North East Linguistic Society ¦¢ MIT ⋅ ìÕ October óþÕ¦

Deconstructing quirky subjects Ethan Poole University of Massachusetts Amherst [email protected]

(Õ) a. Can bind subject-oriented anaphora Main claim Hennii þykir [bróðir sinni ] leiðinlegur DPs exhibit a set of subjecthood properties as a function of how high they her.oZ± thinks brother.•™“ her.§u€ boring raise in the functional sequence. ‘Shei thinks heri brother boring’ [Zaenen et al. Õɘ¢:¦¢þ] Empirical motivation b. Can be PRO Quirky subjects in Icelandic, German, Hindi, Basque, and Laz show that the Égi vonast til [PROi að vanta ekki peninga ] subjecthood properties exhibited by DPs obey the Quirky Subject Hierarchy: I.•™“ hope for PRO.Zhh to lack not money.Zhh binding ≫ PRO ≫ reduced relatives ‘I hope not to lack money’ [Zaenen et al. Õɘ¢:¦¢¦] Analysis in a nutshell  Hindi • Tý is decomposed into two separate heads, Persp(ective)ý to host PRO and Hindi quirky subjects can bind subject-oriented anaphora, but cannot be PRO B(inding)ý to bind anaphora, where fseq = ⟨C ≻ Persp ≻ B ≻ v ≻ V⟩. (e.g. Mohanan ÕÉɦ; Bhatt óþþì): • DPs dišer in (i) how high they raise in the functional sequence and (ii) (ó) a. Can bind subject-oriented anaphora whether they can undergo projecting movement from [Spec, PerspP], the Mujhei [apnei sab rishtedar¯ ] pasand hε˜ highest syntactic position, to form a reduced relative. I.oZ± §u€.“.£ all relative.“ like be.£§«.£ ý ý [NP [PerspP Perspective [BP Binding [ QS ...]]]] ‘I like all my relatives’ [Bhatt óþþì:Õó] b. Cannot be PRO *Ravii [PROi Rina pasand a-n¯ a¯ ] nah˜¯ı Õ Introduction Ravi.•™“ PRO.oZ± Rina.•™“ like come-†•€ •u chah-t¯ a.¯ • Quirky (nonnominative) subjects dišer across languages in whether they display want-„Zf.“.« the full range of properties exhibited by canonical nominative subjects. To Intended: ‘Ravi doesn’t want to like Rina’ illustrate, consider Icelandic and Hindi quirky subjects. • e consensus in the literature is that quirky subjects in languages like Icelandic  Icelandic are “true subjects”, whereas quirky subjects in languages like Hindi are not (e.g. Icelandic quirky subjects are well-known for exhibiting all the properties canon- McFadden óþþ¦; Preminger óþÕÕ, óþÕ¦). ically ascribed to subjects other than bearing nominative case and controlling • e subjecthood status of a quirky subject is determined by a suite of subject- verb agreement, such as binding subject-oriented anaphora and being PRO hood diagnostics, e.g. the diagnostics of Zaenen et al. (Õɘ¢). (Andrews ÕÉßä; Þráinsson ÕÉßÉ; Zaenen et al. Õɘ¢; Sigurðsson ÕɘÉ):

Õ • However, it is clear that quirky subjects in languages like Hindi are not normal quirky subjects is the result of quirky subjects dišering across languages in how obliques. ey possess extra properties that are typical of subjects, but atypical high they raise in the functional sequence. of obliques, such as the ability to bind subject-oriented anaphora. • Structure of this talk • Moreover, given the importance of quirky subjects in our understanding of Õ. I present a suite of crosslinguistic subjecthood diagnostics: binding of subject- the interaction between case, agreement, and movement, we do not want to oriented anaphora, PRO, and reduced relatives. base our theory on something like Icelandic quirky subjects if they might be typologically rare. ó. Applying these diagnostics to quirky subjects in Icelandic, German, Hindi, Basque, and Laz shows that the subjecthood properties exhibited by quirky ⇒ Question subjects constitute an implicational hierarchy, called the Quirky Subject How are subjecthood properties distributed amongst quirky subjects crosslin- Hierarchy (QSH). guistically? For example, one could imagine either of the following distributions: ì. I implement the QSH in the syntax by proposing the Height Conjecture: DPs exhibit subjecthood properties as a function of how high they raise in the functional sequence. ¦. I discuss how the empirical generalisations encoded in the QSH are incom- pÔ patible with the existing views of subjecthood in the literature.

pç pò pÔ ó Subjecthood diagnostics • Standard subjecthood diagnostics pò pç e standard suite of subjecthood diagnostics comes from Zaenen et al. (Õɘ¢) and includes four crosslinguistic tests for subjecthood: Õ. Binding of subject-oriented anaphora • To address this question, I investigated quirky subjects in Icelandic, German, ó. Being PRO Hindi, Basque, and Laz. I used three subjecthood diagnostics: binding of subject-oriented anaphora, being PRO, and undergoing relativisation in reduced ì. Raising-to-object (ECM) relatives. ¦. Conjunction reduction

✳ Quirky Subject Hierarchy • ese diagnostics (or a subset of them) are adopted in the recent literature on e pattern to emerge from the data is that the subjecthood properties crosslin- subjecthood (e.g. Sigurðsson ÕɘÉ, óþþó, óþþ¦; Mohanan ÕÉɦ; Fanselow óþþó; guistically exhibited by quirky subjects obey an implicational hierarchy: Barðdal óþþó; Barðdal & Eyþórsson óþþìa,b, óþþ¢; Bayer óþþ¦; McFadden óþþ¦; Gutiérrez-Bravo óþþä; Rákosi óþþä; Wunderlich óþþÉ; Preminger óþÕÕ, óþÕ¦). (ì) Q¶†§Ží S¶fuh± H†u§Z§h„í binding ≫ PRO ≫ reduced relatives • Moreover, following Zaenen et al. (Õɘ¢), the de facto assumption in the literature is that nominative case and controlling verb agreement are not properties of ⇒ Question that follows subjects. How do we account for this implicational hierarchy?

✳ Answer in this talk A DP exhibits a set of subjecthood properties as a function of how high it raises in the functional sequence. e crosslinguistic variation in the behaviour of

ó • Section outline (˜) Arbitrarily-controlled I select three tests to form a revised suite of subjecthood diagnostics: (i) binding a. [ PROarb to hug Fernanda ] is fantastic of subject-oriented anaphora and (ii) being PRO, taken from Zaenen et al. (Õɘ¢),

in addition to (iii) undergoing relativisation in reduced relatives.Ö b. *[ (Fernanda) to hug PROarb ] is fantastic ⇒ erefore, PRO provides a crosslinguistic subjecthood diagnostic: ó.Õ Binding of subject-oriented anaphora (É) PROD†Z•™«±†h • Many languages have a special subclass of anaphora whose antecedent must If XP can be PRO, XP is a subject. be the subject, which are called «¶fuh±-™§†u•±uo Z•Z£„™§Z (SOAs). English is not such a language, but Danish is: ó.ì Reduced relatives (¦) Danish at Peteri fortalte Michael j om sig selvi~∗ j • In reduced relative clauses (RRCs), the relativised element can only occur in that Peter told Michael about §u€ the subject position: ‘that Peteri told Michael j about himself ~∗ ’ [Vikner Õɘ¢:Õþ] i j (Õþ) a. the Basquei [ ___i giving Stefan the rutabaga ] sig • ere is some variation in whether SOAs need bound locally (e.g. Danish b. *the Germani [ Jon Ander giving ___i the rutabaga ] selv) or nonlocally (e.g. Danish sig), or either (e.g. Finnish itse), but they are crucially always bound by a subject. c. *the rutabagai [ Jon Ander giving Stefan ___i ]

⇒ erefore, SOAs provide a crosslinguistic subjecthood diagnostic: ⇒ erefore, RRCs provide a crosslinguistic subjecthood diagnostic:

(¢) B†•o†• D†Z•™«±†h (ÕÕ) Ruo¶huo RuZ±†êu D†Z•™«±†h If XP can bind subject-oriented anaphora, XP is a subject. If XP can be relativised on in reduced relatives, XP is a subject. • Whereas the Binding Diagnostic and the PRO Diagnostic are taken from Zaenen ó.ó PRO et al. (Õɘ¢) and therefore commonly used in the literature, the Reduced Relative Diagnostic is new. As far as I am aware, it has not before been used to diagnose • PRO can be a subject, but not an object. is generalisation holds regardless subjecthood. of whether PRO is subject-controlled (ä), object-controlled (ß), or arbitrarily- controlled (˜): ì Experiencer subjects (ä) Subject-controlled • Assumptions a. Fernandai wanted [ PROi to hug Megan ] – A ¤¶†§Ží «¶fuh± is a subject marked with a nonnominative case. b. *Fernandai wanted [ (Megan) to hug PROi ] – e «¶fuh± is the highest base-generated argument (working assumption).

(ß) Object-controlled – By these denitions, quirky subjects include experiencers in Icelandic, Ger- man, Hindi, Basque, and Laz; ergatives in Hindi, Basque, and Laz; and a. Fernanda told Hsin-Luni [ PROi to hug Megan ] oblique objects under passivisation in Icelandic, German, and Hindi.

b. *Fernanda told Hsin-Luni [ (Megan) to hug PROi ] ⇒ In this presentation, I will focus on experiencer quirky subjects because they ex- ist in all ve languages under investigation and su›ce to illustrate the empirical Ö In the interest of time, I have omitted the arguments against raising-to-object and conjunction reduction as subjecthood diagnostics. generalisation that emerges from the broader study.

ì • Hindi • Basque Quirky subjects in Hindi can bind SOA, but cannot be PRO or undergo relativi- Quirky subjects in Basque can bind SOA, but cannot be PRO or undergo rela- sation in RRCs. tivisation in RRCs.

(Õó) a. Binding Diagnostic  (Õ¦) a. Binding Diagnostic  Mujhei [apnei sab rishtedar¯ ] pasand hε˜ Jon-ii [bere buru-a ]i gusta-tzen zaio I.oZ± §u€.“.£ all relative.“ like be.£§«.£ Jon-oZ± his head-ou±.•™“ like-†“£€ê Z¶ì(쫁.Zf«-쫁.oZ±) ‘I like all my relatives’ [Bhatt óþþì:Õó] ‘Jon likes himself’  b. PRO Diagnostic  b. PRO Diagnostic *Ravii [PROi Rina pasand a-n¯ a¯ ] nah˜¯ı *Jon-eki [PROi gustatu Miren ] nahi du Ravi.•™“ PRO.oZ± Rina.•™“ like come-†•€ •u Jon-u§ PRO.oZ± like Miren.•™“ want Z¶ì(쫁.Zf«-쫁.u§) Intended: chah-t¯ a.¯ ‘Jon wants to like Miren’ want-„Zf.“.« c. Reduced Relative Diagnostic  Intended: ‘Ravi doesn’t want to like Rina’ *[ ___i Miren gustatu-ta-ko ] gizon-ai ... c. Reduced Relative Diagnostic  ___.oZ± Miren.•™“ like-£±h£-™h man-ou±.•™“ Intended: ‘the man who likes Miren . . .’ *[ ___i cot. lag-a¯ ] lar.ka¯i ... ___.oZ± hurt contact-£€ê boy • Icelandic Intended: ‘the hurt boy . . .’ Quirky subjects in Icelandic can bind SOA and be PRO, but cannot undergo relativisation in RRCs. • German Quirky subjects in German can bind SOA, but cannot be PRO or undergo (Õ¢) a. Binding Diagnostic  relativisation in RRCs. Hennii þykir [bróðir {sinni /*hennari }] leiðinlegur her.oZ± thinks brother.•™“ her.§u€ her boring (Õì) a. Binding Diagnostic  ‘Shei thinks heri brother boring’ [Zaenen et al. Õɘ¢:¦¢þ] [Dem Fritz]i gefällt [das Bild von sichi ] the.oZ± Fritz likes the.•™“ picture of §u€ b. PRO Diagnostic  ‘Fritz likes the picture of himself’ Égi vonast til [PROi að vanta ekki peninga ] I.•™“ hope for PRO.Zhh to lack not money.Zhh b. PRO Diagnostic  ‘I hope not to lack money’ [Zaenen et al. Õɘ¢:¦¢¦] *Fritzi ho [PROi geholfen zu werden] Fritz hopes PRO.oZ± helped to be c. Reduced Relative Diagnostic  Intended: ‘Fritz hopes to be helped’ *[ ___i ekni ] bíll-inni ... ___.oZ± driven car-the.•™“ c. Reduced Relative Diagnostic  Intended: ‘the driven car . . .’ [Einar Freyr Sigurðsson, p.c.] *der [___i der Fritz gefallende] Manni ... the.•™“ ___.oZ± the.•™“ Fritz like.£±h£ man • Laz Intended: ‘the man who likes Fritz . . .’ Quirky subjects in Laz can bind SOA, be PRO, and undergo relativisation in RRCs.

(Õä) a. Binding Diagnostic  [Ham biç’i-s ]i ti-muşii opşa a-limb-en this boy-oZ± head-£™««.ì.•™“ much Z££-love-†“£€ê ‘is boy loves himself very much’ [Demirok óþÕì:óÕ]

¦ b. PRO Diagnostic  ⇒ All DPs obey the QSH Bere-ki [PROi layç’-epe o-limb-u ] gor-um-s Even though the QSH is based on an investigation of quirky subjects, it applies child-u§ PRO.oZ± dog-£.•™“ •“«-love-•“« want-†“£€ê-ì to all DPs: ama a-şk’urin-en – Nominative subjects → All but Z££-fear-†“£€ê.ì On the highest end of the hierarchy, nominative subjects possess all of the ‘e child wants to love the dogs, but s/he fears’ [Demirok óþÕì:ó¢] three properties. c. Reduced Relative Diagnostic  – Objects → None [___i ma limb-eri ] berei On the lowest end of the hierarchy, objects possess none of the three proper- ___.oZ± Õ.•™“ love-£±h£ child.•™“ ties. If a DP possesses none of the properties, it still adheres to the QSH, but ‘the child who has loved me’ [Demirok, p.c.] it is not a subject in any sense. • Summary of empirical ndings – Quirky subjects → Vary e empirical ndings are summarised in the table below: Quirky subjects sit somewhere in the middle, the exact position depending on the type of quirky subject and the language—this variation makes them Binding PRO Reduced relatives the interesting empirical domain of inquiry.    Hindi Typological predictions    • German e QSH predicts the following typology of quirky subjects: Basque       Icelandic (՘) Laz-like Laz   

reduced relatives Icelandic-like ¦ Quirky Subject Hierarchy

• e ndings show that the subjecthood diagnostics do not identify a unied PRO Hindi-like property of subjecthood, but rather each identify a unique property which we canonically associate with subjecthood. binding of SOAs ∅ An implicational hierarchy of subjecthood properties ✳  e QSH predicts that the following three types of quirky subjects exist: An empirical generalisation does emerge however from the data: the subject- hood properties exhibited by quirky subjects obey an implicational hierarchy: Õ. Laz-like quirky subjects: Can be relativised on in RRCs, be PRO, and bind SOAs. (Õß) Q¶†§Ží S¶fuh± H†u§Z§h„í (QSH) binding ≫ PRO ≫ reduced relatives binding ≫ PRO ≫ reduced relatives ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ ó. Icelandic-like quirky subjects: • Notation: For a hierarchy p ≫ p ≫ ⋯ ≫ pn, a DP has property pi+ only if it Ô ò Ô Can be PRO and bind SOAs. has property pi (i.e. pi+Ô → pi ). binding ≫ PRO ≫ reduced relatives ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ • e QSH captures two empirical generalisations: Õ. If a quirky subject can be relativised on in RRCs, it can also be PRO and ì. Hindi-like quirky subjects: bind SOAs. Can bind SOAs. binding ≫ PRO ≫ reduced relatives ó. If a quirky subject can be PRO, it can also bind SOAs. ´¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¶

¢  e QSH predicts that the following three types of quirky subjects do not exist: ¢ e Height Conjecture Õ. Can be relativised on in RRCs, but not be PRO. ✳ Proposal ∗binding ≫ PRO ≫ reduced relatives ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ A DP exhibits a set of subjecthood properties as a function of how high it raises ∗ binding ≫ PRO ≫ reduced relatives in the functional sequence: ´¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¶ ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ ó. Can be relativised on in RRCs, but not bind SOAs. (ÕÉ) T„u Hu†„± C™•uh±¶§u ∗binding ≫ PRO ≫ reduced relatives a. Let fseq be the functional sequence ⟨XÔ ≻ Xò ≻ ⋯ ≻ Xn⟩ such that ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ Xi takes Xi+ as its complement. ∗binding ≫ PRO ≫ reduced relatives Ô ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ b. Let m(x) be the mapping from functional heads to properties ì. Can be PRO, but not bind SOAs. {⟨XÔ, pÔ⟩, ⟨Xò, pò⟩,..., ⟨Xn , pn⟩} such that pÔ ≫ pò ≫ ⋯ ≫ pn. ∗binding ≫ PRO ≫ reduced relatives ± fseq m(x) p ∗binding ≫ PRO ≫ reduced relatives c. Given and , a DP base-merged in [Spec, Xk P] bears i ´¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¸¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¹¶ only if it moves to [Spec, Xi P] through [Spec, X jP] for all j such that k < j < i. • Surveying more languages with quirky subjects A potential criticism of the QSH is that it is based on only ve languages, even • e Height Conjecture maps an implicational hierarchy onto the syntactic though no study of even this relatively small magnitude exists. To mitigate this structure and derives the implicational relationships via movement.ô concern, I have surveyed the literature on quirky subjects in eleven other lan- guages. Insofar as the data are available (e.g. data on RRCs are rarely available), • Illustration of the proposal I have found no exception to the QSH. Take fseq = ⟨X ≻ Y ≻ Z ≻ ⋯⟩ and m(x) = {⟨X, x⟩, ⟨Y, y⟩, ⟨Z, z⟩,...}: – Faroese: Jónsson óþþÉ; Þráinsson óþþß; Þráinsson et al. óþþ¦ (óþ) DP has properties x, y, z (óÕ) DP has properties y and z – Gujarati: Mistry óþþ¦ XP XP – Hungarian: Rákosi óþþä DP XP Xý YP – Kannada: Amritavalli óþþ¦ Xý YP DP YP – Korean: Yoon óþþ¦ – Malayalam: Jayaseelan óþþ¦ ⟨DP⟩ YP Yý ZP

– Marathi: Wali óþþ¦ Yý ZP ⟨DP⟩ ZP – Russian (?): Schoorlemmer ÕÉɦ; Moore & Perlmutter óþþþ; Sigurðsson óþþó ⟨DP⟩ ZP Zý ⋮ – Spanish: González Õɘ˜; Masullo ÕÉÉì; Gutiérrez-Bravo óþþä Zý ⋮ – Telugu: Subbarao & Bhaskararao óþþ¦ • In (óþ), the DP raises to [Spec, XP] through [Spec, YP] and [Spec, ZP] such that it receives all three properties: x, y, and z.

ô I will argue that the QSH maps onto syntactic structure high in the functional sequence. is raises the question about whether such implicational relationships exist for structure lower in the functional sequence as well, e.g. for objects. I leave this topic for future research.

ä • In (óÕ), the DP only raises to [Spec, YP] through [Spec, ZP] such that it receives (ó¢) fseq = ⟨C ≻ α ≻ β ≻ v ≻ V⟩ properties y and z, but not x. • With the functional sequence in (ó¢), the QSH can be derived syntactically ⇒ e Height Conjecture in (ÕÉc) stipulates that a DP moves cyclically through according to the Height Conjecture: quirky subjects vary across languages in each specier until arriving at its targeted landing site. In the following section, whether they raise to [Spec, αP] through [Spec, βP] or just to [Spec, βP]. I implement this stipulation with particular combinations of movement-driving ý ý features. is stipulation will su›ce to illustrate how the Height Conjecture • Identity of α and β ý ý accounts for the QSH, but see the appendix for a more principled explanation. ere are a few plausible options for the identity of α and β . Consider the following two options: Application to the QSH in a nutshell • Õ. αý and βý are heads standardly assumed to be in the functional sequence. e remainder of this section details how the Height Conjecture accounts for the QSH. e analysis comprises three parts: ó. αý and βý are distinct functional heads whose purposes are solely to host Õ. Tý is decomposed into Persp(ective)ý to host PRO and B(inding)ý to bind PRO and bind SOA respectively. SOA where Perspý ≻ Bý. • An instantiation of the rst option is αý = Tý and βý = vý, following Williams ó. Quirky subjects vary in how high they raise in the functional sequence. (óþþì) who argues that control takes place at TP and anaphora binding takes place at vP. is predicts that any XP merged in [Spec, vP] can bind SOA. I leave ì. When a quirky subject can raise to [Spec, PerspP], it may be able to undergo open whether this option is viable given other functionality typically ascribed further projecting movement to form an RRC. to Tý and vý. • ese two dimensions of variation yield three types of quirky subjects: • For the sake of exposition, I assume the second option: αý = Persp(ective)ý and (óó) Hindi-like quirky subjects βý = B(inding)ý. Below, I sketch out what these specialised functional heads do. ý ý [NP [PerspP Persp [BP QS B [ ...]]]] (óä) fseq = ⟨C ≻ Persp ≻ B ≻ v ≻ V⟩   • Persp(ective)ý (óì) Icelandic-like quirky subjects Following Sundaresan (óþÕó), I propose that perspective, an analogue of lo- ý ý [NP [PerspP QS Persp [BP B [ ...]]]] gophoricity, is represented in the syntax as Perspý. Perspý represents a function  of type ⟨e, st⟩ that relates a perspective holder to an eventuality and that com- ý (ó¦) Laz-like quirky subjects bines via Event Identication, like Voice (Kratzer ÕÉÉä):í ý ý [NP QS [PerspP Persp [BP B [ ...]]]] (óß) Perspý = λx.λe.PerspHolder(e, x) [Sundaresan óþÕó:՘þ] J K – e perspectival holder contains the coordinates of the perspectival centre, ¢.Õ Decompose Tý e.g. time and location.

• e QSH suggests that the syntactic position canonically associated with sub- – PRO contains the coordinates of the attitude holder. By raising to [Spec, PerspP], jects, namely [Spec, TP], serves two purposes relevant to subjecthood: it establishes the attitude holder as the perspectival centre. Õ. To host PRO. – Deriving de se e de se interpretation of PRO is the result of identity between the perspec- ó. To establish the binding relationship between SOAs and their antecedents. tival holders in the matrix clause and the embedded clause (i.e. the same ⇒ Tý should be decomposed into two separate functional heads αý to host PRO coordinates). Movement of PRO to [Spec, PerspP] is a necessary, but not de se and βý to establish the binding relationship of SOAs, where αý crosslinguistically su›cient condition for achieving its interpretation. ý projects above β . í Landau (óþÕì) also denes control partly in terms of logophoricity.

ß – Forcing de se in control clauses • I make use of two structure-building (EPP) probes: [●•™“●] and [●o●]. e complement of a control predicate is subject to the Perspectival Centre Constraint (ó˜) at LF such that if a DP does not raise to [Spec, PerspP] in the • Assumption about probing complement of a control predicate, the derivation crashes at the LF interface.– Features probe down into the accessible structure until they encounter a DP. If that DP satises the feature, the feature agrees with the DP; otherwise, the (ó˜) Pu§«£uh±†êZ Cu•±§u C™•«±§Z†•± probe aborts and discontinues searching. erefore, only the highest DP in the e complement of a control predicate must have an explicit perspecti- structure is eligible to move to [Spec, BP] and subsequently to [Spec, PerspP]. val centre. • Hindi-like quirky subjects B(inding)ý • Hindi-like quirky subjects only raise to [Spec, BP] because, in languages like e connection between SOAs and the subject is standardly mediated through Hindi, Bý bears [●o●], but Perspý bears [●•™“●]:Ž Tý. e anaphor agrees with Tý whose specier is its binder (e.g. Reinhart & ý Reuland ÕÉÉì; Kratzer óþþÉ). e same functionality can be transferred to B . (ìÕ) PerspP A simple denotation for Bý is in (óÉ) where r is the special index borne by SOAs such that a DP must move to [Spec, BP] in order to bind DPs bearing index r. ∃ PerspP

ý g g[r→x] (óÉ) B = λP.λx. P Perspý BP J K J K [●•™“●] • Again, these are only sketches of analyses to show that such functional heads  DP BP are feasible. Full-blown theories of control and SOA are beyond the scope of oZ± this project. Bý vP [●o●] ⟨ ⟩ ⋮ ¢.ó Source of variation kÕ: Height in the functional sequence DPoZ±

⇒ Quirky subjects vary in how high they raise in the functional sequence, i.e. to [Spec, PerspP] through [Spec, BP] or just to [Spec, BP]. erefore, the Height • When the quirky subject is PRO, it does not raise high enough to [Spec, PerspP] Conjecture dictates which properties that they bear. for a perspectival centre to be selected such that the Perspectival Centre Con- • Notation straint is not satised. Hence, Hindi-like quirky subjects cannot occur in the [●€●] features trigger Mu§u upon A§uu. [⋆€⋆] features are satised by pure complement of a control predicate. A§uu. (Heck & Müller óþþß)

• Movement is driven by case-relativised probes Probes can be case-relativised to positions on the (Revised) Moravcsik Hierarchy (Bobaljik óþþ˜; Preminger óþÕÕ, óþÕ¦):

(ìþ) T„u (Ruꆫuo)M™§Zêh«†Ž H†u§Z§h„í unmarked ≫ dependent ≫ lexical/inherent [Bobaljik óþþ˜:ìþì] at is, [⋆•™“⋆] agrees with only nominative DPs, [⋆ou£⋆] only agrees with dependent-case and nominative DPs, and [⋆o⋆] agrees with any DP. – Another possibility is that the type of PRO is such that it must raise from its base position to avoid type mismatch, e.g. ⟨st, ⟨e, st⟩⟩ (shiŸing from a proposition to a property). is would require a Ž Existential closure is required to avoid type mismatches, which is why the Perspectival Centre dišerent denotation of Perspý, but it would also account for PRO under nonattitudinal predicates, Constraint requires an explicit perspectival centre to be established in the complement of a control e.g. tell and neglect. predicate. Existential closure will not satisfy the constraint.

˜ • Icelandic-like quirky subjects (ìì) a. the Basquei [ who ___i gave Stefan the rutabaga ] Agent Icelandic-like quirky subjects raise to [Spec, BP] and then to [Spec, PerspP] Goal because, in languages like Icelandic, both Perspý and Bý bear [●o●]: b. the Germani [ who Jon Ander gave ___i the rutabaga ]

c. the rutabagai [ that Jon Ander gave Stefan ___i ] eme (ìó) PerspP – In an RRC, an XP must reach the edge of the clause via other means (i.e. A- DPoZ± PerspP movement) to be relativised on. erefore, only the subject can be relativised on: Perspý BP [●o●] (ì¦) a. the Basquei [ ___i giving Stefan the rutabaga ] Agent ⟨DPoZ±⟩ BP b. *the Germani [ Jon Ander giving ___i the rutabaga ] Goal ý v B P c. *the rutabagai [ Jon Ander giving Stefan ___i ] eme [●o●] ⟨DPoZ±⟩ ⋮ • Laz-like quirky subjects Laz-like quirky subjects are just like Icelandic-like quirky subjects, except they can undergo projecting movement from [Spec, PerspP] to form an RRC:‡ • When the quirky subject is PRO, the Perspectival Centre Constraint is satised (ì¢) DP because PRO has raised to [Spec, PerspP]. Hence, Icelandic-like quirky subjects can occur in the complement of a control predicate. Dý NP

• Due to case-relativised probing, when a nominative DP occurs in structures oZ± PerspP like (ìÕ) and (ìó), it will raise all the way to [Spec, PerspP] through [Spec, BP] ● ● ● ● because a nominative DP satises both [ •™“ ] and [ o ]. ⟨oZ±⟩ PerspP

ý ¢.ì Source of variation kó: Reduced relative formation Persp BP [●o●] ⟨ ⟩ ⇒ When a quirky subject can raise to [Spec, PerspP], it may or may not be able to oZ± BP undergo further projecting movement to form an RRC. Bý vP ● ● • Background: Forming a reduced relative [ o ] ⟨ ⟩ ⋮ To form a relative clause, the relativised element must raise to the edge of the oZ± clause. e relativised element then undergoes short projecting movement to form the head NP (Bhatt óþþä).Ø – e dišerence between a nite relative clause (FRC) and a reduced relative • e ability of RRCs to be formed on quirky subjects in languages like Laz indi- clause (RRC) is that an FRC has a CP layer and an RRC does not. cates that they occupy the specier position of the highest syntactic projection. As predicted, in Laz, quirky subjects can also be PRO and bind SOAs. – In an FRC, the CP layer permits A-movement to the edge of the clause.Ô erefore, any argument position can be relativised on:

Ø Another prevalent analysis of relative-clause formation is that the relativised element is deleted under identity with Ný (e.g. Chomsky ÕÉä¢). Either analysis of relative-clause formation is compatible with ‡ As the case of the quirky subject does not preserve in the matrix clause in Laz, we may need some this paper because movement to the edge of the clause is a prerequisite in both analyses. mechanism of case overwriting. See Deal (óþÕ¦) who proposes such an analysis for relative clauses in Ô I am assuming that A-movement targets [Spec, CP] (see e.g. Sauerland ÕÉɘ). Nez Perce.

É ⇒ Support for a height distinction ä.Õ Gradient view is provides support for quirky subjects being at dišerent heights in dišerent languages. However, it is not as ne-grained a diagnostic as one would like • A view of subjecthood, which is common particularly in the LFG literature, is because it can only distinguish between the highest and not highest position. that subjecthood is gradient or on a continuum (e.g. Mohanan ÕÉɦ; Barðdal óþþó; Barðdal & Eyþórsson óþþìa,b, óþþ¢; Bayer óþþ¦; Keenan ÕÉßä): • An aside: Similarities with Representation eory In Representation eory (RT), each level denes its own particular kind of (ìß) G§Zo†u•± V†uë subject (Williams óþþì:˜þ–Éì). For example, the subject in TP can be controlled A DP is a subject iš it possesses any subset of the maximal set of subject- and the subject in LP (Logic Phrase) is identied with EPP subjects. hood properties.

• Williams (óþþä) exploits this notion to explain the dišerence between subjects • is view is largely a response to the crosslinguistic variation of quirky subjects. in English, which can be PRO, and in Russian, which cannot be PRO. In English, ⇒ Empirical problem the surface subject is the subject in both TP and LP. In Russian, the surface Even though the gradient view is compatible with the QSH, it makes typological subject is only the subject in LP, which is why it cannot be PRO. predictions that do not bear out. If subjecthood is genuinely gradient, for x x • Equating timing in RT with height in standard theory, the Height Conjecture number of subjecthood properties, we expect to nd ò dišerent types of DPs. and the notion of subject in RT share the same intuition, although the reasoning • is typological prediction is incompatible with the QSH. is from a dišerent angle and the implementation dišers.

ä e nature of subjecthood ä.ó Absolute-metric view

• According to the QSH, the distribution of subjecthood properties follows a • Following Zaenen et al. (Õɘ¢), another common view of subjecthood is that absolute metric of subjecthood subset relationship, as illustrated below. the suite of subjecthood diagnostics constitute an (e.g. Zaenen et al. Õɘ¢; Sigurðsson ÕɘÉ, óþþó, óþþ¦; McFadden óþþ¦; Gutiérrez- (ìä) Bravo óþþä; Rákosi óþþä; Preminger óþÕÕ, óþÕ¦): (ì˜) Af«™¶±u-Mu±§†h V†uë A DP is a subject iš it passes every diagnostic.¢

pç pò pÔ • “I am using the term quirky subject to refer specically to instances of non- nominative noun- phrases that pass the full battery of subjecthood diagnostics.” (Preminger óþÕÕ:ÕóÉ)

• Unied property of subjecthood One interpretation of this view is that there is some unied property of sub- jecthood. erefore, a DP should either pass every diagnostic or fail every • We are leŸ to decide what we want to label as “subjects”, but this choice of la- diagnostic because the diagnostics all probe for the same property. belling is inconsequential. What is important is that the subjecthood properties bear a subset relationship to one another. ⇒ is strict interpretation is incompatible with the QSH because in the languages under investigation, excluding Laz, quirky subjects pass only some of the diag- • Section outline nostics, an outcome that this interpretation predicts to be impossible. is section considers precisely what the QSH entails for the dišerent views of subjecthood found in the literature. e conclusion will be that the existing views are incompatible with the QSH. ¢ Given the QSH, this view ešectively amounts to saying that a DP must pass the Reduced Relative Diagnostic to be a subject.

Õþ • Concatenation of properties • For example, N might include being PRO and P might include being able to be Another interpretation is that the conjunction of all the properties makes a DP relativised on in a RRC. a subject: ⇒ Conceptual problem (ìÉ) e designation between subject and nonsubject is arbitrary. We can simply rede- ne the notion of subjecthood to include those languages where we want quirky subjects to be true subjects by including and excluding specic diagnostics. pÔ • For example, if we want Icelandic quirky subjects to be true subjects, we assert that the Binding Diagnostic and the PRO Diagnostic are the two necessary diagnostics. If we want Hindi quirky subjects to be true subjects, we assert that only the Binding Diagnostic is necessary. pò pç ⇒ Not expressive enough e real problem with this weaker absolute-metric view is that it is not expressive enough because it only has two levels. e QSH requires more articulation to ⇒ is view predicts that we should nd DPs that exhibit each property indepen- capture the implicational relationships. Once you start adding these extra layers dently. For example, we should nd a DP that can be PRO, but not bind SOA. to (¦Õ), it starts to look like (ìä). However, this prediction is incompatible with the QSH. • In this sense, the QSH is an improvement on this view given the data. • Weaker version Maintaining this view requires weakening the absolute-metric view such that ß Conclusion only some diagnostics are necessary for subjecthood and others are extra prop- erties that only subjects can possess: • I have presented a suite of crosslinguistic subjecthood diagnostics: binding of subject-oriented anaphora, PRO, and reduced relatives. (¦þ) Af«™¶±u-Mu±§†h V†uë (ëuZŽ) A DP must possess some minimal set of properties to be a subject, but • e behaviour of quirky subjects in Icelandic, German, Hindi, Basque, and Laz subjects may also commonly possess a number of additional properties.Öü on these diagnostics constitutes an implicational hierarchy:

• In other words, there is a set of properties P that subjects can possess, but only (¦ó) Q¶†§Ží S¶fuh± H†u§Z§h„í (QSH) a subset of these properties N does a subject need to possess: binding ≫ PRO ≫ reduced relatives

(¦Õ) • To account for the QSH, I have proposed the Height Conjecture: DPs exhibit a set of subjecthood properties as a function of how high they raise in the functional sequence.

• e QSH is the result of DPs dišering in (i) how high they raise in the func- N P tional sequence and (ii) whether they can undergo projecting movement from [Spec, PerspP], the highest syntactic position, to form a reduced relative.

• e absolute-metric and gradient views of subjecthood are incompatible with the empirical generalisation encapsulated in the QSH. Öü Although no one in the literature explicitly argues for this view of subjecthood, at least as far as I am aware, it is what the absolute-metric view would be forced into given the QSH.

ÕÕ Acknowledgements: González, Nora. Õɘ˜. Object and raising in Spanish. New York: Garland. Many thanks to Rajesh Bhatt, Kyle Johnson, Stefan Keine, Jon Ander Mendia, Ja- Gutiérrez-Bravo, Rodrigo. óþþä. A reinterpretation of quirky subjects and related phenomena son Merchant, Gereon Müller, Ellen Woolford, and audiences at MIT and UMass in Spanish. In New perspectives in Romance linguistics, ed. Chiyo Nishida & Jean-Pierre Y. Amherst for their helpful and insight discussion. For discussion about data and Montreuil, Õóß–Õ¦ó. Amsterdam: John Benjamins. Proceedings of the óäth grammaticality judgements, I am indebted to Rajesh Bhatt (Hindi), Ömer Demirok Heck, Fabian, & Gereon Müller. óþþß. Extremely local optimization. In Western Conference on Linguistics (WECOL òâ) (Laz), Anton Karl Ingason (Icelandic), Stefan Keine (German), René Lacroix (Laz), , ed. Erin Brainbridge & Brian Agbayani, Õßþ–՘ì. Fresno, CA: California State University. Jon Ander Mendia (Basque), and Einar Freyr Sigurðsson (Icelandic). is work Jayaseelan, K. A. óþþ¦. e possessor-experiencer dative in Malayalam. In Non-nominative is supported by the National Science Foundation Graduate Research Fellowship subjects, ed. Peri Bhaskararao & Karumuri Venkata Subbarao, volume Õ, óóß–ó¦¦. Amster- under NSF DGE-þÉþßÉÉ¢. dam: John Benjamins. Jónsson, Jóhannes Gísli. óþþÉ. Covert nominative and dative subjects in Faroese. Nordlyd ìä:Õ¦ó–Õä¦. Keenan, Edward. ÕÉßä. Towards a universal denition of ‘subject’. In Subject and topic, ed. Amritavalli, R. óþþ¦. Experiencer datives in Kannada. In Non-nominative subjects, ed. Peri Charles Li, ìþì–ììì. New York: Academic Press. Bhaskararao & Karumuri Venkata Subbarao, volume Õ, Õ–ó¦. Amsterdam: John Benjamins. Kratzer, Angelika. ÕÉÉä. Severing the external argument from its verb. In Phrase structure Andrews, Avery. ÕÉßä. e VP complement analysis in Modern Icelandic. In Proceedings of and the lexicon, ed. Johan Rooryck & Laurie Zaring, ÕþÉ–Õìß. Dordrecht: Kluwer. the äth Meeting of the North East Linguistic Society (NELS â), Õ–óÕ. Kratzer, Angelika. óþþÉ. Making a pronoun: Fake indexicals as windows into the properties Barðdal, Jóhanna. óþþó. “Oblique subjects” in Icelandic and German. Working Papers in of pronouns. Linguistic Inquiry ¦þ:՘ߖóìß. Scandinavian Syntax ßþ:äÕ–ÉÉ. Landau, Idan. óþÕì. A two-tiered theory of control. Manuscript, Ben Gurion University. Barðdal, Jóhanna, & Þórhallur Eyþórsson. óþþìa. e change that never happened: e story Masullo, Pascual J. ÕÉÉì. Two types of quirky subjects: Spanish versus Icelandic. In Proceedings of oblique subjects. Journal of Linguistics ìÉ:¦ìÉ–¦ßó. of the óìth Meeting of the North East Linguistic Society (NELS òç), ed. Amy J. Schafer, ìþì–ìÕß. Barðdal, Jóhanna, & Þórhallur Eyþórsson. óþþìb. Icelandic vs. German: Oblique subjects, Amherst, MA: GLSA. agreement and expletives. Chicago Linguistic Society ìÉ:ߢ¢–ßßì. McFadden, omas. óþþ¦. e position of morphological case in the derivation: A study Barðdal, Jóhanna, & Þórhallur Eyþórsson. óþþ¢. Oblique subjects: A common Germanic on the syntax-morphology interface. Doctoral Dissertation, University of Pennsylvania, inheritance. Language ˜Õ:˜ó¦–˜˜Õ. Philadelphia, PA. Bayer, Josef. óþþ¦. Non-nominative subjects in comparison. In Non-nominative subjects, Mistry, P. J. óþþ¦. Subjecthood of non-nominatives in Gujarati. In Non-nominative subjects, ed. Peri Bhaskararao & Karumuri Venkata Subbarao, volume ó, ¦É–ßä. Amsterdam: John ed. Peri Bhaskararao & Karumuri Venkata Subbarao, volume ó, Õ–ìÕ. Amsterdam: John Benjamins. Benjamins. Bhatt, Rajesh. óþþì. Experiencer subjects. Handout from MIT course “Structure of the Mohanan, Tara. ÕÉɦ. Argument structure in Hindi. Stanford, CA: CSLI Publications. Modern Indo-Aryan Languages”. Moore, John, & David Perlmutter. óþþþ. What does it take to be a dative subject? Natural Bhatt, Rajesh. óþþä. Covert modality in nonnite contexts. Berlin: Mouton de Gruyter. Language and Linguistic eory ՘:ìßì–¦Õä. Bobaljik, Jonathan. óþþ˜. Where’s phi? Agreement as a post-syntactic operation. In Phi- Preminger, Omer. óþÕÕ. Agreement as a fallible operation. Doctoral Dissertation, MIT, theory: Phi features across interfaces and modules, ed. Daniel Harbour, David Adger, & Cambridge, MA. Susana Béjar, óÉ¢–ìó˜. Oxford: Oxford University Press. Preminger, Omer. óþÕ¦. Agreement and its failures. Cambridge, MA: MIT Press. Chomsky, Noam. ÕÉä¢. Aspects of the theory of syntax. Cambridge, MA: MIT Press. Rákosi, György. óþþä. Dative experiencer predicates in Hungarian. Utrecht: LOT. Chomsky, Noam. óþþß. Approaching UG from below. In Interfaces + recursion = language?, Reinhart, Tanya, & Eric Reuland. ÕÉÉì. Režexitivity. Linguistic Inquiry ó¦:ä¢ß–ßóþ. ed. Uli Sauerland & Hans-Martin Gärtner, Õ–ìÕ. Berlin: Mouton de Gruyter. Sauerland, Uli. ÕÉɘ. e meaning of chains. Doctoral Dissertation, MIT, Cambridge, MA. Chomsky, Noam. óþþ˜. On phases. In Foundational issues in linguistic theory, ed. Robert Schoorlemmer, Maaike. ÕÉɦ. Dative subjects in Russian. In Formal approaches to Slavic Freidin, Carlos Otero, & Maria Luisa Zubizarreta. Cambridge, MA: MIT Press. linguistics: e Ann Arbor meeting, ed. Jindřich Toman, ÕóÉ–Õßó. Ann Arbor, MI: Michigan Deal, Amy Rose. óþÕ¦. Cyclicity and connectivity in Nez Perce relative clauses. Ms., University Slavic Publications. of California, Santa Cruz. Sigurðsson, Halldór Ármann. ÕɘÉ. Verbal syntax and case in Icelandic. Doctoral Dissertation, Demirok, Ömer Faruk. óþÕì. Agree as a unidirectional operation: Evidence from Laz. Master’s Lunds universitet. thesis, Boğaziçi University. Sigurðsson, Halldór Ármann. óþþó. To be an oblique subject: Russian vs. Icelandic. Natural Fanselow, Gisbert. óþþó. Quirky “subjects” and other speciers. In More than words. Language and Linguistic eory óþ:äÉÕ–ßó¦. A FestschriŸ for Dieter Wunderlich, ed. Ingrid Kaufmann & Barbara Stiebels, óóß–ó¢þ. Sigurðsson, Halldór Ármann. óþþ¦. Icelandic non-nominative subjects: Facts and implica- Akademie Verlag. tions. In Non-nominative subjects, ed. Peri Bhaskararao & Karumuri Venkata Subbarao,

Õó volume ó, Õìß–Õ¢É. Amsterdam: John Benjamins. Subbarao, Karumuri Venkata, & Peri Bhaskararao. óþþ¦. Non-nominative subjects in Telugu. In Non-nominative subjects, ed. Peri Bhaskararao & Karumuri Venkata Subbarao, volume ó, ÕäÕ–ÕÉä. Amsterdam: John Benjamins. Sundaresan, Sandhya. óþÕó. Context and (co)reference in the syntax and its interfaces. Doctoral Dissertation, University of Tromsø, University of Stuttgart. Vikner, Sten. Õɘ¢. Parameters of binder and of binding category in Danish. Working Papers in Scandinavian Syntax óì. Wali, Kashi. óþþ¦. Non-nominative subjects in Marathi. In Non-nominative subjects, ed. Peri Bhaskararao & Karumuri Venkata Subbarao, volume ó, óóì–ó¢ó. Amsterdam: John Benjamins. Williams, Edwin. óþþì. Representation theory. Cambridge, MA: MIT Press. Williams, Edwin. óþþä. Subjects of dišerent heights. In Proceedings of FASL Ô¥. Wunderlich, Dieter. óþþÉ. e force of lexical case: German and Icelandic compared. In e nature of the word: Studies in honor of Paul Kiparsky, ed. Kristin Hanson & Sharon Inkelas, ¢˜ß–äóþ. Cambridge, MA: MIT Press. Yoon, James H. óþþ¦. Non-nominative (major) subjects and case stacking in Korean. In Non-nominative subjects, ed. Peri Bhaskararao & Karumuri Venkata Subbarao, volume ó, ó䢖ìÕ¦. Amsterdam: John Benjamins. Zaenen, Annie, Joan Maling, & Höskuldur Þráinsson. Õɘ¢. Case and grammatical functions: e Icelandic passive. Natural Language and Linguistic eory ì:¦¦Õ–¦˜ì. Þráinsson, Höskuldur. ÕÉßÉ. On complementation in Icelandic. New York: Garland. Þráinsson, Höskuldur. óþþß. e syntax of Icelandic. Cambridge, UK: Cambridge University Press. Þráinsson, Höskuldur, Hjalmar P. Petersen, Jógvan í Lon Jacobsen, & Zakaris Svabo Hansen. óþþ¦. Faroese: An overview and reference grammar. Tórshavn: Føroya Fróðskaparfelag.

Õì Appendix: Downwards feature percolation • Application to the QSH Languages dišer with respect to whether Perspý has access to [●•™“●] and • As discussed earlier, the Height Conjecture stipulates that a DP moves cyclically [●o●] and Bý has access to [●•™“●] and [●o●]. through each specier until arriving at its targeted landing site. Until now, I have – Hindi-like languages: abstracted away from this stipulation by using only particular combinations of Perspý:[●•™“●], Bý:[●•™“●] and [●o●] movement-driving features, which derives this cyclic movement. – Icelandic-like and Laz-like languages: • Problem Perspý:[●•™“●] and [●o●], Bý:[●•™“●] and [●o●] Only certain combinations of probes on Perspý and Bý are possible given the Height Conjecture and the QSH: • In any given derivation, Perspý and Bý each receive a probe at random, contin- gent on which probes are accessible to each head given the language. (¦ì) a. Perspý:[●•™“●], Bý:[●•™“●]ÖÖ ⇒ Consider when Perspý receives [●o●] and Bý receives [●•™“●]—the illicit com- ý ● ● ý ● ● b. Persp :[ o ], B :[ o ] bination. Whenever Perspý probes down into the structure, it copies [●o●] onto Bý. As [●•™“●] ∪ [●o●] is equivalent to [●o●], the illicit combination is c. Perspý:[●•™“●], Bý:[●o●] ešectively overwritten. d. *Perspý:[●o●], Bý:[●•™“●]

• e combination [●o●]–[●•™“●] would allow a dative DP to skip over [Spec, BP], i.e. a dative DP that can be PRO but not bind SOA—which violates the QSH. ⇒ Generalisation e probe on Perspý must be a subset of the probe on Bý.

• Assumption Operations apply simultaneously at S£u™¶± (Chomsky óþþß, óþþ˜).

• Proposal To account for this generalisation, I propose a system of downwards feature percolation:

(¦¦) D™ë•ëZ§o« FuZ±¶§u Pu§h™Z±†™• When a functional head αý probes into the structure to satisfy the feature [€], αý copies [€] onto every functional head in its search path until it nds a suitable goal. (¦¢) U•†™• P§™f†• When a head has two probes [€Ô] and [€ò], it probes with the union of these two features [€Ô] ∪ [€ò].

ÖÖ One might consider English to be a [●•™“●]–[●•™“●] language if experiencer arguments of verbs like seem are treated as datives because these arguments are ineligible to be subjects but block another DPs raising over them, i.e. they intervene.

Õ¦