Sleep unconsciousness and the fragmentation of the global workspace

While several mental functions, from visual perception up to working memory, are characterized by parallel computation performed by integrated moduli in the cortex, consciousness is sustained by a serial process of global integration, insofar as a single scene at a time takes place. Rigorous studies on the theoretical physics of second order phase transition (i.e. critical phenomena) show that the so called order parameters, defined as the macroscopic variables (i.e. thermodynamically measurable quantities that integrate the activity of many components) that respond to external fields (i.e. interactions), display, in the absence of interactions, coordinated fluctuations with an intermittent serial structure when the system is at the transition, termed critical state. This "criticality" state is operationally defined by the presence of avalanches with inverse-power-law (scale-free) distribution densities of sizes and inter-event times (duration of metastable states, i.e. states without events). Criticality is a state of maximal complexity, with maximal redundancies in the dynamical patterns, and complex topologies supporting the structure of cross correlations between different areas. The state of criticality has been established by our group in human brain dynamics in basal conditions, by studying abrupt transitions (RTPs, or rapid transition processes) to and from stationary states, via multichannel EEGs (electroencephalograms). RTPs are the events that delimit quasi-stationary (sometimes called metastable) epochs with constant frequencies and amplitude. It remained unsolved whether this complex behavior correlates with consciousness or, alternatively, with a non-task-driven default mode activity of the brain, also present in non-conscious states, such as NREM sleep. This does not mean that default mode and consciousness are in conceptual contrast, but that so far we had not assessed whether our finding of critical neural dynamics was a correlate of either the former or the latter. Here we show that in NREM sleep this serial dynamical behavior breaks down, insofar as the inverse-power-law distributions of the inter-event times are replaced, in the long-time regime, by exponential cutoffs whose time scales correlate with the average time between episodes of neural bistability,marked in EEGs by Sleep SlowOscillations (SSOs). During REM sleep the dynamics turns back to the scale-free behavior observed during the pre-sleep wakefulness. We demonstrate that the dynamics of passage between quasi stationary states in unconscious NREM sleep has a spatial connectivity not significantly different from the spatial connectivity of wakefulness and REM sleep. However, NREM RTPs are not dynamically compatible with a serial (single time) scenario, with a strong discrepancy with respect to the serial behavior of wakefulness and REM sleep, where consciousness takes place. This is ultimately due to the fact that the global workspace, the most used model to describe consciousness, cannot emerge due to NREM neural bistability.