J. Huang (1999), from a comparative perspective, proposes that three properties unify the notion of the passive construction.
(85) Properties of a Universal Notion of Passive (J. Huang 1999: 481 (113):
A. Intransitivization:
The English morphological passive intransitivizes by eliminating the external argument of the predicate; the East Asian passive (of the type represented by Mandarin long passive) intransitivizes by turning the internal argument into a lambda predicate.
B. Argument promotion:
The English be passive does this by moving the object to the subject position;
in the East Asian passive, an Experiencer NP comes to occupy the surface subject position as a result of inchoativization.
C. A missing NP position in the predicate coindexed with the subject:
In the English be-passive an NP trace is created as a result of the promotion process in (113b); in the East Asian passive, this configuration (with an A’
trace or resumptive pronoun) is created by NOP movement followed by predication.
According to (85), the English be-passive is thus a prototypical passive that exhibits all three properties. However, for the English be-passive, the property of (85c) only exists within a movement-based framework. Furthermore, note that even within a derivational approach (85c) is redundant as it is merely the consequence of the promotion of an internal argument, which is part of (85b). Thus, (85c) need not be taken as a universal property of the passive. Next, we examine the property in (85b).
Here again, once we ignore the movement-based characterization, the object’s promotion is merely the consequence of (85a), i.e., the elimination of the external argument, or the logical subject. As pointed out by an anonymous reviewer, the object’s promotion is merely one logical possibility. Indeed, there are languages, for example, Ukraiinian, Kannada, and Irish, where the object in (impersonal) passive is not promoted and retains the accusative case (Carnie 2002: 250-251). Therefore, (85b) also need not, and in fact should not, be seen as a defining property of the passive. All that is needed is (85a), i.e., the ‘elimination’ of the logical subject. However, a better characterization is ‘demotion’ rather than ‘elimination’, since the external argument, or the logical subject, is no longer linked to the subject, the most prominent grammatical function and yet may still surface as a non-subject element.
The single syntactic property that unifies the notion of passive universally thus boils down to the demotion of the logical subject in terms of its syntactic assignment.
This is further confirmed by the functionality of the passive. Givón (2006: 338), for instance, defines the passive clause functionally as ‘the clause-type whereby the agent of the corresponding active is radically de-topicalized and its patient becomes, by default, the only topical argument’. Similarly, Shibatani (1985) describes the prototypical function of passivization as ‘agent de-focusing’. Passivization is therefore crucially different from straightforward topicalization or clefting of the patient, which does not involve the demotion of the logical subject, or the agent (e.g., Keenan and Dryer 2007). The essential property of the passive is thus solely the
de-topicalization of the agent; the topicalization of the patient is only the consequence.57
There are two syntactic options for the demoted external argument in the passive:
It may be overt and appear as a non-subject element, or it may be covert and thus receives null syntactic assignment. There are in turn three possibilities for the demoted logical subject’s overt assignment: a non-subject argument, an adjunct, or an element incorporated into the passive verb (e.g., Keenan and Dryer 2007). A non-subject argument may in turn be either a term (or core) or non-term (noncore) function. Thus, it seems the full range of syntactic choices, other than the subject, is available for the demoted agent. We can conceptualize this range in terms of the
‘depth’ of its demotion, as represented in (86).
(86) Syntactic Choices for the Passivized Logical Subject 1. Term (e.g., object)
2. Non-term (e.g., oblique) Depth of demotion 3. Adjunct
4. Verb-incorporated element 5. Null
All these choices are manifested in the world’s languages. We will illustrate each choice with one or two examples, starting from the bottom. In the English short passive, also known as the truncated passive, the logical subject is syntactically null.
In the Mandarin unmarked passive (e.g., Hsueh 1989, Tan 1991), as shown in (87), as well as the lexicalized bei-V passives in (88) (which do not have an exact active counterpart in modern Mandarin given that the relic V within is no longer a free morpheme), the agent is in fact inadmissible. Likewise, in Tarahumara, an indigenous language in the Uto-Aztecan family of northern Mexico, the demoted agent is entirely inadmissible in the passive. Example in (89) is taken from Valdez (2004), quoted in Givón (2006).
(87) a. Tamen chai-le Lisi-de fangzi. (Mandarin) they demolish-PERF Lee-POSS house
'They demolished Lee’s house.' b. Lisi-de fangzi chai-le.
Lee-POSS house demolished-PERF 'Lee’s house was demolished (*by them).'
57Keenan and Dryer (2007: 325), on the other hand, characterize the function of passives as to
“topicalize” (“fore-ground”, “draw our attention to”) an element…which is not normally presented as topical in the active’. However, I believe Givón’s (2006) view is more insightful, which, for example, is more consistent with the feminist Julia Penelope’s (1990) proposition, made in the book Speaking Freely: Unlearning the Lies of Our Fathers' Tongues, that the passive form, by hiding or placing the actor in the background and the victim in the foreground, e.g., ‘the woman was raped’, is used to obscure agency. The key is thus the drastically reduced prominence of the agent. See also Bohner (2001) for more recent studies on this issue.
(88) a. Tamen zhongyu buhuo xianfan. (Mandarin) they finally capture suspect
‘They finally captured the suspect.’
b. Xianfan zhongyu bei-bu.
suspect finally captured
‘The suspect was finally captured (*by them).’
(89) a. Muéka nechí choná-re. (Tarahumaran) you-ENF me/OBJ hit-PERF
'You hit me.' b. Né-ka choná-ru-re I-ENF hit-PASS-PERF 'I was hit.'
English allows a very limited set of verbs, roughly those expressing authority, to incorporate the agent into the passive form, as in (90), but agent-incorporation is more productive in the Quechua passive, shown in (91) (Keenan and Dryer 2007: 345).
(90) a. The State/NSF/government funds this project.
b. This project is State/NSF/government-funded.
(91) a. Kuru-ø manzana-ta miku-rqa-n. (Quechua) bug-subj apple-do eat-past-3
‘The bug ate the apple.’
b. Kuru miku-sqa-mi manzana-ø ka-rqa-n.
bug eat-ptcpl-comment apple-subj be-past-3 ‘The apple was bug-eaten.’
A similar case may in fact be made of the Taiwanese short passive. According to J. Huang (1999), the apparent short passive of (92a) can only mean that I was hit by a definite third person singular individual as the agent. This incorporated or contracted agent can however appear overtly, as in (92b). In other words, (92a) can only have the meaning of (92b). Likewise, in (93b), the passive auxiliary hong incorporates an indefinite third person as the agent and is the contracted form of (93a).
(92) a. Goa ho pha-tio a. (Taiwanese) I HO hit PRT
I was hit by him/her.
b. Goa ho i pha-tio a.
I HO s/he hit PRT I was hit by him/her.
(93) a. Goa ho lang pha-tio a. (Taiwanese) I HO someone hit PRT
I was hit by someone.
b. Goa hong pha-tio a.
I HONG hit PRT I was hit by someone.
In the English passive, the demoted agent may appear as a syntactic adjunct, in the form of the by-phrase. Some researchers (e.g., Morimoto 1999; Crouch et al 2004) treat the English by-phrase as an oblique argument; however, it should be clear that it is an adjunct. An adjunct is typically optional and appears further away from the head than arguments. As shown in (94) below, the fact that the agent by-phrase is optional and that it appears further from the head than other attested adjunct phrases indicates that it is an adjunct, not a syntactic argument.
(94) a. John kissed Mary (repeatedly) (in the park) (yesterday).
b. Mary was kissed (repeatedly) (in the park) (yesterday) (by John).
Many languages have the demoted agent as a syntactically oblique argument. In the Guarijío example in (95), from Medina-Murillo (2004) as quoted in Givón (2006), the demoted agent still receives syntactic argumenthood and appears as an instrument-marking oblique GF.
(95) a. Owéru wicho-ré wakiá. (Guarijío) women wash-PFV clothes
'The women washed the clothes' b. Wicho-ré-tu wakirá owéru-e.
wash-PFV-PASS clothes women-INST 'The clothes were washed by the women.'
The Guarijío, also a Uto-Aztecan language, is closely related to Tarahumara (see example in (89)), but unlike the latter, it permits the overt expression of the agent via an oblique function. Finally, as Keenan and Dryer (2007: 344) have observed, the overt agent phrase may have no adposition at all, and cited Southeast Asian languages as examples, e.g., Thai, Vietnamese, as well as Mandarin. An example from Haya, a Bantu language, is given in (96), taken from Duranti and Byarushengo (1977), where they suggest that the unmarked agent appears to be an (applicative) object. This is similar to the object status we have argued for, for the unmarked agent phrase in the Mandarin bei passive, shown in (97).
(96) Ebitooke b´ı-ka-cumb-w’ omukˆazi. (Hava) bananas they-past-cook-pass woman
‘The bananas were cooked by the woman.’
(97) Na jian fangzi bei Lisi chai-le. (Mandarin) that CL house BEI Lee demolish-PERF
‘That house got demolished by Lee.’
Finally, it is also important to point out a typological feature regarding the syntactic status of the patient role in the passive. Even though universally the passive involves the demotion of the logical subject, the consequential topicalization of the
patient does not necessarily mean its promotion to the subject syntactically, contra J.
Huang’s (85b). In fact, whether the non-agent topic of the passive clause is its nominative subject is the most general typological feature for passives, which distinguishes between promotional and non-promotional passives (Givón 2006).
English has both promotional and non-promotional passives, the latter in the form of impersonal passives, e.g., it is rumored that John is gay. We can thus conclude quite confidently that the only syntactic property that unifies the notion of passive is the demotion of the logical subject, and we now can determine objectively whether the unified bei construction fits in this universal notion of passive.