The role of the default mode network in component processes underlying the wandering mind

Abstract

Experiences such as mind-wandering illustrate that cognition is not always tethered to events in the here-and-now. Although converging evidence emphasises the default mode network (DMN) in mind-wandering, its precise contribution remains unclear. The DMN comprises cortical regions that are maximally distant from primary sensory and motor cortex, a topological location that may support the stimulus-independence of mind-wandering. The DMN is functionally heterogeneous, comprising regions engaged by memory, social cognition and planning; processes relevant to mind-wandering content. Our study examined the relationships between: (i) individual differences in resting-state DMN connectivity, (ii) performance on memory, social and planning tasks and (iii) variability in spontaneous thought, to investigate whether the DMN is critical to mind-wandering because it supports stimulus-independent cognition, memory retrieval, or both. Individual variation in task performance modulated the functional organization of the DMN: poor external engagement was linked to stronger coupling between medial and dorsal subsystems, while decoupling of the core from the cerebellum predicted reports of detailed memory retrieval. Both patterns predicted off-task future thoughts. Consistent with predictions from component process accounts of mind-wandering, our study suggests a 2-fold involvement of the DMN: (i) it supports experiences that are unrelated to the environment through strong coupling between its sub-systems; (ii) it allows memory representations to form the basis of conscious experience.

Keywords: component process account; default mode network; mind-wandering; perceptual decoupling; resting state functional connectivity.

Fig. 1.

Schematic description of the hypothesized relationship between latent neuro-cognitive components and both task performance and descriptions of experience. TC, task component; NC, neuro-cognitive component; EC, experiential component.

Fig. 2.

Identifying the functional connectivity of different subsystems of the default mode network. Results of a functional connectivity analysis in which the three default mode network subsystems as defined by Yeo et al. (2011) were used as seed regions. These maps are thresholded at Z = 2.3 and are corrected for multiple comparisons at P <0.05 FWE. The maps in the gray sub panel reflect the networks that were used as the seed regions (dorsal medial = blue; core = yellow; medial temporal = green).

Fig. 3.

Identifying the components underlying task performance and experience. The results of decomposition of the multi-dimensional experience sampling data (MDES) and the task battery to produce components of experience (EC) and components of task performance (TC). In both cases, we employed exploratory factor analysis and used varimax rotation. The number of solutions was selected based on the elbow from the scree plot. TC1 = detailed memory retrieval; TC2 = social problem solving; TC3 = external engagement; EC1 = immersive thoughts; EC2 = spontaneous off-task future thoughts; EC3 = modality of thoughts; EC4 = positive thoughts. EC, experiential components; TC, task components; AM, autobiographical memory; SR, semantic retrieval; MEPS, means ends problem solving; TOL, tower of London; RME, reading the mind in the eyes; TOM, theory of mind.

Fig. 4.

Determining the neuro-cognitive components (NC) associated with task performance. Results of multiple regressions using the task components (TCs) as the independent variables and the functional connectivity of the core of default mode network (Yeo 16) and medial-temporal subsystem (Yeo 15) as the dependent variable. These maps were created using a cluster forming threshold of Z = 2.3, and were corrected for the two tailed nature of our statistics, the number of models (3) and for the number of voxels in the brain yielding an α level of P <0.008 FWE. The scatter plots reflect the relationships between the connectivity with the region indicated in red and the relevant TC.

Fig. 5.

Scatter plots showing interactions between task components predicting experience components (based on median splits).

Fig. 6.

The relationship between the components underlying task performance, experience and the connectivity of the default mode network. Multivariate analysis of variance demonstrated that two of the neuro-cognitive components (NCs) were associated with both components that underpin task performance (TC) and those that explain the experience sampling data (ECs). The scatterplots present the correlations between task and brain and are based on 80 individuals (left), and between the NCs and experience which are based on 157 individuals (right)—scores are residualized.

Fig. 7.

Poor external engagement corresponds to coupling between the medial-temporal and dorsal-medial DMN subsystems. Poor performance on tasks with a greater reliance on external engagement is associated with coupling between medial-temporal and dorsal-medial subsystems of the DMN. The left- and right-hand columns represents the unthresholded connectivity pattern for the medial-temporal and dorsal-medial subsystems, respectively. The spatial map in the middle column is thresholded at Z = 2.3 P <0.05 FWE.

Fig. 8.

The default mode network has different modes of connectivity that correspond to different states and that can be organized along the principle gradient. Coupling between the core of the DMN and regions in fusiform/lateral visual cortex were associated with better performance on a task of social problem solving (left). In contrast, decoupling between the medial-temporal subsystem and regions of lateral temporal and prefrontal regions was associated with poor external engagement, but greater propensity for off task future thoughts (right). These correspond to patterns of connectivity that are either focused on integrating information from the unimodal end of the principle gradient into the core or coupling with regions at the heteromodal end of the gradient. The data in the middle column comes from Margulies et al. (2016). The data in the left- and right-hand columns comes from the current study and is thresholded at Z = 2.3 P <0.05 FWE.