Elsevier

Neuropsychologia

Volume 141, April 2020, 107438
Neuropsychologia

Linking individual differences in semantic cognition to white matter microstructure

https://doi.org/10.1016/j.neuropsychologia.2020.107438Get rights and content

Highlights

  • Semantic and affective components of words interact in behavioural performance.

  • White matter microstructure relates to behavioural measures of semantic control.

  • Structural connectivity findings support left-lateralised semantic control.

  • Automatic and controlled semantic retrieval relates to different white matter tracts.

  • White matter connecting vision to semantic regions supports automatic retrieval.

Abstract

Semantic cognition is thought to involve the interaction of heteromodal conceptual representations with control processes that (i) focus retrieval on currently-relevant information, and (ii) suppress dominant yet irrelevant features and associations. Research suggests that semantic control demands are higher when retrieving a link between weakly-associated word pairs, since there is a mismatch between the pattern of semantic retrieval required by the task and the dominant associations of individual words. In addition, given that heteromodal concepts are thought to reflect the integration of vision, audition, valence and other features, the control demands of semantic tasks should be higher when there is less consistency between these features. In the present study, 62 volunteers completed a semantic decision task, where association strength and semantic-affective congruence were manipulated. We used diffusion tensor magnetic resonance imaging to obtain fractional anisotropy (FA) measures of white matter tracts hypothesized to be part of the semantic network. The behavioural data revealed an interaction between semantic-affective congruence and strength of association, suggesting these manipulations both contribute to semantic control demands. Next we considered how individual differences in these markers of semantic control relate to the microstructure of canonical white matter tracts, complementing previous studies that have largely focused on measures of intrinsic functional connectivity. Repeated-measures analysis of covariance showed opposing interactions between semantic control markers and FA of two tracts: left inferior longitudinal fasciculus (ILF) and right inferior fronto-occipital fasciculus (IFOF). Participants with higher FA in left ILF showed more efficient retrieval of weak associations, and more accurate performance for weak associations when meaning and valence were incongruent, consistent with the hypothesis that this left hemisphere tract supports semantic control. In contrast, participants with higher FA in right IFOF were more accurate for trials in which meaning and valence were congruent, and consequently when semantic control demands were minimised. These findings are consistent with recent studies showing that semantic control processes are strongly left-lateralised. In contrast, long-range connections from vision to semantic regions in the right hemisphere might support relatively automatic patterns of semantic retrieval.

Introduction

Our store of conceptual knowledge contains many features and associations, many of which are not relevant at a given moment, and we can flexibly shape retrieval to focus on specific non-dominant properties as required by the circumstances. For example, we know that newspapers are for reading, but when selecting an object to swat a fly, properties such as being rollable and lightweight come to the fore. This capacity for flexible semantic cognition involves the interaction of conceptual representations with control processes (Davey et al., 2016; Jefferies, 2013; Lambon Ralph et al., 2016), and consequently all semantic tasks (including the one used in the current study) are thought to engage these two neurocognitive components (although to varying degrees).

Contemporary theoretical approaches such as the ‘Hub and Spokes’ model suggest that conceptual representations arise from the integration of multiple types of features, such as visual, auditory, motor, verbal, spatial and emotional, within ventral aspects of the anterior temporal lobes (ATL; e.g., Patterson et al., 2007). Consequently, the same heteromodal conceptual representations – interacting with relevant unimodal features (“spokes”) – can support the full range of semantic tasks (e.g., Patterson et al., 2007; Lambon Ralph et al., 2016). Recent work has shown that ventral ATL responds to the meaning of words, and not the modality of presentation – i.e., whether these items were spoken or written (Murphy et al., 2017). This structure also responds to the meaningful combination of different features (Coutanche and Thompson-Schill, 2015; Teige et al., 2018) and shows convergence across inputs presented in different modalities, such as verbal and picture-based tasks (e.g., Spitsyna et al., 2006; Visser et al., 2012). Emotional valence can be thought of as a semantic feature (Kousta et al., 2011; Vigliocco et al., 2009), which is integrated in ATL along with vision and audition (Martin, 2016; Olson et al., 2013; Rice et al., 2015a, 2015b; Skipper and Olson, 2014). There is some evidence that emotion is a particularly important feature for the representation of abstract and social concepts in the ATL (Kousta et al., 2011; Vigliocco et al., 2013, 2009). The uncinate fasciculus is thought to play an important role in the integration of meaning and emotion (Bajada et al., 2015a, 2015b; Moritz-Gasser et al., 2015).

Contemporary accounts of semantic cognition suggest these heteromodal semantic representations interact with control processes that shape retrieval (e.g., Lambon Ralph et al., 2016). Semantic control refers to the ability to focus on specific aspects of knowledge, in line with the current context or goal (Whitney et al., 2011). In neuropsychological studies, degradation of heteromodal knowledge following atrophy of the ATL in semantic dementia is qualitatively distinct from deregulated semantic cognition in patients with damage to left hemisphere regions associated with semantic control (Corbett et al., 2009; Jefferies and Lambon Ralph, 2006; Jefferies et al., 2010; Rogers et al., 2015). When semantic retrieval requires dominant aspects of knowledge to be retrieved for a given concept, or a rich concept with multiple coherent features is retrieved, studies have found stronger activation of ATL (Davey et al., 2016; Bemis and Pylkkänen, 2013). Under these circumstances, the requirement for semantic control may be minimised, since semantic retrieval is relatively stable and self-reinforcing. The hypothesized semantic control network may be recruited more strongly when the task requires non-dominant features or associations to be brought to the fore, or to overcome competition between concepts or semantic features. Although early work particularly focussed on the role of left inferior frontal gyrus (IFG) in semantic control, recent studies have demonstrated that a distributed network underpins this function, including posterior middle temporal gyrus (pMTG). Left IFG and pMTG show common activation across a wide range of manipulations of semantic control demands (Noonan et al., 2013) and inhibitory stimulation of both sites disrupts controlled semantic retrieval (Davey et al., 2015a; Whitney et al., 2011). Left IFG and pMTG show strong intrinsic and structural connectivity (Davey et al., 2016). Moreover, damage or inhibitory stimulation to left IFG elicits compensatory increases in pMTG during tasks requiring controlled retrieval (Hallam et al., 2018; Hallam et al., 2016). This semantic control network lies adjacent to, but appears to be partially distinct from, domain-general executive regions (Davey et al., 2016; Noonan et al., 2013).

There has been recent interest in individual differences in semantic cognition, and how these relate to the structure and function of the brain. These studies have examined functional connectivity (Vatansever et al., 2017) and cortical thickness (Wang et al., 2018) but there has been little work relating individual differences in semantic performance to the microstructure of white matter tracts measured with diffusion tensor MRI. Nevertheless, recent studies have improved our understanding of the white matter connections that support semantic cognition. Binney et al. (2012) examined diffusion-weighted imaging within subregions of ATL and found evidence for graded feature convergence, with relative isolation from any given modality within ventral ATL (see also Bajada et al., 2017a, 2017b; Bajada et al., 2015a, 2015b; Fan et al., 2014). In addition, a recent study delineated the possible functional significance of different tracts within the temporal lobe at the group level by linking their termination regions to neuroimaging meta-analytic data (Bajada et al., 2017a, 2017b). In the current study, we examined the association between tract microstructure and individual differences in semantic performance, focusing in particular on behavioural manipulations of semantic control demands.

More than 60 participants completed a previously-used test of semantic control (Marino et al., 2015): they were asked to select the word that was semantically related to a cue word, when the link between the two words was strong (low control) or weaker (high control). Previous research has shown greater involvement of the distributed semantic control network during weak association trials, when it is necessary to shape retrieval to focus on non-dominant aspects of meaning that are currently relevant (Badre et al., 2005; Davey et al., 2015b, 2016; Hallam et al., 2016; Jefferies, 2013; Lambon Ralph et al., 2016; Noonan et al., 2013; Thompson-Schill et al., 1997; Wang et al., 2018; Whitney et al., 2010). In addition, these strong and weak association trials included cues and targets that were either consistent in their emotional valence (i.e., both positive) or inconsistent (i.e., one word was positive, while the other was negative). The demands on semantic control are expected to be maximal in this task when weakly-associated items are also inconsistent in their emotional connotations. Behavioural work has already shown an interaction between semantic association strength and emotional valence congruence (Luna et al., 2016): participants showed less efficient retrieval of weak (non-dominant) associations when emotional valence was also mismatching between the probe and target. More control might be required when the emotional valence is incongruent with the decision, since such pairs of items are thought to have less coherent features within ATL. In these circumstances, there may be some shift in the balance of semantic processing away from ATL and towards the hypothesized semantic control network. To examine the relationship between individual differences on these metrics of semantic control and white matter, we used diffusion tensor magnetic resonance imaging (Basser et al., 1994), assessing the microstructure of semantically-relevant white-matter tracts by computing fractional anisotropy (FA). FA is an index of directionality of diffusion within each voxel and it is associated with integrity of the axonal membrane and thickness of the myelin sheath (Beaulieu, 2002).

We examined four canonical tracts linked to semantic and language processing. The arcuate fasciculus is thought to support the classical dorsal language pathway (Catani and Mesulam, 2008), given it connects comprehension and production areas of language (Catani et al., 2005). The inferior fronto-occipital fasciculus links posterior and inferior frontal regions through the posterior temporal lobe with the occipital lobe (Catani and Thiebaut de Schotten, 2008). Intraoperative electrostimulation as well as lesion studies have shown evidence for this tract being the essential ventral language pathway (Almairac et al., 2015; Duffau et al., 2009), supporting conceptual processing (Kumaran et al., 2009). The role of the inferior longitudinal fasciculus in semantic processes remains unclear (Bajada et al., 2015a, 2015b). This long tract connects the anterior temporal lobe with the occipital lobe (Catani and Thiebaut de Schotten, 2008), and it has been suggested to form part of an indirect ventral language pathway (Mandonnet et al., 2007). The uncinate fasciculus connects the ventral areas of the IFG and orbitofrontal cortex with the temporal pole (Von Der Heide et al., 2013), and has been suggested to contribute to the executive control of semantic processing (Bajada et al., 2015a, 2015b; Binney et al., 2012) as well as the retrieval of valence (Bajada et al., 2015a, 2015b; Moritz-Gasser et al., 2015; Von Der Heide et al., 2013). The dorsal portion of the uncinate fasciculus is a segmentation of special interest, given it arrives at the pars orbitalis and pars triangularis, which are areas that have been associated with semantic control (Badre et al., 2005; Bajada et al., 2017a, 2017b; Noonan et al., 2013). However, the uncinate fasciculus does not terminate in ventral ATL, which is the region thought to be critical for heteromodal semantic representation (Lambon Ralph et al., 2016; Patterson and Lambon Ralph, 2016) – and this tract does not connect to posterior temporal areas also shown to be important for semantic control alongside IFG (Noonan et al., 2013).

Section snippets

Sample

The sample was composed of 63 Argentinian healthy volunteers aged between 19 and 48 years (M = 27.05, SD = 5.65), of high education level, of which 37 were women. The study protocol was accepted by the Oulton Ethical Committee in accordance with the principles of the Declaration of Helsinki and the participants provided informed consent for their participation in the study.

Procedure

We adapted a task originally designed by Badre et al. (2005) to measure semantic control (see also Whitney et al., 2011;

Behavioural task

Our analysis below considers task accuracy, median response time (RT) and response efficiency (RE; computed as RT divided by accuracy), split by condition. Participants with outlying data (±4 standard deviations) in FA or behavioural measures were excluded from the analysis (N = 1). Reaction time was not significantly correlated with accuracy (r = 0.243, p = .058), although there was a correlation with reaction time variability (calculated as the subject's coefficient of variation in reaction

Discussion

We found that individual differences in the microstructure of white-matter tracts relate to between-participant variation in the ability to shape semantic retrieval to suit the context. Previous research using similar task contrasts of weak and strong semantic associations found dissociations between (i) brain regions implicated in long-term semantic storage and (ii) areas that are engaged when non-dominant aspects of knowledge are needed or when different semantic features are inconsistent (

CRediT authorship contribution statement

Julián Marino Dávolos: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing - original draft, Writing - review & editing, Visualization, Supervision, Project administration, Funding acquisition. Juan Cruz Arias: Methodology, Software, Validation, Formal analysis, Investigation, Data curation, Writing - original draft, Writing - review & editing, Visualization. Elizabeth Jefferies: Conceptualization, Writing - original draft,

Acknowledgements

JMD and JA were funded by the Instituto de Investigación Oulton.

EJ was funded by the European Research Council (FLEXSEM 771863).

References (77)

  • T.R.D.J. Gonzalez Alam et al.

    Meaningful inhibition: exploring the role of meaning and modality in response inhibition

    Neuroimage

    (2018)
  • G.P. Hallam et al.

    Task-based and resting-state fMRI reveal compensatory network changes following damage to left inferior frontal gyrus

    Cortex

    (2018)
  • G.P. Hallam et al.

    Charting the effects of TMS with fMRI: modulation of cortical recruitment within the distributed network supporting semantic control

    Neuropsychologia

    (2016)
  • C.J. Hodgetts et al.

    Distinct contributions of the fornix and inferior longitudinal fasciculus to episodic and semantic autobiographical memory

    Cortex

    (2017)
  • E. Jefferies

    The neural basis of semantic cognition: converging evidence from neuropsychology, neuroimaging and TMS

    Cortex

    (2013)
  • E. Jefferies et al.

    “Pre-semantic” cognition revisited: critical differences between semantic aphasia and semantic dementia

    Neuropsychologia

    (2010)
  • D. Kumaran et al.

    Tracking the emergence of conceptual knowledge during human decision making

    Neuron

    (2009)
  • C. Lebel et al.

    Microstructural maturation of the human brain from childhood to adulthood

    Neuroimage

    (2008)
  • C. Murphy et al.

    Fractionating the anterior temporal lobe: MVPA reveals differential responses to input and conceptual modality

    Neuroimage

    (2017)
  • K. Patterson et al.
  • D. Perrone et al.

    The effect of Gibbs ringing artifacts on measures derived from diffusion MRI

    Neuroimage

    (2015)
  • G. Pobric et al.

    Amodal semantic representations depend on both anterior temporal lobes: evidence from repetitive transcranial magnetic stimulation

    Neuropsychologia

    (2010)
  • T.T. Rogers et al.

    Disorders of representation and control in semantic cognition: effects of familiarity, typicality, and specificity

    Neuropsychologia

    (2015)
  • L.M. Skipper et al.

    Semantic memory: distinct neural representations for abstractness and valence

    Brain Lang.

    (2014)
  • C. Teige et al.

    Dynamic semantic cognition: characterising coherent and controlled conceptual retrieval through time using magnetoencephalography and chronometric transcranial magnetic stimulation

    Cortex

    (2018)
  • D. Vatansever et al.

    Varieties of semantic cognition revealed through simultaneous decomposition of intrinsic brain connectivity and behaviour

    Neuroimage

    (2017)
  • X. Wang et al.

    The structural basis of semantic control: evidence from individual differences in cortical thickness

    Neuroimage

    (2018)
  • F. Almairac et al.

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

    Brain Struct. Funct.

    (2015)
  • C.J. Bajada et al.

    ExTracT : Extracting Tract Terminations Using Diffusion Imaging

    (2015)
  • C. Beaulieu

    The basis of anisotropic water diffusion in the nervous system–a technical review

    NMR Biomed.

    (2002)
  • DK Bemis et al.

    Basic Linguistic Composition Recruits the Left Anterior Temporal Lobe and Left Angular Gyrus During Both Listening and Reading

    Cerebr. Cortex

    (2013)
  • R.J. Binney et al.

    Convergent connectivity and graded specialization in the rostral human temporal lobe as revealed by diffusion-weighted imaging probabilistic tractography

    J. Cognit. Neurosci.

    (2012)
  • Mariem Boukadi et al.

    Test-retest reliability of diffusion measures extracted along white matter language fiber bundles using Hardi-based tractography

    Front. Neurosci.

    (2019)
  • M. Catani et al.

    Perisylvian language networks of the human brain

    Ann. Neurol.

    (2005)
  • F. Corbett et al.

    Different impairments of semantic cognition in semantic dementia and semantic aphasia: evidence from the non-verbal domain

    Brain

    (2009)
  • M.N. Coutanche et al.

    Creating concepts from converging features in human cortex

    Cerebr. Cortex

    (2015)
  • J. Davey et al.

    Automatic and controlled semantic retrieval: TMS reveals distinct contributions of posterior middle temporal gyrus and angular gyrus

    J. Neurosci.

    (2015)
  • J. Davey et al.

    Automatic and controlled semantic retrieval: TMS reveals distinct contributions of posterior middle temporal gyrus and angular gyrus

    J. Neurosci.

    (2015)
  • Cited by (7)

    View all citing articles on Scopus
    View full text