1786 · The spark
Sir William Jones notices something strange
A British judge in Calcutta, studying Sanskrit, saw that it looked uncannily like the Latin and Greek he knew from school — not in a few borrowed words, but in its verb roots and grammar. His conclusion, read to the Asiatic Society, founded a science:
“The Sanscrit language … bears a stronger affinity to both the Greek and the Latin 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.”
That “common source which no longer exists” is what we now call Proto-Indo-European.
The core evidence
Family words match across continents
The words least likely to be borrowed — mother, father, brother — line up sound for sound from Ireland to India:
| Sanskrit | Ancient Greek | Latin | Old Irish | English | |
|---|---|---|---|---|---|
| *ph₂tḗr | pitā́ | patḗr | pater | athair | father |
| *méh₂tēr | mātā́ | mḗtēr | māter | máthair | mother |
| *bʰréh₂tēr | bhrā́tā | phrā́tēr | frāter | bráthair | brother |
1822 · The proof
Grimm's Law: sound change is regular
Jacob Grimm (of the fairy tales) showed the resemblances weren't coincidence but law. Wherever Latin and Greek have p, t, k, the Germanic languages systematically have f, th, h — every time, across thousands of words:
| Shift | Latin / Greek | English |
|---|---|---|
| p → f | pater · pēs/pedis · piscis | father · foot · fish |
| t → th | trēs · tū · tenuis | three · thou · thin |
| k → h | centum · cornū · canis | hundred · horn · hound |
Regularity is what makes reconstruction possible: if changes follow laws, you can run them backwards. This is why an English f reliably points to a PIE *p — try five (Greek pénte) or wolf.
The stable core
Numbers barely change in 6,000 years
Numerals are learned early, used daily, and almost never borrowed — so they preserve the old system beautifully:
| PIE | Sanskrit | Greek | Latin | Lithuanian | English |
|---|---|---|---|---|---|
| *dwóh₁ | dvā́ | dúo | duo | du | two |
| *tréyes | tráyas | treîs | trēs | trys | three |
| *septḿ̥ | saptá | heptá | septem | septyni | seven |
| *déḱm̥ | dáśa | déka | decem | dešimt | ten |
| *ḱm̥tóm | śatám | hekatón | centum | šimtas | hundred |
The great split
Centum and satem
Watch the word for hundred, *ḱm̥tóm, in the table above. The western branches kept the hard k-sound (Latin centum, pronounced kentum); the eastern branches softened it to s-sounds (Sanskrit śatám, Avestan satəm). This single sound choice cleaves the whole family into a western “centum” group (Italic, Celtic, Germanic, Greek) and an eastern “satem” group (Indo-Iranian, Balto-Slavic, Armenian) — early evidence that the family tree had real, datable branches. See the hundred tree.
1879 → 1927 · The confirmation
Saussure predicts sounds no one had ever heard
At 21, Ferdinand de Saussure noticed that PIE roots misbehaved in patterned ways, and proposed the fix: the proto-language must have had extra consonants — now called laryngeals (*h₁ *h₂ *h₃) — that had vanished from every known daughter language, leaving only distortions in neighbouring vowels.
It was pure algebra, and many dismissed it. Then Hittite — recorded on clay tablets in Anatolia around 1600 BCE and deciphered in 1917 — turned out to preserve some of those very sounds, written ḫ, exactly where Saussure's equations demanded them: PIE *h₂ents- “front” appears as Hittite ḫanza. A sound predicted on paper in 1879 was dug out of the ground four decades later. It remains the most celebrated confirmed prediction in linguistics — and it's why the roots on this site are written with those h₁ h₂ h₃ subscripts.