28 Feb 2008

Sporadic phonetic changes in the Indo-European case system

As usual, I'm always inspired by other bloggers as much as the neverending "idea fountain" that is my busy brain. I honestly don't know how anyone could get writer's block when there are so many questions to explore and find answers to.

Phoenix recently pondered on the dialectal variation of Proto-Indo-European *-oso and *-osyo in the genitive case signifying 'of, from'. I'm always fascinated in particular by where words and morphemes come from, especially when working out the details of protolanguages. For me, the origin of this variation of the genitive case ending in Indo-European is sufficiently solved. My take on this "extended" genitive case, using my previously mentioned chronology as guide, is that it probably developed during the middle of the Late IE period by simply agglutinating an endingless locative *ya signifying 'to which' to the original genitive in *-s. The end result in semantics would be 'to which (is the) X' (X = a noun) as a circumlocution for 'of the X'.

The reason behind this seemingly pointless circumlocution would make sense if acrostatic nouns, which fix their accent only on the first syllable of a word no matter what the grammatical case of the noun, had evolved out of protero- and hysterodynamic nouns, which alternate accent between noun stem and case ending depending on grammatical case. So if, for example, there originally was a nominative *ʔékwa-s 'horse' vs. a genitive *ʔekwá-s 'of the horse' in early Late IE and, let's say, a new "acrostatic" rule was imposed on vowel-ending noun stems (a.k.a. thematic noun stems) to fix their accent always on the first syllable regardless of case (thereby getting rid of some clunky root-only alternations of accent), then I would suppose that the resultant homophony of nominative and genitive *ʔékwa-s would be as disturbing for some Indo-European speakers as it is for modern English speakers who are prone to hypercorrect "Thomas' shoe" to "Thomas's shoe" to maintain the same distinction between the nominative subject "Thomas" and its potentially identical possessive form. I might also mention a Guyanese acquaintance who, in his dialect, says 'he own book' for 'his book' where the addition of 'own' to 'he' likewise reinforces a genitive form of the pronoun distinct from the nominative in a similar phenomenon of, shall we say, "unsettling case merger" (i.e. Standard English 'he'/'him'/'his' -> Guyanese English 'he')[1].

After the dialectal variations of the Proto-Indo-European genitive rolled around in my brain for a while, I started pondering on the *-bʰi-/*-mi- isogloss line that seperates some dialects from each other by virtue of an innovation in instrumental plural case endings[2]. It's an interesting fact that both Germanic and Balto-Slavic dialects in the dialect soup known as "Proto-Indo-European" agreed on changing certain case endings with *-bʰi- to *-mi-. This indicates to IEists that the area in which Germanic initially formed was right beside the area in which Balto-Slavic formed. They're two peas in a pod. We know which way the change went too because the postposition *bʰi 'by, near, beside, with'[3] which is the basis of English "by" itself is naturally the origin.

So it provokes the question: Why did *bʰ change to *m in this one case? In Traditional Theory (i.e. where *bʰ is a 'breathy voiced stop') , the phonetics don't work well. It's less likely for a breathy // to become a non-breathy /m/ than it is for a plain voiced /b/ to soften to /m/. It just so happens that in Germanic *bʰ became *b while in Balto-Slavic *bʰ and *b merged so it would then stand to reason that it was after this change that the case endings were free to be mistaken with those in /m/ (like the accusative *-m and genitive plural *-om). However, this is the Traditional account which in all likelihood is inaccurate for several reasons. In my Hybrid Theory (a subtler variant of Glottalic Theory), breathy stops are interpreted as plain ones while 'plain stops' are 'creaky stops' which in some pre-IE stage evolved out of ejective stops. Now we observe instead that Germanic and Balto-Slavic retain conservative phonetics (much like in Glottalic Theory) and that the sporadic change of /b/ to /m/ here was a tempting potential for speakers, right from the beginning. Naturally, in dialects such as Proto-Hellenic (Greek) and Proto-Indo-Iranian where phonation shift from /b/ to // would have occurred, it would no longer be as possible because of the added breathiness of the stop for case endings in *-bʰi- to be mistakingly pronounced as *-mi-. Indeed, we coincidentally see no such innovation in these dialects that would have underwent this shift in pronunciation. So Hybrid Theory would give us a clear post-IE date for this change of *-bhi- to *-mi-, likely occurring sometime after the phonation shift in dialects neighbouring the Germanic/Balto-Slavic sprachbund. Phonation shift would in effect be a barrier to this burgeoning isogloss wave.

Everything seems airtight if I do say so myself.

[1] An example of this is in Peake/Trotz, Gender, Ethnicity and Place: Women and Identities in Guyana (1999), p.157 (see link).
[2] It's a commonly known fact mentioned, for example, in Bloomfield, Language (1994), p.318 (see link). If I recall properly from memory, it was also mentioned for easy reference in Encyclopedia Britannica, volume 22, Languages of the World within the Indo-European section.
[3] The Barnhart Dictionary of Etymology, ed. by Barnhart (1988), p.83 (see link).


  1. Something I don't understand here. How do you explain an alleged alternation between nom. sg. **ékwas and gen. sg. **ekwás in the first place?

  2. Rob: "How do you explain an alleged alternation between nom. sg. **ékwas and gen. sg. **ekwás in the first place?"

    I already explained. Listen closely again! :P I'll start from the beginning.

    My hypothesis is that the genitive singular and a few other case endings were first created in early Proto-Indo-Aegean (c.8500 BCE) by agglutinating postpositions to the noun stem with an optional intrusive *-a- for consonant-ending noun stems. Thus the Pre-Indo-Aegean dative postposition *si "to" became PIAeg *(-a)-sə after Vowel Centralization. (Proto-Indo-Aegean had fixed accent on the initial syllable like in Proto-Uralic.) The nominative singular was at this point still without any case ending (just like in Proto-Uralic).

    Old IE (c.7000 BCE) inherited these case suffixes and also used a definite particle *sa "the" for only definite animate nominative subjects at first. The genitive evolved to *-ásə after PAS (Penultimate Accent Shift) but the nominative case, having no ending at all, placed accent on the second-to-last syllable of the bare noun stem. At this point, both the nominative and the genitive had penultimate accent.

    By early Mid IE, genitive *-ásə was reduced to *-ása since unstressed and *a fell together. Speakers began to perceive postposed *sa now as an agglutinated animate nominative suffix *-sa.

    This agglutination is precisely what obscured the original penultimate accent of the bare nominative singular, making it now a word accented on the antepenultimate syllable (third-to-last syllable). My solution is so efficient that I dare anyone to think up something more ingenious! ;)

    So, like I said already, primary stress accent in Mid IE was much like in Polish and fell on the penultimate syllable (second-from-last syllable) by default unless a suffix was derived from an Old IE agglutinated enclitic in which case the antepenultimate (third-from-last syllable) was chosen.

    By early Late IE, Syncope (deletion of unstressed vowels) produced nominative *-s and genitive *-(á)s. Lastly, Vowel Shift (*a > *o) which occurred just before Proto-IE, yielded the expected case forms.

    Perfect regularity! Glorious! {Glen pats himself on the back for a job well done.}

  3. Okay. Let's take your hypothesis to its logical conclusion.

    *ikwa, *ikwasi

    *əkwa, *əkwasə

    Old IE, before PAS:
    *əkwa-sa, *əkwasə

    Old IE, after PAS:
    *əkwa-sa, *əkwásə

    Mid IE (unstressed vowel merger):
    *əkwasa, *akwása

    Late IE (Syncope):
    *əkws, *kwás

    Am I missing anything?

  4. Well, *ikwa is the correct projection although somehow I doubt that Sea Peoples would retain a word for an animal on the plains. The temptation for Proto-Aegean speakers to forget the word on the islands and only borrow it later from Egyptian, Ugaritic or Luwian during the height of the Minoan civilization would be large, I would think. But I digress.

    Anyways, aside from that you're almost correct. You apply Syncope correctly but I haven't explained the subtleties of the Reduction rule that precedes it. Reduction is when Mid IE unstressed *a is reduced to ultra-short schwa /ᵊ/ in most cases. This is why I was making sure on this blog to say "Syncope reduces *most* instances of MIE unstressed *a to null" and never to say "all" as I once did.

    Here, MIE's nominative form *ékwasa becomes ['ʔekwəsᵊ] after Reduction (not ['ʔekwᵊsᵊ]) because stem-final schwa is not reduced unless it's also word-final in the strong case forms. So Syncope produces early Late IE *ʔékwas.

    Your genitive result **kwás lacks the preceding glottal stop. MIE genitive *akwása becomes [ʔə'kwasᵊ] after Reduction because the first unstressed *a in the MIE form fails to shorten to ultra-short schwa due to Paradigmatic Strengthening which avoids illegal initial clusters in weak stems. Without Paradigmatic Strengthening as an exception to Reduction, the result would have been an eLIE stem with a strong stem *ʔékwa- yet a weak stem **ʔkwá- with an ugly CCC- onset. Instead, the preserved schwa of the first syllable survives Syncope. Pretonic schwa then is fronted to *e. (Paradigmatic Strengthening also explains the origin of PIE's *pód-/*ped- alternation which likewise was not reduced to *pód-/*pd- in its paradigm.)