in short: Is there any discussion available online of the following reinterpretation (due to Alwin Kloekhorst) of the stops in the Indo-European phoneme inventory?

classical inventory    reinterpretation
p  t  ḱ  k  kʷ      →  p: t: kʲ: k: kʷ
b  d  ǵ  g  gʷ      → ˀp ˀt ˀḱ  ˀk ˀkʷ
bʰ dʰ ǵʰ gʰ gʷʰ     →  p  t  kʲ  k  kʷ

what are the arguments backing up Alwin Kloekhorst's statement ?

full details :

I read in Alwin Kloekhorst's monumental thesis, from the Leiden University, about the Hittite lexicon that, speaking about the "Proto-Indo-European phoneme inventory" (title after 1.1, p.29):

(...) I follow Kortlandt (2003: 259) who argues that the traditional 'voiceless' series (*p, *t, *ḱ, *k and *kʷ) in fact were plain fortis stops [p:, t:, kʲ:, k:, kʷ], the traditional 'voiced' series (*b, *d, *ǵ, *g, *gʷ) were lenis (pre-)glottalized stops [ˀp, ˀt, ˀḱ, ˀk and ˀkʷ] and the traditional 'aspirated voiced' stops (*bʰ, *dʰ, *ǵʰ, *gʰ, *gʷʰ) were plain lenis stops [p, t, kʲ, k, kʷ] (1.1, p.30)

It should be noted that although I tried to copy exactly what Kloekhorst wrote, I had to modify the symbols for (pre-)glottalized stops and; I also avoid, for technical reasons, italicizing as the author did. As as sidenote, I think it's worth noting that Kloekhorst's views seem to have nothing to do with the glottalic theory even if Kloekhorst graduated from Leiden University, where the

"proponents [of the glottalic theory] today are historical linguists at the University of Leiden" (source)

Since Alwin Kloekhorst doesn't justify his allegiance to Kortlandt's assumption, I followed the given reference, namely "Kortlandt (2003; 259)" which leads to the article An Indo-European substratum in Slavic? (Languages in Prehistoric Europe (edd. A. Bammesberger & T. Venneman), Heidelberg, 253-260), but this article seems to be out of my reach. Full bibliography here.

So far, I've only read another article from Frederik Kortlandt, From Proto-Indo-European to Slavic, 1994. But what I read is somehow different from Alwin Kloekhorst's views: Proto-Indo-European would have fortes (p, t, ḱ, ...), glottalic (b, d, ǵ, ...) and aspirated (bʰ, tʰ, ǵʰ, ...) stops and "Dialectal Indo-European" (namely Balto-Slavic, Germanic, Albanian, Armenian, Armenian, Indo-Iranian and probably Tocharian) would have fortes (p, t, ḱ, ...), (pre-)glottalized (ˀb, ˀd, ˀǵ, ...) and voiced (bʰ, tʰ, ǵʰ, ...) stops.

Hence my question: what are the arguments backing up Alwin Kloekhorst's statement?

2 Answers 2


If I remember correctly, a major argument in favour of the reinterpreted consonant system is the suspicious rareness of *b in the protolanguage. The absence of /b/ in the presence of /p/ and /bʰ/ is more difficult to explain than the absence of /ˀp/ in the presence of /p:/ and /p/.

The problem is that no Indogermanic language has preserved glottalised stops (they are present in Armenian but this can be explained as a later acquisition due to areal influence of Caucasian languages).

  • 1
    That is the basic argument that started off the glottalic theory in itself; I’m not sure it’s really a central argument in Kloekhorst’s particular view, though. Oct 24, 2017 at 13:56
  • Sindhi has implosives though, ejective consonants also in occur in Ossetian though they mostly in loanwords and in words of unknown origin but they also occur in some native iranic words too e.g "стъалы" /stʼalə/ "star". Sep 4, 2022 at 20:38
  • @LinguisticsFanatic But those are a later acquisition, too. Sanskrit did not have them. Sep 4, 2022 at 20:40

The main evidence comes from Hittite (and to a lesser extent cuneiform Luwian). In essence, when the Hittites adapted the cuneiform writing system from the Akkadians, they thoroughly ignored the difference between the glyphs for voiced and voiceless sounds. Instead, they represented the descendants of PIE "voiced" and "voiceless" stops by adapting the convention for Akkadian geminates: the PIE "voiceless" stops are written with two glyphs, like li-ip-pa- "smear" < *leyp-, while the PIE "voiced" and "voiced aspirated" stops are written with one glyph, like ne-pí- "cloud" < *nebʰ-.

Kloekhorst presents various arguments that this writing convention in Hittite actually did represent a length difference, with fortis stops being pronounced longer than lenis ones. For example, Hurrian shows a similar distinction written in the same way, and Hittite scribes writing Akkadian words make frequent errors in voicing. We also don't see any sign of voicing assimilation; fortis and lenis consonants can sit next to each other with no confusion or change, like in e-ku-ud-du /ekʷtːu/ "he must drink".

He then extrapolates from this back to Proto-Indo-European. The traditional explanation and the most common version of the glottalic theory, he argues, both have significant typological issues. What if, instead of a voicing distinction, early PIE had a length distinction in its stops? In other words, before the Anatolian languages split off from the main trunk, the three series of stops were long, glottalized, and short?

Kloekhorst attributes the first expression of this idea to Kortlandt 2003 ("An Indo-European substratum in Slavic?"), as you mention, which I have also been unable to access. But looking at later papers by Kortlandt that cite it, he suggests that the long/short distinction in stops came from an earlier heavy/light distinction in syllables, hence the prohibition on a fortis stop and a lenis stop in the same root. This, he claims, could have come from an earlier stage of the language with a two-way lexical tone distinction, which could also explain some correlation between stop voicing and accentuation; here he cites Lubotsky 1988 page 170, which isn't listed in his bibliography, but I suspect it's "The System of Nominal Accentuation in Sanskrit and Proto-Indo-European".

This idea isn't widely accepted, because, to put it simply, there isn't much evidence for it. There's no obvious path from voiced stops in PIE to unvoiced aspirates in Greek, but there's also no obvious path from short stops in PIE to unvoiced aspirates. I find Kloekhorst's arguments for a length distinction in Anatolian convincing (and so do Hoffner and Melchert and various others), but I'm not persuaded that there's anything to gain from claiming this was the state of things back in PIE. And Kortlandt's main goal seems to be to connect Indo-European and Uralic; I'm not sure how well his tone-based explanation holds up without the assumption of deriving the PIE stop system from a more Uralic-esque one.

For more details, see chapter 1.3.2 in Kloekhorst's Etymological Dictionary of the Hittite Inherited Lexicon. (You cite this in the question, I know, but a detailed reference may be useful to later readers.)

  • There wouldn’t be a path from voiced stops /b d g/ to the Greek unvoiced aspirates, since those are from the voiced aspirates in the Classical system. A loss of voicing is all that’s needed (bringing the system into alignment with more typologically common distinction types). The path from unvoiced stop to unvoiced aspirated stop is also quite obvious: just add aspiration. But why that should happen only to short stops and not geminates, that’s more problematic to me. Not to mention that having to deal with initial geminates is just another typological rarity for PIE. Sep 5, 2022 at 12:54
  • @JanusBahsJacquet I was thinking of the glottalic theory there (since Kloekhorst references it in developing his new model), which generally gets rid of the aspiration distinction and has voiceless, glottalized, and voiced stops.
    – Draconis
    Sep 5, 2022 at 15:13
  • Oh I see what you mean now – that makes more sense. I do think that Alwin’s hypothesis works better than the glottalic theory on this particular point, though: unvoiced stops becoming aspirated is typologically common, whereas voiced unaspirated to unvoiced aspirated is quite rare (modern Korean notwithstanding). The only downside is that the postulated geminates were somehow exempt (but then again, so are the tense stops in Korean, which are written as doubled plain stops, though phonetically they are the glottalised ones). Sep 5, 2022 at 16:37
  • On the other hand, I’m not aware of any parallels to the development that would be required for the alleged plain unvoiced stops to become the voiced aspirated stops we see directly in Indo-Iranian (and indirectly elsewhere). Sep 5, 2022 at 16:39
  • @JanusBahsJacquet That's a fair point about Korean; I hadn't considered that one. But yes, getting from this set of stops to Indo-Iranian requires some serious convolutions.
    – Draconis
    Sep 5, 2022 at 17:03

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.