Metric Selection Effects in Consciousness Measurement: Lessons from Sleep EEG Analysis
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
Integrated Information Theory (IIT) predicts gradual consciousness changes, while Global Workspace Theory (GWT) emphasizes discrete threshold events. We tested whether sleep stage transitions exhibit these distinct patterns using quantitative EEG analysis.
Methods
We analyzed 622 sleep stage transitions from 12 healthy adults (Wake→N1, N1→N2, N2→N3, REM→Wake) using the Consciousness Gradient Index (CGI = √(φ × ρ) × 10). We employed two approaches: (1) an adaptive method selecting transition-specific metrics (alpha power, spindle density, spectral entropy), and (2) control analyses applying uniform metrics (spectral entropy, Lempel-Ziv complexity) across all transitions.
Results
The adaptive approach showed N1→N2 as the steepest transition (slope = −2.161, d = 2.814, p < 0.001). However, control analyses revealed this pattern to be metric-dependent. With spectral entropy applied uniformly, all transitions showed near-zero slopes (0.001-0.004), with N1→N2 ranking 3rd in magnitude. With Lempel-Ziv complexity, N1→N2 showed the smallest magnitude (0.001). Neither control supported N1→N2 as uniquely steep.
Conclusions
The apparent “dual architecture” resulted from selecting metrics based on known neurophysiological changes at each transition, creating circular reasoning. Control analyses using unbiased metrics showed no meaningful consciousness changes during any transition. This study demonstrates the importance of validation with uniformly-applied metrics and serves as a methodological warning about metric selection effects in consciousness research. The adaptive CGI framework successfully detected known neurophysiological changes but did not reveal fundamental differences in consciousness transition mechanisms.
