Groucho Marxism

It has long been recognized that the distribution of Proto-Indo-European (PIE) velars is highly skewed after *s, with plain velars occurring far more frequently in this position than palatovelars and labiovelars. In Pokorny’s 1959 Indogermanisches Etymologisches Wörterbuch (IEW), there 118 entries with a plain velar occurring after *s or mobile *s, as opposed to 13 entries with a palatovelar and 4 with a labiovelar (here I have included entries with a velar after *z, the voiced allophone of *s). Similarly, in Rix et al.’s 2001 Lexikon der Indomanischen Verben (LIV2) and Wodtko et al.’s 2008 Nomina im Indogermanischen Lexikon (NIL), there are 35 entries with plain velar occurring after *s or mobile *s, as opposed to 7 entries with a palatovelar and 3 with a labiovelar.

How can we account for this distribution? The obvious explanation is that the opposition between the velar series was neutralized after *s in PIE. The problem with this is that we still find all three types of velar occurring in this position. In the case of mobile *s we can posit analogical change to account for this, but that doesn’t work for roots with non-mobile *s. There are 5 entries where a palatovelar occurs after non-mobile *s in IEW: sk’ed- ‘cover’, sk’eH(i)- ‘shimmer’, *sk’er(d)- ‘defecate’, *sk’eu- ‘throw’, and *mosg’hos ‘young’. There are 4 additional entries where a palatovelar occurs in this position in LIV2 and NIL: sk’eH₂i- ‘separate’, sk’eH₂id- ‘split’, sk’end- ‘cover’, and *tresk’- ‘squeeze’. Let us go through each of these in detail.

The root sk’ed- ‘cover’ only has reflexes in Indo-Iranian among the satem languages. Whereas the Iranian reflexes suggest an original palatovelar, the Indic reflexes point to a plain velar. This discrepancy can be explained by positing that palatalization was not blocked after *s in Iranian when the velar was followed by a front vowel. This also applies to sk’end- ‘cover’ which is obviously a nasal-infix variant of the same root. The same explanation can be applied to the root *sk’eH₂i- ‘separate’ if we assume that the laryngeal was lost in the zero-grade forms sk’H₂i- prior to the palatalization in Iranian. It can also be applied to the roots *sk’eH(i)- ‘shimmer’ and *sk’er(d)- ‘defecate’ if we assume that the root was originally *s-mobile and that the Balto-Slavic reflexes are derived from the variant without *s.

The root *mosg’hos ‘young’ only has reflexes in Armenian among the satem languages, where the palatalized reflex can be explained by positing that *-sgh- yields -z- in Armenian. This rule was established by the Australian linguist Robert Woodhouse in 2014, as was the rule that palatalization was not blocked after *s in Iranian when the velar was followed by a front vowel. The root *tresk’- ‘squeeze’ should be reconstructed with a plain velar it only has unpalatalized reflexes in the satem languages. The same applies the root sk’eH₂id- ‘split’. This root is clearly derived from *sk’eH₂i- ‘separate’ (see above), which provides further evidence that the latter root should be reconstructed with a plain velar. The root *sk’eu- ‘throw’ is not probative as it only has reflexes in Balto-Slavic.

Thus none of the entries with a palatovelar after non-mobile *s stand up to scrutiny. What about the examples with a labiovelar in this position? There are no entries where a labiovelar occurs after non-mobile *s in IEW, and just 1 entry where a labiovelar occurs in this position in LIV2 and NIL: *Hosg^w– ‘knot’. In his 2014 paper Woodhouse also provides 3 examples of a labiovelar apparently occurring after non-mobile *s: *sg^whH₂el- ‘stumble’, *sk^welH- ‘split’, and *mosg^w– ‘knot’. However *sg^whH₂el- ‘stumble’ is reconstructed with a mobile *s in LIV2 to account for the Latin reflex; and  *sk^welH- ‘split’ is reconstructed with a plain velar in LIV2 to account for the Hittite reflex. The labiovelars in *Hosg^w– ‘knot’ and *mosg^w– ‘knot’ are reconstructed based on single words from Gaulish and Old Norse respectively and are therefore far from certain.

Thus none of the entries with a labiovelar after non-mobile *s stand up to scrutiny either. The obvious conclusion is that the opposition between the velar series was indeed neutralized after *s in PIE.