In the course of my research for my Indo-European conlang called Wôks Teuteka (https://vellaunos.ca/2021/08/02/woks-teuteko/ ), I have recently established a connection between Greek agathos, Proto-Germanic *gōdaz (English good) and Proto-Celtic *dagos (Welsh da) among others. Here is the entry for the word gado in my Wôks Teuteka wordlist as it now stands:

gado (gados; gadôs) : good [< *h2ghh2-dhos; Lat honor ‘honor, esteem, dignity’ < h2gh(h2)-dho-nos (h2gh(h2)-dho-nes-); PrCelt *dagos (< *h2dhh2-ghos, metathesis of *h2ghh2-dhos (?)) > OIr dag-, Welsh da; *PrGmc *gōdaz (< *h2gheh2-dhos) > Eng good; Gr agathós ‘good, brave, noble, moral’ < *h2(e)ghh2-dhos; PrSlav *godьnъ > Russ godnyj ‘fit, suitable, valid, appropriate’] {for Lat ho– < *ghdho– >>> gəmōn}

This led me to consider the effect of Grassmann’s law in Ancient Greek. The effect of this law is seen in both Ancient Greek and Sanskrit, and it consists of the deaspiration of one of two aspirated consonants in a word. In the case of Greek agathos, the g comes from a deaspirated *gh which normally would have become kh (i.e. **akhathos).

Although there is some dispute about how Grassmann’s law in Ancient Greek is related to Grassmann’s law in Sanskrit, my opinion is that Grassmann’s law operated in the common ancestor of Proto-Greek and Proto-Indo-Iranian, this common ancestor being the eastern variant of North Proto-Indo-European. (By the way, the effect of Grassmann’s law was “nullified” in Proto-Iranian because all aspirated consonants were deaspirated in this language.)

As I’ve indicated elsewhere (https://vellaunos.ca/2021/03/24/the-movements-and-expansions-of-indo-european-language-groups/ ), I believe that Proto-Greek and Proto-Armenian derived from a fusion of the eastern variant of North Proto-Indo-European (Abashevo Culture > Potapovka Culture) with the South Proto-Indo-European that remained on the eastern Pontic Steppe in the Bronze Age (Poltavka Culture).

The common opinion among the experts is that Grassmann’s law must have operated after the devoicing of aspirated consonants in Proto-Greek (and must therefore have occurred after the ancestor of Ancient Greek separated from the ancestor of Sanskrit). The reason for this is that the reflex of the deaspirated consonant in Ancient Greek is usually voiceless rather than voiced. Three examples of this are given at the beginning of the Wikipedia article on Grassmann’s law: https://en.wikipedia.org/wiki/Grassmann%27s_law

I am of the opinion that Grassmann’s law operated on the ancestor of Proto-Greek before the devoicing of the aspirated consonants, and that the Greek agathos is one instance of the effect of this law being retained unchanged.

In my view, it simply would have seemed bizarre to speakers of Proto-Greek to keep a voiced consonant in forms of a word that otherwise has a voiceless consonant (for example, **drikhes instead of trikhes as a plural of thriks ‘hair’, or **daphos instead of taphos ‘grave’ compared to thaptein ‘bury’). In the case of reduplication in verbs, the levelling was done by analogy with verbs whose roots originally had a plain voiceless stop.

It seems quite likely to me that there was regular levelling of these odd voiced consonants in Proto-Greek. But in the case of agathos, the voiced g produced by Grassmann’s law remains because a levelling to a voiceless k was not necessary. There are no forms of agathos or derivatives of agathos that have kh instead of g.