The comparative method and natural experiments

History does not run controlled experiments, but sometimes it runs something close enough. This page gathers conjectures built on accidental pairs: two cases alike in almost everything except the one factor a researcher actually cares about, so the difference between them can be read as a real effect rather than a coincidence. That is the same logic behind anthropology's controlled comparison and an economist's natural experiment, and the pre-print record turns out to hold real candidates for it: two traditions watching the same sky under different conditions, a paper shortage that struck one region and spared its neighbour, naval battles measured against land battles fought by the same side, even a royal decree that changed a script and thereby ran an experiment nobody planned.

Lanchester at Salamis applies a mathematical model built for modern combat losses to compare naval and land battles as two parallel test cases for the same theory. Two copying regimes, two curves and The decreed bestseller both treat a single administrative decision -- two coexisting copying regimes, one top-down decree -- as an accidental natural experiment whose "before" and "after" can be compared directly. The twin page slumps compares a manuscript's two halves, treating one page as the control against which the other's condition can be measured.