General and feature-based semantic representations in the semantic network

Abstract

How semantic representations are manifest over the brain remains a topic of active debate. A semantic representation may be determined by specific semantic features (e.g. sensorimotor information), or may abstract away from specific features and represent generalized semantic characteristics (general semantic representation). Here we tested whether nodes of the semantic system code for a general semantic representation and/or possess representational spaces linked to particular semantic features. in an fMRi study, eighteen participants performed a typicality judgment task with written words drawn from sixteen different categories. Multivariate pattern analysis (MVPA) and representational similarity analysis (RSA) were adopted to investigate the sensitivity of the brain regions to semantic content and the type of semantic representation coded (general or feature-based). We replicated previous findings of sensitivity to general semantic similarity in posterior middle/inferior temporal gyrus (pMTG/ITG) and precuneus (PC) and additionally observed general semantic representations in ventromedial prefrontal cortex (PFC). Finally, two brain regions of the semantic network were sensitive to semantic features: the left pMTG/ITG was sensitive to haptic perception and the left ventral temporal cortex (VTC) to size. This finding supports the involvement of both general semantic representation and feature-based representations in the brain's semantic system. Conceptual and semantic knowledge are fundamental aspects of human cognition and the investigation of the neural substrates underlying these processes is an ongoing topic of research in the cognitive neurosciences. Although current evidence has demonstrated that semantic knowledge is represented in a distributed manner over the brain 1,2 , the manner in which semantic representation is manifest remains a topic of active debate. The association damage to the anterior temporal lobe primary progressive aphasia, herpes encephalitis and lesions led to an emphasis of this brain regions as a critical locus for semantic processing 3,4. However, functional neuroimaging has suggested a broader range of regions are involved in semantic processing 5. A meta-analysis of 120 studies 6 identified a "general semantic network"-a left-lateralized network consisting of seven brain regions that were activated in a variety of semantic tasks: angular gyrus, lateral and ventral temporal cortex, ventromedial prefrontal cortex, inferior frontal gyrus, dorsal medial prefrontal cortex and the posterior cingulate gyrus. However, not all brain regions activated in semantic tasks necessarily represent semantic content, for instance regions may control access to semantic information rather than contain that information themselves 2,7,8. Starting from this assumption, Fairhall and Caramazza (2013) 9 identified a set of regions representing semantic content by means of Multivariate Pattern Analysis (MVPA) and Representational Similarity Analysis (RSA) 10. The authors showed that a left-lateralized netwo...

Combining Concepts Across Categorical Domains: A Linking Role of the Precuneus

“…Previous work investigating the combination of words has emphasised centralised, default mode, semantic systems (Graves et al, 2010;Palliera et al, 2011; for a discussion, see Frankland & Greene, 2020), elements of which are known to contain representations of different categories of objects (Bruffaerts et al, 2013;Devereux et al, 2013;Fairhall & Caramazza, 2013b;Liuzzi et al, 2020). To address whether the combination of concepts across domains continues to rely on distributed category-selective regions or is rather centralised into domain-general semantic systems, we compared the representation based upon single-category sentences to those of combined-category sentences across domain-sensitive ROIs.…”

A Meta-Analysis of fMRI Studies of Semantic Cognition in Children

“…The human capacity to retrieve meaning from words, phrases, and visual objects far exceeds the capacities of other animal species as well as all current state-of-the-art machine learning architectures. Functional magnetic resonance imaging (fMRI) has made it possible to map the brain areas underlying this capacity in the adult brain, showing that semantic information is processed in a distributed fashion across large parts of the cerebral cortex (Humphries et al, 2007; Huth et al, 2012, 2016; Liuzzi et al, 2020; Pulvermüller et al, 2009; Tyler et al, 2003). Most areas of this semantic system act largely amodal, that is, they show similar levels and patterns of activation regardless of whether the sensory input that is being processed comes from the visual domain or from the auditory domain (Deniz et al, 2019; Fairhall & Caramazza, 2013).…”

Convolutional neural networks reveal topic-level representations of sentences in medial prefrontal cortex, lateral anterior temporal lobe, precuneus, and angular gyrus

“…The representation of single words or concepts has been well studied across the brain's semantic system, using methods both investigating the magnitude of the response of specific voxels (Anderson et al, 2017;Deniz et al, 2019;Fernandino et al, 2015;Huth et al, 2016) and relating the semantic distance between word tokens to the distance in observed neural patterns produced by those words (Bruffaerts et al, 2013;Devereux et al, 2013;Fairhall & Caramazza, 2013;Liuzzi et al, 2020). The semantic system can be characterised as a broadly distributed network of left-lateralized frontal, temporal, parietal, and paralimbic regions that show an increased response to semantically richer stimuli.…”

“…A metaanalysis of 120 functional magnetic resonance imaging (fMRI) and positron-emission tomography (PET) studies (Binder et al, 2009) identified seven such brain regions: 1) posterior inferior parietal lobe: angular gyrus (AG) and portions of supramarginal gyrus (SMG), 2) lateral temporal cortex: middle temporal gyrus (MTG) and portions of the inferior temporal gyrus, 3) ventral temporal cortex (VTC): mid-fusiform and adjacent parahippocampal gyrus, 4) dorsomedial prefrontal cortex (dmPFC), 5) inferior frontal gyrus (IFG), 6) ventromedial prefrontal cortex (vmPFC), and 7) the precuneus. Multivariate studies show that the voxel-wise pattern of response within regions of this network are also sensitive to the semantic content of single concepts (Bruffaerts et al, 2013;Devereux et al, 2013;Fairhall & Caramazza, 2013;Liuzzi et al, 2020).…”