15 Oct 2009

Prehistoric isoglosses in Proto-Steppe

As you can see, I've been pondering on Proto-Steppe today. Many people refer to this early hypothetical language set most sensibly around 9,000 BCE as Indo-Uralic and it's called this because it's the common ancestor of both Proto-Uralic (PU) and Proto-Indo-European (PIE) afterall. However I still prefer my own term Proto-Steppe a) because it's more descriptive of the likely region where it was spoken and b) because PIE and PU aren't the only language groups implicated in the grouping. I made this simple isogloss map to show at a glance how I would explain Proto-Steppe's development into the later proto-languages known and studied and it relates, as always, to the unpixelated view of the Wave Model of language change. Thus far, I've been satisfied with a 4-vowel system of *a, *i, *u and , forming a pleasant V-shape when you graph it out on paper using the dimensions of height and backness. V-shaped vowel systems are quite common around the world as far as vowel systems go.

Now to explain the three isoglosses I have on display above. I've been getting the impression for a while that Indo-Aegean (IAeg) and Altaic-Gilyak (AG) must have remained particularly close after diffusion of the Proto-Steppe community because I can think of at least two sure features that they share with each other that couldn't have been inherited from the parent language. One is the wholesale softening of word-final *-t to *-s as seen in the changes on animate plural marker *-it (n.b. further erosion of word-final *-s causes in turn Proto-Altaic *-r₂) and the other is an occasional correspondence of *a in IAeg and AG with Boreal *u in certain key words. I attribute this curious development to an original mid-central schwa which could sit equally in accented positions as well as unaccented ones.

Upon revisiting these ideas, I've just realized an interesting minimal triplet in Proto-Steppe that serves as a simple but effective argument to justify the necessity of at least four reconstructed vowels at this stage:
  • *ta 'from'
  • *tu 'you (sg.)'
  • *tə 'that (near you)'
The first becomes the source of the Indo-European ablative *-ód and Uralic partitive *-ta. The PIE form originated by agglutinating the postposition to the nominal stem in IAeg (thus *-ata), followed by Penultimate Accent Shift in Old IE which took the fixed accent off the initial (*-áta), then Syncope (*-ád̰) and finally Vowel Shift, yielding PIE *-ód with regular rules. The second and last examples show a vital difference between them since *tu becomes *tʷa (> PIE *twe, n.b. vocalism secondarily affected by *me < Proto-Steppe *mi 'I') while *tə becomes IAeg *ta without labialization of the preceding stop (>PIE *to-). This is explained if there was an unrounded vowel distinct from both low central unrounded *a and high back rounded *u, namely the mid central unrounded schwa which fits so nicely into an otherwise common 3-vowel system. Forms that suggest to some long-range linguists the apparent existence of a proximal demonstrative **ti on the Proto-Steppe level[1] are, I figure, caused by later analogical derivation out of inherited *tə since this proximal demonstrative is only evidenced in Boreal and AG while IAeg seems to preserve only *ta 'that' (> PIE *to- and Aegean *ta) with a distinct proximal counterpart *ka (> PIE *ḱo- and Aegean *ka). I take the IAeg evidence to show an original word *ka 'this' in Proto-Steppe since, if this is not so, the source of the IAeg form would remain much more obscure than that of Boreal and AG's *ti vis-a-vis the securely inherited deictic *tə. The Altaic forms with word-initial sibilant in place of expected *t- are surely caused by pre-Altaic palatalization before high front vowels as has also apparently occurred in its second person pronominal forms.

All these speculative ideas while interesting and worthy of discussion are however, of course, subject to some range of interpretation. Debate remains open.

[1] See, for example, page 2 of Frederik Kortlandt's article Indo-Uralic and Altaic [pdf].


  1. I used to support the hypothesis that *t# > *s#, but I couldn't find anything to support the change aside from the supposed plural *-(e)s in PIE. What else points to this change?

  2. Also crucial to remember is the 2nd person objective singular: PIE *-s < Steppe *-t. This ending is in turn certainly due to agglutination of the pronoun *tu, something that I think happened long before Proto-Steppe.

  3. In that case how do you explain 3rd person verbal endings in *-t?

  4. The simple answer: The 3ps *-t derives from an older endingless 3ps in Mid IE (MIE) through the addition of the particle *ta 'that'.

    Now further explanation ensues... At the Mid IE stage, *ta was still used for **all genders and cases** (ie. used for 'that person' as well as 'that thing'). The paradigm was *ta (nominative), *tam (accusative) and *tasa (genitive). The other particle, *sa, was not yet part of the paradigm of *ta, and was undeclinable (ie. couldn't take case suffixes). It was used only with animate nouns specifying *general* deixis (ie. equivalent to English 'the'). As with English 'the', *sa couldn't exist alone and always required a preceding noun.

    By Late IE, *sa became confused with *ta and the functions of the two combined into a single, suppletive system where *sa replaced *ta in the animate nominative. After Vowel Shift, these particles become *so and *to-. Feminine forms of these demonstratives only developed in a post-IE stage after Anatolian split, further obscuring the original pattern.

    So, just as MIE *ta is responsible for both PIE *to- and the 3ps *-t (having lost its vowel due to Syncope), MIE *sa is responsible for both *so and nominative *-s, both of which happen to be restricted to animate gender hinting at its original usage.

    I reconstruct the MIE demonstratives as *ka 'this', *ta 'that', *sa 'the' (for animates only) and *ei 'he, she, it'.

  5. Hello! Just wondering that how are the groupings placed chronologically?

  6. I follow more or less what Allan Bomhard's been saying except I add more detail. I personally divide PIE into three equal stages to make things easier to remember: Old IE (7000-6000 BCE), Mid IE (6000-5000 BCE) and Late IE (5000-4000 BCE). I would date Proto-Steppe to about 9000 BCE, centered in the Asian steppelands, again like Bomhard's position. Unlike Greenberg, Bomhard doesn't include Ainu and I agree with its absence.

    Ultimately, Proto-Nostratic is commonly dated to 15,000 BCE. I disagree with a number of Bomhard's reconstructed "Nostratic roots", especially those where we find only AA and IE evidence since I think he's overlooking important Neolithic borrowings from Proto-Semitic to Mid IE that skew the results very much and make it seem that IE and Semitic share more retained features from an ancestral language than they actually do.

  7. I reconstruct the MIE demonstratives as *ka 'this', *ta 'that', *sa 'the' (for animates only) and *ei 'he, she, it'.

    Not a bad solution. I might take to liking this :)

    As far as "Nostratic roots" go I think I share the same concerns. The various versions of Proto-Nostratic phonology are very elaborate (reminiscient of Caucasian and Khoisan languages) and undergo massive simplification in most daughter languages. You're guaranteed to have lots of false cognates when you rely on only two (or even three) languages and a CVC root structure. This is especially the case when you rely on loose semantics like Starostin did.

  8. Thanks. I thought a great deal about it and it relates to the relationship of PIE and Aegean which share both *ka and *ta, as well as the accusative in *-m that comes with it. The accusative ending survives only in Aegean demonstratives as *-n (Etr. ca & ta with accusatives can & tan). In effect, accusative marking became restricted to definite direct objects only.

    "You're guaranteed to have lots of false cognates when you rely on only two (or even three) languages and a CVC root structure."

    Yes! I have a LARGE bone to pick with Nostraticists who aren't just content to reconstruct a 3-way stop system (which is based on good reasoning) but go off the deep end, reconstructing self-indulgent phonemes like lateral ejectives and such, phonemes that lack necessity and only clutter up the hypothesis with ad hoc add-ons. Sadly, the ones most famed for their Nostratic reconstructions are the ones least schooled on Occam's Razor and the KISS principle.

  9. Hi Glen.

    Excellent work. Pls continue along these lines.


    To me, Gilyak is more in with Uralic - Eskimo-Aleut etc, not with Altaic.

    I feel that Ainu may be much more in line with Austroasiatic, in line with archeological data.

    To me, the closest relative, sister, to Altaic is Uralic. Not relation, sister.