Causal claims of the form "aspirin reduces fever" systematically conceal the conditions required for the claim to hold. Aspirin sitting in a bottle does not reduce fever. A structured configuration of material conditions---including fever itself, an intact biochemical pathway, and an unbroken supply chain---produces the outcome. What gets named as "the cause" is simply the component the investigator chose to manipulate. This paper argues that causation is always the product of such configurations, here called causal containers, and that the traditional distinction between cause and background condition should be replaced with a continuum of substitutability. The cause is not a metaphysically privileged component. It is the least substitutable component within the explored pathway space. Drawing on Gibson's theory of affordances, a distinction is developed between perceiving that an intervention works---sufficient for action---and understanding enough of the surrounding configuration to know whether it will work elsewhere. The randomised controlled trial is reconceived as a container optimised for the first kind of perception at the cost of the second. The position that emerges is architectural realism. Causal structure is real and independent of its descriptions, but navigable by incommensurable research programmes that each track enough of the material reality to enable successful intervention. Causal labelling, however, does more than organise investigative attention. It allocates responsibility and structures ignorance. Components necessary for the widest range of outcomes are systematically the least likely to be named as causes, funded as research, or assigned accountability---an outcome that is structurally incentivised, not incidental. Implementation science, understood in these terms, is the problem of assembling working configurations, not faithfully reproducing named interventions.
Daniel D. Reidpath (Sat,) studied this question.