Comparative Method

Look for consistent, regular sound correspondences across multiple word pairs. If English 'father' and German 'Vater' are cognates, then other English /f/ sounds should regularly correspond to German /v/ sounds (e.g., English 'five', German 'fünf' — actually /f/ corresponds to /f/, showing a partial regularity). The more pairs that follow the same pattern, the stronger the evidence. Also check meaning: related words should have related meanings, not arbitrary similarities.

The method is better at supporting relationships than disproving them. However, if you find no regular, numerous correspondences despite careful comparison, and if the few similarities can be explained by chance or borrowing, you can reasonably conclude non-relationship. The absence of evidence is not absolute proof, but it weakens confidence in genetic connection.

The Comparative Method compares multiple related languages to infer their shared ancestor. Internal Reconstruction works within a single language, finding internal evidence (like irregularities or morphological alternations) that reveal the language's own historical structure. They are complementary: use comparative when you have relatives; use internal reconstruction when you want to reveal the earlier history of a single language.

Reliably, about 6000 years. Beyond that, sound changes accumulate, cognates become obscured or lost, and the 'noise' of chance similarities grows. Some scholars debate whether very distant relationships (like proposed macro-families linking Indo-European, Uralic, and Altaic) can be demonstrated; most mainstream linguists are skeptical of claims beyond 6000–10,000 years without extraordinary evidence.