Feature Uncertainty Predicts Behavioral and Neural Responses to Combined Concepts

Abstract

The cognitive and neural structure of conceptual knowledge affects how concepts combine in language and thought. Examining the principles by which individual concepts (e.g., DIAMOND, BASEBALL) combine into more complex phrases (e.g., "baseball diamond") can illuminate not only how the brain combines concepts but also the key ingredients of conceptual structure. Here we specifically tested the role of feature uncertainty in the modulation of conceptual brightness evoked by adjective-noun combinations (e.g., "dark diamond") in male and female human subjects. We collected explicit ratings of conceptual brightness for 45 noun concepts and their "dark" and "light" combinations, resulting in a measure reflecting the degree of conceptual brightness modulation in each noun concept. Feature uncertainty was captured in an entropy measure, as well as in a predictive Bayesian model of feature modulation. We found that feature uncertainty (i.e., entropy) and the Bayesian model were both strong predictors of these behavioral effects. Using fMRI, we observed the neural responses evoked by the concepts and combinations in a priori ROIs. Feature uncertainty predicted univariate responses in left inferior frontal gyrus, and multivariate responses in left anterior temporal lobe were predicted by degree of conceptual brightness modulation. These findings suggest that feature uncertainty is a key ingredient of conceptual structure, and inform cognitive neuroscience theories of conceptual combination by highlighting the role of left inferior frontal gyrus and left anterior temporal lobe in the process of flexible feature modulation during comprehension of complex language.

The Neural Correlates of Semantic Control Revisited

“…Alternatively, this may be due to the inclusion of nonverbal stimuli. The AG has been specifically associated with sentential and combinatorial processing (Graves et al, 2010;Humphreys and Lambon Ralph, 2014;Price et al, 2015;Solomon and Thompson-Schill, 2020;Branzi et al, submitted).…”

Context Free and Context-Dependent Conceptual Representation in the Brain

“…The results were highly consistent when spherical ROIs based on MNI coordinates from published studies were used instead of these clusters. We selected a left IFG ROI around the peak voxel reported by (Cardillo et al 2012; Solomon and Thompson-Schill 2020) at MNI coordinate (x = −54, y = 24, z = 10; 275 voxels). This site has been linked to semantic selection and modulation of conceptual features.…”

“…Left IFG has long been linked to semantic selection and control processes (Thompson-Schill et al 1997; Jefferies 2013; Noonan et al 2013; Jackson 2020), and is activated during the retrieval of weak semantic associations (Lambon Ralph et al 2017; Jefferies et al 2020), but its role in conceptual combination is largely unknown. One recent study found that left IFG is sensitive to feature uncertainty during comprehension of combined concepts, while ATL reflects the integration of conceptual features (Solomon and Thompson-Schill 2020). Another recent study investigated how the brain resolves semantic ambiguity in homonym comprehension and found that IFG supports context-appropriate meaning (Hoffman and Tamm 2020).…”

“…Multivoxel pattern analysis (MVPA) provides us with a powerful tool to probe how the representation of semantic information in the brain varies according to the context. MVPA studies have started to explore how features combine to form concepts and how word meaning is modified syntactically (Allen et al 2012; Coutanche and Thompson-Schill 2014; Boylan et al 2015; Hoffman and Tamm 2020; Solomon and Thompson-Schill 2020). A recent magnetoencephalography study showed that neural representations of the verb DO were modified across temporal, inferior frontal and inferior parietal regions according to the noun it was combined with (Lyu et al 2019).…”

“…In these situations, semantic control processes might be able to bias the pattern of semantic retrieval in task-appropriate ways by facilitating or inhibiting connections between the heteromodal hub in ATL and task-relevant and task-irrelevant spokes. Multivoxel pattern analysis (MVPA) provides us with a powerful tool to probe how the representation of semantic information in the brain varies according to the context; these studies have started to explore how features combine to construct concepts and how word meaning is modified syntactically (Allen et al 2012;Coutanche and Thompson-Schill 2014;Boylan et al 2015;Hoffman and Tamm 2020;Solomon and Thompson-Schill 2020). For example, a recent magnetoencephalography study showed that neural representations of the noun were modified across temporal, inferior frontal and inferior parietal regions according to the verb it was combined with (Lyu et al 2019).…”