The hypothesis test taught in every introductory course is two rival frameworks spliced together by textbook writers after their authors stopped negotiating. The seam is visible in ordinary use, and the thing it hides is who holds each error.
A number that follows from a definition, a number settled by measurement, and a number that depends on an unstated choice get printed to the same decimal place. Asking what more effort buys separates all three.
Agreement on the text does not produce agreement on the rule for reading it. Textual canons and theology show that disputes often begin after the primary record has been fixed.
A calendar can contain an observed feast day, a legal date and a settlement convention that move differently. Calendar rules and festival records make the authority behind each date part of the datum.
A category code can identify a recurring pattern while saying almost nothing about the particular observation that received it. Folktale indexing keeps accession and type numbers separate; incident systems should too.
A named era makes a gradual or uneven transformation look like a single dated break. Chronology and periodization provide the apparatus for testing whether the boundary belongs to the evidence or to the analyst.
A single start date often compresses conquest, incorporation, succession and current legal form into one number. Dynasties and kingdoms keep the event that supplies the date in view.
Two objects can remain neighbours under a representation while the distances, angles and apparent densities are wrong. Topology supplies invariants; non-Euclidean geometry supplies the metric that the drawing hid.
A total can change because the world changed or because membership was redefined. Set theory made the membership rule explicit because unrestricted collection creates contradictions; data universes need the same discipline.
Digits can travel intact through a system while base, notation, unit or scope changes underneath them. Number systems and notation history separate the value, its representation and the convention that tells a reader how to recover it.
A formal label can preserve rank, precedence and a chain of usage after the authority behind it has moved elsewhere. Heraldry and office records make the entitlement and the name separate fields.
When the reporting unit changes, aggregation can manufacture a discontinuity even if every underlying place changed smoothly. Regression can quantify the apparent effect; historical geography says to repair the unit first.
A battle total can mix killed, wounded, missing, civilian, later and estimated losses under one number. The record is a category system before it is an estimate.