Chapter 2 of Aubrey’s thesis looks at the Greek middle in the context of other languages and their voice systems. How does the Greek middle, and the Greek voice system, ‘fit’ in comparison to other languages and their voices systems (23) rather than the traditional approach which begins from an active-passive contrast and goes from there.
Aubrey begins by distinguishing derived and basic voice systems. English is a derived system, because it is traced from a source, Greek is a basic system because it is not. To put it more simply, the passive in English takes the active structure and “remaps the participants” (25). Greek middles do not work like that – they do not presume the priority of the active and then go about remapping the participants in the event. Other languages are similarly basic or derived, and not all exhibit the same sets of systems (cf. anti-passives, inverse systems, etc. (24)).
Let me give an example (different to Aubrey’s) of that remapping
- Michael wrote (b) the tweets.
- The tweets were written (a) by Michael.
Here the agent (a) is remapped (demoted) to an optional oblique phrase, while the patient (b) is promoted to the subject of the verb. A consequence of having a derived system like this, Aubrey says, is that they can be no passive-only verbs – passives only arise by being derived from actives (26). Secondly, only transitive clauses (i.e. with direct objects) can be passivized. Intransitives cannot.
Thirdly, Aubrey points out that voice alternations in derived systems are “expected to be semantically neutral” (26) that is, switching patient from object to subject does not normally change the meaning of the verb, only the alignment of the participants.
Well, what about a basic middle system? Aubrey’s work here appears to draw primarily on Klaiman and Shibatani. So, in contrast, middles are not derived from their active counterparts – there is no ‘mesofication’ process that turns an active into a middle clause. The agent is still the subject, the patient remains the object. The alternation between active and middle rests in a semantic alternation (27).
Because (θ)η type middles are not ‘passive’, they overlap with -μαι type middles, and they are not derived from active prototypes. Because -μαι and (θ)η type middles are basic and not derived, one doesn’t need to explain middle-only and passive-only verbs. There is no ‘deponency’ problem.
Related to this, because the active-middle contrast is not about syntactic transitivity (e.g. rearranging agent/patient subject/object positions), it means that the middle voice is not restricted t a single set of transitivity. Hence, you find middles with one, and two, arguments. This is an important difference from the derived system, where going from active to passive involves losing an argument:
Michael (1) wrote the tweets (2)
The tweets (1) were written.
In the derived system, one cannot require a second argument. But middle systems can appear as transitive or intransitive, with 1 or 2 arguments .
Thirdly, in contrast to the derived system were a voice alternation is semantically neutral, in a basic system, they are not – shifting between active and middle is a semantic shift, not merely a syntactical rearrangement (30).
Aubrey concludes this subsection, “the descriptive problems in the Greek middle are due more to a misguided use of a derived passive system than to Greek voice operating differently than typologically expected in a basic middle system.” (31) Or, in simple terms, your problem all along was that you kept trying to fit Greek middles into an active<>passive mould, but when you look at Greek middles in light of other active<>middle voice languages, it’s not weird at all.