Cebuano Passives Revisited

Cebuano Passives Revisited

Tanangkingsing, Michael.Huang, Shuanfan.

Oceanic Linguistics, Volume 46, Number 2, December 2007, pp. 554-584(Article)

Published by University of Hawai'i PressDOI: 10.1353/ol.2008.0006

For additional information about this article

Access provided by National Taiwan University (8 May 2013 20:54 GMT)

http://muse.jhu.edu/journals/ol/summary/v046/46.2tanangkingsing.html

http://muse.jhu.edu/journals/ol/summary/v046/46.2tanangkingsing.html

Oceanic Linguistics, Volume 46, no. 2 (December 2007)© by University of Hawai�i Press. All rights reserved.

Cebuano Passives Revisited

Michael Tanangkingsing and Shuanfan Huang

national taiwan university

The view that the gi-clauses and/or their equivalents in other Philippine-typelanguages, speci²cally in Cebuano and closely related Bisayan languages,are active constructions has been widely accepted by a number of Austrone-sian linguists. In a recent study on the gi-verb clauses in Cebuano, however,another linguist reinterprets those with Verb-Patient-Agent (VPA) wordorder as passive. In this paper, we argue against such an interpretation, basedon analyses of the semantics and discourse pragmatics of the gi- and na-clauses in spoken data. A gi- attached to a verb implies a deliberate intentionof an Agent; a na-clause directs attention to the often accidental effect of anaction on a Patient without emphasizing any reference to an Agent. This ren-ders a na-verb construction, especially one in which the Agent is missing, asa much more plausible candidate for passive in Cebuano.

1. INTRODUCTION.1 Recent research has shown that passive constructions in theworld�s languages exhibit a wide range of structural diversity and that there is no singleproperty that all these constructions have in common (Siewierska 1985, Shibatani 1985,Croft 2001). This line of research has demonstrated that there is a rich structural contin-uum from the active to the passive, and that is why there is no universal single propertyfor, and no universally applicable de²nition of, passives. Rather than a separate universalcategory, passive is probably more accurately characterized as a syndrome of featuresand phenomena. Although structural properties of voice constructions vary enormouslyfrom language to language, they can be compared across languages, and typological uni-versals underlying the syntactic space for voice (cf. Croft 2001:312�13) can be inferred.In this paper we hope to contribute to our understanding of the structure of this syntacticspace by examining the voice constructions in Cebuano and argue that the na-verb con-struction, rather than the gi-clause, best satis²es the standard criteria for a basic passiveconstruction in a language: the defocusing of Agents (Shibatani 1985), which is the min-

1. The research reported here was made possible by a research grant from National Science Councilto the second author, NSC 93-2411-H-002-003. Earlier versions of this paper were presented at theSecond Workshop on Discourse and Cognition (National Taiwan University, May 2005), and at theTenth International Conference on Austronesian Linguistics (ICAL-10, Puerto Princesa City, Janu-ary 17�20, 2006). The ²rst author is a native speaker of Cebuano. We would like to acknowledgethe help of our informants, Flora Uy, Irene Joan Lapa Tan, Rowanne Maxilom, and students in ourTypology class, Fuhui Hsieh and Dongyi Lin. We are also grateful to Malcolm Ross, RickyNolasco, Stacy Teng, and the audience at the ICAL-10 for their insightful comments. We arethankful most especially to the two anonymous reviewers, who have provided us very useful anddetailed comments. Any errors are, of course, our own responsibility.

cebuano passives revisited 555

imal integration of A into the syntax of its clause (Comrie 1988:21); low text frequency;and a word order distinct from the active construction.

This paper has a twofold objective. First, we review Payne�s study on Cebuano gi-clauses and propose a different interpretation. Payne (1994) examined written narrativesin Cebuano and reanalyzed the Patient-Focus (PF)2 gi-verb constructions that have a PAword order as passive constructions, because the Ps in his data were said to be highly top-ical and the As downplayed by omission. Native intuition tells us, however, that a gi-morpheme in a great majority of Cebuano verbal clauses (except in �adversative� gi-verbs and ²xed expressions referred to in section 3 below)3 strongly implies an obliga-tory A and that, in gi-constructions, there is an assumed intent and deliberate effort on thepart of the Agent to complete the action denoted by the verb, whether or not the A isovertly expressed. It would therefore seem dif²cult to reconcile the existence of an oblig-atory and effortful A with the concept of passive. To resolve this issue, in section 2 of thispaper, we re-examine the gi-constructions, based on an analysis of natural spoken data.Although there are also gi-verbs that take the -an locative suf²x�the Locative Focusverbs�and function structurally like PF verbs, we follow Payne (1994) in analyzingonly the PF constructions (Goal Focus constructions in Payne�s terms).4

The results of our analyses show that the As in gi-PF clauses are always topicalregardless of the word order of A and P, as also shown in the corresponding PF construc-tions of other Philippine languages and most Formosan languages, such as Tsou (Huang2002) and Saisiyat (Huang, Su, and Sung 2004). Furthermore, we show that these PFclauses behave in ways that run counter to the current understanding of what constitutes apassive construction in languages that can be argued to have one. Three criteria for theidenti²cation of passive clauses are often cited in the literature, namely, defocusing ofagents or the minimal syntactic integration of the A, text frequency, and distinct wordorder. These PF clauses in Cebuano are far too frequent in actual discourse to be consid-ered passive. Moreover, Agent NPs are topical and thus may be expressed as zero, buteven then, they are still overtly expressed in 59 percent (in conversation) and 91 percent(in narratives) of the clauses in our database, showing that As are fully integrated into thesyntax of gi-verb clauses. Because of the obligatory As implied by gi-verbs, the PFclauses that also have more topical Ps than As can be considered inverse constructions atbest, because the As retain their high topicality. We do agree with Payne that word order

2. The abbreviations used in glosses, apart from those that conform to the Leipzig Glossing Rules,are: advrs, adversative marker; af, Agent focus; agt, agent(ive); bc, backchannel; cm, comple-ment marker; conj, conjunction; dm, discourse marker; emph, emphatic marker; evid, eviden-tial marker; exist, existential marker; fil, ²ller; fs, false start; intens, intensi²er; interj,interjection; lf, Locative focus; lk, linker; part, particle; pf, Patient focus; pn, proper noun;pron, pronoun; real, realis marker; spec, speci²c. In addition, portions of the data that will becited in the main body of the paper contain discourse symbols, namely, “@” for laughter; “:”after a syllable for lengthening; and “-” after a syllable or a clause for truncation; square bracketsenclose overlapping utterances. Figures in parentheses indicate the length of pause in seconds inbetween utterances. Three dots indicate pauses between 0.3 and 0.7 seconds.

3. General statements on gi-verbs to be made hereafter in this paper will not include the adversa-tive gi-verbs and the ²xed expressions described in 3.1 and 3.2 respectively.

4. Locative Focus (LF) and Instrumental Focus (IF) constructions have low frequency (7.4 per-cent [N=73] and 1.6 percent [N=16], respectively) in our Cebuano database, and their exclu-sion from the count does not affect the conclusions arrived at in our study.

556 oceanic linguistics, vol. 46, no. 2

is a key factor in identifying passives; but we argue that in Cebuano, the word order forpassive is VP, where the V is marked na-, the P is in the nominative case, and the A isdefocused and thus often missing. In the gi-constructions, however, the A is alwaysaccessible from context even if it is covert, because of the semantics of the gi- marker. Insection 3 we will show that gi-clauses fall into one of the following categories: (a) �adver-sative� clauses in which the main verbs are intransitive and behave syntactically likeAgent-Focus (AF) verbs in being capable of taking only one argument NP; (b) active andinverse constructions depending on the relative topicality of the A and the P; and (c) ²xedexpressions of verbs of naming and saying.

We argue below that a more plausible candidate for passive in Cebuano is to be foundin the non-AF na-construction. In clauses containing the na-af²xed verb, the P nominalserves as an often inadvertent undergoer of an action, and the sentence can be said todirect the hearer�s attention to the effect of the action on the Patient, as opposed to a gi-construction where an �effort-ful� action of the Agent is required (Nolasco 2005),thereby making na- an initially more likely candidate for the passive construction. Recentdescriptions of Philippine languages refer to a class of stative verbs with nominativepatient nominals as true passives, in that they are intransitive and that their As are typi-cally not expressed (Reid and Liao 2004). Furthermore, in his paper on Northern Luzonlanguages, Reid (in press) indicates that in such constructions �when the nominativephrase expressed an undergoer, the verbal morphology re³ected the pre²x reconstructedas Proto�Extra-Formosan (PEF) *ma-, deriving an agentless passive construction.� Infact, As are rare in na-/ma-constructions in some Northern languages of the Philippines.5

In section 4 we identify four types of na-clauses in Cebuano. One of the four na-con-structions identi²ed ²ts the standard criteria of passive, namely, omitted nontopical Assignifying minimal integration of the A into the syntax of the clause, low text frequency,and distinct word order, and we believe that this na-type has developed from the samemorpheme as the aforementioned na-/ma-constructions.

In this paper three voice types are distinguished�active, inverse, and passive�basedon the relative topicality between the A and P arguments in transitive clauses (cf. Coore-man 1982). Active and inverse clauses have A and P arguments that are both topical. Aclause is �active� if the A argument is more topical than the P; and �inverse� if the P ismore topical than the A. A clause is identi²ed as �passive� if the P is topical and the A isnot topical at all. Furthermore, �transitive� clauses refer to those predicated by verbs thattake two core arguments, A and P, while �intransitive� clauses refer to those predicatedby verbs that take only one core argument, the S. In relation to this, we do not considerthe Patient nominals appearing in �intransitive� clauses and marked oblique to be corearguments. In this paper, we refer to these �transitive� and �intransitive� clauses using themore traditional terms Patient-Focus (PF) and Actor Focus (AF) respectively, partly inconformity with Payne (1994). Another term, Nonactor Focus (NAF), is also often foundin recent studies of Formosan and Philippine languages to cover all of the syntacticallytransitive constructions; however, we are restricting our scope to PF clauses only.

The organization of the present study is as follows. In section 2 we will reviewPayne�s study and compare passive and active clauses; in section 3, we will provide our

5. We thank one of the reviewers for directing us to this phenomenon in northern Philippine languages.


own analysis of the gi-clauses that he considers passive; and in section 4, we will discussthe different types of na-constructions. The data base used in our study consists of actualconversations between adult speakers of Cebuano. The texts are approximately 2 hoursand 30 minutes in length, and were collected by the ²rst author between 2001 and 2005.6

2. Gi-CLAUSES: ACTIVE OR PASSIVE? As a preliminary, table 1 shows thecase markers in Cebuano. The view that the PF clause in Philippine-type languages is anactive construction has been widely accepted (Brainard 1994, Brainard and Molen 2005;Cooreman, Fox, and Givon 1984; Mithun 1994; Gerdts 1988, among others). 7 The gi-clauses in Cebuano are now treated as active, or ergative (in the sense of Dixon 1994),where the P argument in a transitive clause is treated in the same way as the S argumentin an intransitive clause, where both the A and P arguments represent central participants,and where neither A nor P can be demoted. As in table 1, both the P and S arguments inCebuano clauses take the same markers ang, si, or sila, while the A argument in transitiveclauses takes another set of case markers, usually ni or nila. The case-marking systemthus shows the language to be morphologically ergative. A number of linguists reject theergative analysis for Cebuano (Shibatani 1988) and other Philippine languages (Himmel-mann 1991 and De Wolf 1988 particularly on Tagalog). As important as these issues are,we will not pursue them further here, as they are beyond the scope of this study.8 Further-more, although we analyze Cebuano as ergative, we prefer to use the typologically more

6. Examples from Saisiyat texts are also taken from the National Taiwan University (NTU) Cor-pus of Formosan Languages.

7. We do not consider Bell�s (1976) �passive analysis� of PF clauses in this paper because it isnot relevant to our position. Our basic position is that PF (or NAF) clauses are active construc-tions, and what we are trying to show is that the gi-clauses, irrespective of word order, not PFclauses in general, cannot be passive. Using Ilokano data, Gerdts (1988) shows major prob-lems concerning the �passive analysis� and proposes the �ergative analysis,� which can begeneralizable to other Philippine languages.

8. Shibatani (1988:96�115) argues that the patient focus construction cannot be indicative of basiccase assignment. Rather, basic case emerges only when the clause is nominalized. In nominal-izations, according to Shibatani, the transitive object P does not coalesce in basic case assign-ment with the intransitive subject S, because the former is assigned goal case, while the case ofthe latter is Actor. We hold, however, that even nominalizations in Cebuano do not contradict theclaim that S and P are treated the same, although it is true that there is syncretism of case mark-ing in nominalizations, resulting in A, S, and P being marked by oblique sa. As for Tagalog,Himmelmann (1991) provides evidence for the nonexistence of the syntactic category �verb� inthe language; hence there could never be any issue with regard to the distinction between transi-tive and intransitive verbs, nor with regard to ergativity. A related issue concerns the subject rela-tion in Philippine languages, particularly Tagalog. Among the many articles written on thiscontroversial issue, Kroeger (1993a, 1993b) uses a wide variety of syntactic tests to show thatthe ang-marked (Nominative) nominal is the grammatical subject of a clause.

TABLE 1. CASE MARKERS IN CEBUANO

nom gen oblperson singular si ni kay

plural sila nila kanilacommon nonspeci²c � ug ug

speci²c ang sa sa


general terms �nominative� (rather than �absolutive�) for the core arguments S and P, and�genitive� (rather than �ergative�) for the core argument A in transitive constructions.

In this section we review Payne�s treatment of gi-clauses in Cebuano (1994) andpresent our ²ndings based on actual spoken data to show that the passive in Cebuanocannot be identi²ed with the gi-clauses, because an Agent in the gi-clauses implies deter-mination, intention, and �effort.�9 If the Agent in gi-clauses is covert, it is not due todemotion but to its high degree of topicality. If no Agent is being implied, the focus turnsto the P nominal, which would then require the use of a na-construction. Based on the setof criteria identi²ed above (namely frequency, degree of syntactic integration of theAgent, and distinct word order), we will demonstrate that the gi-clauses in Cebuano can-not be passive.

2.1 REVIEW OF PAYNE (1994). This subsection will review Payne�s study onCebuano passives where he claims that gi-clauses with a VPA word order function aspassives. We will also look into the degree of topicality of the clausal arguments in theseso-called �passive� constructions.

Payne restricted his examination of the gi-perfective verbs to independent clauses, fortwo main reasons. First, in dependent clauses (relative, adverbial, and complement) aswell as in questions, clefts, and imperatives, the choice of AF vs. PF is determined by thesyntactic environment and therefore cannot be considered a pragmatic option. Second, inindependent clauses in all tense/aspect categories except perfective aspect, the differencebetween AF and PF conveys aspectual or modal nuances that are not directly associatedwith the pragmatic statuses (topicality) of the clausal arguments, as shown in (1):

(1) Aspectual/modal nuances conveyed by verbal af²xes (Payne 1994)

a. Future: basa �read�mo-basa (AF:no promise) vs. basa-hon (PF:promise)

b. Imperfective: minyo �get married�mag-minyo (AF:intention) vs. minyo-on (PF:assertion)

c. Abilitative: luto �cook�maka-luto (AF:ability) vs. ma-luto (PF:intention)

d. Perfect: basa �read�naka-basa (AF:accidentally) vs. na-basa (PF:on purpose)

The ²rst members of each pair above are AF markers, while the second members are PFmarkers. PF conveys a higher degree of transitivity (Nolasco 2005, 2006), and thusnuances of �promise� (-on in [1a]) and �assertion� (-on in [1b]), while AF focuses on thenatural �ability� of the Agent to perform the action indicated by the verb (mo- in [1a],mag- in [1b], and maka- in [1c]). As for the �accidental� nuance indicated in (1d), this is abasic semantic element of the naka�na pair (see table 2): the AF naka conveys an Agent�s�accidental� performance of an action (also �abilitative�), while the PF na- suggests an�accidental� effect of an action on a Patient. In other words, an action is done without anydeliberate purpose or intention.

9. See Nolasco (2005, 2006) for a detailed discussion on the nuances between AF and PF inPhilippine languages. A similar explanation is proposed by Mithun (1994) using the term�involvement� (of absolutives).


In view of (1), Payne considers that it is in the independent, declarative clauses in theperfective aspect where speakers consistently respond to AF and PF as conveying differ-ences in �emphasis� or �topicality� (pragmatic option). These are the clauses that expresstemporally sequenced events in narrative discourse, as in the two extracts below:

(2) a. Ug busa niana dayo-ng adlaw-aand therefore dem immediately-lk day-spec

mi-dala si Hari Rodrigo kaniya ngadto sa palasyo.nfut.af-carry nom pn dat to.there loc palace

�And so that very day King Rodrigo took her to the palace.� (Payne 1994)

b. Ug busa niana dayo-ng adlaw-aand therefore dem immediately-lk day-spec

gi-dala kini ni Hari Rodrigo ngadto sa palasyo.nfut.pf-carry this gen pn to.there loc palace

�And so that very day King Rodrigo took her to the palace.� (Payne 1994)

Payne was correct in stating that AF and PF convey different �nuances.� However, heseems to have overlooked the fact that this difference in emphasis also extends to the pairni- and gi-.10 Aside from having a de²nite and affected Patient and a highly topical Agent,the gi-clause in Cebuano strongly implies that, as in (2b), in addition to the highlightingof the �involvement� of the Patient (Mithun 1994), there is an effort, intention, and deter-mination on the part of the Agent to carry out an action, which is, in this case, �to take herto the palace.� On the other hand, (2a) only indicates an action performed by the Agent,which only highlights the involvement of the Agent �to take somebody to the palace,�without concern for any possible effect on the Patient. It is thus clear that, with the Agentbearing such features, these gi-perfective verbs in independent clauses cannot be passive.Further arguments will be provided in 2.2 and section 3.

10. The perfective (or, more strictly, nonfuture) AF marker in Cebuano is ni- or mi-, which are dia-lectal variations. AF clauses indicate ongoing action (see also Mithun 1994 for a similar case inKapampangan), and it is not surprising that imperfective nag- markers (see table 2) are moreoften attached to AF verbs. In the conversation excerpt below (ia)�which is the instance shownas AF transitive in table 4�although ni-pusil �shot (with a gun)� is semantically transitive, i.e.,the verb requires a semantic object of the action �to shoot�, it is used to convey the act of �shoot-ing�, wherein what is being shot at is not important. If the target of the shooting were critical, (ia)should have been expressed as (ib).

(i) a. Diri=man=siya ni-pusil unya ni-lusot diri.here=par=3sg.nom af.nfut-shoot dm af.nfut-pass.through here�He shot here, then (the bullet) passed through this.�

b. Kini=man iyaha-ng gi-pusil.this=par 3sg.poss-lk pf.nfut-shoot�It was this (the head) that he shot.� (lit. �This is what he shot.�)(constructed based on [ia])

TABLE 2. CEBUANO VERBAL AFFIXES

realis / nonfuture irrealis / future / potentialAF PF AF PF

perfective ni- / mi- gi- mo- -onimperfective nag- gina-V-on mag- paga-V-onnonpurposive naka- na- maka- ma-


2.1.1 Semantic transitivity. It is well known that there is a strong correlationbetween focus and semantic transitivity in Philippine and Formosan languages. This isshown for written data in Cebuano in table 3 and for our conversational data in table 4.AF clauses focus on the action of the Agent and are thus intransitive, while PF gi-clausesmust have an affected Patient and are predominantly transitive. However, in both table 3and table 4, there are PF verbs that are semantically intransitive; we classify them as�adversative verbs.� These gi-af²xes on adversative verbs must not be taken as andclassi²ed together with the PF form. These are discussed further in 3.1.

2.1.2 Constituent order. Payne (1994) found no correlation between focus andconstituent order, as shown in table 5. This would mean that there is equal probability thatthe VPA word order can occur in both AF and PF constructions. Based on numericalsuperiority, the unmarked order is VAP. The more topical argument, which is in this casethe A and more often in clitic form, is located closer to the verb. There are some problemsthat can be raised with respect to table 5, however. As is now recognized among Austro-nesianists, AF clauses, especially in the majority of Formosan and Philippine languages,are syntactically intransitive (Liao 2002, 2004; Starosta 1997, 1998, 1999; among oth-ers), taking at most only one core argument (see table 6). Payne must have missed thenow well-recognized fact that AF clauses with an oblique-marked Patient argument, alsocalled Extended Intransitive Clauses (EIC), are a separate clause type in Cebuano, as wellas in other Formosan and Philippine languages, as the core vs. oblique distinction in these

TABLE 3. FOCUS AND SEMANTIC TRANSITIVITY (Payne 1994)

transitive intransitive totalAF mi- 80 175 255PF gi- 114 11 125

TABLE 4. FOCUS AND SEMANTIC TRANSITIVITYIN SPOKEN CEBUANO DATA

transitive intransitive total

AF mi- 1 36 37

PF gi- 65 16 81

TABLE 5. FOCUS AND CONSTITUENT ORDER (Payne 1994)

vap (unmarked order) vpa

AF mi- 23 6

PF gi- 52 19

TABLE 6. FOCUS AND CONSTITUENT ORDER (CONVERSATIONAL DATA)

SV/VS VAF mi- 29 8

VAP/PAV VA/AV V VP/PV VPA/APVPF gi- 15 12 10 23 11


languages is pretty robust (Huang 2006). Payne apparently treated these EICs in his writ-ten data on a par with ordinary transitive clauses, which we surmise is probably why nodata on single argument clauses in AF were given in table 5.

Moreover, two argument clauses are rare in actual discourse, as Cebuano syntax alsoobeys the One Argument per Clause Constraint (Tanangkingsing 2006), better known asthe Preferred Argument Structure Constraint (Du Bois 1987, 2003). As shown in table 6,VS or SV clauses predominate in AF mi-clauses, and VP and PV clauses outnumber allthe other clause types in PF clauses. Therefore, there is no way to compare the constitu-ent order between mi- and gi-clauses, because what we have obtained from spoken datain table 6 shows markedly different word order patterns. But this should occasion no sur-prise. Being syntactically intransitive, the AF mi-clauses take only Ss. On the other hand,except for VP/PV, all the remaining PF gi-clauses are more or less evenly distributedbetween the other possible word orders.

2.1.3 Topicality: Referential Distance and Topic Persistence. There are twoways to measure the degree of topicality of an argument, namely, Referential Distance(RD) and Topic Persistence (TP). RD measures the number of clauses between one men-tion of a participant and its previous mention in the text. Following Payne (1994), anargument that has been mentioned in a previous utterance would get an RD value of 1.An RD value from 2 to 19 would be considered medium topical; an RD value of 20 ormore would be low in topicality (²rst mentions are given an RD value of 20). Tables 7and 8 show the RD values of the A and P arguments in Payne�s study and in our data,respectively.

As shown in table 7, the PF clauses with VAP word order tend to code highly topicalAs and less topical Ps; but in those with VPA word order, the Ps are more topical than theAs to a signi²cant level (VPA[A/P]χ2 = 12.507, *p=.002), leading Payne to claim that thegi-clauses with P(A) word order function as passives. However, this is not supported byour spoken data, as shown in table 8, where the Ps are not signi²cantly more topical thanthe As under the VPA word order (VPA[A/P]χ2 = 0.305, p=.858). Based on the relationbetween argument topicality and type of construction (Givon 1990), these PF clauses

TABLE 7. REFERENTIAL DISTANCE IN PF CLAUSES (Payne 1994)

P(A) A(P)continuity A P A Phigh 3 (18%) 26 (67%) 47 (70%) 4 ( 9%)medium 6 (35%) 8 (21%) 18 (27%) 10 (23%)low 8 (47%) 5 (13%) 2 ( 3%) 29 (68%)

TABLE 8. REFERENTIAL DISTANCE OF A AND PIN CEBUANO CONVERSATIONAL DATA

P(A) A(P)continuity A P A Phigh 5 (56%) 19 (44%) 19 (83%) 8 (67%)medium 1 (11%) 1 (11%) 3 (13%) 1 ( 8%)low 3 (33%) 6 (44%) 1 ( 4%) 3 (25%)


with a VPA word order in our conversational data can be said to exhibit properties char-acteristic of inverse constructions.

Moreover, as Cebuano obeys the One Argument per Clause Constraint (the quantitypart of the Preferred Argument Structure Constraint), two-argument clauses are rare innatural discourse. For statistical purposes, we have included one-argument clauses in ourcount (in tables 8 and 10): clauses with covert As (including V=P(A);11 V=P; VP; andPV) are grouped with VPA word order and those with covert Ps (including V=A(P);V=A; VA; and AV) with VAP word order.12

TP measures the number of times a participant is mentioned within ten clauses afterany mention, regardless of its syntactic position in subsequent mentions. Again, we fol-low Payne (1994) by assigning a TP value of zero to an argument that is not mentionedagain in the following discourse and that would be referred to as of low importance. A TPvalue between one and three would mean medium importance, and an argument that ismentioned more than four times in the following discourse is highly important and topi-cal. Tables 9 and 10 give the TP values of A and P in Payne 1994 and in our conversa-tional data, respectively.

The data in table 9 show that, although the Ps in PF clauses with PA word order aresigni²cantly more topical than those in PF clauses with AP word order (P[AP/PA]χ2 =11.237, *p=.004), these Ps are not more topical than the As in the PF clauses with PAword order (VPA[A/P]χ2 = 1.858, p=.395). These clauses can thus be said to be function-ally inverse, a result that is consonant with the claim made for the data given in table 8.Table 10, showing the TP values of A and P based on conversational data, also gives asimilar result: the As are more topical than the Ps in VAP word order (VAP[A/P]χ2 =15.168, *p=.001), but the Ps are not more topical than the As in VPA word order (VPA[A/P]χ2 = 1.389, p=.499). However, the preposed Ps in VPA word order are indeed more top-ical than the Ps in VAP word order (P[AP/PA]χ2 = 10.621, *p=.005); hence the preposingof Ps to a position right next to the verb is highly motivated. To summarize, then, our datashow the following topicality hierarchy between the As and Ps in relation to the two word

11. Some arguments are pronominal clitics, which we mark with an equal sign �=� like �V=A� or �V=P.�12. VPA word order refers to a clause where P is contiguous to the verb, while VAP word order

refers to a clause where A is contiguous to the verb.

TABLE 9. TOPIC PERSISTENCE IN PF CLAUSES (Payne 1994)

P(A) A(P)persistence A P A Phigh 10 (59%) 14 (50%) 35 (64%) 8 (19%)medium 3 (18%) 10 (36%) 17 (31%) 13 (31%)low 4 (23%) 4 (14%) 3 ( 5%) 21 (50%)

TABLE 10. TOPIC PERSISTENCE IN SPOKEN CEBUANO DATA

P(A) A(P)persistence A P A Phigh 4 (44%) 12 (46%) 17 (74%) 1 ( 8%)medium 1 (11%) 7 (27%) 2 ( 9%) 1 ( 8%)low 4 (44%) 7 (27%) 4 (17%) 10 ( 83%)


order types in question, which, signi²cantly, also resemble results obtained for two For-mosan languages, Tsou (Huang 2002) and Saisiyat (Huang, Su, and Sung 2004); that is, inPF clauses, As are still as topical, if not more, than Ps.

(3) Topicality: A (AP) > A (PA) = P (PA) > P (AP) → A > PPayne was correct in assuming that a difference in word order, that is, VPA as

opposed to VAP, would probably produce a passive construction, but he was not on theright track in looking for this construction in a gi-marked PF clause. The results shown inthis section based on actual spoken data reinforce the claim that the gi-Patient Focusclauses in Cebuano should therefore be considered active constructions in the language:the gi- morpheme strongly implies a deliberate intention and effort on the part of theAgent (regardless of whether it is overt or implied), with P as a goal.

Furthermore, there seems to exist a discrepancy between oral discourse and writtentexts in that the A is less topical than the P in RD in Payne�s written data, though in ourconversational data the A is consistently more topical than the P. Why should such a dis-crepancy exist? One reason would be that adversative verbs like gi-laylay �be af³icted(with an illness)� (see [11]), which are syntactically intransitive (see 3.1), might have beencounted as transitive, and thus the sa-marked nominals expressing Cause are insteadtreated as Agents. If so, the As in such a clause would have a low RD and a low TP, espe-cially if inanimate, because they are unlikely to become topic. The postposed As in VPAgi-clauses would tend to be common nouns and so marked by sa rather than propernames marked by ni, and they would more likely be ²rst mentions. This would explainwhy the As in VPA clauses in table 7 have low RD values, which is a fairly unusual resultfrom the perspective of conversational data.13

2.2 COMPARISONS WITH ACTIVE CLAUSES. In this section, we usethree criteria often used in the literature to distinguish between active and passive clauses:defocusing of the Agent or lack of syntactic integration of the A, text frequency, and dis-tinct word order (e.g., Shibatani 1988, Comrie 1988, Payne 1982). We will also presentdata from Cebuano to provide additional evidence that the gi-clauses in Cebuano are notpassive but active constructions.

2.2.1 Syntactic integration of Agent. Although the Patient argument possessessubject properties in the gi-clauses, these gi-clauses involve greater integration of the agentphrase into the syntax of the clause (Comrie 1988:9), show no tendency at all towardAgent omission (Shibatani 1988:93),14 and the A argument is most often expressed in pro-nominal form. Shibatani (1988) cited Dryer�s ²gure of 85.1 percent integration (57 clausesout of 67 gi-clauses had As) for Cebuano, and Shibatani himself counted 40 out of 49clauses in his folktale corpus for an 81.6 percentage. Our own data yielded a 59.4 percentintegration (167/281) for conversation, and a 91.1 percent (72/79) for the narratives.

13. A similar case exists in Contemporary Indonesian, which is reported in Verhaar (1988:348�49) to have a �sociolingual� split, where the clause organization in written language and otherprepared speech forms is accusative, while more relaxed conversation utilizes ergative con-structions. We thank Malcolm Ross for pointing us to the reference.

14. The Malagasy clauses called �passive� voice by Keenan and Manorohanta (2001) have around60 percent agent integration. These �passive� voice clauses have been explicitly claimed byKikusawa (to appear) as active constructions, as we have conjectured.


2.2.2 Frequency of PF clauses. The text frequency of PF clauses is also an indica-tion of voice type. Passives are morphosyntactically marked constructions, and a con-struction that is frequently attested in a language cannot be passive (Comrie 1988:9;Payne 1982). This is clearly manifested in Chamorro, where there are both PF and pas-sive clauses. In the PF clauses, the Agents are expressed 80 percent of the time, while inthe passive clauses, only 15 percent of the sentences have overt Agents (Cooreman 1982,also cited in Shibatani 1988:93).

The gi-clauses are not marked constructions in Cebuano in terms of frequency. Inhis folktale corpus, Shibatani counted 46 percent (49 out of 106 clauses) using the PFconstruction (Shibatani 1988:95�96). In our Cebuano conversation data, the propor-tion of gi-clauses is 22.3 percent (220/985),15 while in our narrative data, the proportionis 51.0 percent (79/155).

2.2.3 Distinct word order. The passive is a marked construction that displays aword order that is distinct from other types of constructions in a given language. In thePukapukan (Samoic-Outlier, Polynesian) clauses in (4a) and (4b), the As are positioneddifferently relative to the Ps in passive and ergative clauses: the As are usually positionedfarther from the verb than the Ps in passive clauses (see also 5). Furthermore, in numer-ous languages a speci²c morpheme is recruited to mark [+PASS] on the passive verbs,such as -na in Pukapukan (4) and ki- in Rukai (5).

(4) a. Kai-na loa na tamaliki e te wui aitu pau.eat-pass emph the.pl children agt the pl spirit done

�The children were all eaten by the spirits.�(Pukapukan Passive, Croft 2001:66)

b. Lomilomi ai e tana wawine ma na tama lua tulivae ia.massage pron erg his woman and the.pl boy two knee that

�His two knees were massaged by his wife and the children.(Pukapukan Ergative, Croft 2001:66)

(5) Ki-a-kani kuani umas sa Likulaw.pass-real-eat that man acc leopard

�That man was eaten by a leopard.� (Rukai Passive, Li 1973:193)

There are, of course, languages where a certain construction has already acquired apassive function, but there is as yet no distinct morpheme that has grammaticized into amarker of the passive, because the acquisition of a new function is generally known toprecede the formation of a formal marking for this function. Nevertheless, such a changein function can still be detected from a change in word order. Passives involve a more top-ical P that moves closer to the verb; the A, being less topical and less important, is rele-gated to a position that is farther away from the verb. This reversal of word order can evenprogress to the further omission of the A. In section 4 we will show that this is the case inthe passive constructions in Cebuano.

15. A raw text frequency of more than 20 percent should be too high for a passive construction,which should only account for less than 10 percent of the number of clauses (Wouk 1999:103,among others).


3. CEBUANO gi-CONSTRUCTIONS. If the gi-perfectives found in independentclauses identi²ed by Payne as passives are not passives after all, then what are they? Inthis section, we will show that these �passive� gi-clauses divide into three functionaltypes: adversative clauses, inverse constructions, and ²xed expressions. The ²xed expres-sions are primarily verbs of saying and naming. The inverse type can be distinguishedfrom the active type depending on the relative topicality of the A and P. Nevertheless, theAs in inverse constructions can never be nontopical, as the gi- morpheme stronglyimplies a topical A. The reader must be reminded that the scope of discussion in this sec-tion does not include the active clauses with VAP word order and highly topical As thatform the bulk of gi-constructions.

3.1 ADVERSATIVE VERBS. Rukai, a Formosan Austronesian language, has anadversative verb marker distinct from passive marking. Aside from the passive marker ki-identi²ed by Li (1973), another morpheme kw- is found to attach to verbs in clauseswhere the Experiencer does not have control of an unfortunate event (Zeitoun 2000:96).These verbs indicate adversity, as shown in the examples in (6).

(6) a. Kw-a-lama ku daane-su.advrs-real-burn obl house-2sg.gen

�Your house burned down.�

b. Kw-a-vaLigi ku daedae.advrs-real-typhoon obl road

�The road got affected by the typhoon (or, was damaged by thetyphoon).�

c. Kw-a-Lumaz=aku.advrs-real-hit=1sg.nom

�I got hurt.� (Rukai, Zeitoun 2000:96)

However, in other Formosan languages, such as Thao and Saisiyat, and in Philippinelanguages, adversity is conveyed by the af²xation of the morpheme typically appearingin NAF constructions on a limited set of intransitive verbs. These intransitive verbs indi-cate ailments, aching body parts, and infestation by a certain animal or insect. Below weprovide examples in Thao and Saisiyat. In Thao, the suf²x -in marks adversity in additionto its function of marking PF verbs, as in (7).

(7) a. Wazaqan harbuk-in.lake fog-pf

�The lake is covered with fog (fogged in).�

b. Cicu macuaw kitlhulhuk-in.3sg lots.of kit.pimple- pf

�He has a lot of pimples.� (Thao, Wang 2004:134)

Similarly Saisiyat employs the suf²x -en to indicate adversity. The suf²x -en also marksPF verbs in Saisiyat.

(8) a. A:zaw-en azaw-en o: isaa �oka� ila ka ralom.heat-pf heat-pf fil dm neg pfv nom water

�It�s the drought season; there�s no water.� (lit: [the land] is heated.)


b. ... so: rikrika-en ila ma isaa ana p-in-a-mowa�-an tatimae�if hot-pf pfv like.that regardless cau-pfv-plant-loc vegetable

ma ayo:o� ila saboeh.fut wilt pfV all

�If the weather is hot, the vegetables planted will all wilt.�(Saisiyat, NTU Corpus)

We do not consider these adversative verbs passives for the following reasons. First,the constructions above from (6) through (8) are intransitive clauses with verbs that sharethe same form as Patient Focus verbs. The af²xes on these verbs must not be taken as andclassi²ed together with the usual PF af²x.16 Second, they take a Nominative NP subject,which experiences adversity. They cannot take an Agent NP, as the presence of a geni-tive-marked Agent nominal would render the clause infelicitous; the Agent is insteadexpressed in the verb. And third, these adversative clauses also indicate a negative effectcaused by some meteorological phenomenon, and physical and emotional states ofadversity, as in the Cebuano examples given below, as well as a host of other states ofmisfortune and situations viewed in a negative light.17 The sa phrase indicates the causeof the adversity, but not the Agent.

(9) Gi-kapoy si Pedro (sa trabaho / *ni Juan)nfut.pf-be.tired nom pn loc work gen pn

�Pedro is tired (from work / *by Juan).�

(10) T ay ka-sakit oy [gi-luod=ko]interj nmlz-pain voc nfut.pf-nauseating=1sg.nom

W [@ @ @] @@@T gi-luod=ko dong naku dong

nfut.pf=1sg.nom voc interj voc

T �Oh, hey, it�s too painful. I [feel nauseous].�W [laughing]T �Hey, I feel nauseous.�

(11) Gi-laylay si Rayna Esmeralda sa usa ka sakitnfut.pf-af³ict nom queen pn sA one lk sickness

�Queen Esmeralda was stricken by a sickness.� (Payne 1994)

In (9), the argument Pedro is the undergoer of the �effect� of the work, that is, the feel-ing of tiredness. In (10), the verb marked by gi- denotes a state of adversity of the argu-ment marked in the nominative case, and there is no action that directly affects thisnominative nominal. Any other argument marked by ni or sa can only indicate the causeof the state. Example (11) is taken from Payne (1994): the verb gi-laylay18 merely indi-

16. Not all the languages that have such adversative constructions agree in their use of -in, -un,-an, or si- in these constructions. For example, k-in-abag-an �to suffer gas pain� in Tagalogemploys a locative morpheme -an (Liao 2004:36). This issue is not pursued further as this isnot the main topic of this paper.

17. It is also worth noting that even loan words indicating adversity, such as lawsay �to have loosebowel movement�, also use such a construction, as in this clause commonly spoken amongethnic Chinese Cebuanos: gi-lawsay=ko �I am having loose bowel movement.�

18. The verb gi-laylay �to be af³icted (with an illness)� is considered archaic and is probably onlyfound in written narratives and in the lexical repertoire of the older generation of Cebuanospeakers, especially those from far-³ung areas. Younger speakers and urban residents do notuse this word in daily talk.


cates a state of adversity that is experienced by the queen because of her sickness; the pas-sive sense is merely an effect of the English translation where �a sickness� is being viewedas though it were an agent demoted into a by-phrase.

The type of adversative constructions such as these found in Formosan and PhilippineAustronesian languages seems to be the norm in much of eastern Asia, which Keenanand Dryer (2007) term as �negative effect passives,� as opposed to �positive effect pas-sives� attested in Vietnamese and Korean. 19

Smith (1998:220) argues for treating adversative constructions as a separate grammat-ical category. If it happens that there is an involuntary action, what happens is an �acci-dent,� which implies an adversative notion, as in the nonvolitional permissive causativeconstruction in (12) in Even, a Tungic language. Here the causal agent (the hunter) doesnot wish the event (the deer�s standing up) to occur, but �permits� it due to his inattentive-ness. In the same way, we can say that in adversative clauses in Cebuano, the causal agent(i.e., the experiencer) does not wish an adversative event (say, hunger or thirst) to happen,but �permits� it due to his inability to take care of his own bodily needs.

(12) Bujuheme-Ø buju-m ila-v-ra-n.hunter-nom wild.deer-acc stand.up-advrs-nfut-3sg

�The wild deer stood up; the hunter was negatively affected.�(Even, Malchukov 1993:8, cited in Smith 1998:227)

3.2 FIXED EXPRESSIONS. The second type of gi-verb found in our data is aclosed set of verbs of saying and naming that have fossilized into ²xed expressions, suchas �be said� or �be told�, and verbs of naming, such as �be called� or �be named�. Anexample in English is given in (13). Examples from Cebuano below are taken from Shi-batani (14), Payne (15), and our Cebuano data (16).

(13) It is said that time heals all pain. (Siewierska 1985:238)

(14) Gi-nganlan=siya-g Kapuroy.nfut.pf-name=3sg.nom-obl pn

�He was called Kapuroy.� (Shibatani 1988: 94)

(15) Si Totong gi-ila nga labi-ng hawod sa dama.nom pn nfut.pf-identify LK more-lk best obl game

�Toto is known to be the best at the game of dama.� (Payne 1994)

(16) O syempre sad lala-mao lagi gi-ingon lalaki.yes of.course also fs- that emph nfut.pf-say man

�Right. That�s why (they are) so-called men.�

These are the gi-clauses in Cebuano that most closely resemble what we may call apassive construction. There is a similar phenomenon attested in other languages wheresuch �unspeci²ed subject constructions�20 (Keenan and Dryer 2007) are distinct frompassives. In Kutenai, the unspeci²ed subject suf²x -(n)am only occurs on intransitive

19. There are ²ve verbs in Vietnamese that act as passive auxiliaries in passive constructions. Oneof them is bi �suffer�, an adversative morpheme and another is duoc �enjoy�, used to express apleasant effect of an action on a subject (Keenan and Dryer 2007). The negative effect passivein Korean uses the morpheme tangha �to be subjected to� while the positive effect passive isconstructed with pat �to receive� (Keenan and Dryer 2007).

20. This is also termed Inde²nite Actor Construction by Dryer (1996) in his paper on Plains Cree.


verbs, never on transitive verbs. The subject in such constructions has a meaning roughlyparaphrasable as �by someone/something� or �by them/people (generic)� (Keenan andDryer 2007). Moreover, the examples shown for Kutenai are uniformly verbs of saying,as in Cebuano. We would agree with Payne that the Agent in such constructions as (15) isindeed nontopical.21 However, we have chosen to call them ²xed expressions, becausethese gi- forms are used only with verbs of saying and naming.22 Furthermore, the verbingon �to say� is now normally only used in AF form. As our data show, the threeinstances of the expression gi-ingon �is said� are only con²ned to nominalized construc-tions (see �so-called men� in [16]). Another instance is (17), where ilang gi-ingon is usedas a nominal in an equational clause, or it can also be analyzed as a phrase modifying thefollowing complement clause.

(17) Mao bitaw na ilang gi-ingon nga dali=ra ang pag-pangitathat part that 3pl.poss gi-say comp fast=asp ang nmlz-²nd

ug a- amiga pero lisud pangita-on ang tinuod nga amiga.obl fs friend but dif²cult ²nd-pf ang real lk friend

�That is what everybody says(/thinks), that it is easy to make friends,but it is dif²cult to ²nd real friends.�

3.2 INVERSE GI-CONSTRUCTIONS. The bulk of gi-constructions in Cebuanoare either active or inverse. Inverse as used in this paper does not refer to the constructionwith a formal inverse marker on the verb as in Native American languages. Instead,active and inverse are distinguished based on the relative topicality of Agent and Patient,as de²ned by Cooreman (1982) (see also Brainard 1994; Brainard and Molen 2005). Inan active construction, both the Agent and the Patient are topical, but the Agent is moretopical. In an inverse construction, the Agent and the Patient are also both topical, but thePatient is more topical. As our data attest, the Ps in inverse constructions are predomi-nantly ²rst or second person pronominals (this is especially evident in the na-clauses; see4.2).23 In the following section, we discuss word order patterns that characterize inverseclauses, namely, V=P(A), PV, and V.

21. Aside from Kutenai, there is also an unspeci²ed subject construction (using the morpheme-ukw) in Oneida (Keenan and Dryer 2007). There are, however, also ambiguous cases, as inPlains Cree, where the suf²x -ikawi is argued to be either an unspeci²ed subject suf²x or apassive suf²x (Keenan and Dryer 2007).

22. The �impersonal passives� referred to in Cook 1996 are not the same as the type described inthis paper. In those clauses, the events or situations depicted are not initiated or controlled bya human agent (64�65). Clauses such as (iia) below would require the use of na- instead of gi-in Cebuano (iib).

(ii) a. Sâ u²-tia mâtou i le pogisâ.past bite-Cia we loc the darkness�We were covered by darkness.� (Samoan, Cook 1996:64)

b. Na-ab(o)t-an=mi ug gabii.nfut.na-reach-lf=1pl.excl obl night�We were doing something until night-time.� (lit. �We were reached by night-time.�)

(Cebuano, constructed)

23. The ni- construction in Pendau (Quick 2005) is also considered to be inverse, not passive,based on similar criteria: in the measurement of Referential Distance and Topic Persistence,the As are topical; text frequency is over 20 percent; and the A is rarely omitted. However,further research is required to determine the role of word order in this language.


3.3.1 V=P (A) constructions. In this type of construction, the A nominal may beeither overt or zero. It is the P nominals that are relatively more topical and attach to theverb as a clitic, because they are usually ²rst or second person pronominals. The A nom-inals in such constructions are less animate but nonetheless topical or accessible fromcontext. Examples are given below. 24

(18) L gi-hired=ka=nilanfut.pf-hire=2sg.nom=3pl.gen

J m gi-hire=ko- dili gi-hire=ko ni: miss jerobc nfut.pf-hire=1sg.nom neg nfut.pf-hire=1sg.nom gen pn

L miss a:pn fs

J iya-ng sekretarya: personal alala:y1sg.poss-lk secretary personal bodyguard

L �They hired you?�J �Yes, (they) hired me. No, Miss Jero hired me.�L �Miss uh=�J �as her secretary, personal bodyguard.�

The A in (18) is pronominal and thus topical and accessible from the context. Con-structions such as these can be represented by the schema V=P(A), where the P is a pro-nominal clitic and is nearly always either the speaker or the hearer. The A nominal is ingenitive form and, although not necessarily human, is most likely animate. In caseswhere the A is overt and [+human], as the bolded NP in (19), or zero as in (20), it isaccessible from the context, and it is always so, due to the semantics of the marker gi-,which indicates purposeful participation of an Agent.

(19) Naku ang ako-ng ²rst year sa hotelinterj nom 1sg.poss-lk ²rst year loc hotel

gi-daugdaug=lang=ko sa mga ano- Taiwanese.nfut.pf-bully=only=1sg.nom gen pl fil pn

�Oh, when I was new at the hotel, my Taiwanese colleagues bullied me.�

(20) Suko=kaay=siya suko=gyud=kaay=siya baangry=very=3sg.nom angry=emph=very=3sg.nom part

gi-hold=man=mi-ng duha.nfut.pf-detain=part=1pl.exc-lk two

�He got so angry because we were detained (by the of²cial [mentionedearlier in the conversation] at customs).�

3.3.2 PV clauses. In PV clauses, the P nominal is topicalized, but the A nominal is stilltopical and accessible and not demoted, as in the following examples. In (21), the P argu-ment is topicalized in line 57, in part for contrast with the other character of the story, which

24. Weight of the pronouns could probably affect the word order of pronominal arguments, but itis possible to prepose =nila ‘3pl.gen’ before =ka ‘2sg.nom’ or =ko ‘1sg.nom’ or any othermonosyllabic pronoun for that matter, although it would, of course, sound a bit awkward.Even so, ²rst and second person, which tend to be monosyllabic and so would always precedethird person pronouns, would still be more topical than a third person that is pronominal (as inthe ²rst line of [18]) or a full noun phrase (as in the second line). Based on Obo Manobo data,Brainard and Molen (2006:402) argue that an �Inverse Analysis� is superior to the phonologi-cal and morphological hypotheses in explaining the orders of A and P in transitive clauses.


is also topicalized in line 54. In (22), the P nominal is topicalized as well; the A is accessi-ble from prior discourse. In addition, it is also interesting to note that in lines 62�63, thenarrator switches to English, rendering the gi-clause in lines 60�61 an active construction.

(21) 54. �(0.9) unya ang bata-ang-ang batadm nom child nom nom child

padayon ug pangita sa iya-ng frogcontinue comp ²nd obl 3sg.poss-lk frog

55. � sa- among the treesLOC among the trees

�The child continued to look for his frog in the woods.�

56. �(1.9) unya diri sa:DM here LOC

57. � pag-pangita=niya ang iya-ng- iya-ng pet dog gi:nmlz-²nd=3sg.gen nom 3sg.poss-lk 3sg.poss-lk pet dog FIL

58. i:59. �(1.3) a:60. �(1.5) a: gi-habol sa:-

fil nfut.pf-chase obl

61. �(1.0) @ sa beesobl bees

62. �@ � the bees ra-63. � ran after the- the dog @@@ the puppy

�As for his pet dog, the bees chased after it.�(Narrative data, Frog 2:54�63)

(22) Oy eto-ng- eto-ng MECO gi:-butang sa tradevoc this-lk this-lk pn nfut.pf-place loc trade

unsa ma-y ato-ng kahimtang karon.what part-nom 1pl.poss-lk situation now

�Hey, this of²ce (MECO), (the President) placed it under (the jurisdictionof) the Trade Department. What�s going to happen to us now?�

3.3.3 V clauses. In clauses where both the A and P nominals are covert, both of themare of similar topical status. This type of construction is more frequent in conversation thanin narratives. These verbs are semantically transitive, but they closely resemble adversativeclauses in that they always denote a negative effect on the subject, as shown in (23) and (24).

(23) T unsa=pa mga iba-ng foreign workerswhat=still pl other-lk foreign workers

W Mongolian Mongolian lagipn pn emph

T ni- ni- naa nay Mongolian karonfs fs exist pfv pn now

W oo pero pag-sulod kuno pag- ²rst batch kunoyes but nmlz-enter evid nmlz ²rst batch evid

six days lang gi-rape na kuno @six days only nfut.pf-rape pfv evid


T gi-rape kuno- gi-rapenfut.pf-rape Evid nfut.pf-rape

T �What are the (nationalities of) the other foreign workers?�W �Mongolians.� [emphatically]T �There are Mongolian workers now?�W �Right. But six days after the ²rst batch of workers arrived, one of

them got raped.�T �Raped?�

(24) W tangtang gihapon relo tangtangtake.off still watch take.off

T wa- wa- wala manneg neg neg part

W inig gawas sa X-ray na-walaas leave loc X-ray.machine intr-disappear

T @@ mao gi-suyop gi-suyopright nfut.pf-siphon nfut.pf-siphon

W �(We) took off (our) watches.�T �No, (we didn�t).�W �(Those things) disappear as they pass through the X-ray machine.�T �Right. (Those machines) suck up (those valuables).�

3.4 SUMMARY. Figure 1 shows the relative syntactic space occupied by each typeof gi-construction in a diagram where the vertical axis represents the role of the Nomina-tive NP (S or P), while the horizontal axis represents the form of the A nominal (overt,zero, or absent). When the Nominative NP is S, that is, when the clause is intransitive, theverb taking the gi- form is adversative. On the other hand, when the Nominative NP is P,that is, when the clause is a non-AF construction, the clause may be active or inverse,depending on the relative topicality between the A and the P nominals. The clause isactive when the A nominal is slightly more topical than the P nominal; it is inverse whenthe P nominal is slightly more topical than the A nominal. Fixed expressions are thosewhere the A is zero and the only nominal present is a P. A (systematic) gap exists in thesyntactic space when P is in the nominative case and A is absent: A can never be absentin a non-AF gi-construction, given the semantics of gi- alluded to above.

FIGURE 1. SYNTACTIC SPACE FOR gi-CLAUSES IN CEBUANO

Nom NP

S gi-adversativeclauses

gi-clauses (active; topical A)

P inverse clauses (VOA: V)

²xed expressions

Absent A Zero A Overt A


4. NA-PF CONSTRUCTIONS. If the gi-PF constructions in Cebuano cannot bepassive, as we have argued in the previous two sections, then what can be? In this section,we discuss the discourse pragmatics of the na-clauses and their distinctness from the gi-clauses. Four types of na-constructions are distinguished based on the relative topicalityof the A and P arguments: intransitive (AF construction), active, inverse, and passive.This last type of na-PF clause omits the A, and best ²ts our criteria for the passive.

Like ²gure 1, ²gure 2 shows the relative syntactic space occupied by each type of na-construction in a diagram where the vertical axis represents the role of the NominativeNP (S or P) and the horizontal axis represents the form of the A nominal (overt, zero, orabsent). In the na-intransitive, the nominative NP is S, and there is no A argument. The Snominal is affected by the action denoted by the verb, and the action that is denoted by theverb is not deliberately carried out. On the other hand, when the Nominative NP is P, thatis, when the clause is a non-AF (transitive) construction, the clause may be active orinverse depending on the relative topicality between the A and the P nominals. Theclause is active when the A nominal is slightly more topical than the P nominal; the Anominal is also attached closer to the verb. The clause is inverse when the P nominal isslightly more topical than the A nominal and is more closely attached to the verb. Whenthe A is absent or left unexpressed, which is more likely in na-constructions than in gi-constructions, the na-construction directs the attention to the effect of a nondeliberateaction on the Patient. Here, the A is completely downplayed and inaccessible from thediscourse context. This, we argue, is the passive construction in Cebuano.

As stated above, the semantics of the main verb in a non-AF na-construction (the Pnominal as the nonpurposeful undergoer of an action) is in sharp opposition to a gi-con-struction, where a purposeful action of the Agent is required or implied.25

In (25a) and (26a), the clause indicates an unintended or spontaneous action where aPatient is affected; the occurrence of an Agent is optional and not essential. In (25b) and(26b), there is an effortful and intentional action that originates from an Agent; theabsence of an Agent is due to its topicality, not its unimportance.

25. There is a similar morpheme tafa- in Malagasy (Keenan and Dryer 2007), as in (iiia) below,where the conscious activity of the Agent is downplayed. Clause (iiib) implies that the com-pletion of the action was unexpected: either the action was dif²cult for the agent, unintended,or spontaneous.

(iii) a. Tafa-tsangana ny lai.pass-put.up the tent�The tent is put up.�

(Malagasy, Keenan and Dryer 2007)

b. Tafa-iditra-ko ny omby.pass-enter-1sg.gen the cow(s)�I got the cows in.� (lit: the cows were made to enter by me.)

(Malagasy, Keenan and Manorohanta 2001)

Shibatani (2006:222�28) argues for a spontaneous voice where the form conveys the sense ofan action accidentally brought about, which is not unlike the meaning conveyed by the mor-pheme na- in Cebuano, and further notes that in languages such as Japanese and Indonesian/Malay, the spontaneous voice shares the same marker as the passive. He also shows that thespontaneous form is also attested in many other languages, such as Bengali and Marathi(Indo-Aryan), Tibetan and Newar (Tibeto-Burman), and Tsova-Tush (Batsbi).


(25) activity verbsa. Na-kuha=na=nako ang sulat.

nfut-take=pfv=1sg.gen nom letter

�I have already received the letter.�(There was no prior knowledge that I would receive it.)

b. Gi-kuha=na=nako ang sulat. nfut-take=pfv=1sg.gen nom letter

�I have already taken the letter (on purpose).�(There was a prior intention of taking the letter.)

(26) verbs of perceptual judgmenta. Na-maot-an=ko sa babayi.26

nfut-ugly-lf=1sg.nom sa woman

�I perceive the ugliness of the woman.�(Subject does not have prior knowledge of ugliness of the woman.)

b. Gi-maot-an=ko sa babayi.nfut-ugly-lf=1sg.nom sa woman

�I am consciously aware of the ugliness of the woman.�(Subject is biased against the woman and exerts an effort to emphasizeher ugly appearance.)

This semantic contrast between gi- and na- is also captured in (27). Here, T forgets to takeher juice drink and asks Mario to get it for her. In the ²rst line, she initially utters a gi-af²xed verb gi-biya-an �to leave something behind�, which she does not ²nish andreplaces with a na-af²xed verb na-biya-an. The verb gi-biya-an would imply that sheintended to leave her juice and that it would be her fault that it was not with her. Replacingit with na-biya-an would mean that it was not done on purpose.

(27) T Tawg-on=sa=nako si Mario kaycall-fut.pf=²rst=1sg.gen nom pn because

naa=ko-y gi-biya- na-biya-an.exist=1sg.nom-obl nfut-leave nfut-leave-lf

�I�ll call Mario because I left, � I forgot to bring something.�(a few lines omitted)

T Kuhaa akong juice na-biya-an=nako.get-pf 1sg.poss juice nfut-leave-lf=1sg.gen�(Mario,) get my juice (for me); I (accidentally) left (it) (somewhere).�

26. We thank Ricky Nolasco for providing this example.

FIGURE 2. SYNTACTIC SPACE FOR na-CLAUSES IN CEBUANO

Nom NP

S na-intransitives

active no-clauses

P no-passives inverse clauses



4.1 INTRANSITIVE na-. The intransitive na-verbs have something in commonwith the gi-adversative verbs in that they have PF forms, but function as intransitives, assummarized in table 11. Bodily state verbs with a gi- morpheme have been shown toform adversative verbs (see 3.1). Similarly, those with a na- morpheme are intransitiveand when the na- is replaced by gi-, they become causative.27 The na-intransitive verbsare similar to the gi-adversative verbs in that they cannot take a Genitive Agent phrase;any genitive nominal (ni NP; sa NP) would indicate a location (in bodily state verbs) orcause/stimulus of an emotion (in emotion verbs), as in (28a) and (28b).

(28) a. na-buang / suko=siya ni Pedro.NFUT.INTR-be.crazy / be.mad=3SG.NOM OBL PN

�She�s crazy about/mad with Pedro.�

b. gi-luod=siya ni Pedro / sa dugo.ADVRS-feel.nauseous=3SG.NOM OBL PN OBL blood

�She felt �disgusted� with Pedro / She felt nauseous because of the blood.�

27. Some of these na intransitive verbs (in the right column on table 11) become causative verbswhen na- is changed to gi-, as in gi-hulog �cause to fall (down)� and gi-lumos �cause to drown�.A few require an additional causative pa- morpheme and a nominalizer ka-, as in gi-pa-ka-tawo �cause to become human� and gi-pa-ka-tulog �cause to be in a state of sleep�.

TABLE 11. INTRANSITIVE gi- AND na- VERBS

gi-V (adversative verbs) na-V (intransitive verbs)

Physical (bodily)

gi-atake �have a (heart) attack�gi-duka �feel sleepy�gi-gutom �feel hungry�gi-kapoy �feel tired�gi-katol �feel itchy�gi-laay �feel bored�gi-luod �feel nauseous�gi-sipon �have a running nose�gi-ubo �cough�gi-uhaw �feel thirsty�

na-banhaw �rise from the dead�na-buhi �become alive�na-daot �get bad; get destroyed�na-hagbong �fall�na-hubog �be drunk�na-hulog �fall�na-lumos �drown�na-matay �die�na-panos �be rotten�na-samad �be injured�na-sangit �be hooked�na-sunog �burn�na-takd-an �be contaminated�na-tawo �be born�na-tulog �be asleep�

Mental /emotional

gi-mingaw �miss (s.o.)� na-balaka �be startled�na-bu�ang �be crazy�na-guol �be sad�na-hadlok �be afraid�na-kurat �be astonished�na-lipay �be glad�na-suko �be angry�na-tingala �feel wonder�

Infestingverbs

gi-anay �be infested with termites�gi-kagaw �be infected with germs�gi-kuto �be infected with lice�gi-ulod �be infested with worms�

Activity gi-V �PF usage� na- V �PF usage�


Emotion verbs are mostly formed with a na- morpheme.28 A gi- can only beattached together with a nominalizer ka- in an equational clause; for example, gi-ka-guol �somebody or something that is the cause of sadness of an animate Experiencer�or gi-ka-balaka �somebody or something that is the cause of concern for an animateExperiencer�.29 These are illustrated in (29). The adversative verbs in table 11 indicat-ing �infestation� only use gi-.

(29) a. Na-guol=ko sa kahimtang ni Maria.nfut.intr-be.sad=3sg.nom obl condition gen pn

�I felt sad for Maria�s condition.�

b. Gi-ka-guol=ko ang kahimtang ni Maria.nfut-nmlz-be.sad=1sg.gen nom condition gen pn

�Maria�s condition (is something that) saddens me.�

In the examples below, the na-verbs are intransitive. In (30), the subject of the ²rstverb na-hilom �to become silent� has been personi²ed; in other words, it is the people thathave stopped talking about the news. The same is true with the second verb na-wala �todisappear�. Moreover, a ni phrase would be unacceptable. Example (31) is an instance ofthe use of an emotion verb.

(30) L m: na-hilom=na ang balita mayo unta oy @@fil nfut.intr-silent=pfv ang news good part voc

ma-wala=na=lang [ang gubot sa-]fut.intr-disappear=pfv=only ang chaos loc

J [diri sa Cebu kay-] kuan no kanang: peacefulhere loc pn because kuan q fil peaceful

L peacefulpeaceful

L �The news disappeared. It�s good though @@ No more disorder in-�J �here in Cebu, it�s peaceful, right?�L �Peaceful.�

(31) J mag-lakwatsa=mi mga barkada namo tong high schoolaf-go.out=1pl.excl.nom pl gang 1pl.excl.gen that.time high school

unsa oras=na=kami mo-uli mao na di=man ma-hadlokwhat time=pfv=1pl.excl.nom af-go.home that.way neg=part fut.intr-be.afraid

mao na akong mama=lang ga-ingon unsa o- unsa oras=nathat.way 1sg.poss mother=only af-say what what time=pfv

no dili ba kuan=na kayo gabii[=na kaayo]part neg q kuan=pfv very late=pfv very

L [gabii=na kaayo]late=pfv very

J di=ka ma-hadlok sus hadlok=ka=man noneg=2sg.nom fut.intr-be.afraid interj be.afraid=2sg.nom=part part

28. This is true with the exception of gi-mingaw �to miss somebody�. However, if the more modernterm miss (which is a loanword from English) is used, then na- is more appropriate, as in na-miss=nako si papa �I miss(ed) Papa.�

29. These expressions (gi-ka-V, where V is an emotion verb) are more often used in written andformal speech, such as sermons, of²cial speeches, and news broadcasts.


L @@J to-ng una grabe=pa nay mga marines=pa=gyud

that-lk time serious=still exist pl marines=still=emph

mag-checkpointbya- morag ma-hadlok=ka=gyudaf-checkpoint part like fut.intr-be.afraid=2sg.nom=emph

kong dili=ka taga-didtoif neg=2sg.nom from-there

J �When we went out at night (together with) my high school classmates,what time did we go home? So, we would not be afraid.So, my mother would just say, �What time is it? [Isn�t it too late?]�.�

L �[It�s too late.]�J �You won�t get afraid? You should!�L �Haha�J �At ²rst, the condition was serious. There were marines (assigned)

at checkpoints. You must feel afraid, if you�re not from there.�

Intransitive clauses are the preferred structure in Cebuano oral discourse (Tanang-kingsing 2006), and the subject of a na-clause, which is usually an Experiencer, is over-whelmingly ²rst person, as in the ²rst verb in boldface ma-hadlok in (31), especially inconversations. If they take a second person subject, they may either be asking a question, asin the second verb in (31), or making an inference, as in the third verb in boldface in (31).

4.2 INVERSE na- (VPA; VP). Constructions carrying an inverse voice function asde²ned in Cooreman (1982) and Thompson (1994) are distinguished from the active byword order and are attested in the following languages: Korean (Kwak 1994), Maasai(Payne, Hamaya, and Jacobs 1994), and Modern Greek (Roland 1994).30 Most other lan-guages have developed af²xes, in addition to word order, to distinguish between activeand inverse clauses, as in Mapudungun (Arnold 1994) and Northern Sahaptin (Rude1994). Cebuano has inverse constructions, as we have shown for the gi-perfective clauses,but the inverse function is represented by a marked word order. The P argument is locatedright after the verb, with the A being moved farther to the end of the sentence or left unex-pressed, and the clause is represented by the schema VPA or VP. This attests to the factthat word order may also be a device for signaling an inverse construction (Givon 1994).31

Inverse constructions are rare, and in our data the ratio between those with and without anA argument expressed is about 50:50. Example (32) is taken from our corpus. Both A andP are pronominal and thus topical. The P (second person addressee) is more topical thanthe A (third person). This must also be the main reason why inverse na-constructions arenot so frequent (see table 13). First and second person participants are normally Agents,and it is mainly in inverse clauses where they occasionally function as Patients.

30. In Tongan, the variation between VOS and VSO clauses is attributed to the feature [+FOCUS](Otsuka 2005). It remains to be seen whether either one of the constructions functions asinverse when the topicality metrics are applied to the language data.

31. See also Brainard and Molen (2005) for a description of word order of inverse constructionsin Obo Manobo.


(32) L pananglitan na-meet=kafor.example nfut.intr-meet=2sg.nom

J m:BC

L dilidyud=ko ingon nga- inga- ingani ra gyudnegemph=1sg.nom say comp fs like.this only emph

ang akong- unsa na oy ako-ng kina�iya ba nga-ang 1sg.poss what that voc 1sg.poss-lk personality part lk

J m:bc

L ay: ana di=dyud=ko- di ba parehas sa ubaninterj like.that neg=emph=1sg.nom neg part same obl other

pananglitan ma-meet=na=ka=nilafor.example fut.intr-meet=pfv=2sg.nom=3pl.gen

J m:bc

L �For example, you have met ��J �m=�L �I�m not, just like that, my uh-, what-you-call-it, my personality-�J �m=�L �Oh, I�m not like that, not like others. For example, if you have been

met by them,�J �m=�

4.3 ACTIVE na- VS. PASSIVE na-. As with the intransitive na-construction, theactive na-clauses have As that are ²rst or second person pronominals and so are highlytopical; they are also strongly integrated within the clauses. Because these clauses are PFconstructions, the Ps are also topical and accessible from the context. These active transi-tive clauses account for a large proportion of the na-clauses (see table 13). The possibleword orders for active na-clauses are VAP, PAV, VA (P is usually nonhuman or anabstract referent), and VP.

As for the basic difference between the active and passive na-clauses, as shown inexample (33), the presence of an A would imply that it has been mentioned or is accessi-ble in the preceding discourse. In other words, any A that has not been mentioned or isnot accessible from the preceding discourse would likely make the construction passive.In (33a), because there is no external Agent and most likely it is also absent in the priortalk, this clause is passive. In (33b), because the Agent is overt and has most likely alsobeen mentioned in the prior talk, this clause is active. In addition, (33c) provides a con-trast with (33b). (33c) uses a gi- marker implying that there was a prior intention of thepolice �to catch me,� while such an intention is lacking in (33b), where the police mighthave seen me by accident and �took the opportunity to arrest me.�32

(33) a. Na-dakp-an=ko. (no external Agent)nfut.intr-catch-lf=1sg.nom

�I got caught.� (who it was that caught me is entirely unimportant)


b. Na-dakp-an=ko sa pulis. (with external Agent)nfut.intr-catch=lf=1sg.nom obl police

�The police caught me.� (unintentionally; �the police� is slightly moretopical)

c. Gi-dakop=ko sa pulis. (with intent by the external Agent)nfut.pf-catch=1sg.nom obl police

�The police caught me.� (the police intended to arrest me)

The examples below provide further evidence that the focus of the discourse sur-rounding the na-passive is entirely on the P referent. In (34), the talk is focused on the factthat the topic referent can just be assigned anywhere; as to who assigns him is not impor-tant. If gi- were used instead of na-, then the A would have been mentioned in the preced-ing discourse, with the gi- suggesting a purposeful intention of the A (to assign the Panywhere he wishes). Similarly, in (35), the attention is entirely on the friend who was nothired; if an A were alluded to in the act of hiring the friend, then using a gi-verb wouldhave been much more appropriate.

(34) J kay ako-ng: lolo iyaha-ng papa kuan- kanang-because 1sg.poss-lk grandfather 3sg.poss-lk father kuan fil

sundalo gud soldier emph

L [m:]bc

J [kan]ang bisag asa ma-assignfil anywhere fut.intr-assign

J �because my grandfather, his father was a kuan he (his father) was asoldier.�

L �m=�J �he was assigned anywhere.�

(35) L o: diri=ra=mi nag-kita kay akong- amiga nakofil here=only=1pl.excl.nom af-meet because 1sg.poss friend 1sg.gen

wala=siya na-dawat siya ang nag-anoneg=3sg.nom nfut.intr-accept 3sg.nom ang af-what

J m:fil

L nag-pa-hibawo nga nay [nag- XXX]af-caus-notify comp exist af-

32. A similar situation seems to exist in Kapampangan, and in other languages in the northernareas of the Philippines (Reid and Liao 2004; Reid 2006), where the A argument is absent inpassive clauses. In Kapampangan, there seems to be a distinct verbal morpheme (pre²x me- in[ivb] below) to mark passive, as shown in the examples below cited from Mithun (1994). Ananonymous reviewer points out that, incidentally, the morpheme me- (see example in [ivb]) iscognate with Cebuano na-. They both developed from PEF *mina.

(iv) a. Linutu no reng sagin.cooked 3erg/3pl.abs 3pl.abs banana�She cooked the bananas.� (Kapampangan, Mithun 1994:271 example 55c)

b. Melutu no reng sagin.cooked already/3pl.abs 3pl.abs banana�The bananas have been cooked.� (Kapampangan, Mithun 1994:272 example 55d)


J [unsa=man] diay nag- nag- hiring=ba=sila nag-what=part evid af- af hire=q=3pl.nom af

L una ang mga ka-ila ra nila ang ila ra-ng²rst ang pl recip-know only 3pl.gen ang 3pl.poss only-lk

� hibo:ng=gay=ko ngano ako @@ na-dawatwonder=part=1sg.nom why 1sg.nom nfut.intr-accept

na-dawat=pa=ko unya ang ag ang nag-kuan 33

nfut.intr-accept=still=1sg.nom then ang fs ang af-dummy

sa ako wala=siya na-dawatobl 1sg.nom neg=3sg.nom nfut.intr-accept

L �Yes. We met here, because my friend, she was not hired. She wasthe one-�

J �m=�L �who informed (me) that there [was �XXX]�J �[what] they were hiring?�L �At ²rst, they only looked for people they knew. I even wondered

why I got hired. I got hired, and the one who (told) me (about thisjob), she didn�t get hired.�

4.4 SUMMARY. Based on the preceding analyses then, there is a distinct wordorder pattern for every construction type in both gi- and na-clauses, summarized in table12. When two arguments are overtly mentioned, the argument that is more topical ismost likely to occur closer to the verb. A passive analysis of a gi-clause is problematic inthat both the semantics of the gi-verb and the pragmatics of the gi-clause contradict thelack of emphasis of a passive clause on the A referent. Table 13 shows that the other twocriteria for passive, namely syntactic integration of the A and text frequency, also hold inna-constructions.

33. For a discussion of the functions of the dummy word kuan, see Hsieh and Tanangkingsing (2006).

TABLE 12. CONSTRUCTION TYPES AND WORD ORDER

construction type

relative topicality between a and p

gi-clause na-clause

active A > P VAP VAPPAVVAVP

inverse A < P VPAVP

VPAVP

passive A << P � VP

TABLE 13. na-/gi-VERBS AND SYNTACTIC INTEGRATIONOF A AND RAW FREQUENCY

a integrationin na-clauses

na- frequency (n/%) gi- frequency (n/%)

active A > P 95.2% 42 / 4.3% 40 / 4.1%inverse A < P 50.0% 10 / 1.0% 19 / 1.9%passive A << P 0% 29 / 2.9% n.a.


As shown in table 12, varying constructions are characterized by distinct word orders.Although the VP order with an A unexpressed is found in the active, inverse, and passiveconstructions, the statuses of A in each of these constructions are actually different fromeach other. It is easier to identify VP active clauses, because the A is highly topical, thereferent being either the speaker or the hearer, and so is unexpressed. In the VP inverseclauses, the missing A is accessible from the context and is only slightly less topical thanthe P; in the VP passive clauses, the A is always absent and inaccessible, and the focus isentirely on the nonpurposeful action imposed on the P. The main criterion, then, in distin-guishing between inverse and passive clauses, especially those with covert A arguments,is the accessibility status of the A argument in question. This also shows the value of plac-ing clauses in context rather than treating them in isolation.

Some might further question the status of a morpheme such as nathat appears in asmany as four different types of constructions. A similar suf²x, -Cia, exists in Samoan(Cook 1996) where it is used as a marker for passive, adversative,34 and transitive clauses.It is not unusual, in fact, for a morpheme to mark different functions, as new functionsarise before a formal marking is grammaticized to mark them (Du Bois 2003). More-over, having a �passive without unique passive morphology� (Arka and Kosmas 2005) isnot surprising at all, typologically speaking.35 The passive clause in Cebuano shares a ver-bal morpheme na- with active and inverse clauses, and it is distinguished primarily by thelow topicality of its A arguments and word order. The passive na- in Cebuano can then beseen as a �passive�-marking morpheme, if we consider the comparative evidence that na-passives with absent As are fairly widespread, especially in northern Philippine lan-guages (Reid 2006; Reid and Liao 2004).

5. CONCLUSION. In this paper, we hope to have shown that the gi- and na-clauses inCebuano differ in a number of respects. The gi-clauses normally denote a purposefulaction on the part of the Agent (of course, with exceptions that have been discussed insection 3), while the na-clauses focus on the nonpurposeful effect of an action on thePatient. One type of the so-called gi-passives in Payne�s study could at best be onlytreated as an inverse construction, where the P is slightly more topical. In conversationaldata, the A in such a construction would be expected to be mentioned in the precedingdiscourse, remain topical, and therefore could not be demoted. The other two types ofconstruction that were considered passive in Payne�s study are the adversative gi-verbs,which actually function as intransitive AF verbs, and ²xed expressions consisting of aclosed set of verbs of saying and naming.

We have argued that the passive in Cebuano is associated with one type of na-con-struction. Four types of na-construction are identi²ed. First, the intransitive na-verbsfunction like AF verbs. Such verbs include mental and emotion verbs, which can never

34. These are referred to as �personal passives without a transitive verb stem� by Cook (1996:67�68). They are intransitive verbs that take a patient-like absolutive subject. Examples aretimu�ia �be rained on�, saviligia �be affected by the wind�, lagoia �be covered with ³ies�, andmâfatia �be tired, exhausted, affected� (68), which all seem to re³ect adversity.

35. Arka and Kosmas (2005) present convincing evidence that the le Agent constructions inManggarai, a Central�Malayo-Polynesian subgroup of Austronesian languages spoken onFlores Island in Indonesia, are passive even without any verbal morphology to mark passive.


take gi-, as well as bodily state verbs, which become causative in meaning when attachedwith gi-. Two further types are the regular active transitive clauses and inverse na-con-structions, both of which differ only in the relative topicality of the A and P. Finally, thepassive na-construction has an absent A, which cannot be accessed from the context.

Payne was certainly right in identifying a passive construction based on a distinctword order. However, he was not on the right track when he sought it in a gi-clause, giventhat the semantics of the gi-verbs can never demote the A. We have provided discourse-pragmatic evidence for identifying one type of na-construction as the passive inCebuano. This type of na-construction satis²es the standard criteria for passive: it hasabsent As (i.e., it lacks syntactic integration and thus leads to the defocusing of the A),lower text frequency, and a word order distinct from the active na-clauses.

If we merge the syntactic spaces shown in ²gures 1 and 2, it would look somethinglike ²gure 3: the active clauses would form the bulk of gi- and na- transitive clauses,while some verbs can only be used with gi- and na- intransitively. The other types of tran-sitive gi- and na-constructions would only account for a minor proportion. The syntacticspace of the clauses in Cebuano thus formed comports in general with that for voice con-structions put forth by Croft (2001), save for the intransitive clauses, which were notincluded in Croft�s discussion. There are nonbasic clause types�passive clauses (lackingan A), ²xed expressions (only with verbs of saying and naming), and inverse clauses (theP argument has gained a certain degree of topicality in relation to the A argument)�thatare in decreasing order of distance away from the active end of the syntactic space. Thegi-construction is relatively more stable than the na-construction in being the basic type;it possesses the property of endowing its A argument with high topicality status. The na-construction, on the other hand, reveals some passive characteristics, especially by leav-ing out the A argument.

FIGURE 3. SYNTACTIC SPACE FOR gi-/na-CLAUSES IN CEBUANO

Nom NP

S gi-/na-intransitives

gi-/na-active clauses

inverse clauses

P na-passives²xed expressions



REFERENCES

Arka, I Wayan, and Jeladu Kosmas. 2005. Passive without passive morphology? Evi-dence from Manggarai. In The many faces of Austronesian voice systems: Some newempirical studies, ed. by I Wayan Arka and Malcolm Ross, 87�117. Canberra:Paci²c Linguistics.

Arnold, Jennifer. 1994. Inverse voice marking in Mapudungun. Berkeley LinguisticsSociety 20:28�41. Berkeley: Berkeley Linguistics Society.

Bell, Sarah. 1976. Cebuano subjects in two frameworks. PhD diss., Massachussetts Instituteof Technology.

Brainard, Sherri. 1994. Voice and ergativity in Karao. In Voice and inversion: Typologicalstudies in language, 28, ed. by Talmy Givon, 365�402. Amsterdam : John Benjamins.

Brainard, Sherri, and Ena Vander Molen. 2005. Word order inverse in Obo Manobo. InCurrent issues in Philippine linguistics and anthropology: Parangal kay LawrenceA. Reid, ed. by Hsiu-chuan Liao and Carl R. Galvez Rubino, 364�418. Manila:Linguistic Society of the Philippines and SIL Philippines.

Comrie, Bernard. 1988. Passive and voice. In Passive and voice, ed. by Masayoshi Shiba-tani, 9�23. Amsterdam: John Benjamins.

Cook, Kenneth William. 1996. The CIA suf²x as a passive marker in Samoan. OceanicLinguistics 35:57�76.

Cooreman, Ann. 1982. Topicality, ergativity, and transitivity in narrative discourse:Evidence from Chamorro. Studies in Language 6(3): 343�74.

Cooreman, Ann, Barbara Fox, and Talmy Givon. 1984. The discourse de²nition ofergativity. Studies in Language 8(1): 1�34.

Croft, William. 2001. Radical construction grammar. Oxford: Oxford University Press.De Wolf, Charles M. 1988. Voice in Austronesian languages of Philippine type: Passive,

ergative, or neither? In Passive and voice, ed. by Masayoshi Shibatani, 143�93.Amsterdam: John Benjamins.

Dixon, R. M. W. 1994. Ergativity. Cambridge: Cambridge University Press.Dryer, Matthew. 1996. Passive vs. inde²nite actor constructions in Plains Cree. In Papers of

the Twenty-seventh Algonquian Conference, ed. by David Pentland, 54�79. Winnipeg:University of Manitoba.

Du Bois, John. 1987. The discourse basis of ergativity. Language 63:805�55.��. 2003. Argument structure: Grammar in use. In Preferred argument structure:

Grammar as architecture for function, ed. by John Du Bois, Lorraine E. Kumpf,and William J. Ashby, 11�60. Amsterdam: John Benjamins.

Gerdts, Donna B. 1988. Antipassives and causatives in Ilokano: Evidence for an ergativeanalysis. In Studies in Austronesian Linguistics, ed. by Richard McGinn, 295�321.Athens, Ohio: Ohio University Center for International Studies.

Givon, Talmy. 1990. Syntax: A functional-typological introduction. Amsterdam: JohnBenjamins.

��. 1994. The pragmatics of de-transitive voice: Functional and typological aspectsof inversion. In Voice and inversion, ed. by Talmy Givon, 3�44. Amsterdam: JohnBenjamins.

Himmelmann, Nikolaus P. 1991. The Philippine challenge to universal grammar. WorkingPaper 15. Köln: Institut für Sprachwissenschaft.

Hsieh, Fuhui, and Michael Tanangkingsing. 2006. The empty root in Cebuano and Kava-lan: A counterexample to grammaticalization. Paper presented at the Tenth Interna-tional Conference on Austronesian Linguistics, January 1�20, Puerto Princesa City,Palawan, Philippines. (http://www.sil.org/asia/philippines/ical/papers.html)

Huang, Shuanfan. 2002. The pragmatics of focus in Tsou and Seediq. Language andLinguistics 3(4):665�94.


��. 2006. Transitivity as an emergent category in Formosan languages. Presented at theWorkshop on the Linguistics of the Endangered Languages, April 3�5, Kobe, Japan.

Huang, Shuanfan, Lily I-wen Su, and Li-may Sung. 2004. Syntax and cognition inSaisiyat. ROC National Science Council technical report.

Keenan, Edward L., and Matthew S. Dryer. 2007. Passive in the world�s languages. InLanguage typology and syntactic description, ed. by Timothy Shopen, 243�81.Cambridge: Cambridge University Press. (http://linguistics.buffalo.edu/people/faculty/dryer/dryer/KeenanDryerPassive.pdf)

Keenan, Edward L., and Cécile Manorohanta. 2001. A quantitative study of voice inMalagasy. Oceanic Linguistics 40:67�84.

Kikusawa, Ritsuko. To appear. Transitivity in the analysis of Betsimisaraka Malagasy.In Current issues in Austronesian linguistics: Papers from the 10th ICAL, ed. bySherri Brainard. Studies in Philippine Languages and Cultures (SPLC).

Kroeger, Paul. 1993a. Phrase structure and grammatical relations in Tagalog. Stanford:CSLI Publications.

��. 1993b. Another look at subjecthood in the Philippines. Philippine Journal ofLinguistics 24(2): 1�15.

Kwak, Inhee Lee. 1994. The pragmatics of voice in Korean. In Voice and inversion, ed.by Talmy Givon, 261�82. Amsterdam: John Benjamins.

Li, Paul Jen-kuei. 1973. Rukai structure. Taipei: Academia Sinica.Liao, Hsiu-chuan. 2002. The interpretation of tu and Kavalan ergativity. Oceanic Linguistics

41:14�58.��. 2004. Transitivity and ergativity in Formosan and Philippine languages. PhD

diss., University of Hawai�i at Mânoa.Malchukov, Andrej L. 1993. The syntax and semantics of adversative constructions in

Even. Gengo Kenkyu 103:1�36.Mithun, Marianne. 1994. The implications of ergativity for a Philippine voice system.

In Voice: Form and function, ed. by Barbara Fox and Paul Hopper, 247�77.Amsterdam: John Benjamins.

Nolasco, Ricardo Ma. 2005. What ergativity in Philippine languages really means.Paper presented at the First Taiwan-Japan Workshop on Austronesian Languages.National Taiwan University. June 23�24.

��. 2006. Ano ang S, A at O sa mga wika ng Pilipinas [What are S, A and O in thelanguages of the Philippines?]. Paper presented at the Ninth Philippine Linguis-tics Congress, University of the Philippines, Diliman, January 25�27.

Otsuka, Yuko. 2005. Two passive-like constructions in Tongan. In The many faces ofAustronesian voice systems: Some new empirical studies, ed. by I Wayan Arka andMalcolm Ross, 119�35. Canberra: Paci²c Linguistics.

Payne, Doris, Mitsuyo Hamaya, and Peter Jacobs. 1994. Active, inverse, and passive inMaasai. In Voice and inversion, ed. by Talmy Givon, 283�315. Amsterdam: JohnBenjamins.

Payne, Thomas. 1994. The pragmatics of voice in a Philippine language: Actor-focusand goal-focus in Cebuano narrative. In Voice and inversion, ed. by Talmy Givon,317�64. Amsterdam: John Benjamins.

Quick, Phil. 2005. Topic continuity, voice and word order in Pendau. In The many facesof Austronesian voice systems: Some new empirical studies, ed. by I Wayan Arkaand Malcolm Ross, 221�42. Canberra: Paci²c Linguistics.

Reid, Lawrence A. In press. On reconstructing the morphosyntax of proto-NorthernLuzon. Philippine Journal of Linguistics 37(2).

Reid, Lawrence A., and Hsiu-chuan Liao. 2004. A brief syntactic typology of Philip-pine languages. Language and Linguistics 5(2): 433�90.

Roland, Katy. 1994. The pragmatics of Modern Greek voice: Active, inverse, and passive.In Voice and inversion, ed. by Talmy Givon, 234�60. Amsterdam: John Benjamins.


Rude, Noel. 1994. Direct, inverse, and passive in Northwest Sahaptin. In Voice andinversion, ed. by Talmy Givon, 101�19. Amsterdam: John Benjamins.

Shibatani, Masayoshi. 1985. Passives and related constructions: A prototype analysis.Language 61(4): 821�48.

��. 1988. Voice in Philippine languages. In Passive and voice, ed. by MasayoshiShibatani, 85�142. Amsterdam: John Benjamins.

��. 2006. On the conceptual framework for voice phenomena. Linguistics 44(2):217�69.

Siewierska, Anna. 1985. The passive: A comparative linguistic analysis. London:Croom Helm.

Smith, Tomoko Yamashita. 1998. How �give� and �receive� provide structure for moreabstract notions: The case of benefactives, adversatives, causatives, and passives.Berkeley Linguistics Society 24:219�31. Berkeley: Berkeley Linguistics Society.

Starosta, Stanley. 1997. Formosan clause structure: Transitivity, ergativity, and casemarking. In Chinese languages and linguistics IV: Typological studies of languagesin China, ed. by Chiu-yu Tseng, 125�54. Taipei: Academia Sinica.

��. 1998. Ergativity, transitivity, and clitic coreference in four Western Austrone-sian languages. In Case, typology, and grammar, ed. by Anna Siewierska and Jae-Jung Song, 277�307. Amsterdam: John Benjamins.

��. 1999. Transitivity, ergativity and the best analysis of Atayal case marking. InSelected papers from the Eighth International Conference on Austronesian Linguistics,ed. by Elizabeth Zeitoun and Paul Jen-kuei Li, 371�92. Taipei: Academia Sinica.

Tanangkingsing, Michael. 2006. What intonation units can tell us about Cebuanogrammar. Paper presented at the Ninth Philippine Linguistics Congress, Univer-sity of the Philippines, Diliman, January 25�27.

Thompson, Chad. 1994. Passives and inverse constructions. In Voice and inversion, ed.by Talmy Givon, 47�64. Amsterdam: John Benjamins.

Verhaar, John W. M. 1988. Syntactic ergativity in Contemporary Indonesian. In Studies inAustronesian linguistics, ed. by Richard McGinn, James Coady, Marmo Soemarmo,and Philip Hubbard, 347�84. Monographs in International Studies, Southeast AsiaSeries Number 76. Athens, Ohio: Center for Southeast Asia Studies, Ohio UniversityCenter for International Studies.

Wang, Shan-shan. 2004. An ergative view of Thao syntax. PhD diss., University ofHawai�i.

Wouk, Fay. 1999. Sasak is different: A discourse perspective on voice. Oceanic Linguistics38:91�114.

Zeitoun, Elizabeth. 2000. Rukai reference grammar. [In Chinese.] Taipei: Yuanliou.

National Taiwan UniversityP.O. Box 53-1299Taipei 106 [email protected]

Graduate Institute of LinguisticsNational Taiwan UniversityNo. 1, Roosevelt Road Section 4Taipei 106 [email protected]

mailto:[email protected]

mailto:[email protected]

Cebuano Passives Revisited

Documents