Historical analysis

Define the claim

A historical claim should name a form or construction, the proposed change, the branch or community, and the time depth. “These languages are similar” is too loose to test. “This consonant correspondence recurs in inherited words of two branches” has observable consequences. Distinguish genealogy, borrowing, semantic change and script adoption; each requires different evidence. Include the possibility that the available data cannot decide.

Build a source sheet before drawing arrows. For each token or dictionary form, record the cited edition or comparative entry, writing system, transcription convention, grammatical form, meaning in context, date and confidence. Keep reconstructed protoforms on a separate sheet with their proposed node and author. Note which source first proposed an analysis and which later source merely repeats it.

Test and revise

Align a group of forms and formulate correspondence rules from several sets. Test those rules on sets held aside. Check a proposed morphological innovation against a whole paradigm, not a convenient isolated cell. For a loan proposal, require a plausible donor, route and chronology and compare the expected inherited reflex. For a syntactic or semantic change, seek bridging contexts and control for genre. If a counterexample survives, revise the environment, candidate set or confidence level in public view.

The Indo-European father set and Austronesian *lima set are useful models for this discipline, but neither single set is a proof of its whole family. Their source entries supply forms and editorial choices; the explanatory weight comes from the larger network of correspondences. End with a claim no stronger than that network allows, and cite the particular entries that carry it.

Research exercise

Prepare a short historical dossier on one published cognate set. Include at least four forms from three branches, one correspondence recurring elsewhere, an explicit reconstructed form only if the source supplies it, a contact alternative, a relative chronology if supported, and one unresolved issue. Mark every form as attested, normalized or reconstructed. A successful conclusion can be “the relation remains uncertain” if that follows from an honest test.