The Contrast-Field Account of Conscious Experience: How Self-Model Depth and the Learning–Automation Cycle Shape Consciousness

Abstract

Within the predictive processing framework, most theories of consciousness address either the architecture of conscious access, the representational structure, the information-theoretic measure, or the computational principle. What remains underexplored is the temporal dynamic of consciousness: why it emerges under novelty, recedes under familiarity, and re-emerges under disruption. This paper presents the Predictive Self-Model Consciousness framework (PSMC), which contributes two novel proposals. First, the contrast-field account: the qualitative character of conscious experience is jointly determined by prediction error magnitude and the depth of the automated self-model in the domain where the error occurs, formalised as S(k) = g(ε, π, D(k)). A deeper self-model provides a richer 'screen' against which disruptions register, producing more textured and self-involving experience. This generates a divergent prediction from standard active inference: two organisms with identical prediction errors but different self-model depths will have qualitatively different experiences. Second, the learning–automation cycle: consciousness functions as a temporary adaptive mode whose intensity is modulated by the organism's position in a continuous cycle of novelty, learning, and automation. The paper operationalises the core variable D(k) through three convergent proxies (fMRI representational similarity analysis, ERP cascade depth, behavioural discrimination depth), reviews supporting evidence from expert perception neuroscience, meditation research, motor skill acquisition, and brain complexity studies, formalises the contrast-field hypothesis within precision-weighted prediction error mathematics, and specifies three falsifiable experimental protocols that discriminate PSMC from competing accounts.