Syntactic Ergativity in Balinese: an Argument Structure Based Theory

Size: px
Start display at page:

Download "Syntactic Ergativity in Balinese: an Argument Structure Based Theory"

Transcription

1 [Natural Language and Linguistic Theory, vol. 16, pp , 1998] Syntactic Ergativity in Balinese: an Argument Structure Based Theory Stephen Wechsler University of Texas at Austin I Wayan Arka Udayana University and University of Sydney mailing address: Prof. Stephen Wechsler Department of Linguistics Calhoun 501 University of Texas Austin, Texas (512) (512) (fax)

2 (Wechsler) (Arka) 2

3 1 Syntactic Ergativity in Balinese: an Argument-Structure Based Theory * Abstract: This paper argues that variation between syntactically accusative and ergative clauses in Balinese results from different mappings between argument structure and syntactic realization. Our version of argument structure, essentially the ARG- S feature of Head-Driven Phrase Structure Grammar, is the locus of anaphoric binding conditions. We further assume that passive, causative, and applicative are morpholexical operations on argument structure. This set of assumptions allows us to explain: (i) why conditions on Balinese binding are generally thematic and independent of surface grammatical relations or phrase structure position; and (ii) the specific exceptions to such thematic conditions that arise with raised arguments, direct/oblique alternations, and passive by-phrases.

4 2 0. Introduction: syntactic ergativity. Research on syntax is bringing us ever closer to the universal system underlying the syntactic expression of semantic arguments. At the interface between syntax and lexical meaning lies an initial level, sometimes called argument structure, at which a predicate s semantic or conceptual form is concretized for syntactic expression. The challenge is to discover the universal properties of that level, while still explaining the observed diversity of the world s languages. Syntactically ergative languages in which the mapping from arguments to surface grammatical relations is roughly opposite to the more familiar accusative languages throw this universal interface into sharp relief. Many Austronesian languages have both accusative and ergative clauses, allowing a comparison within a single language. In this paper we look at one such mixed accusative/ergative language, Balinese (Austronesian; Bali and Lombok). In a syntactically accusative language the agent argument of a basic transitive clause shares with the sole argument of a typical intransitive such important interclausal properties as accessibility to relativization, raising, and sometimes control. 1 But in a small handful of syntactically ergative languages the most celebrated example being Dyirbal these subjectoriented properties pick out not the agent but the patient (Dixon 1979, 1994). (When shallower intraclausal properties such as case-marking or agreement drive the grouping, such a language (or clause) is said to be morphologically ergative. Morphologically ergative languages are considerably more common than syntactically ergative languages.)

5 3 In our terminology, arguments characterized by the subject-oriented syntactic properties are called GRAMMATICAL SUBJECTS, or simply SUBJECTS (on this terminology see footnote 4). We will show (in Section 1) that the preverbal NPs in the Balinese sentences (1) and (2) (ia (s)he and bawi-ne the pig, respectively) are both subjects, in this sense: 2 (1) Ia pules. 3 sleep (s)he is sleeping. (2) Bawi-ne punika tumbas tiang. (high register) 3 pig-def that OV.buy 1 I bought the pig. (2) contains a verb in Objective Voice (OV). But the same verb in Agentive Voice (AV), indicated by a homorganic nasal prefix replacing the initial consonant, takes its agent as subject: (3) Tiang numbas bawi-ne punika. (h.r.) 1 AV-buy pig-def that I bought the pig. To distinguish from the AV verb form (here numbas), the unmarked form of the verb is called Objective Voice (OV); hence tumbas in (2) is glossed OV.buy. We argue that the preverbal position is the subject position, regardless of which voice marking appears. If we are correct in that claim, then transitive patient in

6 4 (2) and intransitive sole argument in (1) are grouped together as subjects, so (2) is a syntactically ergative clause. In contrast, (3) is a syntactically accusative clause. We offer and defend a general theory of ergativity in terms of a static mapping from a particular version of argument structure to grammatical relations, i.e., the features determining grammatical expression. Argument structure is the site for binding (universally) and control (of some varieties, in some languages); most other subject properties depend on grammatical relations. Section 1 reviews evidence that the Balinese pre-verbal NP, whether patient (as in (2)) or agent (as in (3)), is the subject. In section 2 we show that clauses like (2) are underived transitives and not passives, and that (3) is not an antipassive construction in the usual sense (the AV morpheme does not demote the Theme). Section 3 introduces binding facts. Section 4 presents a general theory of argument structure and linking. In Sections 4 and 5 this theory is supported by evidence from binding, subject-to-object raising verbs, applicatives, causatives, ditransitives, and passives. In Section 6 we focus on binding in raising constructions. 1. The Balinese preverbal argument as subject. Our theoretical framework includes an argument structure but no D- structure, so the term subject unambiguously denotes, for us, the surface or grammatical subject: the argument picked out by relativization, raising, and other subject-oriented properties. The most prominent argument in argument structure will be called the A-SUBJECT. 4 The A-subject is crucially distinct from the (grammatical) subject. (For example, the Agent I is the A-subject in both (2)

7 5 and (3) above, but it is the (grammatical) subject only in (3); in (2) the grammatical subject is the Patient, the pig. ) In this section we show that the preverbal arguments of OV, AV, and intransitive verbs group together as (surface, grammatical) subjects Relativization. If a language allows relativization at all, it allows subjects to relativize (Keenan and Comrie 1977). In Balinese only the purported subjects can be relativized, as shown in (4)a and (5)a (examples (4) and (5) adapted from Artawa 1994). Objects as in (4)b and (5)b, and obliques, as in (6)b, cannot be relativized. 5 (4)a. anak-e cenik [ane gugut cicing] ento. person-def small [REL OV.bite dog] that the child whom the dog bit b. *cicing [ane anak-e cenik ento gugut]. dog [REL person-def small that OV.bite] the dog that bit the child (5)a. I Warta [ane maca koran]. Art Warta [REL AV.read newspaper] Warta who read the newspaper

8 6 b. *koran [ane I Warta maca]. newspaper [REL Art Warta AV.read] the newspaper that Warta read (6)a. Ia ngejang pipis-ne di paon 3 AV.put money-3poss at kitchen He put his money in the kitchen. b. *di paon ane ia ngejang pipisne at kitchen REL 3 AV.put money-3poss in the kitchen, where he put his money 1.2. Subject-to-subject raising. Raising to subject is known to be restricted to embedded subjects crosslinguistically (Zaenen, Maling, and Thráinsson 1985; Chung 1978). On the other hand there do not seem to be any languages where raising is restricted to Topics only. Balinese has many raising predicates which allow raising only of the purported subject argument (as noted by Artawa 1994). In this intransitive verb example the subject ia (s)he can be raised to the position to the left of the matrix predicate ngenah seem : 6 (7)a. ngenah ia mobog. seem 3 lie It seems that (s)he is lying.

9 7 b. Ia ngenah mobog. 3 seem lie (s)he seems to be lying. Now consider embedded transitive verbs, first an OV verb (8) and then an AV verb (9): (8) a. Ngenah sajan [kapelihan-ne engkebang ci]. seem much mistake-3poss OV.hide 2 It is very apparent that you are hiding his/her wrongdoing. b. Kapelihan-ne ngenah sajan engkebang ci. mistake-3poss seem much OV.hide 2 c.?*ci ngenah sajan kapelihan-ne engkebang. 2 seem much mistake-3poss OV.hide In example (8)a the bracketed clause is complement of the predicate ngenah seem / apparent. Since the embedded verb engkebang hide is in OV form, its theme kapelihan-ne his/her mistake is the subject and its agent ci you is object. Being the embedded subject, the theme kapelihan-ne can be raised, as shown in (8)b, while the embedded agent ci, being a non-subject, cannot raise, as shown in (8)c. In (9)a the embedded verb ngengkebang hide appears in its AV form, so the arguments are reversed from (8)a: now the agent ci is the embedded subject and theme kapelihan-ne is the object.

10 8 (9) a. ngenah sajan [ci ngengkebang kapelihan-ne]. seem much 2 AV.hide mistake-3poss It is very apparent that you are hiding his/her wrongdoing. b. Ci ngenah sajan ngengkebang kapelihan-ne. 2 seem much AV.hide mistake-3poss c.?*kapelihan-ne ngenah sajan ci ngengkebang. mistake-3poss seem much 2 AV.hide As expected, the embedded agent ci, being the subject, can raise (as in (9)b), while the theme kapelihan-ne, being a non-subject, cannot raise (as in (9)c). In short, raising picks the subject of the embedded proposition, regardless of thematic role. To take another example, the predicate enggal quick has all the same raising properties shown above for ngenah seem. (10)a gives the unraised version; in (10)b the embedded subject has been raised, while (10)c shows a failed attempt to raise the embedded object: (10)a. Enggal tiang ngadep umah. quick 1 AV.sell house b. Tiang enggal ngadep umah. 1 quick AV.sell house It is quick for me to sell a house. c. *Umah enggal tiang ngadep. house quick 1 AV.sell

11 9 The raising predicate keweh difficult provides a third example, as in this sentence from a Balinese novel: (11) Peteng-ne ento Nyoman Santosa keweh ngidemang paningalan night-def that Nyoman Santosa difficult AV.close eyes That night, it was difficult for Nyoman Santosa to close (his) eyes. In (11) the embedded subject Nyoman Santosa is raised. The sentence becomes bad if the object paningalan eyes is raised instead: (12) * Peteng-ne ento paningalan keweh Nyoman Santosa ngidemang night-def that eyes difficult Nyoman Santosa AV.close That night, it was difficult for Nyoman Santosa to close (his) eyes. An anonymous reviewer has raised the possibility that our purported raising predicates are really adverbs. While we cannot absolutely eliminate this as a possible analysis, there are two reasons for believing that they are predicates and not adverbs. First, they can appear in only three positions: sentence-initial (13)a, sentence-final (13)b, and immediately following the subject (13)c. Other positions are impossible, as shown in (13)d.

12 10 (13) a. ngenah sajan [ia ngalihang I Nyoman gae di Jakarta] seem much [3 AV.find.APPL I Nyoman job in Jakarta] b. [ia ngalihang I Nyoman gae di Jakarta] ngenah sajan. [3 AV.find.APPL I Nyoman job in Jakarta] seem much c. ia ngenah sajan ngalihang I Nyoman gae di Jakarta. 3 seem much AV.find.APPL I Nyoman job in Jakarta d. ia ngalihang (*ngenah sajan) I Nyoman (*ngenah sajan) gae 3 AV.find.APPL I Nyoman job (*ngenah sajan) di Jakarta. in Jakarta It is very apparent that (s)he found Nyoman a job in Jakarta. This distribution suggests a predicate, which can take a clausal complement (13)a, clausal subject (13)b, or appear in the raising configuration (13)c. This parallels raising predicates English: (14) a. It is likely [that she will find Nyoman a job in Jakarta]. b. [That she will find Nyoman a job in Jakarta] is likely. c. She is likely to find Nyoman a job in Jakarta. The raising predicate keweh difficult has the same distribution:

13 11 (15) a. Keweh [ia ngalihang I Nyoman gae di Jakarta] difficult 3 AV.find-APPL I Nyoman job in Jakarta b. [Ia ngalihang I Nyoman gae di Jakarta] keweh. c. Ia keweh ngalihang I Nyoman gae di Jakarta. d. Ia ngalihang (*keweh) I Nyoman (*keweh) gae (*keweh) di Jakarta It is difficult for him/her to find a job for I Nyoman in Jakarta. In contrast, the placement of uncontroversial adverbial expressions like pidan ago is considerably freer: (16) a. (Pidan) ia ngalih-ang Nyoman gae di Jakarta.. ago 3 AV.find-APPL Nyoman job in Jakarta b. Ia (pidan) ngalihang Nyoman gae di Jakarta. c. Ia ngalihang (pidan) Nyoman gae di Jakarta. d. Ia ngalihang Nyoman (pidan) gae di Jakarta. e. Ia ngalihang Nyoman gae (pidan) di Jakarta. f. Ia ngalihang Nyoman gae di Jakarta (pidan). (S)he found a job for Nyoman in Jakarta a while ago. Similarly, the manner adverb sanget-sanget in a very hard manner / with great effort (lit. very-very ) can intervene between a verb and its object.

14 12 (17) a. (Sanget-sanget) ia nepak gamelan-e. very-very 3 AV.hit gamelan-def b. Ia (sanget-sanget) nepak gamelane. c. Ia nepak (sanget-sanget) gamelane. d. Ia nepak gamelane (sanget-sanget). (S)he was hitting the gamelan really hard. So our purported raising predicates differ from adverbs in distribution. Moreover, the raising-type distribution correlates with the possibility of verbal inflection. For example, (18) shows that the raising predicate enggal quick accepts the causative suffix -ang (on causatives, see Section 5.1) and participates in the OV/AV alternation. Notice how the voice marking on the matrix predicate (ng)enggal-ang cause to be quick determines the position of the subject of the subordinate predicate nyemak take (namely tiang, the first person pronoun): (18) a. Enggal-ang tiang nyemak pipis-ne. OV.quick-CAUS 1 AV.take money-3poss I quickly took his/her money. (Lit. I made it quick to take his/her money. ) b. Tiang ngenggal-ang nyemak pipis-ne 1 AV.quick-CAUS AV.take money-3poss I quickly took his/her money. (Lit. I made it quick to take his/her money. ) The NP tiang clearly bears a grammatical relation to enggal since that NP s position is regulated by voice marking. More specifically, the observed

15 13 alternation parallels what we have seen for transitive predicates, where the agent NP follows the OV-verb but precedes the AV-verb (details of this alternation must wait until Section 4 below). In contrast to enggal, true adverbs cannot be inflected in this way: (19) pidan ago -> *pidan-ang ago-caus ibi yesterday -> *ibi-ang /*ngibiang yesterday-caus (OV/AV) mani tomorrow -> * maniang/*ngemaniang tomorrow-caus (OV/AV) sajan very (much) -> * sajanang/*nyajanang very-caus (OV/AV) The following sentences illustrate the contrast between raising predicates and adverbials: (20) a. Ia enggal kema 3 quick go.there (S)he went there quickly. b. Ia ngengal-ang kema 3 AV.quick-CAUS go.there (S)he made it quick to go there. (21) a. Ia ibi kema. 3 yesterday go.there (S)he went there yesterday. b. *Ia ngibi-ang kema. 3 AV.yesterday-CAUS go.there (S)he made yesterday be the day to go there.

16 14 In short, the correlation between distribution and inflection suggests that certain Balinese elements such as enggal quick, ngenah seem, and keweh difficult are true raising predicates. To return to the main point of this section, raising predicates pick out the preverbal NP for raising, whether it is the agent (of an AV verb) or theme (of an OV verb). This lends support to our contention that this argument is the subject of its clause Subject-to-object raising. In the Balinese subject-to-object raising (SOR) construction, the subject of an embedded predicate is syntactically dependent upon the superordinate predicate. Consider this example (from Artawa 1994, p. 148): (22)a. Nyoman Santosa tawang tiang mulih. (name) OV.know 1 go.home I knew that Nyoman Santosa went home. b. Tiang nawang Nyoman Santosa mulih. 1 AV-know (name) go.home I knew that Nyoman Santosa went home. The SOR verb tawang know takes the embedded subject Nyoman Santosa as a syntactic dependent, hence subject to the same alternations as semantic arguments. Specifically, this raised NP appears in the subject position of an OV SOR verb (as in (22)a) or object position of an AV SOR verb (as in (22)b). (Note

17 15 that in this case there is no possibility of an adverb analysis.) When the downstairs predicate is transitive then only our purported downstairs subject can raise into the higher clause. (23) illustrates this fact for all four combinations of AV and OV on the matrix and embedded predicates (all four sentences have the same logical relations, indicated by the translation below): (23)a. Ia nawang Wayan lakar tangkep polisi. 3 AV.know Wayan FUT OV.arrest police b. Wayan tawang=a lakar tangkep polisi. Wayan OV.know=3 FUT OV.arrest police c. Ia nawang polisi lakar nangkep Wayan. 3 AV.know police FUT AV.arrest Wayan d. Polisi tawang=a lakar nangkep Wayan. police OV.know=3 FUT AV.arrest Wayan He knew that the police would arrest Wayan. In (24) we have attempted to raise the embedded object instead of the subject, again taking all four combinations. This systematically fails:

18 16 (24)a. *Ia nawang polisi Wayan lakar tangkep. 3 AV.know police Wayan FUT OV.arrest b. *polisi tawang=a Wayan lakar tangkep. police OV.know=3 Wayan FUT OV.arrest c. *Ia nawang Wayan polisi lakar nangkep. 3 AV.know Wayan police FUT AV.arrest d. *Wayan tawang=a polisi lakar nangkep. Wayan OV.know=3 police FUT AV.arrest He knew that the police would arrest Wayan. In short, only our purported subject can be raised in the SOR construction Control. Control is another classic subject test: in many languages only a subject can be controlled (Zaenen, Maling, and Thráinsson 1985). 7 Once again, only the pre-verbal argument, whether the Theme of an OV verb or the Agent of an AV verb, can be a controllee: (25)a. Tiang edot [ teka]. 1 want come I want to come.

19 17 b. Tiang edot [ meriksa dokter]. 1 want AV.examine doctor I want to examine a doctor. c. Tiang edot [ periksa dokter]. 1 want OV.examine doctor I want to be examined by a doctor. In contrast with the examples in (25), the agent in the following sentence is not the downstairs subject and cannot be therefore controlled: (26)a. *Tiang edot [dokter periksa ]. 1 want doctor OV.examine (I want to examine a doctor.) b.?*tiang edot [dokter meriksa ]. 1 want doctor AV.examine (I want to be examined by a doctor.) Turning to majanji promise, in this type of commitment relation, the promiser must have semantic control over the action promised (Farkas 1988, Kroeger 1993, Sag and Pollard 1991). The promiser should therefore be the actor of the downstairs verb. This semantic constraint interacts with the syntactic constraint that the controllee must be the subject to predict that the controlled VP must be in AV voice, which places the Agent in subject role. This prediction is borne out:

20 18 (27) a. Tiang majanji maang Nyoman pipis. 1 promise AV.give Nyoman money b. *Tiang majanji Nyoman baang pipis. 1 promise Nyoman OV.give money c. *Tiang majanji pipis baang Nyoman. 1 promise money OV.give Nyoman I promised to give Nyoman money. The same facts obtain for other control verbs such as paksa force Extraposition. A last distinguishing characteristic of subjects is the possibility of extraposition. In marked discourse contexts such as focal contrast, the subject can be extraposed: (28)a. Nasi-ne amah cicing. (SVO) rice-def OV.eat dog A dog ate the rice. b. Amah cicing nasi-ne. (VOS) eat dog rice-def A dog ate the rice. What happened to the rice is that it was eaten by a dog (not some other fate).

21 19 (29)a. Tiang ngae umah. (SVO) (h.r.) I AV.build house I built a house. b. ngae umah tiang (VOS) AV.build house I I built a house. What I did was build a house (not some other activity). Extraposition blindly picks whichever NP is our purported subject, irrespective of voice-marking or thematic role. In contrast, objects cannot be easily extraposed. For example, the benefactive object cannot be easily extraposed (30)b, whereas the agent subject can (30)c. (30)a. Tiang nge-maang Wayan nasi. (S V O rec O th ) 1 AV.give Wayan rice I gave Wayan rice. b.??tiang nge-maang nasi Wayan. (*S V O th O rec ) 1 AV.give rice Wayan c. Nge-maang Wayan nasi tiang. (V O rec O th S) AV.give Wayan rice 1 I gave Wayan rice. What I did was give Wayan rice. Turning to the OV form in (31), the agent tiang I is not the subject; in neutral order it appears post verbally (31)a. Crucially, the agent cannot be extraposed, as

22 20 shown in (31)b. But the benefactive Wayan is the subject, so it can be extraposed (31)c: (31)a. Wayan baang tiang nasi (S V O agt O th ) Wayan OV.give 1 rice I gave Wayan rice. b. * Wayan baang nasi tiang (*S V O th O agt ) Wayan OV.give rice 1 I gave Wayan rice. c. Baang tiang nasi Wayan (V O agt O th S) OV.give 1 rice Wayan I gave Wayan rice. What I did was give Wayan rice. In short, all and only subjects can be extraposed in Balinese. To summarize Section 1: raising to subject, raising to object, control, relativization, and extraposition all support the claim that the NP appearing to the left of the verb is the subject. This is true irrespective of the verb s voice marking. In other words, the voice marking regulates subject selection. 2. Complement nominals The complement Agent of an OV verb. While the Agent of an AV verb must appear as the subject, the subject of a mono- or ditransitive OV verb is any term (NP) argument except the Agent, as shown in (33). The term TERM is used in this paper to refer to a direct argument,

23 21 as distinct from oblique arguments. In Balinese all terms are NP s while obliques are PP s. The Agent of the OV verb instead appears in one of three forms. First, it can take the form of a third person pronominal clitic =a (or =na, if the host is vowel-final) on the verb (low register only): (32) Buku-ne jemak=a book-def OV.take=3 (S)he took the book. (33)a. Bungkusan dadua jemak-ang=a Nyoman Santosa package two OV.take-APPL=3 (name) (s)he brought two packages for Nyoman Santosa. b. Nyoman Santosa jemak-ang=a bungkusan dadua (name) OV.take-APPL=3 package two (s)he brought two packages for Nyoman Santosa Alternatively, the Agent can be realized as an indefinite NP (34). (34)a. I Wayan gugut cicing /*cicinge ento. Art Wayan OV.bite dog /dog that A/*the dog bit Wayan. b. ia alih [dagang celeng uli Badung] 3 OV.look.for trader pig from Badung A pig trader from Badung looked for him/her. Finally, the Agent can be expressed as a post-v free pronoun (see (2) above).

24 Discourse topics, the function of AV/OV, and the definiteness restriction. A full treatment of the definiteness restriction illustrated in (34) must await further research, but we will make some general remarks here. As in many of its Austronesian relatives, Balinese uses the subject (=pivot) position for discourse topics (indeed, many Austronesianists prefer the term topic to subject ). One consequence is that these languages typically have important conditions on the distribution of definites; for example, Tagalog definite patient arguments must appear in nominative case, i.e. must be subjects (= pivots ) (Foley and Van Valin 1984, p ; Kroeger 1993). The discourse correlates of subjecthood also allow us to explain the primary function of the OV/AV alternation: namely, to provide a means for designating the appropriate argument as discourse topic. In the following passage from a Balinese novel, the topic of the discourse is the narrator. Since the narrator is the agent of the seeing action in the first sentence, that verb appears in AV form nepukin, AV.see, so that the agent is the subject (35)a. The narrator is still the discourse topic in the second sentence but is not the agent of the waiting event, so the OV form, ati-ati-na OV.wait-wait-3, must be used (35)b. (35) a. Tiang lakar mulih, bas makelo ten nepukin i meme. 1 will go.home too long NEG AV.see Art mother I will go home, it s too long (for me) not to see (my) mother.

25 23 b. Mirib-ne ati-ati-na, apang ten kadenanga ngambul. apparently-def OV.wait-wait-3 so NEG OV.think=3 run.away so that she doesn t think that I m running away; it seems she is waiting for me. This example illustrates how a speaker often must switch back and forth between OV and AV to preserve topic continuity. (The third person singular pronominal clitic anaphorically refers to mother, and the first person singular pronoun has been pro-dropped in both the see and the wait clauses. The other verb in (35)b, kadenanga OV.think-3, is a subject-to-object raising verb.) Returning to the definiteness restriction on post-v agent NPs, definiteness (or perhaps specificity) and topichood are closely related, so one might expect a tendency for definite NP s to be subjects rather than complements. Given the cross-linguistic tendency for agents to be discourse topics (Dixon 1994), the pull towards topichood is even stronger for a definite agent than a definite non-agent. In other words, the two factors of definiteness and agentivity may conspire to force a definite agent into the subject function, thus prohibiting definite agent complements. However, this cannot be the whole answer. First, non-subject definite agents are possible in passive by-phrases (see Section 5.3). Second, pronouns, which are definite, are not subject to the definiteness restriction (e.g. (2) above). The mystery of the definiteness restriction must be left for future research The postverbal NP arguments as terms. Following Bresnan and Kanerva (1989), Alsina and Mchombo (1993), and others, we assume that alternations such as passive, antipassive, causative, and

26 24 applicative result from morpholexical operations on argument structure which add or suppress roles. But we argue that the ergative/accusative alternation results from applying two different subject selection rules to the same argument structure. These selection rules are encoded in the OV/AV distinction. In this section we motivate this approach by showing how Balinese voice marking differs from typical argument-structure changing rules such as passive and antipassive. Could the OV construction be a passive? This view is untenable, as the OV has none of the properties typical of passives. First, there is no passive verb morphology: OV is morphologically unmarked. Instead, it is the morphologically marked voice (AV) that resembles actives in accusative languages. Second, the post-verbal agent in OV is not the oblique by-phrase characterizing passives, but rather is a term. Prima facie evidence for this claim is that in Balinese, which lacks morphological case, all obliques are expressed as PPs, while terms are NPs. The post-verbal agent is always an NP (or clitic pronoun). Also, passive agent phrases are optional (and often analyzed as adjuncts; Grimshaw 1990), but the OV agent is crucially obligatory. 9 Third, the OV agent launches floating quantifiers (FQ), a property restricted to terms in Balinese (and perhaps in all languages; see Schachter 1977 p. 286ff). This restriction on FQ is illustrated here with intransitive matakon and transitive nakonin, both meaning ask (the latter is an applied form; see Section 5.2 below):

27 25 (36)a. Ia matakon [teken tiang] ibi ajak makejang. 3 ask to 1 yesterday accompany all (i) They all asked me yesterday. (ii)?*(s)he asked us all yesterday. b. Ia nakonin tiang ibi ajak makejang. 3 AV-ask 1 yesterday accompany all (i) They all asked me yesterday (ii) (s)he asked us all yesterday The quantifier ajak makejang can only be construed with terms, yielding only one interpretation for (36)a, but allowing two for (36)b. 10 As shown by the glosses, this quantifier yields a plural interpretation when it quantifies over a singular pronoun. Turning to OV verbs, we find that the complement agent can launch quantifiers: (37)a. Pipise jemak tiang ibi ajak makejang. money OV.take 1 yesterday accompany all (i) We all took the money. (ii) *I took all the money. b. Ia dengokin tiang ibi ajak makejang 3 OV.visit 1 yesterday accompany all (i) We all visited him/her yesterday (ii) I visited them all yesterday Reading (ii) is unavailable in (37)a because ajak makejang only applies to animates. Modulo this restriction, this quantifier can be launched by any term.

28 26 In addition, the post-verbal agent is tightly bound to the verb, like an object. An extraposed subject (see (28)) cannot intervene (from Artawa 1994, p. 30): (38)a. Siap-e uber cicing ke jalan-e. S V O agent PP chicken-def OV.chase dog to street-def A dog chased the chicken into the street. b. uber cicing ke jalan-e siap-e. V O agent S PP c. *uber siap-e cicing ke jalan-e. *V S O agent PP d. *siap-e uber ke jalan-e cicing. *S V PP O agent And it is generally difficult for other complements to intervene, as shown in (38)d. In sum, the post-verbal agent has all the properties of a direct complement of the verb, and no characteristics of obliques or adjuncts. Balinese has a passive construction which, in contrast with OV, has the properties noted above as typical of passives. It is morphologically marked and the agent appears as an optional PP by-phrase. Binding facts further illustrate a crucial difference between passive and the OV: while the OV agent can bind the subject (see (40)-(41) below), a passive by-phrase cannot. (See Section 5.3.) This difference follows from the theory to be presented in Section 5. On that theory, passive alters argument structure while the OV/AV voice distinction does not affect argument structure, but only the mapping between argument structure and grammatical relations.

29 27 Lastly, we consider the possibility that the AV construction could be analyzed as an antipassive. Antipassivization demotes the theme argument, which has the effect in a syntactically ergative system of making the agent a subject (or an absolutive, in an ergative case system). But we must reject the antipassive analysis, as there is no evidence that the theme in a Balinese AV clause is oblique; the arguments given above can be replicated for the theme complement: for example, it is an NP, not a PP; and it too can launch quantifiers: (39) Wayan nakonin tiang ibi ajak makejang Wayan AV-ask 1 yesterday accompany all Wayan asked us all yesterday. In short, the effect of the AV morphology is not to demote the theme, but rather to simply make the agent (actually the A-subject) the subject. This is captured by the subject-selection account proposed in Section 4.2 below. 3. Sketch of an analysis Binding in brief. Binding relations in Balinese depend, roughly speaking (but see below), on the thematic status of the arguments and not on surface configurational relations such as c-command (as assumed in standard GB configurational binding theory) or grammatical relations (as assumed in standard LFG binding theory). 11 This situation is familiar from many Austronesian languages (Andrews 1985, Kroeger 1993, Schachter 1984, inter alia). An Agent asymmetrically binds

30 28 a Theme, regardless of how these roles map onto the grammatical relations. Hence with an OV verb, the complement Agent can bind into a subject: (40) Ragan idane cingakin ida. (h.r.) self OV.see 3 (S)he saw himself/herself. (41) Awakne tingalin=a. self OV.see=3 (S)he saw herself/himself. But the opposite binding relation, where the Theme subject binds an Agent object, is impossible. (42) *Wayan tingalin awakne Wayan OV.see self (lit. Himself saw Wayan.) When we turn to AV, the subject Agent binds a complement Theme: (43) Ia ningalin awakne. 3 AV.see self (S)he saw herself/himself. This suggests that binding is defined at argument structure rather than grammatical relations or c-command at s-structure (Andrews 1985, Schachter 1984, inter alia). We concur with this view, although our version of argument structure differs slightly from some others (see the following section).

31 29 We should also note in passing that Balinese binding observes a linear precedence condition. Balinese allows backward binding of reflexives; see, e.g., (40) above, where the anaphor precedes its binder. This is not always possible however. A full NP does not allow backward binding (44)a. When the antecedent precedes the reflexive as in (44)b, binding is successful. (44) a. *Awakne gugut cicing self OV-bite dog A dog bit itself b. Cicing ngugut awakne dog AV-bite self A dog bit itself. Backwards binding by quantified noun phrases (QNP s) is also strongly resisted. Consider the pair of OV sentences in (45). Both sentences share the same argument structure where the possessive NP is a-commanded by the QNP which binds it. Normally in a ditransitive either the goal or theme can appear as subject (see (59) below). However, when binding applies then the binder must precede the bindee (45)a, otherwise the sentence is bad (45)b.

32 30 (45) a. Sabilang anak cenik i jemak-ang tiang bungkusan-ne i every person small OV-take-APPL 1 package-3poss For every child I brought his/her package. b.?*bungkusan-ne i jemak-ang tiang sabilang anak cenik i. package-3poss OV-take-APPL 1 every person small (For every child I brought his/her package.) As a final example, consider the double oblique constructions with the verb matakon ask in (46): the person asked (i.e goal role) is marked by sig to/at and the thing asked (i.e theme) is marked by unduk about. The PP s can appear in either order: (46) a. Tiang matakon sig Nyoman-e unduk I Wayan-e. 1 ask to Nyoman-DEF about I Wayan-DEF I asked (to) Nyoman about I Wayan. b. Tiang matakon unduk I Wayan-e sig Nyoman-e 1 ask about I Wayan-DEF to Nyoman-DEF I asked (to) Nyoman about I Wyan. (Lit. I asked about I Wayan to Nyoman. ) The sig-pp (goal) can bind the unduk-pp (theme) only when the binder precedes the bindee, but not in the reverse order:

33 31 (47) a. Tiang matakon sig sabilang anak-e i ane teka unduk umah-ne i 1 ask to every person-def REL come about house-3poss I asked (to) every person i who came about his/her i house. b. *Tiang matakon unduk umah-ne i sig sabilang anak-e i ane teka. 1 ask about house-3poss to every person-def REL come * I asked about his/her i house (to) every person i who came. (47)b is bad because the pronominal precedes the operator that binds it. To conclude, Balinese binding obeys the following linear order condition: (48) A pronominal (anaphor or ordinary pronoun) cannot precede a nonpronominal that binds it. Note that this correctly allows cases of backwards binding by pronominals, as in (40). Identical or similar conditions have been proposed for other languages (cp. Langacker 1969, Barss and Lasnik 1986, Bresnan 1995b, inter alia). This linear order condition will not be discussed further in this paper (see Arka and Wechsler 1996 for more on this issue); instead we will focus on ARG-S based binding conditions HPSG binding theory. We represent argument structure with the lexical feature ARG-S whose value is a list (ordered set). This feature for the Balinese verb for see (both OV and AV forms) is roughly as follows (see (51) for more detail):

34 32 (49) [ARG-S < seer, seen >] As Manning (1994) points out, the behavior of binding in syntactically ergative languages constitutes prima facie evidence for the binding theory in Head-Driven Phrase Structure Grammar (HPSG; Pollard and Sag 1992, 1994). Local binding is universally defined on ARG-S rather than phrase structure or grammatical relation. In the following statements of HPSG binding theory, A-COMMAND refers to the ordering relation on ARG-S; e.g. in (49) the seer argument a-commands the seen. 12 (50) HPSG Binding Theory 13 Principle A. An a-commanded anaphor must be locally a-bound. Principle B. A personal pronoun must be locally a-free. Principle C. A non-pronoun must be a-free. Definitions: In the feature [ARG-S < x, y, >], x is said to a-command y. x locally a-binds y = x a-commands y, x and y coindexed x a-binds y = x a-commands some z dominating y, x and y coindexed x is (locally) a-free = x is not (locally) a-bound Ergative and accusative clauses are alike with respect to binding conditions, but differ in the mapping between ARG-S and the VALENCE lists SUBJ (subject) and COMPS (complements), which specify how the arguments are syntactically realized (see below). For example, (51) gives the approximate HPSG lexical signs for ningalin see (AV) and tingalin see (OV) as in (40) and (41) respectively (from Manning and Sag 1995).

35 33 (51) a. ningalin ( AV.see ): b. tingalin ( OV.see ): PHON < ningalin > SUBJ < [1] > VALENCE COMPS < [2] > ARG S < [1]NP i,[2]np j > seeing - relation CONTENT SEER i SEEN j PHON < tingalin > SUBJ < [2] > VALENCE COMPS < [1] > ARG S < [1]NP i,[2]np j > seeing - relation CONTENT SEER i SEEN j The AV and OV verb forms represented in (51)a and b are built on the same stem, specifying phonology, semantic content, and argument structure information. The two verb forms differ only in their valence features and inflectional morphology (see Section 4.2). An HPSG lexical sign is a 2xn attribute-value matrix, that is, a set of features (attribute-value pairs). The value of each feature is either an embedded attribute-value matrix (such as the values of VALENCE and CONTENT), an atom, or a list of elements (such as the values of SUBJ, COMPS, ARG-S, and PHON). 14 The boxed numerals [1] and [2] are tags indicating structure-sharing (reentrancies) between elements in this matrix. For example, in (51)a the tag [1] beside NP i indicates that the element NP i appears in two places at once: as sole element in the list value of the feature SUBJ and as first element in the list value of the feature ARG-S. The lower-case letters i and j are tags for indices, which correspond roughly to the notion of referential index. Referring to (51)a,b, the referential indices associated with the first and second items in the ARG-S list are the SEER and the SEEN arguments, respectively. Note that in (51)a (the accusative verb) the NP i and NP j realizing the SEER and SEEN roles are subject and complement respectively; while in (51)b this mapping is reversed. It is also

36 34 worth noting that (51)b does not represent a passive verb, since NP i is a term argument and not an oblique or adjunct. Items in the list values for SUBJ and COMPS are effectively cancelled from these lists as phrases of the appropriate type are concatenated with the head (cp. categorial grammar), as in this simple Subject-Verb-Complements ditransitive clause: V SUBJ < > COMPS < > (=S) (52) 1 NP V SUBJ < 1 NP> COMPS < > (=VP) V SUBJ < 1 NP> COMPS < 2 NP, 3 NP> 2 NP 3 NP A small system of immediate dominance schemata and linear precedence constraints licenses all phrases. We will not spell them out here (see Pollard and Sag 1994 for one proposal), but note only that the clause structure we posit for Balinese, as for English, is essentially that depicted in (52). Also, note that in Balinese and English ditransitives, the two complement NPs are linearly ordered according to their COMPS list ordering (see Pollard and Sag 1987, p. 174). So far, our main theoretical claim is the following: at the lexical level, a predicate is associated with a hierarchically arranged set of arguments (its ARG-S value). One argument is designated as subject and the rest as a list of complements. We further claim that binding (of reflexive anaphors and pronoun

37 35 binding by quantified NPs) universally takes place at argument structure. We believe this model, which posits nothing beyond the essential design of HPSG, is sufficient to describe ergative, accusative and mixed systems as in Austronesian. Section 4 spells out a general theory governing the mapping between argument structure and valence lists, universally and for Balinese. 4. Linking and the theory of argument structure. The problem of argument realization has two parts: the determinants of argument structure itself (Section 4.1); and the mapping from argument structure to valence features (Section 4.2) Theory of Argument Structure The ordering on argument structure elements closely reflects the predicate s meaning, in two different ways (cp. Manning 1994): (53) Theory of ARG-S ordering. a. terms outrank obliques. b. arguments within each group are ordered by the lexico-semantics of the verb. For the purpose of Balinese, terms are NPs and obliques are PPs (in a casemarking language, NPs in oblique case might be classified as obliques). Regarding (53)b, one might assume a thematic hierarchy like ag < go < th < loc, or adopt an alternative such as Wechsler 1995 or Davis The particular

38 36 system is not crucial and will not be discussed further here; we assume that it is universal, although this is not crucial to our argument. Ranking by the term/oblique distinction as in (53)a reflects the lexicosemantic grounding of the notion oblique, in that obliques are marked by semantically contentful morphology (adposition or case) indicating semantic role type. Governed adpositions and oblique case markers are not merely formal devices but have semantic content which is systematically related to that of the governing verb (Gawron 1986, Wechsler 1994, 1995). Since the morphological resources for indicating oblique relations vary across languages (though governed by tendencies), we do not assume that a given verb (concept) will have the same ARG-S in every language: e.g., benefactive arguments might tend to be obliques in one language, terms in another. Even within a single language, term/oblique alternations are common (dative, locative, and factitive alternations, e.g.) and subject to lexical idiosyncrasy. Thus the ARG-S for a given verb cannot be blindly derived from the verb s meaning alone, but instead derives from knowledge of the lexicon of the language in question, including the morphological resources available for marking obliques, maximum number of terms allowed, and so on. Turning now to syntactic justification, binding provides evidence since it occurs at ARG-S. Take term arguments first. The claim that term arguments are ranked for the purpose of binding relations is well-accepted, and considerable further support from Balinese is given in Section 5 below. In accusative languages argument structure ranking tends to correlate with grammatical relation hierarchies (subj < 1st-obj < 2nd-obj), linear order (at least in subject-initial

39 37 languages, and prior to wh-movement), and c-command. This makes it more difficult to show that argument structure is relevant to binding. But Hellan (1988, p. 162), citing Elisabet Engdahl, gives an ingenious demonstration from Norwegian, an accusative SVO language. Norwegian ditransitives are symmetrical in the sense that either object passivizes (Åfarli 1989), so the equivalent of John was shown Mary is ambiguous ( John was shown to Mary or Mary was shown to John ). Hellan (1988, p. 162) observes that such constructions become unambiguous if we impose a binding relation, as in the following: (54) Barnet i ble fratatt sine i foreldre. (Norwegian) child was from-taken self s parents The child i was deprived of self s i parents. NOT: *The child i was taken from self s i parents. The source (malefactive) argument can bind into the theme, but not vice versa. For us this simply means that the source NP outranks the theme at ARG-S: (55) [ARG-S < NP(ag), NP(source), NP(th)] It is harder to explain in terms of c-command (at s-structure or later) or even ranking of grammatical relations (as in Bresnan 1995b, inter alia), since normally either source or theme can appear as the passive subject.

40 38 Turning to the relative ranking of obliques, we can cite well-known contrasts like the English translations of the following examples (Postal 1971, p. 36); the same facts obtain in Balinese: (56)a Tiang ngomong teken Made i unduk awakne i 1 AV.talkto Made about self I talked to Made i about himself i. b. *Tiang ngomong unduk Made i teken awakne i. 1 AV.talk about Made to self *I talked about Made i to himself i. c. [ARG-S < NP[ag], PP[go] i, PP[th]:ana i >] As expected, the thematically higher goal can bind the lower theme, but not vice versa. Note that (56)b is not ruled out because of the complement order. PP s of this kind can normally occur in either order, in Balinese as in English. The over-arching distinction between terms and obliques (53)a has been supported on the basis of anaphoric binding facts by Hellan (1988, p. 164), Dalrymple (1993, p. 172ff), and Manning (1994). Alternations like the following illustrate the impossibility of a purely thematic account which fails to make this distinction: 15 (57)a. Mary showed the twins i each other i. a'. Mary showed the twins i to each other i.

41 39 b. I sold the slave i himself i b'. I sold the slave i to himself i c. [ARG-S <NP[ag], NP[go] i, NP[th]:ana i >] c'. [ARG-S <NP[ag], NP[th] i, PP[go]:ana i >] In (57)a,b,c the Goal a-commands the Theme while in (57)a,b,c the Theme, being a term, a-commands the oblique Goal. Further evidence for our model of argument structure comes from other processes defined on it: relation-changing operations like applicative, passive, and causative (Section 5 below), and linking rules (Section 4.2) Subject selection ( linking ). Recall that the A-subject of a Balinese AV verb must appear as the (grammatical) subject: (58) Ia meli-ang I Wayan potlot-e ento. 3 AV-buy-APPL ART Wayan pencil-def that (s)he bought Wayan the pencil. Also, for any intransitive verb, whatever its voice marking, the sole term is the subject. On the other hand, the subject of a mono- or ditransitive OV verb is any term (NP) argument except the A-subject. Thus a ditransitive OV verb has two linking possibilities:

42 40 (59)a. Potlote ento beli-ang=a I Wayan. pencil-def that OV.buy-APPL=3 Art Wayan (s)he bought Wayan the pencil. b. I Wayan beli-ang=a potlote ento. Art Wayan OV.buy-APPL=3 pencil-def that (s)he bought Wayan the pencil. c. *Ia beli-ang I Wayan potlot-e ento. 3 OV.buy-APPL Art Wayan pencil-def that (s)he bought Wayan the pencil. The following universal subject and complement linking rules account for this: (60) Subject and complement linking (universal). I. Subjects. a. intransitives: Link subject to the single ARG-S term. b. transitives: Link subject to a term which is: i. the A-subject, if verb is of sort acc-verb; or ii. not the A-subject, if verb is of sort erg-verb. II. Complements. Link any remaining ARG-S items to the COMPS items, preserving order. We sort verbs into two universal types acc-verb and erg-verb, according to which linking rule it follows. These rules are formalized as follows. For any verb (modulo pro-drop, etc.) the ARG-S value is the sequence union (u) of the SUBJ and COMPS values. 16 The ARG-S of an active (OV or AV but not passive) verb matches that of its stem:

43 41 (61) verb: active-verb: SUBJ [1] VALENCE COMPS [2] ARG S [1] [2] STEM v stem ARG S [1] STEM v stem ARG S[1] Active verbs are sorted into accusative and ergative (62)a. For an accusative verb, the A-subject is the SUBJect (62)b. For an ergative verb transitivity is crucial, so verb stems are sorted into intransitive (i.e., exactly one term NP) and (mono- or di-) transitive (i.e., more than one term NP). This is shown in (63). For ergatives, if intransitive, then the single argument is the subject; if transitive, then any NP except the A-subject maps to SUBJ ((62)c): (62) a. active-verbo acc-verb erg-verb SUBJ < [1] > b. acc-verb: c. erg-verb: ARG S < [1], > SUBJ < [1] > SUBJ < [1] > ARG S < [1], > ARG S < [], [1], > STEM intrans stem STEM trans stem (63) a. v-stemo intrans-stem trans-stem b. intrans-stem: [ARG-S <NP[term]> & list(non-term)] (exactly 1 NP) c. trans-stem: [ARG-S <NP[term], NP[term], >] (more than 1 NP) Most languages have only one type: English uses the acc-verb type, Dyirbal the erg-verb type. Many Austronesian languages allow both, including Toba Batak (Schachter 1984), Karo Batak (Woollams 1996), Tagalog (Kroeger 1993, inter alia), and of course Balinese.

44 42 The dots ( ) in (62)b and c are to be interpreted as zero or more arguments of any kind. Focussing for now on acc-verb and the left option for erg-verb, note that when there are no terms among those unspecified arguments, we have an intransitive (since the only term is [1]). Hence intransitives can be of either the acc-verb or erg-verb type. Indeed intransitives are split: some bear the AV homorganic nasal prefix (or sometimes the prefix ma-), while others do not. Artawa (1994, 92ff) notes that the nasal prefix intransitives tend to be agentive while the others tend not to be. In that sense Balinese resembles what Dixon (1979) calls Split S-marking languages, which are also called active languages. Using Dixon s three terms A (agent of transitive), O (theme of transitive), and S (subject of intransitive), we may say that these languages have two classes of intransitive verbs: for some intransitives, the subject patterns with A, for others with O. Among Austronesian languages, Acehnese works this way, according to Durie (1985). Artawa (1994, 92ff) and Clynes (1995) both suggest a Split-S analysis of Balinese. On this view the nasal prefix would be encoding the semantic agentivity (volitionality, perhaps) of the subject, rather than encoding the fact that the subject is the A-subject, as our analysis claims. On the Split-S view we could simplify our analysis by eliminating the disjunction between transitive and intransitive linking sorts in (62)c. Informally, the new rules would be: (64) a. AV: subject is the Agent. b. OV: subject is not the Agent.

45 43 However, the correlation for intransitives between agentivity and the nasal prefix is only a tendency. For example, bangun get up is a bare (unprefixed) intransitive, yet it is semantically agentive: it can be used in imperatives (65)a or modified with volitional adverbials like deliberate (65)b. Most strikingly, semantically identical verbs may be unprefixed in one register, and prefixed in the other. The corresponding high register verb is ma-tangi (ma- is a variant of the nasal prefix for intransitives): (65) a. (cai) bangun! (low register) (you) get up b. Ia ngelap bangun (low register) 3 deliberate get up (S)he got up on purpose. c. (Tu) ma-tangi! (high register) 2 get up (you) get up. Also, not all nasalised intransitive verbs express actions. For example, from the precategorial root tenggel we can get nenggel, meaning in the edge of something/ in the position of a boundary of something : (66) Buku-ne nenggel di mejane book-def at.the.edge.of at table-def The book is at the edge of the table.

46 44 Clearly buku is a not an agent. Other non-agentive nasal intransitive verbs are ngliling roll, nyilem drown, and ngendih flare up, flame. In conclusion, while Balinese has some split-s language tendencies, there are some difficulties with this analysis. Still, it is an intriguing possibility, and we leave it for future research. Note that this issue affects only the rules responsible for linking. The output of those rules, that is, the linking patterns (and hence the processes dependent upon them which we investigate below) are identical under the two approaches. While the linking rules in (60) (=(62)) are assumed to be universal, we also assume that further parochial constraints hold for specific languages. One parameter of variation concerns subject selection in ditransitives where the A- subject cannot be the (grammatical) subject, e.g. because the clause is ergative or passive. As we just saw in (59), Balinese allows either argument of the ditransitive to appear as subject. While Balinese ditransitives are symmetrical, many languages are asymmetrical, allowing only one of the arguments to link to subject. Interestingly, this parameter of variation cuts across closely related languages, as Bresnan and Moshi (1990) showed in their study of comparative Bantu syntax. Investigating applied benefactive constructions, they show that languages from one group (Kichaga, Kinyarwanda, Kihaya, Kimeru, Mashi, Luyia,.) allow either the patient or applied benefactive argument to passivize while languages in another (Chichewa, Kiswahili, Bokamba, Chimwi:ni, Hibena.) allow only the benefactive to passivize. A similar split runs across closely related Austronesian languages. (We are grateful to an anonymous

47 45 reviewer for pointing out this split to us, and supplying the following data.) Like Balinese, Madurese is symmetrical, as shown in these OV sentences: (67) Madurese. a. Ali e-teguh-agi abakna poqlod-a. Ali OV-see-APP 3 pencil-def (s)he showed Ali the pencil. b. Poqlod-a e-teguh-agi abakna (daq) Ali. pencil-def OV-see-APP 3 (to) Ali (s)he showed Ali the pencil. But Bahasa Indonesia and Javanese are both asymmetrical: (68) Bahasa Indonesia. a. Ali di-kirim-i surat itu oleh Hasan. Ali OV-send-APP letter that by Hasan Hasan sent Ali the letter. b. *Surat itu di-kirim-i Ali oleh Hasan. letter that OV-send-APP Ali by Hasan (69) Javanese. a. Ali di-duduh-i dheke potlot-e. Ali OV-see-APP 3 pencil-def (s)he showed Ali the pencil. b. *potlot-e di-duduh-i Ali (karo dheke). pencil-def OV-see-APP Ali (by 3)

48 46 In Bahasa Indonesia and Javanese, only the goal role can be selected as subject (cp. also English passive ditransitives). We will not attempt to incorporate this parameter into our analysis. See Bresnan and Moshi (1990) for one approach. Turning to Balinese-specific grammar, the sort declarations in (70)c and d respectively specify that Balinese uses the universal sort erg-verb for OV and accverb for AV verbs: (70) Balinese verb forms. a. verbo active-verb pass-verb b. active-verb o AV-verb OV-verb c. OV-verb: erg verb PHON Φ STEM PHON Φ d. AV-verb: acc verb PHON n(φ) STEM PHON Φ where the phonology value n(φ) is obtained by replacing the initial consonant in Φ with a homorganic nasal, if Φ is consonant-initial; and n(φ)= +Φ if Φ is vowel-initial. Passive verbs (pass-verb) are discussed in Section 5.3 below. For example, the verb stem beliang buy for, whether serving as the stem for an AV verb (as in (58)) or OV verb (as in (59)), has the following ARG-S and semantic content:

49 47 (71) beliang buy for : ARG S < NP i,np j,np k > buy rel AGENT i CONTENT GOAL j THEME k The A-subject (NP i ) is the agent NP. By (70)d, the AV form meliang is an accverb, so from (60)I.b.i, its agent is the subject and the remaining items are complements (see (58)). By (70)c, the OV form beliang is an erg-verb, so by (60)I.b.ii its agent is not the subject, and instead the subject is either Goal or Theme (see (59)). 17 For completeness, we treat the third person pronominal OV agent clitic =a (Section 2). This agent can bind the subject, as we saw in (41), repeated here: (72) Awakne tingalin=a. self OV.see=3 (S)he saw herself/himself. In lexicalist theories like HPSG the syntax deals only in fully-formed words. A pronominal affix encodes an ARG-S list item of type ppro (personal pronoun) which fails to appear on a valence list. 18 For the agent-clitic verb, we need only specify that this pronoun is the first item in the ARG-S list:

50 48 (73) PHON Φ+a SUBJ [1] < NP > COMPS [2] ARG S [3][< NP: ppro [3rd] > [[1] [2]]] ov verb HOST PHON Φ ARG S[3] (The phonological value Φ+a is obtained by suffixing /-a/ to Φ if Φ is consonant final, or /-na/ if vowel-final. The symbol & represents list concatenation: e.g., <a,b> & <c,d> = <a,b,c,d>. The symbol u represents sequence union; see fn. 17 for explanation.) The agent (NP:ppro) appears in argument structure, so it participates in binding like any pronoun (see, e.g., (72)), but it does not map onto either valence list, so it cannot be expressed as a subject or complement. Instead, the remaining ARG-S items (indicated by [[1]u[2]]) map onto the SUBJ and COMPS lists. Arka (1997) shows that the /-a/ suffix actually leads a double life in Balinese: it can be a pronominal clitic, or it can be a passive morpheme. In its function as passive morpheme it marks the low register counterpart to the (primarily) high register ka- passive. Passives are discussed in Section 5.3 below Subject-to-Object raising verbs. The behavior of subject-to-object raising ( exceptional case-marking ) verbs in Balinese lends support to this analysis based on a syntacticized version of argument structure, as opposed to a theory of ergativity based on regulating the mapping from a pure semantic ( thematic, conceptual, etc.) structure to

51 49 syntax, as in the Ergativity Parameter of Marantz 1984 and Levin 1983, for example. 19 As we saw in Section 1.3 above, raised NPs, which receive no theta role from the matrix verb, participate in the OV/AV alternation like other complements. Consider example (22), repeated here: (74)a. Nyoman Santosa tawang tiang mulih. (name) OV.know 1 go.home I knew that Nyoman Santosa went home. b. Tiang nawang Nyoman Santosa mulih. 1 AV-know (name) go.home I knew that Nyoman Santosa went home. In a sense subject-to-object raising is a misnomer since the raised NP (here, Nyoman Santosa) appears as object if the matrix verb is in AV form, but subject if it is in OV form. Reversing the NP s changes the interpretation accordingly: (75)a. Nyoman nawang tiang mulih. (name) AV.know 1 go.home Nyoman knew that I went home. ( I knew that Nyoman went home.) b. Tiang tawang Nyoman mulih. 1 OV-know (name) go.home You (Nyoman) knew that I went home. ( I knew that you (Nyoman) went home.)

52 50 This alternation follows automatically from our account, assuming that the raised NP and the following open complement (here, the predicate mulih go home ) are both complements of the matrix verb the standard analysis in all lexicalist frameworks, which lack NP-movement (see Pollard and Sag 1994, ch. 3 for extensive argumentation); similarly, some GB treatments posit a ternary branching VP for such constructions (see Carrier and Randall 1992, p. 227; Williams 1983). The verb tawang know has the following ARG-S list and semantic content: (76) tawang: ARG S < NP i,[2]np, AP[SUBJ < [2] >]: j > know relation CONTENT KNOWER i PROPOSITION j The tag [2] indicates that the object NP acts as the predication subject for the AP. 20 In its OV voice a non-a-subject term must be the subject; the only element qualifying for subject is the second item in ARG-S, [2]NP. The result is (74)a. In AV voice the A-subject is the subject, yielding (74)b. Note that [2]NP is not a semantic argument of know, as shown in (76) by the fact that it is not linked to an argument within the semantic content value. Binding facts support this analysis. If coreferent with the matrix subject, the raised NP must be a reflexive (77)a,b rather than a pronominal (77)c,d:

53 51 (77) a. Cang ngaden awak cange suba mati. 1 AV.think myself already dead I believed myself to be dead already. b. Awak cange kaden cang suba mati. myself OV.think 1 already dead I believed myself to be dead already. c. *Cang ngaden cang suba mati. 1 AV.think 1 already dead (*I believed me to be dead already.) d. *Cang kaden cang suba mati. 1 OV.think 1 already dead (*I believed me to be dead already.) These facts are expected since binding is defined on ARG-S: the matrix agent a- commands the raised NP. Binding in SOR and SSR constructions is discussed further in Section 6 below. Constructions of this kind show that ergativity, at least in Balinese, cannot be reduced to a mapping between semantic structure and syntax. Instead a syntacticized version of argument structure appears to be appropriate. For us the raised NP is crucially a member of the raising verb s ARG-S list, so we expect it to participate in the OV/AV alternation like other complements. 5. Operations on argument structure. We have shown at several points that OV/AV voice morphology must be distinguished from morpholexical operations like applicative, causative, and

54 52 passive in that the former influences the mapping from argument structure to valence features, while the latter encode operations on argument structure. In this section we present evidence for this view Causatives. Causativisation adds a causer argument to the argument structure. Adding the causative morpheme -ang to the one-place predicate sakit ill (78)a, yields a transitive verb sakitang hurt. As with any transitive, subject selection depends on the voice marker. Thus, alongside (78)b we can also have the AV form in (78)c: (78)a. Adin cai-ne sakit. younger-sibling 2-DEF ill Your younger sibling is ill. b. Adin cai-ne sakit-ang cai. younger-sibling 2-DEF OV.ill-CAUS 2 You hurt your younger sibling. c. Cai nyakit-ang adin cai-ne. 2 AV-ill-CAUS younger-sibling 2-DEF We analyze Balinese causativisation as adding a Causer argument at the top of the ARG-S list.

55 53 (79) a. base verb: [ARG-S L] b. causative verb: [ARG-S <NPc>&L] (where L is a list and NPc is the causer argument) (As shown below, the causee is sometimes demoted, depending on the type of causative, causing further adjustments to the ARG-S list. We will not treat the effects of causativization in detail here.) As expected, the causee can be bound by the causer, not vice versa. (80)a. Cai nyakit-ang awak cai-ne 2 AV.ill-CAUS yourself You hurt yourself. b. *Awak cai-ne nyakit-ang cai. yourselfav.ill-caus 2 (Lit. Yourself hurt you. ) c. Awak caine sakit-ang cai. yourself OV.ill-CAUS 2 You hurt yourself. d. *cai sakit-ang awak caine. 2 OV.ill-CAUS yourself (Lit. Yourself hurt you.) If the base verb is a transitive verb, the derived causative is a three-place predicate. With the -ang marked causative, the causee is the second term argument and the lower patient becomes oblique:

56 54 (81) Tiang ngugut-ang lalipine sig Nyoman-ne. 1 AV-bite-CAUS snake-def at Nyoman-DEF I made the snake bite Nyoman. We predict that the reflexive oblique theme can be bound by the causee or the causer (82)a, but not vice versa (82)b. (82)a. Ia i ngugut-ang lalipine j sig awakne i /j 3 AV-bite-CAUS snake-def at self He i made the snake j bite himself i /itself j. b. *Tiang ngugut-ang awakne sig lalipi-ne 1 AV-bite-CAUS self at snake-def I made the snake bite itself. With the -in marked causative, a causativized transitive is ditransitive (double object): the causee is the second term and the lower theme the third. (83) Tiang nyuun-in Ni Luh Wayan banten. 1 AV-carry.on.head-CAUS Ni Luh Wayan offering I made Ni Luh Wayan carry the offering. Since the theme banten offering is the last argument, it is a-commanded by the agent causer (a-subj) tiang I and by the causee (second argument) Ni Luh Wayan. We predict therefore that a pronominal in the third argument can be bound from either of these arguments (84)a,b, and the A-subject can bind the

57 55 second argument (84)c. But the lower theme cannot bind the causee (84)d, even if this lower theme is the subject of the clause: (84)a. Sabilang jlema i nyuun-in Ni Luh Wayan banten-ne i every person AV-bear.on.head-CAUS Ni Luh Wayan offering-3poss Every person i made Ni Luh Wayan carry his/her offering i. b. [Sabilang tukang banten] i suun-in tiang banten-ne i every maker offering OV-bear.on.head-CAUS 1 offering-3poss I made every offering-maker i carry his/her offering i. c. Sabilang jlema i nyuun-in pianak-ne i banten every person AV.bear.on.head-CAUS child-3poss offering Every person i made his/her i child carry an offering on the head. d.?*sabilang banten i suun-in tiang [tukang banten-ne] i. every offering OV-bear.on.head-CAUS 1 maker offering-3poss I had every offering i carried by its offering-maker i. In conclusion, causative verbs have the same binding properties as simple verbs, as automatically predicted by our theory of argument structure, together with the assumption that causativization adds a new A-subject Applicatives. Balinese has two applicative verbal suffixes which signal that a term argument has been added to the argument structure. Each suffix is restricted to a range of semantic types for the added role:

58 56 (85) -in: Locative or Source roles -ang: Instrumental, Benefactive, or Recipient roles Sentence (a) uses the non-applied form, so the locative argument appears in a PP. Sentence (b) is the applied (-in suffix) form, in objective voice, while (c) illustrates the applied form in Agentive Voice: (86)a. Cai teka ka kantor-e ene. 2 come to office-def this b. Kantor-e ene teka-in cai. office-def this OV.come-APPL 2 c. Cai neka-in kantor-e ene. 2 AV.come-APPL office-def this You came to this office. Note that the applied locative role is the subject in (b) and the object in (c). We analyze applicativization as an operation which adds a term argument in the second place in ARG-S, that is, immediately following the A-subject. (87) a. base verb: [ARG-S < [1]NP, >] b. applied verb: [ARG-S < [1]NP, NPa, >] (where NPa is the applied argument)

59 57 This rule captures a celebrated cross-linguistic observation about applicatives, and further generalizes that observation to account for applicatives in ergative as well as accusative clauses. The observation, originally due to Marantz (1984), is dubbed Marantz s Generalization by Baker (1988: 246). Baker explains it this way: whenever a verb appears with both extra morphology and an additional NP argument bearing some oblique thematic role (a pretheoretical characterization of applicatives), that additional NP argument will behave like the surface direct object of the complex verb. (Baker 1988, p. 246) As stated here, Marantz s Generalization is true of accusative clauses but not ergative ones: in ergative clauses, the added NP behaves like the surface subject, not direct object. What (87) says instead is that the applied NP must be, so to speak, the logical direct object, that is, the second item in ARG-S. If the clause is accusative the applied role maps to the object, if ergative it maps to the subject. As with causatives, (87) can be implemented by means of a sort declaration for the applied verb, but we will not spell this out here. Applicativization operates on ARG-S and does not itself determine the mapping to Valence lists. Instead it creates the input to the mapping, which is then determined by the AV-verb and OV-verb sorts given above. Hence the applied NP becomes the subject of a monotransitive OV verb (as in (86)b) but the object of an AV verb (as in (86)c). Turning to derived ditransitives, that is, monotransitive roots to which a third term argument has been added, our analysis makes two predictions:

60 58 Prediction I: Linking symmetry. Applied OV ditransitives should be symmetrical in the sense that either non-a-subject term can be the subject. This follows immediately since OV specifies that the SUBJ is any term in ARG-S except the A-subject. The correctness of this prediction is illustrated here: (88)a. Bungkusan dadua jemak-ang=a Nyoman Santosa. package two OV.take-APPL=3 (name) (s)he brought two packages to Nyoman Santosa. b. Nyoman Santosa jemak-ang=a bungkusan dadua. (name) OV.take-APPL=3 package two (s)he brought two packages to Nyoman Santosa. Prediction II: Binding asymmetry. Applied OV ditransitives are asymmetrical with respect to binding: the applied NP can bind the other non-a-subject, but not vice versa. This also follows straightforwardly from the fact that the applied argument is the highest item in ARG-S, excepting the A-subject. Hence it a- commands all non-a-subjects. This is illustrated by the examples of binding by quantified noun phrases (QNPs):

61 59 (89)a. *sabilang potrekan edeng-in=a jleman-ne. every picture OV.show-APPL=3 person-poss (S)he showed every picture i (to) its i owner. b. Sabilang jlema OV.edeng-in=a potrekan-ne. every person show-appl=3 picture-poss (S)he showed every person i his/her i picture. The agent is expressed by the 3rd person clitic. The goal is the applied role, so it a-commands the theme. Hence a QNP goal can bind into a theme (as in (b)) but not vice versa, as shown in (a). Assuming goals thematically outrank themes, then since both are terms the above facts might seem to follow already, independently of our assumption that applicativization adds the second role. From that point of view, even stronger evidence comes from the behavior of locatives, which are generally assumed to fall below Theme on the thematic hierarchy. An applied locative can bind into a theme, but not vice versa: (90)a. Sabilang toko anggeh-in=a baas-ne ane paling luunga every shop OV.charge-APPL=3 rice-poss best In every shop i (s)he bought its i best rice on credit. b. *Sabilang sate jang-in=a bodag-ne. every satay OV.put-APPL=3 basket-poss (S)he put every satay i in its i basket Being applied, the locative falls into the second position in ARG-S, above the theme. Our applicative rule (87) is structurally defined on ARG-S position,

62 60 without reference to thematic role type. Here it leads to an ARG-S ordering where the locative a-commands the theme, departing from the usual hierarchy Passive Passive as an operation on argument structure. Balinese has two passive constructions, one for each register. In the High Passive construction, shown in (91)a, the verb bears the passive ka- prefix, and the A-subject argument optionally appears as a PP by-phrase, headed by the preposition antuk (for high register only) or teken (unmarked, commonly for low register). The Low Passive in (91)b is a historical reanalysis of the OV clitic construction, in which the agent clitic has been reinterpreted as a passive morpheme (evidence will be given below; see also Arka 1997 for further details). Hence many sentences are ambiguous between OV and Low Passive. However, the addition of a by-phrase, as in (91)b, forces the Low Passive interpretation. (91) a. High passive Buku-ne ka-ambil (antuk/teken I Guru). (h.r.) book-def PASS-take (by Art teacher) The book has been taken (by the teacher). b. Low Passive Buku-ne jemak-a teken Wayan. book-def OV.take-PASS by Wayan The book has been taken by Wayan. The High and Low Passives resembles the OV construction, but there are interesting differences as we shall see.

63 61 We analyze the Balinese passive as demotion of the A-subject to an optional by-phrase. By the theory of argument structure (53)a, this oblique automatically follows any terms in the ARG-S list: (92) a. High register passive pass-verb: PHON ka+φ ARG S 1 ( PP[antuk/teken] i ) STEM PHON Φ ARG S NP i 1 b. Low register passive pass-verb: PHON Φ+a ARG S 1 ( PP[teken] i ) STEM PHON Φ ARG S NP i 1 We treat the passive form as a verb rather than a stem (see (70)a). With respect to morphology this correctly predicts that the passive morpheme cannot cooccur with AV marking. With respect to syntax this predicts that either term argument of a passive ditransitive can be the subject, since neither OV nor AV morphology places any constraints on subject selection. Sentence (93)a is an AV ditransitive; (b) and (c) are passives with the recipient and the theme, respectively, as the subject.

64 62 (93)a. Ida sampun numbas-ang okan idane motor anyar (h.r.) 3 already AV.buy-APPL child 3-DEF motor-bike new (s)he bought his child a new motor bike. b. Okan idane sampun ka-tumbas-ang motor anyar (antuk ida) child 3-DEF already PASS-buy-APPL motor-bike new (by 3) His child was bought a new motor bike (by him/her). c. Motor anyar sampun ka-tumbas-ang okan idane (antuk ida) motor-bike new already PASS-buy-APPL child 3-DEF (by 3) A new motor bike was bought (for) his child (by him/her) Binding in Passive and OV clauses. here: Recall that an OV agent can bind into the subject, as in (40), repeated (94) Ragan idane cingakin ida. (h.r.) self 3 OV.see 3 (S)he saw himself/herself. In contrast, a passive by-phrase cannot bind into the subject, as we will show. Simple examples like the following seem to suggest the opposite conclusion, namely that the passive by-phrase can bind into the subject: (95) Ragan idane tan ka-runguang antuk ida. (h.r.) self 3 NEG PASS-care by 3 (s)he does not take care of him/herself.

65 63 However, the apparent binding in (95) turns out to be illusory. To see why, consider the following generalization about Balinese. Whenever a term locally a- commands the anaphor, then the by-phrase can no longer antecede the reflexive. Instead, the local a-commander must bind the anaphor. Sentences in (96) have OV-verbs whereas the ones in (97) have the corresponding passive verbs. The actor (index j) can be a possible binder in (96), but it cannot in (97). (96) OV-verb: a. Anak cenik-e ento i edengin=a j awakne i/j di kaca-ne person small-def that OV.show=3 self at mirror-def b. Awakne i/j edengin=a j anak cenik-e ento i di kaca-ne self OV.show=3 person small that in mirror-def He j showed the child i himself i/j in the mirror (97) High Passive (h.r.): a. Anak alit-e punika i ka-edengin ragane i/*j ring kaca-ne (antuk ida j ) person small that PASS-show self in mirror-def (by 3) b. Ragane i/*j ka-edengin anak alit-e punika i ring kaca-ne (antuk ida j ) self PASS-show person small-def that in mirror-def (by 3) The child i was shown himself i/*j in the mirror (by him j ) If the reflexive is a-commanded but there is no suitable binder, then the by-phrase still cannot bind, and the sentence becomes unacceptable.

66 64 (98)a. OV-verb: Bambang-e ento pulang-in=a i awakne i hole-def that OV.drop=APPL=3 self In the hole (s)he i dumped herself/himself i. b. High Passive (h.r.):?*bambang-e punika ka-pulang-in ragan idane i antuk ida i. hole-def that PASS-drop-APPL self by 3 In the hole, (s)he i dumped herself/himself i. The Low Passive provides a similar contrast with OV-verbs. As noted above, the Low Passive morpheme is homophonous with the agent clitic, but the inclusion of the by-phrase disambiguates, forcing the Low Passive interpretation. Hence (99)a and b form a minimal pair (examples from Arka 1997): (99)a. OV-verb: [Anake cenik ento] i edengin=a j awakne i/j di kacane person small that OV.show=3 self at mirror-def He j showed the child i himself i/j in the mirror. b. Low Passive: Anake cenik ento i edengin-a awakne i/*j di kacane teken ia j person small that show-pass self at mirror-def by 3 The child i was shown himself i/*j in the mirror by him j. c. High Passive (h.r.): Anake alit punika i ka-edengin ragane i/*j ring kacane antuk ida j person small that PASS-show self at mirror-def by 3 The child i was shown himself i/*j in the mirror by him j.

67 65 Just as in the examples above, the agent clitic in (99)a can bind the reflexive while the by-phrase in (99)b cannot. The High Passive in (99)c parallels the Low Passive in (99)b. The above facts actually follow straightforwardly from the binding theory in (50) above (Pollard and Sag 1992, 1994). Principle A applies only to a- commanded anaphors; a non-a-commanded anaphor is exempt from the structural binding theory (BT) and instead is subject to discourse and intervention factors (cp. Giorgi 1984 for a similar theory of exempt anaphors within GB). 22 Since it has no a-commander, the reflexive in (95) is an exempt anaphor, while the reflexives in (97), (98) and (99) are non-exempt. (100) a. ex. (95): ARG-S <NP:ana, NP, PP[antuk]> (exempt from BT) b. ex. (98)b: ARG-S <NP, NP:ana, PP[antuk]> (BT applies) In fact sentences like (95) are possible even with the by-phrase omitted, given proper discourse conditions, as in the following discourse: (101) Ida i inceg pisan ng-runguang sisian idane. (h.r.) 3 busy very AV.care student 3-DEF (S)he i was very busy taking care of his/her students Ragan idane i tan ka-runguang. self 3 NEG PASS-care (s)he does not take care of him/herself. (lit. Herself i is not taken care of. )

68 66 In conclusion, there is no structural binding by passive by-phrases only discourse binding. In contrast, however, an OV agent can bind into the subject, as we saw in (40)/(94) above. Unlike the passive, this binding is unperturbed by the appearance of intermediate arguments, as we saw in (96), (98)a and (99)a. The OV construction allows true structural binding by the agent, while the passive does not. This important difference follows immediately from our main claim that passive alters argument structure (placing the agent at the bottom of the list), while OV voice marking does not (leaving the agent at the top of the list). Instead, the OV/AV alternation reflects the distinction between two universal linking patterns, ergative and accusative Summary. Summarizing this section, we define common operations on argument structure in terms of the position on ARG-S at which an argument is added or suppressed/demoted: 23 (102) Common Operations on Argument Structure. First in ARG-S Add an argument causative Demote an argument passive Second in ARG-S applicative (antipassive) Since binding is defined at ARG-S, these processes crucially affect binding possibilities. But OV/AV morphology affects only the mapping between ARG-S

69 67 and VALENCE features, which does not affect binding in this way. We schematically show the effects of these types of morphology in this chart: (103) Schematic diagram of the effects of verb morphology ARG-S ARG-S ARG-S VALENCE add morphemes encoding operations on ARG-S (passive, causative, applicative, etc.) add morphemes regulating linking between ARG-S and VALENCE (ergative, accusative, etc.) 6. Binding in Balinese SSR and SOR constructions. Subject to subject raising (SSR) and subject to object raising (SOR) constructions provide an interesting proving ground for our theory of binding. As we will see, the theory correctly predicts the facts without modification. However, to apply the HPSG binding theory to raising constructions in a syntactically ergative language we will need to spell out a minor aspect of the interpretation of that theory. As in English, Balinese binding in raising constructions applies to the surface position of the NP. In the English example (104), the raised NP he receives no theta role from seem but still binds the reflexive in the to-pp, and cannot bind a pronominal in that position. Similar facts obtain for QNP binding of the pronoun in (105).

70 68 (104)a. I wonder whether he i seems to himself i to be ugly. b. *I wonder whether he i seems to him i to be ugly. (105)a. Every girl i seems to her i parents to be the prettiest. b. *It seems to her i parents that every girl i is the prettiest. The same facts apply in Balinese SSR (apa is an interrogative complementizer, glossed as Q): (106)a. Takonang tiang apa ia i ngenah sig awakne i jelek sajan. OV.ask 1 Q 3 seem to self bad very I asked (him) whether he seemed to himself to be very ugly. b. *Takonang tiang apa ia i ngenah sig ia i jelek sajan. OV.ask 1 Q 3 seem to 3 bad very (*I asked (him) whether he i seemed to him i to be very ugly.) (107)a. Jani sabilang anak bajang i ngenah sajan sig rerama-ne i now every person young AV.seem very to parent-3poss suba matunangan. already ma-lover Nowadays it is very apparent to her i parents that every young girl i has got a lover.

71 69 b. *Jani ngenah sajan sig rerama-ne i [sabilang anak bajang i suba now seem very to parent-3poss every person young already matunangan]. ma-lover (*Nowadays it is very apparent to her i parents that every young girl i has got a lover.) Within the GB framework, patterns like (104)-(105) have been adduced as evidence that binding can apply later than D-structure. In HPSG these facts follow automatically. Raising, like almost all grammatical phenomena, is handled with the mechanism of structure sharing: the raised NP is simultaneously the subject of the lower predicate and the subject (in SSR) or object (in SOR) of the higher predicate (indicated by the reentrancy tag [1] in (108)). Thus ARG-S includes raised items (whose distribution is tightly regulated in HPSG by the Raising Principle; Pollard and Sag 1994, p. 140). In the ARG-S for ngenah seem, as in (107), the raised NP a-commands the PP:

72 70 (108)a. seem: ARG S < 1 NP, PP[to] i, VP[SUBJ < 1 >]:j > CONTENT seem(i, j) b. ngenah: ARG S < 1 NP, PP[sig] i, VP[SUBJ < 1 >]:j > CONTENT seem(i, j) Hence it is expected that the raised subject can bind into the PP, as we saw in (104)-(107). Note that the raised NP is not linked to any semantic role of the seem relation. Binding in SOR constructions (see Section 5.3) receives exactly the same treatment. Let us consider how this works in Balinese. First note that a subordinate subject reflexive cannot be bound from the matrix clause; a pronoun must appear instead, as in (109). (109) Ia i tusing nawang apa ia/?*awakne i suba ngemaang I Nyoman pipis-ne 3 NEG know Q 3/self PERF give I Nyoman money-3poss She i does not know whether she i has given I Nyoman his money But raising the embedded subject into the matrix clause as in (110) yields the opposite pattern: the reflexive is fine ((110)a,b), the ordinary pronoun unacceptable ((110)c,d) (regardless of whether we use AV (110)a,c or OV (110)b,d on the matrix verb):

73 71 (110)a. Ia i nawang awakne i lakar tangkep polisi. 3 AV.know self FUT OV.arrest police b. Awakne i tawang=a i lakar tangkep polisi. self OV.know=3 FUT OV.arrest police He i knew that the police would arrest self i. c. *Ia i nawang ia i lakar tangkep polisi. 3 AV.know 3 FUT OV.arrest police He i knew that the police would arrest him i. d. *Ia i tawang=a i lakar tangkep polisi. 3 OV.know=3 FUT OV.arrest police He i knew that the police would arrest him i. Recall our supposition that ARG-S is not affected by the OV/AV distinction; instead this morphology regulates the mapping between ARG-S and grammatical expression. Hence the ARG-S of know is identical whether in AV form as in (110)a or in OV form as in (110)b. The semantic content is included here for clarity: (111) tawang: ARG S <NP i,1, VP[SUBJ < 1 NP > ]:j > CONTENT know(i, j) The A-subject NP i restricts the knower argument, while the VP restricts the proposition which is known. The middle argument is the raised NP, which is

74 72 token-identical to the subject specified by the unsaturated VP complement, as indicated by the tag [1]. Crucially, NP i a-commands the raised NP, regardless of whether they map onto the subject and object in OV fashion ([1]=subject, NP i =object) or AV fashion ([1]=object, NP i =subject). This correctly allows for (110)a,b, where the raised [1]NP is a reflexive bound by the A-subject, while ruling out (110)c,d as Principle B violations (since the pronominal is improperly a-bound). Now let s consider the ARG-S list of the embedded predicate, arrest in (110). In an accusative language such as English, the raised argument is always the A-subject of the lower predicate; consequently, if that argument is an anaphor then it is always exempt with respect to the lower predicate, vacuously satisfying Principle A; and if it is a pronominal then it is always free with respect to the lower predicate, thus satisfying Principle B. Turning to Balinese, if that predicate is ergative (OV) then an interesting situation arises: the raised NP corresponds not to the A-subject of the lower predicate, but rather to its logical object, i.e. the arrestee argument in (110). How exactly should the binding theory be interpreted? If a pronoun or anaphor simultaneously appears on two ARG-S lists, is it subject to the binding conditions on both lists? The most sensible interpretation a priori is the following. Principle B dictates the obviation domain for pronouns, so we take the strongest ( universal quantification ) interpretation of Principle B: a pronominal must be free in all ARG-S lists in which the pronominal appears. On the other hand, Principle A specifies where an anaphor must find its antecedent. Since an anaphor takes only one antecedent, it would be illogical to expect it to be bound on more than one

75 73 ARG-S list. Hence we adopt an existential interpretation of Principle A: An a- commanded anaphor must be a-bound on some ARG-S list. This interpretation gives the right results for (110) above. For example, in (110)a,b, the reflexive awakne is a-bound on the matrix predicate s ARG-S list, so it satisfies Principle A, despite the fact that it is a-commanded and unbound on the ARG-S list of the embedded predicate ( arrest ). If we change the voice of the embedded verb from OV to AV, then the unacceptable pronouns in (110)c,d become acceptable: (112)a. Ia i nawang polisi lakar nangkep ia i. 3 AV.know police FUT AV.arrest 3 b. Polisi tawang=a i lakar nangkep ia i. police OV.know=3 FUT AV.arrest 3 He i knew that the police would arrest him i. Since it is a pronoun, the arrestee argument ia him (in (112)a it is the second occurrence of ia) must be free on all ARG-S lists on which it appears. Indeed, this pronoun does not appear on the list for know, and it is free on the list for arrest (as polisi police is not coindexed with it) Conclusion. In this paper we have argued on the basis of Balinese that variation between syntactically accusative and ergative clauses results from different

76 74 mappings between argument structure and syntactic realization of arguments. In our proposed ARG-S list, arguments are first divided into direct and oblique arguments, then thematically ranked. Following standard HPSG, ARG-S also includes certain non-thematic arguments such as raised phrases, and is the locus of anaphoric binding relations. The ARG-S value for a verb represents the output of morpholexical operations on argument structure such as passive, causative, and applicative. This set of assumptions leads to an explanation for the familiar observation that Austronesian binding is generally sensitive to theta-roles and not grammatical relations. Moreover, we explain the precise exceptions to theta-sensitivity: that terms asymmetrically bind obliques, regardless of theta role type; that passive agents cannot bind into the subject (while ergative agents can); and that raised arguments assume their surface grammatical relations for the purpose of binding. * For their comments on earlier drafts or discussion of various aspects of this topic, we thank Lisa Travis, Carol Georgopoulos, Bill Foley, Jane Simpson, Chris Manning, Adrian Clynes, and the audiences at the following conferences: the Third International Conference on Head-Driven Phrase Structure Grammar, Marseille, France, May 21, 1996; the Workshop on Lexical-Functional Grammar, Grenoble, France, August 28, 1996; and the 4th Annual Meeting of the Austronesian Formal Linguistics Association, UCLA, April 26, Our primary informant, in addition to the second author, was Ni Luh Adnyawati, whom we also wish to thank.

77 75 1 These phenomena are interclausal in the sense of Dixon (1979, 1994): they relate a clause to elements outside of it. 2 In morphemic glosses we abbreviate first, second, and third person pronouns with 1, 2, and 3 respectively. Most Balinese pronouns are unmarked for number and gender (except second person pronouns, which are marked for gender). However, for simplicity we will often translate them as singular pronouns in the English paraphrases (e.g., I instead of the more accurate I/we ). 3 Balinese has two registers (speech levels), kasar or low register and alus or high register. Certain lexical forms, including some pronouns, differ in the two styles. In this paper, high register is indicated by h.r. ; low register examples are unmarked. 4 These terms are used inconsistently, in part because of differences in theoretical assumptions. this paper: (grammatical) subject A-subject Dixon 1979, 1994: pivot (logical) subject Guilfoyle et al 1992: topic(spec,ip) subject(spec,vp) Manning 1994 pivot A-subject Bresnan & Kanerva 1989 SUBJ thematic subject 5 In addition to subjects, possessors of subjects can be relativized (as in (i)), but not possessors of objects (as in (ii)): (i) anake [ane panakne ngeling] ento

78 76 person-def REL child-3poss cry that the person whose son is crying (ii) *[Anake [ane tiang ngalih panakne] person-def REL 1 AV.search child-3poss the person whose son I searched 6 The predicate ngenah has the initial nasal, but we do not gloss it as AV since it lacks a non-nasal counterpart. 7 This is not universal. In Dyirbal (Dixon 1979, 1994; Bittner and Hale 1996b), in Tagalog volitive clauses (Kroeger 1993), and in Polynesian languages (Chung 1978), the controllee must be the logical subject, our A-subject, and not necessarily the grammatical subject. But Balinese control, like that of English, picks the grammatical subject. We suspect that this is related to the fact that Balinese, like English but unlike Tagalog or Dyirbal, has fairly rigid word order and configurational structure, so that the VP is a rather salient constituent. We will not pursue this issue here. 8 Clynes (1995) argues that the post-verbal OV agent is incorporated into the verb. This analysis, if tenable, would explain the indefiniteness restriction, since crosslinguistically incorporated nouns tend to receive a generic interpretation. However, Arka (in progress) shows that there are serious problems for the incorporation analysis. For one, the agent can be a complex phrasal NP as in example (34)b. See Arka (in progress) for further syntactic and phonological evidence against incorporation.

79 77 9 The so-called subjective passive (or object preposing ) construction in Bahasa Indonesia has some of the properties listed in this paragraph but has nevertheless been analyzed as a type of passive by Chung (1976) and Guilfoyle, Hung, and Travis (1992). 10 The same restriction to terms applies to other Balinese quantifiers, including onya all and makejang all. In general quantifiers cannot be construed with indefinites (cp.?*sm girls all left.) and Balinese quantifiers, with the exception of ajak makejang, cannot apply to pronouns. Since OV agents must be indefinite or pronominal, ajak makejang becomes our best test. Artawa (1994) claimed that quantifier float picks only subjects in Balinese, but this appears to be incorrect. 11 The more elaborated LFG binding framework in Bresnan 1995a allows for three types of binding condition: (i) linear order; (ii) grammatical relations; (iii) argument structure. Also, a great deal of work in GB has been devoted to thematic conditions on binding. 12 In earlier versions of HPSG (Pollard and Sag 1992, 1994) ARG-S and the VALENCE lists are conflated on one list, called SUBCAT. The ordering relation was termed o-command (for obliqueness command ). Following Manning and Sag 1995, we prefer a-command ( argument command ), since ARG-S diverges from obliqueness in ergative clauses. 13 The HPSG binding theory in (50) (Pollard and Sag 1992, 1994) owes an obvious debt to Chomsky 1981, as recognized by the choice of terminology. However, there are at least three key differences:

80 78 (i) The distinction between a-command, which is valence-theoretic, and c- command, a configurational relation defined on phrase structure, has many consequences for the HPSG theory which GB theory does not share: subject-toobject raising structures must exist; argument PPs are handled straightforwardly (in John talked to Mary about herself, Mary a-commands but does not necessarily c-command herself); etc. (ii) In HPSG, binding domains do not just expand as conditions fail (e.g. when there is no accessible subject). Instead certain anaphors (namely non-acommanded ones) are just exempt from Principle A in the HPSG binding theory. This position is argued to be correct by Pollard and Sag (1992, 1994). (iii) Many GB definitions differ in HPSG or have no correlate at all: there is no [+a,+p] category (PRO) in HPSG, no notion of government, etc. Also the HPSG equivalent of wh-trace is not necessarily an R-expression, but rather has the binding status of its antecedent phrase. 14 PHON, SUBJ, and COMPS have singleton lists in this example. 15 Judgements vary on these examples. Postal (1971, p. 126) finds both (57)b and b unacceptable, while Pollard and Sag (1994, p. 276) finds them both acceptable. Anyway, neither sentence seems significantly worse than the other, while one would expect such an asymmetry on a purely thematic account which ignores the direct/oblique distinction. 16 Given two lists L1 and L2, the list L1uL2 contains all and only the elements of L1 and L2, preserving ordering on L1 and on L2 but freely intermingling their

81 79 elements. For example: <a,b> u <1,2> = <a,b,1,2> <a,1,b,2> <a,1,2,b> <1,a,b,2> <1,a,2,b> <1,2,a,b>. See Reape While AV is the morphologically inflected form and OV the zero form, we assume that neither form is syntactically more basic than the other. It is true, as two reviewers have noted, that we must stipulate which form receives morphology. But this may be an advantage of our account. Cognate languages vary with respect to morphological markedness: in Toba Batak, which has been analyzed along lines very similar to the present analysis (Schachter 1984), both OV and AV are morphologically marked (with di- and mang-, respectively); in Bahasa Indonesia the AV nasal prefix (me)ng- (cognate with the Balinese AV marker) is usually optional and sometimes prohibited (e.g. in extraction environments). 18 More precisely, ppro is the sort for this NP s CONTENT value. The colon in NP:ppro abbreviates the attribute CONTENT. 19 In another respect the present account resembles Marantz 1984 and Levin 1983: all three approaches involve allowing the association lines between argument structure and grammatical relations to cross for ergative clauses. The difference is mainly that argument structure is more syntacticized for us. 20 Predicates can remain unsaturated in HPSG. 21 In particular, Balinese causatives lack the biclausal binding properties found with causative constructions in some languages. See Manning and Sag 1995 for an analysis of the biclausal type.

82 80 22 This theory accounts for picture NP contrasts like: The boys i were surprised by what they saw in the Post Office. (*Mary's) pictures of themselves i were on the wall. When the possessive NP does not appear then the reflexive has no a- commander and is thus exempt. See Pollard and Sag 1992, 1994 for further discussion. (See however Baker 1995 for some problematic reflexives in literary British English.) 23 In studies of languages such as Inuit, causative constructions are not restricted to semantic causation, but also can be formed with verbs of thinking, saying, etc. What these share with semantic causatives is that the added argument is the A- subject. Antipassive does not occur in Balinese. 24 Extending the binding analysis to cover raising constructions, while straightforward in HPSG, turns out to be surprisingly difficult for GB analyses of Austronesian such as Guilfoyle, Hung and Travis (1992). On that analysis thetaroles are discharged under VP and the [Spec,IP] position, to which an argument sometimes moves for Case, is associated with surface subject properties like relativization (cp. our SUBJ list item). The problem is that for some purposes [Spec,IP] must be an A-bar position, while for others it must be an A-position. Specifically, to account for binding in simplex clauses, it clearly must be an A-bar position (since, e.g., the agent binds the theme irrespective of which one is in [Spec,IP]). Turning to raising, we saw above that the raised NP acts, for purposes of the binding within the matrix clause, as if it is in an A-position (see (106)). Since only the lower subject can raise, we know that it has passed through the lower [Spec,IP] position. But movement from A-bar position to A-position is

83 81 improper. Ergo the lower subject must also be an A-position, a contradiction. See Wechsler and Arka 1997 and Travis 1997 for discussion; the latter paper proposes a radical revision of GB binding theory to solve this problem.

84 82 References Åfarli, Tor A.: 1989a, Passive in Norwegian and in English, Linguistic Inquiry, 20.1, Alsina, Alex and Sam Mchombo: 1993, Object Asymmetries and the Chichewa Applicative Construction, in S. Mchombo (ed.) Theoretical Aspects of Bantu Grammar, CSLI Publications, Stanford, Andrews, Avery: 1985, The major functions of the noun phrase, in T. Shopen (ed.) Language Typology and Syntactic Description, vol. I: Clause Structure, Cambridge University Press, Cambridge, England, pp Arka, Wayan: in progress, From Morpho-syntax to Pragmatics in Balinese. PhD dissertation, Department of Linguistics, The University of Sydney, Sydney, Australia. Arka, Wayan: 1997, On the double life of a pronominal form and two passives in Balinese. Departmental Seminar paper, University of Sydney Linguistics Department, Sydney, Australia, May 9, Arka, Wayan and Stephen Wechsler: 1996, Argument Structure and Linear Order in Balinese Binding. Paper read at the Workshop on Lexical-Functional Grammar, Grenoble, France, August 26 28, Artawa, K.: 1994, Ergativity and Balinese Syntax. PhD thesis, La Trobe University, Bundoora, Australia. Baker, C.L.: 1995, Contrast, discourse prominence, and intensification, with special reference to locally free reflexives in British English, Language 71.1,

85 83 Baker, Mark: 1988, Incorporation A Theory of Grammatical Function Changing. University of Chicago Press, Chicago. Barss, Andrew, and Howard Lasnik: 1986, A Note on Anaphora and Double Objects. Linguistic Inquiry 17, Bittner, Maria and Ken Hale: 1996, Ergativity: Towards a Theory of a Heterogeneous Class, Linguistic Inquiry 27.4, Bobaljik, Jonathan: 1993a, Nominally Absolutive is not Absolutely Nominative, Jonathan Mead (ed.), Proceedings of the 11thWest Coast Conference on Formal Linguistics, Stanford Linguistics Association and CSLI, Stanford, California, Bobaljik, Jonathan: 1993b, Ergativity and Ergative Unergatives. In Colin Phillips (ed.) MIT Working Papers in Linguistics 19 Papers on Case and Agreement II, MIT Dept. of Linguistics, Cambridge, Bresnan, Joan: 1995a, Morphology Competes with Syntax: Explaining Typological Variation in Weak Crossover Effects, Manuscript, Stanford. Bresnan, Joan: 1995b, Linear Order, Syntactic Rank, and Empty Categories: On Weak Crossover. In M. Dalrymple, R.M. Kaplan, J.T. Maxwell III, and A. Zaenen (eds.) Formal Issues in Lexical-Functional Grammar, CSLI Publications, Stanford, pp Bresnan, Joan and Jonni Kanerva: 1989, Locative Inversion in Chichewa: A Case study of factorization in Grammar Linguistic Inquiry, 20:1-50. Bresnan, Joan, and Lioba Moshi: 1990, Object Asymmetries in Comparative Bantu Syntax. Linguistic Inquiry 21.2:

86 84 Carrier, Jill and Janet Randall: 1992, The Argument Structure and Syntactic Structure of Resultatives. Linguistic Inquiry 23.2, Chomsky, Noam: 1981, Lectures on Government and Binding The Pisa Lectures. Foris Publications, Dordrecht. Chung, Sandra: 1976, On the Subject of Two Passives in Indonesian, in C. N. Li (ed.) Subject and Topic, Academic Press, NewYork, pp Chung, Sandra: 1978, Case Marking and Grammatical Relations in Polynesian. University of Texas Press, Austin. Clynes, Adrian: 1995, Topics in the Phonology and Morphosyntax of Balinese based on a dialect of Singaraja, North Bali. PhD Thesis, Australia National University. Dalrymple, Mary: 1993, The Syntax of Anaphoric Binding. CSLI Publications, Stanford. Davis, Tony: 1996, Linking and the Hierarchical Lexicon. PhD thesis, Stanford. Dixon, R.M.W.: 1979, Ergativity, Language 55, Dixon, R.M.W.: 1994, Ergativity. Cambridge University Press, Cambridge, England. Durie, Mark: 1985, A grammar of Acehnese on the basis of a dialect of North Aceh. Foris, Dordrecht. Farkas, Donka: 1988, On Obligatory Control. Linguistics and Philosophy 11, Flickinger, Daniel: Lexical Rules in the Hierarchical Lexicon. Doctoral dissertation, Stanford University Linguistics Department.

87 85 Foley, William and Robert Van Valin: 1984, Functional Syntax and Universal Grammar. Cambridge University Press, Cambridge. Gawron, Jean Mark: 1986, Situations and Prepositions, Linguistics and Philosophy 9, Giorgi, Allesandra: 1984, Toward a Theory of Long Distance Anaphors: a GB Approach. The Linguistic Review 3, Grimshaw, Jane: 1990, Argument Structure. MIT Press, Cambridge, MA. Guilfoyle, Eithne, Henrietta Hung, and Lisa Travis: 1992, SPEC of IP and SPEC of VP: two subjects in Austronesian languages. Natural Language and Linguistic Theory 10.3, Hellan, Lars: 1988, Anaphora in Norwegian and the Theory of Grammar. Foris, Dordrecht. Keenan, E. and B. Comrie: 1977, Noun phrase accessibility and universal grammar. Linguistic Inquiry 8, Kroeger, Paul: 1993, Phrase Structure and Grammatical Relations in Tagalog. CSLI Publications, Stanford. Langacker, R.: 1969, On Pronominalization and the Chain of Command. In D. A. Reibel and S. A. Schane, eds., Modern Studies in English, Prentice- Hall, Englewood Cliffs, N.J., pp Levin, Beth: 1983, On the Nature of Ergativity. PhD dissertation, MIT, Cambridge. Manning, Chris and Ivan Sag: 1995, Dissociations Between Argument Structure and Grammatical Relations. Draft, manning/papers/tuebingen.ps.

88 86 Manning, Chris: 1994, Ergativity: Argument Structure and Grammatical Relations, PhD thesis, Stanford. Marantz, Alec: 1984, On the Nature of Grammatical Relations. MIT Press, Cambridge. Pollard, Carl and Ivan Sag: 1987, Information-Based Syntax and Semantics--Vol. 1 Fundamentals. Stanford: CSLI Publications. Pollard, Carl and Ivan Sag: 1992, Anaphors in English and the Scope of Binding Theory, Linguistic Inquiry 23.2, Pollard, Carl and Ivan Sag: 1994, Head-Driven Phrase Structure Grammar. CSLI Publications, Stanford, California; and University of Chicago Press, Chicago. Postal, Paul: 1971, Cross-Over Phenomena. Holt, Rinehart, and Winston, New York. Reape, Mike: 1994, Domain Union and Word Order Variation in German, in John Nerbonne and Klaus Netter and Carl Pollard (eds.), German in Head- Driven Phrase Structure Grammar. CSLI Publications, Stanford University, pp Sag, Ivan and Carl Pollard: 1991, An Integrated theory of Complement Control, Language 67, Schachter, Paul: 1977, Reference-Related and Role-Related Properties of Subjects. In Peter Cole and Jerrold Sadock (eds.) Syntax and Semantics Volume 8 Grammatical Relations, Academic Press, New York, pp

89 87 Schachter, Paul: 1984, Semantic-Role-Based Syntax in Toba Batak, in P. Schachter (ed.) Studies in the Structure of Toba Batak, UCLA Occasional Papers in Linguistics No. 5, UCLA, pp Travis, Lisa: 1997, Theta-positions and Binding in Balinese and Malagasy. Paper presented at the Fourth Annual Meeting of the Austronesian Formal Linguistics Association, University of California at Los Angelos, April 27, Uszkoreit, Hans: 1987, Word Order and Constituent Structure in German. CSLI Publications, Stanford. Wechsler, Stephen: 1994, Preposition Selection Outside the Lexicon. In R. Aranovich, W. Byrne, S. Preuss, and Martha Senturia (eds.) Proceedings of the 13th West Coast Conference on Formal Linguistics, CSLI Publications, Stanford, pp Wechsler, Stephen: 1995, The Semantic Basis of Argument Structure. CSLI Publications, Stanford, California. Wechsler, Stephen and Wayan Arka: 1997, Balinese Binding and Raising. Paper presented at the Fourth Annual Meeting of the Austronesian Formal Linguistics Association, University of California at Los Angelos, April 26, Williams, Edwin: 1983, Against Small Clauses, Linguistic Inquiry 14, Williams, Edwin: 1987, Implicit Arguments, the Binding Theory, and Control, Natural Language and Linguistic Theory 5,

90 88 Woollams, Geoff: 1996, A Grammar of Karo Batak, Sumatra. Pacific Linguistics Series C-130, Department of Linguistics, Australian National University, Canberra. Zaenen, Annie, Joan Maling, and Höskuldur Thráinsson: 1985, Case and Grammatical Functions: the Icelandic Passive, Natural Language and Linguistic Theory 3, [Submitted September 24, 1996] [Revised July 24, 1997]

Movement and Binding

Movement and Binding Movement and Binding Gereon Müller Institut für Linguistik Universität Leipzig SoSe 2008 www.uni-leipzig.de/ muellerg Gereon Müller (Institut für Linguistik) Constraints in Syntax 4 SoSe 2008 1 / 35 Principles

More information

Index. 344 Grammar and Language Workbook, Grade 8

Index. 344 Grammar and Language Workbook, Grade 8 Index Index 343 Index A A, an (usage), 8, 123 A, an, the (articles), 8, 123 diagraming, 205 Abbreviations, correct use of, 18 19, 273 Abstract nouns, defined, 4, 63 Accept, except, 12, 227 Action verbs,

More information

Constraints in Phrase Structure Grammar

Constraints in Phrase Structure Grammar Constraints in Phrase Structure Grammar Phrase Structure Grammar no movement, no transformations, context-free rules X/Y = X is a category which dominates a missing category Y Let G be the set of basic

More information

English prepositional passive constructions

English prepositional passive constructions English prepositional constructions An empirical overview of the properties of English prepositional s is presented, followed by a discussion of formal approaches to the analysis of the various types of

More information

Non-nominal Which-Relatives

Non-nominal Which-Relatives Non-nominal Which-Relatives Doug Arnold, Robert D. Borsley University of Essex The properties of non-restrictive relatives All non-restrictive relative clauses include a wh-word. There are no that or zero

More information

Syntactic and Semantic Differences between Nominal Relative Clauses and Dependent wh-interrogative Clauses

Syntactic and Semantic Differences between Nominal Relative Clauses and Dependent wh-interrogative Clauses Theory and Practice in English Studies 3 (2005): Proceedings from the Eighth Conference of British, American and Canadian Studies. Brno: Masarykova univerzita Syntactic and Semantic Differences between

More information

Noam Chomsky: Aspects of the Theory of Syntax notes

Noam Chomsky: Aspects of the Theory of Syntax notes Noam Chomsky: Aspects of the Theory of Syntax notes Julia Krysztofiak May 16, 2006 1 Methodological preliminaries 1.1 Generative grammars as theories of linguistic competence The study is concerned with

More information

Paraphrasing controlled English texts

Paraphrasing controlled English texts Paraphrasing controlled English texts Kaarel Kaljurand Institute of Computational Linguistics, University of Zurich [email protected] Abstract. We discuss paraphrasing controlled English texts, by defining

More information

Semantics and Generative Grammar. Quantificational DPs, Part 3: Covert Movement vs. Type Shifting 1

Semantics and Generative Grammar. Quantificational DPs, Part 3: Covert Movement vs. Type Shifting 1 Quantificational DPs, Part 3: Covert Movement vs. Type Shifting 1 1. Introduction Thus far, we ve considered two competing analyses of sentences like those in (1). (1) Sentences Where a Quantificational

More information

The compositional semantics of same

The compositional semantics of same The compositional semantics of same Mike Solomon Amherst College Abstract Barker (2007) proposes the first strictly compositional semantic analysis of internal same. I show that Barker s analysis fails

More information

English Appendix 2: Vocabulary, grammar and punctuation

English Appendix 2: Vocabulary, grammar and punctuation English Appendix 2: Vocabulary, grammar and punctuation The grammar of our first language is learnt naturally and implicitly through interactions with other speakers and from reading. Explicit knowledge

More information

Chapter 1. Introduction. 1.1. Topic of the dissertation

Chapter 1. Introduction. 1.1. Topic of the dissertation Chapter 1. Introduction 1.1. Topic of the dissertation The topic of the dissertation is the relations between transitive verbs, aspect, and case marking in Estonian. Aspectual particles, verbs, and case

More information

A Beautiful Four Days in Berlin Takafumi Maekawa (Ryukoku University) [email protected]

A Beautiful Four Days in Berlin Takafumi Maekawa (Ryukoku University) maekawa@soc.ryukoku.ac.jp A Beautiful Four Days in Berlin Takafumi Maekawa (Ryukoku University) [email protected] 1. The Data This paper presents an analysis of such noun phrases as in (1) within the framework of Head-driven

More information

I have eaten. The plums that were in the ice box

I have eaten. The plums that were in the ice box in the Sentence 2 What is a grammatical category? A word with little meaning, e.g., Determiner, Quantifier, Auxiliary, Cood Coordinator, ato,a and dco Complementizer pe e e What is a lexical category?

More information

Morphology. Morphology is the study of word formation, of the structure of words. 1. some words can be divided into parts which still have meaning

Morphology. Morphology is the study of word formation, of the structure of words. 1. some words can be divided into parts which still have meaning Morphology Morphology is the study of word formation, of the structure of words. Some observations about words and their structure: 1. some words can be divided into parts which still have meaning 2. many

More information

Ling 201 Syntax 1. Jirka Hana April 10, 2006

Ling 201 Syntax 1. Jirka Hana April 10, 2006 Overview of topics What is Syntax? Word Classes What to remember and understand: Ling 201 Syntax 1 Jirka Hana April 10, 2006 Syntax, difference between syntax and semantics, open/closed class words, all

More information

According to the Argentine writer Jorge Luis Borges, in the Celestial Emporium of Benevolent Knowledge, animals are divided

According to the Argentine writer Jorge Luis Borges, in the Celestial Emporium of Benevolent Knowledge, animals are divided Categories Categories According to the Argentine writer Jorge Luis Borges, in the Celestial Emporium of Benevolent Knowledge, animals are divided into 1 2 Categories those that belong to the Emperor embalmed

More information

COMPUTATIONAL DATA ANALYSIS FOR SYNTAX

COMPUTATIONAL DATA ANALYSIS FOR SYNTAX COLING 82, J. Horeck~ (ed.j North-Holland Publishing Compa~y Academia, 1982 COMPUTATIONAL DATA ANALYSIS FOR SYNTAX Ludmila UhliFova - Zva Nebeska - Jan Kralik Czech Language Institute Czechoslovak Academy

More information

GMAT.cz www.gmat.cz [email protected]. GMAT.cz KET (Key English Test) Preparating Course Syllabus

GMAT.cz www.gmat.cz info@gmat.cz. GMAT.cz KET (Key English Test) Preparating Course Syllabus Lesson Overview of Lesson Plan Numbers 1&2 Introduction to Cambridge KET Handing Over of GMAT.cz KET General Preparation Package Introduce Methodology for Vocabulary Log Introduce Methodology for Grammar

More information

EAP 1161 1660 Grammar Competencies Levels 1 6

EAP 1161 1660 Grammar Competencies Levels 1 6 EAP 1161 1660 Grammar Competencies Levels 1 6 Grammar Committee Representatives: Marcia Captan, Maria Fallon, Ira Fernandez, Myra Redman, Geraldine Walker Developmental Editor: Cynthia M. Schuemann Approved:

More information

What s in a Lexicon. The Lexicon. Lexicon vs. Dictionary. What kind of Information should a Lexicon contain?

What s in a Lexicon. The Lexicon. Lexicon vs. Dictionary. What kind of Information should a Lexicon contain? What s in a Lexicon What kind of Information should a Lexicon contain? The Lexicon Miriam Butt November 2002 Semantic: information about lexical meaning and relations (thematic roles, selectional restrictions,

More information

Outline of today s lecture

Outline of today s lecture Outline of today s lecture Generative grammar Simple context free grammars Probabilistic CFGs Formalism power requirements Parsing Modelling syntactic structure of phrases and sentences. Why is it useful?

More information

19. Morphosyntax in L2A

19. Morphosyntax in L2A Spring 2012, April 5 Missing morphology Variability in acquisition Morphology and functional structure Morphosyntax in acquisition In L1A, we observe that kids don t always provide all the morphology that

More information

Pupil SPAG Card 1. Terminology for pupils. I Can Date Word

Pupil SPAG Card 1. Terminology for pupils. I Can Date Word Pupil SPAG Card 1 1 I know about regular plural noun endings s or es and what they mean (for example, dog, dogs; wish, wishes) 2 I know the regular endings that can be added to verbs (e.g. helping, helped,

More information

MARY. V NP NP Subject Formation WANT BILL S

MARY. V NP NP Subject Formation WANT BILL S In the Logic tudy Guide, we ended with a logical tree diagram for WANT (BILL, LEAVE (MARY)), in both unlabelled: tudy Guide WANT BILL and labelled versions: P LEAVE MARY WANT BILL P LEAVE MARY We remarked

More information

2. PRINCIPLES IN USING CONJUNCTIONS. Conjunction is a word which is used to link or join words, phrases, or clauses.

2. PRINCIPLES IN USING CONJUNCTIONS. Conjunction is a word which is used to link or join words, phrases, or clauses. 2. PRINCIPLES IN USING CONJUNCTIONS 2.1 Definition of Conjunctions Conjunction is a word which is used to link or join words, phrases, or clauses. In a sentence, most of conjunctions are from another parts

More information

Albert Pye and Ravensmere Schools Grammar Curriculum

Albert Pye and Ravensmere Schools Grammar Curriculum Albert Pye and Ravensmere Schools Grammar Curriculum Introduction The aim of our schools own grammar curriculum is to ensure that all relevant grammar content is introduced within the primary years in

More information

Rethinking the relationship between transitive and intransitive verbs

Rethinking the relationship between transitive and intransitive verbs Rethinking the relationship between transitive and intransitive verbs Students with whom I have studied grammar will remember my frustration at the idea that linking verbs can be intransitive. Nonsense!

More information

Constituency. The basic units of sentence structure

Constituency. The basic units of sentence structure Constituency The basic units of sentence structure Meaning of a sentence is more than the sum of its words. Meaning of a sentence is more than the sum of its words. a. The puppy hit the rock Meaning of

More information

A Survey of ASL Tenses

A Survey of ASL Tenses A Survey of ASL Tenses Karen Alkoby DePaul University School of Computer Science Chicago, IL [email protected] Abstract This paper examines tenses in American Sign Language (ASL), which will be

More information

L130: Chapter 5d. Dr. Shannon Bischoff. Dr. Shannon Bischoff () L130: Chapter 5d 1 / 25

L130: Chapter 5d. Dr. Shannon Bischoff. Dr. Shannon Bischoff () L130: Chapter 5d 1 / 25 L130: Chapter 5d Dr. Shannon Bischoff Dr. Shannon Bischoff () L130: Chapter 5d 1 / 25 Outline 1 Syntax 2 Clauses 3 Constituents Dr. Shannon Bischoff () L130: Chapter 5d 2 / 25 Outline Last time... Verbs...

More information

the subject called the voice. A sentence that begin with the subject or the

the subject called the voice. A sentence that begin with the subject or the 2. ACTIVE AND PASSIVE VOICE In English, the verb form which indicates whether the subject (person or object) of a sentence do something or something has been done on the subject called the voice. A sentence

More information

How the Computer Translates. Svetlana Sokolova President and CEO of PROMT, PhD.

How the Computer Translates. Svetlana Sokolova President and CEO of PROMT, PhD. Svetlana Sokolova President and CEO of PROMT, PhD. How the Computer Translates Machine translation is a special field of computer application where almost everyone believes that he/she is a specialist.

More information

2. SEMANTIC RELATIONS

2. SEMANTIC RELATIONS 2. SEMANTIC RELATIONS 2.0 Review: meaning, sense, reference A word has meaning by having both sense and reference. Sense: Word meaning: = concept Sentence meaning: = proposition (1) a. The man kissed the

More information

[Refer Slide Time: 05:10]

[Refer Slide Time: 05:10] Principles of Programming Languages Prof: S. Arun Kumar Department of Computer Science and Engineering Indian Institute of Technology Delhi Lecture no 7 Lecture Title: Syntactic Classes Welcome to lecture

More information

Syntactic Theory on Swedish

Syntactic Theory on Swedish Syntactic Theory on Swedish Mats Uddenfeldt Pernilla Näsfors June 13, 2003 Report for Introductory course in NLP Department of Linguistics Uppsala University Sweden Abstract Using the grammar presented

More information

Syntax: Phrases. 1. The phrase

Syntax: Phrases. 1. The phrase Syntax: Phrases Sentences can be divided into phrases. A phrase is a group of words forming a unit and united around a head, the most important part of the phrase. The head can be a noun NP, a verb VP,

More information

psychology and its role in comprehension of the text has been explored and employed

psychology and its role in comprehension of the text has been explored and employed 2 The role of background knowledge in language comprehension has been formalized as schema theory, any text, either spoken or written, does not by itself carry meaning. Rather, according to schema theory,

More information

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. Three Reasons for Not Deriving "Kill" from "Cause to Die" Author(s): J. A. Fodor Source: Linguistic Inquiry, Vol. 1, No. 4 (Oct., 1970), pp. 429-438 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/4177587.

More information

UNBOUND ANAPHORIC PRONOUNS: E-TYPE, DYNAMIC, AND STRUCTURED-PROPOSITIONS APPROACHES

UNBOUND ANAPHORIC PRONOUNS: E-TYPE, DYNAMIC, AND STRUCTURED-PROPOSITIONS APPROACHES FRIEDERIKE MOLTMANN UNBOUND ANAPHORIC PRONOUNS: E-TYPE, DYNAMIC, AND STRUCTURED-PROPOSITIONS APPROACHES ABSTRACT. Unbound anaphoric pronouns or E-type pronouns have presented notorious problems for semantic

More information

IP PATTERNS OF MOVEMENTS IN VSO TYPOLOGY: THE CASE OF ARABIC

IP PATTERNS OF MOVEMENTS IN VSO TYPOLOGY: THE CASE OF ARABIC The Buckingham Journal of Language and Linguistics 2013 Volume 6 pp 15-25 ABSTRACT IP PATTERNS OF MOVEMENTS IN VSO TYPOLOGY: THE CASE OF ARABIC C. Belkacemi Manchester Metropolitan University The aim of

More information

Appendix to Chapter 3 Clitics

Appendix to Chapter 3 Clitics Appendix to Chapter 3 Clitics 1 Clitics and the EPP The analysis of LOC as a clitic has two advantages: it makes it natural to assume that LOC bears a D-feature (clitics are Ds), and it provides an independent

More information

the primary emphasis on explanation in terms of factors outside the formal structure of language.

the primary emphasis on explanation in terms of factors outside the formal structure of language. Grammar: Functional Approaches William Croft MSC03 2130, Linguistics 1 University of New Mexico Albuquerque NM 87131-0001 USA Abstract The term functional or functionalist has been applied to any approach

More information

CHAPTER 7 GENERAL PROOF SYSTEMS

CHAPTER 7 GENERAL PROOF SYSTEMS CHAPTER 7 GENERAL PROOF SYSTEMS 1 Introduction Proof systems are built to prove statements. They can be thought as an inference machine with special statements, called provable statements, or sometimes

More information

English Grammar Passive Voice and Other Items

English Grammar Passive Voice and Other Items English Grammar Passive Voice and Other Items In this unit we will finish our look at English grammar. Please be aware that you will have only covered the essential basic grammar that is commonly taught

More information

Early Morphological Development

Early Morphological Development Early Morphological Development Morphology is the aspect of language concerned with the rules governing change in word meaning. Morphological development is analyzed by computing a child s Mean Length

More information

Syntax-driven bindings of Spanish clitic pronouns

Syntax-driven bindings of Spanish clitic pronouns Procesamiento del Lenguaje Natural, núm. 35 (2005), pp. 253-257 recibido 30-04-2005; aceptado 0-06-2005 Syntax-driven bindings of Spanish clitic pronouns Ivan V. Meza Ruiz ICCS University of Edinburgh

More information

Syntactic Theory. Background and Transformational Grammar. Dr. Dan Flickinger & PD Dr. Valia Kordoni

Syntactic Theory. Background and Transformational Grammar. Dr. Dan Flickinger & PD Dr. Valia Kordoni Syntactic Theory Background and Transformational Grammar Dr. Dan Flickinger & PD Dr. Valia Kordoni Department of Computational Linguistics Saarland University October 28, 2011 Early work on grammar There

More information

CS4025: Pragmatics. Resolving referring Expressions Interpreting intention in dialogue Conversational Implicature

CS4025: Pragmatics. Resolving referring Expressions Interpreting intention in dialogue Conversational Implicature CS4025: Pragmatics Resolving referring Expressions Interpreting intention in dialogue Conversational Implicature For more info: J&M, chap 18,19 in 1 st ed; 21,24 in 2 nd Computing Science, University of

More information

INF5820 Natural Language Processing - NLP. H2009 Jan Tore Lønning [email protected]

INF5820 Natural Language Processing - NLP. H2009 Jan Tore Lønning jtl@ifi.uio.no INF5820 Natural Language Processing - NLP H2009 Jan Tore Lønning [email protected] Semantic Role Labeling INF5830 Lecture 13 Nov 4, 2009 Today Some words about semantics Thematic/semantic roles PropBank &

More information

COMPARATIVES WITHOUT DEGREES: A NEW APPROACH. FRIEDERIKE MOLTMANN IHPST, Paris [email protected]

COMPARATIVES WITHOUT DEGREES: A NEW APPROACH. FRIEDERIKE MOLTMANN IHPST, Paris fmoltmann@univ-paris1.fr COMPARATIVES WITHOUT DEGREES: A NEW APPROACH FRIEDERIKE MOLTMANN IHPST, Paris [email protected] It has become common to analyse comparatives by using degrees, so that John is happier than Mary would

More information

MODERN WRITTEN ARABIC. Volume I. Hosted for free on livelingua.com

MODERN WRITTEN ARABIC. Volume I. Hosted for free on livelingua.com MODERN WRITTEN ARABIC Volume I Hosted for free on livelingua.com TABLE OF CcmmTs PREFACE. Page iii INTRODUCTICN vi Lesson 1 1 2.6 3 14 4 5 6 22 30.45 7 55 8 61 9 69 10 11 12 13 96 107 118 14 134 15 16

More information

PTE Academic Preparation Course Outline

PTE Academic Preparation Course Outline PTE Academic Preparation Course Outline August 2011 V2 Pearson Education Ltd 2011. No part of this publication may be reproduced without the prior permission of Pearson Education Ltd. Introduction The

More information

Linguistic Universals

Linguistic Universals Armin W. Buch 1 2012/11/28 1 Relying heavily on material by Gerhard Jäger and David Erschler Linguistic Properties shared by all languages Trivial: all languages have consonants and vowels More interesting:

More information

Double Genitives in English

Double Genitives in English Karlos Arregui-Urbina Department Linguistics and Philosophy MIT 1. Introduction Double Genitives in English MIT, 29 January 1998 Double genitives are postnominal genitive phrases which are marked with

More information

Methodological Issues for Interdisciplinary Research

Methodological Issues for Interdisciplinary Research J. T. M. Miller, Department of Philosophy, University of Durham 1 Methodological Issues for Interdisciplinary Research Much of the apparent difficulty of interdisciplinary research stems from the nature

More information

Slot Grammars. Michael C. McCord. Computer Science Department University of Kentucky Lexington, Kentucky 40506

Slot Grammars. Michael C. McCord. Computer Science Department University of Kentucky Lexington, Kentucky 40506 Michael C. McCord Computer Science Department University of Kentucky Lexington, Kentucky 40506 This paper presents an approach to natural language grammars and parsing in which slots and rules for filling

More information

Overview of the TACITUS Project

Overview of the TACITUS Project Overview of the TACITUS Project Jerry R. Hobbs Artificial Intelligence Center SRI International 1 Aims of the Project The specific aim of the TACITUS project is to develop interpretation processes for

More information

Structure of Clauses. March 9, 2004

Structure of Clauses. March 9, 2004 Structure of Clauses March 9, 2004 Preview Comments on HW 6 Schedule review session Finite and non-finite clauses Constituent structure of clauses Structure of Main Clauses Discuss HW #7 Course Evals Comments

More information

Phrase Structure Rules, Tree Rewriting, and other sources of Recursion Structure within the NP

Phrase Structure Rules, Tree Rewriting, and other sources of Recursion Structure within the NP Introduction to Transformational Grammar, LINGUIST 601 September 14, 2006 Phrase Structure Rules, Tree Rewriting, and other sources of Recursion Structure within the 1 Trees (1) a tree for the brown fox

More information

Quantifier Scope in Formal Linguistics

Quantifier Scope in Formal Linguistics Quantifier Scope in Formal Linguistics E.G. Ruys Utrecht University Yoad Winter Technion/Utrecht University To appear in D. Gabbay (ed.), Handbook of Philosophical Logic Second Edition 1. Introduction

More information

CINTIL-PropBank. CINTIL-PropBank Sub-corpus id Sentences Tokens Domain Sentences for regression atsts 779 5,654 Test

CINTIL-PropBank. CINTIL-PropBank Sub-corpus id Sentences Tokens Domain Sentences for regression atsts 779 5,654 Test CINTIL-PropBank I. Basic Information 1.1. Corpus information The CINTIL-PropBank (Branco et al., 2012) is a set of sentences annotated with their constituency structure and semantic role tags, composed

More information

Structure of the talk. The semantics of event nominalisation. Event nominalisations and verbal arguments 2

Structure of the talk. The semantics of event nominalisation. Event nominalisations and verbal arguments 2 Structure of the talk Sebastian Bücking 1 and Markus Egg 2 1 Universität Tübingen [email protected] 2 Rijksuniversiteit Groningen [email protected] 12 December 2008 two challenges for a

More information

1 Basic concepts. 1.1 What is morphology?

1 Basic concepts. 1.1 What is morphology? EXTRACT 1 Basic concepts It has become a tradition to begin monographs and textbooks on morphology with a tribute to the German poet Johann Wolfgang von Goethe, who invented the term Morphologie in 1790

More information

Pronouns: A case of production-before-comprehension

Pronouns: A case of production-before-comprehension Faculty of Arts University of Groningen Pronouns: A case of production-before-comprehension A study on the the comprehension and production of pronouns and reflexives in Dutch children Petra Hendriks Jennifer

More information

Glossary of literacy terms

Glossary of literacy terms Glossary of literacy terms These terms are used in literacy. You can use them as part of your preparation for the literacy professional skills test. You will not be assessed on definitions of terms during

More information

PUSD High Frequency Word List

PUSD High Frequency Word List PUSD High Frequency Word List For Reading and Spelling Grades K-5 High Frequency or instant words are important because: 1. You can t read a sentence or a paragraph without knowing at least the most common.

More information

Parts of Speech. Skills Team, University of Hull

Parts of Speech. Skills Team, University of Hull Parts of Speech Skills Team, University of Hull Language comes before grammar, which is only an attempt to describe a language. Knowing the grammar of a language does not mean you can speak or write it

More information

Extraposition, the Right Roof Constraint, Result Clauses, Relative Clause Extraposition, and PP Extraposition

Extraposition, the Right Roof Constraint, Result Clauses, Relative Clause Extraposition, and PP Extraposition Extraposition, the Right Roof Constraint, Result Clauses, Relative Clause Extraposition, and PP Extraposition Mark R. Baltin revised version to appear in The Syntax Companion New York University First

More information

10th Grade Language. Goal ISAT% Objective Description (with content limits) Vocabulary Words

10th Grade Language. Goal ISAT% Objective Description (with content limits) Vocabulary Words Standard 3: Writing Process 3.1: Prewrite 58-69% 10.LA.3.1.2 Generate a main idea or thesis appropriate to a type of writing. (753.02.b) Items may include a specified purpose, audience, and writing outline.

More information

Subordinating Ideas Using Phrases It All Started with Sputnik

Subordinating Ideas Using Phrases It All Started with Sputnik NATIONAL MATH + SCIENCE INITIATIVE English Subordinating Ideas Using Phrases It All Started with Sputnik Grade 9-10 OBJECTIVES Students will demonstrate understanding of how different types of phrases

More information

LESSON THIRTEEN STRUCTURAL AMBIGUITY. Structural ambiguity is also referred to as syntactic ambiguity or grammatical ambiguity.

LESSON THIRTEEN STRUCTURAL AMBIGUITY. Structural ambiguity is also referred to as syntactic ambiguity or grammatical ambiguity. LESSON THIRTEEN STRUCTURAL AMBIGUITY Structural ambiguity is also referred to as syntactic ambiguity or grammatical ambiguity. Structural or syntactic ambiguity, occurs when a phrase, clause or sentence

More information

Predicate Logic Review

Predicate Logic Review Predicate Logic Review UC Berkeley, Philosophy 142, Spring 2016 John MacFarlane 1 Grammar A term is an individual constant or a variable. An individual constant is a lowercase letter from the beginning

More information

Written Language Curriculum Planning Manual 3LIT3390

Written Language Curriculum Planning Manual 3LIT3390 Written Language Curriculum Planning Manual 3LIT3390 TABLE OF CONTENTS Language Usage Curriculum... 1 Language Usage I... 2 Language Usage II... 4 Language Usage III... 6 Language Usage IV... 8 Language

More information

Morphemes, roots and affixes. 28 October 2011

Morphemes, roots and affixes. 28 October 2011 Morphemes, roots and affixes 28 October 2011 Previously said We think of words as being the most basic, the most fundamental, units through which meaning is represented in language. Words are the smallest

More information

Chapter 13, Sections 13.1-13.2. Auxiliary Verbs. 2003 CSLI Publications

Chapter 13, Sections 13.1-13.2. Auxiliary Verbs. 2003 CSLI Publications Chapter 13, Sections 13.1-13.2 Auxiliary Verbs What Auxiliaries Are Sometimes called helping verbs, auxiliaries are little words that come before the main verb of a sentence, including forms of be, have,

More information

Word order in Lexical-Functional Grammar Topics M. Kaplan and Mary Dalrymple Ronald Xerox PARC August 1995 Kaplan and Dalrymple, ESSLLI 95, Barcelona 1 phrase structure rules work well Standard congurational

More information

Cambridge Primary English as a Second Language Curriculum Framework

Cambridge Primary English as a Second Language Curriculum Framework Cambridge Primary English as a Second Language Curriculum Framework Contents Introduction Stage 1...2 Stage 2...5 Stage 3...8 Stage 4... 11 Stage 5...14 Stage 6... 17 Welcome to the Cambridge Primary English

More information

The parts of speech: the basic labels

The parts of speech: the basic labels CHAPTER 1 The parts of speech: the basic labels The Western traditional parts of speech began with the works of the Greeks and then the Romans. The Greek tradition culminated in the first century B.C.

More information

English Language Proficiency Standards: At A Glance February 19, 2014

English Language Proficiency Standards: At A Glance February 19, 2014 English Language Proficiency Standards: At A Glance February 19, 2014 These English Language Proficiency (ELP) Standards were collaboratively developed with CCSSO, West Ed, Stanford University Understanding

More information

Course Description (MA Degree)

Course Description (MA Degree) Course Description (MA Degree) Eng. 508 Semantics (3 Credit hrs.) This course is an introduction to the issues of meaning and logical interpretation in natural language. The first part of the course concentrates

More information

Basic Concepts in Research and Data Analysis

Basic Concepts in Research and Data Analysis Basic Concepts in Research and Data Analysis Introduction: A Common Language for Researchers...2 Steps to Follow When Conducting Research...3 The Research Question... 3 The Hypothesis... 4 Defining the

More information

Concepts and terminology in the Simula Programming Language

Concepts and terminology in the Simula Programming Language Concepts and terminology in the Simula Programming Language An introduction for new readers of Simula literature Stein Krogdahl Department of Informatics University of Oslo, Norway April 2010 Introduction

More information

Lecture 16 : Relations and Functions DRAFT

Lecture 16 : Relations and Functions DRAFT CS/Math 240: Introduction to Discrete Mathematics 3/29/2011 Lecture 16 : Relations and Functions Instructor: Dieter van Melkebeek Scribe: Dalibor Zelený DRAFT In Lecture 3, we described a correspondence

More information

Comparative Analysis on the Armenian and Korean Languages

Comparative Analysis on the Armenian and Korean Languages Comparative Analysis on the Armenian and Korean Languages Syuzanna Mejlumyan Yerevan State Linguistic University Abstract It has been five years since the Korean language has been taught at Yerevan State

More information

Writing Common Core KEY WORDS

Writing Common Core KEY WORDS Writing Common Core KEY WORDS An educator's guide to words frequently used in the Common Core State Standards, organized by grade level in order to show the progression of writing Common Core vocabulary

More information

Building a Question Classifier for a TREC-Style Question Answering System

Building a Question Classifier for a TREC-Style Question Answering System Building a Question Classifier for a TREC-Style Question Answering System Richard May & Ari Steinberg Topic: Question Classification We define Question Classification (QC) here to be the task that, given

More information

One Ring to rule them all, One Ring to find them One Ring to bring them all, and in the darkness bind them In the Land of Mordor where shadows lie

One Ring to rule them all, One Ring to find them One Ring to bring them all, and in the darkness bind them In the Land of Mordor where shadows lie One Ring to rule them all, One Ring to find them One Ring to bring them all, and in the darkness bind them In the Land of Mordor where shadows lie OR, THE RING POEM IN OLD CHURCH SLAVIC by Jussi Halla-aho

More information

Culture and Language. What We Say Influences What We Think, What We Feel and What We Believe

Culture and Language. What We Say Influences What We Think, What We Feel and What We Believe Culture and Language What We Say Influences What We Think, What We Feel and What We Believe Unique Human Ability Ability to create and use language is the most distinctive feature of humans Humans learn

More information

2. Argument Structure & Standardization

2. Argument Structure & Standardization 2. Argument Structure & Standardization 1 Some Review So, we have been looking at arguments: What is and is not an argument. The main parts of an argument. How to identify one when you see it. In the exercises

More information

1 Uncertainty and Preferences

1 Uncertainty and Preferences In this chapter, we present the theory of consumer preferences on risky outcomes. The theory is then applied to study the demand for insurance. Consider the following story. John wants to mail a package

More information

THE ENGLISH IT-CLEFT CONSTRUCTION:

THE ENGLISH IT-CLEFT CONSTRUCTION: i THE ENGLISH IT-CLEFT CONSTRUCTION: A ROLE AND REFERENCE GRAMMAR ANALYSIS EMMA LOUISE PAVEY THESIS SUBMITTED FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF SUSSEX MARCH 2004 I hereby declare that

More information

A terminology model approach for defining and managing statistical metadata

A terminology model approach for defining and managing statistical metadata A terminology model approach for defining and managing statistical metadata Comments to : R. Karge (49) 30-6576 2791 mail [email protected] Content 1 Introduction... 4 2 Knowledge presentation...

More information

Language Meaning and Use

Language Meaning and Use Language Meaning and Use Raymond Hickey, English Linguistics Website: www.uni-due.de/ele Types of meaning There are four recognisable types of meaning: lexical meaning, grammatical meaning, sentence meaning

More information

Comprendium Translator System Overview

Comprendium Translator System Overview Comprendium System Overview May 2004 Table of Contents 1. INTRODUCTION...3 2. WHAT IS MACHINE TRANSLATION?...3 3. THE COMPRENDIUM MACHINE TRANSLATION TECHNOLOGY...4 3.1 THE BEST MT TECHNOLOGY IN THE MARKET...4

More information

Sentence Structure/Sentence Types HANDOUT

Sentence Structure/Sentence Types HANDOUT Sentence Structure/Sentence Types HANDOUT This handout is designed to give you a very brief (and, of necessity, incomplete) overview of the different types of sentence structure and how the elements of

More information