A family resemblance
Say the numbers from two to ten in Latin, Greek, Sanskrit and English and similarities start to appear. Latin “tres”, Greek “treis”, Sanskrit “trayas” and English “three” are obviously close. Latin “septem”, Greek “hepta” and Sanskrit “sapta” are nearly the same word, and English “seven” is not far off.
Nineteenth-century scholars turned this resemblance into a method. They found that the correspondences were regular: a sound in one language matched a predictable sound in another, again and again, across the vocabulary. This pointed to a shared ancestor, now called Proto-Indo-European, spoken perhaps several thousand years ago. The forms of that ancestor are reconstructed, not recorded, and they are written with an asterisk to mark them as scholarly reconstructions. The details of the reconstructions are still discussed.
One to ten compared
The table below gives the usual forms in each language, with simplified spelling. Greek forms are the classical ones, and Sanskrit is written in a common transliteration.
| Number | Sanskrit | Greek | Latin | English |
|---|---|---|---|---|
| 1 | eka | heis, hen | unus | one |
| 2 | dva | duo | duo | two |
| 3 | trayas | treis | tres | three |
| 4 | catvaras | tettares, tessares | quattuor | four |
| 5 | panca | pente | quinque | five |
| 6 | sat | hex | sex | six |
| 7 | sapta | hepta | septem | seven |
| 8 | asta | okto | octo | eight |
| 9 | nava | ennea | novem | nine |
| 10 | dasa | deka | decem | ten |
Reading the table
Most rows show an obvious relationship. Some of the differences are regular sound changes.
Greek often has an “h” where Sanskrit and Latin have an “s”. Greek “hex” and “hepta” correspond to Latin “sex” and “septem”, and the same shift is a well-known feature of the language. In the other direction, English “six” and “seven” keep the “s”.
English also took part in the Germanic sound shift, sometimes called Grimm’s law. Under it, an old “p” became “f” and an old “t” became “th”, and an old “d” became “t”. That is why Latin “pater” corresponds to English “father” and why Latin “decem” corresponds to English “ten”, while “three” begins with a “th” where Latin “tres” has a “t”. The “f” of “five”, against the “p” of Greek “pente”, is the same effect.
The word for “five” shows another pattern. The ancestral form is usually reconstructed with a “kw” sound at the end, which Latin kept as “qu” in “quinque”. The initial “qu” there is generally explained as the result of the second “kw” influencing the first. Welsh “pump” has a “p” from the same ancestral sound, and Irish “cuig” has a “c”, which is one reason that Celtic languages are divided into “p-Celtic” and “q-Celtic” groups.
The numbers that do not match
“One” is the instructive exception. English “one” and Latin “unus” go back to the same root, and a related Greek word survives as “oine”, the ace on a die. But Sanskrit “eka” and the Greek “heis” come from different roots. This is a reminder that even very common words can be replaced within a branch of a family, and that finding a match is not guaranteed for every item.
“Nine” has a small curiosity. The Latin form “novem” ends in an “m”, though an older final sound would have been “n”, and this is often explained as influence from “decem”. This explanation is plausible but not certain, and some scholars connect the ancestral “nine” to the word “new”. That link is a debated suggestion and should not be treated as established.
“Four” is also unusual in English. In Greek and Latin the first sound is a “t” or “q”, but Germanic “four” begins with “f”. This does not follow the standard pattern and is usually attributed to influence from the following “five”. A confident statement would be that its beginning is explained by a special development, and that scholars have discussed it.
Beyond ten
The same method works for “hundred”, where English “hundred” and Latin “centum” match, with the Germanic “h” in the place of Latin “c” being another instance of Grimm’s law. Multiples of ten in English are made from the units plus “-ty”, which comes from the same word as “ten”, and the pattern of compounding is much the same in other related languages. These are the same kind of compounds found in number words generally, and they are a good reminder that the units are the old core and the larger numbers were assembled on top.
The tens in Sanskrit, Greek and Latin were built in similar ways. Latin “viginti” (twenty) is often analysed as a form of “two tens”, although the sounds changed enough that the connection needs the comparative method to see. The decimal basis is shared across the family. Counting on ten fingers is a natural guess for why, though the habits of prehistoric counting among the speakers of the ancestral language cannot be observed.
Languages that kept more of the old forms
Not all branches changed at the same pace. Lithuanian, a Baltic language, and Sanskrit are often cited as conservative in some features. Russian and the other Slavic languages show recognisable relatives of the numerals above: “dva”, “tri”, “pyat”, “sem”, “devyat” and “desyat” for two, three, five, seven, nine and ten. Irish and Welsh also keep clear traces. A comparison across ten languages from Ireland to India makes the family connection even harder to dismiss.
What it does and does not show
The shared numerals are among the best illustrations of the idea of a language family. They show that languages spread across a wide area, from Iceland to India, descend from a common source. They do not tell us who the speakers were, where they lived, or what they thought numbers meant. Where the numbers show up later in belief systems, for example in Pythagorean number philosophy, in the Latin names behind Roman numerals or in the Sanskrit texts of India, those are separate histories built on the same words.
Languages outside the family count differently. Hungarian “egy, ketto, harom” and Finnish “yksi, kaksi, kolme” belong to another family altogether, and Basque, which is not related to any of them, uses a base-twenty pattern, a topic taken up under base twenty counting. Comparing them shows the difference between a shared origin and a borrowed idea.
In short, the table is an argument that is built up one regular sound change at a time. It supports the family tree without proving every detail of the ancestral language.