Experimental probes of autopoietic self-maintenance.

Autopoietic closure is a constitutive causal property of a living system: a recursively organized network produces the components and boundary conditions that regenerate that same network. Experiments can reveal, discriminate, or fail to detect this organization, but they do not create or define it. This article develops a categorical account of the relation between autopoietic organization and its experimental representation. Building on a topos-theoretic formulation of viability-localized self-production, the paper distinguishes three levels: the natural biological system, a formal organizational surrogate, and the observer-side profile generated by experimental protocols. A category of organizational models and a category of typed protocols are defined explicitly, together with a realization functor from protocols to formal test objects. The Yoneda lemma then characterizes formal models through all formal tests, while restriction to an experimentally available protocol family yields a partial response profile. Equivalence of restricted profiles establishes organizational equivalence only when that protocol family is conservative on a specified class of candidate models. This condition converts the idea of an adequate assay panel into a precise, model-relative criterion without conflating experimental accessibility with operational closure. The framework also distinguishes diagnostic from predictive uses, accommodates destructive assays, differentiation, environmental context, and one-off experimental histories, and limits its claim about emergence to local-to-global compatibility. A worked reaction-network example shows how perturbation profiles discriminate highly resilient from fragile self-maintaining organizations while the causal closure remains defined by the reaction network itself. The result is an experiment-oriented semantics for testing formal models of autopoietic self-maintenance, not an observational redefinition of autopoiesis.