Why do languages change over time?

Why do languages change over time?

Who owns a language: a dictionary, a teacher, or the people speaking it today? If meaning lives in shared use, perhaps change is not an accident that happens to language but something speakers create whenever their habits shift.

Every generation learns language from imperfectly identical examples. Speakers shorten common expressions, extend old words to new situations, borrow terms from neighboring communities, and favor pronunciations that mark identity or belonging.

Small variations may disappear, but some spread through a network of speakers. Children can regularize patterns; adults can popularize slang; migration and trade can mix vocabularies. Over long periods, many small choices accumulate until older speech feels unfamiliar.

Does this mean change is random? Not entirely — and the nineteenth century found evidence far stronger than "not entirely." If pronunciations drifted word by word, you would expect a scatter: this word shifting, that one not, no pattern. What was found instead was a rule. Compare Latin pater with English father, Latin pēs with foot, Latin trēs with three, Latin cornū with horn, Latin centum with hundred. The same three sounds, p, t and k, have become f, th and h — not in a handful of words but across the whole inherited vocabulary of the Germanic languages. That regularity is Grimm's law, and its importance is not the particular sounds. It is that a sound change is not a set of decisions about words at all. It is a shift in how a sound is produced, and it therefore sweeps through every word containing that sound, whether anyone knows the words are related or not.

Honesty requires the sequel, which is better than the headline. Grimm's law had exceptions that nagged for decades, until Verner noticed they depended on where the accent had fallen in the ancestral word. The exceptions were themselves regular — which is what convinced linguists the regularity was real rather than a tidy approximation, though the strongest version, that sound change admits no exceptions whatever, is still argued over.

Sound change alone would leave a language a wreck of irregularities, since it does not care what grammar it destroys. So notice the second force, pulling the other way. Learners generalize. Faced with a pattern that mostly holds, they extend it to the cases that resist — which is why the past tense of help was once holp and is now helped, and why the plural of cow is no longer kine. This is analogy, and it is a repair mechanism. It also runs backward when a pattern is strong enough: dove for the past of dive is not ancient, it was built on the model of drive and drove.

A third mechanism is quieter and stranger. A learner hears only the surface and must guess the structure beneath it, and sometimes guesses differently from the speaker. Middle English a napron was heard as an apron, and the n changed sides permanently; an ewt went the other way and became a newt. Nobody erred, exactly — each listener drew the only line the sound stream supported. That is reanalysis, and it is how grammar changes rather than merely vocabulary.

Then there is contact. Words cross easily; English took thousands from Norman French. But contact can reach further than vocabulary: they, them and their are borrowings from Norse, and pronouns are supposed to be the part of a language least available for borrowing. Intense enough contact does not just add words, it reshapes the frame.

Language change is not simply decay. Earlier forms were themselves the result of still earlier changes. A form can be stigmatized in one setting and useful in another, while grammar continues to provide structure even as pronunciations and meanings move.

The decay story deserves a firmer answer than that, because it is the intuition almost everyone starts with, and it makes a prediction we can check. If change were erosion, languages would get simpler over time — and no language has been observed to do so overall. English lost nearly all its case endings, which looks like pure loss until you ask what took over their work: a rigid word order Old English did not need, a system of articles it largely lacked, an elaborate machinery of prepositions and auxiliary verbs. The complexity moved. Whether it is genuinely conserved is a live argument, and there is decent evidence that languages learned by many adults shed morphology faster than others — but the direction of travel is nothing like erosion. Every "decayed" form is somebody's future grammar.

Nor is a change a mistake merely because you happen to be alive while it is running. Every feature you consider correct was once an innovation somebody's grandparents disliked, and that complaint has been recorded for as long as we have records.

What about writing — does it not fix the language? It slows things, and it does so unevenly. English spelling was largely settled by printers while the Great Vowel Shift was still running, which is why the spellings preserve a pronunciation nobody now uses and knight keeps letters that stopped being sounded centuries ago. That is the pattern: writing anchors the visible form while the spoken language walks away from it, so literacy widens the gap between spelling and speech rather than closing it. Standardization can also slow the spread of a change by giving everyone a prestige target to imitate. What no academy has ever managed is to stop one.

• How accents split into dialects and languages, and why the line between the two is political.
• Why word meanings sometimes reverse, and why they so often drift toward the negative.
• How the comparative method reconstructs a language nobody wrote down.

Every term the collection defines is gathered in the glossary.