Skip to content
← Experiments

An interactive field guide·

Every word is a fossil.

Languages don't get invented. They drift, one sound and one generation at a time, until Latin becomes French and a steppe dialect nobody wrote down becomes English, Persian and Hindi. Here's the family tree, the map, and the rules.

*ph₂tḗr
4000 BCEProto-Indo-European
  2. →
  3. →
  4. →
  5. →

p → f (Grimm's law), t → d (Verner's law), then d → th in the 1500s, the same change that made modor into mother.

Chapter 1

Do European languages come from Persian?

No. They're cousins. Persian, English, Latin, Greek, Russian and Hindi all descend from one older language, Proto-Indo-European. Persian is a sibling in the family photo, not the grandparent.
Proto-Indo-European
about 4500 BCE · the steppe north of the Black Sea · never written
Indo-Iranian
  • Persian
  • Hindi
  • Kurdish
Greek
  • Greek
Italic
  • Latin
  • Spanish
  • French
Germanic
  • English
  • German
  • Swedish
Balto-Slavic
  • Russian
  • Polish
  • Lithuanian
Celtic
  • Irish
  • Welsh

Persian and English sit on different branches of the same tree. Their shared ancestor was spoken about 6,000 years ago. Old Persian was first written about 2,500 years ago, after Greek and Latin already were: they lived side by side, so one can't be the other's parent.

PersianEnglishLatinSanskrit
pedarfatherpaterpitā
mādarmothermātermātā
barādarbrotherfrāterbhrātā
nāmnamenōmennāma
setārestarstēllatārā
nownewnovusnava
sethreetrēstrayas
  • Why it feels true. In 1786 William Jones noticed that Sanskrit, Greek, Latin and Persian match too well to be chance and must share “some common source”. That source is the ancestor, not Persian.
  • Persian did lend words. English got paradise (a walled garden), bazaar, caravan, pajamas, jasmine and check (from shāh, as in chess) from Persian, mostly through trade. Borrowing, not ancestry.
  • Look-alikes can lie. Persian bad means “bad”, and it's a pure coincidence. Real relatives follow sound rules and often look nothing alike.

Chapter 2

The family tree, and where it walked

Every line is a language, from its first speakers to today (or its last). Dashed means nobody had written it down yet. Drag the year or press play: the map shows who was talking where, and where they came from.

English, Persian, Hindi, Russian, Latin, Greek: one family, about 6,500 years old, spoken natively by nearly half of humanity.

Word lens
  • Proto-Indo-European
  • Anatolian
  • Small branches
  • Indo-Iranian
  • Greek
  • Italic & Romance
  • Celtic
  • Germanic
  • Balto-Slavic
  • not yet written
4500 BCE2500 BCE1000 BCE050010001500nowProto-Indo-EuropeanProto-Anatolian† Hittite† Luwian† TocharianProto-Indo-IranianPrakrits & PaliHindi–UrduBengaliPunjabiRomaniProto-Iranian† AvestanPersianKurdishPashtoOssetianGreek† PhrygianArmenianAlbanianProto-Italic† Oscan & UmbrianLatinItalianFrenchSpanishPortugueseCatalanRomanianProto-Celtic† GaulishBrittonicWelshBretonOld IrishIrishScottish GaelicProto-Germanic† GothicOld NorseIcelandicNorwegianSwedishDanishWest GermanicMiddle EnglishEnglishScotsFrisianLow GermanAfrikaansOld High GermanGermanYiddishProto-Balto-SlavicProto-BalticLithuanianLatvian† Old PrussianProto-SlavicBulgarianOld East SlavicRussianUkrainianPolishCzechSerbo-Croatian

Time runs left to right on a stretched scale: the last 1,000 years get as much room as the 4,000 before them. Dates are approximate. Words at the line ends are each language's best-known form; on the time cursor and the map they appear only once the language was written down (* marks a reconstruction). Tap a line for its story.

Show the tree as a list
  • 4500 BCE – 2500 BCE
    • 4200 BCE – 2000 BCE
      • 2000 BCE – 1200 BCE
      • 2000 BCE – 600 BCE
    • 3300 BCE – 900 CE
    • 2800 BCE – 1800 BCE
      • 1800 BCE – 300 BCE
        • 600 BCE – 1000
          • 1000 – today
          • 1000 – today
          • 1000 – today
          • 600 CE – today
      • 1800 BCE – 1000 BCE
        • 1500 BCE – 400 BCE
        • 1000 BCE – 300 BCE
          • 300 BCE – 900 CE
            • 800 CE – today
        • 1000 BCE – today
        • 1000 BCE – today
        • 1000 BCE – 400 CE
          • 300 CE – today
    • 2700 BCE – 1600 BCE
      • 1600 BCE – 1100 BCE
        • 1100 BCE – 300 CE
          • 300 CE – 1453
            • 1453 – today
    • 2700 BCE – 500 CE
    • 2700 BCE – today
    • 2600 BCE – today
    • 2900 BCE – 800 BCE
      • 800 BCE – 100 CE
      • 800 BCE – 800 CE
        • 800 CE – today
        • 800 CE – today
        • 800 CE – today
        • 800 CE – today
        • 800 CE – today
        • 800 CE – today
    • 2900 BCE – 600 BCE
      • 600 BCE – 500 CE
      • 600 BCE – 600 CE
        • 600 CE – today
        • 600 CE – today
      • 600 BCE – 1200
        • 1200 – today
        • 1200 – today
    • 2800 BCE – 200 CE
      • 100 BCE – 800 CE
      • 0 CE – 750 CE
        • 750 CE – 1350
          • 1350 – today
          • 1350 – today
          • 1350 – today
          • 1350 – today
      • 0 CE – 500 CE
        • 450 CE – 1150
          • 1150 – 1500
            • 1500 – today
            • 1400 – today
        • 400 CE – today
        • 400 CE – 1150
          • 1150 – today
        • 450 CE – 1150
          • 1150 – today
            • 1700 – today
        • 500 CE – 1050
          • 1050 – 1350
            • 1350 – today
          • 1000 – today
    • 2800 BCE – 1000 BCE
      • 1000 BCE – 600 CE
        • 600 CE – today
        • 600 CE – today
        • 600 CE – 1700
      • 1000 BCE – 900 CE
        • 860 CE – 1100
          • 1100 – today
        • 900 CE – 1400
          • 1400 – today
          • 1400 – today
        • 900 CE – today
        • 900 CE – today
        • 900 CE – today
today
GreekFrenchGermanPersianItalianSpanishEnglishRussianHindi–Urdu
Drag to move. Pinch, or Ctrl/⌘ + scroll, to zoom.

todayToday: nearly half of humanity speaks an Indo-European language natively.

Press play to watch 6,500 years go by, or tap any line or dot to follow one language home. Pick a word above to see it on every branch.

Chapter 3

Sound laws: why father isn't pather

Sound change isn't random. When a sound shifts, it shifts in every word at once, like a find-and-replace run across a whole language. Type a Latin word and run Germanic's famous one.
After Grimm's law
father

Real English: father

Around 500 BCE, Germanic speakers shifted a whole set of consonants at once. It hit every word with those sounds, which is why linguists can run it backwards. Latin kept the old sounds, so it acts as a fossil of the “before”.

  • p → f
    pater / father
  • t → th
    trēs / three
  • k → h
    cornū / horn
  • kʷ → hw
    quod / what
  • d → t
    decem / ten
  • g → k
    genū / knee
  • bʰ → b
    Sanskrit bhrātā / brother
  • dʰ → d
    Greek thugátēr / daughter
  • gʰ → g
    Latin hostis / guest

German later ran a second shift of its own (apple → Apfel, ten → zehn), and Turkish has one too: see chapter 5.

Chapter 4

Elma, alma, and the apple that crossed families

A Turkish near-pun turns out to be a map of Eurasian history.

elma · apple. From Proto-Turkic *alma. Turkey's Turkish shifted the first vowel; most other Turkic languages never did.

  • Hungarian says alma too

    Hungarian isn't Turkic: it's Uralic, a cousin of Finnish. But before the Magyars reached Hungary in 895 they lived beside Turkic nomads and borrowed about 300 words. Alma is one.

  • Almaty is apple town

    Kazakhstan's biggest city is named after alma. The wild apple forests of the nearby Tian Shan are where the domestic apple comes from.

  • Latin had the same pun

    Latin mālum (apple) and malum (evil) differ only in vowel length. Genesis never names the fruit; the Latin wordplay probably helped make it an apple.

  • Apple used to mean any fruit

    Old English æppel was fruit in general. French pomme de terre and Dutch aardappel, 'earth apple', still mean potato.

  • Spain renamed it

    Spanish manzana comes from mala Matiana, a Roman apple variety, and pushed out the Latin word.

  • English apple is ancient

    Apple, Apfel, jabłko, obuolys and Welsh afal all come from one northern word, older than English itself.

“Apple” in today's languages, grouped by where the word comes from

*h₂ébōl
old northern word (apple, Apfel, jabloko)
  • úll Irish
  • ubhal Scottish Gaelic
  • afal Welsh
  • aval Breton
  • epli Icelandic
  • eple Norwegian
  • äpple Swedish
  • æble Danish
  • apple English
  • aipple Scots
  • apel Frisian
  • Appel Low German
  • appel Dutch
  • appel Afrikaans
  • Apfel German
  • epl Yiddish
  • obuolys Lithuanian
  • ābols Latvian
  • jabəlka Bulgarian
  • jábloko Russian
  • jábluko Ukrainian
  • jabłko Polish
  • jablko Czech
  • jabuka Serbo-Croatian
mālon / mālum
Greek & Latin (also 'evil' pun)
  • mílo Greek
  • mela Italian
  • măr Romanian
  • mollë Albanian
pōmum
Latin 'fruit' (pomme)
  • pomme French
  • poma Catalan
mala Matiana
a Roman apple variety (manzana)
  • manzana Spanish
  • maçã Portuguese
*alma
Turkic (Hungarian borrowed it)
  • ulma Chuvash
  • elma Turkish
  • alma Azerbaijani
  • alma Turkmen
  • alma Kazakh
  • alma Tatar
  • alma Kyrgyz
  • olma Uzbek
  • alma Uyghur
  • alma Hungarian
sēb
Persian sīb
  • sīb Persian
  • sêv Kurdish
  • seb Hindi–Urdu
  • seb Punjabi
*tappūḥ-
Semitic
  • tuffāḥ Arabic
  • tapuaḥ Hebrew
unrelated / isolate
  • khndzor Armenian
  • omena Finnish
  • õun Estonian
  • ḥazzūrā Aramaic
  • sagar Basque

Want it on a map? Turn on the word lens and pick “apple”.

Chapter 5

The Turkish connection

Turkish is Turkic, a family that has nothing to do with Indo-European. But languages are neighbours as well as relatives, and they borrow.
“şeş beş”

Turkish tavla players call the dice in Persian: yek, dü, se, cihar, penç, şeş. Those are Indo-European numbers, so every call hides an English word. (6-5 is şeş beş: Persian six, Turkish five.)

şeş= 6
  • six English
  • sex Latin
  • hex Greek

Hexagon and şeş come from the same *swéḱs.

penç= 5
  • five English
  • pente Greek
  • pañca Sanskrit

Punjab is 'five waters'. The drink punch probably had five ingredients.

Turkish ran its own sound law

Like Grimm's law in Germanic, the Oghuz branch (Turkish, Azerbaijani) voiced many word-initial t and k. Kazakh and Uzbek kept the old sounds.

Old TurkicKazakhUzbekTurkishAzerbaijani
mountaintaγtautog'dağdağ
tongue, languagetiltiltildildil
irontemirtemirtemirdemirdəmir
day, sunkünkünkungüngün
eyeközközko'zgözgöz
comekel-kel-kel-gel-gəl-

Hungarian borrowed from Turkic

Before 895. And from a specific branch: the “r-Turkic” Oghur languages, which say r where Turkish says z. Chuvash, on the Volga, is the last one alive.

  • alma·elmaapple
  • ökör·öküzoxr/z
  • gyűrű·yüzükringr/z
  • ír·yaz-writer/z
  • kecske·keçigoat
  • kos·koçram
  • bika·boğabull
  • tenger·denizsea
  • sátor·çadırtent
  • kapu·kapıgate
  • gyümölcs·yemişfruit
  • búza·buğdaywheat

Indo-European words hiding in Turkish

Turkish isn't Indo-European, but centuries of Persian and Greek neighbours left cousins of English words all over it.

nam ↔ name
Persian nām

Both from *h₁nómn̥. Namı diğer = 'its other name'.

peder ↔ father
Persian pedar

Both from *ph₂tḗr. Germanic turned p into f; Persian kept it.

birader ↔ brother
Persian barādar

Both from *bʰréh₂tēr.

hafta ↔ seven
Persian haft, 'seven'

A week is 'a seven'. Persian and Greek turned s into h: haft, hepta, septem, seven.

Nevruz ↔ new
Persian now-rūz, 'new day'

Persian now and English new are both *néwos.

efendi ↔ authentic
Byzantine Greek afthéntis, 'master'

Same Greek word, one path through Ottoman courts, one through Latin.

tavla ↔ table
Greek távli, from Latin tabula

The board game is literally 'the table'.

şah mat ↔ checkmate
Persian šāh māt, 'the king is helpless'

Not cognates but one phrase borrowed twice: through Arabic and Old French into English.

And English borrowed back

  • yogurt ← yoğurt
    From yoğurmak, 'to knead'.
  • kiosk ← köşk
    An Ottoman garden pavilion, itself from Persian kušk.
  • horde ← ordu
    'Army, camp', reached English through Polish horda.
  • coffee ← kahve
    Arabic qahwa, through Turkish and Italian.
  • tulip ← tülbent
    'Turban cloth': the flower looks like one. Originally Persian dulband.
  • kebab ← kebap
    Arabic kabāb, made famous by Turkish cooks.
  • shish ← şiş
    'Skewer', as in shish kebab.
  • turquoise
    French for 'Turkish' (stone): it reached Europe through Turkey.
  • bosh ← boş
    'Empty', made popular in English by an 1834 novel (probably).

Chapter 6

English, a layer cake

A Germanic base, a Viking layer, a thick Norman French layer, then Latin and Greek on top. You can see the layers in any sentence.

Tap a word to see where it came from.

  • Old English (Germanic) ×4
  • French (Normans)
  • Latin
  • Greek
  • Old Norse (Vikings) ×2
  • Arabic, Persian & Turkish

Where the dictionary comes from

About 80,000 words of the Shorter Oxford Dictionary (Finkenstaedt & Wolff, 1973).

  • Latin 28%
  • French 28%
  • Germanic 25%
  • Greek 5%
  • Other & unknown 14%

Yet nearly all of the 100 words you use most are Germanic: the, and, is, have, mother, water. English keeps Germanic for the kitchen and French and Latin for the courtroom.

The 1066 menu

English farmers raised the animal; Norman lords ate the meat. (Simplified by Walter Scott's Ivanhoe, but the words are real.) Tap a card.

Chapter 7

Myth or true?

Twelve things people say about languages. Guess before you peek.
0/0 right · 12 left

  1. Most European languages come from Persian.

  2. The oldest written Indo-European language was found in Turkey.

  3. English is basically a Latin language.

  4. Latin evolved from Ancient Greek.

  5. Turkish, Hungarian, Finnish, Mongolian, Korean and Japanese form one 'Ural-Altaic' family.

  6. Sanskrit is the mother of all Indo-European languages.

  7. Persian 'bad' and English 'bad' are the same word.

  8. Whiskey and vodka both just mean 'water'.

  9. Basque was spoken in Europe before Indo-European languages arrived.

  10. Turkish is related to Arabic because it has so many Arabic words.

  11. 'Mama' sounds alike everywhere because all languages share one ancestor.

  12. Languages decay over time.

Sources & caveats

Dates are rounded and many are debated; forms with * are scholarly reconstructions. The steppe homeland is the leading view, not the only one (the main rival puts it in Anatolia around 7000 BCE). Built from: Mallory & Adams, The Oxford Introduction to Proto-Indo-European (2006); Anthony, The Horse, the Wheel, and Language (2007); Fortson, Indo-European Language and Culture (2010); Johanson & Csató, The Turkic Languages; Haak et al., Nature (2015); Etymonline, Wiktionary and Glottolog. Spotted a mistake? Tell me.