On Sinhala subject case marking and A-movement

+ Models

LINGUA-2406; No. of Pages 24

Available online at www.sciencedirect.com

ScienceDirect Lingua xxx (2016) xxx--xxx www.elsevier.com/locate/lingua

On Sinhala subject case marking and A-movement Chao-Ting Tim Chou a,*, Sujeewa Hettiarachchi b a

b

Taipei Medical University, Taiwan University of Sri Jayewardenepura, Sri Lanka

Received 16 September 2013; received in revised form 18 December 2015; accepted 21 December 2015

Abstract The relation between volitive/involitive moods of Sinhala verbs and subject case marking remains unresolved in the scarce generative literature on Sinhala. Previous analyses of subject case marking in this language (Gair, 1990a,b; Inman, 1992; Beavers and Zubair, 2010, 2013) assume that non-nominative cases are lexical/inherent/quirky cases assigned by involitive verbs to the subject NP, whereas nominative case is not tied to any particular lexical semantics, and arises as the default case when the semantic conditions for all available quirky cases fail. We argue that the distinction between nominative and non-nominaive cases in Sinhala should not be characterized as one between default and lexical/inherent/quirky cases. Rather, based on previously unnoted data, we contend that (i) nominative case in Sinhala is a structural case assigned by a finite T, and (ii) A-movement in Sinhala is driven by case valuation (see Epstein and Seely, 2006; Bosˇ kovic´, 2002, 2007), rather than by a universal EPP structural requirement on T (contra Gair, 1990a,b; see Chomsky, 2000, 2001). © 2015 Elsevier B.V. All rights reserved.

Keywords: (In)volitives; Lexical case; Structural case; A-movement; EPP; Sinhala

1. Introduction Sinhala, an Indo-Aryan language spoken in Sri Lanka, has a well-known but uncommon semantic classification of verbs. Each Sinhala verb root occurs in one or two stem classes, commonly known as volitives and involitives (see Gair, 1990a,b; Kahr, 1989; Gair and Paolillo, 1997; Gunasekara, 1999; Inman, 1992; Henadeerage, 2002; Beavers and Zubair, 2010, 2013). The (in)volitivity of a verb denotes the extent to which the referent of the subject or external argument, in a given context, is involved in the action denoted by the verb. For instance, in (1a) with the volitive form of the verb, natənəwa ‘dance’ is both volitional and intentional on the part of the subject Lal. By contrast, in (1b), with the involitive verb, the act of dancing is either non-volitional or unplanned as far as Lal is concerned.1,2,3 * Corresponding author. Tel.: +886 2 27361661x2657. E-mail addresses: [email protected] (C.-T. Tim Chou), [email protected] (S. Hettiarachchi). 1 The following abbreviations are used: ACC = accusative case, COMP = complementizer, DAT = dative case, EMP = emphatic marker, GEN = genitive case, ERG = ergative case, FUT = future tense, INDEF = indefinite NP, INF = infinitival, INST = instrumental case, INVOL = involitive verbs, LOC = locative case, NOM = nominative case, NOMIN = nominalizer, OPT = optative marker, VOL = volitive verbs, PAST = past tense, PRED = predicate, PRES = present tense, PRT = participle, Q = question particle/morpheme, SUBJ = subjunctive, 3SG = third person singular. 2 In this paper, our focus is on standard colloquial Sinhala as spoken in and around the western province of Sri Lanka (Dissanayake, 1976). Our informants (age: 30--40 years old), including one of the authors of this paper, are native Sinhala speakers who use it as their most dominant language in day-to-day communication. However, subject case marking and scopal interpretation in Sinhala can be subject to dialectal variation (see e.g., Beavers and Zubair, 2010, 2013; Dissanayake, 1976). 3 Notice that volitive verbs do not strictly entail volitionality. The semantic/pragmatics of the use of volitive and involitive verbs in Sinhala is not the focus of this paper. See Inman (1992) and Beavers and Zubair (2008, 2010, 2013) for details. http://dx.doi.org/10.1016/j.lingua.2015.12.012 0024-3841/© 2015 Elsevier B.V. All rights reserved.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

2

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

(1)

a.

b.

lal natənəwa. Lal.NOM dance.VOL ‘Lal (actively/voluntarily) dances.’ lal-ʈə nætenəwa. Lal-DAT dance.INVOL ‘Lal (involuntarily) dances.’

This volitive/involitive contrast among Sinhala verbs correlates with different case marking of the external argument of a typical finite clause (see Gair, 1990a,b; Inman, 1992; Kariyakarawana, 1998; Henadeerage, 2002; Jany, 2005; Beavers and Zubair, 2010, 2013): a volitive verb almost always takes a nominative subject4 (1a) and (2), while the subject of an involitive verb most often takes a range of non-nominative cases,5 including dative (1b) and (3), instrumental (4), and accusative (5)6: (2)

lal/*lal-ʈə/*lal-athih/*lal-wə natənəwa Lal.NOM/*-DAT/*-INST/*-ACC dance.VOL.PRES ‘Lal (actively/voluntarily) dances.’

(3)

a.

b.

(4)

a.

b.

(5)

lal-ʈə nætenəwa Lal-DAT dance.INVOL.PRES ‘Lal (involuntarily) dances.’ lal-ʈə induwa-k kiyəuna. Lal.DAT song-INDEF sing.INVOL.PAST ‘Lal (involuntarily) sang a song.’ sita-athih7 karame kæduna. Sita-INST tap.ACC break.INVOL.PAST ‘Sita (involuntarily) broke the tap.’ amma-geh sinhala kæmə hondətə hædenəwa. mother-INST Sinhala food.ACC well make.INVOL.PRES ‘Mother makes Sinhala food well.’

lameya-wə waturə-ʈə wætuna. water-into fall.INVOL.PAST child-ACC ‘The child (involuntarily) fell into water.’

The major goal of this paper is not to give a detailed discussion of the semantic/pragmatic conditions under which each non-nominative subject case is assigned in Sinhala. Rather, this paper investigates the syntactic (structural) conditions of subject case assignment and the driving force of A-movement in Sinhala, especially in light of recent developments in Minimalism (Chomsky, 1995 and thereafter). We focus on the assignment of nominative case with volitives in this paper, and propose that (i) nominative case is a structural case assigned/valued by a finite T, whereas non-nominative case is an inherent case assigned by the involitive verb, and (ii) involitive subjects remain within vP, whereas the subject of a volitive verb must raise to spec-TP for case assignment/valuation by the finite T. Two important consequences of our proposal are that (i) A-movement can be triggered by (structural) case valuation, rather than by a universal EPP requirement on

4 Beavers and Zubair (2013) and Gair (1990a,b) observe that some volitive verbs such as dannəwa ‘know’ can occur with DAT subjects. However, we observe that this is commonly found among Sinhala speakers whose dominant language is Tamil. 5 Beavers and Zubair (2010, 2013), contra Inman (1992), correctly show that some involitives such as bindenəwa ‘break’, ærenəwa ‘open’, and mærenəwa ‘die’ also allow nominative subjects. Our intuition is that nominative subject is only allowed with an inherently involitive verb which lacks a volitive counterpart. However, our focus in this paper is limited to non-nominative subjects, mainly dative subjects, in involitive constructions. We leave it for future research to determine whether nominative with (inherent) involitives patterns like nominative with volitives (as suggested by Beavers and Zubair, 2010, 2013) or they behave like dative and other non-nominative subjects associated with involitives. The major goal of this paper is to show nominative subjects with volitive verbs and dative subjects with involitive verbs pattern differently in a number of contexts to be examined in section 3. 6 Despite the variation shown in (3)--(5), subject case marking in involitive constructions shows some obvious patterns (Gair, 1990a,b; Inman, 1992; Henadeerage, 2002). Dative subjects can occur with both intransitive involitives (3a) and transitive involitives (3b). By contrast, instrumental subjects are only allowed with transitive involitives (4). Accuative subjects in (5) are only found with a certain class of intransitive involitives which are known as ‘P only verbs’ (Gair, 1990a,b) or ‘unaccusatives’ (Beavers and Zubair, 2013). 7 Sinhala has two distinct morphological forms associated with the instrumental case: -athih and -geh (see Gair, 1990a,b).

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

3

Table 1 (6) Examples of volitive and involitive verbs. Gloss

Root

a. ‘wash’ b. ‘suck’ c. ‘make’ d. ‘go’ e. ‘run’ f. ‘give’ g. ‘murder’ h. ‘win’ i. ‘mature’ j. ‘ripe’ k. ‘fall’ l. ‘turn’

hoodəurəhadəyaduwədeminimarədinəpæheidewætehære-

Present stem formation

Past stem formation

Volitive stem

Involitive stem

Volitive stem

Involitive stem

hoodə-nə-wa urə-nə-wa hadə-nə-wa ya-nə-wa duwə-nə-wa de-nə-wa minimarə-nə-wa dinə-nə-wa N/A N/A wæte-nə-wa hære-nə-wa

heede-nə-wa ire-nə-wa hæde-nə-wa yæwe-nə-wa diwe-nə-wa dewe-nə-wa N/A N/A pæhe-nə-wa ide-nə-wa wæte-nə-wa hære-nə-wa

heed-u(w)-a ir-u(w)-a hæd-u(w)-a yæ-u(w)-a diuw(w)-a dunn-a minimær-uw-a din-uw-a N/A N/A wæt-un-a hær-un-a

heed-un-a ir-un-a hæd-un-a yæ-un-a diuww-un-a deww-un-a N/A N/A pæh-un-a id-un-a wæt-un-a hær-un-a

T (contra Gair, 1990a,b; see also Chomsky, 2000, 2001; Epstein and Seely, 2006; Bosˇ kovic´, 2002, 2007), and (ii) structural case valuation can apply independently of w-feature valuation. The remainder of this paper is structured as follows. Section 2 reviews previous analyses of subject case marking in Sinhala. Section 3 discusses empirical challenges for these previous analyses. We propose an empirically preferable analysis in section 4. We then present new data indicating that nominative subjects with volitives and dative subjects with involitives in Sinhala behave differently in a number of other syntactic contexts in section 5. Section 6 summarizes our findings and discusses some of their theoretical implications. 2. Previous analyses of subject case marking in sinhala Table 1, adapted and expanded from Inman (1992:24--25) illustrates key morphological distinctions between volitive stems and involitive ones. First, involitive stems always have front root vowels and a thematic vowel -e- in the present tense, whereas there is no place restriction on the root vowels for volitive stems whose thematic vowel is either -a- or -i- in the present tense. Second, the past tense form of both volitive and involitive stems has a front root vowel, and the thematic vowel is -u(w)-/-i(y)- for the volitives and -un- for the involitive. Third, verbs are inflected for tense by adding the nə-wa suffix for simple present, the -a(a) suffix for simple past, and the -nnə/n̩d̩ə suffix for infinitive. Fourth, notice that the volitive/ involitive distinction in Sinhala is common across transitive (6a--c), intransitive (6d--e), and ditranstive verbs (6f). This fact challenges previous analyses (e.g., Premaratne, 1986; Kahr, 1989) that treat involitives as either passive or medial passive in colloquial Sinhala. If the involitive construction were passive voice in colloquial Sinhala, involitive forms would not be expected with intransitive verbs.8 Fifth, as noted in many previous studies (e.g. Beavers and Zubair, 2010; Chandralal, 2010; Gair, 1990a,b; Inman, 1992), not all verbs exhibit both volitive and involitive moods. Beavers and Zubair (2010) assume that whether a given verb has both forms depends on the lexical semantics of that verb: ‘‘verbs that describe events with no inherent (non-) volitionality can show both forms (p. 4).’’ If a verb denotes an action that requires volition and intention on the part of the referent of its subject, it is an inherent volitive verb that lacks an involitive form (6g--h). Similarly, there are inherent involitive verbs like (6i) and (6j) which denote actions or processes that are not under the volitional or intentional control on the part of the referent of their subject. Thus, such inherent involitive verbs lack corresponding volitive forms.9 Sixth, note that verbal morphology alone is not always a reliable indicator of the volitivity classification of Sinhala verbs.10 For instance, Gair (1990a,b) and Inman (1992) consider wætenəwa ‘fall’ (6k) and hærenəwa ‘turn’ (6l) as inherent involitive verbs that lack volitive forms. This analysis is not surprising in view of the fact that their morphological

8

Gair (1990a,b) and Henadeerage (2002) argue that colloquial Sinhala does not have a true passive construction, and Sinhala involitive constructions are different from prototypical passive constructions in languages like English. For a detailed discussion of the phenomenon, readers are referred to Gair (1990a,b) and Henadeerage (2002). 9 Inman (1992) provides a list of inherent volitive (reponent) and involitive (deponent) verbs in Sinhala. However, his classifications of some verbs as inherent volitives or involitives are problematic. For instance, he assumes dirənəwa ‘wither’, paayənəwa ‘rise’, and morənəwa ‘mature’ are inherent volitives, yet they denote non-volitionality. 10 For a similar observation, see Chandralal (2010:157).

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

4

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

forms pattern together with other involitive verbs. However, despite the morphological resemblance, they can be used as both a volitive and an involitive, as evidenced by the distinct interpretations of the verb and subject case marking in (7) and (8). (7)

a.

b.

(8)

a.

b.

balannə mamə dæn wætenəwa. see.INF I.NOM now fall.VOL.PRES ‘See, I (voluntarily) fall down now.’ putuwe hæpu-not oya-wə wætenəwa. chair bump-if you-ACC fall.INVOL.PRES ‘If you bump against the chair, you will (involuntarily) fall.’ paləmə gawin mamə dakuna-ʈə hæruna. bridge by I.NOM right-to turn.VOL.PAST ‘I (voluntarily) turned to right by the bridge.’ sadde æhuna-mə, ma-wə e pættə-ʈə hæruna. that side-to turn.INVOL.PAST sound hear.PAST-by I-ACC ‘By hearing the sound, I (involuntarily) turned to that side.’

In view of the difficulty arising from identical morphological forms, we suggest adopting the following more reliable tests to distinguish volitive from involitive verbs. First, Inman (1992) shows that only volitive verbs can form an imperative, but not involitives, as shown by (9): (9)

a.

b.

bath kannə! rice eat.VOL.INF ‘Eat rice!’ *bath kæwennə! rice eat.INVOL.INF

Second, Beavers and Zubair (2013:7) note that only volitive verbs are compatible with ‘volitional adverbs’ such as hiθəla ‘deliberately’, as illustrated by (10): (10)

a.

siri hiθəla sinduwa-k kiwwa. Siri.NOM deliberately song-INDEF sing.VOL.PAST ‘Siri deliberately sang a song.’ b. *siri-ʈə hiθəla sinduwa-k kiyəuna. Siri-DAT deliberately song-INDEF sing.INVOL.PAST #‘Siri deliberately (and involuntarily) sang a song.’

These two tests show that neither wætenəwa ‘fall’ nor hærenəwa ‘turn’ is an inherent involitive verb: both can form an imperative (11), and can co-occur with hiθəla ‘deliberately’ (12): (11)

a.

b.

(12)

a.

b.

mamə kiyənə-kotə oya/*oya-wə wætennə. I.NOM ask-when you.NOM/you-ACC fall.VOL.INF Fall down when I ask you to do so. oya/*oya-wə meha-ʈə hærennə. you.NOM/you-ACC here-to turn.VOL.INF ‘You turn here.’ laməya/*laməya-wə hiθəla-mə mette child.NOM/child-ACC deliberately-EMP mattress ‘The child deliberately fell on the mattress.’ mamə/*ma-wə hiθəla-mə e pættə-ʈə I.NOM/I-ACC deliberately-EMP that side-to ‘I deliberately turned to that side.‘

udə-ʈə wætuna. on-to fall.VOL.PAST hæruna. turn.VOL.PAST

Given this brief review of the contrast between volitive and involitive verbs, now we turn to previous analyses of subject case marking in Sinhala. Gair (1990a,b) argues that Sinhala does not have typical case-driven A-movement to spec-IP/ Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

5

TP,11 an operation hypothesized for languages such as English. Instead, a subject NP in Sinhala receives lexical (also called ‘‘inherent’’ or ‘‘quirky’’) case12 from the verb inside vP at D-structure before it moves to spec-IP to satisfy the EPP (Extended Projection Principle). Given the role of lexical semantics associated with subject case marking, Gair (1990a,b) proposes the Strong Lexical Case Assignment Hypothesis to account for subject case marking in Sinhala, in which the case assignment of arguments ‘bears an intimate connection with θ-role’ and ‘is fully specified in the lexicon’ (p.73). Gair (1990a,b) does not extend this strong lexical relation between case and θ-role to nominative case in Sinhala because he treats nominative as the default case that is not tied to any particular lexical semantics. Actually, the nominative is unspecified in a verb’s lexical entry and is filled in later by default rules under his analysis. He departs from the standard treatment of nominative assignment in the GB theory in the sense that it is the volitive verb rather than a finite INFL that assigns nominative to the subject in Sinhala. As a result, all arguments, including the subject NP are not only basegenerated inside vP but also receive (either inherent or default) case from the verb. He contends that ‘weak’ INFL in Sinhala, characterized by the absence of verbal agreement, as illustrated by (13) (Gair and Wali, 1998:140), is unable to assign nominative to a subject NP. (13)

mamə / oyaa / eya / api / oyaala / eyaala pansələ-ʈə yanəwa. I / you / (s)he / we / you-PL / they temple-to go.PRES ‘I/you/(s)he/we/you-all/they go to the temple.’

Thus, Gair (1990b) argues that ‘‘AGR plays no role in Sinhala, to the extent that there is no element within INFL that plays its subject case assignment role’’ (p. 142). To support this line of analysis, Gair (1990b) discusses subject case marking in embedded infinitival clauses as in (14) and (15). (14)

mamə hetə ya-nnə puluwani. I.NOM tomorrow go.VOL-INF be.likely.to-PRED ‘I am likely to go tomorrow.’

(15)

lameya-ʈə æhe-nnə puluwani. child-DAT hear.INVOL-INF be.likely.to-PRED ‘The child is likely to hear.’

Note that the subject case marking in (14) and (15) depends on the volitivity of the embedded predicates (go and hear, respectivey), so Gair (1990b) assumes that the subject NPs in (14) and (15) are the subjects of the embedded infinitival clause. Based on the fact that the subject of an infinitival clause can bear case, Gair (1990a,b) reasons that subject case marking in Sinhala has nothing to do with a finite INFL.13 Following Gair (1990b), Inman (1992) also assumes that nominative is the default case that is assigned to ‘‘a subject NP to which no other case has been assigned’’ (p. 159). The default nominative case, similar to non-nominative case, is assigned by the verb. Henadeerage (2002) takes a similar approach to subject case marking in Sinhala. He also treats volitive verbs on a par with involitives in their ability to assign case to the subject NP. However Henadeerage (2002) differs from Gair (1990b) and Inman (1992) in that he defines nominative as a lexically assigned case under semantic conditioning rather than the ‘‘default case’’ that arises when no more specific case is assigned to the subject by a verb. He assumes that a nominative subject indicates a volitional agent. However, Jany (2005) contends that there is no one-to-one relation between bearing nominative case and being the volitional agent, because the semantic and lexical properties of the verb, in particular volitivity, is not the only determinant of subject case marking in Sinhala. The semantic properties of the argument, such as animacy, semantic role, and definiteness also play a crucial role. For instance, nominative case can also appear on an inanimate argument which is unable to be a volitional agent, as shown in (16). (16)

karekə wegen giya. car.NOM fast go.VOL.PAST ‘The car went fast.’

11

We use Tense/TP and INFL/IP interchangeably in this paper. In this paper, we use lexical, inherent, and quirky case interachangeably to refer to the non-nominative case on the subject NP in Sinhala. 13 However, as we will argue in section 4, the nominative case in (14) is not assigned/valued within the infinitival embedded clause. Therefore, the occurrence of the nominative case in (14) does not constitute a piece of conclusive evidence for the dissociation of nominative case and the finite T in Sinhala. 12

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

6

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

In view of the fact that no unifying property can be identified with the variety of arguments that can be marked with the nominative case, Jany (2005:81--82) concludes that nominative case is the unmarked or default case in Sinhala. Similarly, the sentence (17) cited by Beavers and Zubair (2010:11, (9)) makes clear that the nominative case marking on the subject of volitive verbs does not strictly entail volitionality, as evidenced by the presence of the modifier hitəla nemeyi ‘without intention’, indicating accidental action. (17)

laməya piŋgaanə kæduwa, eet hitəla nemeyi. child.NOM plate broke.VOL.PAST but intention without ‘The child broke the plate unintentionally.’

(Inman, 1992:98, (39))

Beavers and Zubair (2010, 2013) define nominative as a semantically neutral case and the default structural case for subjects that arises only when the semantic conditions for all available quirky cases fail. It is worth pointing out that while Beavers and Zubair (2010, 2013), like Gair (1990a,b), Inman (1992), Jany (2005), treat nominative case as the default case in Sinhala, they (p. 3) explicitly distinguish lexical (dative/accusative/instrumental) cases from the default ‘‘structural’’ (nominative) case, the latter having nothing to do with verbal semantics. In other words, their analysis of nominative case is completely detached from the verb. The characterization of nominative case as the default ‘‘structural’’ case implies that the assignment of nominative is subject to a certain structural requirement, say, occupying the specifier position of a finite T, as we see in a wide range of languages such as English. However, Beavers and Zubair (2010, 2013) do not propose an explicit structural rule for the assignment of nominative case in Sinhala. The gist of previous analyses of Sinhala subject case marking and A-movement is summarized in (18): (18)

a. b.

c.

The EPP feature on T drives both volitive and involitive subjects to spec-TP (Gair, 1990a,b). All prior authors except Beavers and Zubair (2010, 2013) propose that all cases are assigned by V (either as a lexical case or as a semantically-neutral default nominative for everyone but Henadeerage, who says nominative is assigned by V under a specific semantic interpretation). Beavers and Zubair (2010, 2013) propose that nominative is a default structural case for subjects that arises when the subject receives no other case, either lexically or structurally; however, they do not propose an explicit structural rule for the assignment of nominative case in Sinhala.

Building upon Beavers and Zubair’s assumption in (18c), this paper aims to provide a detailed argumentation against (18a) and (18b). In particular, we maintain that a clear-cut distinction between nominative and non-nominative cases should be drawn in terms of structural case and lexical case, rather than in terms of default case and lexical case. In this sense, we adopt part of the core intuition in Beavers and Zubair (2010, 2013) who treat nominative as a default ‘‘structural’’ case (though, again, they do not provide an explicit structural rule governing its assignment). One major goal of this paper is to show how this line of analysis could be formally implemented within the Minimalist framework. Our argument is based on a variety of syntactic properties of Sinhala volitives/involitives, including (i) the scope interpretation of subject quantifiers, and (ii) the interaction between case marking and the interpretation of modals. These heretofore unnoted data lead us to argue for the following three points14: (19)

a. b. c.

Only involitive verbs assign inherent case to their external arguments, while volitive verbs are not lexically related to subject case marking in any way. Subjects that do not receive lexical case within vP raise to the specifier of a finite T to receive the structural nominative case. A-movement in Sinhala is triggered by case valuation, rather than by a universal EPP feature on T.

3. Empirical challenges for existing analyses In this section, we present two empirical challenges to (18a) and (18b). Each of these two empirical challenges suggests that our analysis based on the assumptions in (19) fares better empirically.

14 The new data examined in the next section show that dative subjects with involitive verbs behave differently from nominative subjects with voitive verbs. We leave for future work the question whether this contrast extends to the rest of lexically assigned cases (e.g., instrumental and accusative).

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

7

3.1. Volitivity and the interpretation of quantified NPs A subject universal quantifier scopally interacts with negation only in a sentence with a volitive verb, as shown by (20),15 which is ambiguous between total negation and partial negation.16 By contrast, when we replace the volitive verb in (20) with an involitive counterpart as in (21), only the partial negation interpretation is available. Therefore, (21) is compatible with a context where some but not all children (involuntarily) danced, but cannot be felicitously uttered in a context where no child danced involuntarily, or one in which no child danced at all.17 (20)

(21)

lamai hæməkena-mə nætuwe næhæ. children all.NOM-EMP danced.VOL not ‘All children did not (voluntarily) dance.’ ‘Not all children (voluntarily) danced.’ lamai hæməkena-ʈə-mə nætune næhæ.18 children all-DAT-EMP danced.INVOL not #‘All children did not (involuntarily) dance.’ ‘Not all children (involuntarily) danced.’

[Total negation = all > negation] [Partial negation = negation > all]

[#Total negation = all > negation] [Partial negation = negation > all]

This effect on the scopal interpretation of the subject quantifier is previously unnoted to the best of our knowledge. The contrast between (20) and (21) is not easily explained by previous analyses of Sinhala case marking and A-movement reviewed in the last section. If both nominative and dative cases are assigned within vP (either as the default case or as an inherent case), (20) and (21) are predicted to have the same range of scopal interpretation(s) for the subject quantifier. Specifically, suppose we follow Chomsky’s (2001) Activity Condition in (22), the subject’s uninterpretable/unvalued case feature is valued and then eliminated within the vP, and hence the subject becomes inactive and inaccessible to further Amovement to spec-TP. Thus, both subject quantifiers in (20) and (21) should stay within vP below næhæ which hosts a projection between TP and vP.19 Consequently, (20) is wrongly predicted to have only a partial negation interpretation. (22)

a. b.

Inactivity of an XP: An XP that eliminates its uninterpretable features (case, -wh) is rendered inactive. The Activity Condition: Inactive elements are not accessible for further (syntactic) operations.

15 As correctly pointed out by one of the reviewers, the quantifier hæməkenamə contains an emphatic marker in the form of -mə. This seems to be a property of quantifiers in Sinhala given that even several other Sinhala forms equivalent to ‘all’ have the same or a similar emphatic marker: okko-mə ‘all’, kawuru-th ‘all’. A quantifier cannot be formed in Sinhala without these makers: *hæmo, *okko, *kawuru. 16 Sinhala has several forms of negation, including næhæ, næ, nemei, nemee, nevei (see e.g., Kariyakarawana, 1998). While some of these forms result from dialectal variation, at least some forms have a specific restricted usage (see Kariyakarawana, 1998:49). In our examples, we use the most common colloquial form: næhæ. Gair (1990a,b) and Gair and Paolillo (1997) show that næhæ has the properties of a quasi verb as it can share some inflectional properties with lexical verbs. But in our view, these properties of næhæ are more commonly found in formal spoken Sinhala or literary Sinhala. 17 One reviewer asks about the interpretation of simple negative sentences with involitive verbs like (i).

(i)

maʈə malu kæwn-e næhæ. I.DAT fish.ACC eat.INVOL not ‘I did not (involuntarily) eat fish.’

The native speakers we consulted reported that negative involitive sentences like (i) are ambiguous. It could mean that I did not eat fish regardless of whether I wanted to or did not want to do so. The other meaning is that I ate fish voluntarily. That is, næhæ could negate either the entire verb phrase or only the involitive component of the verb phrase. For the ease of exposition in this paper, we focus only on the interpretation where næhæ negates the entire verb phrase, rather than the (in)volitive component of it. 18 One reviewer points out that the interpretation of (21) does not necessarily follow from a scopal configuration where the negation scopes above all because ‘‘negation > all’’ is semantically weaker than ‘‘all > negation’’ and cannot by itself rule out that all the children danced in (21). To confirm that the interpretation of (21) results from a scopal configuration where the involitive subject scopes below the negation (which is the analysis we propose for (21)), the reviewer suggests using positive polarity items to test this hypothesis, as we will show in the end of section 4.1. 19 Some researchers argue that negation may not be a unique and rigid head occupying a fixed scope position in phrase structure (Boeckx, 2001; Ladd, 1981; Büring and Gunlogson, 2000 for English and von Stechow and Penka, 2003 for German). However, how such a view on the scope position of negation solves the issue at hand in Sinhala remains unclear. In particular, there is no principled account for why negation could have varied scope positions in relation to the nominative subject, but not to the dative subject. For the purpose of this paper, we will assume the hypothesis that negation heads a projection between TP and vP in Sinhala.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

8

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

Alternatively, suppose we follow Gair (1990a,b) and assume that INFL/T in Sinhala is endowed with a universal EPP property that induces obligatory A-movement to spec-TP, and follow Nevins (2005) to abandon the Activity Condition.20 The universal quantifier subject NPs in (20) and (21) should then both raise to spec-TP to satisfy EPP, even though they have already been assigned case within vP, as in (23). Consequently, the negation næhæ c-commands the lower copy of the universal quantifier subject within vP and is also c-commanded by the higher copy at spec-TP. Therefore, (20) and (21) are wrongly predicted to exhibit both total and partial negation interpretations (as does) the English sentence in (24), contrary to fact. (23)

TP lamai hæməkena-məi lamai hæməkena- ə-məi

[TD$INLE]

T’ 3 NegP T[EPP] 3 For EPP vP Neg 3 NP v ti

næhæ case

(24)

nætuwe nætune

Every student did not pass the exam. (negation > every; every > negation)

A clarification about reconstruction is in order before we can conclude that EPP on T does not exist in Sinhala. One of the reviewers notes that the scopal contrast observed in (20) and (21) reminds one of the scopal contrast between nominative and ergative subjects in Hindi/Urdu in (25) (a novel observation made in Nevins and Anand, 2003; Anand and Nevins, 2006).21 (25)

a.

b.

koi shaayer har ghazal likhtaa hai some poet-NOM every song-ACC write.m-IMPF be-PRES ‘Some poet writes every song.’ kisii shaayer-ne har ghazal likhii some poet-ERG every song-NOM write.f.PERF ‘Some poet wrote every song.’

[9 > 8; 8 > 9] [9 > 8; *8 > 9]

While (25a) with the nominative subject admits an inverse scope reading (8 > 9), (25b) with the ergative subject can only be interpreted with surface scope (9 > 8).22 Anand and Nevins (2006) maintain that while nominative case is a structural case based on the w-AGREE relation between T and an NP, ergative case is a lexical case associated with the theta-role of agent in Hindi (see also Woolford, 1997; Ura, 2000). In addition, they assume that EPP on T drives the movement of both nominative and ergative subjects to spec-TP. They claim that the relevant difference between ergative and nominative subjects responsible for scopal rigidity in (25b) is the absence of a w-AGREE relation between the ergative subject and T, which renders the reconstruction of the ergative subject back to spec-vP inapplicable, in accordance with the restriction on reconstruction in (26). (26)

Agreement-allows-Reconstruction: Reconstruction of an XP from [the specifier of, CTC and SH] a head H is possible iff H AGREEs with XP (Anand and Nevins, 2006:10).

20 Note that Gair’s (1990a,b) EPP-driven analysis cannot co-exist with an analysis assuming Chomsky’s (2001) Activity Condition. This is because case assignment/valuation within vP should render the quantifier subject NPs in (23) inactive according to the Activity Condition, and hence the subject NPs should stay within vP. However, this leads to a violation of EPP, incorrectly predicting that (20) and (21) are both ungrammatical. 21 Hindi/Urdu is an aspectually-split-ergative language showing ergative subject marking and agreement with the object in perfective aspects. 22 As one reviewer points out, the Sinhala pattern is like the pattern observed in Hindu-Urdu in one way: non-nominative subjects have only one reading, while nominative subjects have two. However, these two languages are different in that the reading that non-nominative subjects have in Sinhala requires reconstruction (if movement to spec-TP for EPP takes place, as suggested by Gair, 1990a,b), in contrast to Hindu-Urdu where non-nominative subjects cannot reconstruct.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

9

However, we do not see how the presence or absence of w-AGREE with T helps us with the contrast between nominative and dative subjects in (20) and (21). Specifically, suppose we follow Gair (1990a,b) in assuming the presence of EPP on T in Sinhala which drives movement of both nominative and dative subjects to spec-TP, the absence of agreement features on T in Sinhala (which Gair also assumes, based on (13)--(15)23) should render both nominative and dative subjects unable to undergo reconstruction back to a position below negation, wrongly predicting that the partial negation interpretation is unavailable in (20) and (21). Therefore, without the help of reconstruction (as formulated by Anand and Nevins, 2006), once we move both volitive and involitive subjects to spec-TP, in accordance with Gair’s EPP-based analysis of A-movement, there is no way for the subjects to reconstruct back to spec-vP to yield the partial negation interpretation. Notice that, if we assume Sinhala exhibits phonologically null w-AGREE with T, and nominative case is assigned based on such w-AGREE, the nominative subject would be allowed to reconstruct to yield the partial negation interpretation, a correct result. However, an EPP-based analysis would still wrongly predict the total negation interpretation in (21). Therefore, we conclude that the EPP-based analysis of A-movement in Sinhala is not on the right track, and some other force, presumably structural nominative case valuation, is responsible for moving only the volitive subject, and crucially not the involitive one, to spec-TP thereby yielding the total negation interpretation in (20). In effect, this amounts to the claim that nominative case in Sinhala is not assigned within vP (either as the default case in Gair, 1990a,b; Inman, 1992, or as a lexical case in Henadeerage, 2002). Rather, it is a structural case, in line with Beavers and Zubair’s (2010, 2013) core intuition. Defenders of the analyses of nominative case assignment in Gair (1990a,b), Inman (1992), or Henadeerage (2002) may claim that the total negation interpretation in (20) is brought about by quantifier raising (QR, see May, 1985; Hornstein, 1995; Johnson and Tomioka, 1997; Beghelli and Stowell, 1997, among many others) or overt (string-vacuous) scrambling of the volitive subject out of vP to a position scoping over negation, as depicted in (27). (27)

QR or overt (string-vacuous) scrambling? [TD$INLE]

lamai hæməkena-mə [ NegP [ vP t nætuwe] næhæ] children all. NOM-EMP danced.VOL not

An obvious problem with this line of analysis is why QR or overt scrambling selectively applies only to the volitive subject, but not the involitive subject. Below we present additional evidence against this alternative analysis of the total negation reading in (20). First, notice that, unlike the English counterparts, the Sinhala sentences in (28) and (29) are not ambiguous: they only have the surface scope interpretation, and the inverse scope reading (available in English due to the QR of the object NP over the subject trace at spec-vP at LF) is not available, regardless of subject case-marking and volitivity.24,25 (28)

(29)

[kauruhari laməye-k] [guruwaru hæməkena-wə-mə] dækka. some student.NOM-INDEF teacher every-ACC-EMP see.VOL.PAST ‘Some student saw every teacher.’

[9 > 8; *8 > 9]

[kauruhari laməye-ku-ʈə] [guruwaru hæməkena-wə-mə] penuna. every-ACC-EMP see.INVOL.PAST some student-INDEF-DAT teacher ‘Some student saw (involuntarily) every teacher.’

[9 > 8; *8> 9]

23 See section 6 where we provide further evidence based on Rizzi’s (1990) anaphor agreement effect for the absence of agreement features on T in Sinhala. 24 Notice that the case marker generally appears on the quantifier rather than on the noun in a quantifier phrase in Sinhala, with quantifier phrases headed by kauruhari ‘some’ being an exception. 25 In the dialect of Sinhala that we are concerned with here, two operations are needed to generate the inverse scope interpretation in contexts like (28): the head noun of the some NP must be plural, and the order between the noun and the quantifier in the some NP needs to be reversed, as illustrated below:

(i)

[lamai kauruhari] [guruwaru hæməkena-wə-mə] dækka. students.NOM.INDEF some teacher every-ACC-EMP see.VOL.PAST ‘Some students saw every teacher.’ [every > some; *some > every]

Reduplication of the quantifier in the preceding NP as in (ii) also occurs in addition to the two operations in (i) in some dialects. The correlation between these three operations and the inverse/surface scope interpretations is beyond the scope of this paper, so we leave it for future research. (ii) [lamai kauru-kauruhari] [guruwaru hæməkena-wə-mə] dækka. students.NOM.INDEF some-some teacher every-ACC-EMP see.VOL.PAST ‘Some students saw every teacher.’ [every > some; *some > every]

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

10

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

This scope-rigidity property has been found in various SOV languages like Japanese, Hindi/Urdu, Persian, and Hungarian (see Karimi, 2005; Kidwai, 2000; Kiss, 2006). For instance, the inverse scope reading is not available in the Persian sentence (30) nor in the Japanese sentence (31). (30)

Har dâneshjui bâyad ye mas’ala-ro hal bo-ken-e. each student must a problem-râ26 solution SUBJ-do-3SG ‘Each student must solve a problem.’

(31)

Dareka-ga daremo-o aisiteiru. someone-NOM everyone-ACC love ‘Someone loves everyone.’

[8 > 9; * 9 > 8]

[9 > 8; *8 > 9]

Thus, the lack of an inverse scope reading in (28) suggests that the total negation reading in (20) does not arise from the application of QR of the nominative quantifier subject (if QR in fact exists in Sinhala). Furthermore, we contend that even though Sinhala exhibits overt scrambling, the total negation reading in (20) does not arise from overt (string-vacuous) scrambling of the nominative subject quantifier. First, notice that overt scrambling in Persian and Japanese alters scope interpretations, as shown by (32) and (33), respectively (cf. (30) and (31)): (32)

[ye mas’ala-ro]i har dâneshjui bâyad ti hal bo-ken-e. a problem-râ each student must solution SUBJ-do-3SG ‘Each student must solve a problem.’

[8 > 9; 9 > 8]

(33)

Daremo-oi dareka-ga ti aisiteiru. everyone-ACC someone-NOM love ‘Someone loves everyone.’

[9 > 8; 8 > 9]

By contrast, overt scrambling does not affect scopal interpretation in Sinhala, as shown by (34). (34)

a.

b.

[kauruhari laməye-k] [guruwaru hæməkena-wə-mə] dækka. some student.NOM-INDEF teacher every-ACC-EMP see.VOL.PAST ‘‘Some student saw every teacher.’ [guruwaru hæməkena-wə-mə]i [kauruhari laməye-k] ti dækka. teacher every-ACC-EMP some student.NOM-INDEF see.VOL.PAST

[9 > 8; *8 > 9] [9 > 8; *8 > 9]

Importantly, notice that subject case-marking does not affect the observed scope rigidity in Sinhala. Overt scrambling of the object quantifier does not yield inverse scope regardless of whether the subject bears nominative or dative case, as shown by (35) and (36).27 (35)

a.

b. (36)

a.

b.

[kauruhari laməye-k] [sindu hæməeke-mə] kiwwa. some student.NOM-INDEF song every.ACC-EMP sing.VOL.PAST ‘Some student sang every song.’ [sindu hæməkenə-mə]i [kauruhari laməye-k] ti kiwwa. [kauruhari laməye-ku-ʈə] [sindu hæməeke-mə] kiyəuna. some child-INDEF-DAT song every.ACC-EMP sing.INVOL.PAST ‘Some student (involuntarily) sang every song.’ [sindu hæməkenə-mə]i [kauruhari laməye-ku-ʈə] ti kiyəuna.

[9 > 8; *8 > 9] [9 > 8; *8 > 9] [9 > 8; *8 > 9] [9 > 8; *8 > 9]

In view of the absence of inverse scope readings in (34b), (35b), and (36b), we maintain that the total negation reading in (20) does not result from overt (string-vacuous) scrambling of the nominative quantifier subject (over the negation næhæ).

26

According to Karimi (2005:4), râ is a case marker for specific objects in Persian. To get the ‘‘all > some’’ reading in (34)--(36), the operations mentioned in footnote 25 (i.e., the pluralization of the noun and switching the order of the noun and the quantifier in the some NP) are needed. 27

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

11

Thus, we conclude that neither QR nor overt scrambling is responsible for the total negation reading in (20), and previous accounts of subject case marking in Sinhala fail to predict the asymmetry between (20) and (21).28 3.2. Modality and case marking Recall that Gair (1990b) cites (14) and (15) (repeated here as (37) and (38)) to argue for the dissociation between subject case assignment and the finite INFL/T in Sinhala.29 Specifically, he notes that mamə ‘I’ in (37) and lameyaʈə ‘child’ in (38), as the subjects of the embedded verbs, are assigned case within an infinitival clause, suggesting that subject case marking in Sinhala has nothing to do with a finite INFL/T or the presence of w-Agree. (37)

mamə hetə ya-nnə puluwani. I.NOM tomorrow go.VOL-INF be.likely.to-PRED ‘I am likely to go tomorrow.’

(38)

lameya-ʈə æhe-nnə puluwani. child-DAT hear.INVOL-INF be.likely.to-PRED ‘The child is likely to hear.’

However, like modals in many other languages, Sinhala modals like puluwan exhibit an ambiguity between an epistemic (possibility) and a deontic (ability) reading. Note that when (in)volitive verbs are embedded under an epistemic modal as in (37)--(40), case marking on the subject is in compliance with previous analyses: the subject of a volitive verb bears nominative, while that of an involitive verb bears dative case. (39)

lal nata-nnə puluwan. Lal.NOM dance.VOL-INF be.likely.to-PRED ‘Lal is likely to (actively) dance.’

(40)

lal-ʈə næte-nnə puluwan. Lal-DAT dance.INVOL-INF be.likely.to-PRED ‘Lal is likely to (involuntarily) dance.’

Curiously, we find that when volitive verbs are embedded under deontic modals as in (41), dative case is required despite the presence of the embedded volitive verb.30,31 (41)

lal-ʈə/*lal nata-nnə puluwan. Lal-DAT/*Lal.NOM dance.VOL.INF be.able.to ‘Lal is able to (actively) dance.’

In this connection, Inman (1992:132, fn.51) proposes that in Sinhala modal constructions, ‘‘case is first assigned by the complement verb, and then, if the complement verb assigns no case, case is assigned by the modal.’’ But the reason why the volitive verb in (41) fails to assign nominative case to the subject NP in this context, but can do so in other contexts, remains unexplained in his analysis. By contrast, the non-occurrence of nominative case in (41) is not surprising under Beavers and Zubair’s (2010, 2013) analysis which characterizes nominative as a structural default case that arises when the subject receives no other case, either lexically or structurally. The suppression of the nominative case suggests that there must be some other case assigner in (41), which is able to assign case to the subject, either lexically or structurally. We will argue in section 4.2 that (i) deontic modals in Sinhala are control verbs, and (ii) the assigner of the dative case in (41) is the deontic modal.

28 The analysis of the contrast between (20) and (21) proposed in the next section depends on the effect of the Activity Condition in (22). Note that the Activity Condition only prohibits further A-movement of a case-marked NP. Further A0 -movement of a case-marked NP should be allowed, as illustrated by (34b), (35b), and (36b). 29 Notice that the modal puluwani in (37) and (38) is mostly found in either formal spoken Sinhala or written Sinhala. The more natural colloquial form is puluwan. In the discussion below, we will use puluwan, except for the original data cited from previous work. 30 The combination of the deontic/ability reading of the puluwan and the involitive interpretation of the verb is ineffable because it is semantically/ pragmatically odd for someone to be able to do something involuntarily. 31 See also Chandralal (2010) for an observation concerning case marking and modal interpretation.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

12

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

4. Analysis In this section, we show how the assumptions in (19) (repeated here as (42)) can account for the empirical challenges, concerning scope and deontic modals, presented in the previous section. (42)

a. b. c.

Only involitive verbs assign lexical case to their external arguments, while volitive verbs are not lexically related to subject case marking in any way. Subjects that do not receive lexical case within vP raise to the specifier of a finitie T to receive the structural nominative case. A-movement in Sinhala is triggered by structural case valuation, rather than by a universal EPP feature on T.

4.1. Volitivity and the interpretation of quantifier NPs First, the scopal contrast between (20) and (21) follows from our proposal that only involitive verbs assign inherent case to their subject NP. Therefore, once the involitive subject lamai hæməkenaʈəmə ‘all children’ in (21) is base-generated in spec-vP, it receives inherent dative case from the involitive verb within the involitive vP and becomes inactive for further movement to spec-TP for case valuation, thereby yielding only the partial negation reading, as in (43). (43)

TP 3 T’ 3 NegP T 3 [TD$INLE] vP Neg 3 NP V lamai nætune næhæ hæməkena əmə inherent dative case

On the other hand, volitive verbs in Sinhala are not lexically related to subject case marking in any way. Therefore, in (20), the relevant structure of which is shown in (44), the subject NP does not get case within the vP, and hence needs to (and can, given Activity) move to spec-TP to value its case feature as nominative. Consequently, the negation næhæ ccommands the lower copy of the subject quantifier at spec-vP and is also c-commanded by the higher copy at spec-TP, correctly yielding the scopal ambiguity. (44)

TP NP T’ lamai hæməkena-məi 3 NegP TFINITIE [TD$INLE] 3 case-driven vP Neg 2 ti V nætuwe næhæ

Before ending this section, we would like to discuss an interesting prediction of our analysis pointed out by one reviewer. First, notice that ‘‘some’’ in English is called a positive polarity item (PPI) because it cannot be interpreted under the scope of negation, as shown by (45): (45)

John did not buy some books. (oksome > negation; *negation > some)

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

13

In colloquial Sinhala, wh-words can receive an indefinite interpretation meaning ‘‘some X’’ as illustrated in (46).32 (46)

a.

b.

c.

mala mokak-də33 kiwwa. Mala.NOM what-Q say.VOL.PAST ‘Mala said something.’ sara kohe-də giya. Sara.NOM where-Q go.VOL.PAST ‘Sara went somewhere.’ siri kau-də hamu-una. Siri.NOM who-Q meet.VOL.PAST ‘Siri met someone.’

Under the indefinite interpretation, wh-words in Sinhala exhibit similar licensing requirements as the PPIs in English, as shown by the following examples where these wh-indefinites cannot be interpreted under the scope of negation, regardless of the volitivity of the verbs: (47)

a.

b.

(48)

a.

b.

(49)

a.

b.

mala mokak-də kiwwe næhæ. Mala.NOM what-Q say.VOL.PAST not ‘Mala did not say something.’ mala-ʈə mokak-də kiyəune næhæ. Mala-DAT what-Q say.INVOL.PAST not Mala did not involuntarily say something. sara kohe-də giye næhæ. Sara.NOM where-Q go.VOL.PAST not ‘Sara did not go somewhere.’ sara-ʈə kohe-də yæune næhæ. Sara-DAT where-Q go.INVOL.PAST not ‘Sara did not involuntarily go somewhere.’ siri kau-də hamu-une næhæ. Siri.NOM who-Q meet.VOL-PAST not ‘Siri did not meet someone.’ siri-ʈə kau-də hamu-une34 næhæ. Siri-DAT who-Q meet.INVOL-PAST not ‘Siri did not involuntarily meet someone.’

(oksome > negation; *negation > some)

(oksome > negation; *negation > some)

(oksome > negation; *negation > some)

(oksome > negation; *negation > some)

(oksome > negation; *negation > some)

(oksome > negation; *negation > some)

Our analysis predicts that the quantifier subject in involitive contexts like (21) cannot be replaced by a PPI because we contend that an involitive subject, receiving a lexical case from the involitive verb, remains within vP below negation, violating the licensing requirements of PPI. This prediction is borne out, as evidenced by the contrast between (50) and (51)35: (50)

(51)

kau-də potə gatte næhæ someone.NOM-Q book.ACC get.VOL.PAST not ‘Someone did not take the book.’

(oksome > negation; *negation > some)

potə gænune næhæ *kaʈə-də someone.DAT-Q book.ACC get.INVOL.PAST not ‘Someone did not (involuntarily) take the book.’

32

Sinhala wh-words can be used either as wh-interrogatives or indefinite pronouns (see Kishimoto, 2005:5). This is not typologically uncommon. For a similar observation of Chinese wh-words, see Cheng (1991). 33 In some dialects, the polarity use is restricted to a specific form such as mokakdo ‘something’, kaudo ‘someone’ and kohedo ‘somewhere’. 34 The verb hamu ‘meet’ has the same form for volitive and involitive past tense. 35 Notice that there is an intriguing contrast between object PPIs in (47)--(49) and subject PPIs in (50)--(51): some kind of QR can apply to the former to satistify the licensing requirements of PPIs, whereas subject PPIs do not have access to this operation. We do not have a principled account for this contrast, and we leave it for future work.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

14

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

Previous analyses of subject case marking do not predict this contrast between (50) and (51) regarding the distribution of subject NPs formed with the PPI kaudo ‘some’. In particular, Gair’s (1990a,b) postulation of the EPP feature on T in Sinhala would force the movement of both kaudə and kaʈədə in (50) and (51) to spec-TP above the negation. As a result, (50) and (51) should both be grammatical because the PPI kaudo ‘some’ would be out of the scope of næhæ in both sentences. By contrast, our analysis correctly predicts the ungrammaticality of (51) because the subject kaʈədə marked with the lexical dative case remains within vP below næhæ. On the other hand, the subject kaudə in (50) has to move to spec-TP above næhæ for nominative case valuation under our analysis, so no restriction on the occurrence of the PPI is violated. 4.2. Subject case marking in modal constructions We turn to our analysis of the dative case in (41) (repeated below as (52)) where dative case is required despite the presence of the volitive verb. At least under existing analyses (e.g., Gair, 1990a,b; Inman, 1992; Henadeerage, 2002), one cannot attribute this to the infinitival form of the volitive verb natannə because Gair (1990a,b) and others analyze nominative case as a default case whose assignment has nothing to do with finiteness. (52)

lal-ʈə/*lal nata-nnə puluwan Lal-DAT/Lal.NOM dance.VOL.INF be.able.to ‘Lal is able to (actively) dance.’

By contrast, when puluwan is interpreted as an epistemic modal, case marking on the subject is in compliance with the analyses proposed in Gair (1990a,b), Inman (1992), and Henadeerage (2002): the subject of a volitive verb bears nominative (53), while that of an involitive verb bears dative case (54). (53)

lal nata-nnə puluwan. Lal.NOM dance.VOL-INF be.likely.to ‘Lal is likely to (actively) dance.’

(54)

lal-ʈə næte-nnə puluwan. Lal-DAT dance.INVOL-INF be.likely.to ‘Lal is likely to (involuntarily) dance.’

We argue that the contrast between the epistemic reading and the deontic reading of the modal puluwan shown in (52)--(54) can be explained if we assume that the epistemic modals in Sinhala are one-place raising predicates, while deontic modals are two-place control predicates, as schematically shown in (55) and (56), respectively. (55)

Epstemic modals as raising predicates [TP subject TINFINITIVAL (object) verb] modal.EPISTEMIC

(56)

Deontic modals as control predicates Subject [TP PRO TINFINITIVAL (object) verb] modal.DEONTIC

Under the analysis proposed here, the surface subject in sentences with epistemic modals originates from the embedded infinitival clause, while the surface subject of sentences with deontic modals is the thematic subject of the deontic modal. We will show that this raising/control distinction works in tandem with the following assumptions to account for subject case marking in (52)--(54): (57)

a. b. c.

The subject NP in (54) receives inherent dative case from the involitive verb, and stays in the embedded infinitival clause. The subject NP in (53) does not get case within the infinitival clause, and hence moves to the matrix clause to get its case feature valued as nominative by the matrix finite T. Deontic modals assign inherent dative case to their subject (=(52)).

Empirical support for the raising-control distinction in (55) and (56) comes from the placement of adverbs modifying the embedded predicate in modal contexts. Note that the adverb hondətə ‘well’ modifies the embedded verb in (58) and (59). Importantly, the surface subject must precede the adverb when the embedded verb is volitive in (58), whereas the relative order between the involitive subject and the adverb in (59) is flexible. Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

(58)

a.

b.

(59)

a.

b.

lal hondətə natannə Lal.NOM well dance.VOL.INF ‘Lal is likely to dance well.’ *hondətə lal natannə well Lal.NOM dance.VOL.INF

15

puluwan. be.likely.to puluwan. be.likely.to

lal-ʈə hondətə nætennə puluwan. dance.INVOL.INF be.likely.to Lal-DAT well ‘Lal is likely to (involuntarily) dance well.’ hondətə lal-ʈə nætennə puluwan. well Lal-DAT dance.INVOL.INF be.likely.to

Our proposal provides a straightforward explanation for the contrast between (58) and (59). Under our analysis, the involitive subject in (59), receiving lexical dative case from the verb, remains in the embedded spec-vP position, and the adverb adjoins to the embedded vP. In addition, we adopt Chomsky’s (1995) Bare Phrase structure that allows for multiple specifiers, where traditional adjuncts can be treated as (additional) specifiers, eliminating the traditional specifier/adjunct distinction. Therefore, there is flexible ordering between the subject and the adverb in (59). In contrast, the volitive subject in (58) cannot get any case in the embedded clause because neither the embedded volitive verb nor the embedded infinitival T is able to assign case. Therefore, the volitive subject is forced to raise to matrix spec-TP to receive structural nominative case. Consequently, the volitive subject at matrix spec-TP must precede the embedded adverb. Next, notice that when puluwan in (58) receives deontic/ability reading as in (60), the surface subject must precede the embedded adverb. (60)

a.

b.

lal-ʈə hondətə natannə Lal-DAT well dance.VOL.INF ‘Lal is able to dance well.’ *hondətə lal-ʈə natannə well Lal-DAT dance.VOL.INF ‘Lal is able to dance well.’

puluwan. be.able.to puluwan. be.able.to

Given our analysis of deontic modals as control verbs, the surface subject in (60) does not originate from the embedded clause, in contrast to the surface subject in (58), which reaches its surface position in the matrix clause via case-driven Amovement. Rather, it is base-generated in the matrix clause as the thematic subject of the deontic modal. Therefore, assuming strict cyclic application of merge (Chomsky, 1995), the surface subject enters the derivation of (60) later than the embedded adverb does, and hence in no derivational stage of (60) does the embedded adverb precede the surface subject. Additional support for our analysis is related to the interpretation of idiom chunks. Idioms are regarded as one unit at some level of lexical-semantic analysis. If some part of an idiom is separated from the rest of it and the idiomatic interpretation is still available, the separation must be a result of movement of an idiom chunk. Thus, Chomsky (1980) notes that care in the idiom ‘‘take care of’’ can be separated from the rest of the idiom chunk, via A-movement, and maintains the idiomatic interpretation, as exemplified in (61). (61)

a. b.

Excellent care was taken of the orphans. Excellent care seems to have been taken of the orphans.

Crucially, he also notes the contrast between (61) and (62a). He argues that (62a) is ungrammatical because unlike (61), excellent care in (62a) is base-generated at its surface position in tough-constructions, but only a movement relation as in (61) can license separation of an idiom chunk (i.e. not the combination of null operator movement and strong binding as in the analysis of tough-constructions). (62b) is the base-line example showing a grammatical non-idiomatic counterpart to (62a). (62)

a. b.

*Excellent care is hard to take of the orphans. Food is hard to take from the orphans.

The example in (63) shows that the Sinhala expression baləla paninnə mall-en eliyə-tə ‘‘the cat got out of the bag’’ also has an idiomatic interpretation ‘‘the secret was revealed’’. Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

16

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

(63)

baləla mall-en eliyə-tə cat.NOM bag-from out-to ‘The cat got out of the bag.’ ‘The secret was revealed.’

pænna’. get.VOL.PAST [literal reading] [idomatic reading]

Next, we embed (63) under an epistemic modal verb and a deontic modal verb, as shown in (64) and (65), respectively. Notice that the idiomatic interpretation survives the epistemic context (64), whereas the deontic context in (65) only allows for the literal (compositional semantic) interpretation. (64)

baləla mall-en eliyə-tə paninnə puluwan. cat.NOM bag-from out-to get.VOL.INF be.likely.to Literal: ‘The cat is likely to get out of the bag.’ Idiomatic: ‘‘The secret is likely to be revealed.’’

(65)

baləla-ʈə mall-en eliyə-tə paninnə puluwan. cat-DAT bag-from out-to get.VOL.INF be.able.to Literal: ‘The cat is able to get out of the bag.’ #Idiomatic: ‘‘The secret is able to be revealed.’

The absence of the idiomatic interpretation in the deontic context follows from our proposal that deontic modals in Sinhala are control verbs. Thus, the surface subject in (65) does not originate from the embedded clause, so no idiom has ever been formed in the derivation of (65). By contrast, the availability of the idiomatic reading in (64) can be treated on a par with Chomsky’s (61). The idiomatic interpretation survives the epistemic context because the separation of baləla from the rest of the idiom mall-en eliyə-tə paninnə is a result of case-driven A-movement. Based on the raising/control distinction, we also argue that the dative case on the surface subject in (52) is a lexical case assigned by the deontic modal. In the rest of this section, we examine the embedding of modal constructions under ECM verbs to provide empirical support for this analysis. As first noted by Sumangala (1991, 1992), Sinhala allows exceptional (accusative) case marking on the embedded subject by ECM verbs like dannəva ‘know’.36 (66)

mamə [gunapalə/gunapalə-wə aava kiyəla] dannəwa. I.NOM Gunapala.NOM/Gunapala-ACC come.VOL.PAST COMP know.VOL.PRES ‘I know (that) Gunapala came.’

Notice that the accusative case on the embedded volitive subject in (66) disappears when either the ECM context is removed as in (67) or the embedding verb dannəva is replaced by a non-ECM verb like dækka ‘see’ as in (68). This confirms the assumption that the accusative case on the embedded volitive subject comes from the matrix ECM verb dannəva ‘know’. (67)

gunapalə/*gunapalə-wə aava. Gunapala.NOM/*-ACC come.VOL.PAST ‘Gunapala came.’

(68)

mamə [gunapalə/*gunapalə-wə aava kiyəla] dækka. I.NOM Gunapala.NOM/*-ACC come.VOL.PAST COMP see.VOL.PRES ‘I saw (that) Gunapala came.’

Interestingly, ECM in Sinhala shows a volitive-involitive asymmetry. The subject of a volitive verb in (66) and (69) can receive accusative case from the ECM verb, while the subject of an involitive verb can take only dative case (70): (69)

36

mamə [eya/eya-wə natənəwa] dannəwa. I.NOM he.NOM/-ACC dance.VOL.PRES know.VOL.PRES ‘I know (that) he voluntarily dances.’

The nominative-accusative alternation in (66) is beyond the scope of this paper, and we leave it for future work.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

(70)

17

mamə [eya-ʈə/*eya-wə/*eya-ʈə-wə nætenəwa] dannəwa. I.NOM he-DAT/*-ACC/*-DAT-ACC dance.INVOL.PRES know.VOL.PRES ‘I know (that) he involuntarily dances.’

The analysis proposed here (as well as the previous analyses reviewed in section 2) can account for this prohibition of ECM on the embedded involitive subject in (70). ECM is not allowed in (70) because the assignment of the lexical dative case to the embedded subject by the involitive verb renders further structural case valuation (in later higher stages of the syntactic derivation) unnecessary and hence inapplicable (possibly due to the Activity Condition in (22)). Now, consider the embedding of epistemic raising modal constructions in (71) under the ECM verb dannəwa ‘know’ as in (72). (71)

a.

b.

(72)

a.

b.

lal potə ga-nnə puluwan. Lal.NOM book.ACC take.VOL-INF be.likely.to ‘Lal is likely to take the book.‘ lal-ʈə potə gæne-nnə puluwan. Lal-DAT book.ACC take.INVOL-INF be.likely.to ‘Lal is likely to (involuntarily) take the book.’ mamə [lal/lal-wə potə ga-nnə puluwan kiyəla] dannəwa. I.NOM Lal.NOM/-ACC book.ACC take.VOL-INF be.likely.to COMP know.PRES ‘I know that Lal is likely to take the book.‘ mamə [lal-ʈə/*lal-wə/*lal-ʈə-wə potə gæne-nnə puluwan kiyəla] dannəwa. I.NOM Lal-DAT/*-ACC/*-DAT-ACC book.ACC take.INVOL-INF be.likely.to COMP know.PRES ‘I know that ‘Lal is likely to (involuntarily) take the book.’

Notice that only the subject of the volitive verb lal in (72a) can bear the accusative case -wə assigned by the ECM verb, whereas the lexical dative case -ʈə on lal can neither be replaced or be followed by -wə, the same restriction seen in (70). Next, given our analysis of the dative case on the surface subject in (73) as a lexical case assigned by the deontic modal, we predict that the surface subject of deontic modals should also resist ECM, for the same reason ruling out ECM in (70) and (72b). This prediction is borne out, as shown by (74). (73)

lal-ʈə potə ga-nnə puluwan. I-DAT book.ACC take.VOL-INF be.able.to ‘Lal is able to take the book.’

(74)

mamə [lal-ʈə/*lal-wə/*lal-ʈə-wə potə ga-nnə puluwan kiyəla] dannəwa. I.NOM Lal-DAT/*-ACC/*-DAT-ACC book.ACC take.VOL-INF be.able.to COMP know.PRES I know that Lal is able to take the book.’

Thus, the resistance of the accusative case marking -wə on the embedded subject of the deontic modal lends support to our proposal that deontic modals in Sinhala assign lexical dative case. 4.3. Case-driven movement What distinguishes our proposal from others is that we maintain that structural case valuation drives A-movement in Sinhala. Given that the accusative case on the embedded subject in (66) is from the ECM verb (as confirmed by (67) and (68)), one clear prediction of our analysis is that only accusative subject, but not the nominative one, moves out of the embedded clause in (66) to the specifier of the verb phrase hosted by the ECM verb. In what follows, we present four pieces of supporting evidence to show that this prediction is borne out. First, notice that the embedded subject cannot be bound by the matrix subject when it bears accusative, as shown by (75). This is because in order for the embedded subject to get accusative case, it has to move to the matrix spec-vP, where its co-reference with the matrix subject would violate Binding Principle B. (75)

rajəthumaj eyai/j/eya-wəi/*j weerəye-k kiyəla] hithənəwa. king.NOM he.NOM/he-ACC hero-INDEF COMP think.VOL.PRES ‘The king thinks that he is a hero.’

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

18

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

Second, the embedded subject can bind the anaphor in the matrix adverbial only when it bears accusative case, as shown by the contrast between (76) and (77). The embedded subject, if assigned nominative, does not move out of the embedded clause, so the embedded nominative subject is not at a position high enough to bind the anaphor in the matrix adverbial, violating Binding Principle A. (76)

rajathuma æməthiwarun-wə wærəd-i kiyəla thamə-thamange ware-di oppu-kəla. guilty-is COMP each-other’s trial-during prove-PAST king.NOM ministers-ACC ‘The king proved the ministers to be guilty during each other’s trial.’

(77) ??rajathuma æməthiwaru wærəd-i kiyəla thamə-thamamge ware-di oppu-kəla. trial-during prove-PAST king.NOM ministers.NOM guilty-is COMP each-other’s ‘The king proved the ministers to be guilty during each other’s trial.’ Third, the placement of adverbs also lends support to our proposal. Notice that the adverb modəkamə-tə ‘foolishly’ in (78) modifies the ECM verb hithənəwa ‘think’, so it should adjoin to the vP hosted by the ECM verb to take scope over the ECM verb. (78)

minissu modəkamə-tə rajəthuma/rajəthuma-wə weerəye-k kiyəla hithənəwa. people.NOM foolish-ly king. NOM/-ACC hero-INDEF COMP think.PRES ‘People foolishly think that the king is a hero.’

Intriguingly, the embedded subject, if marked accusative, could precede the adverb sitting high in the matrix clause, as in (79a). In contrast, the nominative-marked subject must follow the adverb, as shown in (79b). (79)

a. b.

minissu rajəthuma-wə modəkamə-tə weerəye-k kiyəla hithənəwa. people.NOM king-ACC foolish-ly hero-INDEF COMP think.PRES *minissu rajəthuma modəkamə-tə weerəye-k kiyəla hithənəwa. people.NOM king.NOM foolish-ly hero-INDEF COMP think.PRES

This contrast follows from our proposal because the embedded subject, to get accusative case from the ECM verb in the matrix clause, must raise to spec-vP. Besides, the flexible relative order of the accusative subject with the matrix adverb can be explained if we follow Chomsky’s (1995) proposal of Bare Phrase structure with multiple specifiers, where traditional adjuncts are treated as (additional) specifiers. Consequently, the traditional specifier-adjunct distinction can be eliminated under Bare Phrase structure. Therefore, there is flexible ordering between the (raised) accusative subject and the matrix adverb. Finally, additional evidence comes from Long Distance Scrambling (LDS) in Sinhala. Similar to Japanese (see Saito, 1985; Tanaka, 2002), Sinhala does not allow LDS of an embedded subject to the matrix clause initial position, as shown in (80). But surprisingly, this restriction does not apply to the accusative-marked subject as shown in (81). (80)

(81)

*eyai rajəthuma [ti weerəye-k kiyəla] hithənəwa. hero-INDEF COMP think.PRES he.NOM king.NOM ‘The king thinks that he is a hero.’ eya-wəi rajəthuma [ti weerəye-k kiyəla] hithənəwa. he-ACC king.NOM hero-INDEF COMP think.PRES ‘The king thinks that he is a hero.’

One possible explanation for this contrast based on our proposal is that the embedded subject moves out of the embedded clause for accusative case, and thereby escapes the restriction of LDS of embedded subjects. Before ending this section, a clarification is in order. One may argue that the syntactic properties exhibited by the embedded accusative subject in ECM contexts do not necessarily result from syntactic movement from the embedded clause to the matrix clause. Rather, the accusative subject might be base-generated at the matrix clause, along the lines suggested by Wechsler’s (1995) analysis of Korean ECM. Therefore, the nominative-accusative contrast seen in this section would not constitute conclusive evidence for case-driven movement in Sinhala. We maintain that this basegeneration analysis is not tenable in Sinhala. First, note that the scopal ambiguity of volitive subjects in relation to the negation seen in (20) also holds in the ECM context, as shown in (82). Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

(82)

19

siri lamai hæməkena-mə nætuwe næhæ kiyəla dannəwa. Siri.NOM children all.NOM-EMP danced.VOL not COMP know.PRES ‘Siri knows that all children did not dance.’ [Total negation = all > negation] [Partial negation = negation > all]

Next, if the accusative subject in the ECM context were base-generated in the matrix clause, we predict that it would not scopally interact with the embedded negation. However, this prediction is not borne out, as evidenced by (83). (83)

siri lamai hæməkena-wə-mə nætuwe næhæ kiyəla dannəwa. Siri.NOM children all-ACC-EMP danced.VOL not COMP know.PRES ‘Siri knows that all children did not dance.’ [Total negation = all > negation] [Partial negation = negation > all]

Therefore, the accusative subject in (83) apparently originates from the embedded vP (where it scopes below the negation), and then raises to the matrix clause for case, thereby yielding the total negation interpretation. 5. Further differences between volitive and involitive subjects In this section, we present additional data concerning the syntactic differences between dative subjects with involitives and nominative subjects with volitives. We investigate conjunction reduction, quantifier floating, and control phenomena. These data suggest that these two types of subjects are not equivalent, and that involitive dative subjects show less subject-hood properties, but we leave the comprehensive and formal integration of these data into our proposed analysis for future work. One piece of evidence showing that involitive subjects exhibit less subject-hood properties than volitive subjects comes from conjunction reduction, a subject-hood diagnostic used in Sigurðsson (2004) and Bayer (2004) for Icelandic. The following example from Sigurðsson (2004:142) shows that a non-nominative NP can be the elided argument in a conjunct, implying that it functions as a subject in the coordinated clause. (84)

Hu´n var syfjuð og (henni) leiddist bokin. She.NOM was sleepy and (her.DAT) bored book.the.NOM ‘She was sleepy and found the book boring.’

But in German, in which non-nominative NPs exhibit less subject-hood properties (see Bayer, 2004:57), such reduction yields ungrammaticality. (85)

*Ich war hungrig und (*mich) hat gefroren. I.NOM was hungry and me.ACC has frozen. ‘I was hungry and was cold.’

We observe that Sinhala involitive dative subjects pattern with non-nominative subjects in German: even though conjunction reduction is allowed in a volitive construction (86), the elision of the external argument in an involitive construction (87) yields ungrammaticality. (86)

lal kapət-ii eth (lal) boru kiyanne næhæ. Lal.NOM cunning-is but Lal.NOM lie say.VOL.PRES not ‘Lal is cunning but (he) does not lie.‘

(87)

*lal kapət-ii eth (*lal-ʈə) boru kiyəwenne næhæ. Lal.NOM cunning-is but Lal-DAT lie say.INVOL.PRES not ‘Lal is cunning but does not lie.’

When the verb is in the volitive mood (86), the sentence is grammatical either with or without the deletion of the subject in the conjoined clause. But in contrast, (87) with the involitive verb requires the presence of an overt subject NP, implying that non-nominative case marked NPs in Sinhala exhibit less subject-hood properties. Another piece of evidence that shows involitive subjects exhibit less subject-hood properties is quantifier floating (QF). As Henadeerage (2002:50) observes, in colloquial Sinhala QF applies to both subjects (88) and objects (89). Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

20

(88)

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

a.

b.

(89)

a.

b.

ape iskoo-le guruwəru okko-mə honday-ii. we.GEN school-LOC teacher.PL all.NOM-EMP good-are ‘All the teachers in our school are good.’ guruwərui ape iskoo-le ti okko-mə honday-ii. teacher.PL we.GEN school-LOC all.NOM-EMP good-are ‘All the teachers in our school are good.’ amma nangi-ʈə topi huŋgak dunna. mother.NOM younger.sister-DAT toffee.ACC a lot give.PAST ‘Mother gave a lot of toffees to the younger sister. amma topii nangi-ʈə ti huŋgak dunna. mother.NOM toffee.ACC younger.sister-DAT a lot give.PAST ‘Mother gave a lot of toffees to the younger sister.’

More importantly, we observe that QF, though allowed with volitive subjects (90), is generally disfavored in involitive constructions, as illustrated by (91):37,38 (90)

a.

b.

(91)

adə lamai okko-mə nætuwa. today children all.NOM-EMP dance.VOL.PAST Today, all children (voluntarily) danced.’ lamaii adə ti okko-mə nætuwa. children. today all.NOM-EMP dance.VOL.PAST ‘Today, all children (voluntarily) danced.’

a.

adə lamai okko-ʈə-mə nætuna today children all-DAT-EMP dance.INVOL.PAST Today, all children (involuntarily) danced.’ b. ??lamaii adə ti okko-ʈə-mə nætuna. children today all-DAT-EMP dance.INVOL.PAST ‘Today, all children (involuntarily) danced.’

Last, non-nominative subjects cannot be controlled in some languages such as German and Hindi/Urdu (though allowed in languages like Icelandic, see Sigurðsson, 2004). For instance, Davison (2004) shows that in Hindi/Urdu, a lexically dative marked subject in (92) cannot be controlled as in (93): (92)

(93)

Mujhee aisaa paisaa mil ga-yaa. I.DAT such money.NOM get go-PF.M.SG ‘I got such money.’ *mãi ̃ [PRO aisaa paisaa mil-naa] nahiĩ ̃ caah-tii hũũ. such money.NOM get-INF not want-IMPF.F am I.NOM ‘I don’t want [PRO to get such money].’

The following examples show that dative subjects in Sinhala pattern together with non-nominative subjects in Hindi-Urdu and German: the dative subject in (94) cannot be controlled as in (95).39

37 Kariyakarawana (1998:62, 74), citing the example in (i), maintains that quantifier floating is allowed in some involitive constructions in Sinhala. However, our native speakers strongly disfavored QF in involitive constructions. For an observation similar to ours, see Henadeerage (2002).

(i)

lamayi adə iskoledi ti okko-ʈə-mə diuna. children today school-LOC all-DAT-EMP run.PAST.INVOL ‘Today in school all children ran.’ 38 As one reviewer points out, the contrast in QF does not necessarily tell us anything about subjecthood properties. It could just be the case that QF is bad with case-marked DPs, so it might have nothing to do with the subject/non-subject contrast. 39 Recall that in footnote 30, we provide a semantic/pragmatic account for why involitive verbs cannot be embedded under deontic modals. As one reviewer points out, given that non-nominative subjects cannot be controlled, we have an additional, syntactic explanation for why deontic modals (analyzed as control verbs here) cannot embed involitive verbs: involitive verbs have non-nominative subjects, and such positions cannot host PRO.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

(94)

(95)

21

ma-ʈə thægi læbuna. I-DAT presents.ACC receive.INVOL.PAST ‘I involuntarily received presents.’ *mamə [PRO thægi læbennə] balaporoththu-wenne næhæ presents.ACC receive.INVOL.PAST hope-PRES not I.NOM ‘I do not hope to involuntarily receive presents.’

In sum, the three diagnostics explored in this section suggest that volitive and involitive subjects in Sinhala exhibit different subject-hood properties. Our conjecture is that this asymmetry, as we have proposed in this paper, results from the fact that they occupy two distinct structural positions in syntax. We leave the detailed analysis of the volitive/involitive contrast with respect to conjunction reduction, QF, and control for our future research. 6. Summary and theoretical implications In this paper, we examine previous analyses of subject case marking in Sinhala that characterize nominative case as the default case that arises when the semantic conditions for all available quirky cases fail. We argue that nominative case in Sinhala is instead a structural case assigned by a finite T, and the need for structural case valuation drives the movement of subject NPs of volitive verbs to spec-TP. Three important implications of our proposal are that (i) there is no universal EPP requirement on T in Sinhala (contra Gair, 1990a,b), and (ii) case-valuation can be the driving force of A-movement to spec-TP, and (iii) case-valuation can be dissociated from w-valuation. We have seen how Gair’s EPP-based account of A-movement in Sinhala yields incorrect predictions about the scopal interpretation of subject quantifiers. In this section, we would like to discuss the second and third implications about the motivation of A-movement and the mechanics of case valuation. There are several proposals concerning a motivation for A-movement in the generative literature. Three distinct driving forces identified by different researchers that are responsible for triggering A-movement to spec-TP include (i) the unvalued w-features on T (e.g., Kuroda, 1988; Miyagawa, 2005), (ii) the EPP feature on T (e.g. Chomsky, 1995, 2000, 2001; Lasnik, 1995, 1999, 2001; Nevins, 2005), (iii) the unvalued Case feature on an NP (e.g. Epstein and Seely, 1999, 2006; Alexiadou and Anagnostopoulou, 1998; Bosˇ kovic´, 2002, 2007). The data examined in this paper lead us to conclude that neither EPP nor w-features on T triggers A-movement in Sinhala. Recall that Gair (1990b:142) insightfully maintains that ‘‘AGR plays no role in Sinhala, to the extent that there is no element within INFL that plays its subject case assignment role’’. Here, we provide an additional piece of evidence for the lack of agreement features on T in Sinhala. Note that as is well-known, English does not allow subject reflexives in a finite clause as illustrated in (96): (96)

*John thinks that himself is hard-working.

There are various proposals regarding the ungrammaticality of (96) (see e.g. Chomsky, 1981, 1986; Everaert, 1991; Lebeaux, 1988; Kayne, 1984; Picallo, 1985; Rizzi, 1990; Woolford, 1999; Despic´, 2011). For expository purposes, we focus on one particular set of proposals whose central unifying idea is to attribute the ungrammaticality of (96) to the fact that anaphors cannot be agreed with (cf. Rizzi, 1990, following Picallo, 1985). In particular, Rizzi (1990) proposes the anaphor agreement effect in (97) and argues that the reason why anaphors are barred from the subject position of tensed clauses as in (96) is that anaphors cannot agree. (97)

The anaphor agreement effect40 Anaphors do not occur in syntactic positions construed with agreement.

Citing Icelandic and Italian data, Rizzi (1990) argues that neither nominative Case (see Brame, 1977; Koster, 1978; Anderson, 1982; Maling, 1984; Everaert, 1991) nor the subject position (see Kayne, 1984; Chomsky, 1986) accounts for the ban on subject anaphors. Rather, it is agreement that causes the ungrammaticality (cf. Chomsky, 1981; George and Kornfilt, 1981; Johnson, 1985; Picallo, 1985). He maintains that the anaphor agreement effect ‘‘holds quite systematically 40 There are several analyses aiming at deriving (97). For example, Chomsky (1981:209) regards agreement on T as an accessible SUBJECT for the purposes of determining the binding domain, so a subject anaphor exhibiting agreement must be bound by the agreement on T; however, this leads to an i-within-i violation where the subject anaphor and the agreement on T enter into an infinitive regression relation due to dependence on each other for reference (see also Johnson, 1985 for a similar proposal).

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

22

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

in natural languages’’ (Rizzi, 1990:26). Building upon Rizzi’s work, Woolford (1999) argues that the anaphor agreement effect is universal and can be a diagnostic for the presence or absence of (covert) agreement.41 With this conclusion, consider the grammatical occurrence of subject anaphors in (98). (98)

a.

b.

siri [thaman awanka-i kiyala] hitənəwa. Siri self.NOM honest-is COMP think.PRES ‘Siri thinks that himself is honest.’ mala [thaman parə dannəwa kiyəla] kiwwa. Mala self.NOM way.ACC. know.PRES COMP say.PAST ‘Mala said that self knows the way.’

A clarification is in order before we proceed. Unlike Chinese (see Huang and Liu, 2001), Sinhala does not have the barecompound distinction of the morphological form of reflexives. It only has the bare reflexive thaman ‘self’. Accordingly, one may wonder whether thaman, as an embedded subject, is used as a logophor whose distribution has nothing to do with the presence or absence of w-features on T, and hence sentences in (98) do not constitute an argument for the lack of w-features on T in Sinhala. Note that Huang and Liu (2001) show that the bare reflexive ziji in Chinese can be used either as a logophor or an anaphor. Importantly, when occurring as an embedded subject, ziji is not subject to various logophoric conditions (e.g. under a de se scenario). This observation carries over to the bare reflexive thaman ‘self’ in Sinhala. The sentences in (98) can be uttered under a non-de se scenario in which the coreference between thaman and Siri/Mala is reported purely as speaker’s knowledge from the speaker’s own perspective. This indicates that thaman in (98) can be an anaphor, rather than a logophor, and thus its grammatical occurrence as an embedded subject constitutes an argument against the presence of w-features on T in Sinhala. Given that neither EPP nor w-features on T is responsible for the A-movement to spec-TP in Sinhala, we conclude that case-feature valuation alone can motivate Amovement.42 One important consequence of adopting the case-driven approach to A-movement in languages without w-features on T is that the valuation of the case feature on an NP cannot universally be a ‘‘reflex’’ of w-feature valuation between T and the NP contra Chomsky’s (2000, 2008) system of feature valuation. Chomsky (2000, 2008) has to postulate such reflex valuation for the case feature precisely because he assumes the biconditional of feature valuation and feature interpretability in (99). (99)

Valuation/Interpretability Biconditional (Chomsky, 2001:5) A feature F is uninterpretable [at the Conceptual-Intentional/CI interface, CTC and SH] iff F is unvalued [in the lexicon, CTC and SH].

Given that the case features are clearly uninterpretable on both T and the NP, they are also unvalued under (99). As a result, there can be no direct valuation relation between T and the NP for the case feature because there exists no ‘‘value’’ for the case feature in the relation between T and the NP, and hence case valuation can only be tied to the other probegoal relation between T and the NP, i.e., the w-feature valuation relation. The postulation of the reflex valuation can be eliminated if we abandon (99) and adopt Bosˇ kovic´’s (2011) approach to case valuation in (100).

41 Therefore, the contrast between (96) and other examples showing subject anaphors in the absence of agreement can be compared to that between (ia) and (ib) in English.

(i)

a. b.

*John believes himself is smart. John believes himself to be smart.

The subject anaphor himself is allowed in (ib) precisely because of the lack of w-features on T (morphologically realized by to) of the embedded TP selected by the ECM verb believe. 42 Notice that the subject anaphor thaman is able to bear other types of case, including lexical dative case from involitive verbs (i), and structural accusative case from ECM verbs (ii). (i)

(ii)

mala [thaman-ʈə nætenəwa kiyəla] kiwwa. Mala self-DAT dance.INVOL.PRES COMP say.PAST ‘Mala said that herself (involuntarily) dances.’ mamə [thaman/thaman-wə parə dannəwa kiyəla] dannəwa I.NOM self.NOM/-ACC way know.VOL.PRES COMP know.VOL.PRES ‘I know myself to know the way.’

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

(100)

a. … TuCase[Nom] b. [TD$INLE] …NPuCase[Nom]

……

23

NPuCase[ ] TuCase[Nom]…… tNP

probing and valuation

Importantly, Bosˇ kovic´ (2011) notes that even though the case feature is uninterpretable on both finite T and an NP, the case feature on the finite T is valued, whereas the case feature on an NP is unvalued. The case feature on the finite T is valued because it always assigns nominative case; by contrast, the case feature on an NP is unvalued because it depends on its syntactic context (e.g., it can be valued as accusative if it is valued by v*). Bosˇ kovic´ (2011) combines this novel featural characterization of case valuation, as illustrated in (100a), with his (2007) moving-element-driven approach to movement, and argues that the unvalued case feature on the NP triggers the NP’s A-movement to spec-TP, where the NP can c-command and probe the valued Case feature on T as in (100b). We maintain that the w-independent case-driven A-movement in Sinhala lends support to Bosˇ kovic´’s (2011) system of case valuation. Acknowledgements We are much indebted to Samuel D. Epstein, Acrisio Pires, and three anonymous reviewers for their rigorous and penetrating comments that have led to significant improvement of different versions of this paper. Portions of this research have been presented at the 3rd Workshop on the Formal Approach to South Asian Languages, GLOW in Asia IX, the 10th Workshop on Formal Syntax and Semantics in Taiwan, and the 33rd West Coast Conference on Formal Linguistics. We would like to thank the audiences at those venues for their encouragement and stimulating questions, especially Rajesh Bhatt, Dylan Tsai, Niina Zhang, and Roger Liao. We gratefully acknowledge the financial support from the Predoctoral Fellowship from the University of Michigan and the research grant MOST-104-2410-H-038-001 from the Ministry of Science and Technology in Taiwan. Finally, we take sole responsibility for any remaining errors or inconsistencies. References Alexiadou, A., Anagnostopoulou, E., 1998. Parameterizing word order, V-movement, and EPP-checking. Nat. Lang. Linguist. Theory 16, 491--539. Anand, P., Nevins, A., 2006. The locus of ergative case assignment: evidence from scope. In: Johns, A., Massam, D., Ndayiragije, J. (Eds.), Ergativity: Emerging Issues. Kluwer, Dordrecht, pp. 3--25. Anderson, S., 1982. Types of dependency in anaphors: Icelandic (and other) reflexives. J. Linguist. Res. 2, 1--22. Bayer, J., 2004. Non-nominative subjects in comparison. In: Bhaskararao, P., Subbarao, K. (Eds.), Non-nominative Subjects, vol. 1. John Benjamins, Amsterdam, pp. 49--76. Beavers, J., Zubair, C., 2008. Non-nominative subjects and the involitive construction in Sinhala. Paper presented at the 82nd annual meeting of the Linguistic Society of America, Chicago, IL, January 3--6. Beavers, J., Zubair, C., 2010. The interaction of transitivity features in the Sinhala involitive. In: Brandt, P., Garcia, M. (Eds.), Transitivity: Form, Meaning, Acquisition, and Processing. Benjamins, Amsterdam, pp. 69--92. Beavers, J., Zubair, C., 2013. Anticausatives in Sinhala: involitivity and causer supression. Nat. Lang. Linguist. Theory 31, 1--46. Beghelli, F., Stowell, T., 1997. Distributivity and negation. In: Szabolsci, A. (Ed.), Ways of Scope Taking. Kluwer. Boeckx, C., 2001. Scope reconstruction and A-movement. Nat. Lang. Linguist. Theory 19. Bosˇ kovic´, Z., 2002. A-movement and the EPP. Syntax 5, 167--218. Bosˇ kovic´, Zˇ ., 2007. On the locality and motivation of move and agree: an even more minimal theory. Linguist. Inquiry 38, 589--644. Bosˇ kovic´, Zˇ ., 2011. On valued uninterpretable features. In: Proceedings of NELS, p. 39. Brame, M., 1977. Alternatives to the tensed S and specified subject conditions. Linguist. Philos. 1, 381--411. Büring, D., Gunlogson, C., 2000. Aren’t Positive and Negative Questions the Same? UCLA/UCSC Manuscript. . Chandralal, D., 2010. Sinhala. John Benjamins Publishing Company, Amsterdam. Cheng, L., 1991. On the Typology of Wh-questions Doctoral dissertation. MIT. Chomsky, N., 1980. Rules and Representations. Columbia University Press, New York. Chomsky, N., 1981. Lectures on Government and Binding. Foris Publications, Dordrecht. Chomsky, N., 1986. Knowledge of Language. Praeger, New York. Chomsky, N., 1995. The Minimalist Program. MIT Press, Cambridge. Chomsky, N., 2000. Minimalist inquiries: the framework. In: Martin, R., Michaels, D., Uriagereka, J. (Eds.), Step by Step: Essays in minimalist syntax in honor of Howard Lasnik. MIT Press, Cambridge, MA. Chomsky, N., 2001. Derivation by phase. In: Michael, K. (Ed.), Ken Hale: A Life in Language. MIT Press, Cambridge, MA, pp. 1--52. Chomsky, N., 2008. On phases. In: Freidin, R., Otero, C.P., Zubizarreta, M.L. (Eds.), Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roger Vergnaud. MIT Press, Cambridge, pp. 133--166. Davison, A., 2004. Non-nominative subjects in Hindi/Urdu: VP structure and case parameters. In: Subbarao, K.V., Bhaskararao, P. (Eds.), Nonnominative Subjects, vol. 2. John Benjamins, Amsterdam, pp. 141--168. Despic´, M., 2011. Syntax in the Absence of Determiner Phrase Doctoral dissertation. University of Connecticut, Storrs. Dissanayake, J.B., 1976. National Languages of Sri Lanka 1 -- Sinhala. Department of Cultural Affairs, Colombo.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012

+ Models

LINGUA-2406; No. of Pages 24

24

C.-T. Tim Chou, S. Hettiarachchi / Lingua xxx (2016) xxx--xxx

Epstein, S.D., Seely, T.D., 1999. SPEC-ifying the GF ‘‘subject’’: eliminating A-chains and the EPP within a derivational model. Manuscript. University of Michigan and Eastern Michigan University. Epstein, S.D., Seely, T.D., 2006. Derivations in Minimalism. Cambridge University Press, Cambridge. Everaert, M., 1991. Nominative anaphors in Icelandic: morphology or syntax. In: Abraham, W., Kosmeijer, W., Reuland, E.J. (Eds.), Issues in Germanic Syntax. Mouton de Gruyter, Berlin, pp. 277--305. Gair, J.W., 1990a. Subjects, case and INFL in Sinhala. In: Verma, M., Mohanan, K.P. (Eds.), Experiencer Subjects in South Asian Languages. Center for the Study of Language and Information, Palo Alto. (Reprinted in Gair 1998, pp. 65--86), pp. 13--41. (Reprinted in Gair 1998, pp. 65--86). Gair, J.W., 1990b. Studies in South Asian Linguistics: Sinhala and Other South Asian Languages. Oxford University Press, USA. Gair, J.W., Paolillo, J.C., 1997. Sinhala. Lincom Europa, Munchen. Gair, J.W., Wali, K., 1998. On distinguishing AGR from agr: evidence from south Asia. In: Gair, J.W., Lust, B.C. (Eds.), Studies in South Asian linguistics: Sinhala and Other South Asian Languages. Oxford University Press, pp. 140--152. George, L., Kornfilt, J., 1981. Finiteness and boundedness in Turkish. In: Heny, F. (Ed.), Binding and Filtering. MIT Press, Cambridge, pp. 105--127. Gunasekara, D.M., 1999. A Comprehensive Grammar of the Sinhalese Language. Asian Educational Services, New Delhi. Henadeerage, D.K., 2002. Topics in Sinhala Syntax Doctoral dissertation. The Australia National University, Canberra. Hornstein, N., 1995. Logical form. Blackwell. Huang, C.-T.J., Liu, C.-S.L., 2001. Logophoricity, Attitudes and Ziji at the Interface. In: Cole, P., et, al. (Eds.), Long Distance Reflexives, Syntax and Semantics, vol. 33. Academic Press, New York, pp. 141--195. Inman, M., 1992. Semantics and Pragmatics of Sinhala Involitives Doctoral dissertation. Stanford University. Jany, C., 2005. The relationship between case marking and S, A, and O in spoken Sinhala. In: Proceedings from the workshop on Sinhala linguistics, . Johnson, K., 1985. Some notes on subjunctive clauses and binding in Icelandic. In: MIT Working Papers in Linguistics 6: Papers in Theoretical and Applied Linguistics. MIT Press, Cambridge, pp. 102--134. Johnson, K., Tomioka, S., 1997. Lowering and mid-size clauses. In: Katz, G., Kim, S., Haike, W. (Eds.), Tübingen Workshop on Reconstruction. Kahr, J.C., 1989. Discourse Functions of the Sinhala Passive Doctoral dissertation. Harvard University. Karimi, S., 2005. A Minimalist Approach to Scrambling: Evidence from Persian. Mouton de Gruyter, Berlin. Kariyakarawana, S.M., 1998. The Syntax of Focus and Wh-questions in Sinhala. Karunaratne & Sons Ltd., Colombo. Kayne, R., 1984. Connectedness and Binary Branching. Foris, Dordrecht. Kidwai, A., 2000. XP-adjucntion in Universal Grammar: Scrambling and Binding in Hindi-Urdu. Oxford University Press, New York. Kishimoto, H., 2005. Wh-in-situ and movement in Sinhala questions. Nat. Lang. Linguist. Theory 25, 1--51. Kiss, K., 2006. Quantifier scope ambiguities. In: Everaert, M., Van Riemsdijk, H., Goedemans, R., Hollebrandse, B. (Eds.), The Blackwell Companion to Syntax, vol. 1. . Koster, J., 1978. Locality Principles in Syntax. Foris, Dordrecht. Kuroda, S.Y., 1988. Whether we agree or not: a comparative syntax of English and Japanese. Lingvist. Investig. 12, 1--47. Ladd, R., 1981. A first look at the semantics and pragmatics of tag questions and negative questions. In: Chicago Linguistics Society, . Lasnik, H., 1995. Case and expletives revisited: on Greed and other human failings. Linguist. Inquiry 26, 615--633. Lasnik, H., 1999. Chains of arguments. In: Epstein, S.D., Hornstein, N. (Eds.), Working Minimalism. MIT Press, Cambridge, pp. 189--215. Lasnik, H., 2001. A note on the EPP. Linguist. Inquiry 32, 356--362. Lebeaux, D., 1988. Language Acquisition and the Form of the Grammar Doctoral dissertation. University of Massachusetts, Amherst. Maling, J., 1984. Non-clause-bounded reflexives in Modern Icelandic. Linguist. Philos. 7, 211--241. May, R., 1985. Logical Form: Its Structure and Derivation. MIT Press, Cambridge. Miyagawa, S., 2005. On the EPP. In: McGinnis, M., Richards, N. (Eds.), Perspectives on Phases, MIT Working Papers in Linguistics 49. MIT Press, Cambridge, pp. 201--236. Nevins, A., 2005. Derivations without the activity condition. In: MIT Working Papers on Linguistics 49. , pp. 283--306. Nevins, A., Anand, P., 2003. Some agreement matters. In: Gardner, G., Tsujimura, M. (Eds.), WCCFL 22 Proceedings, pp. 370--383. Picallo, M.C., 1985. Opaque domains Doctoral dissertation. The Graduate Center, City University of New York. Premaratne, A.C., 1986. The verb in early Sinhalese Doctoral dissertation. University of London. Rizzi, L., 1990. On the anaphor agreement effect. Riv. Linguist. 2, 27--42. Saito, M., 1985. Some Asymmetries in Japanese and their Theoretical Implications Doctoral dissertation. MIT. Sigurðsson, H.A., 2004. Icelandic non-nominative subjects: facts and implications. In: Subbarao, K.V., Bhaskararao, P. (Eds.), Non-nominative Subjects, vol. 2. John Benjamins, Amsterdam, pp. 137--160. Sumangala, L., 1991. Inner and outer subjects in Sinhala. Cornell Working Papers in Linguistics 9. Cornell University. Sumangala, L., 1992. Long-distance dependencies in Sinhala: The synatx of focus and wh-quesitons Doctoral dissertation. Cornell University. Tanaka, H., 2002. Raising to object out of CP. Linguist. Inquiry 33, 637--652. Ura, H., 2000. Checking Theory and Grammatical Functions in Universal Grammar. Oxford University Press. von Stechow, A., Penka, D., 2003. N-words and negation in German. Presented at Datenvielfalt und Perspecktivenvielflat zwischen Universalität und Variation, Tübingen, January 2003. Wechsler, S., 1995. Korean ECM: semantic and pragmatic Factors. Paper read at the 69th Annual Meeting of the Linguistic Society of America, New Orleans, January 6. Woolford, E., 1997. Four way case systems: ergative, nominative, objective and accusative. Nat. Lang. Linguist. Theory 15, 181--227. Woolford, E., 1999. More on the anaphor agreement effect. Linguist. Inquiry 30, 257--287.

Please cite this article in press as: Tim Chou, C.-T., Hettiarachchi, S., On Sinhala subject case marking and Amovement. Lingua (2016), http://dx.doi.org/10.1016/j.lingua.2015.12.012