Most consonants in Hittite appear in two variants, conventionally called "voiced" and "voiceless": "voiceless" consonants are written twice in a row, while "voiced" consonants are written only once. For example, the word appa "later" is variously written a-ap-pa or ap-pa, but never *ap-a or *a-pa, while the word apa "that" is always ap-a or a-pa, not *ap-pa. So it's thought that the former was something like /apa/, and the latter /aba/. (Or maybe fortis-lenis, or short-long, or…)

However, this distinction usually only shows up intervocalically: initially, finally, or in clusters, the writing system has no consistent way to express it (as there are no glyphs for lone consonants).

So—is anything known about this contrast in other environments? Do we know if it was actually neutralized in other contexts, or if it just didn't show up in writing?

(We might know this based on, for example, the spelling of clusters: if a form like nahh-teni "you all are afraid" is consistently written *na-ha-te-ni, for example, that's decent evidence that the hh ended up getting voiced; if it's consistently written *na-ah-ha-te-ni, that's decent evidence that it remained voiceless. But I don't know if either of these writings actually happens.)

EDIT: As fdb pointed out, voiceless consonants aren't quite "written twice in a row", since there are no signs for individual consonants in Hittite cuneiform. Rather, two adjacent signs both indicate the same consonant: ap-pa is written with a sign for /ap/ followed a sign for /pa/.

  • To say that "voiceless consonants are written twice in a row" is somewhat misleading. ap-pa is written with two syllable signs "ap" and "pa". Neither can be broken down further; there is no "a" and no "p" in either of them. – fdb Sep 26 '19 at 15:59
  • @fdb Fair; I'll clarify. The intent is that voiceless consonants are written with two adjacent signs both marking the same consonant. – Draconis Sep 26 '19 at 16:00
  • Something along those lines. – fdb Sep 26 '19 at 16:01

This topic is not easy to answer.
To begin with, as a rule, PIE voiceless *t k p are normally written double intervocalically, while PIE voiced *d(h) *g(h) *b(h) are normally written simple in the same position. So we have good reasons to think that Hittite had two series of consonants. The controversy is the nature of the distinction.
The same issue exists in Hurrian.
There are few exceptions to the above rule. One is mekki "much" related to *meg- "big". Hittite has an odd kk in that word.
Now, the cuneiform system does not give the means to write the distinction word-initially or word-finally nor in clusters of consonants. So here, we're in the dark. I think the number of Hittite words attested in Ugaritic script is too small to provide sufficient light. Some Armenian words seem to be of Hittite origin, but I don't know if they are numerous enough to be helpful.
So, the conclusion is that we don't have satisfactory information about Hittite consonants in word-final or word-initial position. It seems probable that the distinction existed but we can't tell for sure.


Melchert claims that "voicing" was not distinguished word-initially or word-finally, with word-initial stops ending up fortis (PIE *geis- > kiš- "become" > reduplicated kikkiš- with a fortis consonant) and word-final stops ending up lenis (PIE *h₁poi-h₁ei-ti > pait "went" > paid=aš "he came" with a lenis consonant).

Kloekhorst, however, argues that this "voicing" was distinguished word-finally, but generally not written due to the limitations of syllabic cuneiform; he explains paid=aš as lenition caused by the enclitic, like how =(y)a "and" fortites a preceding consonant. I know Kloekhorst's ideas about Hittite phonology are generally controversial (to put it lightly), but Watkins does seem to agree with him in general here.

(Note that I'm generally following Melchert rather than Kloekhorst in the phonemic transcriptions, since Kloekhorst's other ideas, like word-initial a- indicating a glottal stop, are much less widely accepted.)

Kloekhorst's argument is, first, that Hittite had phonemic labiovelar stops. He points to graphical alternations like e-ku-zi vs e-uk-zi "drinks" and tar-ku-zi vs tar-uk-zi "dances" as indicating an underlying /egʷ-zi/ and /tarkʷ-zi/, and claims that the u written in these forms doesn't act like /u/ or /w/: a-ku-e-ni /agʷ-weni/ "we drink" but ar-nu-me-ni /arnu-weni/ "we transport" (with /w/ becoming [m] when adjacent to /u/).

Then, if we accept that this spelling can indicate a single phonological consonant (rather than a CV sequence), we see a clear word-final "voicing" distinction in pairs like ne-ek-ku /nekʷ/ "not?" < PIE *ne-kʷe versus e-ku /egʷ/ "drink!" < PIE *h₁egʷʰ.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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