Reconstruction:Proto-Uralic/kakteksa

• *kakteksan

From *kakta (“two”) with a suffix (shared with *ükteksä (“nine”)) of unknown derivation and origin. There are two main theories regarding the suffix, both of which must be rejected due to various problems:

• that the suffix *-eksa represents a combination of *e- (negative verb), *-k- (reflexive affix) and *-sa(n), *-se(n) (third-person possessive suffix) (and possibly a *-n (dual ending) in *kakteksan, but not in *ükteksä), thus originally meaning "is not there", "is non-existent", with the implication being that one or two is missing from ten, hence 10 - 2 and 10 - 1.[1] The comparison is often made with reflexive forms of the form -kse found in Finnic, but even there only in some languages. The narrow distribution makes it unlikely that such a reflexive structure would have existed and been used in numerals all the way back at the Finno-Mordvinic stage.
• that the suffix is originally of the form *teksa and represents a borrowing from (a descendant of) Proto-Indo-European *déḱm̥ (“ten”).[2] This etymology is phonologically dubious, the *-t- is probably rather part of *kakta,[3] and it is unlikely that an old compound "two-ten", would have come to mean "eight", not "twenty" (or even "twelve").[4] In addition, this theory suffers from chronological issues, as the vowel *-e- implies a significantly earlier date of borrowing than the consonantism of *-ks-.[5] In any case, any supposed *teksa would only be attested through this suffix and nowhere else; it pays a superficial similarity to Hungarian tíz, Komi-Zyrian дас (das) and Udmurt дас (das), but these are simply separate borrowings from Iranian.

As morphologically complex numerals in descendant languages are subject to reduction (e.g. Skolt Sami õtmlo (“eleven”), from *ëktë-mumpē-n-lokā-sën, where all of the latter morphemes were reduced to simply -mlo), it seems likely that the suffix is itself reduced from an earlier form, which can possibly never be conclusively reconstructed.

Furthermore, the Finno-Volgaic numeral has been sometimes compared with Eastern Mari кандаше (kandaše) (and *ükteksä respectively with индеше (indeše)), but the Mari suffix appears to reflect an earlier form *-ndVŋs(k)ə-, which would have been formed in parallel from *kakta rather than the Finno-Mordvinic and Mari forms going back to a common proto-form.[6] Forming the numerals for eight and nine by using the numerals for two and one respectively in a subtractive fashion appears to have been a wider West Uralic phenomenon, as such derivatives are also found in the Permic languages (e.g. Komi-Zyrian кӧкъямыс (kökjamys, “eight”) from кык (kyk, “two”)).

Whether a final *-n was originally part of the suffix is unclear.

*kakteksa

1. (Finno-Mordvinic) eight

• Proto-Mordvinic: *kavksə
  • Erzya: кавксо (kavkso)
  • Moksha: кафкса (kafksa)
• Proto-Finnic: *kaktëksa (see there for further descendants)
• Proto-Samic: *kākcē (or borrowed from Finnic) (see there for further descendants)

• Napolsikh, Vladimir. Uralic Numerals is the evolution of numeral system reconstructable? (2003).
• Kallio, Petri (2020–) “*ükteksän”, in Yhteissuomalainen sanasto‎ (in Finnish)

• Entry #1264 in Uralonet, online Uralic etymological database of the Hungarian Research Centre for Linguistics.