Middle Voice

[evmdbm]

Does anyone know of a good description anywhere of the middle voice? I know there is one in Ancient Greek, and doubtless in other languages too but can't get my head round the idea. Not sure how you can have something that is mid-way between passive and active.

[Kuchigakatai]

The thing about the middle voice is that it's a vague concept, and you only get a good definition of an instance of it when you look at a particular language with it.

The vague concept is that the middle voice is a verbal inflection or verbal construction that expresses the reflexive (doing something to yourself), which is, very importantly, inflected or built in a similar way to the way the active or passive voices are (so we're not talking about merely adding a reflexive pronoun like "(my)self" here).

Often, in a particular specific language, this reflexive voice is used to express some kind of passive too (even if there's a true proper passive beside it as well), or perhaps the language uses the same voice to express both reflexive and passive meanings. In both situations, you may also see it called the mediopassive.

Note that Ancient Greek only distinguishes the middle voice and the passive voice in the future and the (past-tense perfective) aorist tense-aspects (in all available moods), and otherwise it has a single mediopassive voice in all the other tense-aspects (present, imperfect, perfect, pluperfect, and the rare future perfect). So the middle-voice indicative aorist ἐλύσαμεν elȳ´sāmen is different from its passive-voice counterpart ἐλύθην ely´thēn, but indicative present λύομαι lȳ´omai is a mediopassive (it serves in both the middle and passive voices). In other languages, mediopassive may mean "a middle-voice form that in some particular situation serves as a secondary passive, while still different from the true passive-voice form".

[zompist]

I looked into this a bit when I was writing about Sanskrit. The middle goes back to Indo-European, but of course the Indo-Europeans neglected to leave us a grammar. My impression is that it develops into different things in different branches, and you're better off asking e.g. "what is the middle in Ancient Greek?" rather than "What is the middle?"

In Sanskrit, FWIW, the middle was originally a reflexive, but came to be used as a mere alternative form. Sanskrit ended up with a variety of forms with no real meaning difference, which was a great convenience in using its very strict metrical forms. There were a few verbs which were used only in the middle, e.g. jan- 'be born'. These suggest that the middle really did have a sort of "not active and not passive" meaning for some verbs. It's certainly a thing that many verbs don't really fit the prototype "thing A does something to thing B" of transitivity.

[bradrn]

I think that what chris_notts said in the ergativity thread is relevant here:

Depending on who's writing, voice seems to cover some mixture of:

1. Valency adjustment, both increasing (causative) and decreasing (passive, anti-passive)
2. Argument rearrangement (regardless of valency)
3. Verbal marking of argument roles and affectedness (middle voice, version ..., which may not change valency at all but do mark subject affectedness, e.g. "he built himself a house" vs "he built a house", or similarly with cutting hair, washing parts of one's body, ...)
4. Verbal marking of pragmatics and promotion to privileged roles for coreference and control purposes (again, may not involve detransitivisation)

In addition to the general vagueness of the middle voice (which I never understood either, until I read the posts in this thread), the usual description of being ‘between active and passive’ only makes sense with some of the definitions given above: if you think of passive as a valency-decreasing or argument-rearranging action, then there is no way to have something ‘between active and passive’, but it makes perfect sense when you consider voice in terms of affectedness (e.g. think of active as focusing on the action itself, and passive as focusing on the resulting state).

(By the way, in the same post, chris_notts mentioned that the non-Indo-European language Fula also has a middle voice. Does anyone know anything about that?)

[evmdbm]

That was what was confusing me - I do think of voice as valency changing. If therefore you delete the subject and promote the object to subject-hood, it's the passive - not the middle. If you don't, it's the active - not the middle. Affectedness does make sense and although I'd noted the extensive syncretism between the middle and passive in Greek, I'll have a look for a good description of it in Greek

Incidentally I think I read somewhere that deponent verbs in Latin (passive in form, but active in meaning) are derived from the Indo-European middle voice.

[Kuchigakatai]

In addition to the general vagueness of the middle voice (which I never understood either, until I read the posts in this thread), the usual description of being ‘between active and passive’ only makes sense with some of the definitions given above: if you think of passive as a valency-decreasing or argument-rearranging action, then there is no way to have something ‘between active and passive’, but it makes perfect sense when you consider voice in terms of affectedness (e.g. think of active as focusing on the action itself, and passive as focusing on the resulting state).

The description "between active and passive" works also in terms of control, like when the subject does do the action but ends up affecting itself without meaning to. This would be opposed to the active voice, where the subject has more control, and the passive voice, where it has no control and does no action at all but rather receives the effect of somebody else's probably intentional action.

Good examples of what I mean would be the Latin deponent verbs, which are verbs that only take "passive inflections" (which actually mostly descend from the IE middle voice), not allowing active-voice inflections. In spite of the "passive inflections", some of them can take a direct object in the accusative. Because they almost always have an English active-voice verb as a translation equivalent, Latin teachers love saying that they're "passive verbs (in form) that are active in meaning", but that's just a comment about English translations (and believe me when I tell you this explanation generates controversy now and then in Latin forums...). Many deponents involve accidental actions of little control in which the subject still does *something*.
nāscor - to be born
morior - to die
lābor - to fall by slipping
fateor - to confess sth, likely reluctantly (with accusative object)
orior - (of people) to wake up; (of the Sun, stars) to rise; (of rivers) to originate somewhere; (of protests, controversies, riots) to arise
expergīscor - (of people) to wake up
patior - to suffer/allow sth (with accusative object, i.e. receiving or allowing something bad voluntarily)
vereor - to revere sth sacred (w/ acc. obj., as in feel fear towards a god, cognate with English 'wary' and 'aware')
veneror - to revere sth sacred (w/ acc. obj.)

Some of the above are quite nicely opposed to other verbs that take the active voice (and have a passive transformation available), expressing greater control by the subject: mē expergefaciō (with a reflexive pronoun) 'to wake up, wake oneself up (voluntarily, as planned)', sinō and permittō 'to allow sth' (per-mittō lit. "through-send", synchronically looking very voluntary), colō 'to worship sth sacred'.

It is interesting that morior 'to die' can be shocking because of its [-control] morphology, and in contrast, all the euphemisms are active, as if the person had had some [+control] choice at the time of death: obeō ("go forwards"), intereō ("go through the middle"), exeō ("go out"), excēdō ("move out"), dēcēdō ("move away"), occumbō ("recline forwards"). (You see something similar in Spanish, where non-reflexive morir 'to die' is both more polite and calmer than reflexive morirse, the latter expressing greater suddenness and emotion.)

The speech acts are also interesting. The deponent ones focus more on the subject, but the active/passive dīcō on what is said. Notice for has little control yet again, and that although baubor/nūgor/fābulor/loquor are perfectly voluntary and in control, they're nevertheless basically atelic...
for - to barely say sth, as a baby or while dying; to say sth while possessed by a god or spirit (w/ acc. obj.)
auguror - to give a prophesy or the will of the gods using bird divination (intransitive)
baubor - (of dogs) to bark while happy (intrans., cf. active lātrō 'to bark while angry')
nūgor - to speak non-sensically or trivially (intrans.)
fābulor - to speak idly in good humour (intrans., from fābula 'tale', so lit. "to do tales")
loquor - to speak (typically without a direct object but takes a complement with dē 'about', focusing on the subject, occasionally w/ acc. obj.; antonym: taceō 'to keep silence')
garriō - to gossip (with active morphology!)
dīcō - to say sth, tell sb sth (typically something of importance, focusing on the object)

Naturally there are also some of those reflexives of affectedness you talk about, where the subject has an interest in what they get from doing the action:
sequor - to follow sth (w/ acc. obj., basically go after sth to get it or get to it for yourself)
adipīscor - to get sth for yourself (w/ acc. obj., historically lit. reach to sth for yourself)
fūror - to steal sth (w/ acc. obj.)
amplexor - to hug sb (w/ acc. obj.)
ōsculor - to kiss sb (w/ acc. obj.)
polliceor - to promise sb sth (w/ acc. obj. for the theme, dat. for the person sth is promised to)
hortor - to encourage sb to do sth (w/ acc. obj. for the person and subjunctive subclause)
nancīscor - to barely manage to get sth (w/ acc. obj.)
proficīscor - to start (no obj., lit. "to make yourself start acting forwards", pro-fic-īsc-or lit. "forwards-do-inchoative-middlevoice")
vēscor - to eat, to eat sth (w/ acc. obj.)
aquor - (of soldiers) to drink water (intransitive), typically while terribly thirsty

Relatedly, verbs with experiencer subjects that nevertheless involve the subject actively taking in and thinking about things:
reor - to think/reckon sth after consideration (w/ acc. obj.)
recordor - to remember and consider sth, call sth to mind (w/ acc. obj.)
opīnor - to hold the opinion that [S] (w/ sentential complement)
tueor - to watch sth to protect it (w/ acc. obj.)

The category of labile verbs in which inanimates can be said to relatively do certain actions, which in English are often ambitransitive with the causative, showing ergative-like behaviour ("the window broke", "I broke the window"; "I moved", "she moved me"), are generally not deponent, but sometimes do use passive inflections while still implying the subject does something within the action (even if with little control):
frangō - to break sth (w/ acc. obj.)
~ frangor - to break (intransitive)
quatiō - to shake sth (w/ acc. obj.)
~ quatior - to shake (intransitive)
moveō - to move sth (w/ acc. obj.)
~ mē moveō - to move (with a reflexive pronoun as the acc. obj., no separate theme argument)
~ moveor - to move (intransitive, synonym of mē moveō)
labefaciō - (of an earthquake) to shake sth (w/ acc. obj., typically a building, house, wall)
~ labefīō - (forced by an earthquake) to shake (intrans.; fīō is the passive form of faciō)
Again, the verbs here with passive morphology are actually in the middle voice and the subject is implied to actively do something. One of the Latin dictionaries I consulted gives this quote from Cicero: quod ipsum ex suā sponte movētur 'what moves by itself out of its own will', where movētur = moveor in the 3SG form. (Note how translating that as "what is moved by itself of its own will" makes no sense in English.) That could've also been sē movētur with a reflexive pronoun.

However, quite a lot of the usual labile verbs are rendered with wholly different stems, using derivational suffixes, or maybe with a reflexive pronoun in the intransitive usage:
fervefaciō - to boil sth (w/ acc. obj., causative compound of ferveō and faciō 'to do/make')
ferveō - to boil (intrans.)
sānō - to heal sth (w/ acc. obj.)
sānēscō - (of a wounded person or a wound) to heal (intrans., with the usually inchoative -ēscō)
ūrō - to burn sth (w/ acc. obj.)
ardeō - to burn (intrans.)
scindit - to tear sth, burst sth (w/ acc. obj.)
sē scindit - to tear, burst (intrans., with refl. pron.)




I think all the above notions are good to apply the middle voice to in conlangs if one is interested in that.

(By the way, in the same post, chris_notts mentioned that the non-Indo-European language Fula also has a middle voice. Does anyone know anything about that?)

I've seen quite a number of languages getting talked about as having a middle voice while incidentally reading about something else, but I don't pay attention to that much. Standard Arabic has a valency derivational affix (valency V, taCaC:aCa) that gets used a lot to express somewhat idiomatic reflexives, especially reflexives of the usually causative valency II (CaC:aCa), so I've seen valency V referred to as expressing the middle voice sometimes.

ʕalima - 'to know sth'
> ʕallama - 'to teach sb sth' ("make sb know sth")
>> taʕallama - 'to learn sth' ("teach sth to oneself", or well, "make oneself know sth")

ħasan - 'beautiful/handsome, good', ħasuna - 'to be beautiful/handsome, good'
> ħassana - 'to make sth good, improve sth'
>> taħassana - 'to improve' (intrans., "improve oneself", or well, "make oneself good")

qadama - 'to be before', qaduma - 'to be ancient'
> qaddama - 'to put sth at the front or before'
>> taqaddama - 'to be before sth; make progress' ("put oneself at/towards the front", or well, "make oneself be at/towards the front")

kalima - 'word, speech'
> kallama - 'to talk to sb' (valency II CaC:aCa derives transitive verbs from nouns too)
>> takallama - 'to speak; to speak [a language]' ("talk for oneself", transitive but focused on the subject)

ʃaraf - 'honour'
> ʃarrafa - 'to honour sb, give honours to sb'
>> taʃarrafa - 'to be honoured' ("give honours to oneself" through one's actions)

By the way, a related affix, valency VI taCaaCaCa, typically expresses reciprocals, especially reciprocals with a plural subject of the dative applicative valency III CaaCaCa (qaasama 'to share with sb' > taqaasama 'to share with each other'). Valency VI taCaaCaCa also creates verbs of pretending to do something ("making yourself be sth", as in maridˤa 'to be ill/sick' > tamaaradˤa 'to fake being ill/sick, make oneself be/appear sick', or ħaaʃaa (underlyingly ħ-aa-ʃ-a-y-a) 'to exclude sth' > taħaaʃaa 'to avoid/shun sth', that is "to make oneself exclude sth"). This suggests that the prefix ta- is probably a grammaticalized reflexive morpheme...

[vegfarandi]

Middle voice is used differently based on language specific criteria. The Icelandic middle voice comes from a fusion with the reflexive pronoun which in Old Icelandic was sik, became -sk, now -st in Modern Icelandic. The voice has a variety of uses. Sometimes it is reflexive in meaning:

Ég klæddi hana – I dressed her
Ég klæddist – I got dressed (I dressed myself)

Sometimes recoprocal:

Ég kyssti hana - I kissed her
Við kysstumst - We kissed (eachother)

Sometimes it causes less obvious semantic changes:

Ég kom – I came, I arrived
Ég komst – I made it

Hann fór – He went, he left
Hann fórst - He was killed (in an accident)

Sometimes it is essentially passive in meaning:

Það verður að segjast – It must be said (lit. it has to say itself)

Some verbs only exist in the middle:

Ég ferðast – I travel (there is no verb *ferða)

And some verbs have no middle. Essentially, the ending -st which forms the Icelandic middle lives in the grayzone between inflection and derivation. Arguments for inflection are that it commonly applies to verbs and through regular and predictable morphological processes. Arguments for derivation are that the resulting semantics are erratic and it does not apply to every single verb.

[chris_notts]

(By the way, in the same post, chris_notts mentioned that the non-Indo-European language Fula also has a middle voice. Does anyone know anything about that?)

My source for this was Klaiman's "Grammatical Voice". IIRC, he also presents a strong/weak verb alternation in some Dravidian languages that is reminiscent of an active/middle alternation. In all of these languages, there are examples of transitive middles and of roots whose middle inflected form is as transitive as its active form. Middle voice languages also often have reflexive pronouns or markers distinct from their middle voice. It seems fairly undeniable that, while there is an overlap between reflexive and middle related to the subject-affectedness of middle voices, the two are potentially semantically distinct and middles, unlike passives, are generally not purely valence rearrangement devices.

[Vardelm]

Not sure I have much useful to add, but since I had to wrap my head around the middle voice a while back, I'll chime in.

For me it helped to just think of the middle as the subject being affected by the verb, much like the passive voice, but information about the existence of an agent is either completely absent or it is understood to also be the subject. This can be either a reflexive, reciprocal, or autobenefactive. It also helped to just remember that English doesn't have a great way of phrasing some (many? most?) middle-voice translations.

I think confusion about the middle voice is driven by several factors. For native English speakers, the middle isn't distinctly marked and isn't a central feature. Languages that do have a middle voice use it in different manners, as other described. Beyond that, I think there are a bunch of linguistic topics surrounding the middle voice that are not labeled in a particularly consistent manner: voice, diathesis, accusative/ergative/unaccusative/unergative verbs, etc. These topics also cause some confusion when it comes to ergativity, too.

[So Haleza Grise]

I think confusion about the middle voice is driven by several factors. For native English speakers, the middle isn't distinctly marked and isn't a central feature. Languages that do have a middle voice use it in different manners, as other described.

English verbs do all kinds of odd an interesting things in terms of valence, it's just that we don't have a morphologically consistent way of indicating that. At the risk of labouring the point, it's language-specific and it will depend on how your language handles transitivity/valence issues generally. E.g a language that has a very strict transitive/intransitive division might need a special voice to handle detransitivisation and the like. You can't really come up with a category of "middle voice" without describing how valence works more broadly (same is true of "passive" or "antipassive").

A language like English that broadly is relaxed about valence categories will find it easier to do the things that a "middle voice" is used for (or any kind of valence-reducing operation) without necessarily resorting to specific syntactic or morphological constructions.

[bradrn]

I think confusion about the middle voice is driven by several factors. For native English speakers, the middle isn't distinctly marked and isn't a central feature. Languages that do have a middle voice use it in different manners, as other described.

English verbs do all kinds of odd an interesting things in terms of valence,

I’d be interested to know if you have any examples of these.

A language like English that broadly is relaxed about valence categories …

What do you mean by this? And what would a language which is stricter about ‘valence categories’ look like?

[So Haleza Grise]

I think confusion about the middle voice is driven by several factors. For native English speakers, the middle isn't distinctly marked and isn't a central feature. Languages that do have a middle voice use it in different manners, as other described.

English verbs do all kinds of odd an interesting things in terms of valence,

I’d be interested to know if you have any examples of these.

A language like English that broadly is relaxed about valence categories …

What do you mean by this? And what would a language which is stricter about ‘valence categories’ look like?

Perhaps "interesting" is overstating it because it's natural from the point of view of English. Basically, for a lot of languages, a verb is either transitive or it's intransitive but definitely not "either".

English verbs like wake, burn, cook, break, lift, boil, wash, shave can be used either transitively, or intransitively. In a lot of languages that's not allowed. Yukulta for example allows only the verb for "burn" to be used either transitively or intransitively. And it's easy to work out which is which, because in Yukulta transitive and intransitive verbs are conjugated differently and the nominal morphology is ergative (in a main clause; there's actually quite a complicated arrangement for other clauses but that's not relevant here). Yukulta is an outlier among Australian languages in having even one verb that works this way - in general, Australian languages have verbs that are either transitive, or intransitive.

In Yukulta you want to use a verb like "cook" you have to use it transitively; essentially you have to say "X is cooking the meat" (X can be unstated) but you can't say "the meat is cooking". That construction is very natural in English but it's impossible in Yukulta.

Nahuatl is another language where valence categories are stricter than in English. In Nahuatl there are different verbs for read depending on whether you mean "the inscription reads X" or "I read the inscription". There are valence-changing operations in Nahuatl - a causative, an applicative - but these are morphological. You can't just, as English does, pick up a verb like "sink" and then say either I sank something, or something sank. You can only use a different verb, or a morphological causative.

A "middle voice" (like the Latin deponent) operates to cover situations where strict valence categories get in the way. For example, in Australian languages broadly, a reflexive (e.g. "I washed myself") is counted as a special intransitive construction. The point I am trying to make is that English doesn't need a special morpheme for what in another language might count as a "middle" construction - e.g. "the jar broke" or "the child washed".

[bradrn]

English verbs do all kinds of odd an interesting things in terms of valence,

I’d be interested to know if you have any examples of these.

A language like English that broadly is relaxed about valence categories …

What do you mean by this? And what would a language which is stricter about ‘valence categories’ look like?

Perhaps "interesting" is overstating it because it's natural from the point of view of English. Basically, for a lot of languages, a verb is either transitive or it's intransitive but definitely not "either".

English verbs like wake, burn, cook, break, lift, boil, wash, shave can be used either transitively, or intransitively. In a lot of languages that's not allowed. Yukulta for example allows only the verb for "burn" to be used either transitively or intransitively. And it's easy to work out which is which, because in Yukulta transitive and intransitive verbs are conjugated differently and the nominal morphology is ergative (in a main clause; there's actually quite a complicated arrangement for other clauses but that's not relevant here). Yukulta is an outlier among Australian languages in having even one verb that works this way - in general, Australian languages have verbs that are either transitive, or intransitive.

In Yukulta you want to use a verb like "cook" you have to use it transitively; essentially you have to say "X is cooking the meat" (X can be unstated) but you can't say "the meat is cooking". That construction is very natural in English but it's impossible in Yukulta.

Nahuatl is another language where valence categories are stricter than in English. In Nahuatl there are different verbs for read depending on whether you mean "the inscription reads X" or "I read the inscription". There are valence-changing operations in Nahuatl - a causative, an applicative - but these are morphological. You can't just, as English does, pick up a verb like "sink" and then say either I sank something, or something sank. You can only use a different verb, or a morphological causative.

A "middle voice" (like the Latin deponent) operates to cover situations where strict valence categories get in the way. For example, in Australian languages broadly, a reflexive (e.g. "I washed myself") is counted as a special intransitive construction. The point I am trying to make is that English doesn't need a special morpheme for what in another language might count as a "middle" construction - e.g. "the jar broke" or "the child washed".

Thanks for clarifying! So you were just talking about the presence/absence of ambitransitive verbs and valency-changing morphemes then. (I did know all of this before, I just didn’t connect it to what you were saying — sorry for making you write all this!)

[So Haleza Grise]

That's OK! I should have expressed myself more clearly, I was too vague and tired.

[sasasha]

Just a little addition to Vegfarandi's Icelandic examples: it might be surmised from that that the middle voice wasn't present in Old Norse and developed only in Icelandic; I don't believe that is exactly correct, though the difference might be down to terminology (e.g. what exactly is meant by Old Icelandic and Old Norse).

The Old Norse middle voice was grammaticalised from both the reflexive pronouns 'mik' and 'sik', giving it separate 1st person singular realisations in -mk, as opposed to -sk, and AFAIR it is found early and was lost in Old East Norse (though I can't cite anything to back that up as my undergrad was a long time ago... apologies if I am misremembering.)

As Veg says, middle voice is a bit of a catch all term and varies from language to language, but the modern Icelandic way to express 'see you later' is a neat example that I often think about when conlanging: við sjáumst - literally 'we will see each other', with the middle voice -st suffix.

It think of it as suggesting 'we will see and be seen', i.e. encapsulating both active and passive meanings in one expression. This idea fits some of Veg's other examples:

I dressed and was dressed
We kissed and were kissed
I came and was come (to a place)
He went and was gone
It must say (itself) and be said

...perhaps suggesting why it can only be applied to certain verbs, because sometimes passive meanings are removed enough from active meanings that this reciprocity is blocked. Just a thought - might I be barking up the right tree, Vegfarandi?

[vegfarandi]

Just a little addition to Vegfarandi's Icelandic examples: it might be surmised from that that the middle voice wasn't present in Old Norse and developed only in Icelandic; I don't believe that is exactly correct, though the difference might be down to terminology (e.g. what exactly is meant by Old Icelandic and Old Norse).

The Old Norse middle voice was grammaticalised from both the reflexive pronouns 'mik' and 'sik', giving it separate 1st person singular realisations in -mk, as opposed to -sk, and AFAIR it is found early and was lost in Old East Norse (though I can't cite anything to back that up as my undergrad was a long time ago... apologies if I am misremembering.)

It's pretty certain that this development was underway as a form of the middle exists in all the Nordic languages. It's logical to assume there was at some point a -þk form for the 2nd person but I'm not sure that's attested. Icelandic and Faroese are the most conservative and retained the -mk forms the longest.

As Veg says, middle voice is a bit of a catch all term and varies from language to language, but the modern Icelandic way to express 'see you later' is a neat example that I often think about when conlanging: við sjáumst - literally 'we will see each other', with the middle voice -st suffix.

It think of it as suggesting 'we will see and be seen', i.e. encapsulating both active and passive meanings in one expression. This idea fits some of Veg's other examples:

I dressed and was dressed
We kissed and were kissed
I came and was come (to a place)
He went and was gone
It must say (itself) and be said

...perhaps suggesting why it can only be applied to certain verbs, because sometimes passive meanings are removed enough from active meanings that this reciprocity is blocked. Just a thought - might I be barking up the right tree, Vegfarandi?

Yeah, this reciprocalish meaning seems to be close to the core meaning of the category, though it certainly doesn't apply to all meanings, in particular with verbs missing an active form. I think by nature of the fuzzy core meaning of the category, the applied meaning tends to shift to more useful/common meanings or other things within the semantic neighborhood of this reciprocal notion.