Indo-Iranian and Balto-Slavic languages are called satem languages, because in them the Proto-Indo-European palatovelars *ḱ, , and *ǵʰ developed into sibilants or affricats, usually into [s]/[z]- or [ʃ]/[ʒ]-type sounds. In the centum languages (Hellenic, Celtic, Italic, and Germanic), the palatovelars merged with the plain velars *k, *g, *gʰ series.

Still, the Balto-Slavic languages share a word that does not follow that rule, in this word the PIE *ḱ changed into *k, and not to *s or *š as it would have been expected. Such a change is characteristic of the centum languages, but not the satem Balto-Slavic.
The word means 'stone, rock' and it developed so:
PIE *h₂éḱmō > Proto-Balto-Slavic *akmō, which gave the modern words:

  • Latvian: akmens
  • Lithuanian: akmuõ
  • Proto-Slavic: *akmy > *kamy (with metathesis) which further gave Russian ка́мень (kámenʹ), Serbo-Croatian kȃm, kami, kȁmēn, Czech kámen, etc. Other IE languages treat the word according to their position within the centum-satem division: Proto-Germanic *hamaraz (English hammer), Sanskrit áçmā 'stone, rock', Old Persian asman- 'stone'.

Is there any plausible explanation for such an inconsistency of an otherwise consistent sound law? All the sources I found just say "unexplained centumization". But are there at least some theories why that happened, any attempts to explain that? Maybe the PIE source of the PBS *akmō had *k in the root and not *ḱ?

  • 2
    More examples: paśu- ~ pekus; śapha- ~ kanopa, копыто; śardha- ~ kerdzius; śmaśru- ~ smakras Feb 16, 2020 at 12:38
  • 1
    @BertBarrois - It looks like копыто doesn't belong here. Isn't it derived from the verb копать 'to dig' < PIE *(s)kep- (“to strike, beat”)? And the Lithuanian kanopa has "n" in the middle... The rest of the words are really impressive.
    – Yellow Sky
    Feb 16, 2020 at 13:13
  • копыто (hoof) is a perfect match in meaning, but Pokorny doesn't cite it, and Vasmer also has doubts. I'd argue that many animals do use hooves to dig. Lithuanian kanopa is anomalous, but the odds against the k...p sequence occurring in an exact synonym must be rather long. Feb 16, 2020 at 15:26
  • For some possible explanations, take a look here: en.wikipedia.org/wiki/Proto-Balto-Slavic_language#Satemization.
    – RuslanD
    Apr 26, 2020 at 3:08
  • 1
    The simplest explanation is probably that the etymologies on Wiktionary are wrong, as they unfortunately often are. It is not a reliable source, it is not meant to be. Nov 9, 2020 at 11:38

2 Answers 2


This has been known for a long time (see e.g. Meillet 1894: 299), cf. Kim 2018 "Balto-Slavic also has several examples of velars continuing PIE palatals (“Gutturalwechsel”)." For instance, Otkupshchikov (Откупщиков 1989/2001), in Ряды индоевропейских гуттуральных, discusses what he calls «непоследовательная сатемность» (inconsistent satemization) in e.g. Lithuanian, Sanskrit, and Slavic.

Hock 2004 discusses these "exceptions" in more detail - he refers to them as examples of "unvollständige Satempalatalisation."

Now it is usually attributed to dialectal variation (as opposed to some earlier accounts due to borrowing).

e.g. Fortson 2010, who incidentally thinks the centum-satem merger was a separate/parallel development in each IE branch, says the following in Chapter 18 Balto-Slavic (section 18.5, pp. 415-416), "A more probable explanation is that there were early dialectal differences within Balto-Slavic with regards to the outcomes of the palatal velars, and that these differences have persisted. This hypothesis, unfortunately, cannot yet be tested."

cf. Kim 2018: Such cases "suggest that pre-PBS exhibited some variation in this regard; perhaps the palatalization of PIE palatal stops began in the east of the (Late) IE-speaking area, in the dialects ancestral to Indo-Iranian, and spread to most but not all pre-PBS dialects" (p. 1975)

Languages are very complex phenomena and they cannot be adequately described with "die Lautgesetze kennen keine Ausnahmen."


The idea that exceptions to the satem changes in Balto-Slavic has to do with dialect borrowing and/or dialect variation isn't very plausible, in my opinion, since no centum dialects are known in Baltic or Slavic. I believe that the exceptions stem from the fact that Balto-Slavic territory was the periphery of the spread of the satem changes, and that the spread was never completed. (This is probably compatible with Hock's view.) That is, the situation (if I'm right) was similar to the Rhenish fan in German-speaking territory, where the High German Consonant Shift didn't hit all words at the farthest extent of its spread in the Rhineland. Messy peripheries are well known to dialect geographers. The situation in Balto-Slavic may also be connected to the fact that the labialized velars were changing to plain velars: if that set of changes had arrived in Balto-Slavic territory before the original plain (pre-?)velars had all undergone the satem changes, some of the new plain velars (from labiovelars) might have been at risk of undergoing the satem palatalizations. This explanation has been part of my historical linguistics classes for years.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.