The distinct and overlapping brain networks supporting semantic and spatial constructive scene processing

Abstract

Scene imagery features prominently when we recall autobiographical memories, imagine the future and navigate around in the world. Consequently, in this study we sought to better understand how scene representations are supported by the brain. Processing scenes involves a variety of cognitive processes that in the real world are highly interactive. Here, however, our goal was to separate semantic and spatial constructive scene processes in order to identify the brain areas that were distinct to each process, those they had in common, and the connectivity between regions. To this end, participants searched for either semantic or spatial constructive impossibilities in scenes during functional MRI. We focussed our analyses on only those scenes that were possible, thus removing any error detection that would evoke reactions such as surprise or novelty. Importantly, we also counterbalanced possible scenes across participants, enabling us to examine brain activity and connectivity for the same possible scene images under two different conditions. We found that participants adopted different cognitive strategies, which were reflected in distinct oculomotor behaviour, for each condition. These were in turn associated with increased engagement of lateral temporal and parietal cortices for semantic scene processing, the hippocampus for spatial constructive scene processing, and increased activation of the ventromedial prefrontal cortex (vmPFC) that was common to both. Connectivity analyses showed that the vmPFC switched between semantic and spatial constructive brain networks depending on the task at hand. These findings further highlight the well-known semantic functions of lateral temporal areas, while providing additional support for the previously-asserted contribution of the hippocampus to scene construction, and recent suggestions that the vmPFC may play a key role in orchestrating scene processing.

Multimodal MEG and microstructure-MRI investigations of the human hippocampal scene network

“…The finding of greater hippocampal theta modulation during scene, than face, odd-one-out discrimination, builds upon prior fMRI studies showing that scene oddity engages the hippocampus (Lee et al, 2013;Lee et al, 2008) and the literature of MEG investigations of the hippocampus (e.g., see Ruzich et al, 2019 for reviews of hippocampal recording with MEG). Notably, when intersecting our scene-related theta modulation maps, we identified a peak in hippocampal theta modulation in the comparison of scene versus face conditions, in the anteromedial hippocampus, aligning with results of fMRI studies that have observed anteromedial hippocampal activity during scene imagery (Dalton & Maguire, 2017;McCormick & Maguire, 2021;Zeidman et al, 2015) and perceptual scene discrimination (Gardette et al, 2022;Hodgetts, Voets, et al, 2017). High-resolution fMRI work suggest that this scene-selective region likely corresponds to the anteromedial subicular complex (Dalton & Maguire, 2017;Hodgetts, Voets, et al, 2017;Read et al, 2024), which is the most connected region of the hippocampus, (in part, though not exclusively, via the fornix) to an extended hippocampal system (Aggleton & Christiansen, 2015), and is known to receive spatial input from cortical areas such as caudal inferior parietal, retrosplenial and parahippocampal cortices (Aggleton & Christiansen, 2015).…”

Allocentric information represented by self‐referenced spatial coding in the primate medial temporal lobe

“…In summation, we hypothesize that the constructive process in the HPC fits the memory to the current spatial context to generate a first-person perspective from the item location in the allocentric space (Yang & Naya, 2020) (Figure 6). The presence of constructive process in the HPC has also been suggested by other research groups based on human imaging studies, but the previously suggested process either combines pieces of memory for remembering the past and imagining the future or combines pieces of perception for scene perception (McCormick & Maguire, 2021;Schacter, 2012;Schacter et al, 2007;Zeidman & Maguire, 2016).…”

Attention and schema violations of real world scenes differentially modulate time perception

“…According to them, semantics are about the object’s fitting into the global context of a scene, while syntactic relations refer to objects being at their expected location in a scene. This notion was further supported by the different neural representations of semantic and syntactic scene violations with the former producing negative deflections in the N300-N400 time window, while the latter positive deflections in the P600 2 , 3 .…”