Imagine a transitivity spectrum of high object-affectedness at one end and high subject-affectedness at the other end.

When describing an event, there may be some freedom in where on the spectrum to go but for different choices, there’s an ordering of where they would be placed relatively on the spectrum. For example, consider:

• I broke the vase
• The vase broke
• The vase was broken by me

These three descriptions of the same event would be placed, relatively, from left to right on the spectrum.

Now consider each of the following pairs. If being used to describe the same event, the first of the pair would be placed on the spectrum (again, relatively) to the left of the second of the pair:

• take / choose (choosing might just be a mental decision but taking involves action)
• destroy / perish
• resolve / deliberate (resolve is a more active step beyond merely deliberating)
• stop / cease
• honor / value (you might value something but honoring it is taking action in response to that value)
• show / appear (you can just appear but you can also actively show someone)

Now in the imperfective, Greek offers two sets of endings that can (and I stress can) be used to capture the distinction between more to the left and more to the right on the spectrum. In the perfective, Greek offers three sets of endings.

However, where the line is drawn between these two or three segments of the spectrum to map them to the different endings is somewhat arbitrary between different words and it isn’t always directly comparable between different tense-aspect forms either. A single set of endings might cover a pretty large part of the spectrum. There is also no “requirement” that a single lexeme use all ending sets available, either. Instead, voice is available as a potential way of conveying the kinds of distinctions in the pairs above and in the three-way distinction in the vase example.

Where distinctions don’t need to be made, it should not surprise us to find only “middle” forms in use, especially in cases of lower object affectedness (like in mental verbs). This does mean in the imperfective there is not a separate form for a passive but passivization is less useful (and hence less likely) in these cases. But it should also not surprise us if some mental verbs use active forms.

It should also not surprise us to find, say, the future using the middle where the present uses the active. If the imperfectives only need a two-way distinction, the perfectives can also make just a two-way distinction even if choosing to use the two middle-passive forms to do so.

And if only a one-way distinction is required, there is nothing odd about a lexical item choosing to use a particular one of any of the three available voice endings (although we would expect broad tendencies to be based on object-affectedness).

The “active” is often described as unmarked with the “middle” marked for subject-affectedness but I think it’s actually helpful to think less about markedness and more about this transitivity spectrum of relative object-affectedness vs subject-affectedness. One can then think of voice as a largely lexically-determined tool for making relative contrasts on this spectrum.

This way of thinking means that the names of voices should probably not be so absolute but somehow be expressed in purely relative terms. The use of “middle” for the middle of the three isn’t bad but “active” and “passive” are highly misleading although they are “more active” and “more passive” than the “middle” when directly contrasting within the same lexeme.