1 Nominalization Source of Ergativity in Tagalog 1 Edith Aldridge, University of Washington 1. Introduction This paper proposes a diachronic source for the syncretism between ergative and genitive case observed in a large number of Austronesian languages. As illustrated by the Tagalog examples below, the subject of a transitive clause in (1a) takes the same ng (pronounced ‘nang’) marker as the possessor in (1b). Tagalog (1) a. B<in>ili ng babae ang isda. <TR.PRV>buy ERG woman ABS fish ‘The woman bought the fish.’ b. isda ng babae fish GEN woman ‘(the) woman’s fish’ Syncretism between ergative and genitive case is also found in a large number of other languages with ergative alignment. A connection – either synchronic or diachronic – between ergativity and clausal nominalization has been proposed for some of these (Bricker 1981 for Mayan; Gildea 1998 for Cariban; Starosta et al. 1982, Kaufman 2007, Ross 2009 for Austronesian; Johns 1992 for Inuit; among others). In this paper, I take a diachronic perspective 1 This paper is write-up of a presentation made at the 14 th Diachronic Generative Syntax conference at the University of Lisbon in 2012. I would like to thank the conference organizers for the opportunity to present the paper. Let me also thank Ian Roberts, Theresa Biberauer, and Tony Kroch for their questions and comments after the presentation.
38
Embed
Nominalization Source of Ergativity in Tagalogfaculty.washington.edu/aldr/pdf/DIGS-14.pdf · Nominalization Source of Ergativity in Tagalog1 Edith Aldridge, University of Washington
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
1
Nominalization Source of Ergativity in Tagalog1
Edith Aldridge, University of Washington
1. Introduction
This paper proposes a diachronic source for the syncretism between ergative and genitive case
observed in a large number of Austronesian languages. As illustrated by the Tagalog examples
below, the subject of a transitive clause in (1a) takes the same ng (pronounced ‘nang’) marker as
the possessor in (1b).
Tagalog
(1) a. B<in>ili ng babae ang isda.
<TR.PRV>buy ERG woman ABS fish
‘The woman bought the fish.’
b. isda ng babae
fish GEN woman
‘(the) woman’s fish’
Syncretism between ergative and genitive case is also found in a large number of other
languages with ergative alignment. A connection – either synchronic or diachronic – between
ergativity and clausal nominalization has been proposed for some of these (Bricker 1981 for
Mayan; Gildea 1998 for Cariban; Starosta et al. 1982, Kaufman 2007, Ross 2009 for
Austronesian; Johns 1992 for Inuit; among others). In this paper, I take a diachronic perspective
1 This paper is write-up of a presentation made at the 14th Diachronic Generative Syntax conference at the University of Lisbon in 2012. I would like to thank the conference organizers for the opportunity to present the paper. Let me also thank Ian Roberts, Theresa Biberauer, and Tony Kroch for their questions and comments after the presentation.
2
and propose that ergative alignment in Austronesian languages resulted from the reanalysis of a
reduced clausal nominalization nP as a verbal projection vP. Crucially, this reanalysis involved
only a category change and did not alter constituency or hierarchical relations within the nP/vP.
2. Synchronic Nominalization?
One main component of the proposal I make in this paper is that the input to the reanalysis was a
reduced clausal nominalization nP. In this section, I provide indirect evidence for this claim by
showing the inadequacies of an approach to modern Tagalog clause structure involving a full
relative clause structure.
2.1. Kaufman’s (2009) Analysis of Tagalog
Kaufman (2009) proposes that Tagalog lacks a v functional category, forcing lexical roots to
merge with n and project a nominal predicate. The external argument of a transitive clause is
treated as a possessor merged in [Spec, n] and assigned genitive case. A DP layer is projected
above this which is selected by a Pred functional head. Pred also selects a null operator in its
specifier which identifies the gap position in the predicate. This missing argument corresponds to
the participant in the event identified by morphology on the verb. This will be the direct object in
a transitive clause like (1a). The operator is then coindexed with the predicate-external subject. T
is treated as a null copula; its specifier houses PredP, while the subject is selected as its
complement.
3
(2) TP (Kaufman 2009) PredP T’ OPi Pred’ T DPi Pred DP ang isda ABS fish binili ng babae bought GEN woman
This proposal accounts for some significant facts of Tagalog syntax. First, it correctly
predicts coordination on an absolutive pivot. The coordinated clauses in (3) can be analyzed as
PredPs, which exclude the absolutive.
(3) [Hu-hugas-an=ko] at [pu-punas-an=mo] ang mga pinggan.
RED-wash-APPL=1S.ERG and RED-dry-APPL=2S.ERG ABS PL dish
‘I will wash and you dry the dishes.’
(2) also predicts the non-extractability of all constituents other than the absolutive, since
these are all contained within the headless relative clause, which is an island to extraction. (4b)
shows that a relative clause can be formed on the absolutive object in a transitive clause.
However, the ergative subject cannot be extracted in this way, as shown in (4c). The absolutive
restriction on A’-extraction is one of the hallmark characteristics of syntactic ergativity (Aldridge
2004, 2008b; Campana 1992; Dixon 1994; Manning 1996; among many others).
4
(4) a. B<in>ili ng babae ang isda.
<TR.PRV>buy ERG woman ABS fish
‘The woman bought the fish.’
b. isda-ng b<in>ili ng babae
fish-LK <TR.PRV>buy ERG woman
‘fish that the woman bought’
c. *babae-ng b<in>ili ang isda
woman-ng <TR.PRV>buy ABS fish
‘woman who bought the fish’
This does not mean, however, that external arguments can never undergo A’-movement. The
external argument can be extracted from an antipassive. An antipassive is a clause which is
semantically transitive, in the sense that it has two nominal arguments. But it is syntactically
intransitive: intransitive morphology appears on the verb; the external argument has absolutive
case; and the direct object has inherent case instead of absolutive. This is shown in (5a). Note
that the inherent case on the direct object is genitive. (5b) shows that the external argument
absolutive in an antipassive can be the head of a relative clause. The genitive object, on the other
hand, is no longer eligible to undergo this movement. Kaufman’s analysis accounts for this, since
the genitive object will be contained within the predicate island.
(5) a. B<um>ili ang babae ng isda.
<INTR.PRV>buy ABS woman GEN fish
‘The woman bought a fish.’
5
b. babae-ng b<um>ili ng isda
woman-LK <INTR.PRV>buy GEN fish
‘woman who bought a/the fish’
c. *isda-ng b<um>ili ang babae
fish-LK <INTR.PRV>buy ABS woman
‘fish that the woman bought’
2.2. Arguments against Kaufman (2009)
The analysis in (2) also makes a number of incorrect predications for Tagalog syntax. For
example, (2) predicts that the absolutive DP always surfaces in clause-final position, but in fact
absolutives can be followed by a wide variety of material. In (6), we see the absoltuive preceding
a genitive object (6a) and a goal PP (6b). Since these constituents are arguments of the verb, I
assume they would be base generated in the PredP on Kaufman’s analysis. Consequently, their
dislocation to clause-final position should invoke an island violation along the lines of (4c) and
(5c) above, contrary to fact.
(6) a. [PredP B<um>ili tDP ] ang babae ng bahay.
<INTR.PRV>buy ABS woman GEN woman
‘The woman bought a house.’
b. [PredP I-bi-bigay=ko tPP ] ang bulaklak kay Maria.
APPL-RED-give=1S.ERG ABS flower to Maria
‘I will give the flowers to Maria.’
6
A second problem regards the licensing of the absolutive DP. In (2), it appears that the only
functional head available to license absolutive case is T, which predicts that absolutives should
not surface in nonfinite environments. (7) shows, however, that absolutive internal arguments are
in fact licensed in nonfinite transitive clauses.
(7) B<in>a-balak ng babae-ng [PRO tulung-an ang lalaki]
TR.RED-plan ERG woman-LK help-APPL ABS man
‘The woman is planning to help the man.’
To summarize this section, I have shown how the nominalist approach to modern Tagalog
accounts for some facts but makes false predictions in other cases. In the next section, I propose
an alternative analysis of modern Tagalog and shows how this analysis accounts for all of the
facts observed in section 2.
3. Analysis of Modern Tagalog
Aldridge (2004, 2008b, 2012b) develops an analysis of Tagalog as a v-type ergative language. In
this type of ergative language, the locus of the syntactic derivation is feature bundles on v. As
can be seen in (8), transitive v has a full complement of case and EPP features. This accounts for
licensing of both the subject and object, as well as the extractability of internal arguments. In
contrast to this, intransitive v lacks these features, with the consequence that T will license the
highest argument in vP, and internal arguments will not be able to escape the VP.
7
(8) v-Type Ergativity
vTr: Inherent ergative case
[uCase:Abs]
[EPP]
vIntr: No case feature vAP: No case feature
TFin: Optional [uCase:Abs]
No [EPP]
In a transitive clause, v is responsible for licensing both ergative and absolutive case.
Following Legate (2002, 2008), Mahajan (1989), and Woolford (1997, 2006), I assume that
ergative case is inherent, assigned by transitive v to its specifier. Transitive v also values
absolutive case on the direct object in VP. Since absolutive case is valued by v in a transitive
clause, the availability of absolutive case for an internal argument in a nonfinite clause is not
surprising, as seen in (7).
In addition, transitive v carries an EPP feature which draws the absolutive object into its
outer specifier. This places the absolutive in the highest specifier within the vP phase edge,
making it eligible to undergo further movement without violating the Phase Impenetrability
Condition of Chomsky (2000, 2001, 2004, 2005) and thereby deriving the absolutive restriction
on A’-extraction observed in section 2.1. (9) illustrates the derivation of a transitive clause. In
addition to the case-licensing processes described above, (9b) shows head movement of the verb
to an aspectual projection above vP, thereby deriving VSO basic Tagalog word order. Note
further that absolutive objects in transitive clauses receive a specific, typically definite
8
interpretation. This is accounted for on Diesing’s (1992) Mapping Hypothesis, which affords a
presuppositional interpretation to objects which have moved out of VP.
(9) a. B<in>ili ng babae ang isda.
<TR.PRV>buy ERG woman ABS fish
‘The woman bought the/*a fish.’
b. TP T AspP V+v+Asp vP DP[ABS] v’ DP[ERG] v’ tV+v[ABS, EPP] VP tV <DP[ABS]>
In intransitive clauses – both simple intransitives and antipassives – v has no case or EPP
features. The direct object in an antipassive receives inherent case from the lexical verb2, which
is morphologically genitive in Tagalog, as noted in section 2. The lack of an EPP feature on
intransitive v further accounts for the inability of genitive objects to move from the VP, as
observed in (5c).
In an intransitive clause, absolutive case must be valued on the subject by T. (8) shows that
the case feature on T appears only optionally. Other factors conspire, however, to ensure that T
2 Although they outwardly resemble transitive clauses in that they generally contain two or more nominal arguments, antipassives have formal characteristics of intransitive clauses (Baker 1988; Campbell 2000; Cooreman 1994; Davies and Sam-Colop 1990; Dixon 1979, 1994; Dryer 1990; England 1988; Kozinsky et al. 1988; Mithun 2000; Palmer 1994; Siegel 1998; Tsunoda 1988; and others).
9
carries an absolutive case feature when v is intransitive. This is because if T does not have a case
feature in an intransitive clause, then the subject will not be licensed and the derivation will crash.
The derivation will also crash if T does carry a case feature in a transitive clause. I assume that
valued case features on functional heads are uninterpretable and therefore must be checked for
the derivation to converge. Consequently, if T carries a case feature in a transitive clause, where
the case needs of both subject and object are satisfied by v, then the feature on T will go
unchecked and the derivation will again crash. In this way, case licensing and dislocation are
both controlled by the feature bundles on v. This is what it means to say that Tagalog is a v-type
ergative language.3
(10) sketches the derivation of an antipassive. As a type of intransitive, antipassive v does not
have a case feature, and T must value absolutive case. Since the external argument is the first DP
in T’s c-command domain, this is the DP which will receive absolutive case. The object is then
dependent on the lexical verb for inherent oblique case, which is genitive. The object also
receives an indefinite, typically nonspecific, interpretation, which I assume is the result of its
being interpreted within VP at LF, as per Diesing’s (1992) Mapping Hypothesis.
(10) a. B<um>ili ang babae ng isda.
<INTR.PRV>buy ABS woman GEN fish
‘The woman bought a fish.’
3 Legate (2002, 2008) also proposes that both T and v play a role in valuing case on absolutives in certain ergative languages. However, she claims that the case valued by T in the syntax is nominative and that valued by v is accusative, as in accusative languages. Absolutive case is treated as the default lexical insertion for these cases in the Morphological Component. I do not adopt this approach for Austronesian languages. because absolutive and accusative cases are morphologically distinct in certain languages. See Aldridge (2012a) for discussion.
10
b. TP T[ABS] AspP V+v+Asp vP DP[ABS] v’ tV+v VP tV DP[GEN ]
Direct evidence for the proposal that T values absolutive case in intransitive clauses comes
from the fact that absolutives cannot surface in nonfinite clauses.
(11) Nagba-balak si Maria-ng [PRO p-um-unta sa Maynila]
INTR.PROG-plan ABS Maria-LK <INTR>go to Manila
‘Maria is planning to go to Manila.’
Before continuing, let me offer additional evidence for the EPP feature on transitive v.
Absolutive objects in transitive clauses like (12a) take wide scope over ergative DPs, while
oblique objects in antipassives like (12b) scope under the external argument. This asymmetry is
accounted for by the proposal that the internal argument moves over the external argument only
in transitive clauses and not in antipassives. Basilico (2003), Benua (1995), and Bittner (1987,
1994) discuss similar scope related phenomena in other ergative languages.
11
(12) a. B-in-asa [ng lahat ng bata] [ang marami-ng libro]
-PRV.TR-read ERG all GEN child ABS many-LK book
‘All the children read many books.’
MANY > ALL
b. Nag-basa [ang lahat ng bata] [ng marami-ng libro]
INTR.PRV-read ABS all GEN child OBL many-LK book
‘All the children read many books.’
ALL > MANY
Somewhat more indirect evidence comes from reflexive binding. A reflexive pronoun
embedded inside the absolutive object can be bound by the ergative subject, as in (13a). A bare
reflexive can surface as the direct object in an antipassive, as in (13b). But a bare reflexive
cannot function as the absolutive object in a transitive clause, as in (13c). This paradigm is
accounted for on the absolutive raising analysis, since the object in a transitive clause moves to a
position where it c-commands the ergative subject. If this object is a bare reflexive, it then c-
commands its intended antecedent and invokes a violation of Binding Principle C. Since
antipassive v does not have an EPP feature, the object does not raise. A bare reflexive is
therefore felicitous in object position in this clause type.
(13) a. P<in>igil ng lalaki ang [DP sarili=niya].
<TR.PRV>control ERG man ABS self =3S.GEN
‘The man controlled himself.’
12
b. Nag-pigil ang lalaki sa sarili.
INTR.PRV-control ABS man DAT self
‘The man controlled himself.’
c. *P<in>igil ng lalaki ang sarili.
<TR.PRV>control ERG man ABS self
‘The man controlled himself.”
Richards (2000) proposes that absolutives undergo covert raising to [Spec, CP], where they
receive a definite interpretation as topics. Evidence from negative polarity licensing, however,
suggests that the landing site is lower. The ergative subject and absolutive object can both be
NPIs, which can be accounted for if the absolutive is interpreted in the vP phase edge.
(14) a. Hindi=niya t<in>anggap ang anumang mungkahi.
NEG=3S.ERG <TR.PRV>accept ABS any proposal
‘He/she didn’t accept any proposal.’
b. Hindi t<in>anggap ng sinuman ang mungkahi=niya.
NEG <TR.PRV>accept ERG anyone ABS proposal=3S.GEN
‘Noone accepted his/her proposal.’
Although absolutives move to the edge of vP and are interpreted there at LF, I assume that
the trace is spelled out in their base positions at PF. Therefore this movement is generally
“covert”. I follow Pesetsky (2000), Bobaljik (2002), and others in assuming that the movement
takes place in the narrow syntax and that either the head or the tail of the chain can be interpreted
13
at the interfaces. Specifically, the head of the chain is interpreted at LF for Tagalog, while the tail
of the chain is generally spelled out at PF.4 This accounts for the fact that the absolutive DP does
not necessarily appear in clause-peripheral position, as shown above in (6). Although absolutives
move to the edge of vP, they are generally spelled out in their base positions and can therefore be
followed by other VP-internal material like goal arguments and genitive objects.
Naturally, the head of the chain can also be spelled out at PF. This happens when A’-
movement proceeds to the CP layer and accounts for the TP coordination example seen above in
(3). I adopt a proposal by Richards (2001) which posits that movement that would otherwise be
“covert” can become “overt” in the case that further movement takes place.
(15) a. [CP [TP Hu-hugas-an=ko] at [TP pu-punas-an=mo] ang mga pinggan]
RED-wash-APPL=1S.ERG and RED-dry-APPL=2S.ERG ABS PL dish
‘I will wash and you dry the dishes.’
b. CP TP C’ ...<DP[ABS]>... DP[ABS] C’ C[EPP] tTP
Regarding the movement of the remnant TP to a higher specifier of CP, I follow Aldridge
(2004) in the assumption that Tagalog topicalization proceeds along the same lines as derivation
of absolutive-final word order in VOS Austronesian languages like Seediq and Atayal. The
4 See Rackowski (2002) and Rackowski and Richards (2005) for other approaches to Tagalog syntax employing “covert” movement of the absolutive to the vP phase edge.
14
absolutive is attracted to a topic position in the CP layer. The remnant TP is subsequently moved
to a higher CP specifier. See Pearson (2001, 2005) for additional evidence that absolutives are
topics in VOS Austronesian languages.
To summarize sections 2 and 3, I have shown that the copula analysis put forth by Kaufman
(2009) accounts for some facts of Tagalog syntax but still suffers from significant shortcomings.
The alternative proposed in section 3 accounts for all of the facts observed in section 2.
At this point, I return to the main focus of this paper, i.e. the ergative/genitive syncretism.
Kaufman’s nominalist approach offers a straightforward account of this fact, since it assumes
that there are no verbal clauses in Tagalog; the external argument of a transitive verb is assigned
genitive case in the nominal projection embedded inside the PredP. On the other hand, my
alternative approach seems at first blush to treat the syncretism between the two types of inherent
case as an accidental homophony. In the remainder of this paper, I investigate a diachronic
explanation of the syncretism. Specifically, I propose that the type of ergativity observed in
Tagalog and other Austronesian languages has its source in a nominalization. However, I
propose that this nominalization was not a full CP relative clause. Rather, I it was a reduced
clause, specifically a nP. The external argument occupied the specifier of this projection, where
it received genitive case. Since n lacks the ability to license structural case, the direct object was
required to move to the edge of nP, where it could value nominative (absolutive) case. In this
way, the n was equipped with two of the features driving the derivation of a transitive clause in a
v-type ergative language, i.e. genitive case for the specifier and an EPP feature. The changes
which resulted in the analysis in (8) for modern Tagalog were 1) the reanalysis of n as v; and 2)
the addition of an absolutive case feature on transitive v. In section 4.2, I present evidence for
each of these changes. Note that not positing a relative clause as the source structure avoids the
15
problems of Kaufman’s (2009) nominalist approach to modern Tagalog. It also allows for
diachronic changes to be straightforwardly accounted for in terms of the featural make-up of the
functional categories of n and v and does not require any structural alterations which radically
affect constituency.
4. Diachronic Source
In this section, I propose a diachronic source for the Tagalog ergative/genitive syncretism as a
reduced clausal nominalization nP. First, I sketch an earlier proposal by Starosta et al. (1982) but
later reject this analysis because the proposed change requires a significant alteration of structure,
i.e. a “radical” reanalysis in the sense of Haspelmath (1998), Whitman (2000), Garrett (2012),
and others. In section 4.2, I spell out the nP proposal and provide support for the two changes
involved in the reanalysis of nP as the v-type ergative vP.
4.1. Previous Approach
Starosta et al. (1982) propose that the type of ergative syntax observed in most Philippine and
Formosan5 languages is the result of diachronic reanalysis of clausal nominalizations as finite
verbal clauses. Specifically, they propose that applicative and transitivity verbal affixes in the
modern languages were nominalizing affixes in Proto-Austronesian (PAN). There was a separate
set of verbal affixes, which included applicatives homophonous with the prepositions *i, *aken,
which are attested in Oceanic languages, as well as some Indonesian, Philippine, and Formosan
5 The term “Formosan languages” refers collectively to the Austronesian languages spoken in Taiwan, but does not indicate a subgroup.
16
languages. For example, an intransitive clause could be constructed containing a verb, subject,
and PP projected by *i, *aken, or another preposition.
(16) S V PP NP ‘climb’ ‘John’ P NP i ‘mountain’ ‘John climbed over the mountain.’ (Starosta et al. 1982:153)
A transitive clause could be formed through “preposition capturing” by suffixing the verb
with the preposition as an applicative.
(17) S V NP NP ‘climb’-i ‘John’ ‘mountain’ ‘John climbed the mountain.’ (Starosta et al. 1982:153)
Preposition capturing could also take place within nominalizations. (18) shows an example
with *-ana, a nominalizing applicative. *-ana attaches to the verb and projects a nominal
category which can then be predicated of a subject.
(18) S NP NP ‘mountain’ N NP (NOM) ‘climb’-ana ‘John’
17
(GEN) ‘The place where John climbed is the mountain.’ (Starosta et al. 1982:157)
The similarity in function between the two kinds of applicative enabled the reanalysis of the
copula construction in (18) as a transitive verbal clause like (19). Nominalizers like *–ana were
likewise reanalyzed as verbal affixes.
(19) S V NP NP ‘climb’-ana ‘John’ ‘mountain’ (GEN) (NOM) ‘John climbed the mountain.’ (Starosta et al. 1982:157)
The resemblance between (18) and Kaufman’s (2009) analysis of Tagalog should be obvious.
The key point is that the nominative (absolutive) subject is located outside the constituent
containing the nominalized verb and external argument possessor. The analysis also assumes that
the reanalysis of (18) as (19) involved a significant alteration of structure, i.e. a “radical”
reanalysis. In (19), the verb and its arguments are all sisters, while in (18) the subject is located
outside the constituent containing the verb and external argument. This is not a problem for the
traditional approach to reanalysis put forth by Langacker (1977) and assumed in more recent
work by Hopper and Traugott (1993), Harris and Campbell (1995), and others. For this approach,
reanalysis only requires structural ambiguity. As long as the surface string is the same, the
presence of an ambiguity allows replacing one phrase marker with a completely different one.
18
(20) Reanalysis
“…change in the structure of an expression or class of expressions that does not involve
any immediate or intrinsic modification of its surface manifestation.” (Langacker
1977:58)
However, there is an obvious shortcoming inherent in such an approach, since there are no
structural constraints imposed on the change. Consequently, it is difficult to make predictions
about the types of changes which are possible in natural languages. Furthermore, Haspelmath
(1998), Whitman (2000), Garrett (2012), Whitman and Paul (2005), Whitman (2009), and others
have shown that many diachronic changes previously analyzed as radical reconstructions can
actually be captured through other means, e.g. grammaticalization (in particular reanalysis of
category labels) or other changes which do not alter constituency. The historical scenario I
propose in section 4.2 takes this type of approach, specifically as changes in the feature bundles
on n (changing it to v) and subsequently to this v (giving it a structural case feature).
Before entering the analysis, it is first necessary to establish the historical point at which the
innovations took place. Wolff (1973) reconstructs the PAN verbal system as closely resembling
modern Philippine languages, which presupposes that the ergative syntax present in the modern
languages must have also been a feature of the proto-language. Ross (1995, 2002) also assumes
this foundation but adds a series of non-indicative verbal affixes to Wolff’s reconstructions.
Starosta et al. (1982), as summarized in the previous subsection, propose that modern
Philippine verbal affixes were nominalizers in Proto-Austronesian, but they do not identify a
subgroup which reflects the reanalysis of the nominalizers as verbal affixes. In later work,
Starosta (1995, 2001) proposed this to be a subgroup excluding Rukai and Tsou. Ross (2009)
19
further excludes Puyuma from this subgroup, which he terms Nuclear Austronesian. This
subgroup contains all Austronesian languages except Puyuma, Rukai, and Tsou. All
Austronesian languages spoken outside of Taiwan – including Philippine languages – belong to
the Malayo-Polynesian subgroup.
PAN subgrouping (Ross 2009:316)
(21) Puyuma
Rukai
Tsou
Nuclear Austronesian (ERG=GEN)
Kanakanavu, Saaroa
Northwest Formosan: Saisiyat, Kulon-Pazih
Atayalic: Atayal, Seediq
Western Plains: Thao, Taokas, Favorlang-Babuza, Papora, Hoanya
Bunun
Paiwan
East Formosan: Basay-Trobiawan, Kavalan, Amis, Siraya
Malayo-Polynesian: all extra-Formosan, including Yami
Returning the discussion to the ergative/genitive syncretism, this can be seen in widely
distributed Austronesian subgroups. Ross (2006) reconstructs PAN genitive case as *n-. (22)
shows an example from Standard Indonesian. Indonesian has undergone numerous innovations
and has largely been reanalyzed an accusative language (Aldridge 2008a, 2012a), so the
20
syncretism appears only in passive clauses, which are the historical remnant of earlier ergative
clauses. In (22a), we can see that the passive agent nya ‘3SG.GEN’ is identical to the possessor in
(22b).
Standard Indonesian
(22) a. Di-tawar-kan-nya rokok ke ujung hidung si penjaga.
PASS-offer-APPL-3SG.GEN cigarette to tip nose PN guard
‘He offered out a cigarette under the tip of the guard’s nose.’ (“Jakarta”)
b. rokok-nya
cigarette-3SG
‘his/her cigarette’
The Formosan language Seediq, which retains the ergative syntax of Proto-Nuclear
Austronesian, reflects the syncretism in the same way that Tagalog does. The third person
singular ergative pronoun na in (23a) is identical to the possessor in (23b).
Seediq (Atayalic, Taiwan)
(23) a. S<n>malu=na ka sapah=nii.
<PRV>build=3SG.ERG ABS house=DEM
‘He/she built this house.’
b. sapah=na
house=3SG
‘his/her house’
21
Though the syncretism is not generally reflected directly in the modern Oceanic languages,
Lynch et al. (2002) do reconstruct it for Proto-Oceanic. (24) shows this form to be *ña.
Proto-Oceanic (Lynch et al. 2002)
(24) a. *ña=kaRat-i=a na manuk a wai
3SG:SUBJ=bite-TR=3SG:OBJ SUBJ chicken OBJ mango
‘The chicken is biting the mango.’ (Lynch et al. 2002:62)
b. *a na-ña Rumaq
ART CL-3SG house
‘her/his house’ (Lynch et al. 2002:77)
But the syncretism is not found in Tsou or Rukai.6 For example, a single set of pronouns is
used in Rukai for transitive and intransitive subjects. There is no substantive difference between
the forms lrao in (25a) and lra in (25b); vowel deletion has taken place in lra because of the
following agreement marker for the object in (25b).
Rukai (Zeitoun 2007:156)
(25) a. maavanao-nga-lrao
DYN.FIN.bathe-already-1SG.NOM
‘I have bathed already.’
6 It can be seen in some dialects of Puyuma, which Ross admits leaves an unanswered question inherent in his proposal.
22
b. o-kelrakelrange-nga-lra-ine ana lalake-‘o
DYN.FIN.beat-already-1SG.NOM-3SG.OBL that child-2SG.GEN
‘I have beaten your child.’
Now note that genitive pronouns are different from the subject forms appearing in finite
clauses. (26a) shows a genitive possessor with a noun. (26b) shows a genitive agent in a
nominalized clause.
Rukai
(26) a. lalake-li (Zeitoun 2007:327)
child-1SG.GEN
‘my child’
b. to’a-dhaac-ae-li (Zeitoun 2007:333)
REAS.NMLZ-DYN.FIN.leave-REAS.NMLZ-1SG.GEN
‘the reason I am leaving’
Clearly, then, the reanalysis of nominalizations as ergative clauses is limited to the Nuclear
Austronesian subgroup.
4.2. Proposal
In this section, I propose two innovations which led to the development of v-type ergativity in
Tagalog. First, reduced clausal nominalizations, nPs, were reanalyzed as verbal vPs. This
23
innovation is reflected across Ross’ (2009) Nuclear Austronesian subgroup. The second
innovation is the reanalysis of the new vP as transitive, with the v acquiring an absolutive case
feature to value with the direct object. This innovation is seen in Philippine languages and also in
Malagasy, so I assume it is associated with the Malayo-Polynesian subgroup or a smaller
subgroup therein.
Turning to the first innovation, I propose that the nP structure was used in Proto-
Austronesian for focus in a reduced cleft construction when an internal argument was focused.
The focused object moved to the edge of nP. Concomitant with this movement, the object valued
nominative case with T. The change involved reanalyzing nP as vP and replacing the motivation
for DP movement from information structure to case.
(27) a. [TP T[NOM] [nP DP[NOM] [nP DP[GEN] [NP N tNOM ]]]] (Movement for focus)
b. [TP T[NOM] [vP DP[NOM] [vP DP[GEN] [VP V tNOM ]]]] (Movement for case)
There are two types of evidence for this innovation. The first comes from languages in the
Nuclear Austronesian subgroup which reflect the new motivation for DP-movement to the edge
of vP. Aldridge (2004, 2008b) proposes that there are two types of ergativity7. One is the v-type
exemplified by Tagalog. The other is T-type, which is widely found among Formosan languages
in the Nuclear Austronesian subgroup. In T-type ergative languages, neither intransitive nor
transitive v carries an absolutive case feature. The sole source of absolutive case is finite T.
Consequently, movement of DPs to the edge of vP is at least in part motivated by the need to
value case.
7 See Legate (2008) and Coon et al. (2011) for similar proposals of an ergative typology.
24
(28) T-type split-ergative language
vTr: Inherent ergative case
[EPP]
No absolutive case feature
vAP: [uCase:Acc]
vIntr: No case feature
TFin: [uCase:Abs]
No [EPP]
Evidence for T-type ergativity comes from the fact that absolutive objects do not surface in
nonfinite clauses. This is because nonfinite T, being defective, does not have the ability to case
license an argument.
Seediq
(29) a. M-n-osa [PRO m-ari patis taihoku] ka Ape.
Intr-Perf-go Intr-buy book Taipei Abs Ape
‘Ape went to buy books in Taipei.’
b. *M-n-osa [PRO burig-un taihoku (ka) patis] ka Ape.
Intr-Perf-go buy-Tr Taipei Abs book Abs Ape
‘Ape went to buy books in Taipei.’
25
Recall from sections 2 and 3 that Tagalog, which is a v-type language, allows absolutive objects
in nonfinite clauses, demonstrating that v is the source for absolutive case in transitive clauses in
Tagalog.
The second type of evidence comes from the fact that some languages in this subgroup, e.g.
the Atayalic language Seediq, still retain the focus position at the edge of vP. (30a) shows
antipassive clause with two VP-internal constituents. Since this is an antipassive, the external
argument has absolutive status. Seediq is a VOS language. Accordingly, the absolutive DP
moves to a topic position in the CP layer. Subsequent to this, the remnant TP moves to a higher
position in the C domain above the absolutive. As in Tagalog, the lexical verb undergoes head
movement to Asp. Within the VP, the direct object precedes the locative. In contrast, if the
locative is expressed as a wh-word, which is inherently focused, then the locative must precede
the object8. Aldridge (2004) accounts for the word order alternation by proposing that the wh-
word in (30b) moves to the edge of vP.
Seediq
(30) a. [AspP M-n-ari [vP ... [VP sapah Purishia]]] ka Pihu.
Intr-Perf-buy house Puli Abs Pihu
‘Pihu bought a house in Puli.’
b. [AspP M-n-ari [vP inu … [VP patis tV twh ]]] Ape?
Intr-Perf-buy where book Ape
‘Where did Ape buy books?’
8 The wh-word does not move further to clause-initial position, because it is trapped inside the fronted TP. See Aldridge (2004) for discussion.
26
The second innovation is illustrated in (31). In Malayo-Polynesian or a subgroup therein, v
was reanalyzed as fully transitive and came to carry an absolutive case feature. Since movement
to the edge of vP was no longer required to value case, the motivation for the movement was
reinterpreted as information structure related and took place only when the object was specific,
as is the case is modern Tagalog, as shown in section 3.
(31) a. [TP T[NOM] [vP DP[NOM] [vP DP[GEN] [VP V tNOM ]]]] (Movement for case)
b. [TP T [vP DP[NOM] [vP DP[GEN] v[NOM] [VP V tNOM ]]]] (Movement for IS)
As discussed in section 3, absolutive case is available in transitive nonfinite clauses in
Tagalog, which is evidence that transitive v can value structural case. If T were the only source
of absolutive case, then absolutive DPs would not be expected to surface in nonfinite clauses,
where T is defective.
Tagalog
(32) Nagba-balak ang babae-ng [PRO tulung-an ang lalaki]
INTR.PROG-plan ABS woman-LK help-APPL ABS man
‘The woman is planning to help the man.’
Malagasy also allows transitive nonfinite clauses, indicating that v values absolutive case in
transitive clauses in this language as well.
27
Malagasy
(33) Kasain-dRasoa [PRO hosasana ny zaza]
intend.TR.ERG Rasoa FUT.TR.wash Det child
‘Rasoa intends to wash the child.’ (Paul and Travis 2006)
5. Precedent: Japanese case-marking and alignment
In the preceding section, I sketched a diachronic account of the reanalysis of a reduced clausal
nominalization nP as a fully transitive verbal vP projection. This reanalysis does not involve an
alteration of c-command relations among constituents in the nP or vP, and it accounts for the
development of the type of ergativity displayed by Tagalog in a straightforward way. What has
not yet been discussed in this paper is evidence for posting nP (rather than a full relative clause
CP) as the input structure to the reanalysis. In this section, I offer indirect evidence from parallel
developments proposed for Japanese.
Modern standard Japanese is an SOV language with morphological case marking.
Nominative is indicated by the postposition ga and accusative by o.
(34) Taroo-ga ringo-o tabe-ta.
Taro-NOM apple-ACC eat-PST
‘Taro ate an/the apple.’
What is interesting for present purposes is that Japanese has undergone a reanalysis similar to
Tagalog in which the case marker for subjects in verbal clauses has been reanalyzed
28
diachronically from an earlier genitive marker. First observe that nominative and accusative
cases in verbal clauses were typically null in Old Japanese of the 8th century.
Old Japanese (Yanagida 2012)
(35) a. 我期大王國所知良之 (Manyoshu 933)
[Wa-ga opo-kimi] kuni siras-u rasi.
I-GEN great-lord country rule-CONCL seem
‘My great lord rules seems to rule the country.’
b. 烏梅能波奈伊麻佐加利奈利 (Manyoshu 933)
[Ume-no pana] ima sakari nar-i.
plum-GEN blossom now at.peak be-CONCL
‘The plum blossoms are now at their peak.’
External arguments surfaced with genitive case in nominalized clauses. Interestingly,
subjects in unaccusative clauses were not genitive but remained bare. Yanagida and Whitman
(2009) analyze Old Japanese as having active alignment in nominalized clauses. What is relevant
to the discussion at hand is the fact that genitive case was assigned only to external arguments,
suggesting that is was indeed an inherent case assigned in the specifier of nP.
29
Old Japanese (Yanagida 2012)
(36) a. 我背子之求流乳母尒 (Manyoshu 2926)
[wa-ga seko-ga motomu]-ru omo-ni
I-GEN lord-GEN ask-ADN nurse-DAT
‘as the wet nurse that my lord asks for’
b. 久木生留清河原尒 (Manyoshu 925)
[pisaki opu]-ru kiyo-ki kapara-ni
catalpa grow-ADN clear-ADN river.bank-on
‘on the banks of the clear river where catalpas grow’
Whitman (2009) proposes that the reanalysis of genitive ga as nominative ga was the result
of the merger of conclusive and adnominal verbal morphology in Late Middle Japanese (15th-16th
centuries). This merger resulted in a structural ambiguity between embedded nominalization (nP)
and verbal projection (vP), allowing the genitive subject marking to be reinterpreted as