Testing Global Separability Across the Sciences: The Ontological Limits of Objectivity, Neutrality, and Mastery

A foundational premise in modern scientific practice is that systems can be treated as separable—isolated without constitutive loss—so that knowledge can be abstracted from context for reliable prediction and intervention. This premise is closely tied to the enduring ideals of objectivity, neutrality, and perspective-independent truth (hereafter O1–O3), even as many philosophers of science and methodologists argue that a literal “view from nowhere” is unattainable. Yet across the sciences, many of the most robust explanations depend on constitutive relations and embeddings that cannot be removed without loss. This paper conducts a systematic cross-disciplinary synthesis and introduces the Relational Necessity Test (RNT)—a falsifiable protocol with named falsifiers and explicit decision rules—by applying a straightforward evaluative criterion: for each domain, ask whether the field’s best-supported explanations can be reconstructed without their candidate constitutive relations/embeddings while still meeting that field’s standards of explanatory adequacy, tracked by pre-stated signatures (predictive fit, invariants, identity conditions, and persistence under re-embedding/scale); a single domain that met these criteria would suffice to vindicate the separability premise. In the present audit, across Cosmology & QFT, Thermodynamics, Chemistry, Biology & Ecology, Cognitive & Neuroscience, and Earth System Science, together with meta-evidence from theory change, no domain does; instead, separability-preserving reconstructions lose adequacy, while relation-dependent accounts retain it. No separability-preserving rival has yet been articulated that matches this cross-domain evidential spread, while also yielding distinctive, independently confirmed predictions. On the present evidential record, separability, therefore, appears as a limiting approximation rather than the structure of reality. Consequently, O1–O3 cannot function as global ontological commitments, although they may remain scope-limited methodological ideals. The paper closes by deriving from this ontological verdict a minimal, auditable standard for responsible intervention in coupled systems.