The word three stays strikingly similar across the Indo-European languages. This makes me wonder... The word one is all over the place. Two is a little easier to chart. However, no matter what language you dig up from the meta-family, you usually aren't going to have much trouble recognizing the word three.
Just look at this handy chart of proto-Indo-European numerals from Wikipedia. Focus particularly on the part excerpted below for your convenience, that which specifically deals with the variant forms of the number three (all rights, of course, reserved to the original authors).
|three||*trei-||Hitt. teriyaš (gen. pl.), Lyc. trei, Ved. tráyas, Av. θrāiiō, Pers. çi/se,|
Osset. ærtæ,ærtæ, Kashmiri tre, Kamviri tre, Gk. τρεῖς,treĩs, Lat. trēs,
Osc. trís, Umbr. trif, ON þrír, Goth. þreis, Eng.þrēo/three, Gm. drī/drei,
Gaul. treis, Ir. treí/trí, Welsh tri, Arm. erek῾/yerek῾/yerek῾, Toch. tre/trai,
OPruss. tri, Latv. trīs, Lith. trỹs, OCS trije, Pol. trzy, Russ. tri, Alb. tre/tre
As you can see, not much variation in threes across the Indo-European language sprachraum. Personally, I have no clue what to make of it. Is it that these low numbers are used more often, and thus stayed more regular? Wouldn't using a word often contribute to, rather than detract from, linguistic drift and sound change? Was "three" more commonly used in trade and diplomacy across linguistic divides? Your guess is as good as mine.