Here is the opening of the Lord's Prayer in Old English, written about a thousand years ago: "Fæder ure þu þe eart on heofonum." It means "Our father, who art in heaven." Every language changes, all the time, and over enough centuries the changes add up to a new language. This level looks at how languages change and how linguists trace their family trees.

Change is normal

Sounds change, meanings change and grammar changes. Meanings drift in surprising directions. The word "nice" once meant foolish, and "silly" once meant blessed or innocent. "Meat" once meant food of any kind, which is why an old phrase like "meat and drink" covers a whole meal.

Grammar changes too, often through grammaticalization, the process by which a full word becomes a grammatical marker. "Going to" once meant only physical movement. It became a way to mark the future ("It's going to rain"), and in speech it is shrinking to "gonna."

Every generation hears change as decay, yet no language has ever decayed into something unusable.

Jones and the hidden family

In 1786 Sir William Jones, an English judge serving in Calcutta, gave a lecture that became famous. He said of Sanskrit, the classical language of India, that it bore to Greek and Latin "a stronger affinity, both in the roots of verbs and in the forms of grammar, than could possibly have been produced by accident; so strong indeed, that no philologer could examine them all three, without believing them to have sprung from some common source, which, perhaps, no longer exists." Jones argued that Sanskrit, Greek and Latin sprang from a common source that might no longer exist. Others had noticed resemblances before him, but his statement launched a century of research.

The comparative method

Resemblance alone proves little. Persian bad and English bad mean the same thing and are unrelated. What counts is regular sound correspondence. Regular sound correspondences, not chance resemblances, are the evidence that languages are related. In the early nineteenth century the Danish scholar Rasmus Rask and the German Jacob Grimm described one of the best known patterns. Grimm's law describes how certain consonants shifted in the Germanic languages, such as p becoming f. Where Latin has p, English and its Germanic relatives usually have f: pater and father, piscis and fish, pes (stem ped) and foot. Where Latin has t, English often has th: tres and three, tu and thou.

By lining up hundreds of such correspondences, scholars reconstructed much of an unwritten ancestor called Proto-Indo-European. Reconstructed forms are marked with an asterisk to show they are inferences, not recorded words.

The Indo-European family

The family that grew from that ancestor stretches from Iceland to India. Its branches include Germanic (English, German, Swedish), Italic (Latin and its descendants, the Romance languages), Celtic, Slavic, Baltic, Greek, Armenian, Albanian and Indo-Iranian (Hindi, Bengali, Persian), along with extinct branches such as Hittite. Nearly half of the world's people speak an Indo-European language as their first language.

Where was the original homeland? The steppe hypothesis places it in the grasslands north of the Black and Caspian Seas, perhaps 6,000 years ago, with speakers spreading partly through horse herding and wheeled vehicles. Ancient DNA studies since 2015 have found large migrations out of the steppe that fit this view. The rival Anatolian hypothesis holds that the languages spread from what is now Turkey with early farming, several thousand years earlier. Some recent studies propose a mix of the two. The question is not settled.

Other families

Indo-European is one family among many. Others include Sino-Tibetan (Mandarin, Tibetan, Burmese), Afro-Asiatic (Arabic, Hebrew, Amharic, Hausa), Niger-Congo (including the Bantu languages, such as Swahili and Zulu) and Austronesian, which spread by sea from Taiwan to places as far apart as Madagascar, Hawaii and Easter Island. A language isolate is a language with no proven relatives, such as Basque. The comparative method reaches back perhaps eight to ten thousand years, so proposals linking major families remain controversial.

Worked example: a cognate or a coincidence?

Latin habere means "to have," and it looks a lot like English have. Surely they are related? Linguists say no, and the reason shows how the method works.

Start with Grimm's law. One part of it says that where Latin has k (written c), Germanic languages usually have h: Latin centum and English hundred, Latin cornu and English horn, Latin canis and English hound. So if English have has a Latin relative, that relative should start with c, not h. And there is one: capere, "to take, to seize." The meanings are close, since taking leads to having, and the sounds match the regular pattern.

What about habere? By the same rule, an English relative would have to start with a different sound, and many scholars link habere instead to the family of English give. The look-alike pair, have and habere, turns out to be a coincidence, like Persian and English bad. The real relatives look less alike but follow the rules. That is the comparative method in miniature: trust the system of correspondences, not the surface resemblance.