Elsevier

NeuroImage

Volume 59, Issue 1, 2 January 2012, Pages 815-823
NeuroImage

Network modulation during complex syntactic processing

https://doi.org/10.1016/j.neuroimage.2011.07.057Get rights and content

Abstract

Complex sentence processing is supported by a left-lateralized neural network including inferior frontal cortex and posterior superior temporal cortex. This study investigates the pattern of connectivity and information flow within this network. We used fMRI BOLD data derived from 12 healthy participants reported in an earlier study (Thompson, C. K., Den Ouden, D. B., Bonakdarpour, B., Garibaldi, K., & Parrish, T. B. (2010b). Neural plasticity and treatment-induced recovery of sentence processing in agrammatism. Neuropsychologia, 48(11), 3211–3227) to identify activation peaks associated with object-cleft over syntactically less complex subject-cleft processing. Directed Partial Correlation Analysis was conducted on time series extracted from participant-specific activation peaks and showed evidence of functional connectivity between four regions, linearly between premotor cortex, inferior frontal gyrus, posterior superior temporal sulcus and anterior middle temporal gyrus. This pattern served as the basis for Dynamic Causal Modeling of networks with a driving input to posterior superior temporal cortex, which likely supports thematic role assignment, and networks with a driving input to inferior frontal cortex, a core region associated with syntactic computation. The optimal model was determined through both frequentist and Bayesian Model Selection and turned out to reflect a network with a primary drive from inferior frontal cortex and modulation of the connection between inferior frontal cortex and posterior superior temporal cortex by complex sentence processing. The winning model also showed a substantive role for a feedback mechanism from posterior superior temporal cortex back to inferior frontal cortex. We suggest that complex syntactic processing is driven by word-order analysis, supported by inferior frontal cortex, in an interactive relation with posterior superior temporal cortex, which supports verb argument structure processing.

Highlights

► Complex syntactic processing activates a left-hemisphere cortical network. ► Inferior frontal cortex drives the input to the syntactic processing network. ► Frontal and posterior temporal cortices interact during syntactic processing.

Introduction

Processing of complex syntactic structures demands more cognitive resources than the processing of relatively simple constructions, and this is associated with locally increased neuronal activation (Just et al., 1996, Stromswold et al., 1996, Caplan et al., 1998). Although reported patterns of activation foci modulated by experimental factors vary between studies, it appears that an important role in sentence processing is played by left-hemisphere inferior frontal cortex, in particular Broca's area (for a comprehensive and critical overview, see Rogalsky & Hickok(2011)). Debate on the precise functional role of (different parts of) this area is ongoing and hypotheses range from relying on specific structure-building or linearization operations (Bornkessel-Schlesewsky et al., 2009, Grodzinsky and Friederici, 2006) to those relying on more general cognitive processes of representational conflict resolution (Novick et al., 2005) or the integration/unification of different types of information into the sentence context (Hagoort, 2005). It has also been claimed that Broca's area supports a working memory component that may underlie any of the above-named processes (Fiebach et al., 2005, Kaan and Swaab, 2002), though this appears to be specifically plausible for the pars opercularis (Rogalsky & Hickok, 2011). Whatever its precise functional role(s), inferior frontal cortex does not operate in isolation, but is part of a larger network involved in sentence processing (Keller et al., 2001). The form of this network, as well as its modulation through syntactic complexity, is still under investigation (see Friederici (2009)).

In addition to inferior frontal cortex, another important role in syntactic processing is played by left-hemisphere posterior superior temporal cortex, where activation has also been shown to increase with syntactic complexity in sentence processing, from early functional imaging studies onwards (e.g. Just et al., 1996, Ben-Shachar et al., 2003). It is quite possible that the specific contribution of the posterior superior temporal cortex to syntactic parsing is in thematic role assignment, based on verb argument structure, the extraction of ‘actorhood’, and/or order preferences with respect to the animacy of potential arguments (Bornkessel et al., 2005, Grewe et al., 2007, Shetreet et al., 2007). This is in line with effects of verb argument structure complexity observed in this area (Den Ouden et al., 2009, Thompson et al., 2007, Thompson et al., 2010a). It has been suggested that Broca's area and posterior superior temporal gyrus together form a network that is responsible for thematic role assignment, a crucial aspect of complex sentence processing which, in English, relies on both the processing of word order and verb argument structure (Friederici, 2009, Friederici et al., 2006).

In an fMRI study examining the neural correlates of syntactic processing and recovery from aphasia, Thompson et al. (2010b) showed a pattern of left-hemisphere activation associated with processing of complex syntactic structures, viz. object-cleft constructions (OC; 1a) compared to subject-cleft constructions (SC; 1b). Using an auditory verification task, in which auditory sentences and visual scenes were presented simultaneously, participants indicated by button-press (yes/no) whether or not the two matched. Sentence types included OC, SC, and simple actives (ACT) (60 trials per condition), pseudorandomly distributed over 4 runs (see Thompson et al. (2010b) for details).

  • 1.

    • a.

      It was the groom that the bride carried. (OC)

    • b.

      It was the bride that carried the groom. (SC)

Unlike subject clefts, object-cleft constructions have a noncanonical word order (in English, an order other than subject–verb–object), and are deemed to be more complex based on formal syntactic theory (e.g. Chomsky, 1977, Chomsky, 1995, Bresnan, 2001), as well as on more general cognitive theories in which object clefts make greater demands on working memory (e.g. King and Just, 1991, Caplan and Waters, 1999, Gibson, 1998, Gordon et al., 2002). Whereas the contrast subtracting object-cleft activation levels from subject-cleft activation only revealed a small cluster of voxels in the left posterior insula, the opposite contrast revealed robust differential activation in a number of perisylvian left-hemisphere areas, including the inferior frontal, middle frontal and precentral gyri, the anterior insula, as well as the middle temporal, posterior superior temporal and angular gyri.

The areas identified by Thompson et al. (2010b) to be involved in processing complex sentences were similar to those reported in other studies (Just et al., 1996, Stromswold et al., 1996, Caplan et al., 1998, Caplan et al., 2001, Caplan, 2001, Cooke et al., 2002, Ben-Shachar et al., 2003). However, the BOLD signal subtraction analyses performed in this and similar studies do not provide insights about connectivity and information flow between these areas. Functional and effective connectivity analyses are required to map the network structure between activated areas, that is, to ascertain which network nodes interact during complex syntactic processing. One particular purpose of the current investigation was to determine which of two cortical areas is a better candidate to provide the driving input to the ‘syntactic network’: (i) posterior superior temporal cortex, likely involved in verb argument structure processing, or (ii) inferior frontal cortex, with its suggested prime role in supporting sequential processing, complex structure building and decomposition, either directly or indirectly through a working memory component. If activation throughout the network turns out to be principally driven by posterior superior temporal cortex, this corroborates the view that sentence processing occurs bottom-up, starting with the lexico-syntactic analysis of prime components, viz., verbs. On the other hand, if the network is driven primarily by inferior frontal activation, this suggests that sentence processing starts from the analysis of the linear order of its lexical components into a hierarchical structure. Ultimately, these processes have to team up, in order to achieve a correct parse for complex sentences.

In this paper we reanalyzed the raw data from Thompson et al. (2010b) by performing a two-stage connectivity analysis: we first used directed partial correlation (dPC) analysis as a hypothesis-free method to limit the model space and we then applied Dynamic Causal Modeling (DCM) to look at driving inputs and modulatory influences on the connections within the preselected models. dPC is a method that in principle allows for detecting effective connectivity, as discussed in Mader et al. (2008). This method has been used successfully by Saur et al. (2010), to investigate the networks underlying different aspects of auditory comprehension. Due to the comparably low temporal resolution of fMRI data the information about the connectivity structure has to be assumed to be contained in the instantaneous interactions. dPC does not depend on prior knowledge about the underlying network structure. It can be applied without assumptions about the network topology under investigation. The statistics that come with dPC analyses “decide” about the presence or absence of interactions, which can be used for the formulation of hypotheses about the network structure, as in the present study.

As a second step in our reanalysis of the Thompson et al. (2010b) data, we used Dynamic Causal Modeling (DCM; Friston et al., 2003) to further specify the preselected models. Through inference from local activation levels, DCM provides parameter estimates that reflect the effective strength and context-dependent modulation of connections between clusters of neurons (Stephan et al., 2010). The method has been used to investigate effective connectivity in areas such as task-related modulations of the network supporting speech comprehension (Leff et al., 2008), developmental changes within the phonological processing network (Booth et al., 2008), modulations of inferior frontal gyrus connectivity associated with lexical and phonological processing (Heim et al., 2009a) and altered connectivity in patients with primary progressive aphasia (Sonty et al., 2007).

One major concern in DCM is the a priori selection of models to be tested. The primary challenge is to reduce the number of relevant models that will be compared, based on theoretical, practical or other data-external considerations, beforehand. Without such a reduction, the number of possibilities is essentially unlimited, due to boundless combinations of different driving inputs, self-modulating nodes and multiple modulations on different connections. For this reason, we raised the cluster size threshold in the subtraction reanalysis of the Thompson et al. (2010b) data, in order to select only the most strongly activated peaks in the potential network, and we let our competing models be constrained by the outcome of the dPC functional connectivity analysis. Further restrictions on the model shape are discussed in Material and methods.

Through serial application of these three methods of data analysis (BOLD subtraction, dPC and DCM), we investigated (i) which of two competing hypotheses about the driving input to the network provided a better fit to the data, viz., models with driving input from posterior superior temporal cortex or from inferior frontal cortex, and (ii) which of the directional connections in the syntactic network is crucially modulated by the processing of complex syntactic structures.

Section snippets

Time series

The background to the fMRI experiment, as well as the participant, task and imaging information, have been published in detail in Thompson et al. (2010b). For further background, we refer the reader to that publication. The data of twelve right-handed volunteers ranging in age from 32 to 79 years (7 females, mean age 54), presented in Thompson et al. (2010b), were used to identify participant-specific activation peaks within a sphere of a 9 mm radius of the group activation peaks (based on the

fMRI

The factorial re-analysis of Thompson et al.'s (2010b) fMRI data, with an elevated cluster size threshold (k = 15), revealed a significant main effect of sentence type, with no main effect of sentence–picture matching and no interaction of sentence type and sentence–picture matching. In further analyses, therefore, matched and mismatched stimulus trials were collapsed. The effect of sentence type was solely driven by the contrast of OC > SC, yielding four clusters of differential activation.

Discussion

As reported by Thompson et al. (2010b), conventional fMRI analysis contrasting object-cleft and subject-cleft sentence processing revealed a left-lateralized group of perisylvian regions that showed increased activation associated with complex syntactic processing. Refined analyses of these data revealed four regions of significant activation located in the inferior frontal gyrus (IFG), premotor cortex (PM), posterior superior temporal sulcus (pSTS) and anterior middle temporal gyrus (aMTG),

Conclusions

Successive application of three methods of neuroimaging data analysis was used to investigate the infrastructure of a neural network supporting complex syntactic processing, as well as the information flow within this network. A conventional BOLD fMRI subtraction paradigm investigating increased activation associated with the processing of object-cleft sentences relative to subject-cleft sentences revealed four major left-hemisphere peaks of activation, viz. in the inferior frontal gyrus, the

Acknowledgments

This study was supported by NIH grant # R01 DC007213-03 (C.K. Thompson) and German Academic Exchange Service (DAAD) grant # D/09/42786 (D. Saur). The authors report no conflicts of interest.

References (57)

  • T. Grewe et al.

    The role of the posterior superior temporal sulcus in the processing of unmarked transitivity

    NeuroImage

    (2007)
  • Y. Grodzinsky et al.

    Neuroimaging of syntax and syntactic processing

    Current Opinion in Neurobiology

    (2006)
  • P. Hagoort

    On Broca, brain, and binding: a new framework

    Trends in Cognitive Sciences

    (2005)
  • E. Kaan et al.

    The brain circuitry of syntactic comprehension

    Trends in Cognitive Sciences

    (2002)
  • A. Kim et al.

    The independence of combinatory semantic processing: evidence from event-related potentials

    Journal of Memory and Language

    (2005)
  • J. King et al.

    Individual differences in syntactic processing: the role of working memory

    Journal of Memory and Language

    (1991)
  • D. Saur et al.

    Combining functional and anatomical connectivity reveals brain networks for auditory language comprehension

    NeuroImage

    (2010)
  • T.M. Snijders et al.

    Effective connectivity of cortical and subcortical regions during unification of sentence structure

    NeuroImage

    (2010)
  • K.E. Stephan et al.

    Bayesian model selection for group studies

    NeuroImage

    (2009)
  • K.E. Stephan et al.

    Ten simple rules for dynamic causal modeling

    NeuroImage

    (2010)
  • K. Stromswold et al.

    Localization of syntactic comprehension by positron emission tomography

    Brain and Language

    (1996)
  • M.K. Tanenhaus et al.

    Sentence comprehension

  • C.K. Thompson et al.

    Neural plasticity and treatment-induced recovery of sentence processing in agrammatism

    Neuropsychologia

    (2010)
  • Y. Benjamini et al.

    Controlling the false discovery rate: a practical and powerful approach to multiple testing

    Journal of the Royal Statistical Society, Series B

    (1995)
  • M. Ben-Shachar et al.

    The neural reality of syntactic transformations: evidence from functional magnetic resonance imaging

    Psychological Science

    (2003)
  • J. Bresnan

    Lexical–Functional Syntax

    (2001)
  • D. Caplan

    Functional neuroimaging studies of syntactic processing

    Journal of Psycholinguistic Research

    (2001)
  • D. Caplan et al.

    Verbal working memory and sentence comprehension

    The Behavioral and Brain Sciences

    (1999)
  • Cited by (82)

    • Neuromodulation of verb-transitivity judgments

      2022, Journal of Neurolinguistics
    View all citing articles on Scopus
    View full text