Internally Headed Relative Clauses in Austronesian Languages

LANGUAGE AND LINGUISTICS 5.1:99-129, 2004 2004-0-005-001-000016-1 Internally Headed Relative Clauses in Austronesian Languages Edith Aldridge SUNY, Stony Brook It is well known that many Austronesian languages have head-initial and head-final relative clauses. This paper reports that the Atayalic language Seediq and the Philippine language Tagalog additionally have internally headed relative clauses. In this paper, I first identify the formal and structural differences among the three types of relative clause in these languages. In particular, I distinguish the head-final from the head-internal type by showing that the former exhibit evidence of remnant TP movement to [Spec, DP] from a [D CP] structure, as per Kayne’s (1994) proposal for head-final relative clause derivation. The head-internal variety, on the other hand, do not show evidence of movement and therefore qualify as internally headed relative clauses. For their derivation, I propose an analysis similar to Basilico (1996), in which the head nominal moves to a clause-internal position where it can be bound by an operator in [Spec, CP]. Support for this quantificational analysis of Seediq and Tagalog internally headed relative clauses is provided by the fact that parallel structures are employed in these languages for other constructions involving binding from external operators, including wh-in situ and clauses containing weakly quantified arguments. Key words: internally headed relative clauses, Austronesian, antisymmetry, remnant movement 1. Introduction It is frequently reported in the descriptive and theoretical literature that Austronesian languages have head-initial relative clauses (Dixon 1988, Georgopolous 1991, Josephs 1975, Lee 1975, Sneddon 1996, Sohn 1975, Topping 1973, Finer 1998). Some Austronesian languages, particularly Philippine and Formosan, have also been cited as having head-final (typically in addition to head-initial) relative clauses (Ramos 1971, Schachter & Otanes 1972, Reid & Liao 2001, Chung 1998, H.-C. Chang 2000, Y.-L. Chang 2000a, b, M.-C. Huang 2000a, b, Wu 2000, Yeh 2000, Zeitoun 2000a, b, c, Li 2000). Donohue (1999) and Tang et al. (1998) discuss relative clauses in Tukang Besi and Paiwan, respectively, where the head can appear to the left or right of the clause. Both suggest that the former could be analyzed as head-internal relatives. However, reporting of unambiguously head-internal relative clauses in Austronesian languages is extremely Edith Aldridge 100 rare. Internally headed relative clauses have been reported to exist in Seediq (Chang 2000a), Puyuma (M.-C. Huang 2000b), and Riau Indonesian (Gil 2000). However, to my knowledge, no analysis has yet been proposed for their structure. The purpose of this paper is to propose an analysis of relative clause, especially internally headed relative clause, structure in the Austronesian languages Tagalog (Philippines) and Seediq (Atayalic, Taiwan). These languages have head-initial (1, 2), head-final (3, 4), and head-internal (5, 6) relative clauses. Numbers (5) and (6) are unambiguously internally headed relative clauses, since the head nominal is positioned inside the clause, between the verb and the agent. (1) S: sapah s-n-malu na tama house -Perf-build Erg father ‘the house Father built’ (2) T: libro-ng b-in-ili ni Maria book-Lk -Perf-buy Erg Maria ‘the book Maria bought’ (3) S: s-n-malu na tama sapah -Perf-build Erg father house ‘the house Father built’ (4) T: b-in-ili ni Maria-ng libro -Perf-buy Erg Maria-Lk book ‘the book Maria bought’ (5) S: s-n-malu sapah na tama -Perf-build house Erg father ‘the house Father built’ (6) T: b-in-ili-ng libro ni Maria -Perf-buy-Lk book Erg Maria ‘the book Maria bought’ Basic word order in Tagalog and Seediq, as is the case in most Philippine and Formosan languages, is verb-initial. In Seediq, the absolutive 1 nominal has a fixed position in the clause, always appearing clause-finally, yielding strict VOS word order. Number (7) is an antipassive; the agent absolutive appears in clause-final position, to 1 By those who take these languages to be accusative, the grammatical role “absolutive” is generally referred to as “subject”. Following my earlier work (Aldridge 1999, 2001, 2002), I treat Tagalog and Seediq as ergative languages. Earlier ergative analyses of Austronesian languages include De Guzman (1988), Gertds (1988), and Payne (1982). Internally Headed Relative Clauses in Austronesian Languages 101 the right of the theme. The absolutive in (8) is the theme, which also appears in clause- final position, following the agent. (7) S: Gaga m-ekan ido ka Pawan. Pres Intr-eat rice Top2 Pawan ‘Pawan is eating rice.’ (8) S: Wada burig-un na Ape ka patis-ni. Past buy-Tr Erg Ape Top book-this ‘Ape bought this book.’ Unmarked word order in Tagalog is VSO, following the thematic hierarchy: Verb- Agent-Theme-Goal. The absolutive nominal has no fixed position, appearing most naturally in its base position. Hence, the agent absolutive in (9) appears to the right of the theme, while the theme absolutive in (10) appears between the agent and the goal. (9) T: B-um-ili si Maria ng libro. -Intr.Perf-buy Abs Maria Obl book ‘Maria bought a book.’ (10) T: I-b-in-igay ni Maria ang libro kay Pedro. App-Perf-give Erg Maria Abs book P Pedro ‘Maria gave the book to Pedro.’ The fact that verb-initial languages like Seediq and Tagalog should have internally headed relative clauses comes as a surprise in light of the assertions of Downing (1978), Keenan (1985), Cole (1987), and others that internally headed relative clauses are found only in verb-final languages. However, I shall show in this paper that internally headed relative clauses do in fact exist in Tagalog and Seediq and that they are licensed in accordance with other syntactic characteristics of these languages. 2. Differences between internally and externally-headed types Before entering the structural analysis of these different types of relative clause, it is first necessary to establish that these types have distinct properties. The head-initial type shown in (1) and (2) should be uncontroversial, but there is need to clarify the 2 Ka is glossed as a topic marker and not as an absolutive marker, since it can precede ergative topics as well as absolutives. In the analysis I propose in section 5, absolutive DPs obligatorily undergo movement to a topic position in the derivation of a declarative clause. Edith Aldridge 102 distinction between the head-internal and head-final types. Descriptively speaking, head-internal relative clauses appear with the head in immediate post-verbal position, as shown in (5) and (6). Those relative clauses where the head follows the verb, but is not immediately adjacent to it, I refer to as head-final relative clauses. This should be intuitively obvious in the case of (3) and (4). However, examples like the following, where clause-internal material follows the head, also pattern structurally with (3) and (4) and not with the internally headed type. (11) S: b-n-ari-na chiiga bulebun ka Ape -Perf-buy-3sErg yesterday banana Top Ape ‘the banana(s) that Ape bought yesterday’ (12) T: i-b-in-igay ng babae-ng kendi sa bata App-Perf-give Erg woman-Lk candy P child ‘the candy the woman gave to the child’ First indication that head-final relatives differ structurally from the internally headed type is provided by the fact that the latter exhibit the definiteness effect cited by Williamson (1987), Culy (1990), and Basilico (1996). It is more natural for heads in final position to be definite, indicating that these heads are located in a position external to the clause. (13) S: k-n-ta-an-na Awe-ni seediq kiya -Perf-see-App-3sErg Awe-Def person that ‘that person whom Awe saw’ (14) S: * k-n-ta-an seediq kiya na Awe-ni -Perf-see-App person that Erg Awe-Def ‘that person whom Awe saw’ Further evidence for the external position of the head in head-final relatives is offered by the position of quantifiers. Quantifiers in head-initial and head-internal relatives precede the entire construction. (15) T: tatlo-ng mangga-ng b-in-ili ni Maria three-Lk mango-Lk -Perf-buy Erg Maria ‘three mangoes that Maria bought’ (16) T: tatlo-ng b-in-ili-ng mangga ni Maria three-Lk -Perf-buy-Lk mango Erg Maria ‘three mangoes that Maria bought’ Internally Headed Relative Clauses in Austronesian Languages 103 It is awkward for the quantifier to appear in this position in head-final relatives. It is much more natural for it to immediately precede the head. (17) T:?* tatlo-ng b-in-ili ni Maria-ng mangga three-Lk -Perf-buy Erg Maria-Lk mango ‘three mangoes that Maria bought’ (18) T: b-in-ili ni Maria-ng tatlo-ng mangga -Perf-buy Erg Maria-Lk three-Lk mango ‘three mangoes that Maria bought’ In contrast to this, it is impossible for the quantifier to appear with the internal head, inside the clause. (19) T: * b-in-ili-ng tatlo-ng mangga ni Maria -Perf-buy-Lk three-Lk mango Erg Maria ‘three mangoes that Maria bought’ Head-final relatives involving stranding exhibit the same pattern. The quantifier can appear adjacent to the head, but not preceding the clause. (20) T: i-b-in-igay ng babae-ng tatlo-ng mangga sa bata App-Perf-give Erg woman-Lk three-Lk mango P child ‘three mangoes which the woman gave to the child’ (21) T:?* tatlo-ng i-b-in-igay ng babae-ng mangga sa bata three-Lk App-Perf-give Erg woman-Lk mango P child ‘three mangoes which the woman gave to the child’ In addition, there are strict constraints on what can follow the head in a head-final relative. As seen above, a PP can be stranded, but not an oblique theme. (22) T: i-b-in-igay ng babae-ng kendi sa bata App-Perf-give Erg woman-Lk candy P child ‘the candy the woman gave to the child’ (23) T:?* b-in-igy-an ng babae-ng bata ng kendi -Perf-give-App Erg woman-Lk child Obl candy ‘the child to whom the woman gave candy’ On the other hand, there is no such restriction in the case of head-internal relatives. Edith Aldridge 104 The head in these cases can be followed by a PP, an ergative DP, an oblique DP, or any combination of these. (24) T: i-b-in-igay na kendi ng babae sa bata App-Perf-give Lk candy Erg woman P child ‘the candy the woman gave to the child’ (25) T: nag-bigay na tao ng kendi sa bata Perf.Intr-give Lk person Obl candy P child ‘the person who gave candy to the child’ (26) T: b-in-igy-an na bata ng babae ng kendi -Perf-give-App Lk child Erg woman Obl candy ‘the child to whom the woman gave candy’ This section has shown that head-final relatives have different properties from the internally headed variety. The next two sections will develop analyses for their derivations. I shall show first that a movement analysis is most appropriate for externally headed relatives. Following that, in section 4, I shall demonstrate that such an analysis is impossible for the internally headed type and show that these require an analysis in which the head nominal does not move to a clause-external position. I present this analysis in section 5. 3. Movement analysis for externally-headed types This section presents an analysis of externally headed relative clauses based on movement of the head nominal to a position outside the clause. The derivation I assume for externally headed relative clauses is that of Kayne (1994). To derive head-initial relative clauses, the head nominal simply moves from its base position within the clause into [Spec, CP]. (27) T: libro-ng b-in-ili ni Maria book-Lk -Perf-buy Erg Maria ‘the book Maria bought’ (28) DP CP booki TP Maria bought ti Internally Headed Relative Clauses in Austronesian Languages 105 For head-final relative clauses, after the head moves to [Spec, CP], the remnant clause further fronts to [Spec, DP]. (29) T: b-in-ili ni Maria-ng libro -Perf-buy Erg Maria-Lk book ‘the book Maria bought’ (30) DP TP CP book tTP The facts introduced in the previous section can be captured straightforwardly in this analysis. Recall that that a quantifier must appear immediately before the relative head and not to the left of the entire clause. (31) T: b-in-ili ni Maria-ng tatlo-ng mangga -Perf-buy Erg Maria-Lk three-Lk mango ‘three mangoes that Maria bought’ (32) T:?* tatlo-ng b-in-ili ni Maria-ng mangga three-Lk -Perf-buy Erg Maria-Lk mango ‘three mangoes that Maria bought’ This is not surprising under the movement analysis. The quantified relative clause in (31) would be derived as follows. The head nominal moves to [Spec, CP] of the clause, below the position of the quantifier. The remnant TP of the clause fronts to [Spec, DP]. This derives the word order in which the clause precedes both the quantifier and relative head. (33) DP TP QP 3 CP mango tTP Turning to head-final relatives involving stranding of clause-internal material, Edith Aldridge 106 these examples argue most strongly for separate analyses of the head-internal and head- external types of relative clause. Specifically, Seediq relative clauses allow a topicalized agent to follow the relative head. (34) S: b-n-ari-na chiiga bulebun ka Ape -Perf-buy-3sErg yesterday banana Top Ape ‘the banana(s) that Ape bought yesterday’ Tagalog relative clauses allow a scrambled PP to follow the head. (35) T: i-b-in-igay ng babae-ng kendi sa bata App-Perf-give Erg woman-Lk candy P child ‘the candy the woman gave to the child’ The stranded word orders in (34) and (35) can be accounted for straightforwardly under Kayne’s (1994) [D CP] analysis. For Seediq, topicalization takes place first, moving the agent into clause-initial topic position. The relative head then moves into [Spec, CP]. Finally, the remnant TP fronts to [Spec, DP]. (36) DP TP CP banana TopP Ape tTP Tagalog PP stranding is handled in the same way. The PP in Tagalog first scrambles to clause-initial position. Then the relative head moves up to [Spec, CP]. Finally, the remnant TP fronts to [Spec, DP]. (37) DP TP CP candy FocP PP tTP Internally Headed Relative Clauses in Austronesian Languages 107 This analysis predicts that stranding in relative clauses should be possible when dislocation of the stranded XP is allowed and impossible otherwise. This prediction is indeed borne out. A'-movement in Tagalog is highly constrained, as it is in a great number of Austronesian languages (Nakamura 1994, Pensalfini 1995, Chung 1998, among many others). Only absolutives are eligible to undergo relativization, topicalization, clefting, and wh-question formation. Hence, a relative clause formed on the theme of a transitive clause in (38) is grammatical, but (39), formed on the agent of the same transitive clause is not. (38) T: libro-ng b-in-ili ni Maria book-Lk -Perf-buy Erg Maria ‘the book Maria bought’ (39) T: * tao-ng b-in-ili ang libro person-Lk -Perf-buy Abs book ‘the person who bought the book’ Aside from this restriction, PPs are allowed to move to preverbal position in Tagalog, where they typically receive a focus interpretation. (40) T: I-b-in-igay ng babae ang kendi sa bata. App-Perf-give Erg woman Abs candy P child ‘The woman gave candy to the child.’ (41) T: Sa bata i-b-in-igay ng babae ang kendi. P child App-Perf-give Erg woman Abs candy ‘The woman gave the candy to the child.’ Other non-absolutive arguments, in contrast, are not able to undergo scrambling. In the ditransitive in (42), the goal is licensed as absolutive of the clause by the applicative affix on the verb. The theme is demoted to oblique status. This non- absolutive theme cannot be scrambled, as shown in (43). (42) T: B-in-igy-an ng babae ng kendi ang bata. -Perf-give-App Erg woman Obl candy Abs child ‘The woman gave the child candy.’ (43) T: * Ng kendi b-in-igy-an ng babae ang bata. Obl candy -Perf-give-App Erg woman Abs child The same pattern can be observed in stranding in relative clauses. As seen above, Edith Aldridge 108 only the absolutive can be the head NP. But a PP can also be stranded to the right of the head. This is not possible for an oblique object. The ungrammaticality can be explained by the inability of the oblique object to scramble before remnant TP fronting. This straightforwardly accounts for the contrast between (44) and (45) first observed in section 2. (44) T: i-b-in-igay ng babae-ng kendi sa bata App-Perf-give Erg woman-Lk candy P child ‘the candy the woman gave to the child’ (45) T:?* b-in-igy-an ng babae-ng bata ng kendi -Perf-give-App Erg woman-Lk child Obl candy ‘the child to whom the woman gave candy’ Interestingly, head-initial relative clauses formed on goal absolutives with oblique objects in situ are perfectly grammatical. The head initial version of (45) is shown in (46), where the oblique object appears in situ inside the clause. This is completely consistent with the analysis proposed in (28). The relative head moves to [Spec, CP], but no other dislocation need take place. Therefore, the oblique nominal can remain in its base position. (46) T: bata-ng [b-in-igy-an ng babae ng kendi] child-Lk -Perf-give-App Erg woman Obl candy ‘the child to whom the woman gave candy’ The Seediq case is parallel to Tagalog. Seediq also exhibits the absolutive restriction on A' extraction. (47) S: sapah s-n-malu na tama house -Perf-build Erg father ‘the house Father built’ (48) S: * seediq s-n-malu ka sapah person -Perf-build Top house ‘the person who built the house’ In addition to this, Seediq declarative clauses can have a topicalized agent appearing in clause-initial position and resumed by a pronoun in the clause. Number (49) shows the ergative agent in situ. Number (50) is the topicalized version. Internally Headed Relative Clauses in Austronesian Languages 109 (49) S: Wada bube-un na Pawan ka dangi-na. Past hit-Tr Erg Pawan Top friend-3sPoss ‘Pawan hit his friend.’ (50) S: Pawan-ni wada-na bube-un ka dangi-na. Pawan-Def Past-3sErg hit-Tr Top friend-3sPoss ‘Pawan hit his friend.’ As in Tagalog, oblique objects cannot be dislocated in this way. (51)