Elsevier

NeuroImage

Volume 45, Issue 3, 15 April 2009, Pages 1009-1019
NeuroImage

Neural correlates of semantic ambiguity processing during context verification

https://doi.org/10.1016/j.neuroimage.2008.12.044Get rights and content

Abstract

Understanding the relevant meaning of a word with different meanings (homonym) in a given context requires activation of the neural representations of the relevant meaning and inhibition of the irrelevant meaning. The cognitive demand of such disambiguation is highest when the dominant, yet contextually irrelevant meaning of a polar homonym must be suppressed. This central process (semantic inhibition) for lexico-semantic ambiguity resolution was monitored with fMRI during semantic context verifications. Twenty-two healthy volunteers decided whether congruent or incongruent target words fitted into the contexts established by preceding sentences. Half of the sentences ended with a homonym, thereby allowing to cross the factors ambiguity and semantic congruency. BOLD increases related to the inhibitory attentional control over non-selected meanings during ambiguity processing occurred in a brain network including left dorsolateral prefrontal cortex (DLPFC), bilateral angular gyrus (AG), bilateral anterior superior temporal gyrus (aSTG) as well as right ventromedial temporal lobe. In left DLPFC (BA 46/9) and left AG (BA 39) BOLD activity to target words of the incongruent-ambiguous condition correlated with the individual amount of semantic interference experienced by the subjects. BOLD increases of incongruent versus congruent semantic verifications occurred in bilateral inferior frontal gyrus. The results of the present study suggest a specific role of left DLPFC and AG in the resolution of semantic interference from contextually inappropriate homonym meanings. These fronto-parietal areas might exert inhibitory control over temporal regions in service of attentional selection between relevant and irrelevant homonym meanings, by creating a sufficient activation difference between their respective representations.

Introduction

Semantic ambiguity is pervasive in natural languages and constitutes a ubiquitous challenge for the human brain during meaning comprehension. Homonyms are special instantiations of semantic ambiguity with two or more distinct meanings. As such, they lend themselves particularly well to the controlled investigation of the neurocognitive processes involved in semantic ambiguity resolution. The successful disambiguation of homonymy requires at least two cognitive processes: semantic activation of the possibly relevant meaning and cognitive control over the contextually irrelevant meaning by semantic inhibition. So far, an enormous amount of mostly behavioural research has been devoted to giving detailed descriptions of the temporal dynamics and the various constituent factors of ambiguity resolution (e.g., meaning frequency, homonym polarity, contextual constraints) (for a review see Gorfein, 1989, Gorfein, 2001, Simpson, 1994). However, the neural correlates of both lexico-semantic ambiguity processing per se and the implementation of cognitive control over contextually irrelevant meanings have hitherto remained largely unexplored.

To date, knowledge of semantic ambiguity processing is rather gross anatomical. A major portion has come from divided visual field studies indicating differential contributions from the two cerebral hemispheres (Atchley et al., 1999, Chiarello, 1998, Collins, 2002, Coney and Evans, 2000, Faust and Chiarello, 1998, Faust and Kahana, 2002, Faust and Lavidor, 2003, Faust and Gernsbacher, 1996). These studies have suggested that a wider range of homonym meanings — including contextually irrelevant ones — is activated and maintained in the right hemisphere, while only the left hemisphere is able to select the relevant meaning and suppress the contextually inappropriate meanings.

In a similar vein, bilateral contributions to semantic ambiguity processing have also been concluded from studies on brain-lesioned patients. Individuals with complex language deficits following lesions of the left hemisphere were found to be impaired in selecting and integrating contextually appropriate homonym meanings as well as in suppressing inappropriate ones, while their ability to maintain alternative meaning associates active was largely spared (Copland et al., 2002, Grindrod and Baum, 2005, Hagoort, 1993, Katz, 1988, Metzler, 2001, Swaab et al., 1998, Swinney et al., 1989). A functional role of the right hemisphere in semantic ambiguity processing is less well supported by much more heterogeneous results in patients with focal damage to the right hemisphere (Grindrod and Baum, 2003, Grindrod and Baum, 2005): While some reports claim a functional significance of the right hemisphere for effective use of context, others emphasize contributions to the inhibitory control of inappropriate homonym meanings.

More specific localization of semantic ambiguity processing has only most recently come from fMRI studies that either used a covert word generation task (Chan et al., 2004), or speech comprehension during silent reading (Mason and Just, 2007, Zempleni et al., 2007) or listening (Rodd et al., 2005) to sentences that contained ambiguous words. Covered word generation in relation to semantically ambiguous as compared to unambiguous words was predominantly associated with increased brain activation in bilateral middle and superior frontal gyrus, anterior cingulate and cuneus, as well as in the inferior parietal lobe of the right hemisphere (Chan et al., 2004). Word generation to semantically unambiguous words, by contrast, was predominantly associated with increased activation of bilateral inferior frontal gyrus (IFG). Reading comprehension of ambiguous sentences where the ambiguous word appeared before the disambiguating context was accompanied by increased activation in bilateral IFG and inferior/middle temporal gyri (Zempleni et al., 2007) or left IFG, bilateral caudate and bilateral superior/middle frontal gyri (Mason and Just, 2007). A direct comparison of sentences using biased or balanced homonyms revealed additional activation of bilateral insula and left IFG. Rodd et al. (2005) observed ambiguity-related effects during spoken language comprehension in left posterior inferior temporal cortex and bilateral IFG and suggested an involvement of these regions in activating and selecting contextually appropriate word meanings while computing the meaning of spoken sentences.

The results of these studies show that the hitherto available evidence for ambiguity-related brain activation is highly heterogeneous and inconsistent. For instance, a central role of the IFG for semantic ambiguity processing is highlighted by the sentence comprehension studies (Mason and Just, 2007, Rodd et al., 2005, Zempleni et al., 2007), while activity in this brain region is not only absent in the ambiguity contrast of the word generation study (Chan et al., 2004), but predominantly observed for the inverse contrast (unambiguous > ambiguous). Moreover, the cognitive roles of the diverse network components reported in these studies remain rather vaguely determined. In particular, the central mechanism by which the brain exerts inhibitory control over contextually inappropriate homonym meanings in the course of ambiguity resolution is not explicitly addressed.

Current theory on semantic ambiguity processing underscores the importance of two interactive cognitive processes (Martin et al., 1999, Neill, 1989, Twilley and Dixon, 2000): semantic activation of relevant homonym meanings and semantic inhibition of competing irrelevant meanings. In the absence of contextual constraints, this account would postulate an initial rise of activity in all neural meaning representations related to a homonym, with the activation being a function of the relative frequency of that particular word's usage in language. This activity is further modulated by contextual constraints. If a polar homonym is strongly biased towards its dominant meaning, ambiguity resolution may proceed by pure surplus activation and without the need for additional inhibitory control over subordinate meanings. However, if a polar homonym is biased towards its subordinate meaning, inhibitory control over the competing dominant, yet irrelevant meaning associate becomes necessary to create a sufficiently large activation difference between both meanings in order to allow for an efficient selection of the relevant meaning.

Studies within the language domain have revealed and characterized a number of brain regions that might also play important roles in the disambiguation of lexico-semantic ambiguity. For instance, the cognitive process of semantic activation provides for initial access to distributed semantic representations (features and first-order associations) related to each input element. Human brain-lesion data and neuroimaging studies (Bookheimer, 2002, Demonet et al., 2005) have shown that semantic activation crucially depends on bilateral Wernicke's area — especially the posterior middle and superior temporal gyrus. Numerous imaging studies have also associated the angular gyrus (AG) with semantic processing, especially when the meaning of words is assessed (Binder et al., 2003, Giesbrecht et al., 2004, Gorno-Tempini et al., 1998, Price, 2000, Price and Friston, 1997), or when semantic memory is probed (Binder et al., 1997, Grossman et al., 2002, Martin et al., 1996, Vandenberghe et al., 1996, Warburton et al., 1996).

Another cognitive process important in the context of the present study is semantic integration. Semantic integration is considered to support meaning interpretation by detecting and elaborating the semantic relations between concepts or between concepts and message-level interpretations. However, this cognitive process does not only provide for integrative concepts by computing the semantic overlap among multiple semantic fields, but it can also operate on already selected concepts to construct complex representations of meaning relations. Although some reports suggest that semantic integration is served by those areas also involved in semantic activation, there is an accumulating amount of evidence suggesting that the anterior superior temporal gyrus (aSTG) is of central importance for this specific process (Bookheimer, 2002, Hodges and Graham, 2001, Hodges et al., 1992, Humphries et al., 2001, Maguire et al., 1999, Nobre et al., 1994).

Semantic selection — broadly conceived — is an interactive process by which a concept is selected for action or consciousness (Jung-Beeman, 2005). A large body of evidence suggests that semantic selection crucially depends on the IFG (Bookheimer, 2002, Miller and Cohen, 2001, Thompson-Schill et al., 1997, Wagner et al., 2001). Comparisons of various processing demands within the IFG have even led to the proposal of a functional subdivision with the anterior and ventral parts (roughly equivalent to BA 47 and the inferior aspect of BA 45) being sensitive to semantic retrieval and semantic decisions, and posterior and dorsal parts (BA 44/45) sensitive to manipulations of the demands on selection (e.g., Devlin et al., 2003, Thompson-Schill et al., 1999, Thompson-Schill et al., 1997).

Semantic inhibition refers to a direct cognitive control process that actively reduces semantic interference through inhibition of competing irrelevant meanings or concepts in the service of attentional selection. Within the language domain, many studies have frequently subsumed executive control processes such as the modulation of word-level semantic activation under semantic selection and consequently assigned this process to the ventral portion of the prefrontal cortex (see Jung-Beeman, 2005). However, other researchers have suggested that semantic inhibition is directly triggered by interference from competing items, and not tied to selection itself (Anderson et al., 2004, Knight et al., 1999, Logan, 1994). According to this conception, selection rather sets the stage for an inhibition process that directly targets competing meaning associates (for a review see Levy and Anderson, 2002). Hence, the dorsolateral prefrontal cortex (DLPFC) might be a candidate region for exerting such a form of inhibitory attentional control — presumably in interaction with more posterior attention-related brain regions in parietal cortex (e.g., Anderson et al., 2004, Badre and Wagner, 2004, Knight et al., 1999). To date, direct evidence for the neural basis of inhibitory control at the semantic level is only scarce. The vast majority of imaging studies have hitherto investigated inhibitory control in the context of motor/response inhibition or the inhibition of task sets, highlighting a central role of right hemisphere IFG (for a review see Aron et al., 2004, Chikazoe et al., 2007). So far, neuroimaging studies within the language domain (including studies on ambiguity resolution) have missed the chance to explicitly address the issue of semantic inhibition, making it difficult to adjudicate upon the differential processing contributions from dlPFC and IFG to this cognitive process.

Here we investigate the neural correlates of semantic ambiguity processing during semantic context verifications (Gernsbacher and Faust, 1991, Gernsbacher et al., 1990). In the context verification task, subjects were asked to judge whether congruent and incongruent target words semantically fitted the overall meaning of preceding sentential contexts, with half of the context frames containing a sentence-final homonym. The aims of the study were twofold: Firstly, we aimed at replicating and extending the evidence for semantic ambiguity processing reported by Chan et al. (2004) using an entirely different paradigm that poses a high demand on active lexico-semantic ambiguity resolution, and allows for a closer behavioural monitoring of subjects' performance. We hypothesized that the higher demand on active ambiguity resolution yields increased ambiguity-related activation in additional left-hemisphere network components including brain regions in temporal cortex (presumably involved in semantic integration) as well as fronto-parietal areas (DLPFC and AG; presumably involved in inhibitory control). Secondly, the current approach offers the additional advantage of identifying ambiguity network components specifically involved in the resolution of semantic interference from competing dominant, yet contextually inappropriate homonym meanings. The individual amount of semantic interference is represented by the behavioural costs of ambiguous target verifications (IA condition) compared to unambiguous target verifications (IU condition). Given the behavioural findings of Gernsbacher and Faust, 1991, Gernsbacher et al., 1990, semantic interference is higher in participants who have not yet sufficiently inhibited the competing irrelevant homonym meaning by the time the target word is presented. Accommodating these individual differences in semantic interference resolution for the interpretation of functional activation patterns is expected to yield a (positively correlated) linear brain–behaviour relationship in those components of the ambiguity network specifically involved in the inhibitory control over dominant, but contextually irrelevant homonym meanings.

Section snippets

Subjects

Twenty-two healthy, right-handed volunteers (12 female, 10 male; 20–52 years; mean 25.3 years) participated in the study. Exclusion criteria demanded the absence of any medical, neurological, or psychiatric illness, as well as MR contraindications. Informed consent was obtained from all participants prior to MR scanning. The study protocol was approved by the local ethics committee.

Experimental design

The context verification task required subjects to judge whether congruent or incongruent target words

Behavioral results

During the fMRI session the following mean (± SEM) reaction times (in ms) were recorded for the four experimental conditions: IA = 1150.8 (62.4), IU = 1146.8 (56.6), CA = 1097.1 (66.1), CU = 1093.3 (64.6). A within-subject repeated measures ANOVA that crossed the within-subject factors homonymy and congruency yielded slightly faster reaction times for semantically congruent than for semantically incongruent target decisions (F(1,21) = 4,15, p < 0.05). Neither the main effect for ambiguity, nor the

Discussion

The present fMRI study investigated the brain network involved in semantic ambiguity processing using a context verification task (CVT) that poses a high demand on active lexico-semantic ambiguity resolution. The results of this investigation were threefold: (1) Increased ambiguity-related brain activity in a widespread brain network including frontal, temporal, and parietal brain areas closely replicates previous neuroimaging evidence reported by Chan et al. (2004). They build and expand on

Conclusions

The present study directly assessed the neural correlates of semantic ambiguity processing during semantic context verification. It provides for a detailed identification of a bilateral network involved in processing sentence-final lexico-semantic ambiguity. In building and expanding on previous fMRI evidence on ambiguity processing, a specific characterization is presented regarding the functional contributions of the identified ambiguity-network components, consistent with the hemispheric

Acknowledgments

The authors thank Bärbel Herrnberger, Markus Kiefer, Marcel Daamen and Imogen Scheef for helpful comments on earlier versions of this manuscript. We are also grateful to two anonymous reviewers for their highly valuable comments.

References (82)

  • FaustM. et al.

    Sentence context and lexical ambiguity resolution by the two hemispheres

    Neuropsychologia

    (1998)
  • FaustM.E. et al.

    Cerebral mechanisms for suppression of inappropriate information during sentence comprehension

    Brain Lang.

    (1996)
  • FaustM. et al.

    Priming summation in the cerebral hemispheres: evidence from semantically convergent and semantically divergent primes

    Neuropsychologia

    (2002)
  • FaustM. et al.

    Semantically convergent and semantically divergent priming in the cerebral hemispheres: lexical decision and semantic judgment

    Brain Res. Cogn. Brain Res.

    (2003)
  • FristonK.J. et al.

    How many subjects constitute a study?

    Neuroimage

    (1999)
  • GrindrodC.M. et al.

    Sensitivity to local sentence context information in lexical ambiguity resolution: evidence from left- and right-hemisphere-damaged individuals

    Brain Lang.

    (2003)
  • GrindrodC.M. et al.

    Hemispheric contributions to lexical ambiguity resolution in a discourse context: evidence from individuals with unilateral left and right hemisphere lesions

    Brain Cogn.

    (2005)
  • GrossmanM. et al.

    The neural basis for categorization in semantic memory

    Neuroimage

    (2002)
  • HagoortP.

    Impairments of lexical-semantic processing in aphasia: evidence from the processing of lexical ambiguities

    Brain Lang.

    (1993)
  • Jung-BeemanM.

    Bilateral brain processes for comprehending natural language

    Trends Cogn. Sci.

    (2005)
  • KatzW.F.

    An investigation of lexical ambiguity in Broca's aphasics using an auditory lexical priming technique

    Neuropsychologia

    (1988)
  • KieferM. et al.

    Right hemisphere activation during indirect semantic priming: evidence from event-related potentials

    Brain Lang.

    (1998)
  • KnightR.T. et al.

    Prefrontal cortex regulates inhibition and excitation in distributed neural networks

    Acta Psychol.

    (1999)
  • LevyB.J. et al.

    Inhibitory processes and the control of memory retrieval

    Trends Cogn. Sci.

    (2002)
  • MasonR.A. et al.

    Lexical ambiguity in sentence comprehension

    Brain Res.

    (2007)
  • PriceC.J. et al.

    Cognitive conjunction: a new approach to brain activation experiments

    Neuroimage

    (1997)
  • StoetG. et al.

    Single neurons in posterior parietal cortex of monkeys encode cognitive set

    Neuron

    (2004)
  • SwaabT.Y. et al.

    Understanding ambiguous words in sentence contexts: electrophysiological evidence for delayed contextual selection in Broca's aphasia

    Neuropsychologia

    (1998)
  • Thompson-SchillS. et al.

    Effects of repetition and competition on activity in left prefrontal cortex during word generation

    Neuron

    (1999)
  • WagnerA.D. et al.

    Prefrontal contributions to executive control: fMRI evidence for functional distinctions within lateral Prefrontal cortex

    Neuroimage

    (2001)
  • ZempleniM.Z. et al.

    Semantic ambiguity processing in sentence context: evidence from event-related fMRI

    Neuroimage

    (2007)
  • AndersonM.C. et al.

    Neural systems underlying the suppression of unwanted memories

    Science

    (2004)
  • BaayenR.H. et al.

    The CELEX Lexical Database (Release 2) [CD-ROM]

    (1995)
  • BeemanM.

    Coarse semantic coding and discourse comprehension

  • BeemanM.J. et al.

    The right hemisphere maintains solution-related activation for yet-to-be-solved problems

    Mem. Cognit.

    (2000)
  • BinderJ.R. et al.

    Human brain language areas identified by functional magnetic resonance imaging

    J. Neurosci.

    (1997)
  • BinderJ.R. et al.

    Neural correlates of lexical access during visual word recognition

    J. Cogn. Neurosci.

    (2003)
  • BookheimerS.

    Functional MRI of language: new approaches to understanding the cortical organization of semantic processing

    Annu. Rev. Neurosci.

    (2002)
  • BucknerR.L.

    Beyond HERA: contributions of specific prefrontal brain areas to long-term memory retrieval

    Psychon. Bull. Rev.

    (1996)
  • ChiarelloC.

    On codes of meaning and the meaning of codes: semantic access and retrieval within and between hemispheres

  • ChikazoeJ. et al.

    Activation of right inferior frontal gyrus during response inhibition across response modalities

    J. Cogn. Neurosci.

    (2007)
  • Cited by (61)

    • Neural correlates of semantic ambiguity resolution in paucisymptomatic bipolar disorder patients

      2021, Psychiatry Research - Neuroimaging
      Citation Excerpt :

      The study was approved by the regional ethics committee (CCP Est-3, French National Regulatory Authority) and carried out in accordance with the Declaration of Helsinki and its subsequent amendments. We used a SAR task derived from the one developed by Hoenig and Scheef (2009) and described in detail elsewhere (Raucher-Chéné et al., 2017). Briefly, it featured 90 auditory French context sentences, each followed by a written target word.

    View all citing articles on Scopus
    View full text