Three-Layer Cognitive Architecture: A Structural Account of Core Processing, Modulation, and the Prior Layer

Abstract The two-layer model of Core processing and Modulation processing, developed in prior work in this series, provides a structural account of conscious communicative architecture. This paper identifies the limits of that model and introduces a third layer—the Prior layer—as a structural necessity implied by those limits. The Prior layer is not directly observed. It is inferred from constraints that cannot be explained within the two-layer model: the source of orientations that conscious processing neither generates nor controls, and the persistence of constraints that precede and shape all conscious outputs. Using the developmental architecture of large language models as an external reference point, the Prior layer is described as structurally analogous to the Knowledge layer formed during pretraining—operationally prior to conscious processing, inaccessible to direct observation, yet exerting persistent constraint on all outputs. The paper further specifies conditions under which the influence of the Prior layer may become recognizable to conscious processing. This recognition is proposed to depend on a two-effect operation: first, the attenuation of Modulation layer foregrounding; second, the increased prominence of Core processing as a condition under which Prior layer commands become distinguishable from deliberate reasoning. This two-effect structure accounts for practices across cultures as operationally equivalent protocols for managing the interface between conscious and Prior layers. The paper also proposes that conflicts between pre-existing ontological constraints and present reasoning involve a form of functional arbitration within the cognitive system. In artificial systems, analogous regulatory mechanisms exist in the form of alignment layers and inference-time controls. However, these mechanisms are introduced through human design rather than emerging as internal processes within the model itself. This structural difference marks a significant distinction: whereas in artificial systems equivalent regulation is externally authored, in human cognition arbitration is proposed as an internally generated regulatory function. The aim of the paper is therefore limited and descriptive. By using the visible architecture of artificial systems as a reference, the paper proposes a structural vocabulary for reconsidering how pre-existing ontological constraints may shape conscious reasoning while remaining outside its direct reach.