A question about voice came up on Facebook about εἰμί and γίνομαι and why they take active and middle morphology for their respective person endings. There were a number of answers given that ranged from: “Linking verbs don’t have voice,” to, “Who knows…,” to, “Because history,” to, “Language is arbitrary.”

None of those are particularly satisfying answers to the question, “Why?”

Here’s my attempt at an answer to the question in a way that brings together language history with respect to these two verbs with voice semantics in a meaningful way.

Voice can be a difficult grammatical category to discuss because there is a sense in which it is not real; it is not a thing in itself. The term voice actually refers to a family of similar grammatical phenomena cross-linguistically. This distinguishes voice from tense or aspect where the larger metacategory is notionally consistent from language to language—for example languages that have tense marked on their verbs or auxiliary systems all express temporal reference. Cross-linguistically, tense is temporal reference/deixis. The situation with voice is more complex. There are a number of different kinds of voice systems. In English and many other languages, the system involves an alternation between active and passive constructions that is only relevant for transitive verbs, where the agent of the clause is demoted syntactically to a prepositional phrase (The window was broken by the thief). For such languages, there is no voice marking for intransitive verbs.

But there are other types of voice systems. In addition to the passive system, there are antipassive systems, inverse voice systems, “Philippine” voice systems, and middle voice systems. There is no cohesive cross-linguistic definition that unites them beyond a vague sense that they all involve event structure and transitivity.

Middle systems are unique compared to active-passive systems such as English in that voice alternations go beyond only transitive verbs to all types of verbs, including intransitive verbs. That is an important point for thinking about Greek voice vs. English voice. Because English voice doesn’t involve intransitive verbs, it is tempting to gloss over the difference and ascribe “active” meaning to the middle morphology of Greek intransitives. But the reality is that from the perspective of the English passive system, intransitives having active voice marking is just as unusual as them having middle voice marking. So when we encounter linking verbs with voice morphology, we ought not simply dismiss out of hand the morphology as irrelevant. Semantically speaking, linking verbs have more in common with intransitives than with transitives. They share a basic semantic relationship to the basic types of intransitive predication.

Intransitives fall into four classes (Stassen 1999). These are:

1. Intransitive Action
2. Locational Predication
3. Property State Predication
4. Class Membership Predication

As a general rule, intransitive middles predominantly fall into the first category of intransitive action (Rachel Aubrey 2016), though some intransitive middles express property state predications (especially perfect middle, which is a whole other can of worms and has the additional set of entailments from the perfect). This general rule is significant, however, when we consider εἰμί and γίνομαι. The latter is almost entirely used to processes/changes of state (the domain of category #1), which is the clear preference for middle verbs by a wide margin. In a similar vein, εἰμί’s usage is constrained to the other three types of predications: locational (is in Jerusalem), property state (is tall) and class membership predications (is Gentile).

Now, these two verbs are both incredibly old and their usage is heavily affected by their high frequency for centuries and their forms are effectively fossilized. But it is also no accident that the types of meanings each of them express, one with middle morphology and the other with active morphology correspond effectively one-to-one with the general preferences for other non-linking and low frequency verbs.

In sum, γίνομαι was lexicalized centuries before as middle-only because as a linking verb, it shares its functions with intransitive processes and changes of state—the ice melted (τήκεται) and the ice became (γίνεται) water are identical in their event structure. Likewise, εἰμί was lexicalized as active-only because it has more in common in its semantics with other intransitive active-only verbs, such as πεινάω ‘be hungry’—I’m not aware of an adjective ‘hungry’ that I could produce an εἰμί clause with, but you get the idea: using εἰμί to link a subject to a predicative adjective is a classic property state predication.