You are on page 1of 9
Passives and inverse constructions Chad Thompson Indiana University, Purdue University at Fort Wayne 1. Introdection ‘Voice constructions such pssives and antipasives can be sen a8 ‘being related to eer constructions on tw different levels. On the one level {s their relationship 10 other constucions which suppress arguments an ‘make the verd less acive. Among these are reflexhes, “middes”, Entcausatives,impersonas, and the Te. On the other level, pasves and antpasives ate related to constructions marking the relative topicality of the agent and 4 non-agent. Most notable among, hese i the inverse con- struction, bot their are others which areas yet unnamed. “Givén (1981) claims thatthe three functions of passives are fo assign {opie sats to non-agent, suppress the agent, and 0 make te verb ess five. Shibtani (1985), onthe other hand, claims that passives are bsi- ally defined only by the suppression of the agent nthe following paper, T ‘iscuss the various types of constructions by which arguments can be sup pressed or promoted, ignoring forthe ost part the issue ofthe depree 10 ‘which the verb active or satve 1 claim, alter Shibetani (1985), that passives are marked by the sup pression ofthe agent. Inverse consiructions on the other hand, are defined. by ther functional (ossly ot structural) promotion of non-agent, Some: ‘one defining passives ap agent suppressing constructions, however, should fot ignore the fact that many languages do use pasies 10 assign topic ‘as to nom agent, aor te fact hat passives in many languages are more ‘Statve that their ative sousterpars, One can 83, however, that in general voice systems (active, pusive, antipasie) are defined by the degree 10 ‘which they suppress arguments, while dizetion sjtems (dec, inverse) nm Chad Thompson ate defined by the degroe to which a nom agent hasan increas in opi ‘over the normal now-aget. If one looks at voice and diection from the point of view of relative topicality, ce terms ative, passive, anipasive, direct, and inverse may be insafcent. These terms would be suficient for most languages, bat i ‘conceivable that some languages would have more constructions related t9 relative topicality than these five terms would cover. I ict, T have came ‘sewhere (Thompson, 1985c) tht Navajo just sich a language "An agent will noally be more topical than a nom-agent ithe sme claus bu the non-agent, especially a patent, wills have some degree of jt. Classes having this relative topicality wil most ikely be active or tie, Any deviation fom ths norm may be marked by some special co struction. Figure below gives the posible increases an decreases inthe relative topicality and the range which passives, antipassives, and inverse onsructions normally cover. Noi that there is some overlap in these ‘The question, then, of what the precise definition of a pase or simi- lar constriction i depends to lange degre on the language under ques tion. In language where passives full ll ree of Givon'sRietions would be a mistake to define the passive oaly as an agen-soppretsing con- struction. On the eter hand, n languages sich os German, where passives fan be made of iavansitivs, it would make no sense to sped of sich pas Ak tie fom | eee uo Noo Ady A al Anipesne fo ao hs ae Pr Reet 1 5 bes mataly 1: Someta, Puce 1. Possible pity elton teen agen an patent Passive and inverse consrucions 2 ‘ves at asining topic status to patient. Furthermore, for languages such se Navajo one may have to speak of several diferent types of pasive (Thompson, 198%; Young and Morgan, 1980), Tn te following sectons, I discus the inverse in fall wide varity of| languages. I discuss in some detail Koyukon (Athabaskan), Pins Cree (Algonquian), and Chepang (Tibeto-Burman). ‘To save time and space, however, {do not discuss these languages in reat detail, but refer the reader 0 the relevant publications 2. Methodology — diagnosing inverse dauses | oui that there are no structural indicators of inverse constructions ‘upon which one can totaly depend: I discuss below a numberof languages in which the inverse has been observed, When viewed ast whole, these sages lack any consistent structural characterises. All that hey share are fopical direct objets, Thete are, however, cers structural diagnosis ‘whic often flow from the function of inverse constructions and which five some indication, usualy vali, that particular strctoreis inverse and not pasive, Fir, while pasives tend to be promotional, inverse constuctons ‘often are aot. Sevond, the agent is nok generally stuctraly suppressed weve constructions: it doesnot appear in an oblique phrase, and itis fener just as ikely tobe a ful sae it sina activeldnect cause, Third, the ver is no more sative nor les Wanslive in inverse constructions chan icisinacivelditect clauses. Finally, while passives tend only to assign topic status to a patient or direct objec, inverse constructions are more likely 10 mark the topicality of any object ‘Given the very broad definition ofthe inverse — it marks an increase in topicality of non-agent —and ofthe passive — itmarks a suppressed ‘gent — one would expect to find languages in which a parca eoastuc- tion & ambiguous as to whether itis pasive or inverse. Chepang and ‘Chamotro (Cooreman, 1984, 1987) ate probably two languages with such intermediate forms. (Given the lack of structural agnostics for inverse constructions and the fact the functions cannot drely be measured, one mus esrt to some ‘ype of indirect measurement of the functions of inverse construction. In the following paper Ilook at previous work which has performed the nece- Fy Chad Thompson sary text counts to ascertain the average relative topicality ofthe agent and ‘om agent in the constustions in question ‘A topical aepument is one which it more important andor more pre itable. ‘The contributors to Givin (ed, 1988) use the measurements of topic persistence to determine local importance. An argument’ topic pet sistence is measured by counting the umber of clauses out of the next ten in which a coreferental argument appears. This measurement is based on the assumption that he more affen a character appears ina given stetch of text the more important itis in chat stretch, Tope persistence as fist wed in Givin (e., 1988) measured the number of following consecutive causes having a coreferenalargumeat. The current method allows for & more ‘oticeable spread inthe average count. Global (extwide) and generic (cultreowide) importance are mea- sured by other methods, some of which ae discussed in Thompson (1989). ‘All ofthe quantitative work by Thompson (1987, 1989, 1989, 1989) uses Some measurement to address these two levels of importance, Since the result for global and generic importance generally epliate those for local importance, Iwill aot be giving the resus inthe Following discussion, but rather refer the rears t the appropiate studs "The conebutors to Giv6n (ed, 19K3)ased the measurement af afer ntl distance to determine the predictability ofan argument. The releren- tial sane ofan argument i calculated by counting the number of clauses back uni one finds coreferental argumeat I coeferental arguments found in the preceding clause, the reterential dstance i ons is coreleren- Gil arguments not found unt our clauses back, the referential distance is four, and so on. The arbitrary boundary ofthe measurement is twenty. If fan argument s new information o if as not been mentioned for ewenty fr moe causes, then the referential distance i twenty. [have found the ‘measurement for importance, topic persistence, to be more vsefl in lok. ‘ng at inverse constrictions than the measurement for predictability, but 1 sve both measurements below where avilable. More discussion on these vo measurements ean he found in the appropiate references, patel Jn Givdn (e., 1988) and Thompson (19598) an inverse construction marks a propesiion in which x nom ageot i ‘more topical than the typical nonagent indirect clauses, then one would expect othe average nomagent in an inverse clause to havea greater topic potsstence (and perhaps a lese referential distance) than the sverage no gent i diner lau. Further f inverse constructions do not sappess Pasive and inverse consrctons st agents, then we Would expect to observe two things in inverse causes con ‘cerning: (1) an inverse agent's average top petsitence should not be rch smaller than the average topic of agents in direct clauses (ne should ‘the referent distance be mush greater), and (2) the agem should be coded ‘bya non-oblique no just as often a indirect clauses. 23.4 The inverse in Koyukon (Athabaskan) In my previous dscasion of Koyukon (Thompson, 19894 and 19896), 1 contrasted inverse constructions with dtecvactive ones, aswell as with impersonals and agentless passives. summarize the more salient results in this ston ‘When both subject and object ae third person, actveldireet clauses tke 4 0 marked subject and ye- (~pi) pronominal object as can be seca in (2) and (8) below. When the object coded by an NP, no Je ref 4s present, a seen i example (1) (1) John gine neence Oyo John to opny50 wall “Sihe walked up to John fctiveliet] 2) Jom yuegh nee-nee-0-30 John xo-0 up-revsse-walk “John walked up to him/her’ [actvelaiet] O) yesh neener-P-y0 oso uprrvisc-wall "Sihe walked up to hie’ [actvedieet] th verse clauses, othe other hand, the subjects coded with ye- and the object with Be, as can be seea in (4) and (8) Below. (© John bush neeneexo Joh too uperrvavatk “Jol walked up to ihr’ inverse} () burgh newyee- nee Oo upssc-reY Prva. ‘she walked upto hier finvere] "Examples ofthe agentes passive and the impersonal ae given below, Ja the pastive (7), the agent has Been suppressed and cannot appear as an 1. The impersonal pronominal subject prefix can, depending on the

You might also like