Groucho Marxism

In an ever-expanding series of recent blog posts I have argued, based on data compiled by Lars Steensland in his 1973 PhD thesis, that Proto-Indo-European – henceforth PIE – had two rather than three types of velar consonant: plain velars and labiovelars. I have also set out the phonological rules by which the plain velars remained intact in the satem languages, and by which labiovelars were delabialized in the centum languages. Although these rules account for a large proportion of the observed reflexes, they do not account for all of them. For a complete two-velar theory, I think we need to invoke another phonological rule: namely, that plain velars were remained intact before *o in the satem languages. In this blog post I will set out the supporting evidence.

Steensland provides 31 examples of words with palatovelars occurring in word-initial position before *o, when not preceded by *s; 20 examples of words with plain velars in this position; and 12 examples of words with labiovelars in this position. He then shows that of the 12 examples of words with labiovelars, the labiovelar in all but 1 could have been introduced analogically from a related where the labiovelar occurred before *e, and concludes that there is no clear evidence of labiovelars occurring before *o in PIE. However, of the 31 examples of words with palatovelars, the palatovelar in all but 9 could also have been introduced analogically in a similar way. Moreover, none of the 9 remaining words stands up to scrutiny.

The first word is *k’oinos ‘hay’, which supposedly yields reflexes in Balto-Slavic and Greek. However, the Greek word may simply be an inflected form of another Greek word derived from a different root. The second word is *k’olH₂mos ‘straw’ which yields reflexes in Balto-Slavic, Italic, Germanic, and Greek. The original paradigm may be reconstructed as *k’elH₂m ~ *k’lH₂mos, in which case the palatovelar could have been introduced analogically from the *e-grade. The ‘gutturalwechsel’ (velar interchange) in the Baltic-Slavic reflexes provides further evidence that this root originally contained a plain velar. The third word is *k’onghos ‘shell’, which supposedly yields reflexes in Indo-Iranian and Greek, but the correspondence is irregular so these are probably false cognates.

The fourth and fifth words are *k’ormnos ‘acid’, which yields reflexes in Balto-Slavic and Germanic; and *k’ormos ‘pain’, which yields reflexes in Balto-Slavic, Indo-Iranian, and Germanic. Both of these words can be traced back to a root *k’er(H)- ‘hurt’, which is attested in the *e-grade in Greek. The sixth word is *g’houHos ‘call’, which yields reflexes in Indo-Iranian and Balto-Slavic; this word can be traced back to a root *g’heuH- ‘call’, which is attested in the *e-grade in Indo-Iranian. The seventh and eighth words are *k’okH₂eH₂ ‘branch’ and *k’onkos, both of which yield reflexes in Indo-Iranian, Balto-Slavic, and Germanic. These words can be traced back to a root *k’ek- ‘sway’ with a nasal infix variant *k’enk-, which are attested in the *e-grade in Albanian and Indo-Iranian respectively.

Hence, the palatovelar could have been introduced analogically in words four to eight. The final word is *k’opH₂elos ‘carp’. This word only has reflexes in the satem languages, and as these were contiguous, can therefore only be traced back as far as ‘proto-satem’ rather than to PIE. We can conclude from the above that Steensland’s data is consistent with the hypothesis that the split between palatovelars and plain velars before *o occurred separately in the satem languages, and therefore that there is no need to reconstruct a separate palatovelar series of consonants before *o in PIE. This hypothesis also has the advantage of explaining numerous instances of ‘gutturalwechsel’: alternations between palatovelars and plain velars across different satem languages.

For example, the word *g’hordos ‘enclosure’ yields both palatovelar and plain velar reflexes in Balto-Slavic, along with a plain velar reflex in Albanian. The underlying root *g’her- is attested in the *e-grade in Indo-Iranian. The gutturalwechsel in *g’hordos ‘enclosure’ can therefore be explained by positing that the word originally contained a plain velar that was blocked from palatalizing before the *o, with the palatovelar reflexes later being introduced analogically from the *e-grade. Similarly, the word *k’oimos ‘home’ yields both palatovelar and plain velar reflexes in Balto-Slavic, and the underlying root *k’ei- is attested in the *e-grade in Indo-Iranian and Balto-Slavic, so the gutturalwechsel can be explained in the same way.

Another example is the word *bherg’hos ‘fortification’, which yields palatovelar reflexes in Indo-Iranian and Armenian, and plain velar reflexes in Balto-Slavic. The velar in the underlying root *bherg’h- would have occurred before *e in several Indo-Iranian and Armenian reflexes, so again the gutturalwechsel can  be explained by positing that the word originally contained a plain velar that was blocked from palatalizing before the *o. These explanations have the advantage of uniting what are usually reconstructed as separate roots with the same meaning: *g’her-/*gher- ‘enclose’, *k’ei-/*kei- ‘settle’, and *bherg’h-/*bhergh- ‘protect’. It seems much more plausible that these each represent a single PIE root with the divergent velar reflexes being the result of secondary developments.