Who Governs What Persists? Retention, Durable Updating, and Continuity Across Cognitive Systems

This preprint develops a conceptual framework for analyzing how control over cognitive continuity is allocated when cognition, memory, evaluation, and structural change are distributed across multiple carriers and extended across time. Research on distributed cognition has shown that cognitive work need not be confined to a single organism or agent. It may depend on interacting persons, artifacts, institutions, technical infrastructures, records, and historically accumulated practices. However, the distribution of cognitive work does not by itself determine who or what controls the conditions under which the wider cognitive organization continues. This work therefore shifts attention from the distribution of cognition to the regulation of persistence. The article distinguishes three analytically separable continuity-governing functions: 1. Retention selection -- control over which distinctions remain persistently available and consequential for later cognition.2. Durable updating -- control over which changes modify the conditions governing future processing, evaluation, coordination, or transformation.3. Continuity classification -- consequential evaluation of whether a transformation preserves, disrupts, or replaces the relevant cognitive organization under a declared comparison basis. The framework differentiates the continuity of a focal carrier from the continuity of the wider cognitive organization in which that carrier participates. Continuation of a prominent component, institution, platform, model lineage, infrastructure, or governance arrangement does not by itself establish continuation of the wider organization. Conversely, alteration or replacement of a focal carrier does not by itself establish organizational discontinuity. The article introduces the concept of cognitive-continuity capture: a self-reinforcing governance condition in which continuation of a focal carrier becomes operationally privileged as a proxy for the continuity of an independently bounded wider cognitive organization. A supported diagnosis requires the conjunction of six conditions: an independently specified focal/wider relation, operationally privileged continuity substitution, leverage spanning retention selection, durable updating, and continuity classification, consequential coupling among these functions, material displacement of independent continuity pathways, and at least one completed temporal feedback loop that strengthens the carrier’s later capacity to reproduce its privileged position. Five propositions specify the framework’s central claims: the allocation of continuity-relevant control is not determined by the distribution of cognitive processing; focal-carrier persistence is insufficient evidence of wider continuity without warranted bridge conditions; locally accepted updates need not compose into a continuity-preserving trajectory; capture is a conjunctive and temporally structured condition; and formal separation of governance roles provides protection only when it produces effective functional independence. The framework is developed across biological, institutional, social, and technical domains without attributing consciousness, personhood, unified agency, or intrinsic interests to distributed cognitive organizations. Artificial intelligence is treated as a conditional stress test rather than as the necessary or primary realization of the theory. The article also distinguishes cognitive-continuity capture from autopoiesis, organizational identity, structural inertia, path dependence, regulatory capture, mission drift, principal–agent failure, power seeking, interruption resistance, and compositional drift. Institutional cases and a stipulated persistent artificial-agent case are used to clarify boundary conditions, operational indicators, evidential requirements, and possible empirical designs. The article concludes with conditional procedural implications for contexts in which independent assessability, corrigibility, and contestability are accepted aims. This preprint presents the complete conceptual, diagnostic, and operational development of the framework. Subsequent journal-oriented versions may develop selected parts of the argument in a shorter form.