Luppi, A. I., Uhrig, L., Tasserie, J., Mediano, P. A. M., Rosas, F. E., Singleton, S. P., Gutierrez-Barragan, D., Gini, S., Castro, P., Signorelli, C. M., Golkowski, D., Ranft, A., Ilg, R., Jordan, D., Muta, K., Hata, J., Okano, H., Liu, Z.-Q., Yee, Y., Destexhe, A., Cofre, R., Menon, D. K., Gozzi, A., Jarraya, B. & Stamatakis, E. A. (2026). Convergent transcriptomic and connectomic controllers of information integration and its anaesthetic breakdown across mammalian brains. Nature Human Behaviour 10(4): 777–802. doi:10.1038/s41562-025-02381-5.
Key takeaways
From the reading, NOTE-1064.
- The measure. ΦR is the original whole-minus-sum Φ with the persistent redundancy added back, since the naive sum of the parts counts it twice. It is the synergy-containing atoms of integrated information decomposition plus the two pure-transfer atoms, computed per region pair and averaged (pp. 778–779, 795–796).
- It falls with responsiveness. Lower under anaesthesia in four species and five of six regimes, with medetomidine–isoflurane in the mouse the exception (p. 781), which a six-region check recovers (p. 791). It returns on recovery in humans.
- It returns with thalamic stimulation. In two macaques under continuous propofol, central thalamic stimulation restored arousal and ΦR, so ΦR “co-varies with the consciousness-suppressing effect of anaesthetics, not with their mere presence in the system” (p. 781). Low-intensity stimulation of the control site raised ΦR somewhat without restoring arousal.
- It beats the rivals. It carries 51% of the explained variance in the macaque arousal score. The original Φ is inconsistent and goes negative; causal density misses recovery (pp. 781–782, 790–791).
- Controllability. The energy to move between successive brain states rises under anaesthesia and correlates negatively with ΦR (pp. 782–784).
- Where and why. The largest regional losses are in unimodal visual and somatomotor cortex, with default-mode losses too (p. 784). They follow cortical PVALB expression in three species (pp. 784–786), and models show PVALB-shaped inhibition is especially suited to lowering ΦR (pp. 787–789).
- What the authors concede. Responsiveness is “an imperfect marker of unconsciousness” (p. 792); pairwise ΦR is “perhaps best viewed as a lower bound” (p. 793); “the consistency was not perfect” (p. 793).
Why this is here
The owner’s exchange of 2026-10-21, “Experience and interpretation”, asked whether experience is constituted by interpretation or produced by it. The exchange cited this study for its finding across four species that integration breaks down under anaesthesia, and that thalamic stimulation reverses the breakdown in macaques. The reading confirms that description, with two qualifications: the reversal rests on two macaques, and what breaks down is a content-free statistical integration that goes with responsiveness, not interpretation in TERM-073’s sense. It measures integration, not the constitution of experience. The record holds the same group’s synergistic workspace paper, LIT-949, which this one cites as ref. 45; it half meets THEORY-240’s promotion test, restoring ΦR by thalamic stimulation without showing that the restored ΦR is on the gateway side. Bears on QUESTION-048, as deficit-and-recovery evidence whose marker is access, and on CLAIM-412 and CLAIM-411, where integration is one candidate condition. Filed from this reading: THEORY-881. It carries no instruction for machine-learning practice.