7 May 2010

Against the *dkmtóm camp

Some Indoeuropeanists reconstruct *-dḱomt- '-ty' and *dḱm̥tóm '100' instead of simply *-ḱomt- and *ḱm̥tóm by analogy with *déḱm̥ '10'. While superficially seductive, there are some good reasons against this.

No evidence for **d in *ḱomt-/*ḱm̥t- whatsoever

This elusive **d has never been directly attested in any Indo-European (IE) cognate of the higher decads, despite there being a plethora of languages to choose from. A lack of proof is enough to dismiss claims by simple Occam's Razor and burden of proof. Yet some scholars persist putting **d- in *-ḱomt- and *ḱm̥tóm, with or without parentheses, before bearing evidence that directly demonstrates this phoneme in these stems, showing that there continues to be a widespread disregard for linguistic methodology and due process in PIE studies.

Laryngeal switcheroo

Since this **d fails inspection, some resort to an alternative *h₁ to account for lengthening in Greek and Latin decadic compounds. Frederik Kortlandt goes so far as to assume that *h₁ (as /ʔ/) comes from earlier **ˀd which he describes as both "preglottalic" and "implosive" interchangeably without anything but circumstantial evidence. He imposes this same change in word-initial position and traces them to original ejectives as per Glottalic Theory[1] (ie. *tʼḱm̥tóm > *ˀdḱm̥tóm > *(ʔ)ḱm̥tóm). I find this implausible and unnecessary[2] and I have already settled on laryngealization (ie. creaky phonation), with derivative word-internal pre-laryngealization, as a superior alternative to wholesale preglottalization in all environments. Both the change of word-initial ejective to preglottalized voiced stop and the ad hoc reduction of *ˀd to here, further compounded by his claim that this laryngeal then explains -e- in Greek hekaton '100'[3], are a series of exotic and contrived fabrications.

Yet even if one wishes to defend **d in a Pre-IE reconstruction, it nonetheless has no business being in a valid root of Indo-European proper. Pushing it to Pre-IE, however, doesn't let the theorist off the hook either. Justification must still be demanded.

Must the vanishing *d be a phonetic phenomenon?

Before any of these considerations, it's presumptious to immediately think that *déḱm̥ and *ḱomt-/*ḱm̥t- must be built on exactly the same stem. A poignant example is Proto-Austronesian (PAN) *sa-puluq 'ten' or literally 'one-ten', built on *esa 'one' and *-puluq 'ten; -ty'. In higher decads, *sa- is naturally replaced by other numbers (cf. *duSa-puluq '20; two tens'). Oddly, this reflects best the actual PIE pattern observed in higher decads where *d- seems to be likewise omitted. Must the disappearance be phonetic instead of morphological? There's no evidence thus far published to support that assumption - the proposed phonetic changes fail miserably and further we have parallel PAN evidence to support a morphological cause.

A solution

Rather than indulging in unnecessary assumptions, let's take the unexpected lack of *d in higher PIE decads as a done deal. Let's also reject simplistic solutions involving a ubiquitous deus ex machina like *h₁. Let's instead acknowledge the obvious that number systems are unlikely to evade analogical levelling over the course of millennia. Instead of taking the lengthening in 30 to 90 at face value as a Proto-Indo-European feature, we accept that the overall PIE data is ambiguous to lengthening and that there are later changes at work.

To go with this topic, I've created my own pdf and placed the link on my Lingua Files page showing how I personally would reconstruct the Indo-European number system. As you can see, my view looks a lot like had been originally reconstructed before idle speculation got the better of IEists by the latter half of the 20th century.

We may blame the lengthening in the Greek and Latin decads on analogy between '30'/'40' and the rest of the decads that originally could not have produced such lengthening. The ultimate source of lengthening? Concatenated neuter forms of '3' and '4' marked with the collective marker, *-h₂, that originally quantified a neuter stem *ḱómt-. Note that the numbers 5 to 9 were unmarked for case and therefore are bare. Once fossilized in '30' and '40' and once laryngeals gave way to compensatory lengthening, the decads from 50 to 90 were prone to analogical lengthening. Latin shows intervening -ā- (eg. quīnqu-ā-gintā) in these stems while Greek shows -ē- (eg. πεντ-ή-κοντα), clearly showing different post-IE strategies employed to normalize obscured decadic terms.

[1] Kortlandt, Baltica & Balto-Slavica (2009), p.68 (see link): "In my view, the original PIE ejectives developed into implosives in all branches except Anatolian and Tocharian, and show traces of glottalization and/or partial merger with the laryngeals in Germanic, Italic, Greek, Armenian, Indo-Iranian, and Balto-Slavic."; Evidence and counter-evidence: Essays in honour of Frederik Kortlandt (2008), p.417 (see link): "In a number of papers Kortlandt (1988a, b; 2000; 2003) has suggested that the ejectives that both he and I reconstruct for Proto-Indo-European changed into preglottalised stops in Proto-Germanic before they became plain voiceless stops in the individual daughter languages."
[2] Fallon speaks against the proposed Germanic change in The synchronic and diachronic phonology of ejectives (2002), p.283 (see link); Doubts are also offered in The new sound of Indo-European: Essays in phonological reconstruction (1989), p.218 (see link).
[3] Venneman, The new sound of Indo-European: Essays in phonological reconstruction (1989), p.28 (see link). Kortlandt's offer solves nothing since **h₁ḱm̥tóm alone can only supply Greek *ekatón (missing h-) while **sm̥-h₁ḱm̥tóm would produce even worse results: *smākatón!


  1. You've got me convinced. There's no proof for the *h1 or *d whatsoever. Even if it seems attractive to reconstruct the *d in Pre-PIE, there's no indication for it whatsoever.

    Might I add that the word 'twenty' and 'thousand' in Vedic sanskrit never seem to show lengthening in the word before it, while other words with initial *HC most definitely do, and often too.

  2. Ah, right. That's an interesting route to argue too. Thanks for mentioning that.

    I've been thinking lately of moraicity and glottal stops, myself, because if we know that glottal stops, by their nature, are incapable of bearing morae, then they shouldn't be able to lend the second mora in the lengthened vowels that they are expected to produce... unless... the glottal stop has already eroded to something more vowel-like!

    In other words, if one is going to blame Winter's Law lengthening on preglottalization (as Kortlandt does), then one is either implying that the preglottalized phoneme in question was already weakened (eg. via laryngealization), or that one is plain wrong.

    Interestingly, I just realized that even *h1 in general could be justified as [h] in mediofinal environments this way, simply by virtue of its reknowned vowel-lengthening properties that shouldn't directly be accomplished through mora-less [ʔ]. Hmmm.