This paper develops an interpretive typology of horizons within a consistency-history framework that distinguishes direct causal access, indirect reconstructive access, recoverability, and public finality. Its central claim is deliberately limited: a horizon does not create, destroy, or finalize a record. It constrains the direct causal channels through which signals and records can reach an observer confined to a worldline or causal patch, and therefore constrains public finality only derivatively. Three horizon types are distinguished. For an evaporating black hole, direct causal access across the horizon is not restored, but in controlled island and quantum-extremal-surface models an interior region becomes reconstructible from the Hawking radiation after the Page transition, with qualifications. This is described as delayed reconstructibility rather than redundant public finality. For an eternally uniformly accelerated observer, a Rindler horizon produces observer-relative causal exclusion: new signals from the causally inaccessible region do not reach that worldline, although no corresponding exclusion holds for suitable inertial observers and indirect records may remain available. For a worldline in a cosmology with finite future conformal time, a cosmological event horizon produces asymptotic causal exclusion of direct signals. Whether beyond-horizon information is reconstructible from static-patch or horizon degrees of freedom remains open. The paper emphasizes that these cases are not physically identical. Their common feature is only that each limits direct causal access within a specified observer-accessible region. The cosmological interpretation is explicitly weaker and more conjectural than the black-hole and Rindler analyses. The framework infers inaccessibility rather than destruction, makes no claim that information is fundamentally lost, derives no horizon entropy, temperature, or information-recovery law, and makes no positive or predictive claim about dark energy. The paper is interpretive and conceptual, has L2 status within the framework’s claim discipline, and makes no independent empirical prediction.
Tomoyuki Uchida (Thu,) studied this question.