An instrument that agrees with expectation too quickly is worth distrusting. Calibration should reduce uncertainty, not flatter prior belief. When every replicate lands near the desired answer, contamination may be conceptual before it is chemical. The first control is to define what result would count as embarrassing.

A hypothesis is stronger when it names the world in which it would be unnecessary. Counterfactuals expose whether a mechanism is doing work or merely borrowing atmosphere. If the conclusion survives after removing the preferred story, the model may be portable. If it collapses, the narrative was load-bearing.