The regularity of sound change

node

The empirical finding on which every family tree ultimately rests: when a sound changes in a language, it changes in every word where it occurs in the same phonetic environment, not word by word. A speech community does not decide to alter one word; an articulation drifts, and the drift applies wherever that articulation appears. The consequence is what makes the comparative method work. Relatives do not resemble each other in scattered vocabulary — they differ from each other by rules, and the rules generate predictions in words nobody has yet looked at. The doctrine was stated at its sharpest by the Leipzig Neogrammarians in 1878, in the formula that sound laws admit no exception, and even linguists who reject the doctrine in its absolute form work as though it were true, because apparent exceptions have so consistently turned out to be conditioned by something not yet noticed — a following consonant, a syllable position, or in Verner's case the position of an accent that the language had since lost. Regularity is a claim about mechanism, not merely a convenience: it is why a demonstration of relatedness can fail, and therefore why it can succeed.

A node is not a place. Drawing it on a map would assert something about the world that no stored fact supports.