Here's the passage, with the grammatical terms in italics and the examples in bold:
For the purposes of discussion, I would like to toss out here that Thrax is conceptualizing diathesis in terms of a prototypical transitive construction, using the verb τύπτω as the example. For the grammatical form where the subject is the Agent, he calls it ἐνέργεια. For the grammatical form where the subject is the Patient, he calls it πάθος. For those forms that don't fit this conception, he gives up and calls it μεσότης. In his example, we have Patient-like subjects of ἐνέργεια-looking perfects πέπηγα διέφθορα, and we have Agent-like subjects of πάθος-looking aorist middles ἐποιησάμην ἐγραψάμην.Dionysius Thrax wrote:διαθέσεις εἰσὶ τρεῖς, ἐνέργεια, πάθος, μεσότης· ἐνέργεια μὲν οἷον τύπτω, πάθος δὲ οἷον τύπτομαι, μεσότης δὲ ἡ ποτὲ μὲν ἐνέργειαν ποτὲ δὲ πάθος παριστᾶσα, οἷον πέπηγα διέφθορα ἐποιησάμην ἐγραψάμην.
All in all, a fairly unsophisticated analysis of the Greek voice system, not used today and not particularly insightful. I would like to suggest, though, that there is one area we have regressed over Thrax, though. We don't really have a clean category for these non-kappa, non-aspirated perfects that seem to change diathesis like πέπηγα διέφθορα. But of these perfects have either aspirated (πέπηχα) or kappa (διέφθαρκα) forms that are "better" behaved diathetically. It is probably not good to lump them in with the middles, as Thrax did, but it is also not good to basically ignore them.