I feel like this statement deserves a slight bit of qualification:
All of this raising more complicated questions regarding the nature of the perfect and its diachronic trajectory in the coming centuries. How does the perfect interact with Ancient Greek voice patterns at a more fine-grained level? So far as I know, these questions have not been engaged with by anyone yet.
That looks like a thesis topic if anyone is in need of one.
There is literature on this question relative to Greek and Indo-European, such as the following:
- Chantraine, Pierre. 1927. Histoire du parfait grec. Paris.
- Claflin, Edith Frances. 1939. ‘Voice and the Ind-European Perfect.’ Language 15, 155-159.
- Kuryłowicz, Jerzy. 1932. ‘Les Désinences moyennes de l’indo-européen et du hittite.’ Bulletin de la Société de Linguistique de Paris 33:1–4.
- Renou, L. 1925. La valeur du parfait dans les hymnes védiques (Société de Lin- guistique de Paris, Collection linguistique, 18), Paris.
- Stang, Christian S. (1932). ‘Perfektum und Medium’. Norsk Tidsskrift for Sprogvidenskap 6: 29–39.
- Wackernagel, Jacob (1904) Studien zum griechischen Perfektum, Güttingen .
But these are all specifically about the early/old active-only perfects relative to PIE. The fact that there are middle-only perfects that develop in the later period and continue to exist into (through?) the Roman period is suggestive of something else entirely: that the semantics of the perfect were sufficiently recognizable that language users sought, to some extent, to still limit the perfect’s transitivity with certain classes of verbs, even while the perfect was extending to itself to more and more transitive events.
It is that particular phenomenon that, so far as I have read, is not discussed. Most studies of the perfect are focused on explaining its meaning. Similarly when its paradigmatic structure is examined, it is done in seclusion or, at best, in terms of how speakers begin to use aorist agreement suffixes with it.