How far back can language comparison reach?
Proposed ancestors of very old language families reach beyond surviving texts. Where can comparison still test a claim, and where does it become a hypothesis?
The limits of language comparison
Campbell and Poser's Language Classification: History and Method reviews how language relationships have been established and evaluates controversial proposals of distant relationship, including Nostratic and Proto-World. The research record keeps those macrofamily hypotheses outside its canonical ancestry paths. Cambridge University Press ↗
Confidence that falls with depth
The record stores Nostratic above Proto-Indo-European at confidence 0.2, Borean above Nostratic at 0.1, and Proto-World at 0.05. Each is labelled a speculative hypothesis. The numbers express how much the comparative method can support at that depth, not a vote against the idea.
A sound pattern or a look-alike word?
Ask what a proposed deep relationship rests on: regular sound correspondences and shared irregular morphology, or lists of look-alike words. The second can be produced by chance, borrowing, or sound symbolism.
When should a hypothesis enter a timeline?
Why does the record allow speculative macrofamilies to be retrieved only when a query explicitly asks for hypotheses? Think about what happens to a timeline if they are mixed in by default.