The phonetic vagueness of laryngeal theory

While I continue to be strictly speaking Not An Indo-Europeanist, I regularly keep reading about comparative Indo-European research just as well. Including not only matters with immediate relevance to Uralic studies, but also the usual controversy honeypots: interpretations of the stop system (glottalic? aspiration where? how many velar series? etc.); and interpretations of the vowel system in relation to ablaut and laryngeal theory. They seem to often form an important “frontier” of sorts in the development of fine-grained historical phonology reconstruction methodologies, if only due to the large amount of attention they receive.

This doesn’t imply I would be particularly impressed with the average state of the field.

In the case of the last-mentioned, one thing that I see come up a lot is that given a certain degree of uncertainty over the original realizations of the laryngeals, almost everyone seems to be still treating them at least to some extent as deus ex machinae, outside of subjection to phonetically meaningful sound changes.

One particular repeat offender seems to be the interaction of laryngeals with syllabic resonants. Consider e.g. the following list of sound developments given by Peter Schrijver (2015), Pruners and trainers of the Celtic family tree:

  • *CRHjV > *CRījV (laryngeals vocalize to *ī between consonant+resonant and a palatal glide)
  • *R̥DC > *RaDC (word-initial syllabic resonants vocalize to resonant + *a before a voiced unaspirated stop + another consonant)
  • *HR̥C > *aRC (syllabic resonants vocalize to *a + resonant after a word-initial laryngeal — including voiced unaspirated stops)
  • *CR̥HV > *CaRV (syllabic resonants vocalize to *a + resonant before laryngeal + vowel)
  • *CR̥HT > *CRaT (syllabic resonants vocalize to resonant + *a before laryngeal + voiceless stop)
  • *CR̥HC > *CRāC (syllabic resonants vocalize to resonant + *ā before laryngeal + other consonant)
  • *N̥ > *aN (remaining syllabic nasals vocalize to *a + nasal)
  • *R̥ > *aR, *Ri (remaining syllabic liquids vocalize to *a + liquid or liquid + *i)

This is pretty much abstract symbol algebra. At best these can be called sound correspondences between Proto-Indo-European and Proto-Celtic. To suggest that a laryngeal or a syllabic resonant would directly change to or excrete *ī in the first case, but *ā in the sixth, is just about equivalent to claiming “a sound change” *dw > *erk- for Armenian. In reality, developments like these surely must have been composed of several stages.

Of course Schrijver is doing only an overview of Celtic historical phonology, and I would predict that some of the primary sources go into more detail. But it strikes me as an overall problem if there is little interest in IE studies in unpacking these kind of sound correspondences. Nowhere have I seen even fairly in-depth introductions to laryngeal theory attempt to explain these kind of developments using the normal tools and frameworks of historical sound change.

It’s not even very difficult at all to see how some elementary order could be imposed on this kind of a mess. We could note that there is e.g. tons of *a-insertion is going on (and I could add the change *CHC > *CaC, which Schrijver skips over, probably on account of being analyzeable as even earlier than Italo-Celtic). It seems likely there has been a single main epenthesis process, followed by diversification in different environments; not from numerous near-identical epentheses. Additionally, the epenthesis seems likely to have been not quite to *a, given some reflexes as *i.

So for the sake of an example, suppose e.g. that early on, all syllabic resonants first break to *əRə. From such a starting point, most of the more complex developments here will be explainable with what are reasonably natural phonetic developments:

  • *R̥DC “>” *RaDC will be simply the loss of word-initial *ə: *əRəDC- > *RəDC- > *RaDC-.
  • *HR̥C “>” *aRC will be explainable as the blocking of the previous change due to an earlier laryngeal, followed by loss of the second schwa: *HəRəC- > *HəRC- (**HRəC) >> *arC-.
  • *CR̥HV “>” *CaRV will be explainable as the loss of a schwa from an open syllable before a full vowel: *CəRəHV > *CəRHV-. It is not clear if the first schwa would be better assumed to have remained due to schwa lowering to *a intervening (> *CaRHV- > *CaRV-), or due to the laryngeal remaining long enough that the loss of schwa from open syllables was no longer operational (> *CəRV- > *CaRV-).
  • *CR̥HC “>” *CRāC appears to show that the second schwa will now remain in a closed syllable, leading to the loss of the first one instead: *CəRəHC- > *CRəHC-. The compensatory loss of laryngeals may have then kicked in around this time: *CRəHC- > *CRə̄C- > *CRāC-.
  • *CR̥HT “>” *CRaT might diverge from the previous due to any number of reasons. One is that medial voiceless *-T- was likely pronounced longer than its voiced counterparts, and could have induced a shortening *ə̄ > *ə.
  • *CRHjV “>” *CRījV (where we probably expect a syllabic resonant in the input?) could be routed thru e.g. a metathesis *Hj > *iH: thus first *CəRəHjV- > *CəRəiHV-. Then assume a monophthongization *əi > *ī, and loss of the first schwa, now found before a full vowel: *CəRəiH- > *CRīHV-. Finally, suppose loss of the stray laryngeal, and epenthesis of *j as a hiatus filler to acquire *CRījV-, as required.

This is but a quick drabble, and I don’t mean to claim that this would be an accurate view of the actual history. But I would like to see more IEists take a stab at developing an analysis of the finer details of laryngeal theory that at least works more like this second set of sound changes.

I’ve already seen some promising work on syllabification in PIE that posits schwa epenthesis already as an original phonological process, but it seems certain that such research could be also linked to numerous the branch-specific historical developments.

My hunch is moreover that this line of query could end up going much further. To my knowledge, even counting barely attested ancient epigraphic languages, no IE language retains any direct evidence of syllabic nasals, or of the phonetically mysterious “syllabic laryngeals”. And if it were to turn out that phonetic vowels can be assumed to have been there all along: what exactly will be benefits of an analysis that claims *[əH] or *[əN] to really have been phonologically plain */H/ or */N/?

As far as I can tell, a lot about this hangs on the urge to group Indo-European ablaut alternations into neater patterns. And I won’t oppose that investigation — but I get the feeling that its proponents fail to show proper respect for the distinction between internal and comparative reconstruction. Alternations along the lines of *sek- : *sk-, *semk- : *sm̥k- certainly have a greater algebraic consistency, but it’s less clear to me if they could be presumed for PIE itself.

(Similarly it’s interesting how numerous introductions to PIE or some individual IE branch will outline laryngeal coloring as an “early sound change”, but neither outline the slightest amount of evidence for dating it as post-PIE, nor clearly assert that the assumed sound changes are pre-PIE, derived by internal reconstruction rather than by comparative evidence.)

So I could ask…: why would we even assume that the stage *s[ə]mk- is the innovation here? Cross-linguistically, the loss of reduced vowels is far more common than their insertion. Yet IE studies instead outline an amazing cornucopia of early epenthesis processes. Another look at the field also reveals several theories about the rise of zero grades from pre-PIE vowel reduction. Still for some reason it seems to have remained overwhelmingly difficult for scholars to put 2 and 2 together and to conclude that many of these “epentheses” are probably archaisms rather than innovations.

Tagged with: , , , ,
Posted in Commentary, Methodology
14 comments on “The phonetic vagueness of laryngeal theory
  1. M. says:

    Syllabic nasals — regardless of whether they had a schwa coloring in their phonetic realization — seem considerably more solid than syllabic laryngeals. This is because, unlike laryngeals, we have direct evidence of nasals in coda position consistently alternating with another sound in zero-grade, i.e. exactly where we would expect the nasal to become syllabic.


    – Greek pépontha “I have suffered” < IE *kWe-kWontH-(H)a vs. épathon “I suffered” < e-kWn.tH-om

    – the realization of the accusative ending after a vocalic stem — Greek lúkon “wolf” (acc. sg.), Latin lupum (id.) — vs. its realization after a consonant stem: Greek póda “foot” (acc. sg.), Latin pedem (id.)

    • M. says:

      Also, the fact that Greek and Indo-Iranian both have -a- in this position (Greek hekatón : Farsi sad : English hundred) suggests that the original sound was not a full, heavy syllable: in other words, it is more likely to have been phonetically *[ən] than *[ən].

      If the original sound were *[ən], then we would have to explain why Greek and Indo-Iranian didn’t simply shift the schwa to another vowel (as the other IE branches did), since neither branch otherwise loses nasals in coda position.

      • j. says:

        This is indeed a good question, though I don’t think it’s a problem in my suggestion specifically. Loss of coda nasals is at least a widely-known and well-attested sound change. So is loss of intervocalic nasals — e.g. Portuguese; or colloquial Finnish menen, panen > meen, paan. But is the direct vocalization of syllabic nasals known from anywhere else than from this alleged early Indo-European development? (Genuine question — they’re not that common to begin with.)

        Also why *a? By far most vocalizations of consonants yield something in their phonetic vicinity: β > u, ɫ > u, ʎ > i, ɣ > i/u, etc. Vocalization to an open vowel normally only appears in the case of gutturals (as in German (ə)ʁ > ɐ). By this precedent I would expect e.g. *m̥ to maintain its labiality and yield **u.

        Then there are the conditional developments *N̥HV, *N̥NV, *N̥V > *aNV. I don’t think I’ve seen evidence elsewhere for PIE to have tolerated geminate sonorants, which makes Siever’s law -type explanations seem off.

        The change is additionally fairly localized in being found in Greek and II, and could be considered a single innovation (either a genetic or an early areal one). So even if we will end up positing actual syllabic [m̩], [n̩] as its input, this could still have already been a further development from earlier [əm] [ən] or the like.

        • M. says:

          Loss of coda nasals is at least a widely-known and well-attested sound change.

          The question is why, if the original pronunciation were *[ən], the nasal would have regularly disappeared in this environment, but not in any other coda environment, stressed or unstressed. (Compare Greek an “against”, sumphorá “gathering together, collection”, etc.)

          But is the direct vocalization of syllabic nasals known from anywhere else than from this alleged early Indo-European development?

          Off the top of my head, I’m not sure, but I may have heard some examples of this in English, where coda nasals in unstressed syllables sometimes become syllabic. E.g., I can imagine the sentence “I can do it” being pronounced [‘aj.kə~.du.wıt], with nasalized schwa.

          Admittedly, it’s hard to tell whether this pronunciation is the outcome of a syllabic nasal, or simply the nasalization of a never-lost vowel.

          Also why *a?

          The vowel may have shifted from *[ə] to *[ʌ] (or a similar central vowel) at an early point, and this *[ʌ] could then have shifted to its closest pre-existing equivalent, short *a. In fact, some modern Indo-Iranian languages show precisely [ʌ] in this position: [ʌ] is the pronunciation of the Hindi sound normally Latinized as short a, according to the transcriptions I’ve seen.

          The same explanation would potentially cover the outcome of syllabic nasals as an in Celtic and Armenian (cf. Welsh cant “100”, Armenian ksan “20” < *wikm̩ti).

          Then there are the conditional developments *N̥HV, *N̥NV, *N̥V > *aNV. I don’t think I’ve seen evidence elsewhere for PIE to have tolerated geminate sonorants, which makes Siever’s law -type explanations seem off.

          It’s hard for me to comment on the (theorized) treatment of syllabic nasals before laryngeals, because I’m currently skeptical of the amount of laryngeals reconstructed for IE to begin with.

          However, in the case of *N̥V > *aNV, this doesn’t seem like an unusual development of *[ən] (again, I meant to put the schwa in superscript, but the browser isn’t letting me do so). When a vowel follows the nasal, it will tend to become the onset of this vowel; the preceding schwa will thus be separated into its own syllable, and expand to the normal length of a vowel in this position.

        • David Marjanović says:

          Vocalization to an open vowel normally only appears in the case of gutturals (as in German (ə)ʁ > ɐ).

          Nope – non-rhoticity is quite a bit older (and therefore more widespread) in German than the shift from [r] to [ʀ] (and then on to [ʁ] in northern Germany). But then, the dialects I have in mind are all Bavarian, and the only reduced vowel there is [ɐ] already…

  2. David Marjanović says:


    To my knowledge, even counting barely attested ancient epigraphic languages, no IE language retains any direct evidence of syllabic nasals, or of the phonetically mysterious “syllabic laryngeals”.

    Are the Lydian syllabic nasals original?

    Some people – like Andrew Byrd, the Great Syllabifier – deny syllabic laryngeals altogether, saying instead that laryngeals couldn’t be syllabic any more than *s could. They just talk about “laryngeals between two other consonants”. I think they’re right.

    I get the feeling that its proponents fail to show proper respect for the distinction between internal and comparative reconstruction

    I think the problem at the root of all this is the continued failure of historical linguists to routinely distinguish phonetic, phonemic and morphophonemic transcriptions. Sometimes you can find all three in a single PIE word.

    This is why Byrd writes his predictable schwa as *°: to make clear that it wasn’t a phoneme (and therefore, despite being a vowel, wasn’t stressable).

    • j. says:

      Are the Lydian syllabic nasals original?

      I don’t know much more about Lydian than that it is not to be confused with Lycian or Ludian, but Melchert in “Historical Phonology of Anatolian” (JIES 1993) seems to posit that any non-close vowel + coda nasal yields ‹ã›, any non-close vowel before an intervocalic nasal yields ‹ẽ›.

  3. David Marjanović says:

    There is some evidence for direct [r̩] > [ɐ] in southeastern German: unstressed der shows up as d’r in an 18th-century song about Prince Eugene of Savoy, but has since become [d̥ɐ].

    • j. says:

      That’s presuming it’s an internal development, and not just the importation of a competing isogloss — similar to e.g. how there are Finnish dialects that up til the early 1900s had /l/ as the weak grade of /t/, but have by now switched to using /r/ or zero (while leaving unalternating /l/ untouched). Do you happen to know if -er as [ɐ] forms a single spread zone across German varieties, or several? I would guess it’s unconnected to English non-rhoticity at least (but I would not be too surprized to be wrong even on that).

      • David Marjanović says:

        I’m sure it’s unconnected to any of the English non-rhoticities (three of which seem to be unconnected with each other: Britain/Australia/NZ/SA, New England/NY, southern US/AAVE); however, –er as [ɐ] is universal in German from sea to shining mountains except 1) in fully rhotic Switzerland and surroundings, where it’s [r̩], and 2) wherever in Germany that strange accent is that has merged [ɐ] into [ɑː].

    • j. says:

      Ah this stuff. I had been wondering when someone would bring it up here. I’ve seen it (it’s from my alma mater after all), but it does not strike me as a laryngeal theory critique that will have much impact. I might be more convinced that this is “trouble” if Pyysalo manages to either acquire supporters from mainstream IE studies, or at least to get one of his critiques into a mainstream journal. (By the latter metric, starting now a journal of his own seems like a concession of being “rogue”.)

      Now, sure, he’s right that *h₁ and *h₃ could be in theory just about replaced by ablaut, if we only looked at *VH / *HV environments. Unfortunately *RH, *HR and *H̥ keep all three distinct: that is cases like Latin datus ~ Greek δοτος, that Pyysalo only manages to treat (at least in his thesis) by special pleading that they would go to different ablaut grades *də-tos / *do-tos; or the likes of *H₃negʷ- > ὄνυξ, where he ends up having to go with ad hoc prefixation. Also Luwic is supposed to distinguish *eh₁ (> *ǣ > ā) as distinct from *ē (remains ē), even though they merge in Hittite and non-Anatolian IE.

      Even most of these could be perhaps still done away with, but I think that would require stronger weapons — something like positing that zero grade is at least partly a post-PIE phenomenon, and that datus < *dətos is a reduction from earlier *dotos, which was preserved unreduced in Greek. Not unthinkable, but I’d like to first see where we can get by less drastic reanalyses.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Enter your email address to follow this blog and receive notifications of new posts by email.

%d bloggers like this: