Hostname: page-component-745bb68f8f-kw2vx Total loading time: 0 Render date: 2025-02-06T07:42:09.555Z Has data issue: false hasContentIssue false
  • English
  • Français

A Criterial Freezing approach to subject extraction in Jordanian Arabic

Published online by Cambridge University Press:  21 June 2017

Marwan Jarrah
Marwan Jarrah*
Affiliation:
University of Newcastle
*
m.a.s.jarrah@newcastle.ac.uk
Rights & Permissions [Opens in a new window]

Abstract

Using the Criterial Freezing approach to movement and chain formation (Rizzi 2005, 2006, 2014; Rizzi and Shlonsky 2006, 2007), this study explores the strategies Jordanian Arabic makes available for subject extraction. I argue that subject extraction in this variety of Arabic is constrained by the postulated D-linking condition of the Subject Criterion – i.e., [spec,SubjP] is filled by an element with the same D(iscourse)-linking status as that of the subject wh-word (D-linked vs. non-D-linked). In case of questions with a D-linked wh-word, [spec,SubjP] can be filled by the D-linked particle illi or a deictic (time-point/place-point) adjunct. Unlike time-point adjuncts, the use of place-point adjuncts to fill [spec,SubjP] is subject to the effects of the Phase Impenetrability Condition (Chomsky 2001), given their low base positions. In contrast, in case of questions with a non-D-linked wh-word, I assume that [spec,SubjP] is filled by an expletive pro.

Résumé

La présente étude s'appuie sur l'approche dite Criterial Freezing (Immobilité critériale) appliquée au mouvement et à la formation des chaînes (Rizzi 2005, 2006, 2014, Rizzi et Shlonsky 2006, 2007), pour analyser les stratégies que rend disponibles l'arabe jordanien afin effectuer l'extraction du sujet. Je soutiens que l'extraction du sujet, dans cette variété d'arabe, est contrainte par le postulat de la condition de la liaison-D du Critère du Sujet – c-à-d. que [spec,SubjP] est rempli par un élément ayant le même statut de liaison-D(iscursive) que le sujet Qu- (soit D-lié, soit non D-lié). Dans le cas des questions contenant un ‘mot Qu-’ qui est D-lié, le Spec de SubjP peut être rempli par la particule D-liée illi, ou alors par un adjoint déictique (point-temps/point-lieu). Contrairement aux adjoints point-temps, l'utilisation des adjoints point-lieu pour remplir [spec,SubjP] est sujette aux effets de la Condition d'Impénétrabilité des Phases (Chomsky 2001), étant donné leurs basses positions de base. Par contre, dans le cas des questions avec un ‘mot Qu-’ non D-lié, j'affirme que [spec,SubjP] est rempli par un pro explétif.

Keywords

D-LinkingCriterial Freezingsubject extractionPhase TheoryJordanian ArabicLiaison-DImmobilité critérialeextraction du sujetthéorie des phasesarabe jordanien
Type
Articles
Information
Canadian Journal of Linguistics/Revue canadienne de linguistique , Volume 62 , Issue 3 , September 2017 , pp. 411 - 448
Copyright
© Canadian Linguistic Association/Association canadienne de linguistique 2017 

1. Introduction

This article investigates subject extraction in Jordanian ArabicFootnote 1 (henceforth, JA) within the Criterial Freezing approach (Rizzi Reference Rizzi, Brugè, Giusti, Munaro, Schweikert and Turano2005, Reference Rizzi, Cheng and Corver2006, Reference Rizzi and Svenonius2014; Rizzi and Shlonsky Reference Rizzi, Shlonsky and Frascarelli2006, Reference Rizzi, Shlonsky, Gärtner and Sauerland2007).Footnote 2 It is clear that JA has different ways of satisfying the Subject Criterion when the subject is extracted. These strategies include the use of the particle illi, time-point adjuncts, place-point adjuncts, and an expletive pro. The choice between them is shown not to be random, but follows from strict conditions, most notably D(iscourse)-Linking. [Spec,SubjP] must be filled by an element whose D-linking status is identical to that of the subject wh-word, a state of affairs that I formulate as the D-linking condition of the Subject Criterion. The word illi, being a D-linked particle, fills [spec,SubjP] in questions with a D-linked subject wh-word. Time-point adjuncts like yesterday and place-point adjuncts (though those are limited to intransitive questions) can also fill [spec,SubjP] in such cases. This is possible because they are D-linked, by virtue of, first, containing a nominal time/place element (cf. Kayne Reference Kayne2005, Stanton Reference Stanton2016) and second, referring to a particular point in discourse. This gives rise to the free variation between illi and such adjuncts with regard to filling [spec,SubjP]. On the other hand, when the subject wh-word is not D-linked, [spec,SubjP] is filled by an expletive pro, a non-D-linked element. All of these ways instantiate Rizzi and Shlonsky's (Reference Rizzi, Shlonsky, Gärtner and Sauerland2007) skipping strategies where the subject is moved directly from its thematic position, and [spec,SubjP] is filled by a different element.

The article is organized as follows. Section 2 discusses subject extraction as well as the theoretical approach taken, namely the Criterial Freezing approach with special reference to the Subject Criterion. Section 3 introduces subject extraction in JA, and section 4 sketches out the main previous approaches to illi, casting doubt on their findings with regard to the grammatical function of illi in questions. Following Criterial Freezing, it is argued that illi is a D-linked XP element that fills [spec,SubjP]. Section 5 elaborates on the assumption that illi is a D-linked element and discusses why illi is sometimes optional when the verb shows [3SG.M] agreement. Section 6 explains the reason why time-point adjuncts and place-point adjuncts can replace illi and the conditions that constrain this replacement. Section 7 examines the use of illi in questions with object extraction, and section 8 is a conclusion.

2. Subject extraction and Criterial Freezing

In this section, I provide a brief background on subject extraction as well as the Criterial Freezing approach.

2.1 Subject extraction

Subject extraction has been the focus of a great deal of synchronic work cross-linguistically (e.g., Taraldsen Reference Taraldsen, Hellan and Christensen1986, Campos Reference Campos1997, Rizzi and Shlonsky Reference Rizzi, Shlonsky, Gärtner and Sauerland2007, Shlonsky Reference Shlonsky, Aboh, Guasti and Roberts2014, Abe Reference Abe2015). A challenge with subject extraction is that it is not straightforward to determine the extraction site of the subject. Is the subject extracted from a postverbal position where it is thematically merged, or from a preverbal position to which it moves for EPP? See Demuth Reference Demuth1995, Potsdam Reference Potsdam, Gärtner, Law and Sabel2006, and Rizzi Reference Rizzi and Svenonius2014 for more discussion. For present purposes, I limit discussion to cases where the subject is questioned, as in the examples in (1) from English and Québec French).

  1. (1)

    1. a. Who saw John?

    2. b. Qui  qui  est  venu?

      who  qui  has  come

      ‘Who has come?’  (Rizzi and Shlonsky Reference Rizzi, Shlonsky, Gärtner and Sauerland2007: 130)

Note that in Québec French the particle qui is used, whereas no similar particle is used in English. The presence of qui indicates that the subject wh-word (also qui) moves to the left periphery in Québec French, while there is no apparent evidence for movement of who in English, (for instance, there is no subject-verb inversion). This difference between English and Québec French is indicative of the fact that the two languages use different strategies to extract the subject. This variability in subject extraction strategies follows from the so-called Subject Criterion which requires each language to resort to some strategy when the subject is extracted. In the following subsection, I discuss this Criterion as well as the Criterial Freezing Approach under which the Subject Criterion is operative.

2.2 Criterial Freezing

Criterial Freezing, first formalized and labelled by Rizzi (Reference Rizzi, Cheng and Corver2006), is defined as a constraint on movement and chain formation. It ensures a three-way mapping between a syntactic phrase, a particular syntactic position, and a particular scope-discourse interpretation (Shlonsky Reference Shlonsky, Aboh, Guasti and Roberts2014: 59). The basic idea behind Criterial Freezing is that an element is first merged in a position in which it is semantically selected, and then it may be internally merged in a position that is dedicated to scope-discourse semantics, resulting in a chain that must terminate in the latter ‘criterial’ position. This restriction on any further movement of the element attracted by criterial features is attributed to what Rizzi (Reference Rizzi, Cheng and Corver2006) calls Criteria, which are defined as configurations in which a head shares a major interpretable feature with its specifier. These features include [Q], [TOP], [FOC], and [R] for questions, topics, foci, and relatives, respectively (Rizzi Reference Rizzi and Shlonsky2015). When a criterion is satisfied by a phrase with the matching criterial feature, the relevant phrase is frozen in place (Rizzi Reference Rizzi, Cheng and Corver2006: 110). Criterial Freezing is thus “an economy condition that ensures a unique correlation of heads of chains, syntactic positions, and specific interpretative properties” (Shlonsky Reference Shlonsky, Aboh, Guasti and Roberts2014: 79). Note that Criterial Freezing is different from a probe-goal relation (Chomsky Reference Chomsky and Kenstowicz2001) in requiring a spec-head relation between the criterial head and a phrase with a matching criterial feature. Additionally, again unlike probe-goal relations, movement is not a reflex of ϕ-Agree in Criterial Freezing.

As an example of Criterial Freezing, Rizzi (Reference Rizzi, Cheng and Corver2006: 112) argues that (2b) below is ungrammatical because the same wh-element which book cannot fulfil the requirements of the indirect question and the main question, even if both are included in the same sentence. (2a) is an intermediate step in the derivation of (2b).Footnote 3

  1. (2)

    1. a. Bill wonders [which book CQ [she read <which book> ]]

    2. b. * Which book CQ does Bill wonder [ <which book> CQ [she read <which book> ]]?

Once the wh-element which book satisfies the Question Criterion in the embedded C-system, it becomes frozen in place in the specifier of Focus Phrase of the embedded question. Which book’s movement to the main C-system is consequently prohibited. Note here that the wh-element which book cannot move to the main C-system, unless it moves first to the embedded C-system, given locality considerations which force successive-cyclic movement. Criterial Freezing is formulated as in (3).

  1. (3) Criterial Freezing (Rizzi Reference Rizzi and Panagiotidis2010: 149)

    A phrase meeting a criterion is frozen in place.

In view of this, Criterial Freezing is thought of as an economy condition that minimizes movement and determines optimally simple chains with unique occurrences of the fundamental ingredients: unique θ-role and unique scope-discourse marking (Rizzi Reference Rizzi, Cheng and Corver2006). A direct consequence of Criterial Freezing is that criteria cannot be satisfied in passing with unpronounced copies; if it could, (2b) would be grammatical because the copy of the wh-element which book would satisfy the Question Criterion in the lower C-system.

I now turn to the Subject Criterion (Rizzi Reference Rizzi, Brugè, Giusti, Munaro, Schweikert and Turano2005, Reference Rizzi, Cheng and Corver2006, Reference Rizzi and Svenonius2014; Rizzi and Shlonsky Reference Rizzi, Shlonsky and Frascarelli2006, Reference Rizzi, Shlonsky, Gärtner and Sauerland2007), a criterion responsible for movement of lexical items to [spec,SubjP].

2.3 The Subject Criterion

Rizzi and Shlonsky (Reference Rizzi, Shlonsky, Gärtner and Sauerland2007) formulate the Subject Criterion, drawing on the observation that subjects have a particular interpretive property, on a par with topics. This interpretive property is the ‘aboutness’ relation that links subjects with predicates as it links topics with comments. Movement to the subject position thus has interpretive consequences: the argument selected as the subject is the starting point in the description of an event, which is presented as being about the selected argument (Rizzi Reference Rizzi and Panagiotidis2010: 151). Subjects, especially when they appear preverbally, correspond to given information. This interpretive property of preverbal subjects gives rise to the assumption that the subject position cannot be filled by copies of moved elements. This constraint on any movement out of the subject position and the requirement to fill it is implemented as the Subject Criterion, which is satisfied once the specifier position of a dedicated functional phrase, namely Subject Phrase (SubjP) is filled. SubjP forms a bridge between the C-domain and the T/I-domain. Rizzi and Shlonsky (Reference Rizzi, Shlonsky, Gärtner and Sauerland2007: 149) formulate the Subject Criterion as in (4).

  1. (4) Subject Criterion:

    The functional head Subj attracts a nominal to its specifier and determines the subject-predicate articulation.

Any phrase that occupies [spec,SubjP] therefore meets a Criterion; hence this phrase resists any further movement to a distinct and higher criterial position (Rizzi and Shlonsky Reference Rizzi, Shlonsky, Gärtner and Sauerland2007: 149). In this regard, Rizzi and Shlonsky (Reference Rizzi, Shlonsky, Gärtner and Sauerland2007) propose that the classical EPP can be advantageously reanalysed as the Subject Criterion. As a result, expletives of various kinds are taken as direct evidence for the Subject Criterion. In structures where there is no external argument of the verb, the subject position must be expressed by a non-referential pronominal element in non–null-subject languages. This is shown for English in (5) and for French as in (6).

  1. (5)

    1. a. There came a man.

    2. b. It seems that John left.

  1. (6) Il  semble  que  Jean  est  parti.

    It  seems  that  Jean  is  left

    ‘It seems that Jean has left.’  (adapted from Rizzi Reference Rizzi, Cheng and Corver2006: 119)

On the other hand, the Subject Criterion poses a problem for subject extraction. That is because the Subject Criterion is normally satisfied by the thematic subject. Once the thematic subject satisfies this criterion, it cannot move farther to an upper position with a different criterial property, such as [spec,FocP] (Abe Reference Abe2015: 159). What this implies is that when the subject meets the requirement of the Subject Criterion by moving to [spec,SubjP], the subject cannot be extracted, as in fact it can be in questions. This is a state of affairs that every language must deal with (Biberauer et al. Reference Biberauer, Holmberg, Roberts and Sheehan2009: 20). In this respect, Rizzi and Shlonsky (Reference Rizzi, Shlonsky, Gärtner and Sauerland2007) show that different languages use different strategies to deal with this problem. They argue that such strategies fall into two broad categories:

  1. (7)

    1. a. Fixed-subject strategies:

      The subject does not move but remains in its freezing position in [spec,SubjP], and a well-formed Aʹ-construction involving the subject is obtained, either with no movement at all (i.e., resumption, as in Hebrew) or by movement of a larger constituent, including the frozen subject (i.e., clausal pied-piping, as in Imbabura Quechua).

    2. b. Skipping strategies:

      The subject moves, but it is allowed to skip the freezing position and is extracted directly from its thematic position or from some other predicate-internal position, as in Italian and French.

Examples of these strategies are given below.

In Hebrew, a resumptive pronoun can satisfy the Subject Criterion as in (8a) below. Any further movement of the resumptive pronoun renders the respective sentence ungrammatical, as in (8b, c) (The resumptive pronoun that fills [spec,SubjP] is in boldface. Examples are adapted from Rizzi and Shlonsky Reference Rizzi, Shlonsky, Gärtner and Sauerland2007: 120):

  1. (8)

    1. a. kaniti  et  ha-šulxan  še  xana  amra  še  dalya  ta'ana  še

      (I) bought acc the-table that Hannah said that Dalya claimed that

      hu ya'ale harbe kesef.

      he will cost a lot money

      ‘I bought the table that Hannah said that Dalya claimed that will cost a lot of money.’

    2. b. * kaniti  et  ha-šulxan  še  xana  amra  še  hu Dalya  ta'ana

      (I) bought  acc  the-table  that  Hannah  said  that  he  Dalya  claimed

      še  <hu>  ya'ale   harbe  kesef.

      that   he  will cost  a lot  money

    3. c. * kaniti  et  ha-šulxan  še  hu  xana  amra  še  <hu>  Dalya

      (I) bought  acc  the-table  that  he  Hannah  said  that  he  Dalya

      ta'ana  še  <hu>  ya'ale  harbe  kesef.

      claimed  that  he will cost  a lot  money

Hebrew resorts to the resumptive strategy to fill [spec,SubjP] when the subject is extracted or relativized. On the other hand, Imbabura Quechua employs clausal pied-piping of the whole CP that contains [spec,SubjP] when the subject satisfies another criterion. Consider the following examples, adapted from Rizzi and Shlonsky (Reference Rizzi, Shlonsky, Gärtner and Sauerland2007: 124):

  1. (9)

    1. a. * pi  -taj  Maria  -ka chayamu-shka  -ta  kri -n?

      who  q Maria  top  arrive-nmlz acc  believe agr

      ‘Who does Maria believe (that) has arrived?’

    2. b. [pi  chayamu-shka  -ta  -taj]  Maria  kri  -n?

      who  arrive-nmlz  acc  q Maria  believe  agr

      ‘Who does Maria believe (that) has arrived?’

      Lit. ‘[Who has arrived] does Maria believe?’

The question in (9a) is ungrammatical because the subject wh-word pi ‘who’ satisfies the Subject Criterion in the embedded clause pi chayamu-shka-ta-taj ‘who has arrived’. Then pi moves to [spec,FocP] in the main clause C-system, violating as such the Subject Criterion. On the other hand, in the grammatical (9b), the Subject Criterion is not violated because the whole embedded clause is pied-piped when the subject wh-word pi moves to the upper C-system. The pied-piping option allows the subject to bypass Criterial Freezing. Hebrew and Imbabura Quechua represent cases of fixed subject strategies: resumptive and clausal pied-piping, respectively.

On the other hand, in null-subject languages, the subject is extracted directly from its thematic position to the left periphery, and [spec,SubjP] is filled by an expletive pro. Consider the Italian example in (10).

  1. (10) Chi  credi  che  vincerà?

    who  think  comp  will.win

    Who do you think will win? (adapted from Rizzi and Shlonky (Reference Rizzi, Shlonsky, Gärtner and Sauerland2007: 126)

An expletive pro is assumed to merge in [spec,SubjP], allowing the thematic subject to escape the effects of Criterial Freezing.Footnote 4

Few studies have been conducted on subject extraction in Arabic and related dialects, despite the fact that Arabic is a null-subject language with two apparently unmarked word orders – SVO and VSO – which makes subject extraction more complicated to account for. The surveyed literature related to subject extraction in Arabic focused largely on a different specific issue: the agreement discrepancies between these two unmarked word orders (Kenstowicz Reference Kenstowicz, Jaeggli and Safir1989, Mohammad Reference Mohammad, Eid and McCarthy1990, Soltan Reference Soltan and Boeckx2006, and Aoun et al. Reference Aoun, Benmamoun and Choueiri2010). No comprehensive study has directly examined how the subject is extracted in Arabic. Along these lines and set within the cartographic framework (cf. Rizzi Reference Rizzi and Haegeman1997, Cinque Reference Cinque1999; Belletti Reference Belletti and Rizzi2002, among many others), the present study provides a theoretical account of subject extraction in JA, a language that has never been investigated in this regard. It will become clear that JA appeals to a set of skipping strategies in avoiding violations of the Subject Criterion, that is, the subject is allowed to skip the freezing position (i.e., [spec,SubjP]) and is extracted directly from its thematic position, as in French and Italian. [spec,SubjP] is filled by either the particle illi, a time-point adjunct, a place-point adjunct or an expletive pro, depending on whether the subject wh-word is D-linked. It is the role of D-linking that makes JA special, as JA imposes D-linking-related restrictions on the category occupying [spec,SubjP].

I turn now to the basic facts of subject extraction in JA.

3. Subject extraction facts in JA

In JA, subject extraction is sensitive to a series of factors, including:

  1. i. the form of the verb (overtly inflected for agreement or not);

  2. ii. the presence of a time-point adjunct or a place-point adjunct between the subject wh-word and the verb;

  3. iii. the type of the question (existential vs. non-existential).

When the verb shows [3SG.M] agreement, the particle illi is optionally used when the subject is extracted (I will show in section 4 that illi is a D-linked particle with an XP categorial status). Consider the examples in (11), where the verb shows [3sg.m] agreement:Footnote 5

  1. (11)

    1. a. miin  (illi)  ʔaχað  l-mafatiiħ  imbaariħ?

      who  illi  took.3sg.m  def-keys  yesterday

      ‘Who took the keys yesterday?’

    2. b. miin  (illi)  rawwah?

      who  illi  went home.3sg.m

      ‘Who has gone home?’

On the other hand, when the verb is inflected for agreement other than [3SG.M], illi is obligatorily used, unless there is a time-point adjunct or a place-point adjunct between the subject wh-word and the main verb, as in (12).

  1. (12)

    1. a. miin  *(illi)  ʔaχað-at  l-mafatiiħ  imbaariħ?

      who  illi  took-3sg.f  def-keys  yesterday

      ‘Who took the keys yesterday?’

    2. b. miin  imbaariħ  ʔaχað-at  l-mafatiiħ?

      who  yesterday  took-3sg.f  def-keys

      ‘Who took the keys yesterday?’

    3. c. miin  illi  rawwaħ-at  bakiir  imbaariħ?

      who  illi  went.home-3sg.f  early  yesterday

      ‘Who went home early yesterday?’

    4. d. miin  imbaariħ  rawwaħ-at  bakiir?

      who  yesterday  went.home-3sg.f  early

      ‘Who went home early yesterday?’

In (12a) and (12c), the subject is extracted, and the verb is inflected for agreement other than [3sg.m]; illi is therefore required. On the other hand, there is no need for illi in (12b) and (12d) as the time-point adjunct imbaariħ ‘yesterday’ shows up between the subject wh-word miin and the main verb ʔaχaðat and rawwaħat, respectively. It should be noted here that in transitive sentences, the fronted adjunct that replaces illi must be a time-point adjunct; otherwise the question is ungrammatical. For instance, if the time-point adjunct imbaariħ is replaced with the locative adjunct biddukaanah ‘in the shop’, the resulting question is ungrammatical, unless illi appears to the right of the locative adjunct, as shown in (13).

  1. (13) miin  bi-ddukaanah  *(illi)  ʔaχað-at  l-mafatiiħ?

    who  in-def-shop  illi  took-3sg.f  def-keys

    ‘In the shop, who took the keys?’

This restriction does not hold of intransitive sentences, as can be seen in (14), where illi is optional.

  1. (14) miin  bi-d-dukaanah  (illi)  wigʕ-at?

    who  in-def-shop  illi  fell down-3sg.f

    ‘Who fell down in the shop?’

Note that illi is obligatory with other adjuncts, such as frequency or manner adjuncts, even in intransitive sentences. This is illustrated in (15).

  1. (15)

    1. a. miin  ʕaadatan  * (illi)  ʔibtagaʕ  bi-l-midrasih?

      who  usually  illi  fall down. 3sg.  in-def-school

      ‘Who usually falls down at the school?’

    2. b. miin  ʔibwagaaħa  *(illi)  ħakat?

      who  rudely  illi  spoke.3sg.f

      ‘Who spoke rudely?’

It can also happen that illi and a fronted time-point adjunct or a place-point adjunct occur together following the subject wh-word, provided that illi shows up to the right of the fronted adjunct, forcing a topicalization reading of the adjunct, as in (16).

  1. (16) miin  imbaariħ  illi  ʔaχað-at  l-mafatiiħ?

    who  yesterday  illi  took-3sg.f  def-keys

    ‘Yesterday, who took the keys?’

Finally, when the subject is extracted in an existential sentence, illi is prohibited in the presence of the expletive fiih. In clauses without fiih, illi or a time-point or place-point adjunct is required. Consider the declarative example in (17a), and the possible questions in (17b–d).

  1. (17)

    1. a. fiih  zalamih  bi-daar-na.

      exp  man  in-house-our

      ‘There is a man in our house.

    2. b. miin  (*illi)  fiih  bi-daar-na?

      who  illi  exp  in-house-our

      ‘Who is in our house?’

    3. c. miin  *(illi)  bi-daar-na?

      who  illi  in-house-our

      ‘Who is in our house?’

    4. d. miin  hala  bi-daar-na?

      who  now  in-house-our

      ‘Who is in our house right now?’

All facts related to the use of illi while the subject is extracted can be summarized in Table 1, below.

Table 1: The use of illi in questions with subject extraction

In what follows, I develop an account of these observations, proposing a unified analysis from which all of them follow.

4. illi and the Subject Criterion

In this section, I capitalize on the assumption that illi fills [spec,SubjP] in JA, beginning with some background information, and a brief summary of what previous approaches have said about illi in connection with subject extraction.

4.1 The grammatical function of illi

As shown in section 3, one way a subject can be extracted in JA is for illi to appear to the left of the verb. Several studies on some Arabic dialects, including Egyptian Arabic (Wahba Reference Wahba1984, Gad Reference Gad2011), Gulf Arabic (Holes Reference Holes1990), Palestinian Arabic (Shlonsky Reference Shlonsky1992, Reference Shlonsky, Ouhalla and Shlonsky2002), and Najrani Arabic (Fakih Reference Fakih2014), discussed the function of illi in questions and relative clauses.Footnote 6 As can be seen from looking at these studies, no consensus has been reached on the actual syntactic function of this particle. In general, there are two approaches to the function of illi. Several researchers take illi to be a complementizer, occupying the head position of the CP (i.e., the head of the Force Phrase in the sense of Rizzi's Reference Rizzi and Haegeman1997 split CP hypothesis), while others assume that illi heads the Focus Phrase. In what follows, a sketch of these two approaches is provided.

4.1.1 Previous approaches to illi

Treating illi as a complementizer is by and large the most commonly held view. For instance, Osman (Reference Osman1990: 50) characterizes illi as an invariant complementizer when it is used in questions in Egyptian Arabic. He claims that illi has no morphology and heads the CP complement of a relative clause. He refers to questions with illi as relativized wh-questions. Shlonsky (Reference Shlonsky1992: 451) differentiates between two complementizers in Palestinian Arabic: ʔinno and ʔilli. He assumes that ʔinno is a complementizer that signals regular subordination, whereas ʔilli only heads CPs that serve as predicates. Thus, the use of illi in relative clauses of all kinds, clefts, and interrogative clauses is, for him, justified. In addition, following Rizzi's (Reference Rizzi1990) feature system for classifying complementizers, Shlonsky (Reference Shlonsky1992) postulates that ʔinno is a [-predicational] C°, while ʔilli is a [+ predicational] C°.

The same view is reiterated by Shlonsky (Reference Shlonsky, Ouhalla and Shlonsky2002: 139) commenting on the strategies employed in the formation of constituent questions (e.g., which man…?) in Palestinian Arabic. There, Shlonsky asserts that in forming constituent questions, a fronted wh-constituent is followed by the complementizer ʔilli. He introduces the following example:

  1. (18) ʔani  bint  ʔilli  l-ʔasad  ʔakal-ha  mbaarih?

    which  girl  comp  def-lion  ate.3sg.m-3sg.f  yesterday

    ‘Which girl did the lion eat yesterday?’

In (18), the fronted wh-constituent ʔani bint ‘which girl’ is followed by ʔilli, which is assumed to be a complementizer.

Illi has been treated as a complementizer in questions and relative clauses in many studies of other Arabic dialects such as Syrian Arabic (Brustad Reference Brustad2000), Moroccan Arabic (Benmamoun Reference Benmamoun2000), Modern Standard Arabic (Soltan Reference Soltan2010), Jordanian Arabic (Al-Momani Reference Al-Momani2015) and Najdi Arabic (Lewis Reference Lewis2013). For these studies, illi is a phonologically reduced counterpart of Modern Standard Arabic's relative complementizer ʔallaði.

On the other hand, neither Osman (Reference Osman1990) nor Shlonsky (Reference Shlonsky1992) provides a theoretical motivation for using illi in direct questions. Treating illi as a complementizer seems ad hoc without independent evidence. For example, although some studies treat illi as a complementizer in questions, no argument has been advanced explaining why it is obligatory in questions with subject extraction but optional in questions with object extraction.Footnote 7 Likewise, it is clearly evident from the data, some of the above-mentioned studies, especially in Palestinian Arabic and Egyptian Arabic, that illi is optional in questions with subject extraction when the verb shows [3SG.M] agreement. However, none of the studies cited provide an account of this pattern. Rather, the main emphasis has been placed on the pragmatic motivations of such questions (Shlonsky Reference Shlonsky, Ouhalla and Shlonsky2002: 141).

These reasons, along with other arguments (mentioned below) have led some authors to question the assumption that illi is a complementizer when it is used in questions. These studies have argued that illi is a focus particle heading the Focus Phrase, a separate layer within the Split CP domain (cf. Rizzi Reference Rizzi and Haegeman1997).

For example, Gad (Reference Gad2011: 173) argued that illi is not a complementizer in Egyptian Arabic, due to the difference in distribution between illi and other complementizers. Gad bases her argument on two interrelated observations. First, the complementizer inn precedes the clausal complements of some verbs, as shown in (19a) below, while illi introduces headless relative clauses which occur in an argument position, as shown in (19b), and is used in questions, as in (19c).Footnote 8

  1. (19)

    1. a. Mona  sadda'i-t  inna-ha  faaz-t  bi-l-gayza

      Mona  believed-3sg.f  that-3sg.f  won-3sg.f  with-def-prize

      ‘Mona believed that she won the prize.’

    2. b. illi  ʕirif  ħall  il-fazuura  kisib  filuus.

      that  knew.3sg.m  answer  def-puzzle  won.3sg.m  money

      ‘The one who knows the answer of the puzzle won money.’

    3. c. miin  illi  Mona  itgawwiz-t-uh?

      who  focus  Mona  married-3sg.f-3sg.m

      ‘Who did Mona marry?'

Second, Gad argues that inn and illi cannot be used interchangeably. If illi and inn are substituted for each other in (19), the resulting sentences are sharply ungrammatical. Although such differences can be accommodated with Shlonsky's (Reference Shlonsky1992) analysis following Rizzi's (Reference Rizzi1990) feature system (–predicational vs. + predicational), they are nonetheless taken by Gad (Reference Gad2011) to be an argument against illi being a complementizer. Furthermore, following Cheng (Reference Cheng1997), Gad rejects the possibility that illi is a wh-particle when it is used in questions, on the assumption that a wh-question cannot have both a wh-phrase and a wh-particle. She assumes instead that illi is a relative pronoun heading the Focus Phrase in wh-questions. Under her proposal, there is a strong Focus feature that forces wh-movement to occur before Spellout, causing the wh-phrase to appear to the left of illi. Additionally, in order to account for the restriction that adjunct wh-phrases cannot co-occur with illi, she claims that illi and the argumental wh-phrases carry [+nominal] features, and that illi and the wh-word have to agree in categorial features. See Fakih (Reference Fakih2014) for a similar analysis of Najrani Arabic, arguing that illi is a morphological realization of the strong Focus feature in the Focus Phrase.

However, this proposal for illi leaves several questions unanswered. For example, why is illi as a Focus head not used (or not realized) in clauses with argument focalization? Why is illi optional in questions with object extraction but obligatory in most questions with subject extraction? Is the illi used in questions different from the one that introduces relative clauses? If so, what are the differences? If not, then relative clauses would effectively be derived by focalization, contrary to current generative reasoning on the derivation of relative clauses. In recent syntactic theory, relative clauses are marked by the feature [REL] on the head of the Force Phrase whose specifier serves as a host for overt wh-pronouns or a null relative operator (Vries Reference de Vries2002, Arsenijević Reference Arsenijević2009, Sullivan Reference Sullivan2016, among others). In contrast to these two alternatives, I propose that illi is used in questions with subject extraction in order to obviate the effects of Criterial Freezing. This proposal is argued for in the following subsection.

4.1.2 illi as a [spec,SubjP] element

I argue that illi is obligatorily used in questions with subject extraction in order to satisfy the requirement of the Subject Criterion that [spec,SubjP] be filled. This amounts to saying that illi as an XP rather than X0. The Subject Criterion forces [spec,SubjP] to be filled by some constituent. In declarative sentences, the normal occupant of [spec,SubjP] is the thematic subject, hence the unmarked word order is SVO. Since the verb in JA bears rich inflectional morphology, it follows that it moves to T° (Bobaljik Reference Bobaljik2000, Holmberg and Roberts Reference Holmberg and Roberts2013).Footnote 9 The appearance of the subject to the left of the tensed verb as well as the strict adjacency required between them is indicative of a spec–head relation between them. The verb moves to T°, while the subject moves to [spec,SubjP]. This being so, the declarative sentence in (20) has the derivation in (21).

  1. (20) l-wlaad  ʔaχað-u  l-mafatiiħ.

    def-boys  took.3m.pl  def-keys

    ‘The boys took the keys.’

  1. (21)

The subject lwlaad ‘the boys’ moves to [spec,SubjP] to satisfy the Subject Criterion.

On the other hand, in cases of subject extraction, the subject would not move to [spec,SubjP]. If it did, it would resist any further movement due to the effects of Criterial Freezing. As indicated above, [spec,SubjP] is a criterial position, and criterial requirements cannot be satisfied in passing (Rizzi Reference Rizzi, Cheng and Corver2006). If the Subject Criterion could be satisfied in passing, a question like the one in (22) would be grammatical, contrary to fact:

  1. (22) * miin  ʔaχað-u  l-mafatiiħ?

    who  took-3m.pl  def-keys

    ‘Who took the keys?’

One way of rescuing (22) is by inserting illi after the subject wh-word:

  1. (23) miin  illi  ʔaχað-u  l-mafatiiħ?

    who  illi  took-3m.pl  def-keys

    ‘Who took the keys?’

I propose that illi acts as a facilitator for the subject lwlaad, enabling it to move directly to the left periphery directly from its thematic position. The requirement that [spec,SubjP] be filled is then satisfied by illi, and there is no need for the subject to move to [spec,SubjP].

The question that arises at this point concerns the presence of agreement suffixes on the verb. If, as proposed here, the subject wh-word miin in (23) does not move to [spec,SubjP], why does the agreement suffix –u [3M.PL] appear on the verb ʔaχað? According to much related literature on Modern Standard Arabic and some other Arabic dialects, this suffix is a morphological manifestation of spec-head agreement between the subject and T°, because the subject moves to [spec,TP] (here [spec,SubjP]) (Benmamoun Reference Benmamoun, Adger, Pintzuk, Plunkett and Tsoulas1998, Harbert and Bahloul Reference Harbert, Bahloul, Shlonsky and Ouhalla2002). Under this approach, the present analysis does not predict any agreement, since the subject does not move to [spec,SubjP]. Abstracting slightly from the specific assumptions made by Benmamoun (Reference Benmamoun, Adger, Pintzuk, Plunkett and Tsoulas1998) and Harbert and Bahloul (Reference Harbert, Bahloul, Shlonsky and Ouhalla2002), and following Chomsky (Reference Chomsky and Kenstowicz2001 et seq.), I propose that this suffix is a morphological manifestation of agreement between T° and the subject in [spec,vP]. This proposal is compatible with the notion that agreement between the subject and T° takes place before subject extraction. No spec-head agreement happens between the subject and T°, because if it did, the subject would have had to remain frozen in [spec,SubjP] due to the Subject Criterion. It would thus have been unable to move to left periphery to satisfy the Question Criterion.

Following Chomsky (Reference Chomsky, Martin, Michaels and Uriagereka2000, Reference Chomsky and Kenstowicz2001) on agreement, I argue that agreement between the subject and T° is established while the former is in situ through a probe-goal relation. Chomsky (Reference Chomsky, Martin, Michaels and Uriagereka2000, Reference Chomsky and Kenstowicz2001) reformulates Agree so that movement is seen as a last resort. Instead of viewing Move as the basic tool for valuing uninterpretable features, Move is only triggered to save a derivation that would otherwise be doomed to failure. Instead, uninterpretable/unvalued features are valued by the operation Agree, as formulated in (24).

  1. (24) The probe α agrees with the goal ß providing that:

    1. a. α has uninterpretable Φ-features;

    2. b. ß has matching interpretable Φ-features;

    3. c. ß is active by virtue of having an unvalued Case feature;

    4. d. α c-commands ß;

    5. e. there is no potential goal ϒ intervening between α and ß. (Chomsky Reference Chomsky and Kenstowicz2001: 122)

With this reasoning of the operation Agree in place, the presence of the agreement suffixes attached to the verb directly follows. With reference to the question in (23), reproduced below as (25), let's start from the point where the subject wh-word miin enters the derivation.

  1. (25) miin  illi  ʔaχað-u  l-mafatiiħ?

    who  illi  took-3m.pl  def-keys

    ‘Who took the keys?’

The wh-word miin, bearing interpretable ϕ-features, merges externally with little v′, forming the v*P. Next, v*P merges with T°, yielding TP. TP then merges first with Subj0 and then with C°, which comes with a set of unvalued ϕ-features, an EPP feature, and an interpretable tense feature, projecting CP. Following Chomsky's (Reference Chomsky, Gärtner and Sauerland2007) proposal of feature inheritance, C°’s features are inherited by T°.Footnote 10 T° then searches for an element with matching interpretable ϕ-features. This search is triggered by the fact that T°’s uninterpretable features are inherently unvalued, but need to be assigned a value in the course of the derivation. They play no role in semantic interpretation and consequently must be deleted before convergence at LF (Chomsky Reference Chomsky, Gärtner and Sauerland2007: 18). T°’s search is limited in domain because it is governed by the so-called the Phase Impenetrability Condition, is stated in (26).

  1. (26) The Phase Impenetrability Condition (PIC)

    In Phase α with head H, the domain of H is not accessible to operations outside α, only H and its edge are accessible to such operations (Chomsky Reference Chomsky and Kenstowicz2001: 14).

The terms of the PIC can be diagrammed as follows:

  1. (27)

T° searches for elements with interpretable ϕ-features within its visible domain as defined by the PIC, i.e., its complement, the edge of the v*P phase, and the head v*. The domain of T°’s search extends only as far as the v head; the boxed material in (28) has been transferred and is no longer accessible.

  1. (28)

Probing down, T° finds the subject, miin, as an active goal with matching interpretable features and unvalued structural Case. A probe-goal relation is established between T° and miin. As a result, the uninterpretable features of T° are valued as [3M.PL], determining the form of the verb at PF; at LF, such features are deleted. As an outcome of this valuation, the agreement marker –u [3M.PL] is realized on the verb ʔaχað.Footnote 11

One remark on the ϕ-features of the subject wh-word is in order. As is clear from all examples above, the subject wh-word miin occurs invariably with all agreement markers.Footnote 12 The actual specification of interpretable ϕ-features of miin can be made overt once a probe-goal configuration is established between it and T°. In question (29a) below, the speaker asks about the girl who took the keys. This is shown by the agreement marker –t [3SG.F] suffixed to the verb. In (29b) the speaker asks about the girls who took the keys, resulting in the [3 F.PL] agreement marker –n.

  1. (29)

    1. a. miin  illi  ʔaχað-at  l-mafatiiħ?

      who  illi  took-3sg.f  def-keys

      ‘Who took the keys?’

    2. b. miin  illi  ʔaχað-n  l-mafatiiħ?

      who  illi  took-3f.pl  def-keys

      ‘Who took the keys?’

Note that it is not sufficient, for the derivation to converge, that the uninterpretable ϕ-features on T° be valued. The demands of the Subject Criterion must also be satisfied. As shown above, in declarative clauses the Subject Criterion is normally satisfied by movement of the subject to [spec,SubjP]. Nonetheless, such a derivation is not available with subject extraction, because the subject wh-word miin would resist movement from [spec,SubjP], as indicated in (30).

  1. (30)

Given that miin has an uninterpretable [Q] feature, it must raise to [spec,FocP], attracted by the [Q] feature on Foc° to satisfy the Question Criterion (cf. Rizzi Reference Rizzi, Cheng and Corver2006). Accordingly, at one point of the question derivation, the Subject Criterion and Question Criterion both need the subject wh-word miin to meet their requirements at the same time. This state of affairs can be seen as a conflict, since both criteria are active and must be satisfied. One might entertain the possibility that miin satisfies the Subject Criterion and hence does not undergo a CP-related movement satisfying the Question Criterion. However, this possibility is ruled out. On the basis of JA data, the Question Criterion wins in attracting the subject wh-word. This indicates that JA employs some strategy to satisfy the Subject Criterion.

Some empirical evidence for movement of miin to the CP-system comes from its position relative to the CP-particle ʃikil. Current research on particles such as ʃikil, which signals the speaker's attitude toward the sentence, confirms the CP-related status of such particles (Struckmeier Reference Struckmeier2014). As shown from the contrast between the two sentences in (31), miin must appear to the left of the CP-particle ʃikil, indicating that miin indeed moves to the left periphery, satisfying the Question Criterion.Footnote 13

  1. (31)

    1. a. miin   ʃikil-hum  illi  ʔaχað-u  l-mafatiiħ?

      who   prt-3m.pl  illi  took-3m.pl  def-keys

      ‘Who has apparently taken the keys?’

    2. b. * ʃikil-hum  miin  illi  ʔaχað-u  l-mafatiiħ?

      prt-3m.pl  who  illi  took-3m.pl  def-keys

      Intended: ‘Who has apparently taken the keys?’

The subject wh-word miin must appear in a position to the left of the discourse particle ʃikil, implying its movement to CP. This discussion does not imply that the requirements of the Subject Criterion need not be met. Rather, they must be satisfied by some other mechanism ensuring that [spec,SubjP] is filled. If the subject wh-word cannot meet the requirements of the Subject Criterion, illi, as an XP, is used instead. Under this analysis, the question in (25), reproduced below as (32), has the structure in (33).

  1. (32) miin  illi  ʔaχað-u  l-mafatiiħ?

    who  illi  took-3m.pl  def-keys

    ‘Who took the keys?’

  1. (33)

Some evidence for illi being in [spec,SubjP] can be induced similarly by its position relative to the CP-particle ʃikil. Under no circumstances can illi appear to the left of ʃikil. Consider the following ill-formed example:

  1. (34) * miin  illi   ʃikil-ha  ʔaχað-at  l-mafatiiħ?

    who  illi  prt-3sg.f  took-3sg.f  def-keys

    Intended: ‘Who has evidently taken the keys?’

As a discourse-related particle, ʃikil occupies a discourse-related position. If this position is the lower Topic Phrase (in the sense of Rizzi Reference Rizzi and Haegeman1997), then illi is positioned below the lower Topic Phrase. Further evidence that illi is below CP comes from its position relative to the position of topicalized elements that appear in JA to the left of illi. For instance, in (35) the topicalized time-point adjunct imbaariħ ‘yesterday’ appears to the left of illi but to the right of the displaced subject wh-word miin, indicating that imbaariħ is fronted to the Topic Phrase, below the Focus Phrase (Rizzi Reference Rizzi and Haegeman1997).

  1. (35) miin  imbaariħ  illi  ʔaχað-at  l-mafatiiħ?

    who  yesterday  illi  took-3sg.f  def-keys

    ‘Yesterday, who took the keys?’

The same adjunct imbaariħ can appear to the left of miin, landing in the upper Topic Phrase, above the Focus Phrase (Rizzi Reference Rizzi and Haegeman1997).

  1. (36) Imbaariħ  miin  illi  ʔaχað-at  l-mafatiiħ?

    yesterday  who  illi  took-3sg.f  def-keys

    ‘Yesterday, who took the keys?’

Additionally, miin can be sandwiched between two topicalized elements, both of which must precede illi, as in (37).

  1. (37) Imbaariħ  miin  ʔibsurʕah  illi  ʔaχað-at  l-mafatiiħ?

    yesterday  who  quickly  illi  took.3sg.f  def-keys

    ‘Yesterday, who quickly took the keys?’

Following the assumption that there is a single Focus Phrase in a single clause (É. Kiss Reference É Kiss1995, Rizzi Reference Rizzi and Haegeman1997), examples (35)–(37) are bone fide evidence for the lower position that illi occupies. Assuming that miin is in [spec,FocP], imbaariħ ‘yesterday’ is positioned in the upper Topic Phrase (labelled as TopP (1) in (38)), and ʔibsurʕah is in the lower Topic Phrase (labelled as TopP (2)).

  1. (38)

Further evidence that illi occupies [spec,SubjP] comes from the observation that the fronted direct object lmafatiiħ ‘the keys’ does not appear between illi and the verb. This shows that there is no syntactic position available for the fronted direct object between illi and the verb, as demonstrated by the ill-formed question in (39a). The direct object can be preposed to a position directly to the left of illi, as in (39b), or even to the left of the subject wh-word miin, as in (39c).

  1. (39)

    1. a. * miin  illi  l-mafatiiħ  ʔaχað-u-hin?

      who  illi  def-keys  took-3m.pl-3f.pl

      ‘Who took the keys?’

    2. b. miin  l-mafatiiħ  illi  ʔaχað-u-hin?

      who  def-keys  illi  took-3m.pl-3f.pl

      ‘The keys, who took them?’

    3. c. l-mafatiiħ  miin  illi  ʔaχað-u-hin?

      def-keys  who  illi  took-3m.pl-3f.pl

      ‘The keys, who took them?’

In view of this, illi is situated below CP but above T0. One conclusive piece of evidence for illi being in [spec,SubjP] comes from its complementary distribution with the expletive fiih in existential questions. When fiih is present, illi is no longer required. Consider the sentence in (40).

  1. (40) miin  fiih  bi-daar-na?

    Who  exp  in-house-our

    ‘Who is in our house?’

If illi is inserted either to the left or to the right of fiih, the resulting question is ungrammatical.

  1. (41) miin  (*illi)  fiih  (*illi)  bi-daar-na?

    who  illi  exp  illi  in-house-our

    ‘Who is in our house?’

However, illi can replace fiih in existential questions, a reliable sign that illi fills the same position as fiih. Thus, illi and fiih compete for the same structural position, complying with the strong requirements of the Subject Criterion.Footnote 14

  1. (42) miin  illi  bi-daar-na?

    who  illi  in-house-our

    ‘Who is in our house?’

Further evidence that illi occupies [spec,SubjP] is that it appears in embedded clauses when long-distance subject extraction occurs. Consider the sentence in (43).

  1. (43) miin  ʔaboo-i  fakkar  illi  sarag  ʔis-siyaarh

    who  father-my  believed.3sg.m  illi  stole.3sg.m  def-car

    ‘Who did my father believe that stole the car?’

In (43), the embedded subject miin undergoes long extraction from its base position to the left periphery of the matrix clause, moving successive-cyclically via [spec,ForceP] in the embedded C-system, which is not a criterial position. Following the Criterial Freezing approach, the embedded subject can move no farther if it occupies [spec,SubjP] of the embedded clause. In order for the subject wh-word miin to move out of the embedded clause, the embedded [spec,SubjP] must be filled by another element, leaving the subject free to be extracted. It is thus clear that illi in (43) is used to escape the effects of Criterial Freezing.

Gad's (Reference Gad2011) proposal that illi is a focus particle is therefore not borne out, as far as JA is concerned. Rather, illi is used to satisfy the requirements of the Subject Criterion. In addition, the findings for JA thus far contradict the long-standing view (Shlonsky Reference Shlonsky, Ouhalla and Shlonsky2002) that illi is an overt complementizer in questions.

The question that now arises is why illi is optional in questions if the verb displays [3SG.M] agreement. I argue that the answer to this question lies in the assumption that illi is a D-linked element that satisfies the Subject Criterion only when the subject wh-word is D-linked, a proposal I discuss in the next section.

5. illi as a D-linked particle

As shown in section 3, illi is more or less optional when the verb bears [3SG.M] agreement, as shown in (44).

  1. (44)

    1. a. miin  (illi)  ʔaχað  l-mafatiiħ  imbaariħ?

      who  illi  took.3sg.m  def-keys  yesterday

      ‘Who took the keys yesterday?’

    2. b. miin  (illi)  rawwah?

      who  illi  went home.3sg.m

      ‘Who has gone home?’

The sentences in (44), in which the main verb displays [3SG.M] agreement, contrast sharply with cases where the verb shows agreement other than [3SG.M] and illi cannot be omitted. Why does the absence of illi not make the question ungrammatical, even without a time-point or place-point adjunct? First and foremost, note that we cannot assume that the subject wh-word in such cases does not move to CP, but moves only to [spec,SubjP]. In (45), miin must appear to the left of the CP-particle ʃikil, and must therefore itself be in the left periphery.

  1. (45)

    1. a. miin   ʃikil-uh  ʔaχað  l-mafatiiħ?

      who  prt-3sg.m  took.3sg.m  def-keys

      ‘Who has apparently taken the keys?’

    2. b. * ʃikil-uh  miin  ʔaχað  l-mafatiiħ?

      prt-3sg.m  who  took.3sg.m  def-keys

      ‘Who has apparently taken the keys?’

It is clear that the Question Criterion is satisfied by the subject wh-word miin, even when illi does not appear, as in (44). The question to ask here is how the Subject Criterion is met under such cases. One possibility is that since JA is a pro-drop language, [spec,SubjP] is filled by an expletive pro, which satisfies the Subject Criterion. Under that view, the question then is why expletive pro seems to be available only when the verb shows only [3SG.M] agreement. I assume that the wh-word used in such cases has no ϕ-features (being non-D-linked, i.e., non-referential, as will be shown later). The uninterpretable ϕ-features of T° are accordingly valued as default, namely [3SG.M]. In line with this possibility, (44a) has the derivation shown in (46).

  1. (46)

Now, we must ask why illi is still available as a means of satisfying the Subject Criterion when the verb shows [3SG.M] agreement. Once this issue is resolved, a better account of subject extraction in questions will be possible. I begin with the observation that questions with illi differ in meaning from questions without it. When illi is present, there is a presupposition that is not made when illi is absent (cf. Shlonsky's Reference Shlonsky, Ouhalla and Shlonsky2002 similar assumptions about Palestinian Arabic). With illi, the speaker implies that the entity doing the action belongs to a set whose members are known to him/her. Consider the examples in (47).Footnote 15

  1. (47)

    1. a. miin  illi  ʔaχað-at  l-mafatiiħ?

      who  d-prt  took-3sg.f  def-keys

      ‘Who took the keys?’

    2. b. miin  illi  ʔaχað-u  l-mafatiiħ?

      who  d-prt  took-3m.pl  def-keys

      ‘Who took the keys?’

The speaker in (47a) presupposes that the person who took the keys is female rather than male, while in (47b) the speaker implies that who took the keys is a plurality of male people. In both cases, the speaker implies some discourse knowledge that forms the basis of the presupposition. One piece of evidence that questions with illi indicates a presupposition comes from the fact that they cannot be used in out-of-the-blue contexts (Frey Reference Frey, Lohnstein and Trissler2004, Rizzi Reference Rizzi, Brugè, Giusti, Munaro, Schweikert and Turano2005, Adger Reference Adger2007). Consider the example in (48).

  1. (48) ʃuuh  (*illi)  sˤaar?

    what  d-prt  happened.3sg.m

    ‘What happened?’

The question in (48) is ungrammatical with illi despite the observation that illi is in most cases optional when the verb bears [3SG.M] agreement. The ungrammaticality of (48) with illi is readily accounted for if the use of illi requires a context in which the speaker presupposes some discourse. Following this line of thought and assuming that presupposition is a discourse effect (Boeckx and Grohmann Reference Boeckx, Grohmann, Adger, de Cat and Tsoulas2004), I propose that illi is used in questions only where the subject wh-word is D-linked, that is, illi implies the existence of a set of contextually determined entities from which the speaker is asking the hearer to choose (Fernández Reference Fernández2009: 119).

Some convergent evidence for this contention can also be found with subject extraction of wh-constituents such as which boy, whose car, etc. (following Shlonsky's Reference Shlonsky, Ouhalla and Shlonsky2002 terminology). Wh-constituents are necessarily D-linked by virtue of their complement noun, which determines the set of relevant entities from which the speaker is asking the hearer to choose. When the subject is a D-linked constituent, illi must be inserted (or a time-point or place point adjunct must appear between the wh-phrase and the verb, an observation I return to in section 6) even if the verb bears [3SG.M] agreement. Consider the sentences in (49).

  1. (49) miin  z-zalamah  *(illi)  ʔaχað  l-mafatiiħ  imbaariħ?

    who  def-man  d-prt  took.3sg.m  def-keys  yesterday

    ‘Which man took the keys yesterday?’

Although the verb in (49) shows [3SG.M] agreement, illi is required. Here, the speaker asks about a man (not a woman, a child, etc.) who took the keys yesterday. Using the D-linked wh-constituent miin zzalamah ‘which man’, the speaker specifies the set of the entities (i.e., men) from which he/she asks the hearer to choose.

Further evidence for the correlation between D-linking and the use of illi comes from cases where the wh-word used is modified. In such cases, illi must again be used even if the verb shows [3SG.M] agreement. Consider (50), where the wh-word is modified by tˤawiil ‘tall’.

  1. (50) miin  l-tˤawiil  *(illi)  ʔaχað  l-mafatiiħ  imbaariħ?

    who  def-tall  d-prt  took.3sg.m  def-keys  yesterday

    ‘Which tall one took the keys yesterday?’

In (50), the speaker delimits the set of candidates for the answer to the question. With an appropriate context in mind (e.g., the speaker is addressing schoolboys) the speaker assumes that a tall person must have taken the keys, perhaps presupposing that they are kept in a place accessible only to tall people. The use of the nominal modifier tˤawiil thus presupposes the place where the keys are placed and thus provides some discourse background on the given question.

Another piece of evidence that D-linking forces the use of illi in question with subject extraction comes from answers to questions with illi when the verb bears [3SG.M] agreement. For instance, it is only felicitous to answer the question in (51) (containing illi) with a [3SG.M] entity, whereas the answer for the question in (52) (without illi) can be any entity irrespective of its ϕ-content.

  1. (51) miin  illi  ʔaχað  l-mafatiiħ?

    who  d-prt  took.3sg.m  def-keys

    ‘Who took the keys?’

    1. a. ibin  ʕamm-i

      son  uncle-my

      ‘my male cousin’

    2. b. ?? binit  ʕamm-i

      daughter  uncle-my

      ‘my female cousin’

    3. c. ?? ʔisˤħaab-i

      friends-my

      ‘my friends’

  1. (52) miin  ʔaχað  l-mafatiiħ  imbaariħ?

    who  took.3sg.m  def-keys  yesterday

    ‘Who took the keys yesterday?’

    1. a. ibin  ʕamm-i

      son  uncle-my

      ‘my male cousin’

    2. b. binit  ʕamm-i

      daughter  uncle-my

      ‘my female cousin’

    3. c. ʔisˤħaab-i

      friends-my

      ‘my friends’

In (51), which includes illi and where the verb displays [3SG.M] agreement, the answers with [3SG.M] content are more felicitous than the answers with any other ϕ-content. That is because the speaker presupposes that whoever took the key is a [3SG.M] element. In contrast, in (52) which does not include illi and where the verb again displays [3SG.M] agreement, answers with any content are felicitous. That is because the speaker does not presuppose the existence of a contextually determined set of entities from which the speaker is asking the hearer to choose. The [3SG.M] form of the verb is in such cases the result of a lack of agreement, where the verb has been assigned the default form.

In view of these pieces of evidence (subject extraction with D-linked wh-constituents, subject extraction with D-linked modified wh-words, and the possible answers for questions with illi whose verb bears [3SG.M] agreement), I argue that SubjP is still projected in questions where the verb displays default agreement. The requirement that [spec,SubjP] be filled is satisfied by an expletive pro whose null PF content does not affect its licensing in [spec,SubjP]. These observations constitute ‘The D-linking condition on the Subject Criterion’, which I formulate as in (53).

  1. (53) D-linking condition on the Subject Criterion:

    [Spec,SubjP] is filled by an element with the same D-linking status as the subject wh-word (D-linked vs. non-D-linked)

JA appears to be unique among null-subject languages in obeying the D-linking condition on the Subject Criterion.Footnote 16 Additionally, due to the requirement of matching D-linking between [spec,SubjP] and the wh-word, I propose that the condition in (53) is implemented in the derivation as agreement.

Along these lines, we can now see why illi seems to be optional when the verb displays [3SG.M] agreement. Logically speaking, when asking a question, the speaker has two options, depending on the discourse context. It could be that he/she is really asking about a male singular entity. If so, illi is required since the subject wh-word is D-linked, and the [3SG.M] form of the verb is an agreement-produced form. Alternatively, it could be that the speaker does not have in mind any particular set of entities from which he/she is asking the hearere to choose. The subject wh-word is therefore non-D-linked, and lacks agreement features. The [3SG.M] form of the verb in this case is the PF reflex of invariant default agreement. Here, illi is not an available strategy, given the D-linking condition on the Subject Criterion. [spec,SubjP] is filled by an expletive pro, a non-referential, non-D-linked element.

It is clear that the interaction with D-linking does not arise when the verb displays agreement other than [3SG.M]. In such cases the wh-word is always D-linked. The speaker here refers to a contextually identified set. For instance, when the verb is marked as [3F.PL], the speaker has identified girls as a set from which the hearer is being asked to choose a subset. Even if the speaker does not have a specific girl in mind, he/she singles out girls from the whole context, which might also include men. Thus, when the verb shows any inflected (i.e., non-default) agreement, the speaker narrows down the options from which the hearer must choose, presupposing some discourse-related background to the question.

A major line of evidence for the relation between D-linking and satisfaction of the Subject Criterion comes from the other mechanism JA provides to satisfy the Subject Criterion; that is, a time-point or place-point adjunct, to which I now turn.

6. Place-point and time-point adjuncts

A crucial observation for my analysis is the fact that the need for illi is obviated in cases like (54), even if the verb is inflected for non-default agreement:

  1. (54) miin  imbaariħ  ʔaχað-at  l-mafatiiħ?

    who  yesterday  took-3sg.f  def-keys

    ‘Who took the keys yesterday?’

In such cases, a time-point or place-point adjunct appears to the left of the verb. As shown in section 3, this adjunct must be a time-point adjunct if the sentence is transitive. For instance, if imbaariħ ‘yesterday’ is replaced with a locative PP like min l-xzaanih ‘from the closet’ as in (55) below, illi must be used if the verb is inflected for non-default agreement. Note that the locative in (55) is construed as a topicalized element when it precedes illi.Footnote 17

  1. (55) miin  min  l-xzaanih  *(illi)  ʔaχað-at  l-mafatiiħ?

    who  from  def-closet  d-prt  took-3sg.f  def-keys

    ‘Here/from the closet, who took the keys?’

In intransitive questions, on the other hand, replacing imbaariħ ‘yesterday’ with a locative adjunct as in (56b) preserves grammaticality even if the verb shows non-default agreement. This implies that the restriction on the use of locative adjuncts without illi is relaxed in intransitive questions. Note that the locative adjunct is not construed as a topicalized element in such cases, but rather has an unmarked interpretation with no topicalization or focalization.

  1. (56)

    1. a. miin  imbaariħ  wigʕ-at?

      who  yesterday  fell down.3sg.f

      ‘Who fell down yesterday?’

    2. b. miin  min  sˤ-sˤatˤiħ  wigʕ-at?

      who  from  def-floor  fell down.3sg.f

      ‘Who fell down from the (upper) floor?’

Following Holmberg (Reference Holmberg2000) and Rizzi and Shlonsky (Reference Rizzi, Shlonsky and Frascarelli2006), I assume that adjuncts in JA might play a subject-like role.Footnote 18 When there is a subject gap, adjuncts can move out of their base position, filling [spec,SubjP]. There is no need thus for illi, as the Subject Criterion is satisfied by a fronted adjunct.

One piece of evidence that the fronted adjuncts in (56) above are not in CP but rather in [spec,SubjP] comes from the fact that they do not have a discourse-scope reading (i.e., topicalization) in these examples. No intonational break is required after the time-point adjunct imbaariħ in questions like (56a) above, nor does the adjunct have a discourse-scope reading (i.e., it is not a topic nor a focus). This suggests that the adjunct imbaariħ is not in CP. Furthermore, the direct object cannot appear between the fronted adjunct and the main verb, as in (57).

  1. (57) * miin  imbaariħ  l-mafatiiħ  ʔaχað-at-hin?

    who  yesterday  def-keys  took-3sg.f-3f.pl

    Intended: ‘Yesterday, who took the keys?’

If imbaariħ ‘yesterday’ is in [spec,SubjP], then there is no structural position between it and the main verb ʔaχað.

However, this discussion does not imply that such adjuncts cannot be topicalized. They can be fronted directly to the left periphery if illi occupies [spec,SubjP]. Consider the examples in (58), where the adjunct imbaariħ is construed as a topicalized entity.

  1. (58)

    1. a. imbaariħ,  miin  *(illi)  ʔaχað-at  l-mafatiiħ?

      yesterday  who  d-prt  took-3sg.f  def-keys

      ‘Yesterday, who took the keys?’

    2. b. miin  imbaariħ  *(illi)  ʔaχað-at  l-mafatiiħ?

      who  yesterday  d-prt  took-3sg.f  def -keys

      ‘Yesterday, who took the keys?’

If topicalized imbaariħ ‘yesterday’ is positioned either to the right or to the left of the subject wh-word miin, illi must be used. Apart from the slight interpretive differences between the questions in (58), in both questions the time-point adjunct imbaariħ is construed as a topicalized entity. In (58a), an intonational break is required, represented by a following comma. In (58b), the speaker is concerned about the person who took the keys yesterday, not another day, implying, for instance, that whoever took the keys yesterday is different from the person who usually takes them.

At this point, there are two questions that deserve consideration; (1) why can only a time-point adjunct or a place-point adjunct replace illi in intransitive questions and (2) why can only a time-point adjunct replace illi in transitive questions? As for the first question, recall that other types of adjuncts (manner, frequency, etc.) do not behave like time-point or place-point adjuncts with respect to illi, as illustrated in (59).

  1. (59) miin  ʔibwagaaħa/ ʕaadatan  *(illi)  ħakat?

    who  rudely/usually  d-prt  spoke.3sg.f

    ‘Who rudely/usually spoke?’

The grammaticality of the question in (59) with illi suggests that time-point and place-point adjuncts have special properties that license them in [spec,SubjP], unlike other types of adjuncts. I assume that time-point and place-point adjuncts can occupy [spec,SubjP] because they contain some nominal category PLACE and TIME, respectively (Kayne Reference Kayne2005, Stanton Reference Stanton2016). Note here that I distinguish time-point adjuncts and place-point adjuncts from other types of temporal/locative adjuncts, because when other temporal/locative replace illi, the result is ungrammatical. As can be seen in (60), illi must be used in conjunction with the fronted durational temporal adjunct l-χamsat ʔayyaam ‘during/for five days’, and with the locative adjunct l-χamsat ʔimtaar ‘for five meters’.

  1. (60)

    1. a. miin  l-χamsat  ʔayyaam  *(illi)  ʔaχað-at  l-mafatiiħ?

      who  for-five  days  d-prt  took-3sg.f  def-keys

      ‘During/for five days, who took the keys?’

    2. b. miin  l-χamsat  ʔimtaar  *(illi)  ʃaal  l-kursi?

      who  for-five  meters  d-prt  carried-3sg.f  def-chair

      ‘For five meters, who carried the chair?’

What the examples in (60) show is that in order to replace illi, an adjunct must have a nominal category and be referential (i.e., it must refer to a specific point in the discourse).

As for the second question – why fronted time-point adjuncts replace illi in transitive and intransitive questions, while place-point adjuncts do so only in intransitive questions – I propose that time-point adjuncts can function as fillers of [spec,SubjP] because they are visible to Subj°. Subj° attracts the adjuncts that are located within its accessible domain to fill [spec,SubjP]. Since time-point adjuncts are, by definition, adjoined to TP, they are visible to Subj° regardless of the valency of the lexical verb. On the other hand, locative adjuncts are low, most probably adjoined to VP, which is contained within a lower phase, that is, v*P. The complement of v* undergoes transfer when C° enters the derivation (Felser Reference Felser2004). Locative adjuncts are thus inaccessible to Subj° because phase impenetrability prevents Subj° from attracting phrases located lower than [spec,v*P]. In contrast, in intransitive questions where there is no lower v*P phase, all adjuncts adjoined to VP are accessible to Subj°. In order to adopt this approach, however, we need to show evidence that first, there is no v*P phase in intransitive sentences in JA, and second, place-point adjuncts are base-generated in the lower v*P in transitive sentences.

The first of these can be supported by the behaviour of so-called floating quantifiers. As has been demonstrated cross-linguistically, floating quantifiers are diagnostics of movement (Sportiche Reference Sportiche1988, Speas and Yazzie Reference Speas, Yazzie, Jelinek, Golla, Midgette, Rice and Saxon1996, Costantini Reference Costantini2010). Consider the examples in (61), which include the quantifier kull in two different positions within the same question.

  1. (61)

    1. a. eeʃ  l-wadʒbih  illi  ʔakal-ha  l-walad  kull-ha

      what  def -meal  d-prt  ate.3sg.m-3sg.f  def-boy  all-3sg.f

      ‘Which meal did the boy eat all of?’

    2. b. eeʃ  l-wadʒbih  illi  ʔakal-ha  kull-ha  l-walad

      what  def-meal  d-prt  ate.3sg.m-3sg.f  all-3sg.f  def-boy

      ‘Which meal did the boy eat all of?’

In (61a) kull surfaces in the canonical position of the direct object, as a complement of VP. In (60b), it appears in the intermediate position where the direct object lands en route to the left periphery. This position is the outer [spec,v*P] which is positioned to left of the subject and to the right of verb in T0. The direct object must land in [spec,vP] because of the PIC, which prevents movement from the non-edge of a phase. Evidence that this position is the outer [spec,v*P] comes from questions with an overt auxiliary in T0, such as kaan ‘was’ in (62). When T0 contains kaan, the main verb surfaces to the right of kull, indicating that it has not raised to T0.

  1. (62) eeʃ  l-wadʒbih  illi  kaan  kull-ha  l-walad  yookil-ha

    what  def-meal  d-prt  was  all-3sg.f  def-boy  ate.3sg.m-it

    ‘Which meal did the boy eat all of it?

In contrast, in passive sentences, kull cannot appear between the auxiliary and the verb, as shown in (63).

  1. (63)

    1. a. eeʃ  l-wadʒbih  illi  kaant  titaakal  kull-ha

      what  def-meal  d-prt  was  ate.3sg.m.pass  all-3sg.f

      ‘Which meal was eaten all?’

    2. b. * eeʃ  l-wadʒbih  illi  kaant  kull-ha  titaakal

      what  def-meal  d-prt  was  all-3sg.f  ate.3sg.m.pass

      ‘Which meal was all eaten?’

The ungrammaticality of (63b) suggests that there is no syntactic position between T0 and the verb, filled by a copy of the direct object and available to host kull. I interpret this as evidence that there is no v*P phase in passive questions in JA.

The same pattern is found in questions with unaccusative verbs. Consider (64), where the quantifier ʃwai ‘some’ cannot appear between the auxiliary and the main verb, in what would be the inner phase edge if there were an inner phase.

  1. (64)

    1. a. eeʃ  l-gitʔaar  illi  kaan  yaʔħtaraq  ʃwai  minn-uh

      what  train  d-prt  was  burn.3sg.m.imp  some  from-it

      ‘Some of which train was burning?’

    2. b. * eeʃ  l-gitʔaar  illi  kaan  ʃwai  minn-uh  yaħtariq

      what  train  d-prt  was  some  from-it  burn.3sg.m.imp

      Some of which train was burning?

I conclude that there is no v*P in unaccusative or passive questions in JA.Footnote 19 As for the second question, some evidence that place-point adjuncts are base-generated in v*P in transitive sentences comes from the observation that place-point adjuncts must be construed as topicalized when they appear pre-verbally in transitive questions. I appeal here to evidence from double-complements constructions. In such constructions, the verb takes two internal arguments (normally a direct object and a PP), as in (65).

  1. (65) l-walad  ħatˤtˤ  l-muʁallaf  ʔa-tˤ-tˤaawlih

    def-boy  put.3sg.m  def-envelope  on-def-table

    ‘The boy put the envelope on the table.’

The place-point locative PP ʔatˤtˤaawlih ‘on the table’ in (65) is base-generated as a complement of the verb ħatˤeet ‘put’. If the locative PP ʔatˤtˤaawlih ‘on the table’ is deleted in (65), the resulting sentence is ungrammatical, given that the selectional requirement of the verb ħatˤeet ‘put’ for a locative internal argument is not met. If the locative PP ʔatˤtˤaawlih ‘on the table’ is fronted to the left periphery, illi must be used when the subject is questioned, as shown in (66).

  1. (66) miin  ʔ-atˤ-tˤaawlih  *(illi)  ħatˤtˤ  l-muʁallaf

    who  on-def-table  d-prt  put.3sg.m  def-envelope

    ‘On the table, who put the envelope?’

The fronted PP ʔatˤtˤaawlih ‘on the table’ is construed as a topic, attracted by the Topic Criterion. If the topicalized locative PP ʔatˤtˤaawlih ‘on the table’ satisfies the Subject Criterion, then the Topic Criterion will be violated, and vice versa. As such, fronted locatives in transitive questions have a discourse-bound reading. This contrasts with locative PPs in intransitive questions where the locative can intervene between the subject wh-word and the verb without forcing the locative to have a topicalization construal. The asymmetry between the behaviour of time-point and non-time-point adjuncts in questions with subject extraction is thus of key importance as a diagnostic of phases in JA.

Having investigated the main facts of subject extraction in JA, we now turn to cases where illi is used in questions with object extraction. This exploration is important, given that under the analysis developed in the previous sections, illi is not expected to appear in questions with object extraction, since the subject is still available to fill [spec,SubjP].

7. illi and object extractionFootnote 20

In the previous sections, I argued that illi is an XP element that occupies [spec,SubjP] and, hence, facilitates subject extraction. It is therefore not expected that illi should occur in questions with object extraction. That is because there is no Object Criterion and, most importantly, the subject is available to move to [spec,SubjP] to satisfy the Subject Criterion, eliminating any need for illi. However, JA data indicates that illi might appear in questions with object extraction. The sentence in (67) is an example of object extraction; notice that illi appears before the verb.

  1. (67) miin  illi  ʃaaf-ha  l-walad  imbaariħ

    who  d-prt  saw.3sg.m-3sg.f  def-boy  yesterday

    ‘Who did the boy see yesterday?’

This sentence might look like a counterargument to the analysis of illi as a D-linked element occupying [spec,SubjP]. However, on closer examination of sentences like these, the use of illi can be shown to follow from the proposed analysis of illi, and even to support it. Before showing how (67) is consistent with my analysis of illi, I first explore the status of the pronominal clitic (which marks the ϕ-content of the extracted object) on the verb. This clitic is required when illi is used. Questions with object extraction can be formed without illi only if there is no clitic, as illustrated in (68).

  1. (68) miin  (*illi)  l-walad  ʃaaf(*-ha)?

    who  d-prt  def-boy  saw.3sg.m-3sg.f

    ‘Who did the boy see?’

There seems to be a connection between the object clitic appearing on the verb and the use of illi. To make this connection explicit, I argue that the object clitic appears on the verb when the extracted object wh-word is D-linked. Evidence for this assumption comes from the observation that object questions with illi (and a clitic) must be used when the object wh-word is a D-linked constituent (which man) or modified (which tall one) which were argued in section 5 to be D-linked. This is illustrated in (69).

  1. (69)

    1. a. miin  l-binit  *(illi)  ʃaaf*(-ha)  l-walad  imbaariħ

      who  def-girl  d-prt  saw.3sg.m-3sg.f  def-boy  yesterday

      ‘Which girl did the boy see yesterday?’

    2. b. miin  tˤ- tˤawiilih  *(illi)  ʃaaf*(-ha)  l-walad  imbaariħ

      who  def-tall  d-prt  saw.3sg.m-3sg.f  def-boy  yesterday

      ‘Which tall one(f) did the boy see yesterday?’

Given the observation that D-linked object wh-words require a clitic on the verb, it can be postulated that the clitic is a copy of the D-linked object wh-word with reduced PF content. I propose that illi is required in such sentences because the subject is forced to remain in situ. This is because of the intervention of the object wh-word. Given the effects of the PIC, the object wh-word first moves to the outer [spec,v*P] and then moves to [spec,FocP]. Assuming the Copy Theory of movement (Chomsky Reference Chomsky1995), when the object moves to [spec,FocP] it leaves a copy in the outer [spec,v*P]. This copy blocks the movement of the subject to [spec,SubjP]. The object wh-word carries, among others, [wh] and [d-link] features, which I assume block the movement of the subject to [spec,SubjP]. If the subject is D-linked, the copy of the object in [spec,v*P] invokes a (feature-based) Relativized Minimality violation against the subject movement to [spec,SubjP] (Rizzi Reference Rizzi and Belletti2004), since the object wh-word is also D-linked. One might ask here why the D-linked subject does not block movement of the object wh-word to [spec,v*P]. I appeal here to Starke's (Reference Starke2001) proposal that the intervention effect induced by α can be overcome if the moved constituent has an additional feature (see Landau Reference Landau2008, Lahne Reference Lahne2008 for similar arguments). The extracted object has an additional [wh] feature that the subject lacks. As a result, the object wh-word overcomes any intervention effect caused by the subject.

On the other hand, if the subject is not D-linked, the copy of the object in [spec,vP] still invokes an intervention effect against the subject, since its featural make-up is richer than that of the non-D-linked subject (Haegeman Reference Haegeman2010, Belletti and Rizzi Reference Belletti, Rizzi, Piattelli-Palmarini and Berwick2013). One consequence of this analysis is that the subject must remain in situ. Since the verb moves to T0 in JA, the subject is expected to surface in postverbal position, as it does in (67) and (69). The structure in (70) (irrelevant details omitted) shows the intervention effect caused by the boxed object wh-word, blocking the subject from moving to [spec,SubjP].

  1. (70)

Since the subject cannot move to [spec,SubjP], that position is instead filled by the D-linked element illi to secure. Following this line of analysis, the generalization is that illi appears when the subject is unable to occupy [spec,SubjP].

On the other hand, when the object wh-word is not D-linked, the object no longer invokes an intervention effect against the subject, which is now able to move to [spec,SubjP], eliminating the need for illi. In such cases, as illustrated in (68), repeated here as (71), the subject surfaces to the left of the tensed verb, which has moved to T0.Footnote 21

  1. (71) miin  (*illi)  l-walad  ʃaaf(*-ha)?

    who  d-prt  def-boy  saw.3sg.m-3sg.f

    ‘Who did the boy see?’

The subject lwalad appears in a pre-verbal position which is, according to my analysis, [spec,SubjP]. Note that the use of illi makes the question ungrammatical.Footnote 22

Everything else being equal, the main strategies used in JA to satisfy the Subject Criterion can be reported in Table 2 (uninterpretable features of T0 = uT0).

Table 2: Strategies used in JA to satisfy the Subject Criterion

In view of the subject extraction facts in JA, it can be argued that JA makes use of certain skipping strategies in order to escape the Subject Criterion effects (Rizzi and Shlonsky Reference Rizzi, Shlonsky, Gärtner and Sauerland2007). What is special about JA is that these strategies are sensitive to the D-linking status of the subject wh-word. [spec,SubjP] can only filled by an element whose D-linked status is identical to that of the subject wh-word. When the subject wh-word is D-linked, [spec,SubjP] must be filled by a D-linked element, which can be the D-linked particle illi or a time-point or place-point adjunct whose inner structure contains a nominal TIME/PLACE category that makes it D-linked. On the other hand, if the subject wh-word is not D-linked, an expletive pro is used to escape the Subject Criterion effects.

8. Conclusion

I have shown that subject extraction in Jordanian Arabic is consistent with Criterial Freezing and the Subject Criterion. The main claim made here is that illi is used in JA to escape the Subject Criterion effects in cases where the subject wh-word is D-linked, under the D-linking condition on the Subject Criterion. Illi is base-generated in [spec,SubjP], ensuring convergence when the subject is extracted from its thematic position. Discourse particles, topicalized elements, direct objects, and the expletive fiih were all used to show that illi is in [spec,SubjP]. Additionally, I have argued that [spec,SubjP] in JA can be filled by a time-point or place-point adjunct instead of illi in transitive questions. I argue that these two types of adjuncts contain a nominal referential category, which qualifies them as D-linked elements. I also explored why time-point adjuncts, unlike place-point adjuncts, can replace illi in (in)transitive questions with a D-linked subject wh-word. The answer boils down to the Phase Impenetrability Condition, which prevents Subj° from attracting locative adjuncts from the lower, v*P phase. Furthermore, I proposed that in questions where the subject wh-word is not D-linked (in which verbs display the default form, i.e., [3SG.M]), [spec,SubjP] is filled by a non–D-linked expletive pro, in compliance with D-linking condition of the Subject Criterion.

Footnotes

An earlier version of this paper was presented at the 9th Days of Swiss Linguistics at Geneva University and at the 46th Poznań Linguistic Meeting at Adam Mickiewicz University, Poznań. I would like to thank the audience at these meetings for their useful input. In addition, I would like to thank my PhD supervisors Prof. Anders Holmberg and Dr. Geoffrey Poole for their invaluable comments on earlier draft of this paper. I am also indebted to two CJL reviewers for their criticism and helpful remarks which have led to significant improvements.

1 Existing only as a spoken variety of Arabic, Jordanian Arabic belongs genetically and typologically to the Semitic language family (cf. Al-Sarayreh Reference Al-Sarayreh2013: 11). It is spoken by the population of the Hashemite Kingdom of Jordan, with approximately seven million speakers.

2 1,2,3: Person; acc: Accusative Case; agr: Agreement; d-prt: Discourse Particle def: Definite; EPP: Extended Projection Principle; exp: Expletive; f: Feminine; imp: Imperfective; m: Masculine; nmlz: Nominalizer; pass: Passive; PIC: Phase Impenetrability Condition; pl: Plural; q: Question marker; sg: Singular; top: Topic.

3 Unpronounced lower occurrences are shown in angle brackets.

4 The Subject Criterion helps to explain two observations: that subjects are harder to move than objects – there is no Object Criterion – and that the subject position is an obligatory component of the form of clauses (Diercks Reference Diercks2010, Polinsky et al. Reference Polinsky, Gallo, Graff, Kravtchenko, Morgan, Sturgeon, Sprouse and Hornstein2013).

5 JA examples in this paper depended on the researcher's intuition and were verified by 25 JA speakers. Note that illi is glossed as ‘illi’ up to the point it is made clear that it is a D-linked particle.

6 Some Arabic dialects have different phonological variants of illi. For instance, halli and yalli are variants of illi in Syrian Arabic, whereas the phonetic alternants of illi in Moroccan Arabic are bəlli and lli (cf. Galal Reference Galal2004: 46).

7 In JA, illi can be used in question with object extraction, an observation I return to later in section 7.

8 The glosses in (22) are Gad's.

9 Benmamoun (Reference Benmamoun2000) argues that the main verb raises to T° in the past and future tenses in the Arabic sentence, given the specification of these two tenses with [+D] and [+V] features. On the other hand, the main verb remains in v° in the present tense, which is specified only for [+D]. For Benmamoun, movement of the verb to T° depends on whether tense bears a [V] feature. However, see Al-Balushi (Reference Al-Balushi2012) for refutation of this assumption.

10 Commenting on the conceptual importance of feature inheritance, Richards (Reference Richards2007: 569) claims that feature inheritance is the only way of ensuring that C°'s uninterpretable features can indeed be valued at the same time that it is transferred. He adds that since C° will not be spelled out until the following phase level, its uninterpretable features must descend onto its complement, namely T°.

11 It should be noted that JA does not exhibit SVO/VSO agreement interactions like those in Standard Arabic. The verb fully agrees with the subject, whichever word order is used.

12 When the subject is intended to be non-human, the subject wh-word used is either ʃuu or eeʃ which are interchangeably used with all inflected forms of the verb.

13 The discourse particle ʃikil indicates that the speaker is not certain of the truth of the propositional content of his/her utterance, though he/she has indirect evidence for it. I translate it as apparently (as opposed to surely).

14 Note that examples (40–42) imply that illi not a head, since it alternates with the expletive fiih, which clearly occupies the subject position,

15 Henceforth, I gloss illi as d-prt, a shorthand for discourse particle.

16 I set aside the question of whether D-linking condition on the Subject Criterion is a language-specific lexical property.

17 An anonymous reviewer asks why the locative can escape the inner phase boundary and then move to a higher topic position in (55), but cannot fill [spec,SubjP] in transitive clauses. I think that answering this question will require viewing Aʹ-chain formation as a consequence of agreement plus attraction. For instance, why does the Topic criterion attract the topicalized element rather than some other element in the sentence? One possibility is that the attracted element has a feature that matches the attractor. Such a feature would be uninterpretable; thus the attracted element can move to a higher position if the feature is not deleted/valued in situ. This might be why locatives move to the edge of the v*P phase when they are topicalized, since their proposed uTOP feature cannot be valued/deleted in situ by a higher Topic probe, given the effects of PIC. The locative would then be attracted to the left periphery by the Topic Criterion. If they do not have such a feature, they have no reason to leave their position, and thus are inaccessible to probes from higher phases (recall that Subject Criterion does not require a specific feature on a par with topics/foci, etc.). If the Subject Criterion attracted the topicalized locative when the locative moves to [spec,v*P], the locative would be unable to move subsequently to [spec,TopP], given Criterial Freezing. The derivation would then crash, because the Topic Criterion is not satisfied. I leave this issue open pending further research.

18 Cross-linguistically, an increasing number of authors argue that [spec,SubjP] can be filled by adjuncts (Bobaljik Reference Bobaljik2002, Holmberg and Hróarsdóttir Reference Holmberg and Hróarsdóttir2004, Landau Reference Landau2007, among many others).

19 An anonymous reviewer notes that Legate (Reference Legate2003) has argued for a v*P phase in passive, unaccusative, or raising constructions. Legate (Reference Legate2003) drew on reconstruction effects, quantifier raising, and parasitic gaps to argue that unaccusative and passive VPs are phases. However, Legate later argues that the reconstruction data do not demonstrate the existence of passive, unaccusative, and raising vP phases (Legate Reference Legate and Gallego2012). As for quantifier raising, JA data suggests that there is no v*P phase in passive, unaccusative, and raising constructions. I set parasitic gaps aside, because such gaps are not allowed at all in JA, as shown in (i).

  1. i. eeʃ  l-ktaab  illi  l-walad  giriih  gabul-maa  yiʃtaari-*(h)

    which  def-book  d-prt  def-boy  read.3sg.m  before-that  bought.3sg.m- it

    Intended: ‘Which book did the boy read before he bought?’

20 It is beyond the scope of the current article to provide a detailed account of object extraction in JA. I discuss here only the cases where illi is used in object extraction. There are several other observations which need further research and which are set aside due to limitations of space.

21 One might wonder why the subject does not block the movement of the object to the left periphery when the latter is not D-linked. I suggest that the subject does not block object wh-word movement because they are different in featural make-up, so no relativized minimality violation arises, and neither the subject nor the object has a richer featural make-up. The subject has [d-link], whereas the object has [wh].

22 A reviewer asks where JA illi stands with regard to the much-discussed contrast of complementizers in Québec French and whether there is a link between the two (unrelated) languages in terms of subject extraction. Given the observation that in Québec French, que turns into qui when the subject is extracted, it appears that illi and qui are both skipping strategies that provide a way around criterial freezing effects. On the other hand, illi but not qui is restricted to D-linked questions.

References

Abe, Jun. 2015. The EPP and subject extraction. Lingua 159: 1–17.Google Scholar
Adger, David. 2007. Stress and phasal syntax. Linguistic Analysis 33(3/4): 238–259.Google Scholar
Al-Balushi, Rashid. 2012. Why verbless sentences in Standard Arabic are verbless. The Canadian Journal of Linguistics 57(1): 1–30.Google Scholar
Al-Momani, Islam. 2015. Overt and null subject pronouns in Jordanian Arabic. Advances in Language and Literary Studies 6(4): 1–10.Google Scholar
Al-Sarayreh, Atef. 2013. The licensing of negative sensitive items in Jordanian Arabic . Doctoral dissertation, University of Kansas.Google Scholar
Aoun, Joseph E., Benmamoun, Elabbas, and Choueiri, Lina. 2010. The syntax of Arabic. Cambridge: Cambridge University Press.Google Scholar
Arsenijević, Boban. 2009. Clausal complementation as relativisation. Lingua 119(1): 39–50.CrossRefGoogle Scholar
Belletti, Adriana. 2002. Aspects of the low IP area. In The structure of IP and CP. The cartography of syntactic structures Vol. 2 , ed. Rizzi, Luigi, 16–52. Oxford: Oxford University Press.Google Scholar
Belletti, Adriana, and Rizzi, Luigi. 2013. Ways of avoiding intervention: Some thoughts on the development of object relatives, passive and control. In Rich languages from poor inputs, ed. Piattelli-Palmarini, Massimo and Berwick, Robert C., 115–126. Oxford: Oxford University Press.Google Scholar
Benmamoun, Elabbas. 1998. Spec-head agreement and overt case in Arabic. In Specifiers: Minimalist approaches, ed. Adger, David, Pintzuk, Susan, Plunkett, Bernadette, and Tsoulas, George, 110–125. Oxford: Oxford University Press.Google Scholar
Benmamoun, Elabbas. 2000. The featural structure of functional categories: A comparative study of Arabic dialects. Oxford: Oxford University Press.Google Scholar
Biberauer, Theresa, Holmberg, Anders, Roberts, Ian, and Sheehan, Michelle. 2009. Parametric variation: Null subjects in minimalist theory. Cambridge: Cambridge University Press.Google Scholar
Bobaljik, Jonathan David. 2000. The rich agreement hypothesis in review. Ms., McGill University.Google Scholar
Bobaljik, Jonathan David. 2002. A-chains at the PF-interface: Copies and covert movement. Natural Language and Linguistic Theory 20(2): 197–267.CrossRefGoogle Scholar
Boeckx, Cedric, and Grohmann, Kleanthes. 2004. ‘SubMove: Towards a unified account of scrambling and D-linking. in Peripheries, ed. Adger, David, de Cat, Cécile, and Tsoulas, George, 241–257. Dordrecht: Kluwer.Google Scholar
Brustad, Kristen E. 2000. The syntax of spoken Arabic: A comparative study of Moroccan, Egyptian, Syrian, and Kuwaiti dialects. Georgetown: Georgetown University Press.Google Scholar
Campos, Héctor. 1997. On subject extraction and the anti-agreement effect in Romance. Linguistic Inquiry 28(1): 92–119.Google Scholar
Cheng, Lisa Lai-Shen. 1997. On the typology of wh-questions. New York: Garland Publishing.Google Scholar
Chomsky, Noam. 1995. The minimalist program. Cambridge, MA: MIT Press.Google Scholar
Chomsky, Noam. 2000. Minimalist inquiries: The framework. In Step by step: Essays on minimalist syntax in honor of Howard Lasnik, ed. Martin, Roger, Michaels, David, and Uriagereka, Juan, 89–155. Cambridge, MA: MIT Press.Google Scholar
Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A life in language, ed. Kenstowicz, Michael, 1–52. Cambridge, MA: MIT Press.Google Scholar
Chomsky, Noam. 2007. Approaching UG from below. In Interfaces + recursion = language? Chomsky's minimalism and the view from syntax-semantics, ed. Gärtner, Hans-Martin and Sauerland, Uli, 1–30. Berlin: Mouton de Gruyter.Google Scholar
Cinque, Guglielmo. 1999. Adverbs and the universal hierarchy of functional projections. Oxford: Oxford University Press.Google Scholar
Costantini, Francesco. 2010. On infinitives and floating quantification. Linguistic Inquiry 41(3): 487–496.Google Scholar
Demuth, Katherine. 1995. Questions, relatives, and minimal projection. Language Acquisition 4(1/2): 49–71.CrossRefGoogle Scholar
Diercks, Michael J. K. 2010. Agreement with subjects in Lubukusu . Doctoral dissertation, Georgetown University.Google Scholar
É Kiss, Katalin, ed. 1995. Discourse configurational languages. Oxford: Oxford University Press.Google Scholar
Fakih, Abdul-Hafeed Ali. 2014. Subject Wh-Movement in Najrani Arabic and Minimalism. International Journal of Linguistics 6(5): 89–108.CrossRefGoogle Scholar
Felser, Claudia. 2004. Wh-copying, phases, and successive cyclicity. Lingua 114(5): 543–574.CrossRefGoogle Scholar
Fernández, Ángel Luis. 2009. On the composite nature of subject islands: A phase-based approach. SKY Journal of Linguistics 22: 91–138.Google Scholar
Frey, Werner. 2004. Notes on the syntax and the pragmatics of German left-dislocation. In The syntax and semantics of the left periphery, ed. Lohnstein, Horst and Trissler, Susanne, 203–233. Berlin: Mouton de Gruyter.CrossRefGoogle Scholar
Gad, Rehab Farouk. 2011. A syntactic study of WH-Movement in Egyptian Arabic within the Minimalist Program . Doctoral dissertation, University of Leeds.Google Scholar
Galal, Mohamed. 2004. Relativizer illi in Arabic dialects. Kansas Working Papers in Linguistics 27: 44–65.Google Scholar
Gallego, Ángel J., and Uriagereka, Juan. 2007. Conditions on sub-extraction. In Coreference, Modality, and Focus, ed. Eguren, Luis and Fernández-Soriano, Olga, 45–70. Amsterdam: John Benjamins.Google Scholar
Haegeman, Liliane. 2010. The internal syntax of adverbial clauses. Lingua 120(3): 628–648.Google Scholar
Harbert, Wayne, and Bahloul, Maher. 2002. Postverbal subjects in Arabic and the theory of agreement. In Themes in Arabic and Hebrew syntax, ed. Shlonsky, Ur and Ouhalla, Jamal, 45–70. Dordrecht: Kluwer.CrossRefGoogle Scholar
Holes, Clive. 1990. Gulf Arabic. London: Routledge.Google Scholar
Holmberg, Anders. 2000. Scandinavian stylistic fronting: How any category can become an expletive. Linguistic Inquiry 31(3): 445–483.Google Scholar
Holmberg, Anders, and Hróarsdóttir, Thorbjörg. 2004. Agreement and movement in Icelandic raising constructions. Lingua 114(5): 651–673.CrossRefGoogle Scholar
Holmberg, Anders, and Roberts, Ian. 2013. The syntax–morphology relation. Lingua 130, 111–131.Google Scholar
Kayne, Richard S. 2005. Movement and silence. Oxford: Oxford University Press.CrossRefGoogle Scholar
Kenstowicz, Michael. 1989. The null subject parameter in modern Arabic dialects. In The null subject parameter, ed. Jaeggli, Osvaldo and Safir, Kenneth J., 263–275. Dordrecht: Kluwer.Google Scholar
Lahne, A. 2008. Local modelling of long distance agreement. Paper Presented at the 31st GLOW Conference, University of Newcastle, 26–28 March 2008.Google Scholar
Landau, Idan. 2007. EPP extensions. Linguistic Inquiry 38(3): 485–523.Google Scholar
Landau, Idan. 2008. Two routes to control: Evidence from case transmission in Russian. Paper Presented at the 31st GLOW Conference, University of Newcastle, 26–28 March 2008.Google Scholar
Legate, Julie Anne. 2003. Some interface properties of the phase. Linguistic Inquiry 34(3): 506–515.Google Scholar
Legate, Julie Anne. 2012. The size of phases. In Phases: Developing the framework, ed. Gallego, Ángel J., 233–250. Berlin: De Gruyter Mouton.Google Scholar
Lewis, Robert. 2013. Complementizer Agreement in Najdi Arabic . Doctoral dissertation, University of Kansas.Google Scholar
Mohammad, Mohammad. 1990. The problem of subject-verb agreement in Arabic: Towards a solution. In Perspectives on Arabic Linguistics, ed. Eid, Mushira and McCarthy, John, 95–127. Amsterdam: John Benjamins.Google Scholar
Osman, Mariam Hussein. 1990. The Syntax and logical form of wh-interrogatives in Cairene Egyptian Arabic . Doctoral dissertation, University of Washington.Google Scholar
Polinsky, Maria, Gallo, Carlos Gómez, Graff, Peter, Kravtchenko, Ekaterina, Morgan, Adam Milton, and Sturgeon, Anne. 2013. Subject islands are different. In Experimental syntax and island effects, ed. Sprouse, Jon and Hornstein, Norbert, 286–310. Cambridge: Cambridge University Press.Google Scholar
Potsdam, Eric. 2006. The cleft structure of Malagasy wh-questions. In Clause structure and adjuncts in Austronesian languages, ed. Gärtner, Hans-Martin, Law, Paul, and Sabel, Joachim, 195–232. Berlin: Mouton de Gruyter.Google Scholar
Richards, Marc. 2007. On feature inheritance: An argument from the Phase Impenetrability Condition. Linguistic Inquiry 38(3): 563–572.CrossRefGoogle Scholar
Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, MA: MIT Press.Google Scholar
Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed. Haegeman, Liliane, 281–337. Dordrecht: Kluwer.CrossRefGoogle Scholar
Rizzi, Luigi. 2004. Locality and the left periphery. In Structure and Beyond, ed. Belletti, Adriana, 223–251. Oxford: Oxford University Press.Google Scholar
Rizzi, Luigi. 2005. On some properties of subjects and topics. In Contributions to the XXX Incontro di Grammatica Generativa, ed. Brugè, Laura, Giusti, Giuliana, Munaro, Nicola, Schweikert, Walter, and Turano, Giuseppina, 203–224. Venice: Cafoscarina.Google Scholar
Rizzi, Luigi. 2006. On the form of chains: Criterial positions and ECP effects. In Wh-movement: Moving on, ed. Cheng, Lisa Lai-Shen and Corver, Norbert, 97–134. Cambridge, MA: MIT Press.Google Scholar
Rizzi, Luigi. 2010. On some properties of criterial freezing. In The complementizer phase: Subjects and operators, ed. Panagiotidis, E. Phoevos, 17–32. Oxford: Oxford University Press.Google Scholar
Rizzi, Luigi. 2014. Some consequences of Criterial Freezing. In Functional structure from top to toe: The cartography of syntactic structures 9, ed. Svenonius, Peter, 19–46. New York: Oxford University Press.Google Scholar
Rizzi, Luigi. 2015. Cartography, criteria, and labeling. In Beyond the functional sequence: The cartography of syntactic structures 10, ed. Shlonsky, Ur, 314–338. New York: Oxford University Press.CrossRefGoogle Scholar
Rizzi, Luigi, and Shlonsky, Ur. 2006. Satisfying the subject criterion by a non-subject: English locative inversion and Heavy NP shift. In Phases of interpretation, ed. Frascarelli, Mara, 341–361. Berlin: Walter de Gruyter.Google Scholar
Rizzi, Luigi, and Shlonsky, Ur. 2007. Strategies of subject extraction. In Interfaces + recursion = language? Chomsky's minimalism and the view from syntax-semantics, ed. Gärtner, Hans-Martin and Sauerland, Uli, 115–160. Berlin: Mouton de Gruyter.Google Scholar
Shlonsky, Ur. 1992. Resumptive pronouns as a last resort. Linguistic Inquiry 23(3): 443–468.Google Scholar
Shlonsky, Ur. 2002. Constituent questions in Palestinian Arabic. In Themes in Arabic and Hebrew syntax, ed. Ouhalla, Jamal and Shlonsky, Ur, 137–159. Dordrecht: Kluwer.Google Scholar
Shlonsky, Ur. 2014. Subject positions, subject extraction, EPP and the Subject Criterion. In Locality, ed. Aboh, Enoch Oladé, Guasti, Maria Teresa, and Roberts, Ian, 58–85. New York: Oxford University Press.Google Scholar
Soltan, Usama. 2006. Standard Arabic subject-verb agreement asymmetry revisited in an Agree-based minimalist syntax. In Agreement systems, ed. Boeckx, Cedric, 239–265. Amsterdam: John Benjamins.CrossRefGoogle Scholar
Soltan, Usama. 2010. On strategies of question-formation and the grammatical status of the Q-particle huwwa in Egyptian Arabic wh-questions. University of Pennsylvania Working Papers in Linguistics 17(1): 1–10.Google Scholar
Speas, Margaret, and Yazzie, Evangeline Parsons. 1996. Quantification and the position of noun phrases in Navajo. In Athabaskan language studies: Essays in honour of Robert W. Young, ed. Jelinek, Eloise, Golla, Victor, Midgette, Sally, Rice, Keren, and Saxon, Leslie, 35–80. Albuquerque: University of New Mexico Press.Google Scholar
Sportiche, Dominique. 1988. A theory of floating quantifiers and its corollaries for constituent structure. Linguistic inquiry 19(3): 425–449.Google Scholar
Stanton, Juliet. 2016. Wholesale late merger in Ā-movement: Evidence from preposition stranding. Linguistic Inquiry 47(1): 89–126.Google Scholar
Starke, Michal. 2001. Move dissolves into Merge: A theory of locality . Doctoral dissertation. University of Geneva.Google Scholar
Struckmeier, Volker. 2014. Ja doch wohl C? Modal particles in German as C-related elements. Studia Linguistica 68(1): 16–48.Google Scholar
Sullivan, Michael D. Sr. 2016. Relativization in Ojibwe . Doctoral dissertation, University of Minnesota.Google Scholar
Taraldsen, Knut Tarald. 1986. Som and the binding theory. In Topics in Scandinavian syntax, ed. Hellan, Lars and Christensen, Kirsti Koch, 149–184. Dordrecht: Reidel Google Scholar
de Vries, Mark. 2002. The syntax of relativization. Utrecht: LOT.Google Scholar
Wahba, Wafaa Aabdel-Faheem Batran. 1984. Wh-Constructions in Egyptian Arabic . Doctoral dissertation, University of Illinois, Urbana-Champaign.Google Scholar
Figure 0

Table 1: The use of illi in questions with subject extraction

Figure 1

Table 2: Strategies used in JA to satisfy the Subject Criterion