Increased neuronal communication accompanying sentence comprehension

https://doi.org/10.1016/j.ijpsycho.2005.03.013Get rights and content

Abstract

The main purpose of this study was to examine large-scale oscillatory activity and frequency-related neuronal synchronization during the comprehension of English spoken sentences of different complexity. Therefore, EEG coherence during the processing of subject–subject (SS)- and more complex subject–object (SO)-relatives was computed using an adaptive fitting approach of bivariate auto-regressive moving average (ARMA) models which enabled the continuous calculation of coherence in the course of sentence processing with a high frequency resolution according to the dynamic changes of the EEG signals.

Coherence differences between sentence types were observed in the theta (4–7 Hz), beta-1 (13–18 Hz) and gamma (30–34 Hz) frequency ranges, though emerging during the processing of different parts of these sentences: gamma differences were evident mainly during the relative clause while theta and beta-1 differed significantly following the end of the relative clause. These findings reveal no simple one to one map between EEG frequencies and cognitive operations necessary for sentence comprehension. Instead, they indicate a complex interplay and dynamic interaction between different EEG frequencies and verbal working memory, episodic memory, attention, morpho-syntactic and semantic–pragmatic analyses, which though distinct often co-occur.

Introduction

During the course of a sentence, the language system must perform a series of analytic and integrative functions including auditory perception, phonological analysis, lexical access, morpho-syntactic, prosodic, semantic and pragmatic analyses in order to arrive at some meaning. Forming a coherent percept requires both serial and parallel integration across different aspects of language together with more general cognitive components. This so-called binding problem for language is a matter of intense debate. Little is known about how the human brain copes with these requirements of the language system and how it integrates the activity of different neuronal resources involved in the different aspects of sentence processing.

Over the past 20 years this neurophysiological binding problem has been theoretically addressed and empirically investigated within cognitive domains such as visual object perception (e.g., Singer, 2002; for review), focusing mainly on the ways in which the brain integrates signals, separated in space and time, to yield a unified sensory experience. According to the temporal correlation hypothesis, binding occurs in the temporal domain by virtue of neurons synchronizing their discharges (Singer, 2002). Synchronization phenomena, frequently found within small frequency ranges of the neuronal signals investigated, are being increasingly recognized as a key feature for establishing communication between different brain regions (Singer, 1999).

The neurophysiological binding problem within language–specifically sentence processing–has rarely been addressed even though determining mechanisms of neuronal integration in language is a key question within Cognitive Neuroscience. Only a handful of studies have investigated the role that brain oscillations in different frequency bands may play in sentence processing (Bastiaansen and Hagoort, 2003, Braeutigam et al., 2001, Roehm et al., 2001); these investigated power changes during sentence reading. Only a few EEG studies have investigated neuronal synchronization processes accompanying online sentence processing in the visual (Haarmann et al., 2002) and auditory domain (Mueller et al., 1997b, Weiss and Mueller, 2003, Weiss et al., 2001, Weiss et al., 2002). Almost all of the EEG studies aimed at providing information about the neurophysiological mechanisms accompanying sentence processing used event-related brain potentials (ERPs) (e.g., Brown and Hagoort, 1999, Kutas, 1997, for review). Even though the analysis of ERPs provides data with exquisite temporal resolution, it is of limited value for studying the processes involved in large-scale synchronization of different brain areas over several milliseconds or seconds. By contrast, fMRI data collected during sentence processing reveals activation in various brain areas such as left frontal and temporo-parietal as well as right hemispheric regions but with a temporal resolution that is not high enough to investigate certain syntactic processes that take place on the order of milliseconds (e.g., Just et al., 1996, Sakai et al., 2001, for review). In any case, these data do not by themselves reveal much, if anything, about the functional organization of information within regions or the cooperation across activated regions during the comprehension act (though see initial attempts assessing temporal correlation of specific prefrontal regions during sentence comprehension task, Homae et al., 2003).

To date, neuronal interaction has been inferred primarily from analyses of electrophysiological recordings. A number of different mathematical approaches exist for extracting information on frequency-based cooperation between neuronal structures during various cognitive tasks in healthy humans (e.g., Bressler and Kelso, 2001, Nikolaev et al., 2001, Varela et al., 2001, Schack and Weiss, 2003, Schack and Weiss, 2005). One well-known algorithm for assessing neuronal interaction or coupling during language processing is the computation of EEG coherence (e.g., Weiss and Mueller, 2003, for review). Coherence (C) at a frequency (w) for two signals x and y is derived from the smoothed cross-spectrum amplitude |Gxy(w)| and the two corresponding smoothed power spectra, Gxx(w) and Gyy(w), C2xy(w) = |Gxy(w)|2/Gxx(w)Gyy(w). The coherence function provides a measure of the linear synchronization between two signals as a function of frequency (Nunez et al., 1997, Petsche and Etlinger, 1998, Rappelsberger, 1998); it is very useful when synchronization is limited to some particular frequency bands, as it is typically the case in EEG signals.

In the current study we investigated the dynamic pattern of EEG coherence during sentence processing in order to gather information about frequency-related transient neuronal co-operation of brain oscillations correlated with syntactic analysis and verbal working memory processes. We investigated EEG coherence coincident with the processing of English subject–subject (SS)- and subject–object (SO)-relative sentences. In SS-relative sentences, such as The fireman who speedily rescued the cop sued the city over working conditions, the subject of the main clause (The fireman) is also the subject and agent of the relative clause. Such sentences have consistently been found to be easier to process than SO-relatives, in which the subject of the main clause is the object and patient of the relative clause (e.g., The fireman who the cop speedily rescued sued the city over working conditions). Children find it easier to comprehend subject relatives than object relatives (e.g., Tavakolian, 1981) as do young adults (e.g., King and Just, 1991) and patients with aphasia, probable Alzheimer disease or fronto-temporal dementia (e.g., Cooke et al., 2003, Grossman et al., 2003). This greater processing difficulty has been attributed by several researchers to the greater working memory (WM) demands of SO-sentences for which the main clause noun phrase (NP; The fireman) has to be maintained in memory over longer stretches of time until its role becomes clear and processing can resume. Another potentially difficult aspect of SO-relatives compared to SS-relatives is that the grammatical role played by the main clause NP in the former changes in the course of sentence processing.

In behavioral studies with sentence materials similar to those we used in our study, SO-sentences were associated not only with more comprehension errors but also slower word-by-word reading times at and just following the end of the relative clause, predominantly at the main clause verb (King and Just, 1991). Similar results have been observed in French (Holmes and O'Regan, 1981), German (e.g., Schriefers et al., 1995), Dutch (Mak et al., 2002) and Japanese (Miyamoto and Nakamura, 2003) but not in Chinese (Hsiao and Gibson, 2003). Across almost all the studies, reading time data show no evidence for a processing difference (e.g., greater WM load) until the end of the relative clause, just at that point when the load may begin to decrease.

In contrast, ERP measurements reveal a neural processing difference between subject and object relatives much earlier in the sentence, specifically shortly after the reader encounters the relative clause. King and Kutas (1995), for example, recorded a greater left fronto-central negativity for written SO-sentences as compared to SS-sentences continuing through the processing of the main clause (post-RC). ERPs indicated a difference between the two sentence types as soon as there was a WM load difference between them. Moreover, in accordance with the largest reaction time and reading time effects, these ERPs to visually presented SS- and SO-sentences showed reliable differences after the end of the relative clause (post-RC) at the main clause verb. SO-sentences elicited a larger left-anterior negativity (LAN) which was taken to index some aspect of working memory load at this point (King and Kutas, 1995). Comparable ERP differences for the reading of German SS- and SO-relatives were observed in sentence-length ERPs when the relative clauses had an unambiguous syntactic structure such as in the English materials (Muente et al., 1997).

In order to determine whether these effects were specific to sentence reading or were modality independent, Mueller et al. (1997a) examined sentence-length ERPs of participants listening to the King and Kutas sentences presented as natural speech. ERP effects were comparable to those with the written sentences, though generally more widespread and somewhat more pronounced over right hemispheric leads. Increased right hemispheric involvement during the processing of the more complex SO-sentences was demonstrated via fMRI measurements (Just et al., 1996).

In the current study our aim was to examine differences in EEG coherence between center-embedded SS- and SO-relatives using a specific spectral analysis technique that affords coherence estimates continuously across time, for a variety of frequency bins (Schack et al., 1995a, Schack et al., 1995b). Previously, we examined the EEG coherence in this sentence material with a different approach by the means of Fourier Transform which yielded initial indications (1) that coherence at left frontal sites changes considerably during sentence processing and (2) that specific frequency bands might play different roles in the information transfer during sentence processing (Weiss et al., 2001, Weiss et al., 2002, Weiss and Mueller, 2003). However, the temporal resolution of that analysis method was relatively low (1 s) and only activity in lower frequency bands was analyzed (< 18 Hz). In the present study, therefore, we chose to use an adaptive fitting procedure for bivariate autoregressive-moving-average (ARMA) models with time-varying parameters, which allows continuous calculation of coherence with a frequency resolution that accurately tracks the dynamic changes of the EEG signal during sentence processing.

Our second aim was to describe the pattern of oscillations within distinct frequency ranges possibly engaged in processing different subcomponents of language processing within the brain. Recent studies in humans strongly suggest that theta varies with episodic or working memory processes (e.g., Klimesch, 1999, Sarnthein et al., 1998, Weiss et al., 2000, Weiss and Rappelsberger, 2000) and theta power tends to change during the course of visual sentence processing (Weiss et al., 2002, Roehm et al., 2001). Thus, we hypothesized that theta band coherence would differ for SO- and SS-relatives as soon as there is a working memory load—i.e., from the beginning of the relative clause and last until the sentence end. There are only a handful of studies that point to the possible roles that activity in other frequency ranges may play in language or language-related processing. Gamma oscillations were affected by episodic verbal memory (e.g., Fell et al., 2003b, Schack and Weiss, 2003, Schack and Weiss, 2005) and correlated with lexical processing (Pulvermueller et al., 1997), semantic integration in sentence processing (Braeutigam et al., 2001), selective attention (Fell et al., 2003a, Fell et al., 2003b) and task complexity (Simos et al., 2002). The role of lower beta frequencies seems even more diverse, having been correlated with semantic word processing (Weiss and Rappelsberger, 1996), syntactic analysis during sentence comprehension (Mueller et al., 1997b), general sentence comprehension (Roehm et al., 2001), and semantic working memory demands (Haarmann et al., 2002). We thus hypothesized that gamma and/or beta frequencies reflect more than working memory load and therefore would show different coherence patterns across different parts of the sentence comparisons and in different brain regions.

Finally, we were interested in the possible differential participation of the left and right hemispheres and the involvement of frontal and parieto-temporal regions as indicated by previous ERP- and fMRI data (Just et al., 1996, King and Kutas, 1995, Mueller et al., 1997a). We thus examined whether signals at left- and right-hemispheric anterior and posterior electrode positions would show increased interaction during sentence processing, in particular the processing of the more complex SO-relatives.

Section snippets

Participants

Twenty-four university students (12 f, 12 m) participated in the experiment. All of the participants were right-handed according to the Edinburgh Handedness Inventory, monolingual English native speakers between 19 and 35 years (Ø = 23.3 ± 3.5). After applying strict criteria for rejecting trials with muscle artifact (as this can adversely affect spectral analysis, especially in the gamma band), 18 participants' data were available for the coherence analysis.

Stimuli and experimental procedure

Two hundred sixteen syntactically and

Results

Fisher-z-transformed coherence values were submitted to repeated measures ANOVA. Separate ANOVAs were conducted for three different sentence intervals (pre-RC, RC, post-RC), for each of the four frequency bands and for each of three different brain regions. The different brain regions (anterior, posterior, anterior–posterior) were selected since results of previous ERP studies (King and Kutas, 1995, Mueller et al., 1997a) suggested the participation of frontal and posterior temporal brain

Discussion

The main finding in the present study was that EEG coherence did reliably differ during the processing of SS-relatives and SO-relatives both across the relative clause (RC) and in the post-relative clause region (post-RC). Moreover, the particular frequency bands (theta, beta-1, gamma) within which these coherence differences were observed were a function of the sentence interval examined.

Acknowledgements

This study was supported by the Austrian Science Foundation (Herta Firnberg-project T127) and the German Science Foundation (SFB 360).

References (63)

  • F. Hsiao et al.

    Processing relative clauses in Chinese

    Cognition

    (2003)
  • M.J. Kahana et al.

    Theta returns

    Curr. Opin. Neurobiol.

    (2001)
  • J. King et al.

    Individual differences in syntactic processing. The role of working memory

    J. Mem. Lang.

    (1991)
  • W. Klimesch

    EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis

    Brain Res. Rev.

    (1999)
  • W.M. Mak et al.

    The influence of animacy on relative clause processing

    J. Mem. Lang.

    (2002)
  • W. Marslen-Wilson et al.

    The temporal structure of spoken language understanding

    Cognition

    (1980)
  • A.R. Nikolaev et al.

    Correlation of brain rhythms between frontal and left temporal (Wernickes's) cortical areas during verbal thinking

    Neurosci. Lett.

    (2001)
  • P.L. Nunez et al.

    EEG coherency I: statistics, reference electrode, volume conduction, Laplacians, cortical imaging, and interpretation at multiple scales

    Electr. Clin. Neurophysiol.

    (1997)
  • K.L. Sakai et al.

    Sentence processing in the cerebral cortex

    Neurosci. Res.

    (2001)
  • B. Schack et al.

    The sensitivity of instantaneous coherence for considering elementary comparison processing: Part I. The relationship between mental activities and instantaneous EEG coherences

    Int. J. Psychophysiol.

    (1999)
  • B. Schack et al.

    Phase-coupling of theta–gamma EEG rhythms during short-term memory processing

    Int. J. Psychophysiol.

    (2002)
  • H. Schriefers et al.

    The processing of locally ambiguous relative clauses in German

    J. Mem. Lang.

    (1995)
  • W. Singer

    Cognition, gamma oscillations and neuronal synchrony

  • R.W. Thatcher et al.

    Cortico-cortical associations and EEG coherence: a two-compartmental model

    Electr. Clin. Neurophysiol.

    (1986)
  • S. Weiss et al.

    The contribution of EEG coherence to the investigation of language

    Brain Lang.

    (2003)
  • S. Weiss et al.

    EEG coherence within the 13–18 Hz band as a correlate of a distinct lexical organisation of concrete and abstract nouns in humans

    Neurosci. Lett.

    (1996)
  • S. Weiss et al.

    Long-range EEG synchronization during word encoding correlates with successful memory performance

    Cogn. Brain Res.

    (2000)
  • C.M. Brown et al.

    The Neurocognition of Language

    (1999)
  • D. Caplan et al.

    Verbal working memory and sentence comprehension

    BBS

    (1999)
  • M. Essl et al.

    EEG coherence and reference signals: experimental results and mathematical explanations

    Med. Biol. Eng. Comput.

    (1998)
  • J. Fell et al.

    Human memory formation is accompanied by rhinal-hippocampal coupling and decoupling

    Nat. Neurosci.

    (2001)
  • Cited by (93)

    • Cross-linguistic differences in case marking shape neural power dynamics and gaze behavior during sentence planning

      2022, Brain and Language
      Citation Excerpt :

      In the current study, the planning of sentences with ergative-marked agents in Basque led to a pattern of agent fixations characterized by a shifted peak compared to nominative sentences in German (Fig. 4), to event-related synchronization of power in the theta and alpha frequency bands (Figs. 5 and 6), and to shorter speech onset latencies, compared to nominative sentences (Table 1). Increases of neural activity in the theta band are indices of the processing of syntactic and semantic dependencies (Bastiaansen, van Berkum, & Hagoort, 2002b; Bastiaansen, van Berkum, & Hagoort, 2002a; Hald, Bastiaansen, & Hagoort, 2006; Weiss et al., 2005, i.a.) as well as working memory engagement (Karrasch et al., 2004; Krause et al., 2000; Pavlov & Kotchoubey, 2020; Riddle et al., 2020; Sauseng et al., 2004). Crucially, theta band synchronization supports the “structured retrieval of choice-relevant information around decision points” and is thus a neural mechanism that coordinates the integration of multiple types of information across brain networks (Womelsdorf, Vinck, Leung, & Everling, 2010, p. 10).

    • The Bilingual Lexicon, Back and Forth: Electrophysiological Signatures of Translation Asymmetry

      2022, Neuroscience
      Citation Excerpt :

      While oscillatory measures register transient (de)couplings between cortical cell assemblies (Buzsáki and Draguhn, 2004; Buzsaki, 2006), functional connectivity metrics capture statistical co-dependencies indexing segregated neural processing (Rubinov and Sporns, 2010; Friston, 2011). In particular, modulations of both measures in the theta, alpha, and beta bands seem sensitive to fine-grained effects during L1 and L2 processing (Weiss et al., 2005; Bastiaansen et al., 2010; Bialystok et al., 2012; Kielar et al., 2014; Pérez et al., 2015; Elmer and Kühnis, 2016; Vilas et al., 2019; Birba et al., 2020), as seen even in studies that examined FT and BT separately (Grabner et al., 2007; Dottori et al., 2020). Indeed, modulations in some such bands (viz., theta) correlate positively with word translation speed, suggesting a putative role during the task (Dottori et al., 2020).

    View all citing articles on Scopus
    View full text