Schematic of the proposed SAPHIRE neurocognitive architecture.

Below, specialised modules characterised by robust redundant interactions process information about the environment. Information is then collected by workspace gateways through synergistic interactions [Step (i)]; synergistic interactions integrate information within the synergistic global workspace [Step (ii)]; workspace broadcasters spread the integrated information back to the specialised modules, through redundant interactions [Step (iii)], for further processing and to guide behaviour. Orange connections represent redundant interactions, and violet connections represent synergistic interactions. Grey connections represent interactions between the system and its environment.

Gateways and broadcaster regions identified by their network connectivity profiles.

(A) Group-average matrices of synergistic interactions between regions of the 454-ROI augmented Schaefer atlas (39, 43). (B) Group-average matrices of redundant interactions. We highlighted modular allegiance to the canonical resting-state networks by using the colour scheme shown in between A and B. (C) Regions are identified as gateways (violet) or broadcasters (orange) based on the difference between rank of participation coefficient for synergy and redundancy, (only shown for brain regions identified as belonging to the synergistic global workspace). Violet indicates synergy rank > redundancy rank, corresponding to workspace regions that combine information of many brain modules (gateways); orange indicates the opposite, identifying workspace regions that broadcast information to many modules. Inset: illustration of the synergistic workspace. Legend: DMN, default mode network. Som, somatomotor network. Vis, visual network. VAN, ventral attention network. DAN, dorsal attention network. FPN, fronto-parietal control network. Lim, limbic network. Sub, subcortical network (comprised of 54 regions of the Tian 2020 atlas 52). These results were also replicated using an alternative parcellation with 232 cortical and subcortical nodes (SI Figure 1).

Loss of consciousness induces similar reorganisation of cortical integrated information across anaesthesia and disorders of consciousness. Top: Brain regions exhibiting overall NBS-corrected increases (red) and decreases (blue) in integrated information exchange when consciousness is lost. (A) DOC patients minus awake healthy volunteers; (B), propofol anaesthesia minus pre-induction wakefulness; (C) propofol-anaesthesia minus post-anaesthetic recovery. (D) Overlaps between the three contrasts in (A-C), showing increases and decreases that are common across anaesthesia and disorders of consciousness.

Synergistic core of human consciousness.

(A) Surface projections indicate brain regions that play the role of broadcasters (orange) or gateways (violet) in the synergistic workspace, and regions that exhibit an overall significant reduction in integrated information across anaesthesia and disorders of consciousness (blue). Network representation: edges indicate reduced integrated information during both propofol anaesthesia and disorders of consciousness, between gateway (violet) and broadcaster (orange) nodes of the workspace. (B) Circular graph representation of significant reductions in integrated information (ΦR) between brain regions, observed in all three contrasts, divided into canonical resting-state networks. Legend: DMN, default mode network. Som, somatomotor network. Vis, visual network. VAN, ventral attention network. DAN, dorsal attention network. FPN, fronto-parietal control network. Lim, limbic network. Sub, subcortical network (comprised of 54 regions of the Tian 2020 atlas).