Family Trees for Words: How Linguists Map Language History
The English word "night" looks almost nothing like the Latin word "nox." But say them side by side with German "Nacht," Dutch "nacht," and Sanskrit "naktam," and a pattern jumps out. All five words trace back to a single ancestor word, *nรณkสทts, spoken by people who lived roughly 6,000 years ago and left no writing behind. Linguists reconstructed that word using the same logic biologists use to build family trees for species -- and today they use computer algorithms borrowed straight from evolutionary biology to do it faster and more precisely.
What You'll Learn
- How linguists prove two words in different languages share a common ancestor - What "sound laws" are and why they let us predict word forms - How computational phylogenetics builds language family trees the same way biologists build species trees - Why a 2003 computer study changed how we date the origin of English
Words Have Ancestors Too
A cognate is a word in one language that shares a direct ancestor with a word in another language, the way cousins share a grandparent. English, German, Latin, Greek, Russian, and Hindi all descend from a single unwritten language linguists call Proto-Indo-European (PIE), spoken by nomadic herders somewhere near the Black Sea around 4500-2500 BCE. Just as a biologist can compare DNA in a fox, a wolf, and a dog to figure out how closely related they are, a linguist compares sounds and word structure across languages to figure out how closely related they are -- and how long ago they split apart.
An unwritten, reconstructed ancestor language. No PIE text has ever been found -- linguists rebuilt roughly 1,500 of its root words purely by comparing its modern descendants, spoken today by nearly half the world's population.
Cognates and Sound Laws: The Clues Linguists Use
In 1822, Jacob Grimm (yes, of the fairy tales) noticed something strange: wherever Latin had a "p" sound, English tended to have an "f" sound in the same word slot. Latin pater -> English father. Latin pes/pedis -> English foot. Latin piscis -> English fish. This isn't a coincidence -- it's a rule, now called Grimm's Law, describing exactly how certain consonants shifted as the Germanic branch split off from the rest of the Indo-European family roughly 500 BCE. Sound laws like this are the fingerprints linguists use to prove two words are cognates rather than coincidental look-alikes.
From Manual Trees to Computer Trees
For 200 years, linguists built language family trees by hand, comparing word lists one pair at a time -- slow, and open to disagreement. In 2003, researchers Russell Gray and Quentin Atkinson borrowed a tool from disease-outbreak tracking called Bayesian phylogenetics: the same statistics doctors use to trace how a virus mutates and spreads through a population. They fed the computer a database of 87 languages and 2,449 cognate word sets, and let the algorithm calculate the most likely family tree and the most likely dates for each branching point. The computer's answer: Indo-European began splitting apart around 8,000-9,500 years ago in Anatolia (modern Turkey) -- thousands of years earlier than most linguists had assumed, and tied to the spread of farming rather than horse-riding herders. The debate over which theory is right is still active today.
Flashcards โ click each card to reveal the answer
Case Study: Counting to Three Across the Family
Numbers are some of the most stable words in any language, which makes them perfect for spotting family resemblances. Compare "three": English three, German drei, Latin tres, Greek treis, Sanskrit trayas, Russian tri. Every single one starts with a t/dr/tr sound and ends in a vowel-plus-s pattern. That is not a coincidence across six languages spanning from Ireland to India -- it is the fingerprint of one ancestor word, *trรฉyes, inherited and gradually reshaped by each daughter language's own sound laws.
According to Grimm's Law, if a Latin word has a "p" sound, what sound would you expect in the corresponding English cognate?
What made the Gray and Atkinson 2003 study different from 200 years of prior Indo-European research?
Build a Five-Word Family Tree
Pick five languages (at least one must be English). Look up the words for "mother," "three," "is/be," "star," and "new" in each language using a dictionary site or Wiktionary. Write each set of five translations in a row, then circle any sounds that repeat in the same position across languages (like the p/f pattern in this lesson). Sketch a simple branching tree diagram showing which of your five languages look most closely related to each other based on your circled patterns, and label one guess for why.
Want to keep learning?
Sign up for free to access the full curriculum โ all subjects, all ages.
Start Learning Free