In a previous blog post I argued that the highly skewed distribution of velars after *s in Proto-Indo-European (PIE), with plain velars occurring far more frequently here than palatovelars and labiovelars, can be explained by positing that the opposition between the velar series was neutralized in this position. Another position where the distribution of velars is highly skewed is after *u. In Pokorny’s 1959 Indogermanisches Etymologisches Wörterbuch (IEW), there 103 entries with a plain velar after *u, as opposed to 16 with a palatovelar and 6 with a labiovelar. And in Rix et al.’s 2001 Lexikon der Indomanischen Verben (LIV2) and Wodtko et al.’s 2008 Nomina im Indogermanischen Lexikon (NIL), there are 32 entries with plain velar after *u, as opposed to 5 with a palatovelar and 0 with a labiovelar.
The obvious explanation that the opposition between the velar series was also neutralized after *u in PIE runs into the problem that we find all three types of velar occurring in this position. However a closer inspection of the data suggests that this might not be the case. One of the entries in IEW where a palatovelar occurs after *u – *peug’- ‘stab’ – only has reflexes in the Centum languages and therefore cannot be securely reconstructed with a palatovelar. Two more entries – *bhreuk’- ‘stroke’ and *ghreug’- ‘rub’ – only have reflexes in Balto-Slavic and therefore cannot be securely reconstructed for PIE, particularly as the first violates a known root structure constraint whereby voiceless and voiced aspirate stops cannot co-occur in the same root.
Of the 15 remaining entries, three should be reconstructed with a medial laryngeal – as *bhuHg’- ‘goat’, *gheuHg’h- ‘hide’, and *gʷreuHg’h- ‘grind’ – to account for the long vowel reflexes, so the palatovelar would not have been in contact with *u in these roots. Two more entries should be reconstructed with a plain velar after *u – as *k’euk- ‘mix’ and *reug- ‘roar’ – to account for the Balto-Slavic reflexes. The entry *leug’- ‘break’ shows divergent reflexes in the satem languages, with plain velar reflexes in Indo-Iranian and Albanian and palatovelar reflexes in Baltic (the Armenian reflexes are not probative as all velars were palatalized after *u in Armenian). This root is reconstructed with a plain velar in LIV2, which creates a connection with the root *leug- ‘bend’ and suggests the proposed Baltic reflexes are unrelated.
Similarly, the entry *leuk’- ‘light’ has divergent reflexes in the satem languages, with plain velar and palatovelar reflexes in both in Indo-Iranian and Balto-Slavic. The acute tone in the Baltic reflexes points to a variant with a medial laryngeal, *leuHk’-, where the palatovelar would not have been in contact with the *u. That leaves us with 6 entries where a palatovelar occurs after *u: *gneug’- ‘knot’, *ieug’- ‘excite’ *meuk’- ‘scratch’, *peuk’- ‘stab’, *puk’- ‘press’, and *(s)leug’- ‘slide’. To this list we may add *seuk’- ‘suck’ from LIV2, although again this has divergent reflexes in the satem languages, with plain velar reflexes in Baltic and palatovelar reflexes in Slavic. Thus it seems that palatovelars do occur after *u, although they are rare in this position.
Four of the entries in IEW where a labiovelar occurs after *u – *bhleugʷ- ‘overflow’, *bhoukʷ- ‘hum’, *H₂eukʷ- ‘cooking pot’, and *ugʷ- ‘damp’ – were shown not to require the reconstruction of a labiovalar by the Australian linguist Robert Woodhouse in a 2017 article. Another entry, *H₁ugʷh- ‘speak’, is a zero-grade variant of the root *H₁uegʷh- where the labiovelar would not have occurred after *u, and indeed the reflexes of the zero-grade form show the expected delabialization. The final entry, *ugʷsen- ‘bull’, contains no evidence for a labiovelar whatsoever; the labiovelar in this word is reconstructed based entirely on a speculative connection with the root *uegʷ- ‘damp’. Thus it seems safe to conclude that labiovelars do not occur after *u.
To sum up then, there is evidence for only two types of velar occurring after *u: palatovelars and plain velars. In a series of previous blog posts I have hypothesized that PIE only contained two types of velar consonant: plain velars and labiovelars. This hypothesis is consistent with the evidence if we assume that plain velars palatalized after *u in the satem languages and labiovelars delabialized after *u in the centum languages. This would imply that the delabialization after *u occurred separately in the individual centum languages and cannot be traced back to PIE. This is not problematic however as there is independent evidence of delabialization occurring in this position in all of the individual centum languages, and the assumption that it occurred in PIE itself is therefore superfluous.
Leave a Reply