Language and Cognition: Festschrift in Honor of James H-Y. Tai on His 70th Birthday. Jung-hsing Chang (ed.) 2011, Taipei: The Crane Publishing. Transitivity, Ergativity, and the Status of O in Tsou Henry Y. Chang Academia Sinica This paper addresses issues concerning transitivity and ergativity in Tsou. I identify Tsou as an ergative language and recast the two sets of case marking forms as a tripartite case distinction, namely, absolutive vs. ergative vs. oblique. Due to its ergative propensity, a transitive object O in the language is required to move from its base position within the vP to a higher position above the vP. In the meantime, I reanalyze the so-called voice markers as transitivity markers, which spread from the sentence-initial auxiliary to the intermediate adverbial verbs and then finally to the lexical verbs, giving rise to a typologically rare property of transitivity concord. The transitive category is comprised of two subcategories, i.e., the simple transitive structure and the applicative structure, either of which represents not only the typical functions of its kind but also its atypical functions. Keywords: Tsou, voice, transitivity, ergativity, applicative, object shift 1. Introduction1 This paper addresses the following issues from a cross-linguistic perspective: (i) Transitivity: Are voice markers transitivity markers in Tsou? (ii) Ergativity: Is Tsou an ergative language? (iii) A/S/O status: What is the syntactic status of the three primitives A, S, and O in Tsou? These issues are closely related. Syntactic transitivity is the cornerstone of the ergative-accusative division, as hinted in the basic common labels for linguistic typology: S stands for intransitive subject, A for transitive subject, and O for transitive object (Dixon 1979, 1994). In other words, the establishment of syntactic transitivity is a prerequisite for the identification of ergativity/accusativity. The paper follows this standard procedure. 1 Part of this paper was presented at the public lecture series, Cairns Institute, James Cook University, Australia, 3 September 2010 and at the colloquium series at Institute of Linguistics, Academia Sinica, 15 November 2010. I am grateful to Alexandra Aikhenvald, Bob Dixon, Huei-ju Huang, Hsiu-chuan Liao, Chia-jung Pan, Jackson Sun, Jane Tang, Stacy Teng, and Elizabeth Zeitoun for their valuable comments. My gratitude also goes to an anonymous reviewer for his/her useful suggestions. I am indebted to my major language consultants Pasuya Tiakiana and Mo’o Peongsi. I thank my research assistants Sihwei Chen and Chia-fen Wu for their typographic assistance. Any remaining error is my own responsibility. Henry Y. Chang In the Austronesian literature, the study of transitivity and ergativity dates back to Starosta and his associates’ earlier work (Starosta, Pawley, and Reid 1982). There are a number of recent publications in appreciation of this line of research, to name just a few: Liao (2004), S. Huang (2005), Ross and Teng (2005), Aldridge (2004, 2005, 2008). Despite the increasing interests, there is relatively little attention paid to the syntactic status of the three primitives A, S, and O, in particular, their syntactic positions. This paper attempts to fill the gap, with special focus on pinpointing the syntactic status of O in Tsou. Tsou is an Austronesian language spoken in the south-western highlands of Taiwan. It is the representative language of the Tsouic subgroup, one of the tenth major branches in the Austronesian family (Blust 1999). It has a population of around 4, 000 and is moderately endangered. The paper is organized as follows. Section 2 is concerned with research questions (i-ii), with special reference to markers of transitivity and case in Tsou. Section 3 distinguishes two types of O in terms of explicit criteria and explores the syntactic status of O in Tsou. Section 4 reaches a conclusion. 2. Transitivity and ergativity Transitivity has two senses in the linguistic literature, i.e., semantic transitivity and syntactic transitivity. Syntactic transitivity is determined by the number of core arguments, while the determinants of semantic transitivity are of various sources. In their seminal work, Hopper and Thompson (1980) enumerate ten factors that affect semantic transitivity: participant number, kinesis (action), aspect, punctuality, volitionality, affirmation, mode, agency, affectedness, and individuation. In this paper, the term transitivity is referred to syntactic transitivity unless otherwise specified. With respect to grammatical marking of core arguments and interclausal association, languages can be classified into two broad types: ergative vs. accusative.2 An ergative language aligns O and S, whereas an accusative language groups A and S together (Dixon 1979, 1994). I have made it clear about the first part of Dixon’s terminology in connection with transitivity. I am to clarify the second part of his terminology in relation to the terms subject and object. Dixon (2010a) suggests the following criteria for identifying transitive subject A and transitive object O: 2 It is important to distinguish between valid and invalid diagnostics for linguistic typology. As noted in Dixon (1994) and Aikhenvald & Dixon (forthcoming), only grammatical marking of core argument and interclausal association are useful for linguistic typology. Grammatical operations that are purely semantically governed and show no cross-linguistic variations (e.g., imperative deletion and reflexive binding) should not be taken as criteria for classifying languages. 278 Transitivity, Ergativity, and the Status of O in Tsou (1) Criteria for identifying A and O (Dixon 2010a: 76) a. A: “that argument whose referent could initiate and control the activity” b. O: “something that is saliently affected by the activity” However, he relaxes the rigid criteria in (1) in his second volume of the same title and suggests that the following principles should be in place: (2) Criteria for identifying A and O (Dixon 2010b: 126-133) a. A: “that role which is most likely to be related to the success of the activity” b. O: “that participant which is not most likely to be related to the success of the activity” With the relaxation, the criteria become more inclusive; at the same time, they become vaguer though, hence less useful. It might be helpful to note that the identification of subject and object typically follows the principles in (1), but might be determined by some other principles in particular situations. As noted in Dowty (1991), Croft (1998), and Levin and Rappaport (2005), in addition to [volitional] and [affected], the semantic properties [sentient] and [causative] might also play an important role in determining the syntactic realization of an argument. Putting Dixon’s and Dowty/Croft/Levin & Rappaport’s insights together, I devise the following flow chart for identifying A and O. Figure 1: Step-by-step diagnostics for identifying A and O Take English for illustration. In (3), the action verb beat requires a volitional 279 Henry Y. Chang beater. By the procedural diagnostics, the sentence should go along Path 1 and take the agent John as A and the affectee Mary as O, as indicated by the subscript capitals. (3) John beat Mary . A O Note that an affectee refers to any participant which is highly affected by the activity. As suggested in Hopper and Thompson (1980), Givón (1993), and Beavers (forthcoming), the affectedness of a participant can be measured by the semantic property [change of location/state/possession] and the pragmatic property [specific]. These semantic/pragmatic criteria are crucial in the identification of O, in particular, when there is more than one participant up for the O slot. In (4a), the affectee should be the vase, given that it normally breaks into pieces if it is hit by a strong stick. A less common scenario is that the vase is sturdy but the stick is weak. When the two collide, it might be the stick rather than the vase that is damaged. Situations like this are illustrated in (4b), where the stick is treated as an affectee and hence appears in the O function. (4) a. John hit the vase with a stick. A O b. John hit the stick against a vase. A O Likewise, the semantic/pragmatic properties also figure in dative alternations. As is well-known, the dative construction is more wide-spread than its double-object counterpart across languages. The reason for the asymmetry might be that a giving event typically encodes a change of location for the gift from the giver to the recipient. Thus, the gift is normally taken as an affectee, hence occurring in the O function, as in (5a). By contrast, the double-object counterpart is in place only if the recipient is specific in interpretation and has the capacity of acquiring the gift, as in (5b-d). (5) a. John gave the book to Mary. A O b. John gave Mary a book. A O c. *John gave students the book. A O d. *John gave New York the book. A O In contrast, the sentence in (6) should go along Path 2 and then Track 1. In this example, no volitional participant is present. The subject the 9/11 terrorist’s attacks 280 Transitivity, Ergativity, and the Status of O in Tsou denotes an event rather than a volitional participant. Nonetheless, the sentence involves causation: the verb shock is a causative-inchoative verb, referring to an event where a participant is caused to become shocked. By the step-by-step diagnostics indicated in Figure 1, the causer the 9/11 terrorist’s attacks should be identified as A and the causee Americans as O. (6) The 9/11 terrorist’s attacks shocked Americans . A O On the other hand, the sentence in (7) does not involve volition/causation and should go along Path 2 and then Track 2. In this example, the verb dislike describes a mental state rather than a change of state. In light of the step-by-step diagnostics, the experiencer John should appear in the A function and the theme/stimulus Mary in the O function. (7) John dislikes Mary . A O With the above background information in mind, we can move on and explore syntactic transitivity and ergativity in Tsou. 2.1 Ergativity 2.1.1 Tsou as an ergative language As mentioned above, the transitivity of a sentence hinges upon the number of core arguments it takes. A core argument is defined as a syntactically prominent argument. There are many diagnostics for determining the syntactic prominence of an argument. Among them, case-marking, verbal agreement and obligatory control are widely recognized as reliable tests for core-argumenthood. These are exactly what I am going to apply in dealing with transitivity and ergativity in Tsou. In Tsou, an agent can trigger verbal agreement on the auxiliary, as illustrated in (8). (8) Tsou a. mi-ta m-ongsi ’e pasuya i i INTR-3S INTR-cry ABS PN3 ‘Pasuya is crying.’ 3 This paper adopts Leipzig Glossing Rules in glossing examples and assigning abbreviations, but with the following amendments: BA=benefactive applicative, CL=classifier, COS=change of state, DIM=diminutive aspect, INST=instrumental case, LA=locative applicative, PN=personal name, PLN=place name. 281 Henry Y. Chang b. i-ta teaph-a to kexpx ta pasuya ’e cxyx4 i i TR-3S put.into-TR OBL backpack ERG PN ABS lunch.box ‘Pasuya put the lunch box into his backpack.’ c. i-ta teaph-i to cxyx ta pasuya ’e kexpx i i TR-3S put.into-LA OBL lunch.box ERG PN ABS backpack ‘Pasuya put a lunch box into his backpack.’ d. i-ta teoc-neni to evi ta pasuya ’e pexcngx i i TR-3S chop-BA OBL tree ERG PN ABS axe ‘Pasuya chopped a tree with the axe.’ By this test, the personal name Pasuya that represents the agents in (8a-d) should be identified as a core argument. Moreover, the personal name Pasuya should be treated as an intransitive S in (8a), given that it is the sole core argument in the sentence. This also suggests that the auxiliary mi and the verbal marking m- can be analyzed as intransitive indicators. Note also that the theme in (8b), the goal in (8c), and the instrument in (8d) are all marked in the same way as the agent in (8a) – all marked by ’e. Given the establishment of the agent in (8a) as a core argument, the ’e-marked arguments in (8b-d) might be treated as core arguments as well. This treatment will receive empirical support from other syntactic tests such as relativization, a point I will return to in section 3.2.1. It follows that the sentences in (8b-d) should be all transitive sentences and take two core arguments. Now, it is time to identify the A and O in (8b-d). As mentioned above, these sentences should be transitive sentences. The verbs teapha/teaphi ‘put into’ and teocneni ‘chop’ all involve volitional agents. In light of the step-by-step diagnostics, the agents in these sentences should be all identified as A. In regard to the O function, it is typically borne by a transported theme, as in (8b). Still, other participants can take over the function if they show higher degree of affectedness or topicality in the event, as in (8c-d). The above observations pave way for identifying the typological status of Tsou. The alignment of O with S in morphological marking illustrated in (8) leads to the conclusion that Tsou should be an ergative language.5 In an ergative language, O and S typically appear in the absolutive case, as opposed to A, which occurs in the 4 For typographic convenience, I substitute x for the high unrounded central vowel i and ng for the velar nasal throughout the paper. 5 At this point, I do not distinguish between morphological and syntactic ergativity. Nonetheless, as will become clearer in the subsequent sections, in addition to being morphologically ergative, Tsou should also be syntactically ergative, given that syntactic processes such as relativization groups S and O, excluding A. 282 Transitivity, Ergativity, and the Status of O in Tsou ergative case. Given the alignment, morphological markers like ’e in (8) can be treated as an absolutive case marker and those marking the agents in (8b-d) should be analyzed as ergative case markers. This is the rationale underlying the case labels given in (8). In addition to absolutive case and ergative case, there is still a third one glossed as oblique in (8). In the next section, I will argue that there is a three-way case distinction attested in Tsou. 2.1.2 A three-way case distinction In a traditional analysis, case markers are divided into two sets with respect to their morphological shapes in Tsou, as shown in Table 1 (Tung 1964, Zeitoun 1992, 2000, 2005, M. Chang 2004, among others). Case markers ’e/si/ta/’o/na/co are treated as nominative, whereas ta/to/no/nca are analyzed as oblique. However, the dichotomous analysis is problematic, regardless of whether the labels nominative and genitive are adequate or not. There is solid evidence that what were taken as oblique case markers should be split into two classes: ergative/genitive vs. oblique. Table 1: The case-marking system in Tsou (Tung 1964: 147, 218) Oblique Nominative Being seen by both (the speaker and the hearer) near ta ’e middle ta si distant ta ta Not being seen by both (the speaker and the hearer) but having been seen by the speaker to ’o and having not been seen by the speaker no na but known to both ne but actually being somewhere nearby nca co Non-existent ci First, ergatively marked agents and experiencers are syntactically prominent in transitive sentences in Tsou. Note that ergative arguments can trigger verbal agreement, as illustrated above in (8). Second, as pointed out in H. Chang and Tsai (2001) and H. Chang (2004), transitive agents and experiencers in Tsou can antecede a missing argument in complement clause. As in (9), the transitive agent (represented by the pronominal suffix -ta) controls the reference of the missing argument 283 Henry Y. Chang (represented by the empty pronoun PRO) in the complement clause. (9) Tsou a. i-ta ahxy-a [PRO pa-bonx to yeokx] ’o oko i i TR-3S force-TR CAUS-eat OBL fish ABS child ‘He forced the child to eat fish.’ b. *mi-ta ahxyx [PRO m-ngoi to oko] i i INTR-3S force(INTR ) INTR-leave OBL child Intended for ‘He forced the child to leave.’ In contrast, an oblique argument is not eligible for either of the above-mentioned commanding operations: they do not control verbal agreement and the reference of an embedded missing argument. Third, ergative arguments differ from oblique arguments in bound pronoun realization in Tsou. As shown in Table 2 below, absolutive arguments can occur as bound pronouns as well as free-standing pronouns.6 Like absolutive arguments, ergative arguments can also be represented by bound pronouns, which I take as verbal agreement, while they do not occur as free-standing pronouns. By contrast, bound pronouns do not encode oblique arguments in Tsou at all. Table 2: The pronominal system of Tsou (cf. Zeitoun 2005) Free (ABS/OBL) Bound (ABS/ERG/GEN) 1S a’o -’o/’u 2S suu -su/ko 3S taini -ta/taini (visible) -si (invisible) 1P (inclusive) a’ati -to 1P (exclusive) a’ami -mza/mio 2P muu -mu 3P hin’i (visible) -hin’i (visible) hee (invisible) -he (invisible) An important generalization follows: bound pronouns refer exclusively to core 6 Unlike bound pronouns, free pronouns do not represent ergative arguments in Tsou. In addition to the absolutive case, they can also occur in the oblique case. For example: te-’o uso suu hosoni. IRR-1S go(INTR) 2S later ‘I will go to your place later.’ 284 Transitivity, Ergativity, and the Status of O in Tsou arguments in Tsou. Last but not the least, ergative arguments are differentiated from oblique arguments in the auxiliary selection. As noted by Tung (1964) and Zeitoun (2005), almost every sentence starts with an auxiliary in Tsou. The auxiliary can be classified into two categories with respect to their temporal/modal anchoring: realis vs. irrealis. The realis category can be further divided into two subclasses with respect to transitivity: transitive vs. intransitive, as indicated in Table 3. Table 3: The TAM system in Tsou (cf. Zeitoun 2005: 279) Reality Realis Irrealis Transitivity INTR TR INTR/TR mi-, i- Habitual Predictive Hypothetical Counterfactual Immediate mio, mo ------- te-, nte ------- tena, ta- mo(h)-, o(h)- la- ------- ------- nto(h)-, ntoso Remote moso Interestingly, ergative arguments only co-occur with transitive auxiliary, while oblique arguments can pattern with both subclasses. As indicated in the glosses in (10b-d), the ergatively case-marked agent amo goes with the transitive auxiliary i-, but is absent from the intransitive sentence in (10a). In (10a), the agent amo appears in the absolutive case instead. (10) Tsou (based on Zeitoun 2000: 93-4) a. mo mo-si ta pangka to emi ’o amo INTR-3S INTR-put OBL table OBL wine ABS PN ‘Father put wine on the table.’ b. i -si si-a ta pangka to amo ’o emi TR-3S put-TR OBL table ERG father ABS wine ‘Father put the wine on the table.’ c. i -si si-i ta amo ta emi ’o pangka TR-3S put-LA ERG father OBL wine ABS table ‘Father put wine on the table.’ d. i -si si-eni ta emi ta amo na a’o TR-3S put-BA OBL wine ERG father ABS 1S.ABS ‘Father put wine for me.’ 285 Henry Y. Chang If both the ergative arguments and the oblique arguments in (10) are treated alike as being in the oblique case, as in the traditional analysis, the distinction in the auxiliary selection (mo vs. i-) will be left unexplained. In contrast, in the tripartite case-marking analysis adopted here, the distinction falls out naturally. Ergative arguments are in the A function and expected to go alongside with the transitive auxiliary.7 It is by now evident that ergative case should be established and distinguished from oblique case in Tsou. It is concluded that Tsou makes a three-way contrast in case marking: absolutive vs. ergative vs. oblique. Accordingly, I reorganize Table 1 as Table 4. Table 4: A three-way case-marking distinction in Tsou Ergative/ Oblique Absolutive Genitive Being seen by both (the speaker and the hearer) near ta ta ’e middle ta ta si distant ta ta ta Not being seen by both (the speaker and the hearer) but having been seen by the speaker to to ’o and having not been seen by the speaker no no na but known to both -- ne8 -- but actually being somewhere nearby nca co Non-existent ci Before leaving this section, a clarification is in order. Like many ergative languages, a possessor is marked in the same manner as an ergative argument in Tsou. For instance: 7 As suggested by an anonymous reviewer, an ergative argument might also differ from an oblique argument in specificity/definiteness in Tsou—the former is normally specific or definite, whereas the latter tends to be non-specific. 8 For the time being, I follow the traditional analysis and identify ne as a locative case marker (Zeitoun 2000, 2005). Still, it should be noted that ne can co-occur with an absolutive case marker, as illustrated below. (i) mo yupa cum’u ’o ne tfuya ho tapangx INTR RECP near ABS NE PLN CONJ PLN ‘Tfuya and Tapangx are geographically close to each other.’ This poses a challenge for the case-marking analysis. I leave this for future research. 286