A gradient from long-term memory to novel cognition: transitions through default mode and executive cortex

Abstract

Human cognition is flexible, supporting decisions that are novel as well as those that arise from long-term memory. Traditionally, these aspects of cognition are ascribed to dichotomous neural systems supported by default mode (DMN) and multiple-demand (MDN) networks. In reality, however, most situations are neither completely familiar, nor entirely new, highlighting the need to understand how cognition is constrained in a graded fashion. A contemporary account proposes a functional connectivity gradient along the cortical surface that captures the transition from heteromodal DMN, through MDN regions, to unimodal regions. We asked whether brain responses changed systematically along the connectivity gradient as we parametrically varied the global semantic similarity of items within a feature matching task to create a 'task gradient', from conceptual combinations that were highly overlapping in long-term memory to trials that only shared the goal-relevant feature. We found the brain's response to the task gradient varied systematically along the connectivity gradient, with the strongest response in DMN when the probe and target items were highly overlapping conceptually. This graded functional change was seen in multiple brain regions and within individual brains, and was not readily explained by task difficulty. Moreover, the gradient captured the spatial layout of networks involved in semantic processing, providing an organizational principle for controlled semantic cognition across the cortex.How is human cortex organized to produce a spectrum of cognition, from efficient memory-based decisions to more novel patterns of thought? We addressed this question by asking participants to match words on the basis of a specific semantic feature, such as colour, while parametrically varying other features. This created a 'task gradient' from strong to weak overlap in long-term memory. We investigated whether this task gradient co-varied with a previously-described functional connectivity gradient which captures transitions along the cortical surface from perception and action, through executive control areas, to heteromodal regions of default mode network. We found the neural response to the task varied systematically along this connectivity gradient providing an organizational principle for semantic cognition across the cortex.

Both default and multiple-demand regions represent semantic goal information

“…Similarly, our results conflict with the claim that DMN regions only support relatively 'automatic' aspects of semantic retrieval (Vatansever et al, 2017). Although DMN regions show a greater response when strong associations are contrasted with weak associations (Davey et al, 2015;Teige et al, 2018, when meaningful inputs share a higher number of features (Wang et al, 2020) and when multiple inputs are convergent in meaning (Westerlund and Pylkkänen, 2014;Lanzoni et al, 2020), other studies find these regions also support controlled aspects of semantic cognition (Krieger-Redwood et al, 2016;Murphy et al, 2018), in line with our observation that DMN regions can decode current goals.…”

“…DMN is highly heteromodal, and thought to support information integration (Simony et al, 2016;Lanzoni et al, 2020), relevant to episodic retrieval (Sestieri et al, 2011) and semantic cognition (Binder & Desai, 2011;Wirth et al, 2011). These observations suggest that semantic DMN regions might extract information about global conceptual similarity in long-term memory (Murphy et al, 2017;Wang et al, 2020). By this view, patterns of response within DMN regions might be driven by conceptual similarity, irrespective of task.…”

Bipartite Functional Fractionation within the Default Network Supports Disparate Forms of Internally Oriented Cognition

“…2018 , 2019a ). These results have led to the contemporary accounts that the brain is equipped with adaptive machinery, implemented by the DN and SN, that supports cognition when it has to rely on internally constructed representations and external stimuli are useless or unavailable ( Wang et al. 2020 ).…”

Distinct and Common Neural Coding of Semantic and Non-semantic Control Demands

“…Contemporary accounts of brain organization suggest that neural function is organized along a connectivity gradient from unimodal regions of sensorimotor cortex, through executive regions to transmodal default mode network (Margulies et al 2016;Huntenburg et al 2018). Wang et al (2020) showed this gradient can capture the orderly transitions between MDN, SCN and DMN in semantic processing.…”