Cowgill's law

Cowgill's law, named after Indo-Europeanist Warren Cowgill, refers to two unrelated sound changes, one occurring in Proto-Greek and the other in Proto-Germanic.

Cowgill's law in Greek

In Proto-Greek, Cowgill's law says that a former /o/ vowel becomes /u/ between a resonant (/r/, /l/, /m/, /n/) and a labial consonant (including labiovelars), in either order.

Examples:

• núks "night" < PIE *nokʷts (cf. Lat. nox, Ved. nák < *nakts, Goth. nahts, Hitt. gen. sg. nekuz /nekʷts/)
• phúllon "leaf" < PIE *bʰolyom (cf. Lat. folium)
• múlē "mill" < PIE *mol-eh₂- (cf. Lat. molīna)
• ónuks "nail" (stem ónukh-) < early PG *onokʷʰ- < PIE h₃nogʷʰ- (cf. OE nægl < PGerm *nag-laz)

Note that when a labiovelar adjoins an /o/ affected by Cowgill's law, the new /u/ will cause the labiovelar to lose its labial component (as in núks and ónuks/ónukh-, where the usual Greek change *kʷ > p has not occurred).

Cowgill's law in Germanic

Cowgill's law in Germanic has no relation to Cowgill's law in Greek other than having been named after the same person. It says that a PIE laryngeal /h₃/, and possibly /h₂/, turns into /k/ in Proto-Germanic when directly preceded by a sonorant and followed by /w/. This law is still controversial, although increasingly accepted. Donald Ringe (2006) accepts it; Andrew Sihler (1995) is noncommittal.

Examples are fairly few:

• *kwikwaz "alive" (whence English quick) < PIE *gʷih₃-wos (cf. Lat. vīvus)
• *unkw- "us two" (cf. Goth. unkis) < PIE *n̥h₃we (cf. Gk. nṓ; Ved. āvā́m acc. du. "us two" < *āva-ám)
• Possibly OE tācor "husband's brother" < PIE *dayh₂wer

If it becomes generally accepted, the relative chronology of this law could have consequences for a possible reconstructed phonetic value of h₃. Since Germanic /k/ results from earlier PIE /g/, and since the change occurred before Grimm's law applied (according to Ringe), the resulting change would be actually h₃w > gʷ. This would have been more likely if h₃ was a voiced velar obstruent to begin with. If h₃ was a voiced labiovelar fricative as is occasionally suggested, the change would therefore have been: ɣʷw > ɡʷ.