It seems to me that some words that have -p- in stem in Latin have clearly reconstructible -k- based on other Indo-European languages. Some examples include

  • *u̯lpes - *u̯lkos ("wolf")
  • *u̯esper - *u̯eskeros ("evening")
  • *apa - aka ("water")

The "k" reconstruction is supported by the majority of languages, still many authors consider the roots with "p" as genuine and/or separate. For example, Katz tryed to proove the connection between *u̯esper, Hettite *u̯esp "cloth of the dead" and Greek ospros "pulse".

Mallory & Adams reconstructed the PIE water deity as *H2epom Nepots instead of evident *Akam Nepot(s) (note that the laryngeal here is not supported by the Nostratic evidence).

Starostin lists both *ap- and *aka; *welp- and *welk- (and also *lup-) as separate roots.

I wonder why the correspondence between *k and *p is not evident to some authors?

  • 6
    I think you should clearly identify the Latin and the Greek stems here first, because I see some Latin /k(ʷ)/ sounds there. Latin: vulpes, vesper, aqua, iecur. | In many cases, it is even the other way around: Latin has /kʷ/, while Greek has /p/: quinque/pente, equus/hippos, relinquo/leipô, etc.; in similar cases, Latin has /kʷ/ where Greek has /t/: qui/ti, quattuor/tettares, -que/te. In all these cases, the /kʷ/ is said to be the original PIE sound, which changed into /p/ or /t/ in Greek depending on the subsequent vowel: /t/ before e or i, /p/ before o (but more changes happened in between).
    – Cerberus
    Commented Aug 8, 2012 at 5:48
  • I'm also find the examples rather confusing, sorry. How is *apa relevant? I only know that as a form in Romanian.
    – legatrix
    Commented Aug 8, 2012 at 12:04
  • Cerberus, indeed, but consider the first two examples. It seems that many scholars either reconstruct two roots for each or consider the p-variant in its own.
    – Anixx
    Commented Aug 8, 2012 at 13:17
  • 3
    I believe Oscan/Umbrian also had a p reflex for kʷ. Commented Aug 8, 2012 at 15:10
  • 2
    Which Starostin, Sergei or his son George?
    – Alex B.
    Commented Aug 8, 2012 at 18:57

3 Answers 3


(a slightly updated answer)

Latin "lupus" is usually believed to be a borrowing from Sabellic - PIE *kw corresponds to Sabellic (Osco-Umbrian) *p. Other examples of this phenomenon is Latin popina vs. Latin coquina or Latin Pontius vs. Quintius. Another stop in Sabellic borrowings: Latin "bos". And then metathesis is usually proposed for Greek "lykos".

As for Latin "vesper", you might not be seeing a bigger picture. Compare Greek "hesperos", Armenian "gisher", Hittite "ispant-", and Sanskrit "ksap-". It has been suggested that the PIE root was *ue(k)spero, a derivative of PIE *kwsep- "night", the zero grade *kwsp-, the consonant cluster was simplified in different daughter languages with various outcomes.

Of course, you can dispute the link between Latin "vesper" (Greek "hesperos" and the rest), on the one hand, and Lithuanian "vakaras" and OCS večerъ, on the other hand. OCS večerъ has no clear etymology, and various hypotheses have been proposed. The hypothesis I mentioned above clearly has one weak point: it offers no explanation of what might have happened to "sp" in Balto-Slavic; although it was suggested that some sort of taboo was involved there (see Havers 1946, Specht 1944) or possibly another word (Muehlenbach 1923-1932). There is another proposal that OCS večerъ is a not a simplex, but a root *vek- with a suffix erъ (like OCS severъ). Or maybe it is related to Lithuanian ukanas "cloudy" and ukti "overcast" etc. So, the last two proposals do not posit any connection between OCS večerъ and Latin "vesper" (and other comparanda mentioned above).

It's not clear how the hypothesized PIE form *weskwer- would explain the Balto-Slavic data (what happened to *-s-?). Nor it is entirely clear why two roots with the same meaning would exist in PIE.

As for Latin "quinque" and PIE *penkwe, it is usually suggested that in Italic as well as in Celtic (!) languages *p was assimilated to the following *kw (distance or non-adjacent assimilation). You can find it even in intro textbooks, e.g. Campbell 2004: 29. Another example of this assimilation: PIE *pekw- vs. Latin coquina.

You need to understand that in historical linguistics, esp. in etymological research, there is no ultimate theory that explains everything, a theory that cannot be refuted. Quite often several hypotheses coexist, with different degrees of probability. And quite often these hypotheses are ad-hoc, that is they are created to explain one or, at best, a couple of words. This is undesirable but, I'm afraid, unavoidable, at least at the moment.

  • How *kwsper could give "ч" in Russian вечер? Only *weskʷer could do.
    – Anixx
    Commented Aug 8, 2012 at 15:44
  • PIE *uekspero - *večerъ. Where do you see a problem here?
    – Alex B.
    Commented Aug 8, 2012 at 15:53
  • 1
    @Annix, you completely misunderstood my post. Greek h and Armenian g correspond to PIE *w (*u).
    – Alex B.
    Commented Aug 8, 2012 at 16:00
  • 1
    @Anixx, again another misunderstanding. Please re-read my post carefully. PIE *kw corresponds to Latin *kw and Osco-Umbrian *p. Latin Quintius coexisted with a Sabellic borrowing Pontius.
    – Alex B.
    Commented Aug 8, 2012 at 16:07
  • 1
    Aeolic Greek followed a similar path as the (p-)Celtic and Oscan/Umbrian languages, e.g. πέμπε "five". Commented Aug 8, 2012 at 16:48

Alex B has given a wonderfully detailed answer, but I'd like to sum up some extra information for future readers.

The general consensus, to my understanding, is that the forms with *kʷ are the original ones. They're found in a wider variety of languages, and /kʷ//p/ is a common sound change, which pretty much never happens the other way around.

The change /kʷ//p/, as mentioned, shows up a lot in unrelated places. The two main sub-branches of Celtic are known as P-Celtic and Q-Celtic because the former went through this change and the latter didn't; in more modern times, Romanian lost the Latin labiovelars in the same way (Spanish cuatro ~ Romanian patru). Most relevantly here, it happened in Attic all dialects of Greek (Attic hippos ~ Latin equus), and in the whole East Italic branch (Oscan, Umbrian, a few others—Latin quattuor ~ Oscan petora).

So to my understanding, all the alternate forms with /p/ are later developments. For example, I haven't seen any solid evidence for a PIE root **h₂-p "water"; forms like Romanian apa come straightforwardly from Latin aqua, with the regular loss of labiovelars. Where there's an alternation within the same language, it can usually be explained by borrowing: Latin seems to have borrowed lupus "wolf" from Oscan or another East Italic language, where the change from *kʷ to /p/ was also regular.

  • 1
    Yes. Separate from that is the (remote) assimilation /p- -kʷ/ -> /kʷ- -kʷ/ in Latin, seen in *penkwe -> quinque, *pekwo -> coquo, and according to some sources *percu -> quercus.
    – Colin Fine
    Commented Mar 27, 2019 at 23:58
  • (Not just Attic Greek, but all Greek dialects, though with differences of detail.)
    – TKR
    Commented Mar 28, 2019 at 21:10
  • @TKR Didn't some dialects turn it into /t/ in all environments or something like that? (I'm not great at dialects.)
    – Draconis
    Commented Mar 28, 2019 at 21:11
  • 1
    No, Aeolic turned it into [p] in all environments (the boukolos rule aside); with other dialects it's basically [t] before front vowels, [p] elsewhere, though again with complications.
    – TKR
    Commented Mar 28, 2019 at 21:58
  • @TKR Interesting! Edited that in.
    – Draconis
    Commented Mar 28, 2019 at 22:00

I'm not sure I understand the question correctly.

*gw > b is a known change in Celtic. E.g. for Irish "woman" from the same root as Greek gyne, Gothic quno, English queen.

*kw > p occurred so late in Old Irish, that it is actually attested. One might speculate on the nature of this change, whether it's complete coincidence.

I had noted something similar in a comment somewhen, for Latin or Greek if I remember correctly, but no general rule.

  • 1
    It happens in P-Celtic, but also in Greek, "Oscan/Umbrian Italic" (which probably has a better name but I don't know it), probably others. It notably doesn't happen in Latin.
    – Draconis
    Commented Mar 27, 2019 at 20:44
  • oh, I meant \kw (updated accordingly), having recently read linguistics.stackexchange.com/questions/14407/…. I'm glad *gw wasn't a wrong note, I even remembered the example. However, I'd still like to find my old comment, that I referred to.
    – vectory
    Commented Mar 27, 2019 at 22:13

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.