Map 107A depicts the geographical distribution of passive constructions. Only two values are represented:
|Go to map|
|There is a passive construction||162|
|There is no passive construction||211|
A construction has been classified as passive if it displays the following five properties:
it contrasts with another constuction, the active;
the subject of the active corresponds to a non-obligatory oblique phrase of the passive or is not overtly expressed;
the subject of the passive, if there is one, corresponds to the direct object of the active;
the construction is pragmatically restricted relative to the active;
the construction displays some special morphological marking of the verb.
A prototypical example of the passive, as defined above, is given in (1b) from Swahili.
‘Hamisi cooked the/some food.’
‘The food was cooked (by Hamisi).’
(1b) is an example of a synthetic passive, where the lexical verb exhibits some form of marking, here the suffix -a, which is absent from the active. Synthetic passives contrast with periphrastic or analytical passives, in which the special verbal morphology involves the use of a participal form of the lexical verb and an additional auxiliary verb, as in the English translation of (1b) and also in the Polish (2b).
(2) Polish (own knowledge)
‘Intensive rain destroyed the harvest.’
‘The harvest was destroyed by intensive rain.’
In Swahili, Polish and English the subject of the active may be expressed in the form of an oblique constituent or remain unexpressed. In many languages only the latter is possible: the subject of the active cannot be overtly present in the passive.
The three examples of passive clauses given so far are personal passives, i.e. passives with an overt lexical subject. Personal passives are typically seen as involving a process of agent demotion (from subject to oblique role or total suppression) and a process of patient promotion (from direct object to subject). There are also passive clauses which involve only agent demotion. These are called impersonal passives. An example of an impersonal passive is given in (3b) from Kannada (Dravidian; southern India).
‘Someone denounced this resolution.’
‘This resolution was denounced.’
We see that the accusatively case-marked direct object nirNayavannu of the active (3a) retains its accusative case marking in the passive (3b). Moreover, the passive auxiliary a:gu ‘become’ is always in the third person singular neuter and thus shows no agreement with nirNayavannu. The direct object is thus not promoted to subject. This becomes even clearer when we compare the impersonal passive in (3b) with the Kannada personal passive in (4b), in which the direct object of the active appears in the nominative case and controls agreement with the passive auxiliary.
‘The boys flew the flag.’
fly.. . . .
‘The flag was flown by the boys.’
It is also important to note that in the Kannada impersonal passive, unlike the personal passive, it is not possible to express an overt agent. This, however, is not an integral feature of the impersonal passive. For instance, in Lithuanian, which also has both a personal and an impersonal passive, an overt agent can be expressed in both constructions. An example of the impersonal passive with an overt agent is given in (5b).
‘The children slept in the garden.’
‘The children slept in the garden.’ (Lit. By the children was being slept in the garden.)
In both Kannada and Lithuanian the impersonal passive co-exists with the personal passive. This is also the case in Dutch, German, Hindi, Icelandic, Spanish and Turkish. But there are languages which have only an impersonal passive, such as Kolami (Dravidian; Andhra Pradesh, India), Ute (Numic, Uto-Aztecan; Colorado) Tukang Besi (Western Malayo-Polynesian, Sulawesi; Sulawesi, Indonesia) and Zuni (isolate; New Mexico). Languages with only impersonal passives have been classified as exhibiting a passive on a par with languages with personal passives.
In languages which have no passive construction, agent demotion or suppression can be achieved by other means. Some languages simply allow the subject to be omitted. As shown in (6b), Awtuw (Sepik; Papua New Guinea) is such a language.
‘He has cooked food.’
‘Someone has cooked food.’
In other languages, what would be expressed in English by an agentless passive is rendered by the use of an explicit impersonal or indefinite subject, such as the German man or French on, or simply the word for ‘persons’ or ‘people’ as in, for instance, Gude (Chadic, Afro-Asiatic; Nigeria and Cameroon).
‘He was beaten up.’
Still other languages achieve the same end by using the third person singular or plural form of the verb. The latter is illustrated in (8) from Paamese (Oceanic; Vanuatu), in which the impersonal interpretation is confined to clauses lacking a corresponding third person plural independent pronoun.
‘There is drinking going on at Vauleli.’
There are several constructions which manifest some, but not all, of the five properties listed earler as definitional for passive constructions, and which therefore have not been classified here as passives. First, there are anticausative (or middle, or mediopassive) constructions, as in (9b) from Gumawana (Oceanic; Papua New Guinea).
‘Yesterday we trapped many fish.’
‘The fish are already trapped.’
Anticausative constructions resemble agentless passives in having a subject which is semantically a patient rather than an agent. However, in the anti-causative, unlike the passive, there is no covert agent. The situation or event is depicted as being brought about spontaneously without the involvement of an agent. That this is indeed so is evinced by the fact that it is not possible to add to an anticausative construction an agentive manner adverb such as deliberately or on purpose. Compare the English passive (10a) with the anticausative (10b).
The door was opened deliberately.
The door opened (*deliberately).
Second there are constructions called inverses (see, e.g., Cooreman 1987, Givón 1994). Inverse constructions are best known from the Algonquian languages, in which the direct voice is used if the agent is more topical or ontologically salient than the patient, and the inverse if the patient is more topical or ontologically salient than the agent. Traditionally the more topical or salient participant is called the proximate and the less salient or topical one the obviative. The direct/inverse opposition is illustrated in (11) from of Plains Cree (Algonquian; Canada).
‘The man scares the dog.’
‘The man scares the dog.’
The inverse is similar to the passive in functional-pragmatic terms (Givón 1994). In both constructions the patient is more topical than the agent. However, whereas in the passive the agent is extremely non-topical or indeed simply suppressed, in the inverse the agent retains considerable topicality. Accordingly, the two constructions differ with respect to the properties of the agent. The agent in the passive, if expressed, is a syntactic adjunct. In the inverse, on the other hand, it is a syntactic argument. This is evinced by the obligatoriness of the agent in the inverse as opposed to the passive and by the ability of the agent of the inverse, for example, to determine verbal agreement or participate in various syntactic processes. The properties of the agent have therefore been used here as criterial for distinguishing the passive from the inverse.
Given the above, I have also not treated as passive the so-called non-actor focus constructions in the Philippine languages, illustrated in (12b) from Cebuano (Western Malayo-Polynesian, Meso-Philippine; Philippines).
‘The man will buy a book.’
‘The man will buy the book.’
There has been a longstanding controversy on whether or not the Philippine focus system should be considered to be a voice opposition and if so of what type - active/passive, ergative/antipassive or even direct/inverse (see e.g. Siewierska 1984: 79-86 and the references cited there). The issue is a complex one and cannot be done justice to here. My main arguments against a passive analysis of non-actor focus clauses are: they exhibit a very high text frequency; the agent is typically overt and manifests some properties associated with syntactic arguments as opposed to adjuncts; they are semantically highly transitive in the sense of Hopper and Thompson (1980); and the verb does not exhibit special marking, as it is also explicitly marked in the actor focus construction. In sum, non-actor focus clauses do not appear to be pragmatically restricted vis-à-vis their actor-focus counterparts.
Passive constructions occur in 44 % of the languages in the sample. They are most common among the languages of Eurasia and Africa. They are also regularly found in the Americas, particularly North America. They are somewhat less frequent in Southeast Asia and the Pacific. In Australia they are attested only in a couple of Tangkic languages spoken in the Gulf of Carpentaria and a few Ngayarda languages in south-western Western Australia. In New Guinea they seem not to occur at all.
In Eurasia passives are frequent everywhere apart from the Caucasus and the Tibetan languages of India and Nepal. In Africa passives are highly common among the Nilo-Saharan languages, and only slightly less so in Afro-Asiatic. Of the Niger-Congo languages in the sample only about half display passive constructions. Passives are less frequent particularly around the coast of West Africa. In North America passives are found mainly in the western part of the continent. In South America they occur chiefly among the languages of the Amazon basin; they are particularly conspicuous by their absence among the languages along the west coast of the continent.