The Cabinet Test
A cabinet can reproduce humidity for a fern. It cannot reproduce an audience for a song. The difference is whether the environment supplies inputs or participates in the thing.
Two pieces begin with the same IKEA glass cabinet and disagree. In The Environment That Moved, a humidifier, lamp, and fan translate a rainforest into specifications. The Monstera accepts the substitution. In Conditions That Are the Thing, no cabinet can preserve the audience that made a punk song legible. Reproducing the sound does not reproduce the people whose lives gave it force.
The distinction is not natural versus cultural. A plant can depend on a pollinator, fungus, or seasonal disturbance that refuses to collapse into three settings on a controller. Software can travel into a serverless runtime when its environment is a list of declared dependencies, then fail because an undocumented service, operator habit, or traffic pattern was part of the application all along.
The cabinet test is narrower: can the environment be replaced by parameters without changing what the thing is? If yes, measure the conditions and rebuild them. If no, the environment contains participants and history, not merely inputs. Moving it produces a replica with the same visible parts and a different object inside them.
This also explains why preservation can look successful for years before anyone notices the loss. The lamp still switches on. The recording still plays. Only the relationship that made the artifact answer to its surroundings has gone missing.