On 2 February 1786, a British judge in Calcutta named William Jones gave a lecture that is often treated as the birth of modern linguistics. He had been studying Sanskrit, the ancient language of Hindu scripture, and he noticed that its words and grammar looked too much like Greek and Latin to be chance. They must, he said, have "sprung from some common source, which, perhaps, no longer exists."
He was right, though he was not the first to suspect it, and he also wrongly lumped in Egyptian, Japanese and Chinese. The common source is now called Proto-Indo-European. Nobody ever wrote it down. It was spoken roughly 5,000 to 6,000 years ago, and over the following millennia it split into the languages now spoken by about 3.4 billion people, some 42% of humanity. English is one small twig on that tree.
Pick a word. Every branch of the family below shows what that word is in some of its languages. Green means the word is inherited from the same ancestral word, a cognate. Amber means that language replaced the old word with a different one somewhere along the way, which happens all the time.
same ancestral word a different word took over
Three things are worth noticing as you click around.
The word for "three" survives in every language here. Numbers and close family words are some of the most stable words in any language, which is exactly why linguists use them as evidence.
"Father" gets replaced a lot. Russian otéts, Polish ojciec and Albanian atë all come from a different ancient word, a child's "daddy" word that became the normal term. Welsh tad is a similar nursery word. Families keep their bones but swap their clothes.
Cognates can look nothing alike. Armenian hayr is the same word as English father. The p at the start became an h in Armenian and an f in English, and the vowels and endings drifted. The match is in the pattern, not the look.
Linguists don't trust words that look alike. They trust sounds that change in the same way, again and again.
Why does English have f where Latin has p? Because at some point in the first millennium BCE, the ancestors of the Germanic languages shifted a whole set of consonants at once. Rasmus Rask spotted the pattern in 1818, and Jacob Grimm (yes, of the fairy tales) set it out in 1822. It was the first time anyone showed that sound change is regular rather than random.
| Shift | Latin keeps the old sound | English shifted it |
|---|---|---|
| p → f | pater, pēs, piscis | father, foot, fish |
| t → th | trēs | three |
| k → h | centum, cornū, canis | hundred, horn, hound |
| d → t | decem, duo, dēns | ten, two, tooth |
| g → k | genū | knee (the k was once spoken) |
Latin c was always pronounced k. Many English words that look like the Latin ones, such as paternal, century and dental, were borrowed from Latin much later, so they never went through the shift. That is why English has both tooth and dental.
The flip side: some words look related and aren't. English have and Latin habēre ("to have") look and mean the same, but they come from two different ancient roots. English have is actually related to Latin capere, "to take," which fits Grimm's Law (k → h). And Persian bad means "bad," but the two words have completely separate histories. With thousands of words and a limited set of sounds, some coincidences are guaranteed.
Being spoken in Europe does not make a language Indo-European, and being spoken in Asia does not rule it out. Test yourself.
This part is genuinely debated. The most widely accepted answer is the steppe hypothesis: Proto-Indo-European was spoken by herding peoples on the grasslands north of the Black and Caspian Seas, roughly 4000 to 3000 BCE, who then spread with horses and wheeled wagons. The reconstructed vocabulary fits that picture. It has words for wheels, wagons, cattle, horses and snow. Since 2015, ancient DNA studies have found large population movements out of the steppe at about the right time, which strengthened the case.
The main rival, the Anatolian hypothesis, placed the origin in what is now Turkey around 8000 BCE, spreading with early farming. It has lost most of its support. Some recent genetic work suggests an even older, earlier stage of the language may have begun south of the Caucasus, with the steppe as a later staging ground. This debate is still live, so treat any confident map with caution.
Indo-European is only one family. Ethnologue counts about 7,164 living languages, and linguist Lyle Campbell counts some 406 separate families once you include isolates, languages with no known relatives at all, like Basque. Indo-European has the most speakers, but not the most languages. Niger–Congo in Africa and Austronesian, which stretches from Madagascar to Hawaii, each contain well over a thousand.
Some families you may meet as a learner: Sino-Tibetan (Mandarin, Cantonese, Tibetan, Burmese), Afroasiatic (Arabic, Hebrew, Amharic, Hausa), Uralic (Finnish, Estonian, Hungarian), Turkic (Turkish, Kazakh, Uzbek), Dravidian (Tamil, Telugu, Kannada), Austronesian (Indonesian, Tagalog, Māori), and Japonic and Koreanic, which are each small families of their own.
Most languages of Europe, Iran and northern India descend from one unwritten ancestor, Proto-Indo-European, spoken roughly five to six thousand years ago. The evidence is not words that look alike but sounds that change in regular, predictable ways, like the p → f shift that turned pater's cousin into father. Families keep their core words (numbers, family terms, body parts) longest, swap others freely, and are only one of several hundred families on Earth.
Try this: if you are learning another Indo-European language, take ten basic words (numbers, body parts, family) and look for the English cousin of each. Spanish pie and English foot, German Zahn and English tooth: each one you spot is a word you've half-learned already.
Sources. Every word and cognate verdict in the explorer was checked against the etymologies in Wiktionary, which cite standard etymological dictionaries (de Vaan for Latin, Beekes for Greek, Derksen for Balto-Slavic, the Indo-Iranian EWAia, and others). Speaker figures (over 3.4 billion first-language speakers, 42%) are from Wikipedia's "Indo-European languages." William Jones's 1786 "Third Anniversary Discourse" and its errors are from Wikipedia's article on him. Grimm's Law dates (Rask 1818, Grimm 1822) are from Wikipedia's "Grimm's law." The homeland debate summary follows Wikipedia's "Proto-Indo-European homeland" (steppe theory most widely accepted; Anatolian hypothesis largely abandoned). Language and family counts: Ethnologue 27th edition (2024) and Lyle Campbell, "How many language families are there in the world?" (2019).