PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of scirepAboutEditorial BoardFor AuthorsScientific Reports
 
Sci Rep. 2017; 7: 11202.
Published online 2017 September 11. doi:  10.1038/s41598-017-11632-z
PMCID: PMC5593879

Temporally and spatially distinct theta oscillations dissociate a language-specific from a domain-general processing mechanism across the age trajectory

Abstract

The cognitive functionality of neural oscillations is still highly debated, as different functions have been associated with identical frequency ranges. Theta band oscillations, for instance, were proposed to underlie both language comprehension and domain-general cognitive abilities. Here we show that the ageing brain can provide an answer to the open question whether it is one and the same theta oscillation underlying those functions, thereby resolving a long-standing paradox. While better cognitive functioning is predicted by low theta power in the brain at rest, resting state (RS) theta power declines with age, but sentence comprehension deteriorates in old age. We resolve this paradox showing that sentence comprehension declines due to changes in RS theta power within domain-general brain networks known to support successful sentence comprehension, while low RS theta power within the left-hemispheric dorso-frontal language network predicts intact sentence comprehension. The two RS theta networks were also found to functionally decouple relative to their independent internal coupling. Thus, both temporally and spatially distinct RS theta oscillations dissociate a language-specific from a domain-general processing mechanism.

Introduction

Language comprehension in general remains remarkably stable across the lifespan; sentence comprehension, however, is known to decline with age1, 2. This age-related decline holds particularly for working memory-intensive sentences3, which predominantly tax the encoding and retention of verbal information across increased intervals36.

The ability to encode working memory-intensive sentences, such as structurally complex, ambiguous, or long sentences79, is a major determinant of language comprehension. That is, language comprehension requires the encoding and retention of verbal information over either increased hierarchical or sequential distances while simultaneously processing intervening verbal material, potentially interfering with already encoded information. Therefore, the encoding success can be affected quantitatively by the number of words (i.e., capacity-based3) as well as qualitatively by the memory strength (i.e., decay-, interference- and structure-based1012). On complex sentences, readers with a high memory span perform faster and more accurately than low-span readers1, 8, 9, 13. This difference is supported by electrophysiological studies showing that event-related brain potentials (ERPs) associated with sentence processing difficulty14 increase for low-span as compared to high-span readers9. Likewise, processing differences have been found during the disambiguation of ambiguous sentences, with larger ERP differences for high-span readers than for low-span readers7. Such differences are particularly pronounced when ambiguities are extended7, possibly suggesting3 that low-span readers’ verbal working memory resources are exhausted by the increased duration of an ambiguity4, 5, 9, 15.

Given its critical role for sentence comprehension, the decline of verbal working memory capacity with age16 could also underlie the decline in sentence comprehension in the elderly17. When sentence processing is working memory-intensive, older adults perform worse than young adults. Syntactically complex sentences are processed slower and less accurately by older as compared to young adults2, 1822. Likewise, older adults comprehend ambiguous sentences less accurately than young adults, behaving much like low-span readers across age groups23. Finally, increasing sentence duration also decreases older adults’ sentence comprehension accuracy18, 19, 24: Poorer repetition accuracy was found for 8-word sentences compared to 5-word sentences in older but not in young adults24.

Despite substantial behavioural evidence for the age-related decline in working memory-intensive sentence comprehension, its neural underpinnings as measured by electrophysiological methods remain highly controversial. In healthy young adults, event-related theta power changes have not only been linked to verbal working memory in general25, 26 but also specifically to the verbal working memory demands during sentence comprehension6, 27. That means, theta power seems to be modulated both by domain-general and domain-specific working memory demands. For example, during general verbal working memory tasks, increased theta power has been found to predict better performance. In comparison, during sentence comprehension, increased theta power has been found for more compared to less complex sentences28 showing that the theta rhythm may be modulated by memory-taxing sentences29, too. The increase in theta power may be linked to increased synchronisation within the functional network of working memory29. In contrast to those event-related studies mentioned afore, in the resting state (RS), decreases in theta power seem to predict better verbal working memory3032. This may imply that, in the brain at rest, less specificity and more variability in frequency-specific neural synchrony may predict optimal functioning at task. This may be due to better readiness for settling into a new state during task-related processing. This is in line with proposals from previous studies33. In sum, that means, there is a critical gap in our understanding of the role of spontaneous neural activity, recorded in the brain at rest, particularly for the functional integrity of the language network. Our study is a first attempt to bridge this gap linking changes in RS theta power to working memory-intensive sentence comprehension.

Given the association between RS theta power and verbal working memory, and given the association between verbal working memory and sentence comprehension across the lifespan, RS theta power could be a plausible lifespan predictor of working memory-intensive sentence comprehension. This hypothesis, however, faces a paradox in the light of three findings (Fig. 1): (a) while RS theta power decreases across the lifespan34, (b) and low theta power in the RS indicates good verbal working memory, verbal working memory and sentence comprehension do not improve with age (c1) but instead deteriorate in older adults (c2). Hence, the association between RS theta power, verbal working memory, and sentence comprehension appears to change across the lifespan.

Figure 1
Schematic overview of paradox as a consequence of two coinciding patterns: (a) an age-related decrease in resting-state theta power and (b) lower resting-state theta power predicting good working memory performance. Taking (a) and (b) together, one would ...

Here we resolve this paradox, extending previous findings30: First, we established RS theta as the electrophysiological marker of working memory-intensive sentence comprehension across three age groups (young, middle-aged and older adults). To this end, we correlated RS theta power across the age trajectory with performance on a working memory-intensive sentence comprehension task. Second, we associated age-related decreases in RS theta power to domain-general cognitive abilities supporting sentence comprehension. This approach enabled us to assess the functional integrity of the brain networks underlying working memory-intensive sentence comprehension. Our findings established a crucial role of RS theta power for successful sentence comprehension that is preserved across the lifespan in spite of an age-related cognitive decline.

Results

The working memory-intensive sentence comprehension task used in this study assessed encoding success through a combination of a retrieval sentence and a comprehension question (see Table 1 and Methods). Behavioural performance was quantified through calculating d-prime scores as well as the percentage of correctly comprehended sentences (Fig. 2).

Table 1
Stimulus materials; encoding sentence with examples of both subject and object retrieval sentences as well as comprehension questions and subsequent feedback.
Figure 2
Behavioural results: (a) average d-prime scores; (b) average percentage of correctly comprehended sentences; both for the three age groups (25, 43 and 65 years of age, error bars indicate the standard error). For both measures, age groups differed in ...

The ANOVA of the behavioural data showed a main effect of age group, F(2,54) = 11.74, p = 5.86 × 10−5, η2 = 0.30, indicating that d-prime decreases with age (Fig. 2a). Similarly, the ANOVA regressing age group against the percentage of correctly comprehended sentences indicated that sentence comprehension declines with age, F(2,54) = 12.07, p = 4.66 × 10−5, η2 = 0.31 (Fig. 2b).

In addition to the working memory-intensive sentence comprehension task, a cognitive screening was performed to, on the one side, ensure healthy ageing of our participants and, on the other side, examine domain-general cognitive abilities associated with the working memory-intensive sentence comprehension task. The results showed significant correlations (FDR-corrected) between d-prime and measures of verbal working memory as well as attention (see Methods): the non-word repetition task (r = 0.45, p = 5 × 10−4), the digit span forward (r = 0.04, p = 5 × 10−4), the digit span backward (r = 0.51, p = 1 × 10−5) and the Auditory Flankers Task (r = −0.41, p = 2 × 10−3). All of those measures declined with age (non-word repetition: F = 0.45, p = 5 × 10−4; digit span forward: F = 0.04, p = 5 × 10−4; digit span backward: F = 0.51, p = 1 × 10−5; Auditory Flankers: F = −0.41, p = 2 × 10−3; Table 2). Given that those domain-general measures are all significantly correlated with the sentence comprehension task, this may suggest a role for the age-related decline in domain-general cognitive abilities in the increased difficulties in working memory-intensive sentence comprehension with age. This proposal is substantiated by our resting state findings.

Table 2
Age-related declines in domain-general cognitive abilities; reporting means (standard deviations) as well as the statistical output from the ANOVAs regressing age group against the neuropsychological measures.

Resting state EEG data from each subject were analysed in the frequency domain using the fast fourier transform (FFT). Statistical analyses were performed hypothesis-driven on the RS data in the theta frequency band. However, to control for effects within neighbouring frequency bands, we also looked at the delta (1–4 Hz), alpha (8–12 Hz), and beta bands (12–25 Hz), none of which showed any additional effects. For source localisation of the theta effect, we used a beamformer in the frequency domain. An analysis of covariance (ANCOVA) on the RS data at each electrode, regressing age group and RS theta power against d-prime, showed main effects of age group at all electrodes (all F(2,51) between 12.92 and 19.91, all p between 4.05 × 10−7 and 2.89 × 10−5, all η2 between 0.33 and 0.44; false discovery rate (FDR) corrected, and main effects of RS theta power (Fig. 3a) at electrodes TP9 (F(1,51) = 12.28, p = 9.62 × 10−4, η2 = 0.19; FDR corrected) and P9 (F(1,51) = 13.25, p = 6.36 × 10−4, η2 = 0.20; FDR corrected; Fig. 3b). There were no interaction effects at any electrode (all F(2,51) between 0.80 and 1.68, all p between 0.20 and 0.92) indicating that the relation between theta and d-prime was not affected by any age-related changes. The main effect of RS theta was source localised predominantly to the left-hemispheric dorso-frontal cortex (i.e., at the intersection of the left pre- and postcentral and middle frontal gyrus as well as the pars triangularis) and, less consistently, to the right anterior inferior and middle temporal gyrus (all F(1,51) between 4.05 and 11.34, p between 0.002 and 0.05, uncorrected; Table 3; Fig. 3c).

Figure 3
Mechanisms supporting sentence comprehension; (a–c) Overview of EEG results supporting a language-specific mechanism (ANCOVA on log RS theta power at each electrode, regressing age group and RS theta power against d-prime) (a) scatter plot of ...
Table 3
Source-level results (p < 0.05, uncorrected) of both the language-specific effect for RS theta power predicting sentence comprehension as well as the domain-general effect for RS theta.

The ANOVA at each electrode, regressing age group against RS theta power, showed main effects of age group (Fig. 3d) at 36 centro-frontal and -posterior electrodes (all F(2,54) between 3.76 and 7.92, all p between 9.60 × 10−3 and 0.03, all η2 between 0.12 and 0.23; Fig. 3e), showing that RS theta power decreases with age. This age-related effect was source-localised to bilateral midline regions (i.e., a region shared by bilateral supplementary motor areas, the left superior frontal gyrus and the right mid-cingulum; all F(2,54) between 3.17 and 12.65, all p between 3.00 × 10−5 and 0.05, uncorrected; Table 3, Fig. 3f). To assess whether this effect is related to a more specific cognitive function, a post-hoc ANCOVA was performed at each electrode regressing age group against RS theta power controlling for domain-general cognitive abilities earlier shown to support working memory-intensive sentence comprehension. Indeed, we found that the inclusion of the non-word repetition task in the model dissolved the effect of age group on theta (all F(2,51) between 0.47 and 7.54, FDR-corrected), indicating its relation to age-related theta decreases over midline regions. All other domain-general cognitive measures that were correlated with the experimental task did not impact the relationship between age group and RS theta power.

We reasoned that if the language-specific and the domain-general effect indeed reflect distinct functional networks, subserving distinct cognitive sub-processes of sentence comprehension, the networks should also exhibit a high degree of functional coupling internally, but a significantly lower degree of functional coupling with each other. To follow this post-hoc hypothesis, a source-level coherence analysis on the reconstructed dipole time courses was performed. The results showed that source-level coherence within each of the two networks was higher than coherence across those two networks (internal coherence within the language network: t(56) = −5.91, p = 2.10 × 10−7; internal coherence within the domain-general network: t(56) = −4.02, p = 1.75 × 10−4). In addition, the coherence of sources within the language network did not significantly differ from the coherence of sources within the domain-general network (t(56) = −1.11, p = 0.27). An additional source separation procedure showed that this result is not confounded by volume conduction but instead the time courses of the two networks cluster around different independent components (z = 3.76, p = 1.68 × 10−4). Together, this strongly supports a functional independence of the two resting state networks (Fig. 4).

Figure 4
Source-level coherence analysis shows that the coherence of sources within the language-specific network and also within the domain-general network is higher than the coherence of sources across those two networks (*** = p < 0.001, ...

In sum, we found both a performance-related, language-specific and an age-related, domain-general theta effect with distinct scalp topographies and neural generators. While the scalp topographies already hint at different underlying networks, the source localisation allows for a stronger claim that the scalp-level effects were generated by at least partially non-overlapping brain networks and reflect two distinct mechanisms—one specifically underlying sentence comprehension, the other related to an age-related impairment of domain-general cognitive functions supporting sentence comprehension.

Discussion

This study addresses a long-standing paradox: Prior research found RS theta power to be negatively correlated with verbal working memory abilities, but also to decrease with age. While this would predict performance in working memory-intensive sentence comprehension to increase with age, this is certainly not the case. Our results resolve this paradox, suggesting that RS theta oscillations dissociate two spatially and temporally distinct networks supporting different mechanisms. On the one side, RS theta power over left-hemispheric dorso-frontal and right anterior temporal regions was found to predict sentence comprehension abilities independently of age. On the other side, we found an age-related decline in RS theta power over central recording sites to be associated with verbal working memory. Moreover, such domain-general cognitive abilities supporting sentence comprehension could be functionally decoupled from the language-specific sentence comprehension abilities. Together this suggests that both older and young adults rely on the same core sentence processing mechanism35 located within the dorsal perisylvian language network3638—but that older adults exhibit a performance decline in working memory-intensive sentence comprehension due to changes within the association network subserving domain-general supplementary cognitive functions22.

Across the age trajectory, the results revealed an association between working memory-intensive sentence comprehension and RS theta power generated in regions frequently associated with syntactic processing3740: predominantly left-hemispheric dorso-frontal regions and, to a minor degree, the right anterior temporal lobe. This association is independent of age as no interaction was found between RS theta power and age group to explain the behavioural performance. Therefore, while many other cognitive abilities decline with age16, the results of the current study suggest that the core sentence comprehension network stays largely intact along the age trajectory. Possibly, the preservation of this performance-related mechanism throughout the lifespan hints at a compensatory role for age-related structural alterations of the language-relevant left-hemispheric dorso-frontal brain regions by functional activation shifts. Previous studies have shown a progressive longitudinal decrease in prefrontal grey and white matter volume across the age trajectory41, which coincides with functional activation shifts from posterior to anterior brain regions42. To substantiate this proposal, further investigation of the structural neuroanatomy underlying our theta generators is needed.

While the association between dorso-frontal RS theta power and working memory-intensive sentence comprehension can explain the behavioural results within each age group, it does not explain the substantial age-related decline in working memory-intensive sentence comprehension. Instead, we found an indication of spatially and temporally distinct RS theta oscillations that may instead drive this age-related decline in sentence comprehension performance. This effect is most pronounced over the central electrodes, Fz, Cz, Fpz and Pz, with neural generators predominantly in bilateral frontal to central midline regions. We could relate this age-related RS theta power decrease to variations in general verbal working memory performance, more precisely, to variations in the repetition of non-words of increasing syllable length indicating the capacity of the phonological store. Phonological storage, in turn, has been shown to play a role in working memory-demanding sentence comprehension43. Therefore, the age-related decline in RS theta power over bilateral frontal to central midline regions associated with working memory differences possibly contributes to the decline in working memory-intensive sentence comprehension. In fact, this is in line with previous findings associating theta over electrode Fz generated by sources along the cingulate as well as the medial frontal cortex with working memory4447. Such frontal midline theta (FMT) was found stronger for high compared to low performers in working memory tasks48. More specifically, FMT has been mainly linked to tasks involving order memory, but not item memory44. This could explain why the results only showed a link to the non-word repetition task but not to the other working memory tasks included in the cognitive screening (see Methods). The nature of this non-word repetition task involves the memory for the order of phonologically distinctive syllables together constituting individual non-words of increasing length. Hence the order of the syllables is crucial to remember the item. However, generally the link between age-related RS theta power decreases and functional variation in verbal working memory has to be tested more directly by future studies to validate our proposal.

The distinction between the age-related, domain-general and the performance-related, language-specific theta effect can also be observed from their distinct generators. In the mammalian brain, theta-band oscillations can be generated by the hippocampocortical loop—either by the hippocampus itself or by cortical regions within this loop49. On the one hand, hippocampal theta, associated with memory formation50, may be related to the age-related decline in RS theta power connected to domain-general functioning. On the other hand, cortical regions within the hippocampocortical loop may underlie the language-specific mechanism observed here. This also implies that the location of a theta generator determines its functionality being either language-specific or domain-general. However, this needs further investigation with the help of spatially fine-grained methodologies. Furthermore, it suggests that future studies should use paradigms which manipulate the stimuli in such a way that language-specific processes can be disentangled from domain-general working memory processes. Such paradigms would ensure the dissociation of different underlying oscillatory networks. This way the field of neuronal oscillations could move forward to categorising frequency bands not only by their function but also by their neural generators.

In sum, this study is the first to identify two spatially as well as temporally distinct theta band oscillations subserving distinct functions during sentence processing. The present results show that RS theta power across left-hemispheric dorso-frontal language regions underlies a language-specific mechanism and hence may serve as robust electrophysiological marker of intact sentence comprehension across the lifespan. Our results, moreover, suggest a dissociable domain-general mechanism associated with a decline of supplementary cognitive functions supporting working memory-intensive sentence comprehension—reflected in decreased RS theta power over bilateral midline regions. In sum, we suggest that working memory-intensive sentence comprehension declines with age due to underlying changes in domain-general brain networks, despite the preservation of the core sentence processing network across the lifespan.

Methods

Participants

Data from 57 healthy right-handed51 participants divided into three equally-sized age groups (young: 9 male; mean age: 25 years; SD: 1 year; middle-aged: 9 male; mean age: 43 years; SD: 2 years; older: 8 male; mean age: 65; SD: 3 years) were analysed. Participants were native speakers of German, had normal or corrected-to-normal vision, and did not suffer from neurological disorders or hearing loss (hearing threshold  25 dB as assessed by standard audiometry, Oscilla® SM910-B, Aarhus, Denmark). Participants were tested for cognitive impairment (Mini-Mental State Examination 252), assuring that no participant suffered from dementia. Education was matched across age groups (≥14 years of education). All participants gave written informed consent prior to any testing. The study was approved by the ethics committee of the University of Leipzig. The experiment was conducted according to the approved guidelines.

Neuropsychological Measures

On the first day of testing, all participants underwent a cognitive screening to ensure healthy ageing. The screening included measures of working memory (digit span forward and backward53, non-word repetition task54, and Counting Span55), verbal intelligence (similarities task and vocabulary task), non-verbal intelligence (matrices and block tasks53), and auditory attention (Auditory Flankers56).

Stimuli

The experimental items for the working memory-intensive sentence comprehension task were 128 encoding sentences, each consisting of two conjoined clauses containing one subject and two direct objects each. Subjects and objects were all animate in order to avoid associated processing differences57. Additionally, subjects and objects differed in grammatical gender within clause, which later served as a retrieval cue. The gender of subjects and objects was counterbalanced across the stimulus set. Subjects and objects were matched for word length (i.e., number of syllables) and word frequency class within and across sentences58. This minimised processing differences due to both word length59 as well as word frequency60. Furthermore, this ensured similar encoding demands for each word, which are affected by both syllable count61 and lexical frequency62. Subjects’ encoding performance was assessed through a combination of a retrieval sentence and a comprehension question (see Table 1). The retrieval sentence operationalised retrieval via two pronouns, one of the subject’s and one of the object’s gender, thus referring to a unique subject or object within the encoding sentence. In the example retrieval sentence Die von ihr Angekündigte war nervös (The one (she) announced by her was nervous), the phrase die von ihr is pointing to exactly one noun of the encoding sentence. Thereby, ihr (her) is pointing to the subject, die Moderatorin (the presenter) who acts upon the female object, die Künstlerin (the artist), referred to by die (she). The subsequent comprehension question assessed retrieval success (and thus, indirectly, encoding success of the encoding sentence) by asking either for the correct item (Was the artist nervous?), a gender lure (Was the actor nervous?) or a word category lure (Was the presenter nervous?). The introduction of lure questions aimed at keeping the amount of correct yes- and no-responses equal. Moreover, the lures necessitated whole sentence processing as category lures emphasised the encoding of both the conjoined clauses whereas gender lures emphasised the encoding of the gender information of all nouns.

Taking into account the number of clauses (i.e., 2), of objects within clauses (i.e., 2) and of subjects per sentence (i.e., 2), 8 variants of each encoding sentence were created in order to counterbalance the gender information of each noun. Those 8 variants were combined with 4 variants of the retrieval sentence varying in gender information of the 2 determiners ensuring that the retrieved items were half male and half female. Hence there were 32 combinations for each of the 128 experimental items which were distributed via Latin Square across 32 lists. This distribution also accounted for the various retrieval types (subject or object retrieval), the retrieval position within the encoding sentence, as well as question types (correct or lure) across lists (within items) and also within lists (across items). With 19 participants in each age group not all 32 variants of each experimental items could be presented. However, this does not pose a limitation as balancing the factors within each list assured equivalent conditions for each participant.

In addition to the experimental items, 64 filler items were included to avoid habituation effects and the buildup of experimental strategies. In comparison to the experimental items, the encoding sentences of the filler items were syntactically more complex (i.e., object-relative clauses and topicalisations). In order to further maximise the difference to the experimental items where gender information was introduced in form of suffixes (e.g. act-or – act-ress), in filler sentences gender information was introduced in form of nouns of biological gender (e.g. uncle – aunt). However, while syntactic differences were maximised, other features like syllable count and word frequency were matched to the experimental items in order to disguise the interests of this study. Furthermore, all content words across experimental and filler stimuli were uniquely used in order to avoid memory consolidation effects, which could have confounded the results. The 64 filler items were added to the 128 experimental items for each of the 32 lists, which were then pseudo-randomised.

Procedure

The experiment was conducted over two days within a period of no more than 7 days. On the first day, participants underwent audiometer testing and cognitive screening (see Neuropsychological Measures). On the second day, we recorded the electroencephalogram (EEG) at rest and then during the working memory-intensive sentence comprehension task. The sentence comprehension task involved an encoding and a retrieval sentence as well as a comprehension question, and feedback to the subject (see Table 1), in this order. Auditory stimuli were presented via headphones (Sennheiser, HD202). To ensure the same hearing level for all participants, the volume was adjusted to 38 dB above the individual hearing threshold (as determined by the method of limits63 on the day of the experiment before setting up the EEG). Visual stimuli were presented in white font (Arial, size 30) on a grey screen (17 inch, Sony Multiscan E220). An experimental trial started with the auditory presentation of an encoding sentence followed by a jittered pause of 1.0–1.5 seconds. A visual retrieval sentence was then presented, querying either a subject or an object from the preceding stimulus sentence (Table 1). A jittered pause of 1.5–2.0 seconds followed. A comprehension question followed, asking for the correct answer (CORR; Table 1) in 50% of trials, for a gender lure (GEN-L; Table 1) in 25% of trials and for a syntactic category lure (CAT-L; Table 1) in 25% of trials. The participants had a response time window of 4 seconds to respond with either “yes” or “no” by pressing either of two single-button response boxes placed individually under their left and right index finger. Button assignment was counterbalanced across participants. After the response was given, visual feedback was provided in form of a happy or sad emoticon. A trial ended with a jittered inter-trial interval of between 1.5–2.0 seconds.

Data acquisition

Resting state EEG from each subject was acquired for 5 minutes with the subject’s eyes closed (EC) followed by 5 minutes with the subject’s eyes open (EO), with eye opening prompted by a pure tone of 500 Hz presented at 85 dB. EC and EO were both recorded to enable the calculation of subjects’ individual theta frequency (see Data analysis). Data were acquired within a pass-band from DC to 270 Hz at a sampling rate of 1,000 Hz from 63 electrodes. The setup was referenced against the left mastoid and grounded to the sternum. The vertical and horizontal electro-oculogram (EOG) was acquired with bipolar electrodes below and above the right eye as well as at the outer canthi of both eyes, respectively. The scalp electrodes were placed according to the international 10–20 system in an elastic cap (Electro Cap International, Inc., Eaton, OH, USA) connected to a 72-channel Refa amplifier (TMS International B.V., Odenzaal, The Netherlands). Electrode impedances were kept below 5 kΩ.

Data analysis

All analyses were performed in MATLAB® (The MathWorks, Inc., Natick, MA, USA). EEG data were analysed using the Fieldtrip toolbox64. First, two differently filtered data sets were generated from the raw RS EEG data (EC and EO): First, a band-pass filtered data set for statistical analysis (zero-phase two-pass fourth-order 0.5–45 Hz Butterworth filter); second, a high-pass filtered data set, optimal for independent component analysis (ICA; zero-phase finite-impulse-response one-pass 3624th-order 1-Hz Kaiser filter65). Both data sets were re-referenced to the average of all electrodes (excluding the EOG) and segmented into 2-second epochs to facilitate artefact detection. Artefact detection was performed on the high-pass-filtered data set, involving a two-step procedure: First, muscle artefacts were detected using a semi-automatic, distribution-based approach (z > 9) and rejected after visual inspection (EC: mean rejection rate = 8.55%, SD = 7.36%; EO: mean rejection rate = 11.22%, SD = 6.97%), taking into account the waveform morphology. In a second step, the data were down-sampled to 300 Hz to accelerate the subsequent ICA. The first 40 independent components (IC) were calculated and those reflecting heartbeat, eye movements or eye blinks as well as electrode noise were detected based on visual inspection of the components’ waveform morphology, power spectrum, and scalp topography. Artefact ICs were removed from the band-pass filtered data (mean number of rejected ICs = 11.70, SD = 1.99; same for all age groups, F(2,56) = 2.26, p = 0.11), which thereafter was used for all further analyses. After removing A1 and A2 due to leftover noise, the data were re-referenced to the average of all remaining electrodes, and the mean potential within trial was subtracted. The data set was then split into EC and EO. The number of RS trials was pseudo-randomly reduced to the common minimum of 90 trials across all participants—first, trials from the first 30 seconds of the recordings were chosen, and then, trials from the remainder of the data were selected randomly. Using a Hanning taper, frequency analysis via FFT was then performed in steps of 1 Hz from 1–25 Hz on the EC and EO data. Oscillatory power was log-transformed. Because of substantial inter-individual variance in individual band peak frequencies within and across age groups66, each subject’s individual alpha peak frequency was detected in order to determine the individual frequency range of the theta band26, 67. First, we preselected an extended alpha frequency range (i.e., 7–14 Hz) and electrodes of interest (Pz, P3, P4, P5, P6, P7, P8, P9, P10, POz, PO3, PO4, PO7, PO8, O1, O2, and Oz) separately for the EC and EO data68. Second, the EO power spectrum within this range was subtracted from the EC power spectrum. Third, the maximum peak within the difference spectrum was defined as the individual alpha peak frequency (IAF)69. IAF was then used as anchor point to constrain the individual theta frequency band into a range from (IAF – 6 Hz) to (IAF – 2 Hz) (adapted from26, 70). Subsequent RS analyses were then performed on the EC data71.

The effects (see Results) were source-localised using a frequency-domain beamformer (dynamic imaging of coherent sources, DICS72). In a first step, a volume conductor was made based on a template Boundary Element Method (BEM) head model73. In a second step, this template head model was joined with template sensor positions in accordance to the electrode set up used in this study to create a forward model within a 1-cm-spaced three-dimensional grid that was used for all participants. In a third step, the spatial filter estimated from the cross-spectral density matrix of each subject the power of the sources along every point of the grid, with a signal-to-noise ratio of 3%. Finally, the Neural Activity Index74 was computed to remove the center-of-head bias.

For source-level coherence analysis, three-dimensional time courses of source moments were derived from the single-segment preprocessed EEG data by multiplying the single-segment EEG with the individual spatial filter from the DICS beamforming analysis. The source grid was masked for the significant grid points (p < 0.05, uncorrected), and three-dimensional dipole moments were extracted for each volume-conductor grid point within the language-specific and the domain-general effects, respectively. The time course for each grid point then underwent singular value decomposition, arriving at a single time course per grid point6, 72. On these time courses, the same FFT frequency analysis that was used for the sensor-level data was performed, now restricted to the individual theta band. Finally, coherence analysis was performed on the complex Fourier-spectra75 for all pairs of dipoles within and across the two source-level effects. However, as coherence between areas decays with increasing inter-area distance due to volume conduction76, the results may be confounded by larger inter-network and smaller intra-network source distances. For this reason, a supplementary analysis was performed using a blind source separation algorithm within individual. To this end, first, the single-trial source-level time courses from the coherence analysis were downsampled to 100 Hz to speed up later source separation. Second, the time courses were band-pass filtered (finite-impulse-response two-pass 266th-order filter77) to match the individual theta band range. Third, an ICA was used to extract 10 independent components, aiming to show that the grid points of the two separate networks would indeed cluster around two different independent components, indicating their functional dissociation. Fourth, the component patterns were masked for all grid points included in the two networks. For each grid point, we then determined the independent component that was maximally represented at the grid points.

Statistical analysis

Two performance measures were computed: the percentage of correctly comprehended sentences and d-prime scores. D-prime scores were calculated by subtracting the z-transformed false alarm rate (FA) from the z-transformed hit rate (H) (d-prime = z(H) − z(FA)). FA or H of 0 was corrected by 1/N and FA or H of 1 was corrected by (N-1)/N, with N being the number of trials78. Analyses of variance (ANOVAs) regressed age group separately against both performance measures. All further analyses were done on the basis of d-prime, as it provides the more sensitive behavioural measure. D-prime was related to the EEG data to assess the relevance of RS theta power for working memory-intensive sentence comprehension. Analyses of covariance (ANCOVAs) were computed for each electrode with age group as factor and RS theta power as covariate of interest, both predicting d-prime79. To assess age-related differences in RS theta power, an ANOVA was computed at each electrode with age group as factor predicting theta power. To account for multiple comparisons, the false discovery rate (FDR80) procedure was used to correct p-values across electrodes. Additionally, those age-related differences in RS theta power were related to variations in domain-general cognitive functions. For this purpose, in a first step, the experimental measure, d-prime, was correlated with the performance measures of the cognitive screening. In a second step, ANCOVAs were performed at each electrode regressing age group against RS theta power controlling for those measures significantly related to d-prime (after FDR-correction). Measures dissolving the relation between age group and RS theta power indicate their link to age-related RS theta power differences. Finally, in analogy to the sensor-level data, an ANCOVA was performed, with source power predicting behavioural performance depending on the age group. Similarly, an ANOVA was performed with age group as factor predicting the source power. The anatomical labels for the significant clusters (uncorrected) were determined from the MNI atlas81. For statistical analysis of source-level coherences, a series of paired-samples t-test was performed, separately comparing the mean coherence of dipole moments within each network (i.e., internal coherence of all dipole pairs within the language network, internal coherence of all dipole pairs within the domain-general network) to the mean coherence of dipole moments across the two networks (i.e., coherence between all pairs that included one dipole from each the language network and the domain-general network). In addition, we also compared the internal coherence of the language network to the internal coherence of the domain-general network. This comparison was substantiated by an additional statistical analysis on source separated component data. We first assessed the difference in the distributions of independent components across networks using within-participant Wilcoxon rank sum tests. The single-subject test statistics from these tests were then submitted to group-level analysis using a one-sample Wilcoxon signed-rank test, hypothesising that the two networks cluster around two different components.

Acknowledgements

The authors thank David Poeppel for his invaluable and constructive comments on this manuscript. We thank Maren Grigutsch for her comments on parts of the analysis. Also, we are obliged to thank Cornelia Henschel, Marie-Theres Götze, and Johanna Tegtmeyer for their contribution to data acquisition.

Author Contributions

Author Contributions

All authors have planned and designed this study. C.B. and B.V. collected the data. C.B. analysed the data. C.B., L.M. and A.F. wrote the manuscript.

Notes

Competing Interests

The authors declare that they have no competing interests.

Footnotes

Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Friederici AD, Schriefers H, Lindenberger U. Differential Age Effects on Semantic and Syntactic Priming. Int. J. Behav. Dev. 1998;22:813–845. doi: 10.1080/016502598384180. [Cross Ref]
2. Caplan D, Dede G, Waters G, Michaud J, Tripodis Y. Effects of age, speed of processing, and working memory on comprehension of sentences with relative clauses. Psychol. Aging. 2011;26:439–450. doi: 10.1037/a0021837. [PubMed] [Cross Ref]
3. Just MA, Carpenter PA. A capacity theory of comprehension: individual differences in working memory. Psychol. Rev. 1992;99:122–149. doi: 10.1037/0033-295X.99.1.122. [PubMed] [Cross Ref]
4. Gibson, E. The dependency locality theory: a distance-based theory of linguistic complexity. in Image, language, brain (ed. Y. Miyashita, A. Marantz & W. O’Neil) 95–126 (MIT Press, Cambridge, MA, 2000).
5. Lewis RL, Vasishth S, Van Dyke JA. Computational principles of working memory in sentence comprehension. Trends Cogn. Sci. 2006;10:447–454. doi: 10.1016/j.tics.2006.08.007. [PMC free article] [PubMed] [Cross Ref]
6. Meyer L, Grigutsch M, Schmuck N, Gaston P, Friederici AD. Frontal-posterior theta oscillations reflect memory retrieval during sentence comprehension. Cortex. 2015;71:205–218. doi: 10.1016/j.cortex.2015.06.027. [PubMed] [Cross Ref]
7. Friederici AD, Steinhauer K, Mecklinger A, Meyer M. Working memory constraints on syntactic ambiguity resolution as revealed by electrical brain responses. Biol. Psychol. 1998;47:193–221. doi: 10.1016/S0301-0511(97)00033-1. [PubMed] [Cross Ref]
8. King J, Just MA. Individual differences in syntactic processing: The role of working memory. J. Mem. Lang. 1991;30:580–602. doi: 10.1016/0749-596X(91)90027-H. [Cross Ref]
9. Vos SH, Gunter TC, Kolk HH, Mulder G. Working memory constraints on syntactic processing: an electrophysiological investigation. Psychophysiology. 2001;38:41–63. doi: 10.1111/1469-8986.3810041. [PubMed] [Cross Ref]
10. Gordon PC, Hendrick R, Johnson M. Memory interference during language processing. J. Exp. Psychol. Learn. Mem. Cogn. 2001;27:1411–1423. doi: 10.1037/0278-7393.27.6.1411. [PubMed] [Cross Ref]
11. Tan Y, Martin RC, Van Dyke JA. Semantic and Syntactic Interference in Sentence Comprehension: A Comparison of Working Memory Models. Front. Psychol. 2017;8:198. doi: 10.3389/fpsyg.2017.00198. [PMC free article] [PubMed] [Cross Ref]
12. Van Dyke JA, McElree B. Retrieval interference in sentence comprehension. J. Mem. Lang. 2006;55:157–166. doi: 10.1016/j.jml.2006.03.007. [PMC free article] [PubMed] [Cross Ref]
13. Mecklinger A, Schriefers H, Steinhauer K, Friederici AD. Processing relative clauses varying on syntactic and semantic dimensions: an analysis with event-related potentials. Mem. Cognit. 1995;23:477–494. doi: 10.3758/BF03197249. [PubMed] [Cross Ref]
14. Kaan E, Harris A, Gibson E, Holcomb P. The P600 as an index of syntactic integration difficulty. Lang. Cogn. Process. 2000;15:159–201. doi: 10.1080/016909600386084. [Cross Ref]
15. McElree B, Foraker S, Dyer L. Memory structures that subserve sentence comprehension. J. Mem. Lang. 2003;48:67–91. doi: 10.1016/S0749-596X(02)00515-6. [Cross Ref]
16. Dobbs AR, Rule BG. Adult age differences in working memory. Psychol. Aging. 1989;4:500–503. doi: 10.1037/0882-7974.4.4.500. [PubMed] [Cross Ref]
17. Salthouse TA. The Aging of Working Memory. Neuropsychology. 1994;8:535–543. doi: 10.1037/0894-4105.8.4.535. [Cross Ref]
18. Kemper S. Imitation of complex syntactic constructions by elderly adults. Appl. Psycholinguist. 1986;7:277–287. doi: 10.1017/S0142716400007578. [Cross Ref]
19. Kemper S. Syntactic complexity and elderly adults’ prose recall. Exp. Aging Res. 1987;13:47–52. doi: 10.1080/03610738708259299. [PubMed] [Cross Ref]
20. Obler LK, Fein D, Nicholas M, Albert ML. Auditory comprehension and aging: Decline in syntactic processing. Appl. Psycholinguist. 1991;12:433–452. doi: 10.1017/S0142716400005865. [Cross Ref]
21. Stine-Morrow EA, Ryan S, Leonard JS. Age differences in on-line syntactic processing. Exp. Aging Res. 2000;26:315–322. doi: 10.1080/036107300750015714. [PubMed] [Cross Ref]
22. Wingfield A, Grossman M. Language and the aging brain: patterns of neural compensation revealed by functional brain imaging. J. Neurophysiol. 2006;96:2830–2839. doi: 10.1152/jn.00628.2006. [PubMed] [Cross Ref]
23. Kemper S, Crow A, Kemtes K. Eye-fixation patterns of high- and low-span young and older adults: down the garden path and back again. Psychol. Aging. 2004;19:157–170. doi: 10.1037/0882-7974.19.1.157. [PubMed] [Cross Ref]
24. Wingfield A, Poon LW, Lombardi L, Lowe D. Speed of processing in normal aging: effects of speech rate, linguistic structure, and processing time. J. Gerontol. 1985;40:579–585. doi: 10.1093/geronj/40.5.579. [PubMed] [Cross Ref]
25. Karrasch M, Laine M, Rapinoja P, Krause CM. Effects of normal aging on event-related desynchronization/synchronization during a memory task in humans. Neurosci. Lett. 2004;366:18–23. doi: 10.1016/j.neulet.2004.05.010. [PubMed] [Cross Ref]
26. Klimesch W. EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis. Brain. Res. Rev. 1999;29:169–195. doi: 10.1016/S0165-0173(98)00056-3. [PubMed] [Cross Ref]
27. Bastiaansen MC, van Berkum JJ, Hagoort P. Event-related theta power increases in the human EEG during online sentence processing. Neurosci. Lett. 2002;323:13–16. doi: 10.1016/S0304-3940(01)02535-6. [PubMed] [Cross Ref]
28. Weiss S, et al. Increased neuronal communication accompanying sentence comprehension. Int. J. Psychophysiol. 2005;57:129–141. doi: 10.1016/j.ijpsycho.2005.03.013. [PubMed] [Cross Ref]
29. Bastiaansen MC, van Berkum JJ, Hagoort P. Syntactic processing modulates the theta rhythm of the human EEG. Neuroimage. 2002;17:1479–1492. doi: 10.1006/nimg.2002.1275. [PubMed] [Cross Ref]
30. Heister D, et al. Resting-state neuronal oscillatory correlates of working memory performance. PLoS One. 2013;8:e66820. doi: 10.1371/journal.pone.0066820. [PMC free article] [PubMed] [Cross Ref]
31. Reichert JL, Kober SE, Witte M, Neuper C, Wood G. Age-related effects on verbal and visuospatial memory are mediated by theta and alpha II rhythms. Int. J. Psychophysiol. 2016;99:67–78. doi: 10.1016/j.ijpsycho.2015.11.004. [PubMed] [Cross Ref]
32. Roca-Stappung M, et al. Healthy aging: relationship between quantitative electroencephalogram and cognition. Neurosci. Lett. 2012;510:115–120. doi: 10.1016/j.neulet.2012.01.015. [PubMed] [Cross Ref]
33. Garrett DD, Kovacevic N, McIntosh AR, Grady CL. Blood oxygen level-dependent signal variability is more than just noise. J. Neurosci. 2010;30:4914–4921. doi: 10.1523/JNEUROSCI.5166-09.2010. [PubMed] [Cross Ref]
34. Vlahou EL, Thurm F, Kolassa IT, Schlee W. Resting-state slow wave power, healthy aging and cognitive performance. Sci. Rep. 2014;4:5101. doi: 10.1038/srep05101. [PMC free article] [PubMed] [Cross Ref]
35. Tyler LK, et al. Preserving syntactic processing across the adult life span: the modulation of the frontotemporal language system in the context of age-related atrophy. Cereb. Cortex. 2010;20:352–364. doi: 10.1093/cercor/bhp105. [PMC free article] [PubMed] [Cross Ref]
36. Catani M, Jones DK, ffytche DH. Perisylvian language networks of the human brain. Ann. Neurol. 2005;57:8–16. doi: 10.1002/ana.20319. [PubMed] [Cross Ref]
37. Friederici AD. The brain basis of language processing: from structure to function. Physiol. Rev. 2011;91:1357–1392. doi: 10.1152/physrev.00006.2011. [PubMed] [Cross Ref]
38. Price CJ. The anatomy of language: a review of 100 fMRI studies published in 2009. Ann. N. Y. Acad. Sci. 2010;1191:62–88. doi: 10.1111/j.1749-6632.2010.05444.x. [PubMed] [Cross Ref]
39. Meyer L, Obleser J, Kiebel SJ, Friederici AD. Spatiotemporal dynamics of argument retrieval and reordering: an FMRI and EEG study on sentence processing. Front. Psychol. 2012;3:523. [PMC free article] [PubMed]
40. Saur D, et al. Ventral and dorsal pathways for language. Proc. Natl. Acad. Sci. USA. 2008;105:18035–18040. doi: 10.1073/pnas.0805234105. [PubMed] [Cross Ref]
41. Raz N, et al. Regional brain changes in aging healthy adults: general trends, individual differences and modifiers. Cereb. Cortex. 2005;15:1676–1689. doi: 10.1093/cercor/bhi044. [PubMed] [Cross Ref]
42. Davis SW, Dennis NA, Daselaar SM, Fleck MS, Cabeza R. Que PASA? The posterior-anterior shift in aging. Cereb. Cortex. 2008;18:1201–1209. doi: 10.1093/cercor/bhm155. [PMC free article] [PubMed] [Cross Ref]
43. Pettigrew C, Hillis AE. Role for Memory Capacity in Sentence Comprehension: Evidence from Acute Stroke. Aphasiology. 2014;28:1258–1280. doi: 10.1080/02687038.2014.919436. [PMC free article] [PubMed] [Cross Ref]
44. Hsieh LT, Ranganath C. Frontal midline theta oscillations during working memory maintenance and episodic encoding and retrieval. Neuroimage. 2014;85:721–729. doi: 10.1016/j.neuroimage.2013.08.003. [PMC free article] [PubMed] [Cross Ref]
45. Onton J, Delorme A, Makeig S. Frontal midline EEG dynamics during working memory. Neuroimage. 2005;27:341–356. doi: 10.1016/j.neuroimage.2005.04.014. [PubMed] [Cross Ref]
46. Scheeringa R, et al. Frontal theta EEG activity correlates negatively with the default mode network in resting state. Int. J. Psychophysiol. 2008;67:242–251. doi: 10.1016/j.ijpsycho.2007.05.017. [PubMed] [Cross Ref]
47. Scheeringa R, et al. Trial-by-trial coupling between EEG and BOLD identifies networks related to alpha and theta EEG power increases during working memory maintenance. Neuroimage. 2009;44:1224–1238. doi: 10.1016/j.neuroimage.2008.08.041. [PubMed] [Cross Ref]
48. Hsieh LT, Ekstrom AD, Ranganath C. Neural oscillations associated with item and temporal order maintenance in working memory. J. Neurosci. 2011;31:10803–10810. doi: 10.1523/JNEUROSCI.0828-11.2011. [PMC free article] [PubMed] [Cross Ref]
49. Miller, R. Cortico-hippocampal Interplay and the Representation of Context in the Brain. (Springer Verlag, Berlin, 1991).
50. Daselaar SM, et al. Parahippocampal activation during successful recognition of words: a self-paced event-related fMRI study. Neuroimage. 2001;13:1113–1120. doi: 10.1006/nimg.2001.0758. [PubMed] [Cross Ref]
51. Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia. 1971;9:97–113. doi: 10.1016/0028-3932(71)90067-4. [PubMed] [Cross Ref]
52. Folstein, M.F., Folstein, S.E., White, T. & Messer, M.A. MMSE-2: Mini-mental state examination 2nd Edition. Psychological Assessment Resources. (Inc., Lutz, FL., 2010).
53. Aster, M.v., Neubauer, A. & Horn, R. eds Wechsler Intelligenztest für Erwachsene (Pearson Assessment, Frankfurt, 2006).
54. Welte V. Der Mottier-Test, ein Prüfmittel für die Lautdifferenzierungsfähigkeit und die auditive Merkfähigkeit. Sprache Stimme Gehör. 1981;5:121–125.
55. Case, R., Kurland, M. D. & Goldberg, J. Operational efficiency and the growth of short-term memory span. J. Exp. Child Psychol. 33 (1982).
56. Chan JS, Merrifield K, Spence C. Auditory spatial attention assessed on a flanker interference task. Acta Acust. united Ac. 2005;91:554–563.
57. Bonin P, Gelin M, Bugaiska A. Animates are better remembered than inanimates: further evidence from word and picture stimuli. Mem. Cognit. 2014;42:370–382. doi: 10.3758/s13421-013-0368-8. [PubMed] [Cross Ref]
58. Goldhahn, D., Eckart, T. & Quasthoff, U. Building Large Monolingual Dictionaries at the Leipzig Corpora Collection: From 100 to 200 Languages. in Proceedings of the Eighth International Conference on Language Resources and Evaluation (2012).
59. Spoehr KT, Smith EE. The role of syllables in perceptual processing. Cogn. Psychol. 1973;5:71–89. doi: 10.1016/0010-0285(73)90026-1. [Cross Ref]
60. Allen M, Badecker W, Osterhout L. Morphological analysis in sentence processing: An ERP study. Lang. Cogn. Process. 2003;18:405–430. doi: 10.1080/01690960244000054. [Cross Ref]
61. Luo H, Husain FT, Horwitz B, Poeppel D. Discrimination and categorization of speech and non-speech sounds in an MEG delayed-match-to-sample study. Neuroimage. 2005;28:59–71. doi: 10.1016/j.neuroimage.2005.05.040. [PubMed] [Cross Ref]
62. Halgren E, et al. N400-like magnetoencephalography responses modulated by semantic context, word frequency, and lexical class in sentences. Neuroimage. 2002;17:1101–1116. doi: 10.1006/nimg.2002.1268. [PubMed] [Cross Ref]
63. Herrmann B, Henry MJ, Haegens S, Obleser J. Temporal expectations and neural amplitude fluctuations in auditory cortex interactively influence perception. Neuroimage. 2016;124:487–497. doi: 10.1016/j.neuroimage.2015.09.019. [PubMed] [Cross Ref]
64. Oostenveld R, Fries P, Maris E, Schoffelen JM. FieldTrip: Open source software for advanced analysis of MEG, EEG, and invasive electrophysiological data. Comput. Intell. Neurosci. 2011;2011:156869. doi: 10.1155/2011/156869. [PMC free article] [PubMed] [Cross Ref]
65. Winkler I, Debener S, Muller KR, Tangermann M. On the influence of high-pass filtering on ICA-based artifact reduction in EEG-ERP. Conf. Proc. IEEE Eng. Med. Biol. Soc. 2015;2015:4101–4105. [PubMed]
66. Koepruner V, Pfurtscheller G, Auer LM. Quantitative EEG in normals and in patients with cerebral ischemia. Prog. Brain Res. 1984;62:29–50. doi: 10.1016/S0079-6123(08)62168-8. [PubMed] [Cross Ref]
67. Clark RC, et al. Spontaneous alpha peak frequency predicts working memory performance across the age span. Int. J. Psychophysiol. 2004;53:1–9. doi: 10.1016/j.ijpsycho.2003.12.011. [PubMed] [Cross Ref]
68. Grandy TH, et al. Peak individual alpha frequency qualifies as a stable neurophysiological trait marker in healthy younger and older adults. Psychophysiology. 2013;50:570–582. doi: 10.1111/psyp.12043. [PubMed] [Cross Ref]
69. Lansbergen MM, Arns M, van Dongen-Boomsma M, Spronk D, Buitelaar JK. The increase in theta/beta ratio on resting-state EEG in boys with attention-deficit/hyperactivity disorder is mediated by slow alpha peak frequency. Prog. Neuropsychopharmacol. Biol. Psychiatry. 2011;35:47–52. doi: 10.1016/j.pnpbp.2010.08.004. [PubMed] [Cross Ref]
70. Caso, F. et al. Quantitative EEG and LORETA: valuable tools in discerning FTD from AD? Neurobiol. Aging, 2343–2356 (2012). [PubMed]
71. Finnigan S, Robertson IH. Resting EEG theta power correlates with cognitive performance in healthy older adults. Psychophysiology. 2011;48:1083–1087. doi: 10.1111/j.1469-8986.2010.01173.x. [PubMed] [Cross Ref]
72. Gross J, et al. Dynamic imaging of coherent sources: Studying neural interactions in the human brain. Proc. Natl. Acad. Sci. USA. 2001;98:694–699. doi: 10.1073/pnas.98.2.694. [PubMed] [Cross Ref]
73. Oostenveld R, Stegeman DF, Praamstra P, van Oosterom A. Brain symmetry and topographic analysis of lateralized event-related potentials. Clin. Neurophysiol. 2003;114:1194–1202. doi: 10.1016/S1388-2457(03)00059-2. [PubMed] [Cross Ref]
74. Van Veen BD, van Drongelen W, Yuchtman M, Suzuki A. Localization of brain electrical activity via linearly constrained minimum variance spatial filtering. IEEE Trans. Biomed. Eng. 1997;44:867–880. doi: 10.1109/10.623056. [PubMed] [Cross Ref]
75. Rappelsberger P, Petsche H. Probability mapping: power and coherence analyses of cognitive processes. Brain Topogr. 1988;1:46–54. doi: 10.1007/BF01129339. [PubMed] [Cross Ref]
76. Winter WR, Nunez PL, Ding J, Srinivasan R. Comparison of the effect of volume conduction on EEG coherence with the effect of field spread on MEG coherence. Stat Med. 2007;26:3946–3957. doi: 10.1002/sim.2978. [PubMed] [Cross Ref]
77. Parks T, McClellan J. Chebyshev approximation for nonrecursive digital filters with linear phase. IEEE Trans Circuit Theory. 1972;19:189–194. doi: 10.1109/TCT.1972.1083419. [Cross Ref]
78. Macmillan, N. A. & Creelman, C. D. Detection Theory: A User’s Guide (Lawrence Erlbaum Associates, New York, 2005).
79. Eid, M., Gollwitzer, M. & Schmitt, M. Statistik und Forschungsmethoden. (Beltz, Basel, 2010).
80. Benjamini Y, Hochberg Y. Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. J. Roy. Stat. Soc. B. Met. 1995;57:289–300.
81. Mazziotta J, et al. A probabilistic atlas and reference system for the human brain: International Consortium for Brain Mapping (ICBM) Philos. Trans. R. Soc. Lond. B. Biol. Sci. 2001;356:1293–1322. doi: 10.1098/rstb.2001.0915. [PMC free article] [PubMed] [Cross Ref]

Articles from Scientific Reports are provided here courtesy of Nature Publishing Group