Arousal refers to changes in brain-body state underpinning motivated behavior but lacks a proper definition and taxonomy. Neuroscience and psychology textbooks offer surprisingly different views on what arousal is, from a global brain-wide modulation of neuronal activity to a multidimensional construct, with specific brain-body patterns tuned to a given situation. The huge number of scientific articles mentioning arousal (~50,000) highlights the importance of the concept but also explains why such a vast literature has never been systematically reviewed so far. Here, we leverage the tools of natural language processing to probe the nature of arousal in a data-driven, comprehensive manner. We show that arousal comes in seven varieties: cognitive, emotional, physiological, sexual, related to stress disorders, to sleep, or to sleep disorders. We then ask whether domain-general arousal exists at the cortical level, and run meta-analyses of the brain imaging literature to reveal that all varieties of arousal, except arousal in sleep disorders for lack of data, converge onto a cortical network composed of the presupplementary motor area and the left and right dorsal anterior insula. More precisely, we find that activity in dysgranular insular area 7 (Jülich atlas), the region with the highest convergence across varieties of arousal, is also specifically associated with arousal. The domain-general arousal network might trigger the reorganization of large-scale brain networks-a global mechanism-resulting in a context-specific configuration-in line with the multidimensional view. Future taxonomies of arousal refining the alignment between concepts and data should include domain-general arousal as a central component.
Keywords: arousal; meta-analysis; natural language processing; resting-state neural networks; semantic analysis.