The Correctness Norm and the Five Elements of a Code: A Structural, Origin-Blind Definition of Encoded Information

"Code" names anything from a Huffman encoding to the rings of a tree, and so demarcates nothing. This paper isolates what separates a code from mere physical correlation and proposes an origin-blind definition. Its central claim is that the watershed is a correctness norm: a fact making one reading right and another wrong, irreducible to covariation. I argue, via the disjunction problem and the underdetermination of decoding rules by channel statistics, that covariation does not fix such a norm, and that selected function — though it may fix a norm — does not by itself fix a separable one; separability is tested by asking whether the system can be read wrongly without its physics changing. Around this core I assemble four conditions from independent fields — discrete alphabet (Goodman), syntax (Morris), separable arbitrary mapping (Peirce, Saussure), independent interpretation layer (Turing, von Neumann) — and a fifth distinguishing a bare code (architecture) from an exercised one (architecture plus function). The definition is a calibrated detector with built-in epistemic asymmetry: it certifies non-codes readily, certifies codes conjunctively, returns "inconclusive" when warranted, and is falsifiable by a single counterexample exhibiting the architecture.