The middle voice

From Glottopedia
Revision as of 09:32, 30 March 2008 by Sven Siegmund (talk | contribs) (+cat syntax +cat valency)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search
This is a survey article. For the corresponding dictionary article, see middle voice.

Based on her typological study, Kemmer (1993) sees the middle voice as a verb form denoting a transitive situation conceptualized as a single entity acting on itself, being both actor and undergoer. Reflexive forms, in contrast, denote situations conceptualized as one complex entity (or two separate entities) where the actor(-part of the entity) is acting on the undergoer(-part of the entity) and actor is coreferential with undergoer.

In other words, what a middle voice marker marks is a transitive situation performed by a single entity on itself (or for itself) whereas a reflexive marker marks the coreferentiality of two entities in a transitive situaton.

Often in a language, the use of middle marker is extended to mark anticausative situations, where the semantic role of the agent or the initator of the event is downplayed od virtually non-existent. Also often, particularly in languages without special middle marker, the use of reflexive markers is exteded to middle situations.


Classical Greek (Greek, Indo-European):

λούο-μαι τάς χεῖρας
loúo-mai tās cheĩras
‘I wash my hands.’

Semantics of Middle situations

Body actions is a large group of situation types which in language after language are coded somehow special. Kemmer divides body actions in four subgroups: self-induced motion (go, walk, fly, ...), change in body posture (lay down, sit down, stand up, ...), non-translational motion (twist, bow, stretch, ...) and grooming (wash, dress, shave, ...). The former are more like intransitive events, the latter transitive events.

The table below which is intended to visualize the spectrum of body actions uses the macroroles Initiator (Agent or Experiencer) and Endpoint (Patient, Recipient or Beneficiary)

self-induced motion change in body posture nontransl. motion grooming
Endpoint self-partitipation great some more some low
Conceptualized as action
towards other person
even more unlikely unlikely likely even more likely
Endpoint = Initiator required ... ... expected

Languages differ in how they employ different markers to treat this continuum. The left edge of the table is likely to be coded by simple intransitive verbs, the right edge is likely to be coded as transitive verbs with reflexive morphemes to signal the coreference of Initiator and Endpoint. If a language has a distinct middle voice marker, it will employ it somewhere in between.

The degree of distinguishability of participants is an important feature which distinguishes one-participant-events, middle events, reflexive events and two-participant-events.

Middle Reflexive Two-partici-
Participant distinguishability -- - + ++
Number of participants 1 1 1 2
Number of semantic roles 1 2 2 2
Conceptualization Referential entity is conceptualized as single entity Referential entity is conceptualized as single entity Referential entity is conceptually split into subparts, the one acting on the another (asymmetric interaction) Asymmetric interaction or relation

Middle marking systems

Languages mark middle situations differently. Kemmer describes the following middle marking systems.

No middle marking

Middle situations are treated like any other, either intransitive or transitive.

One-form system

The reflexive marker is used to mark middle situations. German: sich verlieben ‘fall in love’ (In these systems there is an inevitable mismatch between syntax and semantics. The semantics is intransitive, whereas syntactically this is coded like a transitive.)

Two-form cognate system

The middle voice marker is a reduced form of the reflexive marker. Russian -sja > sebja

Two form non-cognate system

The Latin middle voice suffix -r is not related to the reflexive pronoun se

Compounded Reflexive system

This is one where the reflexive marker consists of the middle marker plus another morpheme. It is found in Fula, for example. The middle marker -o is contained in the reflexive marker (i)t-o.


Direct Middle

Russian (Slavic, Indo-European):

я каждый день мою-сь
Ja každyj den’ moju-s’
I every day wash-MID
‘I wash every day.’

Latin (Italic, Indo-European)

wash (oneself)

Indirect Middle

Indirect middles are forms of transitive or ditransitive verbs conceptualized with actor and beneficiary/recipient being a single entitiy (again, in contrast, indirect reflexive forms mark such verbs for the corefentiality of actor and beneficiary/recipient, conceptualized as two separate parts of a complex entity or as two entities.)

Classical Greek (Kemmer 1993, Table 7):

hairei-sthai, (hairein ‘take’)
‘choose’ (grammaticalization of: ‘take for oneself’)

Turkish (Kemmer 1993, Table 7):

‘acquire’ (grammaticalization of: ‘do/make for oneself’)

Logophoric Middle

A rare phenomenon, reported only for Old Norse and Modern Icelandic is the logophoric middle, where a middle marked verb form is used to signal the coreference of actor-like roles between a matrix-clause and a dependent clause.

The construction is similar to the logophoric reflexive constructions like this one:

Old Norse (Nygaard 1905:195 cited in Kemmer 1993: 91)

Svasi kvað sik vera þann Finn-inn
Svasi said REFL.ACC that.ACC Finn.ACC.DEF
‘Svasii said hei was that Finn.’ (used in contrastive contexts: He himself and not the other guy.)

The logophoric middle construction does not have the reflexive pronoun, instead, the verb is in the middle voice and the nominal complement is in the nominative case.

Old Norse (Dyvik 1983:93 cited in Kemmer 1993: 91)

þú sagði-sk vera goðr læknir.
You said-MID good.NOM doctor.NOM
‘You said that you were a good doctor.’

Naturally Reciprocal Middle

In languages which have a distinct middle voice marker, it is used to mark naturally reciprocal events. These events are necessarily or very frequently semantically reciprocal (‘meet’, ‘fight’, ‘kiss’). The other transitive verbs which can be marked by the normal reciprocal construction cannot usually take the middle marker, because the semantics is not naturally reciprocal (and if the verb can take both reciprocal and the middle marker, the function of the reciprocal marking is contrastive or emphatic).

Examples of the natural reciprocal middle from Kemmer (1993, table 9) are Bahasa Indonesia ber-kelahi ‘quarrel’ ber-tjakap ‘converse (tjakap ‘talk’) Latin osculo-r ‘kiss’, amplecto-r ‘embrace’, Sanskrit saṃvadat-e ‘speak together’, Old Norse hitta-sk ‘meet’.

Other Middle Situation Types

Verbs of emotion (also emotive speech actions), cognition and perception are in many languages likely to occur in the middle voice, often being deponent (i.e. these verbs can occur only middle marked). Their affinities to middle voice are

  • the inherent affectedness of the initiator (an actor-like semantic role)
  • no conceptual separation of the initiator and endpoint (an undergoer-like semantic role). Both are conceptually one and the same entity: the human mind.

Examples of such verb forms are found in Kemmer (1993: 131–140). Here are some of them: Old Norse hata-sk ‘hate’, Latin misereo-r, Changana ku ti-tsakela ‘be/bacome glad’, Turkish döv-ün- ‘lament’, Bahasa Indonesia ber-pikir ‘be cogitating’, Classical Greek hypischnei-sthai ‘undertake, promise’.

Passive-like middle is an older term for what we would nowadays call anticausative. In languages without a distinct anticausative marker, the middle voice marker or the reflexive marker is used to express anticausative. Examples of such anticausatives are


The book sells well.


Cela ne se dit.
That simply isn’t said.
One does not say that.

Kanuri (Kemmer 1993: 147)

It’s not eaten/edible

Spontaneous events are commonly expressed by verbs in the middle voice. They are understood as processes ocurring without human cause. For instance growing, melting, rusting. The difference between these situations and the typical middle situations is the lack of a volitional initation of the event by the Endpoint. Some examples: Old Norse gróa-sk ‘grow’, Latin morio-r ‘die’, Sanskrit jāyat-e ‘be born’.


  • Dyvik, Helge. 1983: ???
  • Kemmer, Suzanne. 1993: The Middle Voice. John Benjamins Publishing Company, Amsterdam/Philadelphia. Series: Typological Studies in Language (TSL), Givón, T. et al. (eds.)
  • Kissling, Hans Joachim. 1960. Osmanisch-Türkische Grammatik (Porta Linguarum Orientalium, neue Serie III.) Wiesbaden. Harrassowitz.
  • Nygaard, Marius. 1905. Norrøn syntax. Kristiania: Aschehoug.