Why Observation Cannot See the Whole Observation, Self-Reference, and the Limits of Internal Representation

We propose an axiomatic framework for variational analysis in which a single scalar functional E is fixed independently of the choice of representation. After an admissible repre- sentation is chosen, the field variables admit a decomposition into a retained component and a complementary component. The basic non-exposure axiom states that the complementary component is not itself a retained variable; it may influence the retained equations only through declared summary maps or induced operator corrections. The paper proves three central structural theorems, supported by a preparatory invariance theorem. The first central theorem is a quotient-descent statement: when the complementary sector enters the density only through the declared summary F = DI II , the functional descends to Γ∞(EI )/ ker(DI ), kernel directions are variationally silent, and the comple- mentary Euler–Lagrange expression factors through D∗ I . The second is a scoped reduction theorem: within the declared semilinear first-derivative scope and a controlled equivalence class generated by four explicit moves, an admissible density reduces to a coercive quadratic leading part plus a single summary-mediated bilinear interface, with all remaining terms in an admissible lower-order sector (a sufficiency statement, not an absolute uniqueness claim). The third is a three-construction compatibility theorem: the same fixed E supports three stan- dard variational formulations associated with one density — the stationary Euler–Lagrange formulation, the dissipative L2-gradient-flow formulation, and a conservative hyperbolic time-extension formulation — with R1 criticality coinciding with R2 fixed points on the same declared variation subspace, and with retained-sector R1 criticality coinciding with time-independent retained-sector R3 solutions, while R2 is dissipative and R3 is conservative under the usual time-translation hypothesis. The supporting principal-class invariance theorem records that the same four equivalence moves preserve the principal variational class across R1, R2, R3, securing the consistency of the equivalence relation used by the reduction and compatibility theorems. Energy conservation is therefore formulation-dependent, not a primitive property of E. Supporting these are: a weak Euler–Lagrange template for local first-derivative functionals, explicit Euler–Lagrange formulas for the canonical representative, an absorption criterion for the bilinear interface, and an algebraic consistency check showing that eliminating the complementary component yields a bounded retained-side correction that introduces no new differential derivative order; under the ellipticity hypothesis, the pseudodifferential principal symbol is preserved. No spectral, geometric, or physical interpretation of this elimination is made here; such interpretations belong to specialized later work. No unification theorem, physical prediction, or application-specific result is claimed.