Source
Luppi, Mediano, Rosas et al. (2024), eLife 12:RP88173, LIT-949, read in NOTE-799: Results (Figs. 2, 4, 5 and their robustness analyses), Discussion and its limitations, and Methods (Eqs. 4–8, the participation-coefficient split, and the network-based statistic with the composite shared-effect test).
What was actually shown
The workspace and its two sides (HCP, N = 100). Pairwise synergy and redundancy were computed between 454 regions with Gaussian MMI-ΦID.
- Workspace. The regions whose synergy rank exceeds their redundancy rank by nodal strength. This definition is taken from the authors’ earlier paper.
- The split. Within the workspace, regions whose participation coefficient ranks higher for synergy than for redundancy are gateways; the rest are broadcasters.
- The measured result. Gateways lie mainly in the default mode network, and broadcasters mainly in the executive control network, with lateral prefrontal cortex prominent. This replicates at a second parcellation and with data-driven modules.
The consciousness contrast. Pairwise ΦR = I(past; future) − I(X_past; X_future) − I(Y_past; Y_future) + Red was compared across three contrasts:
- propofol at Ramsay 5 against pre-induction wakefulness;
- propofol against post-anaesthetic recovery;
- 22 DOC patients against awake controls.
A shared-effect test required all three contrasts to change. The edges whose ΦR consistently fell link bilateral precuneus, medial prefrontal and inferior parietal cortex, with some left temporal cortex. All of these regions are gateways in the HCP classification. The decreases in medial prefrontal and posterior cingulate or precuneus survive a coarser parcellation and stricter thresholds. The net sum of F-scores over significant edges is −25.37.
What could have failed. The consistent decreases could have fallen on broadcasters, or on no workspace region. The original whole-minus-sum Φ, on the same data, found no common decrease at all.
What this does not say
- Not that consciousness is gateway integration. The authors disclaim that ΦR “represent[s] a direct cause of consciousness, or [is] identical to it”. The result is a correlation in resting-state BOLD.
- Not that the workspace has an input side and an output side. Every workspace region is assigned to one class by the sign of a relative rank difference, so a two-part workspace is guaranteed by the method. What is measured is where the classes fall and which class carries the change.
- Not anything about the access of a content. No stimulus, report or per-consumer measure is involved. It cannot show that a content reaches some systems and not others, the result QUESTION-061 asks for.
- Not that broadcasters gain integration in unconsciousness. The increases near lateral prefrontal cortex are not robust: “no region consistently showing increases”.
- Not IIT’s Φ. The measures are pairwise between regional time series, undirected, and built on IIT 2.0’s Φ. They are not the integrated information of the whole system.
- Not a clean contrast of conscious and unconscious states. The DOC group includes minimally conscious patients. The authors concede that some propofol participants may have been “disconnected but still conscious”. Resting state may overweight the DMN against the FPN.
- Not fully specified. The shared-effect test’s details are deferred to a future publication. The propofol sample is given as both N = 15 and n = 16.
Connections
- QUESTION-061 (Deferred). The result bears on the question only by
showing that, at rest, the workspace is not uniform with respect to the
state of consciousness: the consistent loss is on the input side. It is
not evidence about content access. No
answers:relation is filed. - THEORY-096 (GNW). It agrees that lateral prefrontal cortex acts as a broadcaster. It locates the consciousness-tracking change in default-mode gateways instead, and it does not test ignition or exclusivity.
- THEORY-098. The reductions under propofol in thalamic, auditory and somatomotor regions fit the authors’ reading of thalamocortical isolation: a loss of input to the core. That is a state effect, and it does not bear on sensory recurrence as the basis of experience.
- CLAIM-003 (Deferred). It does not bear: the measure is undirected and content-free.
- LIT-025 reviews the ΦID framework whose pairwise measures are used here.