Network Neuroscience and Personality

Abstract

Personality and individual differences originate from the brain. Despite major advances in the affective and cognitive neurosciences, however, it is still not well understood how personality and single personality traits are represented within the brain. Most research on brain-personality correlates has focused either on morphological aspects of the brain such as increases or decreases in local gray matter volume, or has investigated how personality traits can account for individual differences in activation differences in various tasks. Here, we propose that personality neuroscience can be advanced by adding a network perspective on brain structure and function, an endeavor that we label personality network neuroscience. With the rise of resting-state functional magnetic resonance imaging (MRI), the establishment of connectomics as a theoretical framework for structural and functional connectivity modeling, and recent advancements in the application of mathematical graph theory to brain connectivity data, several new tools and techniques are readily available to be applied in personality neuroscience. The present contribution introduces these concepts, reviews recent progress in their application to the study of individual differences, and explores their potential to advance our understanding of the neural implementation of personality. Trait theorists have long argued that personality traits are biophysical entities that are not mere abstractions of and metaphors for human behavior. Traits are thought to actually exist in the brain, presumably in the form of conceptual nervous systems. A conceptual nervous system refers to the attempt to describe parts of the central nervous system in functional terms with relevance to psychology and behavior. We contend that personality network neuroscience can characterize these conceptual nervous systems on a functional and anatomical level and has the potential do link dispositional neural correlates to actual behavior.

Keywords: brain connectivity; conceptual nervous system; connectome; personality traits; resting-state fMRI.

Figure 1

Illustration of the workflow of connectome reconstruction. A connectome combines information from cortical gray and white matter (A). Diffusion imaging and computational reconstruction methods are applied to obtain a detailed map of white matter fiber connections (B). The cortical ribbon is parcellated into a set of non-overlapping regions of interest, (C) illustrates Freesurfer’s Desikan-Killiany atlas. Connectome reconstruction combines the information from (B) and (C) and determines whether two ROIs are touched by any white matter fibers or not. (D) Shows two examples: the left and right superior frontal gyrus (left) are densely connected (E, left) while the left medial orbitofrontal and superior parietal gyrus (D, right) are only sparsely connected (E). Results can be displayed in a connectivity matrix. Rows and columns are regions of interest from the whole brain parcellation. Matrix elements indicate the presence (white) or absence (black) of a connection (F). The left matrix groups brain regions according to hemisphere and cortical vs. subcortical, the right matrix according to their allegiance to network modules.

Figure 2

Circos plot (Krzywinski et al., 2009) depiction of a structural connectome. In total, 82 brain regions of Freesurfer’s Desikan atlas are ordered according to hemisphere and lobe, and arranged on a circle. The red heat map illustrates degree centrality, a measure that quantifies the brain region’s number of connections (darker shades indicate higher degree). The brain areas in dark gray are the brain regions with the highest degree (top 15%) that qualify as possible hub regions. Data are taken from Markett et al. (2017b).

Figure 3

Functional connectivity from blood oxygen level dependent (BOLD) functional magnetic resonance imaging is estimated by correlating the extracted BOLD time series from regions of interest or single voxels (left panel). The right panel shows prominent resting-state functional connectivity networks as revealed by independent component analysis (data courtesy of the authors). The networks in the top row correspond from left to right to the visual, the default mode, the somatomotor, and the dorsal attention network. The networks in the bottom row correspond to the left and right fronto-parietal, the frontal, and the salience network. All networks are displayed in radiological convention (left is right).

Figure 4

Toy networks for the illustration of measures from network theory. The global efficiency of a network can be quantified by its characteristic path length (CPL). CPL is the average of all shortest path length in a network. The shortest path between any two nodes in a network equals the minimum number of edges that have to be transversed to reach one node from the other. In the network in (A), the shortest path equals 1 between node 1 and node 2, equals 2 between node 1 and node 4, and 4 between node 1 and node 7. The CPL of the network is 2.0476. The toy graph in (B) illustrates degree centrality and betweenness centrality. Degree centrality of a node equals the number of its connection to other nodes. The gray nodes have a degree of one, the black nodes a degree of four, and the blue and red node a degree of 5. In general, it is assumed that high degree nodes are more important for the network as a whole. Betweenness centrality is a further centrality measure that can capture different information on the importance of a node. Betweenness reflects the amount of shortest paths between any two nodes that travel through a given node. The red node, for instance, has a higher betweenness than the blue nodes (even though they have the same degree). All shortest paths between any gray and any black node travel through the red and the blue node, however, all shortest paths between any two black nodes travel through the red but not through the blue node, hence its higher betweenness. The networks in (C) illustrate the clustering coefficient, which is regarded a measure of local efficiency. The numbers indicate the clustering coefficient for the red node in the network. The clustering coefficient gives the ratio of neighboring nodes that share a connection themselves over the total number of connections that neighbors could share. In the first graph, none of the black nodes (the neighbors) share a connection, hence the clustering coefficient of 0. In the last network, all neighbors are directly connected to each other (six connections between neighbors), hence the clustering coefficient of one.