller and Wagner, 2002). Previous EEG/MEG studies on oscillatory brain activity have established “subsequent memory effects” (SME) in various frequency bands. Successful memory encoding has been found to be accompanied by increased synchrony in the theta and gamma band as well as by decreased synchrony in the alpha and beta band (Gruber et al., 2004; Hanslmayr et al., 2009, 2011; Klimesch et al., 1996a, 1996b; Osipova et al., 2006). Most importantly, Osipova et al. (2006) found that increased frontal theta power coincided with enhanced gamma power in posterior cortex regions for later remembered versus later forgotten stimuli. The authors speculated that frontal theta oscillations might reflect top-down processes
U. Friese et al. / NeuroImage 66 (2013) 642–647
modulating gamma band activity related to neuronal representations in posterior regions. However, this hypothesis was not tested directly. On the basis of Osipova et al. (2006), the main objectives of our study were (a) to confirm subsequent memory effects in theta, alpha, and gamma frequency bands, (b) to identify the cortical generators of these effects, and (c) to explore if posterior gamma power is coupled to the phases of frontal theta oscillations during the encoding of new information. This type of cross-frequency coupling would provide strong evidence for the notion that theta-oscillations play a pivotal role in long-range communication between brain regions involved in memory encoding processes. Materials and methods Participants A total of 26 healthy university students (23 female, M = 21.3 years, SD = 3.7) received either monetary compensation or course credits for participation. None of the participants reported any history of neurological or psychiatrical disorder. All participants had normal or corrected-to-normal vision. Informed consent was obtained from all participants according to the Declaration of Helsinki.
643
condition with the lowest number of trials. Data of three participants with less than 15 trials per condition were discarded. For the remaining 23 participants on average 26.1 trials (SD = 7.7 trials) per condition were available.
Data analysis: Spectral changes in electrode space Total spectral changes of oscillatory activity were calculated for each trial using Morlet wavelets (Bertrand and Pantev, 1994) with seven cycles. Further analyses were based on a time window of 400–1300 ms after stimulus onset. This segment was chosen after visual inspection of time-by-frequency plots. Moreover, we wanted to exclude possible interference from early visual responses (Osipova et al., 2008). For broadband gamma oscillations, single-trial power spectra were averaged within 50–80 Hz. As theta and alpha signals vary largely across individuals (Klimesch, 1999), we selected frequencies for each participant individually on the basis of all trials across all conditions. The mean theta frequency was 5.1 Hz (SD =1.2 Hz) and the mean alpha frequency was 10.2 Hz (SD = 1.7 Hz). To evaluate SMEs (SR minus SF) in electrode space, we averaged data for each subject and electrode within the respective frequency band and time frame and subjected individual means to paired sample t-tests.
Stimuli and procedure The stimuli consisted of 150 pictures of living (e.g. plants, animals) and 150 pictures of nonliving objects (e.g. clothes, tools). Pictures covered a visual angle of 6.2° 6.2° and were presented at the center of a 21 inch computer screen. In the encoding session, 200 pictures (100 of each category) were presented in random order and had to be classified as living or nonliving objects. The presentation sequence was blank screen (1 s), fixation point (1.0–1.5 s), target stimulus (2 s), question mark until response. Participants used response keys with different fingers of their right hand. The encoding session was followed by a 15 min filler task in which participants solved mathematical equations. Afterwards 200 pictures from the encoding session were randomly intermixed with 100 new stimuli, and participants were to indicate whether they “remember” the object, they “know” the object, or if the object was not shown in the first part of the experiment and therefore “new”. This so-called remember/know procedure allows to distinguish between familiarity-based recognition and controlled recollection (Tulving, 1985). “Know”-responses are assumed to rely more on familiarity while “remember”-responses are rather associated with recollection. Given our main interest in successful memory encoding processes, we focused on remember-responses to obtain “purer” measures of recollection memory and to enhance SMEs (Paller and Wagner, 2002). Hence, we discarded trials with know-responses and limited EEG data analyses to comparisons of subsequently remembered (SR) and subsequently forgotten (SF) items. Electrophysiological recordings A BioSemi Active-Two amplifier system with 128 active electrodes was used at 512 Hz sampling rate. Additionally, horizontal and vertical electro-oculograms were recorded. Common mode sense (CMS) and driven right leg (DRL) electrodes served as recording reference and ground. EEG data were re-referenced to the average of all electrodes and segmented into epochs from −500 to 2000 ms with regard to stimulus onset. Eye blink artifacts were removed using an independent component based procedure (Delorme and Makeig, 2004). Further artifacts were removed with SCADS (Junghoefer et al., 2000). For the analysis of gamma-band oscillations, we applied a correction of saccade-related transient potentials (COSTRAP; Hassler et al., 2011). This procedure has been shown to remove possible artifacts caused by miniature eye movements (Yuval-Greenberg et al., 2008). For all analyses, the number of trials per condition was matched for each subject according to the
Data analysis: Spectral changes in scource space To localize the generators of oscillatory EEG activity, single-trial VARETA analyses were conducted (for details see Gruber et al., 2006). This method is based on a discrete spline distributed inverse model and reveals the estimates of the primary current densities (PCDs) in source space which are the spatially smoothest solutions compatible with the observed scalp topographies. PCD estimates for SR and SF were subjected to voxel-wise paired sample t-tests. Data analysis: Phase-amplitude coupling For the assessment of phase-amplitude coupling between lowfrequency theta/alpha and high-frequency gamma oscillations, we used the modulation index (MI) described in detail by Tort et al. (2010). The time series data for this analysis were created by averaging data epochs (400–1300 ms) across selected electrodes for SR and SF. Electrode selection was based on significant SMEs in the different frequency bands (as described above). In brief, coupling between low-frequency phase signals and high-frequency amplitude signals was computed as follows. The signals in question were filtered at the respective frequencies, and the Hilbert transform was used to derive phase information of the low-frequency signal as well as the amplitude envelope of the high-frequency signal. Next, a composite time series was constructed which represents the amplitude of the gamma oscillation at each phase of the low-frequency signal. Low frequency phases were binned into twelve 30° intervals (0–360°) and the mean amplitude of the gamma oscillation in each bin was derived. These mean values were normalized by dividing each bin value by the sum over all bins. The resulting “amplitude distributions” are uniform if no phase-amplitude coupling is evident. Deviations from uniformity can then be quantified by means of a discrete and normalized (between 0 and 1) version of the Kullback-Leibler divergence between the empirical phase-to-amplitude distribution and the uniform distribution (i.e. the modulation index, MI). Tort et al. (2010) have shown that the MI does not depend on the absolute level of the signal, and it as also relatively insensitive to changes of noise levels. These properties make the MI especially suited for evaluating the intensity of phaseamplitude coupling. For all subjects, we calculated MI-values separately for SR and SF trials, and tested for differences between conditions using non-parametric Wilcoxon signed rank tests.
644
U. Friese et al. / NeuroImage 66 (2013) 642–647
Results
Cross-frequency coupling
Behavioral data
To test the hypothesis if parietal gamma-band activity was modulated by frontal theta-band activity (Osipova et al., 2006), we first averaged raw signals across electrodes with significant SMEs (see Fig. 2), i.e. right frontal electrodes for the theta phase signals and central parietal electrodes for the gamma amplitude signals. The theta phase signals were band-pass filtered from 5 to 8 Hz, and gamma amplitude signals were band-pass filtered from 50 to 80 Hz using default EEGlab two-way least-squares FIR filtering procedures. Individual modulation indices were calculated for each participants and condition. Fig. 3, left panel, illustrates phase-amplitude relations of two exemplary participants as amplitude distributions for two complete phase cycles (0–720°). For participants P20, the amplitude distributions for forgotten items (left) are more uniform than for remembered items (right), whereas for participant P17 both distributions appear to be largely comparable. More formally, the modulation indices for SR and SF conditions of all participants are plotted in Fig. 3, right panel. Data from the two exemplary participants have been highlighted. For participants located above the diagonal (e.g., P20), the MI was greater in the remember condition than in the forgotten condition. Confirming the visual inspection of the amplitude distribution, the MIs of participant P17 fall close to the diagonal. Overall, MIs in the remembered conditions were significantly larger than in the forgotten condition (z = -2.37, p b 0.02) indicating enhanced cross-frequency phase-amplitude coupling between frontal theta and parietal gamma oscillations during successful memory encoding. It is important to note that the differences in cross-frequency coupling between conditions could be biased by power changes of the phase-modulating frequency. Since theta power was higher in the SR condition, estimations of theta phase might be improved in comparison to the SF condition. Consequently, estimations of cross-frequency coupling could be biased. To rule out this possibility, we assessed the relationship between MIs, theta power, and also gamma power independently from the conditions as follows. Single-trial data of each participant were sorted regardless of condition by mean theta power averaged across the selected electrodes. We divided the sorted data into quartiles such that the first quartile (Q1) contained 25% of all trials with lowest theta power, and the forth quartile (Q4) contained 25% of all trials with highest theta power. For each quartile, we derived MIs, mean theta power averaged across frontal electrodes, and mean gamma power averaged across parietal electrodes. Fig. 4A depicts the corresponding means across participants in quartiles of theta power (Q1 to Q4). Naturally, there is a noticeable increase of mean theta power (gray bars) but also an increase of parietal gamma power (white bars) across quartiles. In contrast, MIs do not differ across quartiles (black squares). Corresponding 1-way analyses of variance across quartiles resulted in significant effects for theta power (F3,93 = 88.0, p b .01, linear trend: F1,31 = 93.7, p b .01) and gamma power (F3,93 = 13.4, p b .01, linear trend: F1,31 = 19.1, p b .01), whereas no significant differences across quartiles were found for MIs (F3,93 b 1). Hence, it is unlikely that our finding of increased cross-frequency coupling in the remembered versus the forgotten condition is due to theta power differences between the conditions per se. Lastly, to explore the cross-frequency coupling pattern at the level of individual electrodes, we chose the frontal electrodes for which theta power was larger for SR versus SF as seeds, and calculated MIs between each seed electrode and all other scalp electrodes. In Fig. 4B, electrode pairs associated with significant differences MISR − MISF (p b .01, two-tailed t-tests) are indicated by solid lines. (No significant results were found for the opposite direction.) These results show that with respect to the selected frontal region, the pattern of increased theta-gamma phase-amplitude coupling is indeed confined to primarily parietal and occipital brain areas.
On average, 48.5% (SEM = 3.6%) of the encoding session items were remembered and 22.1% (SEM =3.6%) were reported new. Counting remember responses to old items as hits and remember responses to new items as false alarms (M = 2.8, SEM= 1.0%), mean discrimination performance for remembered items was significantly larger than chance (d′ = 1.99; t22 = 17.2, p b .001).
SMEs in theta-, alpha-, and gamma-band power Time-by-frequency representations of oscillatory activity are depicted in Fig. 1 as power changes relative to the 200 ms pre-stimulus baseline activity. After stimulus-onset, the average theta power clearly increased, whereas spectral power decreased in the alpha-band. Average gamma power showed a maximum approximately between 50 and 65 Hz throughout the measurement interval. Subsequent memory effects in oscillatory power were evident in theta-, alpha-, and gamma-band. The left hand side of Fig. 2 shows grand mean spherical spline interpolated topographical maps of spectral power (400–1300 ms) separately for SR and SF conditions. At marked electrodes the subsequent memory effect, i.e. the difference SR–SF, was significant with p b .01 (corresponding to a false discovery rate of q b .1; Benjamini and Hochberg, 1995). The right hand side of Fig. 2 illustrates the estimated VARETA source localizations of the SMEs. Statistical maps are thresholded at t-values associated with p b .01 and projected onto the smoothed surface of an MRI template brain. In the theta-band, SMEs were evident at right frontal electrodes with corresponding source localization in the right frontal cortex. Alpha-band suppression was larger for remembered versus forgotten items at two large electrode clusters in bilateral anterior and posterior scalp regions. The sources of these SMEs were found in bilateral middle frontal gyri and in primary visual cortex areas. Gamma-band SMEs occurred at central parietal electrodes with sources located in left parietal and occipital cortex as well as in right inferior parietal cortex.
1.3
18
rel. signal change
Frequency (Hz)
14 10 6 2
0.7 0
400
800
1200
1600 1.06 rel. signal change
Frequency (Hz)
80 70 60 50 40
0.94 0
400
800
1200
1600
Time (ms) Fig. 1. Grand mean time-by-frequency plots of spectral power for frequencies up to 20 Hz and for the gamma range (40–100 Hz). Low frequency activity was averaged across all electrodes, gamma-band activity was averaged across parietal electrodes. Amplitude values are expressed as signal changes relative to the 200 ms pre-stimulus baseline.
U. Friese et al. / NeuroImage 66 (2013) 642–647
Forgotten
Remembered
645
Remembered – Forgotten
Theta
Alpha
Gamma
+ signs: electrodes with significant difference Theta +/- 3, Alpha +/- 6, Gamma +/- 0.4 (µV2)
2.5
Theta/Gamma
4.0
T -2.5
Alpha
-4.0
Fig. 2. The left panel depicts subsequent memory effects in electrode space. Topographical maps of average oscillatory power are shown for theta-, alpha-, and gamma-band (400– 1300 ms) in the forgotten and remember conditions. Amplitude values are expressed as signal change to the 200 ms pre-stimulus baseline. SMEs are associated with p b .01 at marked electrodes. In the right panel, source space SMEs are illustrated as t-values (thresholded at p b .01) projected onto the surface of an average MRI template brain.
Discussion This study provides first strong evidence for a functional link between low- and high-frequency oscillations during the encoding of new memories. Phase-amplitude coupling between frontal theta oscillations and parietal gamma oscillations was found to be larger for subsequently remembered versus subsequently forgotten items. This coupling might reflect a mechanism by which theta-mediated control processes modulate gamma-band related processes implicated in the
reactivation and maintenance of neuronal object representations (Osipova et al., 2006; Palva et al., 2005). Overall, our findings regarding SMEs in theta, alpha, and gamma power show considerable concordance with other EEG/MEG studies. Osipova et al. (2006), using a similar design in an MEG experiment, observed a gamma-band SME at posterior parietal and occipital sensors, as well as a theta-band SME at right temporal sensor sites. In that study, the source of the gamma-band SME could be located in visual cortex whereas no reliable source was found for the theta SME.
frontal theta
parietal gamma
Modulation index (MI) Forgotten
SR > SF
Remembered
P20
P20 P17
P17 SF > SR 0.005
0.005
0.01
MI
0.10
Amplitude
Remembered
0.01
0.05 Phase
0
360 720
Forgotten Fig. 3. The cartoon head illustrates phase-amplitude coupling with frontal theta phase modulating parietal gamma power. The left panel displays modulation indices for remember and forgotten conditions of all participants. Above the diagonal line, MIs are larger for remember versus forgotten (SR >SF), below the diagonal line, MIs are larger for forgotten versus remember (SF > SR). The right panel shows phase-amplitude distributions for two complete phase cycles (0–720°) of the two participants (P17, and P20) in the forgotten and remembered conditions. The extent of cross-frequency coupling can be inferred from deviations of the uniform distribution.
646
U. Friese et al. / NeuroImage 66 (2013) 642–647
A 0.016
Theta
Modulation index (MI)
0.8
Gamma
Power (µV2)
MI
0
Q1
Q2
Q3
Q4
0
B Frontal seed electrodes
Significant CFC: Remembered > Forgotten Fig. 4. (A) Mean theta-gamma modulation indices (MI) and mean theta/gamma power in quartiles of theta power. Calculations were based on all trials across conditions sorted by theta power at frontal electrodes. Bars illustrate increase of power in theta band (gray) and gamma band (white) over quartiles (Q1-Q4). Black squares indicate that corresponding cross-frequency modulation indices (MI) do not increase with theta power. (B) Theta-gamma phase-amplitude coupling was calculated between frontal seed electrodes and all other scalp electrodes. Solid lines indicate significant cross-frequency coupling differences (SR > SF) between electrode pairs (p b 0.01, uncorrected).
We identified sources in posterior parietal cortex regions for the gamma-band SME and right frontal cortex for the theta-band SME. Previous studies on EEG gamma-band activity are difficult to interpret because of potential contamination with miniature eye movement artifacts which are not accounted for by traditional artifact removal procedures (Yuval-Greenberg et al., 2008). Here, we used the COSTRAP-algorithm (Hassler et al., 2011) to detect and remove these artifacts. Furthermore, we matched the number of trials each participant contributed to the analyses of SR and SF conditions, and we found individual modulation indices to be uncorrelated to these trial numbers. Hence, different signal-to-noise ratios between conditions cannot explain the results. Although it is in principle possible that the phase-amplitude cross-frequency coupling found here could also be caused by phase-phase frontal-posterior theta coupling in combination with phase-amplitude posterior theta-gamma coupling, inspections of the inverse solution maps (Fig. 2) do not suggest such a relationship (i.e., no significant theta activity in posterior areas and no significant gamma activity in frontal areas). Moreover, differences between cross-frequency coupling were not biased by differences in theta power per se (Fig. 4A). Our findings are compatible with increased gamma-band activity in posterior cortex regions that have been observed in MEG studies during incidental learning (Friese et al., 2012; Kaiser et al., 2004; Osipova et al., 2006). Evidence for theta-band SME in frontal areas has also been reported by Sederberg et al. (2003) using intracranial EEG in epileptic patients and by Hanslmayr et al. (2009) in an EEG study. Alpha-band SMEs have also been demonstrated in former investigations (Hanslmayr et al., 2009; Klimesch et al., 1996b), as well as the co-occurrence of EEG theta synchronization and alpha desynchronization during memory encoding (Klimesch et al., 1997; Molle et al., 2002). The consistent findings of SMEs in different frequency ranges raise the question which different functional aspects might be reflected by each frequency band during memory encoding. Furthermore, if certain cognitive processes are preferentially associated with oscillatory activity in a specific frequency band, what is the
functional significance of interactions between frequency bands? A striking model of the role of cross-frequency interactions in working memory has been put forward by Lisman and Idiart (1995). They propose that objects are cortically represented by consecutive gamma cycles which are nested within a given theta cycle. Capacity limitations of short-term memory then arise as a consequence of the number of gamma cycles that fit into one theta cycle. Supporting evidence for this model has been generated with invasive recordings in humans (Axmacher et al., 2010) but also by some EEG/MEG studies. Gamma power was found to be correlated with theta phase in a working memory task (Demiralp et al., 2007), and theta-gamma synchronization was modulated by attention (Sauseng et al., 2008, 2009). Furthermore, a relationship between individual short-term memory capacity and individual theta/gamma cycle length ratios has been demonstrated (Kaminski et al., 2011). Beyond the working-memory domain, cross-frequency coupling between theta phase and gamma amplitude in various brain regions has also been found during word recognition (Mormann et al., 2005) and in a variety of cognitive tasks (Canolty et al., 2006). Cross-frequency coupling has also been detected in various brain regions of rodents (e.g., Tort et al., 2008, 2009) and nonhuman primates (Lakatos et al., 2005, 2008; Tsunada et al., 2011). The diversity of tasks, brain areas, and species in which cross-frequency couplings have been observed suggests that the interactions between oscillatory activity in different frequency bands serve universal mechanisms of neuronal communication. Lakatos et al. (2005), e.g., propose that cross-frequency couplings reflect a general mechanism by which lower frequency oscillations modulate the exitability of local neuronal assemblies. Based on theoretical considerations and empirical findings, low frequency oscillations are likely to play an important role in large-scale communication between brain areas (Canolty and Knight, 2010; Jensen and Colgin, 2007; Palva and Palva, 2007). Our results indicate that successful memory encoding is accompanied by increased interaction of frontal and parietal brain regions via theta-phase to gamma-amplitude coupling. This finding is compatible with the idea that during encoding frontal areas exert top-down influence on posterior cortex areas. Frontal cortex regions are thought to subserve encoding processes by the selection of goal-relevant information and by the integration of different pieces of information in working memory (Blumenfeld and Ranganath, 2007). Gamma-band activity in posterior cortex regions has been associated with processes related to cortical object representations (Tallon-Baudry et al., 1997). Within the network of brain areas that act in concert to facilitate memory encoding, these representations in posterior regions might reflect the informational counterpart of the frontal control system. Although in agreement with the lines of evidence stated above, these interpretations should nevertheless be considered with caution. Measures of cross-frequency coupling are descriptive and do not necessarily reflect causal relations between oscillations in different frequency bands. Future research will be necessary to establish the causal links and underlying mechanisms of cross-frequency phase-amplitude coupling.
Conclusion Our study is the first to demonstrate increased cross-frequency coupling between EEG-recorded frontal theta and parietal/occipital gamma oscillations during the formation of new memories. Presumably, this coupling reflects interactions between a frontal control system and cortical representations activated and maintained in posterior brain regions.
Acknowledgments This study was supported by grants to T.G. from the German Research Foundation (DFG). We are grateful to three anonymous reviewers for their constructive criticism.
U. Friese et al. / NeuroImage 66 (2013) 642–647
References Axmacher, N., Henseler, M.M., Jensen, O., Weinreich, I., Elger, C.E., Fell, J., 2010. Crossfrequency coupling supports multi-item working memory in the human hippocampus. Proc. Natl. Acad. Sci. U. S. A. 107 (7), 3228–3233. Benjamini, Y., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. 57 (1), 289–300. Bertrand, O., Pantev, C., 1994. Stimulus frequency dependence of the transient oscillatory auditory evoked response (40 Hz) studied by electric and magnetic recordings in humans. In: Pantev, C., Elbert, T., Lütkenhöner, B. (Eds.), Oscillatory Event-Related Brain Dynamics, pp. 231–242. Blumenfeld, R.S., Ranganath, C., 2007. Prefrontal cortex and long-term memory encoding: an integrative review of findings from neuropsychology and neuroimaging. Neuroscientist 13 (3), 280–291. Canolty, R.T., Knight, R.T., 2010. The functional role of cross-frequency coupling. Trends Cogn. Sci. 14 (11), 506–515. Canolty, R.T., Edwards, E., Dalal, S.S., Soltani, M., Nagarajan, S.S., Kirsch, H.E., et al., 2006. High gamma power is phase-locked to theta oscillations in human neocortex. Science 313 (5793), 1626–1628. Delorme, A., Makeig, S., 2004. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J. Neurosci. Methods 134 (1), 9–21. Demiralp, T., Bayraktaroglu, Z., Lenz, D., Junge, S., Busch, N.A., Maess, B., et al., 2007. Gamma amplitudes are coupled to theta phase in human EEG during visual perception. Int. J. Psychophysiol. 64 (1), 24–30. Engel, A.K., Fries, P., Singer, W., 2001. Dynamic predictions: oscillations and synchrony in top-down processing. Nat. Rev. Neurosci. 2 (10), 704–716. Fell, J., Axmacher, N., 2011. The role of phase synchronization in memory processes. Nat. Rev. Neurosci. 12 (2), 105–118. Freunberger, R., Werkle-Bergner, M., Griesmayr, B., Lindenberger, U., Klimesch, W., 2011. Brain oscillatory correlates of working memory constraints. Brain Res. 1375, 93–102. Friese, U., Supp, G.G., Hipp, J.F., Engel, A.K., Gruber, T., 2012. Oscillatory MEG gamma band activity dissociates perceptual and conceptual aspects of visual object processing: a combined repetition/conceptual priming study. Neuroimage 59 (1), 861–871. Gruber, T., Tsivilis, D., Montaldi, D., Müller, M.M., 2004. Induced gamma band responses: an early marker of memory encoding and retrieval. Neuroreport 15 (11), 1837–1841. Gruber, T., Trujillo-Barreto, N.J., Giabbiconi, C.M., Valdes-Sosa, P.A., Müller, M.M., 2006. Brain electrical tomography (BET) analysis of induced gamma band responses during a simple object recognition task. Neuroimage 29 (3), 888–900. Hanslmayr, S., Spitzer, B., Bauml, K.H., 2009. Brain oscillations dissociate between semantic and nonsemantic encoding of episodic memories. Cereb. Cortex 19 (7), 1631–1640. Hanslmayr, S., Volberg, G., Wimber, M., Raabe, M., Greenlee, M.W., Bauml, K.H., 2011. The relationship between brain oscillations and BOLD signal during memory formation: a combined EEG-fMRI study. J. Neurosci. 31 (44), 15674–15680. Hassler, U., Trujillo Barreto, N., Gruber, T., 2011. Induced gamma band responses in human EEG after the control of miniature saccadic artifacts. Neuroimage 57 (4), 1411–1421. Jensen, O., Colgin, L.L., 2007. Cross-frequency coupling between neuronal oscillations. Trends Cogn. Sci. 11 (7), 267–269. Jensen, O., Kaiser, J., Lachaux, J.P., 2007. Human gamma-frequency oscillations associated with attention and memory. Trends Neurosci. 30 (7), 317–324. Junghoefer, M., Elbert, T., Tucker, D.M., Rockstroh, B., 2000. Statistical control of artifacts in dense array EEG/MEG studies. Psychophysiology 37 (4), 523–532. Kaiser, J., Buhler, M., Lutzenberger, W., 2004. Magnetoencephalographic gamma-band responses to illusory triangles in humans. Neuroimage 23 (2), 551–560. Kaminski, J., Brzezicka, A., Wrobel, A., 2011. Short-term memory capacity (7+/−2) predicted by theta to gamma cycle length ratio. Neurobiol. Learn. Mem. 95 (1), 19–23. Klimesch, W., 1999. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain Res. Brain Res. Rev. 29 (2–3), 169–195. Klimesch, W., Doppelmayr, M., Russegger, H., Pachinger, T., 1996a. Theta band power in the human scalp EEG and the encoding of new information. Neuroreport 7 (7), 1235–1240.
647
Klimesch, W., Schimke, H., Doppelmayr, M., Ripper, B., Schwaiger, J., Pfurtscheller, G., 1996b. Event-related desynchronization (ERD) and the Dm effect: does alpha desynchronization during encoding predict later recall performance? Int. J. Psychophysiol. 24 (1–2), 47–60. Klimesch, W., Doppelmayr, M., Pachinger, T., Ripper, B., 1997. Brain oscillations and human memory: EEG correlates in the upper alpha and theta band. Neurosci. Lett. 238 (1–2), 9–12. Lakatos, P., Shah, A.S., Knuth, K.H., Ulbert, I., Karmos, G., Schroeder, C.E., 2005. An oscillatory hierarchy controlling neuronal excitability and stimulus processing in the auditory cortex. J. Neurophysiol. 94 (3), 1904–1911. Lakatos, P., Karmos, G., Mehta, A.D., Ulbert, I., Schroeder, C.E., 2008. Entrainment of neuronal oscillations as a mechanism of attentional selection. Science 320 (5872), 110–113. Lisman, J.E., Idiart, M.A., 1995. Storage of 7+/−2 short-term memories in oscillatory subcycles. Science 267 (5203), 1512–1515. Molle, M., Marshall, L., Fehm, H.L., Born, J., 2002. EEG theta synchronization conjoined with alpha desynchronization indicate intentional encoding. Eur. J. Neurosci. 15 (5), 923–928. Mormann, F., Fell, J., Axmacher, N., Weber, B., Lehnertz, K., Elger, C.E., et al., 2005. Phase/amplitude reset and theta-gamma interaction in the human medial temporal lobe during a continuous word recognition memory task. Hippocampus 15 (7), 890–900. Nyhus, E., Curran, T., 2010. Functional role of gamma and theta oscillations in episodic memory. Neurosci. Biobehav. Rev. 34 (7), 1023–1035. Osipova, D., Takashima, A., Oostenveld, R., Fernandez, G., Maris, E., Jensen, O., 2006. Theta and gamma oscillations predict encoding and retrieval of declarative memory. J. Neurosci. 26 (28), 7523–7531. Osipova, D., Hermes, D., Jensen, O., 2008. Gamma power is phase-locked to posterior alpha activity. PLoS One 3 (12), e3990. Paller, K.A., Wagner, A.D., 2002. Observing the transformation of experience into memory. Trends Cogn. Sci. 6 (2), 93–102. Palva, S., Palva, J.M., 2007. New vistas for alpha-frequency band oscillations. Trends Neurosci. 30 (4), 150–158. Palva, J.M., Palva, S., Kaila, K., 2005. Phase synchrony among neuronal oscillations in the human cortex. J. Neurosci. 25 (15), 3962–3972. Sauseng, P., Klimesch, W., Gruber, W.R., Birbaumer, N., 2008. Cross-frequency phase synchronization: a brain mechanism of memory matching and attention. Neuroimage 40 (1), 308–317. Sauseng, P., Klimesch, W., Heise, K.F., Gruber, W.R., Holz, E., Karim, A.A., et al., 2009. Brain oscillatory substrates of visual short-term memory capacity. Curr. Biol. 19 (21), 1846–1852. Sederberg, P.B., Kahana, M.J., Howard, M.W., Donner, E.J., Madsen, J.R., 2003. Theta and gamma oscillations during encoding predict subsequent recall. J. Neurosci. 23 (34), 10809–10814. Tallon-Baudry, C., Bertrand, O., Delpuech, C., Permier, J., 1997. Oscillatory gamma-band (30–70 Hz) activity induced by a visual search task in humans. J. Neurosci. 17 (2), 722–734. Tort, A.B., Kramer, M.A., Thorn, C., Gibson, D.J., Kubota, Y., Graybiel, A.M., et al., 2008. Dynamic cross-frequency couplings of local field potential oscillations in rat striatum and hippocampus during performance of a T-maze task. Proc. Natl. Acad. Sci. U. S. A. 105 (51), 20517–20522. Tort, A.B., Komorowski, R.W., Manns, J.R., Kopell, N.J., Eichenbaum, H., 2009. Thetagamma coupling increases during the learning of item-context associations. Proc. Natl. Acad. Sci. U. S. A. 106 (49), 20942–20947. Tort, A.B., Komorowski, R., Eichenbaum, H., Kopell, N., 2010. Measuring phase-amplitude coupling between neuronal oscillations of different frequencies. J. Neurophysiol. 104 (2), 1195–1210. Tsunada, J., Baker, A.E., Christison-Lagay, K.L., Davis, S.J., Cohen, Y.E., 2011. Modulation of cross-frequency coupling by novel and repeated stimuli in the primate ventrolateral prefrontal cortex. Front. Psychol. 2, 217. Tulving, E., 1985. Memory and consciousness. Can. Psychol. (Psychol. Can.) 25, 1–15. Varela, F., Lachaux, J.P., Rodriguez, E., Martinerie, J., 2001. The brainweb: phase synchronization and large-scale integration. Nat. Rev. Neurosci. 2 (4), 229–239. Yuval-Greenberg, S., Tomer, O., Keren, A.S., Nelken, I., Deouell, L.Y., 2008. Transient induced gamma-band response in EEG as a manifestation of miniature saccades. Neuron 58 (3), 429–441.
Contents lists available at SciVerse ScienceDirect
NeuroImage journal homepage: www.elsevier.com/locate/ynimg
Successful memory encoding is associated with increased cross-frequency coupling between frontal theta and posterior gamma oscillations in human scalp-recorded EEG Uwe Friese a, b,⁎, 1, Moritz Köster b, c, 1, Uwe Hassler b, Ulla Martens b, Nelson Trujillo-Barreto d, Thomas Gruber b a
University Medical Center Hamburg-Eppendorf, Department of Neurophysiology and Pathophysiology, Hamburg, Germany Institute of Psychology, University of Osnabrueck, Germany c Institute of Cognitive Science, University of Osnabrueck, Germany d Cuban Neuroscience Center, Havana, Cuba b
a r t i c l e
i n f o
Article history: Accepted 1 November 2012 Available online 8 November 2012 Keywords: EEG Oscillations Cross-frequency-coupling Gamma-band Theta-band Memory encoding
a b s t r a c t Although previous studies have established that successful memory encoding is associated with increased synchronization of theta-band and gamma-band oscillations, it is unclear if there is a functional relationship between oscillations in these frequency bands. Using scalp-recorded EEG in healthy human participants, we demonstrate that cross-frequency coupling between frontal theta phase and posterior gamma power is enhanced during the encoding of visual stimuli which participants later on remember versus items which participants subsequently forget (“subsequent memory effect,” SME). Conventional wavelet analyses and source localizations revealed SMEs in spectral power of theta-, alpha-, and gamma-band. Successful compared to unsuccessful encoding was reflected in increased theta-band activity in right frontal cortex as well as increased gamma-band activity in parietal-occipital regions. Moreover, decreased alpha-band activity in prefrontal and occipital cortex was also related to successful encoding. Overall, these findings support the idea that during the formation of new memories frontal cortex regions interact with cortical representations in posterior areas. © 2012 Elsevier Inc. All rights reserved.
Introduction Impressive evidence has been accumulated that synchronized neuronal activity serves the integration of spatially distributed processing in the brain (Engel et al., 2001; Varela et al., 2001). In particular, oscillations in the gamma band range ( > 30 Hz) are thought to reflect processes related to the activation and maintenance of neuronal object representations (Jensen et al., 2007). Low frequency oscillations in the theta band,~ 5 Hz, and in the alpha band, 8–12 Hz, have been related to top-down control modulating the processing of neuronal object representations (Palva et al., 2005). Mostly based on animal research or studies with invasive recordings in epilepsy patients, several authors have suggested that the coordinated interplay of low- and high-frequency oscillations is likely to be crucial for various memory processes (Canolty and Knight, 2010; Fell and Axmacher, 2011; Freunberger et al., 2011; Nyhus and Curran, 2010; Palva and
⁎ Corresponding author at: University Medical Center Hamburg-Eppendorf, Department of Neurophysiology and Pathophysiology, Martinistraße 52, D-20246 Hamburg, Germany. Fax: +49 40 7410 57752. E-mail address: [email protected] (U. Friese). 1 These authors contributed equally to this work. 1053-8119/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.neuroimage.2012.11.002
Palva, 2007). So far, only a few studies using electroencephalography (EEG) or magnetencephalography (MEG) in humans have demonstrated cross-frequency coupling between low- and high-frequency oscillations in working memory and attention tasks (Demiralp et al., 2007; Palva et al., 2005; Sauseng et al., 2008, 2009). In the present study, we employed a subsequent memory paradigm to investigate whether low- and high-frequency oscillations in scalp-recorded EEG data interact during successful encoding of new memories. In a subsequent memory paradigm, neuronal activity during the encoding of stimuli which are remembered in a subsequent memory test, is contrasted versus activity during the encoding of stimuli which are later forgotten (Paller and Wagner, 2002). Previous EEG/MEG studies on oscillatory brain activity have established “subsequent memory effects” (SME) in various frequency bands. Successful memory encoding has been found to be accompanied by increased synchrony in the theta and gamma band as well as by decreased synchrony in the alpha and beta band (Gruber et al., 2004; Hanslmayr et al., 2009, 2011; Klimesch et al., 1996a, 1996b; Osipova et al., 2006). Most importantly, Osipova et al. (2006) found that increased frontal theta power coincided with enhanced gamma power in posterior cortex regions for later remembered versus later forgotten stimuli. The authors speculated that frontal theta oscillations might reflect top-down processes
U. Friese et al. / NeuroImage 66 (2013) 642–647
modulating gamma band activity related to neuronal representations in posterior regions. However, this hypothesis was not tested directly. On the basis of Osipova et al. (2006), the main objectives of our study were (a) to confirm subsequent memory effects in theta, alpha, and gamma frequency bands, (b) to identify the cortical generators of these effects, and (c) to explore if posterior gamma power is coupled to the phases of frontal theta oscillations during the encoding of new information. This type of cross-frequency coupling would provide strong evidence for the notion that theta-oscillations play a pivotal role in long-range communication between brain regions involved in memory encoding processes. Materials and methods Participants A total of 26 healthy university students (23 female, M = 21.3 years, SD = 3.7) received either monetary compensation or course credits for participation. None of the participants reported any history of neurological or psychiatrical disorder. All participants had normal or corrected-to-normal vision. Informed consent was obtained from all participants according to the Declaration of Helsinki.
643
condition with the lowest number of trials. Data of three participants with less than 15 trials per condition were discarded. For the remaining 23 participants on average 26.1 trials (SD = 7.7 trials) per condition were available.
Data analysis: Spectral changes in electrode space Total spectral changes of oscillatory activity were calculated for each trial using Morlet wavelets (Bertrand and Pantev, 1994) with seven cycles. Further analyses were based on a time window of 400–1300 ms after stimulus onset. This segment was chosen after visual inspection of time-by-frequency plots. Moreover, we wanted to exclude possible interference from early visual responses (Osipova et al., 2008). For broadband gamma oscillations, single-trial power spectra were averaged within 50–80 Hz. As theta and alpha signals vary largely across individuals (Klimesch, 1999), we selected frequencies for each participant individually on the basis of all trials across all conditions. The mean theta frequency was 5.1 Hz (SD =1.2 Hz) and the mean alpha frequency was 10.2 Hz (SD = 1.7 Hz). To evaluate SMEs (SR minus SF) in electrode space, we averaged data for each subject and electrode within the respective frequency band and time frame and subjected individual means to paired sample t-tests.
Stimuli and procedure The stimuli consisted of 150 pictures of living (e.g. plants, animals) and 150 pictures of nonliving objects (e.g. clothes, tools). Pictures covered a visual angle of 6.2° 6.2° and were presented at the center of a 21 inch computer screen. In the encoding session, 200 pictures (100 of each category) were presented in random order and had to be classified as living or nonliving objects. The presentation sequence was blank screen (1 s), fixation point (1.0–1.5 s), target stimulus (2 s), question mark until response. Participants used response keys with different fingers of their right hand. The encoding session was followed by a 15 min filler task in which participants solved mathematical equations. Afterwards 200 pictures from the encoding session were randomly intermixed with 100 new stimuli, and participants were to indicate whether they “remember” the object, they “know” the object, or if the object was not shown in the first part of the experiment and therefore “new”. This so-called remember/know procedure allows to distinguish between familiarity-based recognition and controlled recollection (Tulving, 1985). “Know”-responses are assumed to rely more on familiarity while “remember”-responses are rather associated with recollection. Given our main interest in successful memory encoding processes, we focused on remember-responses to obtain “purer” measures of recollection memory and to enhance SMEs (Paller and Wagner, 2002). Hence, we discarded trials with know-responses and limited EEG data analyses to comparisons of subsequently remembered (SR) and subsequently forgotten (SF) items. Electrophysiological recordings A BioSemi Active-Two amplifier system with 128 active electrodes was used at 512 Hz sampling rate. Additionally, horizontal and vertical electro-oculograms were recorded. Common mode sense (CMS) and driven right leg (DRL) electrodes served as recording reference and ground. EEG data were re-referenced to the average of all electrodes and segmented into epochs from −500 to 2000 ms with regard to stimulus onset. Eye blink artifacts were removed using an independent component based procedure (Delorme and Makeig, 2004). Further artifacts were removed with SCADS (Junghoefer et al., 2000). For the analysis of gamma-band oscillations, we applied a correction of saccade-related transient potentials (COSTRAP; Hassler et al., 2011). This procedure has been shown to remove possible artifacts caused by miniature eye movements (Yuval-Greenberg et al., 2008). For all analyses, the number of trials per condition was matched for each subject according to the
Data analysis: Spectral changes in scource space To localize the generators of oscillatory EEG activity, single-trial VARETA analyses were conducted (for details see Gruber et al., 2006). This method is based on a discrete spline distributed inverse model and reveals the estimates of the primary current densities (PCDs) in source space which are the spatially smoothest solutions compatible with the observed scalp topographies. PCD estimates for SR and SF were subjected to voxel-wise paired sample t-tests. Data analysis: Phase-amplitude coupling For the assessment of phase-amplitude coupling between lowfrequency theta/alpha and high-frequency gamma oscillations, we used the modulation index (MI) described in detail by Tort et al. (2010). The time series data for this analysis were created by averaging data epochs (400–1300 ms) across selected electrodes for SR and SF. Electrode selection was based on significant SMEs in the different frequency bands (as described above). In brief, coupling between low-frequency phase signals and high-frequency amplitude signals was computed as follows. The signals in question were filtered at the respective frequencies, and the Hilbert transform was used to derive phase information of the low-frequency signal as well as the amplitude envelope of the high-frequency signal. Next, a composite time series was constructed which represents the amplitude of the gamma oscillation at each phase of the low-frequency signal. Low frequency phases were binned into twelve 30° intervals (0–360°) and the mean amplitude of the gamma oscillation in each bin was derived. These mean values were normalized by dividing each bin value by the sum over all bins. The resulting “amplitude distributions” are uniform if no phase-amplitude coupling is evident. Deviations from uniformity can then be quantified by means of a discrete and normalized (between 0 and 1) version of the Kullback-Leibler divergence between the empirical phase-to-amplitude distribution and the uniform distribution (i.e. the modulation index, MI). Tort et al. (2010) have shown that the MI does not depend on the absolute level of the signal, and it as also relatively insensitive to changes of noise levels. These properties make the MI especially suited for evaluating the intensity of phaseamplitude coupling. For all subjects, we calculated MI-values separately for SR and SF trials, and tested for differences between conditions using non-parametric Wilcoxon signed rank tests.
644
U. Friese et al. / NeuroImage 66 (2013) 642–647
Results
Cross-frequency coupling
Behavioral data
To test the hypothesis if parietal gamma-band activity was modulated by frontal theta-band activity (Osipova et al., 2006), we first averaged raw signals across electrodes with significant SMEs (see Fig. 2), i.e. right frontal electrodes for the theta phase signals and central parietal electrodes for the gamma amplitude signals. The theta phase signals were band-pass filtered from 5 to 8 Hz, and gamma amplitude signals were band-pass filtered from 50 to 80 Hz using default EEGlab two-way least-squares FIR filtering procedures. Individual modulation indices were calculated for each participants and condition. Fig. 3, left panel, illustrates phase-amplitude relations of two exemplary participants as amplitude distributions for two complete phase cycles (0–720°). For participants P20, the amplitude distributions for forgotten items (left) are more uniform than for remembered items (right), whereas for participant P17 both distributions appear to be largely comparable. More formally, the modulation indices for SR and SF conditions of all participants are plotted in Fig. 3, right panel. Data from the two exemplary participants have been highlighted. For participants located above the diagonal (e.g., P20), the MI was greater in the remember condition than in the forgotten condition. Confirming the visual inspection of the amplitude distribution, the MIs of participant P17 fall close to the diagonal. Overall, MIs in the remembered conditions were significantly larger than in the forgotten condition (z = -2.37, p b 0.02) indicating enhanced cross-frequency phase-amplitude coupling between frontal theta and parietal gamma oscillations during successful memory encoding. It is important to note that the differences in cross-frequency coupling between conditions could be biased by power changes of the phase-modulating frequency. Since theta power was higher in the SR condition, estimations of theta phase might be improved in comparison to the SF condition. Consequently, estimations of cross-frequency coupling could be biased. To rule out this possibility, we assessed the relationship between MIs, theta power, and also gamma power independently from the conditions as follows. Single-trial data of each participant were sorted regardless of condition by mean theta power averaged across the selected electrodes. We divided the sorted data into quartiles such that the first quartile (Q1) contained 25% of all trials with lowest theta power, and the forth quartile (Q4) contained 25% of all trials with highest theta power. For each quartile, we derived MIs, mean theta power averaged across frontal electrodes, and mean gamma power averaged across parietal electrodes. Fig. 4A depicts the corresponding means across participants in quartiles of theta power (Q1 to Q4). Naturally, there is a noticeable increase of mean theta power (gray bars) but also an increase of parietal gamma power (white bars) across quartiles. In contrast, MIs do not differ across quartiles (black squares). Corresponding 1-way analyses of variance across quartiles resulted in significant effects for theta power (F3,93 = 88.0, p b .01, linear trend: F1,31 = 93.7, p b .01) and gamma power (F3,93 = 13.4, p b .01, linear trend: F1,31 = 19.1, p b .01), whereas no significant differences across quartiles were found for MIs (F3,93 b 1). Hence, it is unlikely that our finding of increased cross-frequency coupling in the remembered versus the forgotten condition is due to theta power differences between the conditions per se. Lastly, to explore the cross-frequency coupling pattern at the level of individual electrodes, we chose the frontal electrodes for which theta power was larger for SR versus SF as seeds, and calculated MIs between each seed electrode and all other scalp electrodes. In Fig. 4B, electrode pairs associated with significant differences MISR − MISF (p b .01, two-tailed t-tests) are indicated by solid lines. (No significant results were found for the opposite direction.) These results show that with respect to the selected frontal region, the pattern of increased theta-gamma phase-amplitude coupling is indeed confined to primarily parietal and occipital brain areas.
On average, 48.5% (SEM = 3.6%) of the encoding session items were remembered and 22.1% (SEM =3.6%) were reported new. Counting remember responses to old items as hits and remember responses to new items as false alarms (M = 2.8, SEM= 1.0%), mean discrimination performance for remembered items was significantly larger than chance (d′ = 1.99; t22 = 17.2, p b .001).
SMEs in theta-, alpha-, and gamma-band power Time-by-frequency representations of oscillatory activity are depicted in Fig. 1 as power changes relative to the 200 ms pre-stimulus baseline activity. After stimulus-onset, the average theta power clearly increased, whereas spectral power decreased in the alpha-band. Average gamma power showed a maximum approximately between 50 and 65 Hz throughout the measurement interval. Subsequent memory effects in oscillatory power were evident in theta-, alpha-, and gamma-band. The left hand side of Fig. 2 shows grand mean spherical spline interpolated topographical maps of spectral power (400–1300 ms) separately for SR and SF conditions. At marked electrodes the subsequent memory effect, i.e. the difference SR–SF, was significant with p b .01 (corresponding to a false discovery rate of q b .1; Benjamini and Hochberg, 1995). The right hand side of Fig. 2 illustrates the estimated VARETA source localizations of the SMEs. Statistical maps are thresholded at t-values associated with p b .01 and projected onto the smoothed surface of an MRI template brain. In the theta-band, SMEs were evident at right frontal electrodes with corresponding source localization in the right frontal cortex. Alpha-band suppression was larger for remembered versus forgotten items at two large electrode clusters in bilateral anterior and posterior scalp regions. The sources of these SMEs were found in bilateral middle frontal gyri and in primary visual cortex areas. Gamma-band SMEs occurred at central parietal electrodes with sources located in left parietal and occipital cortex as well as in right inferior parietal cortex.
1.3
18
rel. signal change
Frequency (Hz)
14 10 6 2
0.7 0
400
800
1200
1600 1.06 rel. signal change
Frequency (Hz)
80 70 60 50 40
0.94 0
400
800
1200
1600
Time (ms) Fig. 1. Grand mean time-by-frequency plots of spectral power for frequencies up to 20 Hz and for the gamma range (40–100 Hz). Low frequency activity was averaged across all electrodes, gamma-band activity was averaged across parietal electrodes. Amplitude values are expressed as signal changes relative to the 200 ms pre-stimulus baseline.
U. Friese et al. / NeuroImage 66 (2013) 642–647
Forgotten
Remembered
645
Remembered – Forgotten
Theta
Alpha
Gamma
+ signs: electrodes with significant difference Theta +/- 3, Alpha +/- 6, Gamma +/- 0.4 (µV2)
2.5
Theta/Gamma
4.0
T -2.5
Alpha
-4.0
Fig. 2. The left panel depicts subsequent memory effects in electrode space. Topographical maps of average oscillatory power are shown for theta-, alpha-, and gamma-band (400– 1300 ms) in the forgotten and remember conditions. Amplitude values are expressed as signal change to the 200 ms pre-stimulus baseline. SMEs are associated with p b .01 at marked electrodes. In the right panel, source space SMEs are illustrated as t-values (thresholded at p b .01) projected onto the surface of an average MRI template brain.
Discussion This study provides first strong evidence for a functional link between low- and high-frequency oscillations during the encoding of new memories. Phase-amplitude coupling between frontal theta oscillations and parietal gamma oscillations was found to be larger for subsequently remembered versus subsequently forgotten items. This coupling might reflect a mechanism by which theta-mediated control processes modulate gamma-band related processes implicated in the
reactivation and maintenance of neuronal object representations (Osipova et al., 2006; Palva et al., 2005). Overall, our findings regarding SMEs in theta, alpha, and gamma power show considerable concordance with other EEG/MEG studies. Osipova et al. (2006), using a similar design in an MEG experiment, observed a gamma-band SME at posterior parietal and occipital sensors, as well as a theta-band SME at right temporal sensor sites. In that study, the source of the gamma-band SME could be located in visual cortex whereas no reliable source was found for the theta SME.
frontal theta
parietal gamma
Modulation index (MI) Forgotten
SR > SF
Remembered
P20
P20 P17
P17 SF > SR 0.005
0.005
0.01
MI
0.10
Amplitude
Remembered
0.01
0.05 Phase
0
360 720
Forgotten Fig. 3. The cartoon head illustrates phase-amplitude coupling with frontal theta phase modulating parietal gamma power. The left panel displays modulation indices for remember and forgotten conditions of all participants. Above the diagonal line, MIs are larger for remember versus forgotten (SR >SF), below the diagonal line, MIs are larger for forgotten versus remember (SF > SR). The right panel shows phase-amplitude distributions for two complete phase cycles (0–720°) of the two participants (P17, and P20) in the forgotten and remembered conditions. The extent of cross-frequency coupling can be inferred from deviations of the uniform distribution.
646
U. Friese et al. / NeuroImage 66 (2013) 642–647
A 0.016
Theta
Modulation index (MI)
0.8
Gamma
Power (µV2)
MI
0
Q1
Q2
Q3
Q4
0
B Frontal seed electrodes
Significant CFC: Remembered > Forgotten Fig. 4. (A) Mean theta-gamma modulation indices (MI) and mean theta/gamma power in quartiles of theta power. Calculations were based on all trials across conditions sorted by theta power at frontal electrodes. Bars illustrate increase of power in theta band (gray) and gamma band (white) over quartiles (Q1-Q4). Black squares indicate that corresponding cross-frequency modulation indices (MI) do not increase with theta power. (B) Theta-gamma phase-amplitude coupling was calculated between frontal seed electrodes and all other scalp electrodes. Solid lines indicate significant cross-frequency coupling differences (SR > SF) between electrode pairs (p b 0.01, uncorrected).
We identified sources in posterior parietal cortex regions for the gamma-band SME and right frontal cortex for the theta-band SME. Previous studies on EEG gamma-band activity are difficult to interpret because of potential contamination with miniature eye movement artifacts which are not accounted for by traditional artifact removal procedures (Yuval-Greenberg et al., 2008). Here, we used the COSTRAP-algorithm (Hassler et al., 2011) to detect and remove these artifacts. Furthermore, we matched the number of trials each participant contributed to the analyses of SR and SF conditions, and we found individual modulation indices to be uncorrelated to these trial numbers. Hence, different signal-to-noise ratios between conditions cannot explain the results. Although it is in principle possible that the phase-amplitude cross-frequency coupling found here could also be caused by phase-phase frontal-posterior theta coupling in combination with phase-amplitude posterior theta-gamma coupling, inspections of the inverse solution maps (Fig. 2) do not suggest such a relationship (i.e., no significant theta activity in posterior areas and no significant gamma activity in frontal areas). Moreover, differences between cross-frequency coupling were not biased by differences in theta power per se (Fig. 4A). Our findings are compatible with increased gamma-band activity in posterior cortex regions that have been observed in MEG studies during incidental learning (Friese et al., 2012; Kaiser et al., 2004; Osipova et al., 2006). Evidence for theta-band SME in frontal areas has also been reported by Sederberg et al. (2003) using intracranial EEG in epileptic patients and by Hanslmayr et al. (2009) in an EEG study. Alpha-band SMEs have also been demonstrated in former investigations (Hanslmayr et al., 2009; Klimesch et al., 1996b), as well as the co-occurrence of EEG theta synchronization and alpha desynchronization during memory encoding (Klimesch et al., 1997; Molle et al., 2002). The consistent findings of SMEs in different frequency ranges raise the question which different functional aspects might be reflected by each frequency band during memory encoding. Furthermore, if certain cognitive processes are preferentially associated with oscillatory activity in a specific frequency band, what is the
functional significance of interactions between frequency bands? A striking model of the role of cross-frequency interactions in working memory has been put forward by Lisman and Idiart (1995). They propose that objects are cortically represented by consecutive gamma cycles which are nested within a given theta cycle. Capacity limitations of short-term memory then arise as a consequence of the number of gamma cycles that fit into one theta cycle. Supporting evidence for this model has been generated with invasive recordings in humans (Axmacher et al., 2010) but also by some EEG/MEG studies. Gamma power was found to be correlated with theta phase in a working memory task (Demiralp et al., 2007), and theta-gamma synchronization was modulated by attention (Sauseng et al., 2008, 2009). Furthermore, a relationship between individual short-term memory capacity and individual theta/gamma cycle length ratios has been demonstrated (Kaminski et al., 2011). Beyond the working-memory domain, cross-frequency coupling between theta phase and gamma amplitude in various brain regions has also been found during word recognition (Mormann et al., 2005) and in a variety of cognitive tasks (Canolty et al., 2006). Cross-frequency coupling has also been detected in various brain regions of rodents (e.g., Tort et al., 2008, 2009) and nonhuman primates (Lakatos et al., 2005, 2008; Tsunada et al., 2011). The diversity of tasks, brain areas, and species in which cross-frequency couplings have been observed suggests that the interactions between oscillatory activity in different frequency bands serve universal mechanisms of neuronal communication. Lakatos et al. (2005), e.g., propose that cross-frequency couplings reflect a general mechanism by which lower frequency oscillations modulate the exitability of local neuronal assemblies. Based on theoretical considerations and empirical findings, low frequency oscillations are likely to play an important role in large-scale communication between brain areas (Canolty and Knight, 2010; Jensen and Colgin, 2007; Palva and Palva, 2007). Our results indicate that successful memory encoding is accompanied by increased interaction of frontal and parietal brain regions via theta-phase to gamma-amplitude coupling. This finding is compatible with the idea that during encoding frontal areas exert top-down influence on posterior cortex areas. Frontal cortex regions are thought to subserve encoding processes by the selection of goal-relevant information and by the integration of different pieces of information in working memory (Blumenfeld and Ranganath, 2007). Gamma-band activity in posterior cortex regions has been associated with processes related to cortical object representations (Tallon-Baudry et al., 1997). Within the network of brain areas that act in concert to facilitate memory encoding, these representations in posterior regions might reflect the informational counterpart of the frontal control system. Although in agreement with the lines of evidence stated above, these interpretations should nevertheless be considered with caution. Measures of cross-frequency coupling are descriptive and do not necessarily reflect causal relations between oscillations in different frequency bands. Future research will be necessary to establish the causal links and underlying mechanisms of cross-frequency phase-amplitude coupling.
Conclusion Our study is the first to demonstrate increased cross-frequency coupling between EEG-recorded frontal theta and parietal/occipital gamma oscillations during the formation of new memories. Presumably, this coupling reflects interactions between a frontal control system and cortical representations activated and maintained in posterior brain regions.
Acknowledgments This study was supported by grants to T.G. from the German Research Foundation (DFG). We are grateful to three anonymous reviewers for their constructive criticism.
U. Friese et al. / NeuroImage 66 (2013) 642–647
References Axmacher, N., Henseler, M.M., Jensen, O., Weinreich, I., Elger, C.E., Fell, J., 2010. Crossfrequency coupling supports multi-item working memory in the human hippocampus. Proc. Natl. Acad. Sci. U. S. A. 107 (7), 3228–3233. Benjamini, Y., Hochberg, Y., 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Stat. Soc. 57 (1), 289–300. Bertrand, O., Pantev, C., 1994. Stimulus frequency dependence of the transient oscillatory auditory evoked response (40 Hz) studied by electric and magnetic recordings in humans. In: Pantev, C., Elbert, T., Lütkenhöner, B. (Eds.), Oscillatory Event-Related Brain Dynamics, pp. 231–242. Blumenfeld, R.S., Ranganath, C., 2007. Prefrontal cortex and long-term memory encoding: an integrative review of findings from neuropsychology and neuroimaging. Neuroscientist 13 (3), 280–291. Canolty, R.T., Knight, R.T., 2010. The functional role of cross-frequency coupling. Trends Cogn. Sci. 14 (11), 506–515. Canolty, R.T., Edwards, E., Dalal, S.S., Soltani, M., Nagarajan, S.S., Kirsch, H.E., et al., 2006. High gamma power is phase-locked to theta oscillations in human neocortex. Science 313 (5793), 1626–1628. Delorme, A., Makeig, S., 2004. EEGLAB: an open source toolbox for analysis of single-trial EEG dynamics including independent component analysis. J. Neurosci. Methods 134 (1), 9–21. Demiralp, T., Bayraktaroglu, Z., Lenz, D., Junge, S., Busch, N.A., Maess, B., et al., 2007. Gamma amplitudes are coupled to theta phase in human EEG during visual perception. Int. J. Psychophysiol. 64 (1), 24–30. Engel, A.K., Fries, P., Singer, W., 2001. Dynamic predictions: oscillations and synchrony in top-down processing. Nat. Rev. Neurosci. 2 (10), 704–716. Fell, J., Axmacher, N., 2011. The role of phase synchronization in memory processes. Nat. Rev. Neurosci. 12 (2), 105–118. Freunberger, R., Werkle-Bergner, M., Griesmayr, B., Lindenberger, U., Klimesch, W., 2011. Brain oscillatory correlates of working memory constraints. Brain Res. 1375, 93–102. Friese, U., Supp, G.G., Hipp, J.F., Engel, A.K., Gruber, T., 2012. Oscillatory MEG gamma band activity dissociates perceptual and conceptual aspects of visual object processing: a combined repetition/conceptual priming study. Neuroimage 59 (1), 861–871. Gruber, T., Tsivilis, D., Montaldi, D., Müller, M.M., 2004. Induced gamma band responses: an early marker of memory encoding and retrieval. Neuroreport 15 (11), 1837–1841. Gruber, T., Trujillo-Barreto, N.J., Giabbiconi, C.M., Valdes-Sosa, P.A., Müller, M.M., 2006. Brain electrical tomography (BET) analysis of induced gamma band responses during a simple object recognition task. Neuroimage 29 (3), 888–900. Hanslmayr, S., Spitzer, B., Bauml, K.H., 2009. Brain oscillations dissociate between semantic and nonsemantic encoding of episodic memories. Cereb. Cortex 19 (7), 1631–1640. Hanslmayr, S., Volberg, G., Wimber, M., Raabe, M., Greenlee, M.W., Bauml, K.H., 2011. The relationship between brain oscillations and BOLD signal during memory formation: a combined EEG-fMRI study. J. Neurosci. 31 (44), 15674–15680. Hassler, U., Trujillo Barreto, N., Gruber, T., 2011. Induced gamma band responses in human EEG after the control of miniature saccadic artifacts. Neuroimage 57 (4), 1411–1421. Jensen, O., Colgin, L.L., 2007. Cross-frequency coupling between neuronal oscillations. Trends Cogn. Sci. 11 (7), 267–269. Jensen, O., Kaiser, J., Lachaux, J.P., 2007. Human gamma-frequency oscillations associated with attention and memory. Trends Neurosci. 30 (7), 317–324. Junghoefer, M., Elbert, T., Tucker, D.M., Rockstroh, B., 2000. Statistical control of artifacts in dense array EEG/MEG studies. Psychophysiology 37 (4), 523–532. Kaiser, J., Buhler, M., Lutzenberger, W., 2004. Magnetoencephalographic gamma-band responses to illusory triangles in humans. Neuroimage 23 (2), 551–560. Kaminski, J., Brzezicka, A., Wrobel, A., 2011. Short-term memory capacity (7+/−2) predicted by theta to gamma cycle length ratio. Neurobiol. Learn. Mem. 95 (1), 19–23. Klimesch, W., 1999. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain Res. Brain Res. Rev. 29 (2–3), 169–195. Klimesch, W., Doppelmayr, M., Russegger, H., Pachinger, T., 1996a. Theta band power in the human scalp EEG and the encoding of new information. Neuroreport 7 (7), 1235–1240.
647
Klimesch, W., Schimke, H., Doppelmayr, M., Ripper, B., Schwaiger, J., Pfurtscheller, G., 1996b. Event-related desynchronization (ERD) and the Dm effect: does alpha desynchronization during encoding predict later recall performance? Int. J. Psychophysiol. 24 (1–2), 47–60. Klimesch, W., Doppelmayr, M., Pachinger, T., Ripper, B., 1997. Brain oscillations and human memory: EEG correlates in the upper alpha and theta band. Neurosci. Lett. 238 (1–2), 9–12. Lakatos, P., Shah, A.S., Knuth, K.H., Ulbert, I., Karmos, G., Schroeder, C.E., 2005. An oscillatory hierarchy controlling neuronal excitability and stimulus processing in the auditory cortex. J. Neurophysiol. 94 (3), 1904–1911. Lakatos, P., Karmos, G., Mehta, A.D., Ulbert, I., Schroeder, C.E., 2008. Entrainment of neuronal oscillations as a mechanism of attentional selection. Science 320 (5872), 110–113. Lisman, J.E., Idiart, M.A., 1995. Storage of 7+/−2 short-term memories in oscillatory subcycles. Science 267 (5203), 1512–1515. Molle, M., Marshall, L., Fehm, H.L., Born, J., 2002. EEG theta synchronization conjoined with alpha desynchronization indicate intentional encoding. Eur. J. Neurosci. 15 (5), 923–928. Mormann, F., Fell, J., Axmacher, N., Weber, B., Lehnertz, K., Elger, C.E., et al., 2005. Phase/amplitude reset and theta-gamma interaction in the human medial temporal lobe during a continuous word recognition memory task. Hippocampus 15 (7), 890–900. Nyhus, E., Curran, T., 2010. Functional role of gamma and theta oscillations in episodic memory. Neurosci. Biobehav. Rev. 34 (7), 1023–1035. Osipova, D., Takashima, A., Oostenveld, R., Fernandez, G., Maris, E., Jensen, O., 2006. Theta and gamma oscillations predict encoding and retrieval of declarative memory. J. Neurosci. 26 (28), 7523–7531. Osipova, D., Hermes, D., Jensen, O., 2008. Gamma power is phase-locked to posterior alpha activity. PLoS One 3 (12), e3990. Paller, K.A., Wagner, A.D., 2002. Observing the transformation of experience into memory. Trends Cogn. Sci. 6 (2), 93–102. Palva, S., Palva, J.M., 2007. New vistas for alpha-frequency band oscillations. Trends Neurosci. 30 (4), 150–158. Palva, J.M., Palva, S., Kaila, K., 2005. Phase synchrony among neuronal oscillations in the human cortex. J. Neurosci. 25 (15), 3962–3972. Sauseng, P., Klimesch, W., Gruber, W.R., Birbaumer, N., 2008. Cross-frequency phase synchronization: a brain mechanism of memory matching and attention. Neuroimage 40 (1), 308–317. Sauseng, P., Klimesch, W., Heise, K.F., Gruber, W.R., Holz, E., Karim, A.A., et al., 2009. Brain oscillatory substrates of visual short-term memory capacity. Curr. Biol. 19 (21), 1846–1852. Sederberg, P.B., Kahana, M.J., Howard, M.W., Donner, E.J., Madsen, J.R., 2003. Theta and gamma oscillations during encoding predict subsequent recall. J. Neurosci. 23 (34), 10809–10814. Tallon-Baudry, C., Bertrand, O., Delpuech, C., Permier, J., 1997. Oscillatory gamma-band (30–70 Hz) activity induced by a visual search task in humans. J. Neurosci. 17 (2), 722–734. Tort, A.B., Kramer, M.A., Thorn, C., Gibson, D.J., Kubota, Y., Graybiel, A.M., et al., 2008. Dynamic cross-frequency couplings of local field potential oscillations in rat striatum and hippocampus during performance of a T-maze task. Proc. Natl. Acad. Sci. U. S. A. 105 (51), 20517–20522. Tort, A.B., Komorowski, R.W., Manns, J.R., Kopell, N.J., Eichenbaum, H., 2009. Thetagamma coupling increases during the learning of item-context associations. Proc. Natl. Acad. Sci. U. S. A. 106 (49), 20942–20947. Tort, A.B., Komorowski, R., Eichenbaum, H., Kopell, N., 2010. Measuring phase-amplitude coupling between neuronal oscillations of different frequencies. J. Neurophysiol. 104 (2), 1195–1210. Tsunada, J., Baker, A.E., Christison-Lagay, K.L., Davis, S.J., Cohen, Y.E., 2011. Modulation of cross-frequency coupling by novel and repeated stimuli in the primate ventrolateral prefrontal cortex. Front. Psychol. 2, 217. Tulving, E., 1985. Memory and consciousness. Can. Psychol. (Psychol. Can.) 25, 1–15. Varela, F., Lachaux, J.P., Rodriguez, E., Martinerie, J., 2001. The brainweb: phase synchronization and large-scale integration. Nat. Rev. Neurosci. 2 (4), 229–239. Yuval-Greenberg, S., Tomer, O., Keren, A.S., Nelken, I., Deouell, L.Y., 2008. Transient induced gamma-band response in EEG as a manifestation of miniature saccades. Neuron 58 (3), 429–441.