Elsevier

NeuroImage

Volume 114, 1 July 2015, Pages 294-302
NeuroImage

The language skeleton after dissecting meaning: A functional segregation within Broca's Area

https://doi.org/10.1016/j.neuroimage.2015.04.011Get rights and content

Highlights

  • IFG represents language structure in posterior areas and meaning in anterior areas

  • BA 44 sensitive to pure syntactic information, rather than to lexical information

  • BA 45 co-activates in syntactic contrast only if semantic information present

  • Meaningful word units engage temporal cortex and structural engage the syntax network

  • Evidence for segregated neural representations for language structure and meaning

Abstract

Broca's area is proposed as a crucial brain area for linguistic computations. Language processing goes beyond word-level processing, also implying the integration of meaningful information (semantics) with the underlying structural skeleton (syntax). There is an on-going debate about the specialisation of the subregions of Broca's area—Brodmann areas (BA) 44 and 45—regarding the latter aspects. Here, we tested if syntactic information is specifically processed in BA 44, whereas BA 45 is mainly recruited for semantic processing. We contrasted conditions with sentence structure against conditions with random order in two fMRI experiments. Besides, in order to disentangle these processes, we systematically removed the amount of semantic information available in the stimuli. This was achieved in Experiment 1 by replacing meaningful words (content words) by pseudowords. Within real word conditions we found broad activation in the left hemisphere, including the inferior frontal gyrus (BA 44/45/47), the anterior temporal lobe and posterior superior temporal gyrus (pSTG) and sulcus (pSTS). For pseudowords we found a similar activation pattern, still involving BA 45. Among the pseudowords in Experiment 1, we kept those word elements that convey meaning like un- in unhappy or -hood in brotherhood (i.e. derivational morphology). In Experiment 2 we tested whether the activation in BA 45 was due to their presence. We therefore further removed derivational morphology, only leaving word elements that determine syntactic structure (i.e. inflectional morphology, e.g. the verb ending -s in he paints). Now, in the absence of all semantic cues, including derivational morphology, only BA 44 was active. Additional analyses showed a selective responsiveness of this area to syntax-relevant cues. These findings confirm BA 44 as a core area for the processing of pure syntactic information. This furthermore suggests that the brain represents structural and meaningful aspects of language separately.

Introduction

Language is more than words. It is also the combination of content words (categories that convey meaning, e.g. nouns and verbs) and their derivational forms (e.g. brotherhood derived from brother), which are particularly relevant for sentence meaning, i.e. carry semantic information. No less important are their inflectional forms and function words, which are the elements that convey the relationships between them and hence the global sentence structure, i.e. carry morphosyntactic information. Thus, during language processing, the skeleton of syntactic information interacts with semantics. Early fMRI studies (e.g. Dapretto and Bookheimer, 1999, Sakai et al., 2002) showed a dissociation of these two dimensions, but no consensus has been reached so far as to which brain areas deal with the processing of the different types of linguistic information. This holds in particular for the involvement of the inferior frontal gyrus (IFG) (e.g. Rogalsky and Hickok, 2011).

Broca's area in the IFG, which consists of the posterior pars opercularis (BA 44) and the more anteriorly located pars triangularis (BA 45) (Amunts et al., 1999), has long been identified as one of the classical language areas. However, the specific function of Broca's area and, in particular, of its subparts remains to be specified. Review articles, meta-analyses (Hagoort and Indefrey, 2014) and models of language processing (Friederici, 2011) suggest that BA 44 is more likely to be involved in the processing of syntax, whereas BA 45 seems to be more highly engaged by the processing of semantics, especially at the sentence level (Hagoort, 2005, Vigneau et al., 2006, Price, 2010, Friederici, 2011). However, results across different languages and paradigms do not allow for a clear specification of the function of Broca's area's subregions.

The relevance of the distinction of these two neighbouring brain areas, BA 44 and BA 45, also lies in the striking difference in their anatomical connectivity with the temporal cortex, which is crucial for sentence comprehension. Whereas BA 44 and the posterior temporal cortex are connected by dorsal fibre tracts via the superior longitudinal fasciculus (SLF) and the arcuate fasciculus (AF), BA 45 and the superior and middle temporal gyri (STG/MTG) are connected by ventral fibre tracts running via the extreme capsule fibre system (ECFS) (for an overview see Friederici and Gierhan, 2013). Indeed, the tractography-based neuroanatomical parcellation of Broca's area consistently leads to a subdivision into an anterior and a posterior portion corresponding to BA 45 and 44 respectively (Anwander et al., 2007). Furthermore, while the dorsal pathway has been associated with the processing of complex syntactic structures (Friederici et al., 2006a, Friederici et al., 2006b, Brauer et al., 2011, Wilson et al., 2011), the ventral pathway seems to be consistently involved in semantic comprehension (Saur et al., 2008, Griffiths et al., 2013, Almairac et al., 2014). These findings on brain connectivity are consistent with the aforementioned functional segregation within Broca's area.

However, not all functional brain imaging studies that focused on syntactic processing report major activation only in BA 44 (as it is the case of Friederici et al., 2006b, Newman et al., 2010, Kinno et al., 2008), but report additional activation in BA 45 (Tyler and Marslen-Wilson, 2008, Tyler et al., 2010, Santi and Grodzinsky, 2007, Santi and Grodzinsky, 2010, Fedorenko et al., 2011, Pallier et al., 2011). The reason for this difference may be that these studies differed in the stimulus material or in the specific task demands, but it may also be due to the experimental languages they used. In fact, German studies show a clear activation of BA 44 for various syntactic manipulations (Friederici et al., 2006b, Makuuchi et al., 2009, Obleser et al., 2011), whereas English studies frequently showed the activation of BA 44 and additionally BA 45 (Caplan et al., 2008, Santi and Grodzinsky, 2010, Tyler et al., 2010, but see Newman et al., 2010).

One possible account for this distinction is that German and English strongly differ in the linguistic features that are crucial for sentence processing, and in particular the processing of sentence structure. In German, word order is relatively free, with sentence building blocks (called constituents) being able to occupy different positions in the sentence and the verb being the only fixed element. The flexibility in word order is counterbalanced by rich morphosyntactic information—word elements that convey the relations established between constituents (inflectional morphology), e.g. express “who did what to whom” in form of case marking (Haider, 2010). In turn, English speakers mainly rely upon word order to identify the syntactic roles of constituents, since morphosyntactic information is comparatively scarce (Thompson, 1978). We therefore postulate that whereas the activation of BA 44 happens in response to pure syntactic information (in particular, morphosyntactic inflections and function words), the activation of BA 45 is present mainly as a result of parallel sentence-level semantic processing based on the meaning of words and their derivational morphology (like un- in unhappy or -hood in brotherhood). EEG studies could indeed show that semantic processing (in the form of N400) could be blocked by syntactic violations revealed by an ELAN (Friederici et al., 1999, Hahne and Jescheniak, 2001). These findings would explain why we would only find parallel semantic processing in stimuli with sentence structure. Furthermore, these additional semantic computations should be most salient in the absence of rich morphosyntactic information as it is the case of English when compared to German.

In order to test whether the differential activation in BA 44 and BA 45 reported in the literature is a function of the language type (English or German) and/or the function of the respective information type available in the stimulus material, we designed a close replication of the English study of Tyler et al. (2010) using German as the experimental language. In the English study—an fMRI adaptation from a well-established behavioural word monitoring paradigm (Marslen-Wilson and Tyler, 1980)—subjects were presented with three conditions: first, a normal prose (NP) condition; second, a nonsense sentence condition—anomalous prose (AP)—in which all content words of a normal sentence were replaced by semantically implausible words of the same lexical category; third, a condition consisting of the same word material of AP sentences in a permutated order that disrupted syntactic structure—random word order (RWO). The authors of this fMRI study reported a widespread activation for syntax based on a contrast between AP and RWO. This activation pattern, however, could result from the additional recruitment of semantic processes, since semantic information was widely available in these conditions. In the present study we therefore progressively removed the semantic information of sentences in order to find the central hub for pure syntactic processing in language. Our study hence included additional conditions with pseudowords, words that respect the phonological rules of a language but are devoid of word meaning. The conditions of this study not only comprised those tested by Tyler et al. (2010), but also their pseudoword counterparts respectively called Jabberwocky Prose (JP) and Jabberwocky in Random Order (JRO). Two fMRI experiments were conducted in order to disentangle different aspects of semantics by systematically removing meaning from the stimuli, first in the form of content words and then of derivational morphemes.

Section snippets

Participants

Twenty-three participants (11 females, 26.3 ± 3.3 years of age, M ± SD) took part in fMRI Experiment 1. 32 participants (16 females, 27.3 ± 3.2 years of age, M ± SD) were recruited for fMRI Experiment 2. All participants in both experiments were right-handed, as assessed by an abridged version of the Edinburgh Inventory (Oldfield, 1971), and were native speakers of German. They had normal or corrected to normal vision, and did not report any history of neurological, psychiatric, or hearing disorder. All

Behavioural data

In Experiment 1, the mean accuracy rate was 97.9% (SD = 4.2%) and in Experiment 2, it was 95.1% (SD = 6.3%). In Experiment 1, only one participant performed under mean-2SD and was excluded from further analysis. In Experiment 2, a total of 5 participants with accuracy rate under the same threshold of mean-2SD were excluded. In spite of the very high accuracy rates in the monitoring task, it was possible to observe significant differences in reaction times and accuracies across stimulus conditions.

Discussion

The two experiments show that Broca's area (BA 44 and BA 45, extending anteriorly to BA 47) is activated as a whole during sentence processing as long as syntactic and semantic information is available. Once all semantic information is deleted and only pure syntactic information is left—provided by function words, inflectional morphology and word order—only BA 44 is activated. This finding provides evidence for a functional segregation of Broca's area into BA 44 and BA 45 with a functional

Acknowledgements

This worked was funded by a grant from Fundação Para a Ciência e a Tecnologia (FCT) to TG (SFRH/BD/77908/2011).

The authors would like to thank Emiliano Zaccarella, Lisa Knoll and Corinna Bonhage in the preparation of the experiment and Alfred Anwander for anatomical expertise. We would also like to thank Kerstin Flake for her help with the picture design.

Conflict of interest

The authors declare that there are no conflicts of interest.

References (63)

  • D.R. Gitelman et al.

    Modeling regional and psychophysiologic interactions in fMRI: the importance of hemodynamic deconvolution

    NeuroImage

    (2003)
  • P. Hagoort

    On Broca, brain, and binding: a new framework

    Trends Cogn. Sci.

    (2005)
  • A. Hahne et al.

    What's left if the Jabberwock gets the semantics? An ERP investigation into semantic and syntactic processes during auditory sentence comprehension

    Cogn. Brain Res.

    (2001)
  • S. Heim et al.

    Distributed cortical networks for syntax processing: Broca's area as the common denominator

    Brain Lang.

    (2003)
  • S. Heim et al.

    The role of the left Brodmann's areas 44 and 45 in reading words and pseudowords

    Cogn. Brain Res.

    (2005)
  • W. Marslen-Wilson et al.

    The temporal structure of spoken language understanding

    Cognition

    (1980)
  • S.D. Newman et al.

    The effect of semantic relatedness on syntactic analysis: an fMRI study

    Brain Lang.

    (2010)
  • J. Obleser et al.

    Dynamic assignment of neural resources in auditory comprehension of complex sentences

    NeuroImage

    (2011)
  • R.C. Oldfield

    The assessment and analysis of handedness: the Edinburgh inventory

    Neuropsychologia

    (1971)
  • K.L. Sakai et al.

    Selective priming of syntactic processing by event-related transcranial magnetic stimulation of Broca's area

    Neuron

    (2002)
  • A. Santi et al.

    Working memory and syntax interact in Broca's area

    NeuroImage

    (2007)
  • A. Santi et al.

    fMRI adaptation dissociates syntactic complexity dimensions

    NeuroImage

    (2010)
  • M. Vigneau et al.

    Meta-analyzing left hemisphere language areas: phonology, semantics, and sentence processing

    NeuroImage

    (2006)
  • S.M. Wilson et al.

    Syntactic processing depends on dorsal language tracts

    Neuron

    (2011)
  • F. Almairac et al.

    The left inferior fronto-occipital fasciculus subserves language semantics: a multilevel lesion study

    Brain Struct. Funct.

    (2014)
  • K. Amunts et al.

    Broca's region revisited: cytoarchitecture and intersubject variability

    J. Comp. Neurol.

    (1999)
  • A. Anwander et al.

    Connectivity-based parcellation of Broca's area

    Cereb. Cortex

    (2007)
  • H. Baayen et al.

    The CELEX database on CD-ROM

    Linguistic Data Consortium

    (1993)
  • P. Boersma et al.

    Praat 4.0: A System for Doing Phonetics by Computer (Computer Software)

    (2001)
  • M. Bozic et al.

    Neurocognitive contexts for morphological complexity: dissociating inflection and derivation

    Lang. Linguist. Compass

    (2010)
  • J. Brauer et al.

    Neuroanatomical prerequisites for language functions in the maturing brain

    Cereb. Cortex

    (2011)
  • Cited by (113)

    • The role of left temporo-parietal and inferior frontal cortex in comprehending syntactically complex sentences: A brain stimulation study

      2023, Neuropsychologia
      Citation Excerpt :

      Effects of tDCS for left IFG stimulation have been reported at all linguistic levels, including artificial grammar learning (de Vries et al., 2010), lexico-semantic control (Ihara et al., 2015) phonetic/semantic fluency (Cattaneo et al., 2011; Iyer et al., 2005), sentence comprehension (Lum et al., 2019), but also verbal working memory (Fregni et al., 2005). Moreover, general improvement of sentence comprehension for a-tDCS over left IFG was reported, while the expected specific effect for complex sentences was not seen (Giustolisi et al., 2018). Likewise, when applied over posterior areas (TP), a-tDCS augmented a broad array of functions, comprising semantic composition (Price et al., 2016), novel word learning (Floel et al., 2008; Meinzer et al., 2014), phonological facilitation (Pisoni et al., 2017) and picture naming (Sparing et al., 2008).

    View all citing articles on Scopus
    View full text