revisiting the persian ezafe construction: a roll-up movement analysis

Available online at www.sciencedirect.com

ScienceDirect

Revisiting the Persian Ezafe construction:A roll-up movement analysis

www.elsevier.com/locate/linguaLingua 150 (2014) 1--24

Arsalan Kahnemuyipour *

University of Toronto, Mississauga, Department of Language Studies, North Building, 3359 Mississauga Road North, Room 239,Mississauga, ON L5L 1C6, Canada

Received 4 September 2013; received in revised form 11 July 2014; accepted 16 July 2014Available online

Abstract

This paper explores the Persian Ezafe construction, a construction which has received significant attention in the syntactic literature inthe past few decades. Descriptively, Ezafe is an unstressed vowel -e (-ye after vowels) which appears between a noun and its modifier(N-eMod), and is repeated on subsequentmodifiers, if they are present, except the last one (N-eMod1-eMod2-eMod3). This paper takesa fresh look at the distribution of the Ezafe vowel, with a special emphasis on its correlation with the order of elements in the noun phrase.After the close connection between word order and the absence/presence of Ezafe is established, the paper considers alternative waysthis relation can be captured and argues for a roll-up movement account of this construction, which takes the base order of the nounphrase in Persian to be head final, with the surface order derived via phrasal movement to specifiers of intermediate functional projectionsin a roll-up fashion. In developing and arguing for this analysis, we will also have a closer look at several constructions in Persian, such asthe superlative and the colloquial definite marker, and will account for various word order restrictions in the Persian noun phrase.© 2014 Elsevier B.V. All rights reserved.

Keywords: Ezafe; Persian; Word order; Roll-up; Linker

1. Introduction

The Persian Ezafe construction has intrigued linguists and has been the subject of various analyses over the pastseveral decades. Descriptively, Ezafe is an unstressed vowel -e (-ye after vowels) which appears between a noun and itsmodifier (N-e Mod), and is repeated on subsequent modifiers, if they are present, except the last one (N-e Mod1-e Mod2-eMod3) (see section 2). Several authors have investigated the nature and function of the Ezafe vowel. Some have analyzedit as a case marker (Samiian, 1983, 1994; Larson and Yamakido, 2008), some as a vowel inserted at PF to identifyconstituenthood or to link it to its modifiers (Ghomeshi, 1997; Ghaniabadi, 2010), and yet others as a (phrasal) affix to markthe presence of a syntactic dependent (Samvelian, 2007, 2008).

This paper takes a fresh look at the distribution of the Ezafe vowel in Persian, with a special emphasis on its correlationwith the order of elements in the noun phrase. After the close connection between word order and the absence/presenceof Ezafe is established, the paper considers alternative ways this relation can be captured and argues for a roll-upmovement account of this construction. In doing so, the paper also deals with the head vs. phrasal status of the post-nominal modifiers, their structural relation to the head noun and accounts for various word order restrictions in thePersian noun phrase. The paper is organized as follows. Section 2 provides a brief overview of the distribution of Ezafe.In section 3, we will have a closer look at the relation between word order and the presence of the Ezafe in three domains

* Tel.: +1 905 828 5497.E-mail address: [email protected].

http://dx.doi.org/10.1016/j.lingua.2014.07.0120024-3841/© 2014 Elsevier B.V. All rights reserved.

http://crossmark.crossref.org/dialog/?doi=10.1016/j.lingua.2014.07.012&domain=pdf

http://crossmark.crossref.org/dialog/?doi=10.1016/j.lingua.2014.07.012&domain=pdf

http://www.sciencedirect.com/science/journal/00243841

http://dx.doi.org/10.1016/j.lingua.2014.07.012

mailto:[email protected]

http://dx.doi.org/10.1016/j.lingua.2014.07.012

A. Kahnemuyipour / Lingua 150 (2014) 1--242

of Persian (morpho)syntax: the superlative, adjectives in the formal/literary register and compound formation. In section 4,we explore ways of capturing the distribution of the Ezafe marker and its correlation with word order both with reference toPersianandcross-linguistically. In this context, the proposal byLarsonandYamakido (2008),which takesEzafe tobea casemarker, is discussed at some length and several arguments against it are provided. Instead, a roll-up movement analysis ofthe Persian Ezafe construction is developed. This analysis takes the base order of the noun phrase in Persian to be headfinal, with the surfaceorder derived via phrasalmovement in a roll-up fashion.According to this analysis, theEzafe is seenasa reflex of this phrasal roll-upmovement. As such, the assumption that all bare or phrasal post-nominalmodifiers in theEzafeconstructionaremergedas (part of) phrases in thespecifiersof functional projectionabove thenoun is crucial to thisanalysis.In this light, theproposalbyGhaniabadi (2010),whomakesastrict distinctionbetweenbareadjectivesandphrasalmodifiers,treating the former as heads adjoined to N and the latter as phrasal adjunctions to the NP, is considered at some length andseveral empirical arguments are provided against it. Section 5 discusses two cases where the Ezafe construction issomewhat surprisinglymissing, namely the superlative (and ordinal) constructionsandbare post-nominal adjectives used inthe context of the colloquial definitemarker. A syntactic account is suggested for each case. Section 6 concludes the paper.

2. The distribution of Persian Ezafe

In this section, we will look at the distribution of the Ezafe vowel, considering both contexts where Ezafe is present andwhere it is absent. Ezafe is present with post-nominal modifiers as shown in (1). (1a) shows an example of a noun followedby an attributive noun.1 (1b) is an example of a noun followed by an attributive adjective. With two attributive adjectives,both the noun and the first adjective will be marked with an Ezafe, as shown in (1c). The Ezafe vowel also appears on anoun followed by a possessive noun (or pronoun), as shown in (1d). In a prepositional phrase, some prepositions aremarked by the Ezafe vowel, as exemplified in (1e). When a pronoun is modified by an adjective, the pronoun will take theEzafe marker, as in (1f). The Ezafe even appears on a first name followed by a family name, as shown in (1g).2 Finally,more than one post-nominal modifier can be used, in which case the Ezafe vowel appears on the head noun and all themodifiers except the last one. This is already seen in (1c) with two attributive adjectives. Another example with twoadjectives and a possessor is given in (1h).

(1)

1 Standefinite.

2 Whe3 The

example

a. mm‘th

4 NoteEzafe m

Presence of Ezafe with post-nominal modifiers
a.
dard PTheren speasameacontra

ard-ean-Eze lonethat tharker

(ye)

ersianis a mking adjectivsting w

tanhaonlyly mane formis relev

kif-e

does not harker of debout a verye can beuseith (2e) is g

a b.

’in (2a) is gant here.

charm

ave a definite article. Wfiniteness, used in collowell known person, thdpost-nominally and tak

iven in (ia) below. Both th

tanhaa mard-e taonly man-Ez on‘the only lonely man

rammatical with -e if -e

b.

ithoutquial Pe Ezafeeon the pre-n

nhaaly’is taken

(ye)

the indersianis oft

emeanominal

to be

mard-e

efinite artic, which is den droppeding ‘‘lonely’’and post-no

the homoph

chaaq
a bag-Ez leather a man-Ez fat
‘a/the leather bag’
‘a/the fat man’ c. sag-e qahveyi-ye gonde d. ketaab-e Ali/man
le, thiscus: e.g.. Thenminal

onou

dog-Ez
brown-Ez big book-Ez Ali/I ‘big brown dog’ ‘Ali’s/my book’
e.
posht-e dar f. man-e divune behind-Ez door I-Ez crazy ‘behind the door’ ‘crazy me’
g.
Ali-ye Mohammadi Ali-Ez Mohammadi
h.
gorbe-ye iraani-ye sefid-e Maryam cat-Ez Persian-Ez white-Ez Maryam ‘Maryam’s white Persian cat’
Crucially, Ezafe only appears with post-nominal modifiers. Ezafe never appears on a bare noun (2a) or on pre-nominalelements such as numerals (2b), demonstratives (2c), quantifiers such as each/no (2d), the adjective ‘‘only’’ (2e)3 or anycombination of these elements (2f).4

e unmarked interpretation of a modified noun phrase issed in section 5.2.Ali Daayi (famous Iranian soccer player).ounwill have tobemarkedwithanEzafe, asexpected. Anadjectives can appear on the samenoun, as shown in (ib).

s definite colloquial marker (see section 5.2). Only -e as

(2) Absence of Ezafe with bare nouns or pre-nominal modifiers

A. Kahnemuyipour / Lingua 150 (2014) 1--24 3

5 Tha

a.

nks to

ketaab-(*e)

an anonymous reviewer for
b
b.

ringing

do

this e

(taa) -- (*e)

xample to my a

ketaab
book-Ez two classif.-Ez book
‘two books’
c. in-(*e) ketaab d. har/hich-(*e) ketaab-(i)
ttention.

this-Ez
book each/no-Ez book-indef. e. tanhaa-(*e) mard f. in-(*e) do-(*e) ketaab
only-Ez
man this-Ez two-Ez book ‘the only man’ ‘these two books’
When there is a combination of pre-nominal and post-nominal modifiers, the contrast with respect to the appearance ofEzafe is evident. As shown in (3), Ezafe appears in the post-nominal domain and is absent pre-nominally.

(3)
Combination of pre-nominal and post-nominal modifiers a. in do ketaab-e qatur-e jaaleb
this
two book-Ez thick-Ez interesting ‘these two interesting thick books’
b.
in tanhaa so’aal-e bi-ma’ni this only question without-meaning ‘this only meaningless question’
c.
in ketaab-haa-ye baa-arzesh this book-pl.-Ez with-value ‘these valuable books’
d.
har se ketaab-e aabi every three book-Ez blue ‘all three blue books’
To sum up, the above facts establish a clear correlation between the presence of Ezafe and the order of nominalelements. The noun marks a clear boundary for the Ezafe marker: all elements preceding it lack the Ezafe, while the nounitself and all elements following it (except the final one) are marked with the Ezafe marker. In the next section, we explorethis correlation in three other domains in the (morpho)syntax of Persian.

3. Ezafe and word order: further evidence

In this section, we explore the correlation between word order and the presence of Ezafe in the domains of thesuperlative, the formal/literary register and compound formation in Persian. The dominant pattern in all these casesconfirms the pattern we found in section 2, i.e. when the modifier comes before the (head) noun, no Ezafe vowel is used.

3.1. The superlative

Unlike the simple and comparative forms of the adjective, which are post-nominal and are preceded by the Ezafemarker, the superlative form is always used pre-nominally without the Ezafe vowel. This construction is fully productive.Some examples are given in (4), where the simple/comparative forms are also provided to illustrate the contrast.

(4)
a. bad-tarin ketaab vs. ketaab-e bad(-tar) bad-super. book book-Ez bad-comp. ‘the worst book’ ‘bad/worse book’
b.
qashang-tarin maashin vs. maashin-e qashang(-tar) nice-super. car car-Ez nice(-comp.) ‘the nicest car’ ‘nice(r) car’
It should be noted that one can come across examples in Persian which appear to involve the superlative adjectivefollowed by theEzafemarker anda plural noun. This construction exemplified in (5)may be seen as a counterexample to theabove generalization that the superlative form of the adjective is always pre-nominal and never takes the Ezafe vowel.5

(5) qashang-tarin-e doxtar-*(haa)


6 Intethe girls

7 Natiman" w

8 It apsuggessimilar r

9 UseRahnavAnother10 Tak11 Tak

restingly, unlike ‘some’, th’.ve speakers have the intuithout using the Ezafe vowpears that these pre-nomts some type of compounestrictions discussed abd in reference to Zahra Rard is currently under houcandidate of the same

en from Nima Youshij’s Aen from the Toronto-bas

beautiful-super.-Ez
girl-pl. ‘the most beautiful of the girls’
It is worth noting, however, that in this case the superlative adjective is not modifying the head noun and is instead used asa noun itself in a partitive construction, reflected in the English translation given in (5) (see also Ghaniabadi, 2010). Thenoun phrase in (5), therefore, can be interpreted as either singular or plural, depending on whether the set of beautiful girlsout of the larger set of girls has one or moremembers. If we drop the Ezafe in (5), we would be dealing with a true case of asuperlative adjective modifying the plural noun and only the plural interpretation would be available, as expected. Notethat Ezafe is only one of the two forms used in Persian to mark partitivity. The other one is the preposition az ‘of, from’. In(5) above, the use of az ‘of’ in place of the Ezafemarker would lead to ungrammaticality. Meanwhile, there are other caseswhere the opposite is true, with az being the only possibility: ba’zi az/*e doxtar-haa some of/Ez girl-pl ‘some of the girls’.6 Insome contexts, either Ezafe or az ‘of’ seem fine: do taa-ye/az shomaa two classif.- Ez/of you.pl ‘two of you’. I amabstracting away from this partitive use of the Ezafe in this paper and leave the question of whether it may be subject to thesame type of analysis proposed in this paper for the Ezafe construction for future research.

I return to a more detailed discussion of the superlative construction exemplified in (4) in section 5.1.

3.2. Evidence from the formal/literary register

In the formal (and particularly the literary) register, adjectives can be used pre-nominally without the Ezafe vowel ratherproductively. These formal uses can be readily found inmedia interviews, academic lectures, newspaper articles or formalcorrespondence. Still, it would be somewhat unusual to havemore than a few cases in a single discourse. There are somerestrictions on these pre-nominal adjectives. Typically, nomore than one adjective is used and the adjective itself does nottake a complement or is not even modified (e.g. by an intensifier). These restrictions may be violated in literary use, butwould be seen as marked even in such contexts. I leave a closer examination of such restrictions and a possibleexplanation for them for future research. A few examples of these prenominal adjectives are provided in (6) below. Toillustrate the contrast more clearly, the more colloquial forms with post-nominal modifiers, where Ezafe is needed, are alsoprovided for (6a) and (6b).7,8

(6)
a. bichaare xalq
e quantifier

ition that withel answereinal adjecti

d formation.out compouahnavard, ase arrest alopresidentialLetter to a

ed Persian w

vs.

‘all’ is

out thd: xubves aIf so,ndingpromng wielectioPrisoeekly

xalq-e

only good with

e Ezafe vowel(good) mard (ct as a unit withe restrictionsin section 3.3.inent figure ofth her husbandn, Mehdi Karrner. TranslatioShahrvand, A

bichaare

Ezafe, and not w

, the adjective haman). Note that xth the following hon the number a

the 2009 post-eleMir Hossein Mououbi, has also ben is my own.pril 2012, issue

ith az

s to goubmaread nnd for

ctionsavi, hen un

no. 13

‘of, fro

befored is nooun whm of th

uprisinimselfder ho

80, pa

poor/pitiable
people people-Ez poor/pitiable ‘poor people’
b.
sabz baanu9 vs. baanu-ye sabz green lady lady-Ez green ‘the green lady’
c.
shab o ruz-e to dar aan tang hesaar, va shab o
m’: ha

the nt useden fue adj

g in Ira preuse a

ge 4.

ruz-e

me-ye/*az

oun. A Peras a lexicrther modiective may

an, knownsidential carrest.

man
night and day-Ez you in that tight fence and night and day-Ez I andar del-e in baaz hesaar-i ke be zaaher na chenaan zendaan
doxtar-haa a

sian speakeralized compofied, see foralso be view

as the Greenndidate of th

ast10

ll-Ez/of

asked tound in Pexampleed in the

Moveme dispute

in
heart-Ez this open fence-rel. that to appearance not such prison is ‘Your day and night in that tight cell, and my day and night in the heart of this open cell, which, in appearance,is not such a prison.’
d.
. . . bi-shomaar iraani-haa-yi ke dar saraasar-e jahaan paxsh shode-and11
girl-pl

say "ersia(6e).cont

ent. Zd ele

‘all of

goodn.This

ext of

ahraction.

without-number
Iranian-pl.-rel. that in whole-Ez world spread become.past.part.-3pl. ‘the innumerable Iranians who have spread all around the world’


12 The13 In awould l14 The

e.

Persiall theseose theform w

bozorg

n title of thecases, oneidiomatic mith Ezafe is

mard-e

1970 Arthcould formeaning.typically u

kuchak12

i

ur Penn mova syntactic n

sed in contra

ie staroun ph

st to s

ring Drase,

ib-e de

ustin Hsimilar

raxti a

big
man-Ez little ‘little big man’
3.3. Evidence from compound formation

In Persian compounds consisting of a modifier and a head N, the dominant pattern is one where the adjective(or modifying noun) comes before the noun, and no Ezafe is used. Some examples are shown in (7) (some of these aredue to Ghomeshi, 1996 and Lazard, 1992)13:

(7)
a. gol-aab b. bozorg-mard flower-water big-man ‘rose-water’ ‘great man’
c.
ketaab-xune d. kaar-xune book-house work-house ‘library’ ‘factory’
e.
siyaah-cheshm f. bad-zaat black-eye bad-nature ‘having black eyes’ ‘mean person’
Persian compounds are dominantly right-headed, as seen in (7). Another very productive construction which further
underlines this property is the agentive compound. Note that simplex agentive nouns are typically formed by the additionof the agentive marker -ande to the present stem of the verb (e.g. forush-ande sell-er ‘seller’). These nouns can be usedwith post-nominal modifiers in a syntactic phrase where the Ezafe vowel is necessary. There is, however, a compoundformation process in which the agentive noun loses the -ande marking and is combined with a modifying element.Crucially, these compounds are head-final and the Ezafe is absent. The contrast between the syntactically formedphrases and the morphological compounds is shown in (8).
(8)
a. forushande-ye (in) ketaab vs. ketaab-forush seller-Ez this/t he book book-sell(er) ‘the seller of this book’ ‘book-seller’
b.
negaarande-ye (in ) maqaale vs. ruznaame-negaar writer-Ez th s/the article newspaper-write(r) ‘writer of this article’ ‘journalist’
While the head-final pattern illustrated in (7)--(8) is more dominant, there are some ‘compounds’ with post-nominal
adjectives (or modifying nouns) without the Ezafe marker. Some such examples and their counterparts using Ezafe aregiven in (9) (from Ghomeshi, 1996; Lazard, 1992):
(9)
a. maadar-bozorg vs. maadar-e bozorg mother-big/grand mother-Ez big/grand ‘grandmother’ ‘big/grand mother’
offman.to the ones i

pple-Ez tree

b.
aab-porteqaal vs. aab-e sib water-orange water-Ez apple ‘orange juice’ ‘apple juice’
c.
pesar-amu ? pesar-e amu son-uncle son-Ez uncle ‘cousin’ ‘uncle’s son’
d.
sib-zamini vs. ? sib-e zamini14
apple-ground
apple-Ez ground ‘potato’
The compounds in (9)haveseveral properties thatdistinguish themfrom theones in (7). Forone, asstatedpreviously, thehead-final formation in (7) is much more productive than the apparently head-initial forms in (9). Moreover, the forms in (9)

n section 2, as N-Ez A. Meanwhile, these syntactic phrases

(lit. tree apple) ‘apple’.


typically occur with very frequent forms only. This seems to indicate that most (if not all) of these forms originate in syntacticphraseswith theEzafe (N-EzMod.)whichhave lost their Ezafe vowel over timedue to frequent use.Theexamples in (9b) areespecially revealing, given that traditionally ‘orange juice’wasmuchmorepopular than ‘apple juice’.15Finally, thehead-initialforms exemplified in (9) typically have a more transparent meaning than the head-final ones. As such, while the head-initialforms can be replaced with the productive and syntactically formed compounds with the Ezafe (i.e. N-Ez Mod) without anychange inmeaning (e.g.pesar-amuvs.pesar-eamu tomean ‘cousin’), turningahead-final form intoasyntactic compoundbyreversing the order of noun and modifier and adding the Ezafe will likely result in the loss of ‘special’ or ‘idiomatic’meaning(e.g. ketaab-xune book-house ‘library’ vs.??xune-ye ketaab ‘house of books’).16

Given these differences, I take the head-initial compounds in (9) to have a different status from the head-final onesexemplified in (7)--(8). I suggest that the left-headed compounds in (9) are ‘historical’ compounds, i.e. they originated as asyntactic phrase in the Ezafe construction, but have lost the Ezafe vowel over time due to frequency of use. Thesyntactically formed and hence semantically more compositional source of these compounds, leads to the moretransparent meaning noted above. The productive right-headed compounds, on the other hand, are ‘true’ compounds,which reflect the base order of modifiers and nouns in Persian. This compound formation can be seen as the result ofsome kind of morphological merger. This morphological merger occurs between two heads, typically A and N, which willbe involved in the rest of the syntactic derivation as a single unit [A N]. Given that this compound formation involves themerger of two heads, phrasal modifiers are barred from this construction. I will show below how the Ezafe construction canbe derived from the base head-final order. The crucial point I am making here is that the productive head-final compoundformation of Persian should be taken as an indication for the base order of nouns and modifiers in Persian.17 This view isperhaps best understood in a framework that takes all word formation to occur in the syntactic component by rules ofsyntax -- ‘‘syntax all the way down’’, as held, for example, by the proponents of DistributedMorphology (Halle andMarantz,1993; Marantz, 1997, and subsequent authors).18

Before we end this section, it is worth noting that Ghaniabadi (2010) treats the use of the post-nominal adjective(without the Ezafe marker) in the context of the colloquial definite marker, illustrated in (10) below, as a case ofcompounding on a par with the compounds discussed in this section. I leave a discussion of such cases to section 5.2,where I argue that they should not be treated as compounds and will provide an alternative analysis for them.19

(10)

15 In faccan be enjuice-Ez16 The fohard to k‘grandmo17 Note18 I am aa Distribu19 Than20 For awhich ma

ketaab

*

t, with the mcounteredlettuce ‘lettrm in (9a)now whethther’ couldthat, in a lbstractingted Morphks to an andiscussiony be seen

gerun-e
book expensive-def. ‘the expensive book’
In this section, we have looked at three phenomena in Persian syntax and morphology which further establish thestrong correlation between the order of the modifier and the noun and the presence of the Ezafe vowel in Persian. Thiscorrelation is schematized in (11). On the one hand, pre-nominal modifiers do not take the Ezafe marker (11b) (vs. (11d)).This was seen in the previous section with pre-nominal elements such as the numeral and the demonstrative. In thissection, we saw the same pattern with the superlative as well as the productive formation of compounds and the(somewhat restricted) use of pre-nominal adjectives in the formal/literary register. Crucially, the Ezafe never shows upwith pre-nominal modifiers in Persian (11d). On the other hand, attributive nouns and adjectives most productively followthe noun with the use of the Ezafe (11a). With post-nominal modifiers, the use of Ezafe is almost obligatory, with theexception of what I have termed ‘historical’ compounds in this section (11c).

(11)
a. N-Ez
ore rec. Still, ifuce juicmight beer the fobe obta

anguageaway froology peonymouand anaas a co

Mod

ent popyou take’).an excrm withined mlike Enm a fulrspects revielysis ofunter-e

Most common: Fully productive syntactically
b. Mod N Only option with superlative, productive in compounds & formal register c. N Mod Limited to ‘historical’ compounds20
d.
Mod-Ez N NEVER!
ularity of other types of juices, e.g. apple juice among many others, more head-initial forms without the Ezafee a fruit/vegetable, whose juice is not typically consumed, the only possible form is N-Ez N (e.g. aab-e kaahu

eption to this. In present day Persian, the form with the Ezafe will not convey the meaning ‘grandmother’. It isout the Ezafe is historically related to the one with the Ezafe and whether at an earlier stage the meaningore easily from the form with Ezafe.glish, where modifiers are consistently pre-nominal, compounds, too, are head-final.l analysis of compound formation, as it is tangential to the main focus of this paper, the Ezafe construction. Forive, see Harley (2008).wer for pointing out Ghaniabadi’s (2010) approach to this construction and its relevance for my analysis.a bare adjective following the noun without the Ezafe in the context of the colloquial definite marker (see (10)),xample to (11c), see section 5.2.


The picture in (11) establishes a systematic relation between the order of the noun and its modifiers and the presenceof the Ezafe vowel in Persian. There is a clear asymmetry in this respect between pre-nominal and post-nominal modifierswhich requires an explanation. If we extend this relation between the presence or absence of the Ezafe and the order ofthe noun and its modifiers to a more cross-linguistic perspective, we should expect the Ezafe to occur only in languageswith a head-initial noun phrase and not a head-final one (in accordance with (11d)). This expectation may very well beborne out (see Samvelian, 2008:342). All the West Iranian languages with Ezafe such as Persian and Kurdish dialects(e.g. Hawrami and Zazaki) have a head-initial noun phrase. Ezafe is not found in languages such as Germanic (e.g.English or German) or Romance (e.g. French or Italian) with head-final noun phrases.21 In the following section, I discusstwo ways of capturing the correlation between word order and Ezafe established so far and argue in favor of one whichtakes the base structure of the noun phrase to be head-final and derives the surface order via roll-up movement.

4. Toward an explanation of the Ezafe-word order correlation

In the previous two sections, we established a correlation between the order of elements within the noun phrase andthe presence/absence of Ezafe. This correlation was shown to be (nearly) perfect in Persian -- Ezafe appears in the post-nominal domain and is absent in the pre-nominal domain. From a cross-linguistic perspective too, the presence of theEzafe seems to be connected to the head-initial syntax of the noun phrase. Any account of the Persian Ezafe constructionshould provide an explanation for this striking correlation, an issue which plays a central role in the analysis developed inthis paper. This section deals with the question of how this correlation between word order and the presence of Ezafe maybe captured.

Word order variations are commonly captured in the generative tradition via syntactic movement (displacement).22

There are twoways one can approach the particular problem at hand, i.e. the variation with respect to the order of the nounand its modifiers. One is to take the post-nominal (Persian) order as basic and derive the pre-nominal (English) order(Larson and Yamakido, 2008). The other possibility is to take the pre-nominal (English) order as basic and derive the post-nominal (Persian) order (this paper). The other aspect of the correlation noted so far has to do with the presence of theEzafe marker. If we capture the word order variation using movement, a question arises as to how we can relate themovement with overt morphology in light of the two possibilities just discussed. For the option that takes the post-nominal(Persian) order to be basic, there is complementarity between movement and overt morphology, i.e. in the post-nominal(Persian) order where the Ezafe is present, no movement takes place; when there is no overt morphology (Ezafe),movement derives the pre-nominal order of modifiers (English). For the second option which takes the pre-nominal(English) order to be basic, movement and overt morphology go hand in hand, i.e. when there is no movement as is thecase with pre-nominal (English) order, there is also no overt morphology (Ezafe); the movement that derives the post-nominal (Persian) order is accompanied with overt morphology (Ezafe).

Both ideas with respect to the relation between syntactic movement and the presence of overt morphology haveprecedence in the syntactic literature. The complementarity of movement and (overt) morphology is perhaps mostprofoundly alluded to in Cheng (1997) account of the typology of wh-questions. According to Cheng, languages eitherhave a wh-particle in C or move a wh-phrase to SpecCP. The first option is what we find with wh-in situ languages such asChinese, and the second option occurs with wh-movement languages such as English.23 This correlation betweenabsence of morphology and movement is also used in the realm of case. In particular, Larson and Yamakido (2008)develop an analysis of Ezafe in Persian and its absence in English which is framed in this complementarity betweenmovement and the presence of (overt) morphology. Under this view, in Persian, the modifiers are case-marked by theEzafe marker. In English, where Ezafe is not available, themodifier has to move to get case-licensed.24Wewill look at thisanalysis of the Persian Ezafe constructionmore closely in section 4.1, where several empirical arguments against it will beprovided.

21 The cross-linguistic claim about the presence or absence of the Ezafe certainly requires further investigation. Note, however, that the claim isonly about the absence of Ezafe in languages with head-final noun phrases. Thus, the existence of languages with head-initial noun phrases andno overt realization of something akin to the Ezafe marker cannot be taken as evidence against this generalization. Of course, in Romancelanguages, where some modifiers can appear post-nominally, no ‘Ezafe’ is used. Also, Samvelian (2008) points to the optionality or absence ofEzafe in some (North) Western Iranian languages or Southern Kurdish dialects. The correspondence between syntactic relations and overtmorphological realization is never perfect, as known in other domains such as case.22 There is a large body of work, particularly inspired by Kayne (1994), attempting to account for this type of word order asymmetries by positing auniversal base order and various (restrictions on) movement patterns (see, e.g. Cinque, 1996, 1999, 2002, 2005, 2010, Belleti, 2004, Rizzi, 2003).23 It is worth noting that Cheng’s generalization does not appear to be borne out for Persian, a language with no wh-particle and yet no wh-movement to SpecCP (see Kahnemuyipour, 2001).24 Larson and Yamakido (2008) do not discuss the cases of Persian pre-nominal adjectives that were considered in section 3. If adjectives arebase-generated post-nominally, it is not clear how they end up in the pre-nominal position in these cases. Given that some of these occur at thelevel of word formation, it may be difficult to attribute their alternative order to case.


The view which makes a direct connection between overt morphology and overt movement has been alluded to in therealm of agreement, where it is suggested that richness of morphological agreement depends on whether or not overtmovement has been involved in the establishment of a particular syntactic relation (see e.g. Kayne (1994) with respect toHungarian adpositions, Aoun et al. (1994) with respect to Arabic subject-verb agreement, etc.) This is the approach I amdeveloping in this paper. Under this view, in English there is no overt movement in establishing the relation between themodifier and the noun, and as such there is no overt morphology, while in Persian, the overt morphology (Ezafe) isrealized as a reflex of the overt establishment of this relation via phrasal movement.25 Likewise, when there is nomovement in Persian (e.g. pre-nominal adjectives), there is no Ezafe. The analysis developed in section 4.2 is situated inthis general approach to the correlation between overt morphology and overt movement.26

The rest of this section is organized as follows. In section 4.1, we will look at Larson and Yamakido’s (2008) analysis ofthe Persian Ezafe construction which takes the base order of the Persian noun phrase to be head-initial. We have alreadyseen some suggestive evidence in section 3 which seems to favor a head-final base order for the Persian noun phrase.Meanwhile, in section 4.1, further arguments against Larson and Yamakido’s analysis will be provided. This paves theway for a roll-up movement analysis of the Ezafe construction. Meanwhile, a crucial aspect of the analysis developed inthis paper is that the modifiers, whether bare or phrasal, are (part of) XPs located in the specifiers of functional projectionsabove the noun. We therefore need to establish the phrasal status of the modifiers involved in the Ezafe construction. Wealso need to show that a strict division between bare adjectives and phrasal modifiers with respect to their head/phrasalstatus and their structural relation to the head, a la Ghaniabadi (2010), is unwarranted. This is the topic of section 4.2. Insection 4.3, a roll-up movement analysis of the Ezafe construction is developed, where the base order for the noun phraseis taken to be one where the noun starts off in the head-final position, and the Ezafe construction is the result of roll-upphrasal movement in a manner discussed below.

4.1. Ezafe as case marker: Larson and Yamakido (2008)

In this section, we will look at Larson and Yamakido’s (2008) analysis of the Persian Ezafe construction. Their paperalso discusses the structure of the English noun phrase (contrasting it with Persian), but here I only focus on their analysisof Persian with references to English only where relevant.

Following Samiian (1994), Larson and Yamakido (2008) suggest that Ezafe is a case-marker inserted to case-license[+N] elements, namely APs, NPs (including attributive and possessive ones), and nominal PPs. According to theirproposal, relative clauses (CPs) and non-nominal PPs do not require case and can thus appear in their base positionwithout the need for a licensing Ezafe.27 While some of the discussions in sections 2 and 3may already raise some issuesfor Larson and Yamakido’s analysis, in this subsection, I focus on two empirical problems with their analysis, namely theirtreatment of relative clauses (CPs) and (non-nominal) PPs (see also Samvelian, 2007, 2008). Both of these problemscarry over to Samiian (1994), who also takes Ezafe to be a case marker.

I start with Larson and Yamakido’s (2008) treatment of relative clauses. According to their proposal, relative clauses,being CPs, do not require case and as a result should appear in their base position without the Ezafe marker. While theydo not discuss reduced relative clauses in Persian, based on their discussion of reduced relative clauses in English, onemight expect their prediction to hold of reduced relative clauses in Persian as well. Samvelian (2008) providescounterexamples to this prediction, repeated below with minor modifications.

(12)

25 The E(e.g. Kurdnoun (seebe taken26 In sec27 FromMeanwhicase.

a.

zafe hish diaKurmas suption 5.a crosle, as th

in

as anlects saji exaport fo2, I pros-linguey lac

javaan-e

invariant formuch as Hawramples in (14)r the agreemepose that it isistic perspectik an overt case

[az

in Pemi, Kurmbelow, ant viewonly p

ve, theymarker

suis

rsian (-e or -yeanji, and Zazaklso Holmberg anof Ezafe. I leavhrasal movemenargue that langlike the Persian

bargashte]
this young-Ez from Switzerland returned ‘this young man who has returned from Switzerland’
b.
aks-e [chaap-shode dar ruznaame]
after voweli) the form od Odden, 20e a closer lot and not heuages suchEzafe, the [+

picture-Ez
published in newspaper ‘the picture published in the newspaper’
Larson (2009) argues that reduced non-finite relative clauses behave like nouns in many languages suggesting thatexamples like (12) should not be seen as surprising. Instead, their nominal behavior should be all the reason for us to

s). In some other Iranian languages/dialects with richer morphologyf the Ezafe varies depending on phi features and other properties of the05; Larson and Yamakido, 2006; Samvelian, 2007, 2008). These mayok at these dialects for future research.ad movement that leads to the appearance of Ezafe.as English also have a head-initial base order for their noun phrases.N] elements, including adjectives will have to undergo movement to get


expect them to be case-marked by the Ezafe. This argument undermines the status of examples such as (12) as evidenceagainst Larson and Yamakido’s (2008) analysis. Meanwhile, under this view, finite restrictive relative clauses are nevernominal and as such are predicted not to take the Ezafe marker. While this may appear to be true of Modern Persian,historical and typological data seem to militate against this generalization.

From a historical perspective, the Persian Ezafe is seen as a descendent of the Old and Middle Persian ‘relativeconnector’, used to connect the noun with the post-nominal restrictive relative clause (Samvelian, 2007; Skjærvø, 2009,also Moyne and Carden, 1974).28 According to Skjærvø (2009), in Middle Persian, there are two ways to modify the noun.The modifying adjective can either be placed before the head noun, which he describes as the older and rarerconstruction, or it can be placed after the head noun by means of the ‘relative connector’ ī, which he describes as ‘‘thedescendent of the Old Persian relative connector and forerunner of the New Persian ezāfe (CNCT)’’ (p. 221). In (13a), wesee an example of the pre-nominal adjective order without the Ezafe. In (13b), we see an example of the post-nominalorder with the Ezafe (or what he glosses as CNCT or ‘connector’). It is worth noting that these examples provide historicalsupport for the correlation made in the current paper between the order of the noun and the modifier and the presence/absence of the Ezafe marker. Meanwhile, the most relevant example for our discussion about the relation between Ezafeand finite relative clauses is given in (13c). We can see that the same form ī used in (13b) and also in (13c) to mark amodifying adjective is used to mark a relative clause (glossed as REL).29

(13)

28 Many29 In theof the Mi30 Dari P31 In theand Sora

a.

thanksse examddle Peersianse examni data

weh

to Azitaples, I arsian daseemsples, I a

.

dēn

Taleghanm adherinta. OBLp sto exhibit tm adherin

i for bringingg to the transctands for Obhe same behg to the trans

this fact to mription and glique-plural.avior (Ilhan Ccription conve

y attentiolossing co

agri, pc).ntions us

n and for an interesnventions used in th

ed in the original sou

ting discue original

rce to av

ssion.source to avoi

oid potential m

d poten

isrepre

good
religion ‘the good religion’ (Middle Persian, Skjærvø, 2009:222)
b.
sāh ī wazurg king CNCT great ‘a/the great king’ (Middle Persian, Skjærvø, 2009:222)
c.
ān mēnōy ī-s gannā-īh ī dām-ān ī ohrmazd azis būd that spirit REL-him foul-ness CNCT creature-OBLp CNCT Ohrmazd from was ‘that spirit, from whom stench came to Ohrmazd’s creatures’
(Middle Persian, Skjærvø, 2009:225)

Furthermore, even today, in many West Iranian languages, closely related to Persian, Ezafe is used for both post-nominal modifiers as well as finite restrictive relative clauses (see Samvelian, 2008, also Bassols-Codina, 1992).30 In(14), we see examples from two dialects of Kurdish, namely Sorani and Kurmanji, both of which use Ezafe for post-nominalmodifiers, but also to introducea fully finite relative clause.31 In (14a), we seeanexample of Ezafe used inSoraniwith a post-nominal PP modifier, while in (14b) we see its use with a finite relative clause. The example in (14c) is fromKurmanji and shows the use of the Ezafe with the post-nominal adjective and possessor. We see in (14d) that,in Kurmanji, the Ezafe is also used for a finite relative clause. Note that Ezafe agrees in person, number and gender inKurmanji.

(14)
a. xânu-y la sar şâx house-Ez at on mountain ‘the house on the mountain’ (Sorani, Samvelian, 2008:346)
b.
aw şâr-a-y (ka) dît-mân that town-def-Ez (that) see.past-1pl ‘the town that we visited’ (Sorani, Samvelian, 2008:347)
c.
mâl-â mazin-â Narmîn-ê house-Ez.Fem.Sg big-Ez.Fem.Sg Narmin-OBL.Fem ‘Narmin’s big house’ (Kurmanji, Samvelian, 2008:344)
d.
mirov-ê ku min dît-î man-Ez.Masc.Sg that I.OBL see-Past ‘the man who I saw’ (Kurmanji, Samvelian, 2008:347)
tial misrepresentation

sentation of Kurmanji


In Modern Persian, restrictive relative clauses are introduced by the particle -i, known as the relative particle, distinct inform from the Ezafe vowel -e. Therefore, Persian grammarians often state that the Ezafe vowel cannot be used with arestrictive relative clause, a point picked up by Larson and Yamakido (2008). The example in (12a) involving a reducedrelative clause is repeated in (15a) with the full relative clause counterpart given in (15b).

(15)

32 It shoapplies togiven in (

(i) a.

33 An anformer cacollapsinEzafe maup, may bmanifesteclauses i34 The E

a.

uld alsothe reli).

ketaabbook-d‘the bionymon be exg themrker, I de mergd in thn Persizafe m

in

be noative p

-(*e)-(ef.-Ezg bookus revtraposboth uo noted at de graman is barker

javaan-e

ted that the Ezarticle, pointin

y)e bozorgbig

’iewer points oed, while the lnder the cateintend to suggifferent heighmatically condeyond the scis optional for

[az

afe canng perha

b.

ut a diffeatter cangory of rest thatts (in a sitionedope of tha few o

suis

ot be used alongps to the connec

ketaab-(*e)-(y)ibook-def.-Rel‘the book that yrence between fnot. S/he takesequiring the Ezathey are identicachema such as (allomorphy betwis paper.f these prepositi

bargashte]

with the Persiantion between the

ke to xarithat you bouou bought’ully finite relativethis syntactic difffe. It should bel syntactically. Th30) below) and heen -e and -i. A

ons, for exampl

this
young-Ez from Switzerland returned ‘this young man who has returned from Switzerland’
b.
javaan-i [ke az suis bargashte (ast)] young-REL that from Switzerland returned is ‘the young man who has returned from Switzerland’
The nature of this relative particle has always puzzled Persian linguists. In light of the above discussion concerning thehistory of the Ezafe marker and its status in other closely related West Iranian languages, the relative marker may beanalyzed as an allomorph of the Ezafe marker, perhaps conditioned by the syntactic context of the full relative clause. Assuch, the full finite relative clause may be seen as yet another counterexample for Larson and Yamakido’s (2008) strictgeneralization about the absence of Ezafe in this context.32,33

While Larson and Yamakido (2008) may face some issues in dealing with relative clauses in Persian and other relatedlanguages, the strongest evidence against their treatment of Ezafe as a case marker in Modern Persian comes fromprepositional phrases. Following Karimi and Brame (1986) (see also Karimi and Brame, 2013), Larson and Yamakido(2008) correctly classify Persian prepositions into two categories, one noun-like and the other ‘true’ prepositions. Thenoun-like prepositions such as ru ‘on’, zir ‘under’, baalaa ‘above’, paayin ‘below’ have nominal properties. For example,they can be pluralized and can be used in other nominal structures. Interestingly, they also take the Ezafe marker, whenfollowed by a noun, as shown in (16).34 This stands in sharp contrast to the true prepositions such as dar ‘in’, be ‘to’ andbaa ‘with’, which would never take the Ezafe, as shown in (17).

(16)
a. zir-e sandali under-Ez chair ‘under the chair’
b.
baalaa-ye miz above-Ez table ‘above the table’
(17)
a. dar(*-e) ganje in closet ‘in the closet’
b.
baa(*-ye) maadar with mother ‘with(the) mother’
The facts in (16)--(17) are not problematic for Larson and Yamakido. Similarly to other treatments of Ezafe (Ghomeshi,1997; Kahnemuyipour, 2000; Ghaniabadi, 2010, this paper), the presence or absence of Ezafe on the prepositions in(16)--(17) follows from their status as nouns or true prepositions. Crucially, however, Larson and Yamakido claim that true

(colloquial) definite marker (see section 5.2). The same restrictionEzafe marker and the relative particle. The relevant examples are

d-ight-2sg.

clauses and reduced relative clauses in Persian, namely that theerence between these two constructions as a potential problem fornoted that by suggesting that both constructions may involve thee two constructions are certainly different in their structural make-ave different syntactic properties. In fact, one difference is alreadycloser examination of the syntax of reduced and fully finite relative

e ru ‘on’.


(non-nominal) PPs following a noun do not need to be case-licensed and should therefore appear without the need of alicensing Ezafe preceding them. This is contrary to fact, as shown in (18) (see also Samvelian, 2007, 2008). In theseexamples, the PPs are obligatorily preceded by the Ezafe vowel, similarly to any other post-nominal modifier of a noun,such as an adjective. This is a serious problem for Larson and Yamakido’s analysis of Ezafe as case.

(18)

35 In secCrucially,36 If thefinite rela37 Whilethe PP sefuture res

(i) a.

b.

a.

tion 5.2no Ezrelativetive clawe havem toearch

bahsarguporsques

Mohammad

, I provide a heaafe appears in thparticle used wituse will constitue seen exampleallow both possibbut it does appe

(-e) bâment-Ez withesh(-e) aztion-Ez from

[aks*(-e)

d-movement anis context.h fully finite relate yet anothers of nouns takinilities, with or war that there is

HassanHassanra’is-jomhurpresident

dar

alysis

tive clexampg PPsithoua sub

ganje]-raa

of post-nomina

auses is analyzle of a phrasal(as adjuncts) r

t the Ezafe. Sotle semantic dif

be

l adje

ed asmodequirme exferen

Ali

ctive

an aifier uing thampce be

dad
Mohammad picture-Ez in closet-acc. to Ali gave ‘Mohammad gave the picture in the closet to Ali.’
b.
ne-mi-tavaanest-am tasmim be-gir-am [sobh-haa-ye
s used in the cont

llomorph of the Ezsed in the Ezafee Ezafe marker,les are given in (itween the two o

baa

ext of t

afe maconstrusome n) belowptions.

maadar]-raa

he colloquial defin

rker as suggestedction.ouns with a more. I leave the analy

neg.-dur.-can.past-1sg.
decision subj.-get-1sg morning-pl.-Ez with mother acc. bishtar dust daar-am yaa [sobh-haa-ye baa kabutar-haa]-raa more friend have-1sg or morning-pl.-Ez with pigeon-pl.-acc. ‘I could not decide whether I liked the mornings with Mother more or the mornings with the pigeons.’ (Yek ruz maande be eyd-e paak (novel), Z. Pirzaad, p. 80, cited in Samvelian, 2008)
In this subsection, we looked at Larson and Yamakido (2008), an analysis which attempts to account for cross-linguistic
word order variations in the nominal domain by taking the post-nominal (Persian) order to be basic. Several argumentsagainst their analysis of the PersianEzafe constructionwere provided. This leads us to consider the other approach to thesewordorder variations, i.e. onewhich takes the pre-nominal (English) order ofmodifiers asbasicandderives the post-nominal(Persian) order. There are two possibilities discussed in the literature for deriving post-nominal modifiers from a base-generated pre-nominal structure in different languages. The first one involves headmovement and the second one phrasalmovement.Within the first typeof approach,most notably,Cinque (1994) developsa (partial) headmovement analysisof theRomance DP, where all adjectives start off pre-nominally as heads but some surface in a post-nominal position due to thehead movement of the noun to higher projections (see also Ritter, 1991 for Hebrew, among others). Alternatively, manyscholars (see, for example,Cinque, 2005, 2010;Pearce, 2002;Shlonsky, 2004, amongmanyothers) have taken the derivedpost-nominal order of modifiers to be the result of phrasal movement from a base order which takes the modifiers to be XPsplaced in specifiers of functional projections above the noun. It should be noted that many head movement analysesproposed previously have been reformulated in a phrasal movement framework for theoretical reasons (e.g. Cinque, 1994/1996 vs. Cinque, 2010). The present paper sides with the latter approach and offers an account of this kind for the PersianEzafe construction in section 4.3.35 Meanwhile, in order to justify this choice for Persian, it is important to first establish thehead vs. phrasal status of the modifiers and their structural relation to the noun. This is the topic of the following section.
4.2. Persian post-nominal modifiers: heads or phrases

The head vs. phrasal status of modifiers used in the Persian Ezafe construction has been amatter of some controversyin the literature on the topic. While it was widely assumed for a long time that all the elements in the Ezafe Domain (exceptfor the possessive noun phrase) are non-maximal (see, e.g. Samiian, 1994; Ghomeshi, 1997; Kahnemuyipour, 2000), thehead status of the elements within the Ezafe domain is undermined once we expand our dataset. We have already seencases of PPs (in (18)) and (reduced) relative clauses (in (12)) that seriously question this assumption.36 In fact, a closerlook at Persian adjectives reveals that there are some adjectives that can appear maximally, not only involving intensifiersbut also complements, as shown in (19) (see Samvelian, 2007, 2008 for more examples).37

(19)
a. saalon-e [por az jam’iyyat]-e sinemaa hall-Ez full of population-Ez cinema ‘the movie theater filled with people’
ite marker in Persian.

in section 4.1, a fully

thematic relation withsis of such cases for


38 We wstructure39 Thanwith resp40 Ghandistinctionconsisten41 The eGhaniabawhich tak

b.

*

ill see iinvolviks to anect to tiabadi (betwecy withxampledi (201es plac

raftaar-e

n section 5.2 thatng the colloquial danonymous reviehe status of post-2010) is written inen lower and uppethe rest of the pas are slightly mod0), the elided mate after the ellipsis

[xeyli

the ambigefinite mwer for ranominal mthe Distrir case leper.ified moserial with.

dur

uousarker -ising thodifie

butedtters is

tly tostriketh

az

status oé in Peis aspers in PeMorphoirreleva

reflect trough d

entezaar]-e

f a bare adjective arsian.ct of Ghaniabadi’s prsian.logy framework andnt to our purposes a

he transcription cono not contain the Ez

shahrdaar
behaviour-Ez very far from expectation-Ez mayor ‘the mayor’s totally unexpected behaviour’
The existence of phrasal modifiers illustrated above rules out the possibility of treating all post-nominal modifiers as
heads.Aswewill see in section 4.3, in this paper, post-nominalmodifiers are taken to beXPswhich reside in the specifiers offunctional projections above the noun. Under this view, in accordance with Bare Phrase Structure (Chomsky, 1995, andsubsequent authors), a bare adjective is treated as A/AP and can occupy a structural position similar to that of an AP with acomplement.38Thisuniformanalysisofbareadjectivesandphrasalmodifiers inPersianhasbeenchallengedbyGhaniabadi(2010).39 As a starting point, Ghaniabadi assumes the strict ordering of post-nominal modifiers in Persian provided in (20).40
(20)
N A AP PP Possessor
Ghaniabadi takes all the post-nominal modifiers to be right-adjoined in the structure of the noun phrase, but he treats
bare adjectives and phrasal modifiers in radically different ways. For him, bare adjectives are heads which are head-adjoined to the noun, whereas AP and PPmodifiers are phrasal elements in the specifiers of functional projections abovethe NP. It is this structural distinction between the two types of modifiers which is at the heart of Ghaniabadi’s proposal.The suggested fixed order between the bare adjectives and the phrasal modifiers shown in (20) follows from this structuraldifference for Ghaniabadi. Before we can proceed with the uniform treatment of bare adjectives and phrasal modifiers inthe Ezafe construction as XPs (base generated in the specifiers of functional projections above the noun in a right-branching structure), we need to assess whether Ghaniabadi’s proposed distinction between these two categories iswarranted. For this purpose, we need to review his evidence. The main argument Ghaniabadi (2010) offers for thisdistinction comes from a type of noun ellipsis in Persian (what he refers to as the Empty Noun Construction), where thehead noun is elided leaving behind one or more modifiers. He argues that his proposed structural difference betweenbrae and phrasal modifiers is the source of a difference in behaviour between the two classes of modifers in the context ofthe Empty Noun Construction (seeGhaniabadi, 2010 for details). His major claim is that this type of ellipsis is only possiblewith bare adjectives and not with AP or PP modifiers. In other words, a head noun can be elided (along with other head-adjoined elements) only if the remnant is a bare adjective (A) and not an AP or a PP. Some of his examples showing therelevant points are given in (21)--(23), where strikethrough shows ellipsis.41 (21a) is an example of an elided noun with asingle remnant adjective; (21b) shows an example where the noun is elided along with a bare adjective leaving behind yetanother bare adjective. According to Ghaniabadi, the ellipsis of the noun along with another bare adjective is possible,because these adjectives are recursively head-adjoined to the noun, making it possible to elide the noun with one or morebare adjectives as long as what is left behind is another bare adjective and not a phrasal one. The examples in (22) and(23) show the suggested impossibility of this type of ellipsis with AP and PP modifiers, respectively.
(21)
a. Sajjaad pirhan-e aabi pushid, Sina pirhan qermez
s a head or

roposal as

uses italiciznd capital le

ventions usafe marke

Sajjad
shirt-EZ blue wore.3sg Sina shirt red ‘Sajjad wore a blue shirt, Sina a red one.’ (Ghaniabadi, 2010: p. 61)
b.
Sina kif-e charm-i-ye kuchik-o baa kif charmi
a ph

a po

ed lotters

ed ir bec

bozorg

rase allow

tential prob

wer-casehave bee

n this papeause he ta

avaz

s it to take pa

lem for assum

letters n and an used in (20)

r. Note also tkes Ezafe to b

kard
Sina bag-Ez leather-adj-Ez small-Acc. with bag leather big exchange did.3sg ‘Sina exchanged the small leather bag with a big one.’ (Gha niabadi, 2010: p. 79)
(22)
a. keshvar-haa-ye [AP negaraan-e afzaayesh-e qeymat-e naft] ettelaa’iyye-i saader
rt in he

ptions

, in plato avoid

hat in te a late

kard-and
Country-pl.-Ez worried-Ez increase-Ez price-Ez oil statement issue did-3pl ‘The countries worried about the increase of the price of oil issued a statement.’
b.
keshvar-haa [AP negaraan-e afzaayesh-e qeymat-e naft] ettelaa’iyye-i saader kard-and ‘The ones worried about the increase of the price of oil issued a statement.’ (Ghania badi, 2010: p. 69)
ad-movement in a

made in this paper

ce of N and A. Theconfusion and for

he examples frominsertion process


(23)

42 Note43 The eexampleswith the rone founungrammin these eamong spconstruct

a.

*

that agxample. This pight prod (25aatical alided ceakersions sh

kafsh-aa-ye

reement with ans in (24) and (25rompted me to vsody. The result) grammatical and one found (2onstructions, aswhich is worthould only be jud

[PP

inani) areerifywas snd (25b) slsuggexploged u

tu(-ye)

mate plural sbased on mythe examplesomewhat dif5b) slightlyightly degradeested by Gharing in futurender the righ

vitrin-e

ubject is optionnative judgmenwith seven otheferent for (25). Odegraded, oned but (25a) ungniabadi (2010),research. In tet pragmatic con

maqaaze]

al in Persian,t. An anonymr native speakf the seven sfound bothrammatical. Tis on the wronsting these ellditions and th

xeyli

hence tous revers. Allpeakersslightlyhese reg trackiptical ce corre

qashang-e

he singular agriewer raised soof these seven, three judged bdegraded, onesults confirm th. Meanwhile, itonstructions, itsponding proso

shoe-pl-Ez
inside-Ez window-Ez shop very beautiful-is ‘The shoes inside the window shop are very beautiful.’
b.
kafsh-aa [PP tu(-ye) vitrin-e maqaaze] xeyli qashang-e ‘The ones inside the window shop are very beautiful.’ (Ghaniabadi, 2010: p. 70)
It should be noted that these cases of ellipsis, like many other cases discussed in the literature, are licensed only under
certain pragmatic conditions. Ghaniabadi discusses these conditions at some length, but they can be roughlycharacterized as the elided noun having a syntactic and semantic identity relation with an antecedent and the remnantadjective being (contrastively) focused. It is important to ensure that these conditions are met in trying to come up withexamples for this type of noun ellipsis in other contexts. Moreover, as Ghaniabadi (2010) points out, for reasons that arenot very well understood, there may be lexical restrictions on which modifiers may license these elided nouns. Forexample, Ghaniabadi gives the Persian adjectives jaaleb ‘interesting’, zibaa ‘beautiful’ and didani ‘worth watching’ asexamples of adjectives which do not license the elliptical construction even in their bare form. The strict pragmatic andlexical restrictions on these elliptical constructions already undermine any strong conclusion about the head vs phrasalstatus of the modifiers based on the ungrammaticality of a few examples involving one or the other type of remnant.Meanwhile, for the very least, one would need to explore a wider range of contexts and lexical choices to assess thevalidity of Ghaniabadi’s claim with respect to the impossibility of noun ellipsis with phrasal modifiers as remnants. In (24)and (25), we see grammatical examples of noun ellipsis with a modifying PP and AP as remnants, respectively. For clarityof judgment, the required contrastive context is provided. Ellipsis is marked with a strikethrough.42
(24)
Context: There are some books on the table and some under the table. The speaker has organized them and iscontrasting their location with respect to their ownership. The preposition zir ‘under’ is contrastively focused andprosodically prominent. ketaab-aa-ye [PP ru-ye miz] maal-e man-e, ketaab-aa-ye book-pl-Ez on-Ez table property-Ez I-is book-pl-Ez [PP zir-e miz] maal-e to-e under-Ez table property-Ez you-is ‘The books on the table are mine, (the ones) under the table are yours.’
(25)
Context: There are two jars, one filled with wine and the other with vinegar. The speaker in contrasting the contentof the jars, with serke ‘vinegar’ contrastively focused and prosodically prominent. a. Ali tong-e [AP por az sharaab]-o bardaasht, tong-e
Ali
jar-Ez full of wine-Acc. took jar-Ez [AP por az serka]-ro gozaasht
full
of vinegar-Acc. left b. ‘Ali took the jar filled with wine, and left the one filled with vinegar.’
Ali
tong-e [AP por az sharaab]-o bardaasht, Ali jar-Ez full of wine-Acc. took Hassan tong-e [AP por az serka]-ro Hassan jar-Ez full of vinegar-Acc. ‘Ali took the jar filled with wine, and Hassan the one filled with vinegar.’
The examples in (24) and (25) provide strong evidence against Ghaniabadi’s claim that the noun ellipsis construction is
ungrammatical with a phrasal AP or PP remnant and in turn undermines his attempt to treat the bare adjectives andphrasal modifiers differently in terms of their structural relation with the head noun, with the bare adjectives recursivelyhead adjoined to the noun and the phrasal ones in specifier positions above the NP.43 In fact, once a wider range of data isconsidered, Ghaniabadi’s basic assumption about the respective order of bare and phrasal adjectives, shown in (20), isseriously challenged. Recall that in Ghaniabadi’s system, bare adjectives, which are head-adjoined to the noun, are
eement in (24).me questions about the status of thesespeakers found (24) to be grammaticaloth (25a) and (25b) to be grammatical,found (25b) grammatical and (25a)

at an outright ban on phrasal remnantsalso points to some individual variationis important to bear in mind that thesedy.


strictly ordered before phrasal adjectives. The examples in (26) show that a different order, one where the phrasaladjectives appear before the bare ones is also possible.44,45

(26)

44 Notenoun theproblem,45 WhileMeanwhiFor more46 Thesethe plurareading,47 The u‘rich peopquite prod

a.

that thisbare ain the ebare ale, thison theexam

l markethe phrse of plle’, xosuctive i

maqaale-ye

alternative ordedjective is modifxamples in (26),nd phrasal adjecis not related to tposition of adju

ples were providr is pluralizing thase would be touralized adjectivehgel-aa beautiful-nPersian.Recen

por

r mayying. Iadjectives ahe phrnct aned bye whtally gs, whipl. ‘betly, the

az

be ds it mtivesre noasald argan anole Prammch I sautifuterm

eshtebaah-e

ispreferred in maodifying the heahave been choset strictly ordered,status of PP but mument PPs in thonymous revieweP (or the elided natical.uggest should bel people’, divune-hsabz-aa ‘the gree

mozaxraf
article-Ez full of mistake-Ez terrible ‘the terrible article with a lot of mistakes’
b.
xune-ye xaali az sakane-ye fakasani house-Ez empty of resident-Ez decrepit ‘the decrepit house without any residents’
In fact, with the order given in (26), one can even elide the noun along with the phrasal modifier AP, leaving the bare
adjective behind. An example is given in (27). This example provides further evidence against Ghaniabadi’s distinctionbetween bare and phrasal adjectives. Recall that according to Ghaniabadi, only bare adjectives can be elided along with ahead noun (cf. (21b)) as they are considered to be head-adjoined to the noun. This example also further establishes thepossibility of placing a phrasal adjective before a bare one.
(27)
Sina kif-e [AP por az pul]-e kuchik-o
ny contexd noun on whichPP modiay rathee nominar. It is imoun in G

reanalyzeaa crazyns’has be

baa

ts for pr the imare unafers har followl domaportanhaniab

d as ze-pl. ‘craen coin

kif-e [AP por az pul]-e

*

arsing reasons, as it might set up anmediately preceding noun which ismbiguously taken to modify the heave to follow both, as indicated also infrom the same kind of universal ordein, see Cinque (2010) and Kayne (2t to note that (28b) is ungrammaticaladi’s analysis) and not the noun m

ro-derived nominals, is quite commonzy people’ (cf. English *riches, *beautied to refer to the supportersof theGre

bozorg

ambiguitypart of thed noun forGhaniabaring discu000, 2004for the releaqaaze ‘st

in Persianfuls,??crazenMovem

avaz

with respectAP? To avoipragmatic readi’s schema inssed in sectio).vant reading wore’. With the

: puldaar-aa ries). This procent (see footno

kard
Sina bag-Ez full of money-Ez small-Acc. with bag-Ez full of money-Ez big exchange did ‘Sina switched the small bag full of money with the big one.’
In addition to the noun ellipsis construction discussed above, Ghaniabadi (2010) provides evidence from plural
marking, showing that it can only appear on bare adjectives and never on phrasal modifiers. Ghaniabadi analyzes thisconstruction, which I will hereafter refer to as pluralized adjectives, as a combination of noun ellipsis (of the kind discussedabove) and local dislocation of the stranded plural marker to the adjective. He uses this type of evidence as further supportfor the structural divide between bare adjectives and phrasal modifiers. Some examples are given in (28).46
(28)
a. kafsh-aa-ye gerun ! gerun-aa shoes-pl-Ez expensive expensive-pl ‘the expensive shoes’ ‘expensive ones’
b.
kafsh-aa-ye [PP tu-ye vitrin-e maqaaze] ! [PP tu-ye vitrin-e maqaaze]--haa shoes-pl-Ez in-Ez window-Ez store ‘the shoes inside the shop window’ ‘the ones inside the shop window’
While the facts in (28) are unquestionable, the impossibility of placing the plural marker on phrases can find a different
explanation, such as the selectional restrictions of the plural suffix. In fact, in light of the discussion of the noun ellipsiscases above undermining a structural divide between the bare adjectives and phrasal modifiers, I propose that thepluralized adjectives can be analyzed as zero-derived nominal forms which can therefore take the plural marking.Consequently, the impossibility of placing the plural marker on the phrasal elements is reduced to the selectionalrestrictions of the nominalizing zero affix, namely its inability to select phrases. Under this view, in true cases of nounellipsis involving plural nouns, the plural marker is elided along with the noun. An example of this kind was given in (24)(for more examples of this kind, see Ghaniabadi, 2010). When the plural marker appears on the modifier, we are notdealing with ellipsis, but rather the plural form of a nominalized adjective.
It is worth noting that pluralized adjectives seem to exhibit different properties from the noun ellipsis cases, providingsupport for a structural distinction between the two phenomena. For instance, pluralized adjectives appear to be lessconstrained pragmatically and also in terms of lexical choice.47 Thus, for instance, of the three adjectives noted byGhaniabadi (2010) (see above) as examples of adjectives incompatible with noun ellipsis, namely jaaleb ‘interesting’,zibaa ‘beautiful’ and didani ‘worth watching’, the last one is perfectly fine in the plural didani-haa ‘those worth watching’,

to thed thissons.(20).n 4.3.

herelatter

ich-pl.ess iste 9).


and the second one is degraded with the default plural marker --haa (?zibaa-haa), but fine with the plural marker typicallyused for humans, zibaa-yaan ‘the beautiful ones’. This already suggests a distributional difference between the nounellipsis construction and the pluralized adjectives.

Further support for a distributional (and consequently a syntactic) distinction between noun ellipsis and pluralizedadjectives comes from examples such as the ones in (29) where noun ellipsis is ungrammatical (29a,c) but the parallelpluralized adjective construction is grammatical (29b,d), pointing once again to the more constrained nature of the formercompared to the latter. If the pluralized adjective construction involved a noun ellipsis step, as suggested by Ghaniabadi,there is no reason why the examples just involving noun ellipsis should be ungrammatical. Note that the ungrammaticalityof the noun ellipsis cases cannot be attributed to the elision of the plural marker along with the head noun. We havealready seen this possibility in (24) (see Ghaniabadi, 2010 for more examples of elided plural nouns). In fact, if myproposal is on the right track, in true noun ellipsis constructions, the plural marker is obligatorily elided with the head noun,and all cases where the plural marker appears on the modifiers should be treated as cases of nominalization.

(29)

48 The wderived alove’, or t49 The aellipsis wpropertiesexamplemarker --éand not n

(i) unthat‘That

50 Whilethis sectiodeletion udeleted a

a.

ord aas a nouhe comnalysisith singof the

of a no(see soun el

maashcar-EzbeautifI deven as itnder idlong w

daaneshju-haa-ye

sheq does not have a pen, but it is primarily an aparative --tar to form aaof pluralized adjectivesular nouns may also inso-called adjective to seminalized adjective withection 5.2), not even prelipsis.

in-e xoshgel maal-ebeautiful property

ul car is yours and thelop an alternative analyswould take us far afieldentity (Merchant, 2001,ith the head noun (and

aasheq-e

rfect equivaledjective, suppsheq-tar ‘moas cases ofvolve nominae if it is behavsingular refesent in the fir

to-e-Ez you-isugly one is mis of the Ezaffrom the mainamong otherthe plural ma

futbaal

nt in its adjeorted by thre in love’.nominalizatlization ofing as a nourence. Itsst conjunct.

o birixt-and ugly-dine.’e constructfocus of ths) where Erker, if pres

jaa-shun-o

ctival sense in Englise fact that it can take

ion (as opposed to ethe adjective. This in (as is the case whenominal status is higThis example should

é maal-e mef. property-Ez I-i

ion below, I will not ae paper. Meanwhile,zafe is present in theent) should give us

be
student-pl-Ez loving-Ez football place-their-Acc. to *
daaneshju-haa-ye
aasheq-e basketbaal daadand48
student-pl-Ez
lover-pl-Ez basketball gave b. daaneshju-haa-ye aasheq-e futbaal jaa-shun-o be
student-pl-Ez
loving-Ez football place-their-Acc. to aasheq-aa-ye basketbaal daadand lover-pl-Ez basketball gave ‘The football loving students gave their place to basketball lovers.’
c.
daaneshju-haa-ye jadid-e puldaar be jaa-ye student-pl-Ez new-Ez rich to place-Ez *
daaneshju-haa-ye
ghadimi-ye bipul umad-an student-pl-Ez old-pl-Ez poor came-3pl
d.
daaneshju-haa-ye jadid-e puldaar be jaa-ye student-pl-Ez new-Ez rich to place-Ez ghadimi-haa-ye bipul umad-an old-pl-Ez poor came-3pl ‘The new rich students replaced the old poor ones.’
While a full analysis of the pluralized adjective construction and its differences with noun ellipsis is beyond the scope ofthis paper, the above noted differences seriously undermine treating this construction on a par with noun ellipsis.49

Many interesting and important questions remain with respect to the correct analysis of the constructions discussedabove and their various restrictions, but the above discussion shows that the division of the Ezafe domain in Persian intotwo domains, one for bare adjectives (headmodifiers) and another one for phrasal modifiers, as suggested by Ghaniabadi(2010), is unnecessary and unwarranted.50 In addition, as stated previously, the established phrasal status of somemodifiers in Persian rules out the possibility of treating them uniformly as heads, making a head-movement analysis of theEzafe construction impossible (but see section 5.2 and footnote 37). All of this paves the way for a uniform treatment ofbare adjectives and phrasal modifiers as XPs occupying specifiers of functional projections above NP. In what follows, Idevelop a phrasal movement analysis of the Ezafe construction using what is known in the literature as roll-up movement

h. It is often translated as ‘lover’ due to its ability to be zerothe deadjectival nominalizer --i to form aasheq-i ‘being in

llipsis) raises the possibility that some cases of apparents in fact a desirable outcome. The key is to assess then it takes the plural marking) or not. In (i) below, we see anhlighted by the fact that it is taking the colloquial definitetherefore be treated as a case of a nominalized adjective

an-es

ttempt to provide an account of noun ellipsis discussed inI would expect that an analysis based on some form of PFparallel structures, and the second occurrence of it gets

the right result.


(see, for example, Cinque, 2005, 2010; Pearce, 2002; Shlonsky, 2004, among many others).51 In the roll-up movementapproach, there are additional functional projections (with null heads) mediating the noun and modifiers, withcomplementsmoving to the specifiers of these intermediate projections in a roll-up fashion. This option takes themodifiersto be XPs occupying specifier positions rather than heads as in the head-movement option, and as such avoids theproblem raised for that approach.52

4.3. Roll-up movement analysis of the Persian Ezafe construction

In his seminal work on the syntax and semantics of adjectives, Cinque (2010) develops a system in which the base orderof the noun phrase is universally head final. In this system, adjectives have two possible sources: direct and indirectmodification. According to Cinque, direct modification is lower in the structure, rigidly ordered and is associated withindividual-level, nonrestrictive and nonintersective readings. Indirect modification, is higher structurally, has a reducedrelative clause as its source, is not rigidly ordered and has stage-level, restrictive and intersective readings.While a detaileddiscussion of these properties is beyond the scope of this paper, to illustrate Cinque’s basic classification, we can considersome English examples. In visible visible star, the first visible means ‘visible now’ (stage-level), whereas the second onemeans ‘generally visible’ (individual-level). An example such as his unsuitable acts couldmean ‘those of his actswhichwereunsuitable’ (restrictive) or ‘all of his acts which happened to be unsuitable’ (nonrestrictive). Finally, in the beautiful beautifuldancer, the first occurrence of beautiful refers to the beauty of the person (intersective) and the second one to the beauty ofthe dancing (nonintersective).53A slightly revised version of Cinque’s general schema given in (30) is adopted here.54,55

(30)

51 A var2006), mderived bit would h52 Holmpossibiliti53 I refe54 I am uof Numbe55 It is wsuch as s

Structure of DP (adapted from Cinque, 2010)

[TD$INLINE]

DP

Dem

Numeral

AP

APIndirect modification APs

AP

AP

APDirect Modification APs

AP

NP

iant of the ‘‘roll-up’’ movement, known as ‘Intraposition’ movement (e.g. Rackowski and Travis, 2000; Kahnemuyipour and Massam,ay also be considered. In this approach, the modifiers are located in heads of functional projections above N/V and the inverse order isy successive movement of complements to the empty specifiers in their sharedmaximal projection. I am not pursuing this option here, asave to take the modifiers to be heads and as such would face the same problem as the head movement analysis discussed above.berg and Odden (2005) propose a ‘roll-up’ derivation of the Izafe construction in Hawrami which differs in many details from thees discussed in this paper. I am abstracting away from their proposal here.r the reader to Cinque (2010) where he lays out these as well as some other semantic differences in detail.singNumeral for Cinque’s (2010)Num to avoid confusion withNum as the locus of the number feature (plural vs. singular), and the headr Phrase (Ritter, 1992).orth noting that whether the source of the adjective ordering is due to a universal base order (a la Cinque) or anchored in semantic notionscope (see, for example, Ernst, 2001) is beside the point here. The crucial point for the discussion below is the existence of such an order.


According to Cinque (2010), any variation of the order shown in (30) is the result of phrasal movement in a roll-upfashion. I extend the roll-up analysis to the Persian Ezafe construction. As such, the Persian DP is taken to be head-final,with the NP merged at the bottom of the tree structure and the APs residing in the specifiers of projections above it.56 TheDemonstrative (Dem) and the Numeral are heads higher up in the tree structure in accordance with (30). In addition, asdiscussed above, there are intermediate projections enabling the roll-up derivation. The relevant structures and roll-upmovements are shown schematically in (31), where the projections hosting the APs are marked as XP, YP, etc. and theintermediate projections are marked as AgrPs (following Cinque, 2010, amongmany others).57 Under this view, the Ezafemarker can be seen as the surface realization of the suggested inversion process, i.e. a linker in the sense of den Dikken(2006). Crucially, the height of the movement corresponds to the realization of the Ezafe marker. So, for example, the‘overt’movement stops below elements that are high in the universal schema such as numerals and demonstratives andas such these pre-nominal elements do not have the Ezafemarker. This connection between the universal schema for theDP and the elements bearing the Ezafe provides further support for the proposed phrasal roll-up analysis.

It is worth noting another difference between the elements that are involved in the Ezafe construction and those that arenot. While the former are phrases, the latter are heads. Thus, one can see the head-phrase distinction as the underlyingreason for why the inversion process stops, resulting in pre-nominal elements, all heads with no Ezafe marking, and post-nominal elements, all phrases marked with the Ezafe. If we take the movement of the heads whose specifiers host theadjectives to the Agr heads above them (as shown in (31), e.g. X to Agrx) as a crucial step in the roll-up derivation, then onecan see the head status of high elements such as Dem or Numeral as blocking this movement, and thus disabling the roll-up movement.

(31)

56 I am athe EzafePersian,57 Recaintermedi

Deriving the Ezafe construction via roll-up movement

[TD$INLINE]

DemP

Dem NumeralP

Numeral

AgrxP

Agrx XPEz(-e)

AP

X AgryP

Agry YPEz(-e)

AP

Y NP/NumP

ssuming that number is realized on N and moves as part of the NP complex. Note that the plural suffix is tagged along with the noun inconstruction. If one takes number (Num) to head its own projection (Ritter, 1992), then the roll-up movement should start with NumP inwith N raising to Num first.ll that in some Iranian languages/dialects, Ezafe shows morphological agreement, providing support for the Agreement status of theseate projections. See example (14) above and footnote 25.


Under this view, the order of elements found in a language like English can be seen to reflect the base generated orderof elements and as involving no movement at all (see, for example, Cinque, 2010). Alternatively, one can take English toinvolve the same movements ‘covertly’. The Ezafe marker then finds an explanation in the context of the correlationbetween (overt) movement and morphology discussed above.

The roll-up derivation shown in (31) combined with the base structure in (30) leads to some predictions about the orderof adjectives in Persian. According to Cinque (2010), as discussed previously, direct modification adjectives are rigidlyordered while indirect modification adjectives are not. In addition, Cinque argues that many adjectives can have either adirect modification source or an indirect one. As such, they might show variability with respect to the order they appear in.For example, while ‘‘the big red dog’’ is the preferred order in English, as it reflects the order of adjectives within the domainof direct modification, ‘‘the red big dog’’ is also acceptable because ‘‘red’’ can be used as an indirect modifier too. Thesame is true of Persian. Thus, both sag-e qermez-e bozorg (dog-Ez red-Ez big) and sag-e bozorg-e qermez (dog-Ez big-Ez red) are acceptable.58 Crucially, if unambiguously direct modification adjectives are used, then English shows a strictorder and the roll-up analysis makes the prediction that Persian should exhibit the mirror-image order. This prediction isborne out, as shown in the examples in (32).59 Note that each of these examples involves one unambiguously directmodification adjective which has to appear closer to the noun (for the relevant meaning). Thus, for example, the‘‘classificatory’’ adjectives ‘nuclear’ and ‘theoretical’ in (32a, c) and the ‘‘adverbial’’ adjective ‘heavy’ in (32b) have to beclose to the noun. The other adjective may itself be an example of an unambiguously direct modifier (for example, ‘former’in (32b)) or may allow an indirect source as well. The important point is that these particular adjectives are rigidly ordered(even when both are direct modification), with English reflecting their base order. Crucially, Persian shows the mirror-image order of English.60

(32)

58 Somewords, innot totallythe base59 This d60 An anas we takprogramreferenceassume athe elemepredictionelementsEzafe. Th61 This ifor both E

a.

speakEnglishlost. Worder aiscussonymoe the inwhich as thereright-bnts. Thaboutare rige anals ungranglish

fizikdaan-e

ers report an acce, even if both ordhy should this bend those whichion is inspired byus reviewer pointdirect modificatiossumes a basein) has as its goalranching structure purpose of thethe order of adjeht-adjoined, strucysis in this papermmatical for the rand Persian.

hasteyi-ye

ptability differeers are possibl? It would be inare derived viaCinque’s (201s out that the dn adjectives tomerge order ofan attempt to ae and is incompabove discussctives, whichturally lower, ahas attemptedelevant sense.

javaan
Physicist-Ez nuclear-Ez young fizikdaan-e javaan-e hasteyi *‘a young nuclear physicist’
b.
mashrubxor-e sangin-e
ne

a

in

qabli
drinker-Ez heavy-Ez former/previous mashrubxor-e qabli-ye sangin61 *‘the former/previous heavy drinker’
c.
zabaanshenaas-e nazari-ye
ce between P, one is highlteresting to sroll-up movem0) presentatiota in (32) cantake scope ovelements in account for varatible with onion was to shs borne out. Fd take the Ezto explain thIn the given or

baa-este’daad
linguist-Ez theoretical-Ez with-talent
?
zabaanshenaas-e baa-este’daad-e nazari ?‘talented theoretical linguist’
d.
dalil-e asli-ye qeyre-qaabele-qabul reason-Ez main-Ez unacceptable
?
dalil-e qeyre-qaabele-qabul-e asli ?‘the unacceptable main reason’
In this subsection, I developed a roll-up analysis of the Persian Ezafe construction which captures the strongcorrelation between the presence/absence of the Ezafe marker and the order of the modifers in reference to the noun.According to this analysis, pre-nominal elements are heads, are not involved in the roll-up derivation and do not carry the

ersian and English with respect to the variability of the order of adjectives. In othery preferred in the unmarked case, but in Persian, the preference is undermined, ifee if there is a general difference between languages (or adjectives) which reflectent. I leave a more thorough investigation of this issue for future research.n of similar facts in English and Italian.also be captured by a right-adjunction analysis of post-nominal modifiers as longer direct modification adjectives. This is correct but a few points are in order. Theright-branching structure (also known as cartography, see Shlonsky, 2010 and

ious possible and impossible word orders across languages. This approach has toe which allows a mixture of left- and right-adjunction, reflecting the surface order ofow that by adopting this type of framework, the roll-up analysis makes a certaininally, the mixed approach to adjunction has nothing to say about why certainafe marker, while others are left-adjoined, structurally higher and do not take theis correlation.der, ‘‘heavy’’ could only have a predicative meaning, i.e. heavyweight. This is true


Ezafe marker. Post-nominal elements, on the other hand, are phrases, whose surface position is the result of the roll-upderivation, leading to the appearance of the Ezafe marker.62 In the next section, we will consider two cases where theEzafe construction is somewhat surprisingly absent, namely the superlative (and ordinal) constructions and bare post-nominal adjectives used in the context of the colloquial definite marker.

5. The missing Ezafe

In this section we will look at two cases where the Ezafe construction is not instantiated, even though it may beotherwise expected. An account of these cases will be provided. The first case, discussed in section 5.1, occurs with thesuperlatives (and ordinals). We will see that superlatives are pre-nominal, and seemingly high in the structure, and assuch the fact that they are not part of the Ezafe domain may not be surprising. Meanwhile, the mere fact that their simpleand comparative counterparts are involved in the Ezafe construction begs an explanation for their distinct behavior. Insection 5.2, we will consider the case of post-nominal modification without the Ezafe (what I call Modification WithoutEzafe or MWE) which occurs in the context of the Persian colloquial definite marker.

5.1. Superlatives and ordinals in Persian

In section 3, we briefly considered superlative adjectives in Persian. The superlative, as shown in (33), is consistentlypre-nominal and as such in not part of the Ezafe construction (see also Samiian, 1983; Ghomeshi, 1997; Ghaniabadi,2010).

(33)

62 Scopeadjectivethe nounelement ipossesso(at least)for usingDworak a

a.

and bs and pseemsn Persr is thereflecteMcGinnnd Mc

jaaleb-tarin

inding facts seemossessor to be postto take scope over

ian, can occur in acomplement of thed in the c-commandis’s (2013) BindingGinnis (2012). The

ketaab

to provide furthe-nominal in a righand bind the el

possessive consanaphor in a righrelations of thePreservation, orright-adjunction a

vs.

r argumt branchement cltructiont-branchbase ordiginally pnalysis

ketaab-e

ents againsting structureoser to the nsuch as xoding structureer of elemenroposed forof the Ezafe

jaaleb-tar
interesting-SUP book book-Ez interesting-COMP
‘the most interesting book’ ‘more interesting book’
b. ajib-tarin so’aal vs. so’aal-e
anwithoun-e A, asts inTagacon

ajib-tar
strange-SUP question question-Ez strange-COMP
‘the strangest question’
‘stranger question’
The behaviour of the superlative is particularly interesting given that the derivationally related simple and comparativeforms of the adjective are post-nominal and require the Ezafe. The comparative forms are given in (33) for comparison.This derivational relation is particularly important in the context of a non-lexicalist theory which takes word formation to bepart of syntax (i.e. syntax all the way down, a la Distributed Morphology, Halle and Marantz, 1993, and subsequentauthors.) This view undermines the possibility of treating the superlatives, on the one hand, and simple and comparativeadjectives, on the other, as distinct lexical items which are merged in different positions of the structure (see, for example,Ghaniabadi, 2010). This raises a question as to how the superlative form (unlike its simple and comparative counterparts)ends up in the pre-nominal position, leading to its exclusion from the Ezafe domain.

Cinque (2010, 31--32) discusses superlatives as one of the cases where the strict order of direct modificationadjectives can be violated. Thus, for example, while a shape adjective precedes a color adjective in the unmarked order(e.g. a long white plane), this order is reversed with the superlative form of the adjective (e.g. the whitest long plane). Morestrikingly, this reversal of order is observed even with the more rigidly ordered non-intersective adjectives (cf. anoccasional hard worker vs. the hardest occasional worker). Cinque (2010:32) suggests that the superlative morpheme ismerged high in the functional structure of the DP and the relevant adjective is attracted to this high position (see alsoMatushansky, 2008, among others).

Extending Cinque’s analysis to Persian, I propose that in the formation of the superlative, the adjective is attracted tothe superlative morpheme which is itself high in the DP structure. As such, the superlative adjective, like other highelements such as the demonstrative or the numeral lies outside of the Ezafe domain, i.e. it is consistently pre-nominal andlacks the Ezafe marking.

account such as Larson and Yamakido (2008) which takes theno movement involved. In Persian, the element further away fromand the noun itself. Thus, for example, xod ‘self ’, an anaphoric

li self-Ez Ali ‘Ali himself (Lit. Ali’s self)’. This is unexpected if theit cannot bind the anaphor in this configuration. This possibility isthe account proposed here (xodi-e Ali ti), which can be accountedlog, and applied to binding facts at the clausal level in Persian instruction can also capture these facts (see Ghomeshi, 1997).


The behavior of ordinals in Persian lends further support to the analysis of the superlative. There are two ways toexpress an ordinal phrase in Persian, as shown in (34).

(34)

63 In theplural and64 For a65 Ghaninstance,syntacticthese conwhich trepave the

a.

plural,definitreatmiabadi (he notecompostructiats Ghaway fo

moshkel-e

the colloquial dete features (seeent of the [N Ad2010) notes furts that while in syunds are more cons as incorporaniabadi’s ‘‘syntar a better explan

chaarom
problem-Ez fourth ‘the fourth problem’
b.
chaarom-in moshkel fourth-in problem ‘the fourth problem’
Once again, we see the correlation between the order of noun andmodifier and the presence of Ezafe. In (34a), where theordinal is post-nominal, there is an Ezafe marker. Interestingly, the same morpheme -in used with the superlative is usedin (34b) leading to the same effect observed with the superlative: pre-nominal ordinal and no Ezafe marker. This providesfurther support for breaking down the superlative marker -tarin into the comparative marker -tar and -in, with -in having theproperty of merging in a high position, and attracting the correspondingmodifier, either the comparative or the ordinal. Theattracted modifier will consequently fall outside the domain of the Ezafe construction.

In the following section, we will consider a case of post-nominal modifiers appearing without the Ezafe marker in thecontext of the colloquial definite marker -e.

5.2. Modifying nouns in the context of the colloquial definite marker

It was noted in section 3.3 that there are some cases of post-nominal modifiers used without the Ezafe marker in thecontext of the Persian colloquial definite marker -e.63 It should be noted that while this suffix appears homophonous withthe Ezafe marker, it is distinct from it as the definite marker is always stressed and appears as -he after words ending in-e and as -(h)a when it is not word-final, e.g. if followed by the accusative marker -raa. The Ezafe does not exhibit thesephonological properties. The relevant point for the purposes of this paper is that in the context of this definite marker, withsome restrictions to be discussed below, a post-nominal modifier can appear without the Ezafe in apparent violation of thegeneralizations made in (11) above. An example of the colloquial definite marker without a modifier is given in (35a) andwith a modifier in (35b), repeated from (10) above. I will hereafter refer to the construction exemplified in (35b) asModification Without Ezafe (or MWE for short).

(35)
a. ketaab-e book-def. ‘the book’
b.
ketaab gerun-e book expensive-def. ‘the expensive book’
Ghaniabadi (2010) calls examples such as (35b) syntactic compounds and treats them on a par with the other cases ofcompounds, both the right-headed compounds (e.g. ketaab-xune book-house ‘library’) and the left-headed compounds orwhat I referred to as historical compounds (e.g. sib-zaminiapple-ground ‘potato’), taking themall to be cases of incorporatedacategorial Roots (in the sense of Distributed Morphology, see also Harely, 2008) formed through head adjunction.64

Meanwhile, this approach leaves unexplained the clear distributional difference between the compounds discussed insection 3.3 and the post-nominal modification without the Ezafe. While the former can be used in any kind of nominalconstruction, the latter is limited to the context of the colloquial definitemarker, exemplified in (35b) above. In (36a-b), we seeexamples of the right-headed and left-headed compounds used with other post-nominal elements, and (36c) shows thatpost-nominal modification cannot be used without the Ezafe in the absence of the definite colloquial marker.65

finite marker is realized as the default Persian plural marker -haa, which in these cases can be taken to carry bothGhomeshi, 2003; Ghaniabadi, 2010; Hamedani, 2011).j] complex in these contexts as a compound, see also Samiian (1983) and Ghomeshi (2003).her differences between what he calls syntactic compounds exemplified in (35b) and the other compounds. Forntactic compounds up to two modifiers can be incorporated, in the other compounds there is a limit of one. Also,ompositional and transparent in meaning than the other compounds. Meanwhile, given his uniform treatment ofted acategorial Roots, it is not clear what these differences would follow from. The analysis proposed belowctic compounds’’ (our MWE) as a result of syntactic head-movement (s opposed to true compounding) mightation of these differences.

(36) a. ketaab-xune-ye ghadimi-ye zibaa


*

66 I findungrammungramm(like me)

*

*

*

*

the useatical.atical., one f

of the intensifier xeI have consulted seveThe other three sharound it marginal and

yli ‘very’ in (39)n other native sed my judgmenone outright un

book-house-Ez
old-Ez beautiful ‘the beautiful old library’
b.
sib-zamini-ye zard-e xoshmaze apple-ground-Ez yellow-Ez tasty ‘the tasty yellow potato’
c.
ketaab-gerun-e jaaleb
finepet agra

book-expensive-Ez
interesting ‘the interesting expensive book’
The above facts clearly show that treating MWE construction on a par with other types of compounds in Persian is onthe wrong track. Below, I propose an analysis of the colloquial definite marker construction which accounts for theappearance of the post-nominal adjective without the Ezafe. Before that, we should look at some other properties of theMWE construction. The first point to note is that MWE can only occur with bare adjectives and not with phrasal modifiers,as shown in (37) (cf. (35b), see also Ghaniabadi, 2010).

(37)
ketab ru-(ye) miz-e book on-Ez table-def. ‘the book on the table’
In addition, the modification of the noun can extend to two adjectives (though with some restrictions that are notvery well understood) but never to more than two adjectives (see also Ghaniabadi, 2010, cf. Ghomeshi, 2003).An example with two adjectives is given in (38a). (38b) shows that adding yet another adjective would lead toungrammaticality.

(38)
a. ketaab siyaah bozorg-e book black big-Ez ‘the big black book’
b.
ketaab siyaah bozorg gerun-e book black big expensive-def. ‘the expensive big black book’
At least for some speakers, the adjective can itself be modified with the intensifier xeyli ‘very’, but not other intensifierssuch as ziyaadi ‘too’ or besyaar ‘very’. This is shown in (39).66

(39)
ketaab (*)xeyli/?(*)ziyaadi/*besyaar bozorg-e book very/too/very big-def. ‘the very/too/very big book’
If the adjective is modified with an intensifier, then a second adjective cannot be used. This is shown in (40).

(40)
a. ketaab xeyli bozorg gerun-e book very big expensive-def. ‘the expensive very big book’
b.
ketaab bozorg xeyli gerun-e book big very expensive-e ‘the very expensive big book’
Given that the MWE construction is confined to the context of the colloquial definite marker, any attempt to understandthis construction and its properties should start with an analysis of the colloquial definite marker. I take the locus of the

and the use of ziyaadi ‘too’ somewhat degraded. An anonymous reviewer finds bothakers. Of those seven, four agreed with the anonymous reviewer and found both formsbout xeyli, but showed more variation with ziyaadi: one found it somewhat degradedmmatical. I leave a closer examination of these variations for future research.


colloquial definite marker --e to be D (see also Ghomeshi, 2008).67 I further take the head noun to move to D, as shown in(41), leading to the realization of the definite marker on the noun.68

(41)

67 Thehere. Whthe demclassif. bdefinite m68 This(p. 70).69 In orspeakerintensifieconstrucdue to tpropertie

[TD$INLINE]

DP

D N-e ketaab

‘book’

The question is what happens if there are modifiers between N and D. In section 4.3, we took bare and phrasaladjectives to occupy specifier positions of functional projections above the noun (XP, YP, etc. in (31) above). I nowsuggest that bare adjectives have the option of being merged in the head position of these functional projections. Whenthey do, as expected by the Head Movement Constraint (Travis, 1984 and subsequent authors), the movement of N to Dwill go through the intervening heads, which house the bare adjectives, and as a result the bare adjectives will be part ofthe complex head formed in D. This process is shown for the example in (35b) in (42) below, where the merge position ofthe elements is shown between angled brackets.

(42)

[TD$INLINE]

DP

D YP

Y D <Y> <N>-e gerun ketaab

N Yketaab gerun‘book’ ‘expensive’

Let us see how wemight be able to understand the properties of the MWE construction discussed above in the contextof this proposal. The first point to make is that the absence of the Ezafe marker in this context is due to the fact that weare not dealing with the type of roll-up phrasal movement discussed in section 4.3. When the modifiers (includingbare adjectives) merge as phrases in the specifiers of the intervening functional projections, this will lead to the Ezafeconstruction, with iterative Ezafe vowels as a reflex of the phrasal roll-up movement. The Ezafe construction, unlike theMWE construction, has no limit on the number or type of modifiers it allows. Meanwhile, theMWE construction is the resultof head movement, which does not lead to the appearance of the Ezafe vowel. This construction also has a set ofproperties which have been discussed in this section. Given that the MWE construction is the result of head-movement, itis incompatible with phrasal modifiers (see (37) above). The other properties shown in (38)--(40) are related to restrictionson the size of the complex head formed in D (see (42)). Thus, for example, there is a limit of two bare adjectives that can beused in this construction (see (38)). Meanwhile, if the adjective is itself modified by an intensifier, then further modificationis blocked, suggesting that perhaps there is a limit of two on the number of words that can adjoin to the head noun (see(39)--(40)).69 Finally, it should be noted that in theMWE construction, primary stress, which is word-final in Persian, always

exact position of this definite head may need further investigation, in particular in the context of the head movement analysis proposedile the definite marker is very restricted with respect to the elements it can appear with, it is absolutely fine (and possibly preferred) withonstrative (in ketaab-e this book-def.) and its plural allomorph -haa (see footnote 63) can appear with a numeral (do taa ketaab-aa twoook-def.pl ‘the two books’). If the demonstrative and numeral are taken to occupy head positions as suggested in this paper, then thearker should be in a lower position to get the right result. I leave a closer examination of the exact locus of this head for future research.

is also in line with Ghomeshi’s (2003) suggestion that the colloquial definite marker -e takes an N0 and turns it into D0: [N0 + eDEF] ! D0

der to account for the possibility of modifying the bare adjective with an intensifier in the MWE construction (see footnote 66 regardingvariation), we will have to assume that the intensifier is head-adjoined to the adjective (see also Ghaniabadi, 2010). The marginality of ther ziyaadi ‘too’ (see (39)) may be related to the fact that it has three syllables. The impossibility of the intensifier besyaar ‘very’ in thistion may be due to a register mismatch. While besyaar is only used in the formal register, the MWE construction is only used colloquially,he presence of the colloquial definite marker. These suggestions lead to many questions with respect to the MWE construction, itss and restrictions which need to be further investigated.


falls on the definite marker. This can follow from the fact that this construction involves the formation of a complex wordwith the expected word-final stress.70

In this section we have looked at a construction which involves post-nominal bare adjectives without the Ezafe marker,what I calledModificationWithout Ezafe (MWE). I showed that it cannot be treated on a par with other cases of compoundsin Persian and discussed the properties of this construction at some length. I then proposed a head-movement analysis ofthis construction which introduces a new way of understanding this construction and its constraints. Many issues remainwith respect to the exact nature of the properties discussed in this section and how they can best be modeled.

6. Conclusion

In this paper, we have established that there is a near-perfect correlation between the order of the noun and othernominal elements and the presence of the Ezafe marker in Persian, with the noun clearly demarcating the distribution ofthe Ezafe marker: no Ezafe on elements surfacing before the noun and an Ezafe for every element following it. It wasargued that these facts are best captured in a system which takes the merge position of the noun in the DP to be final andthe surface order derived via phrasal movement in a roll-up fashion. In this system, modifiers involved in the Ezafeconstruction, are uniformly merged in the specifiers of functional projections above the NP regardless of whether they arebare or phrasal. Under this view, the Ezafe is seen as a reflex of the roll-up movement. Additional empirical support for thisanalysis was provided. In order to have a better understanding of the Ezafe construction and further evaluate theproposals made in this paper, a closer examination of other languages in the Iranian family and beyond is needed.

Acknowledgements

I would like thank the audiences at Tehran University, ‘‘Tales of the Missing Link’’Workshop at NYU and the CLA 2012annual meeting for their helpful questions and comments. I am particularly grateful to the two anonymous reviewers ofLingua for extensive comments on my submission.

References

Aoun, J., Benmamoun, E., Sportiche, D., 1994. Agreement word order, and conjunction in some varieties of Arabic. Linguist. Inq. 25, 195--220.Bassols-Codina, S., 1992. La phrase relative en kurde central. (Ph.D. Dissertation). Université de Paris, pp. 3.Belleti, A., 2004. Structures and Beyond -- The Cartography of Syntactic Structures, vol. 3. Oxford University Press, Oxford.Cheng, L., 1997. On the Typology of Wh-Questions. Garland, New York.Chomsky, N., 1995. The Minimalist Program. MIT Press, Cambridge, MA.Cinque, G., 1994. On the evidence for partial Nmovement in theRomanceDP. In: Cinque, G., Koster, J., Pollock, J.-Y., Rizzi, L., Zanuttini, R. (Eds.),

Paths Towards Universal Grammar: Studies in Honor of Richard S. Kayne. Georgetown University Press, Washington, DC, pp. 85--110.Cinque, G., 1996. The ‘antisymmetric’ programme: theoretical and typological implications. J. Linguist. 32, 447--464.Cinque, G., 1999. Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford University Press, Oxford.Cinque, G., 2002. The Functional Structure of DP and IP -- The Cartography of Syntactic Structures, vol. 1. Oxford University Press, Oxford.Cinque, G., 2005. Deriving Greenberg’s universal 20 and its exceptions. Linguist. Inq. 36, 315--332.Cinque, G., 2010. The Syntax of Adjectives: A Comparative Study. LI Monograph 57. MIT Press, Cambridge.Dikken, M.den., 2006. Relators and Linkers: The Syntax of Predication, Predicate Inversion, and Copulas LI Monograph 47. MIT Press,

Cambridge.Dworak, C., McGinnis, M., 2012. Binding in Farsi ditransitives. In: Talk Presented at the Annual Meeting of the Canadian Linguistic Association

(CLA). Wilfrid Laurier University, Waterloo.Ernst, T., 2001. The Syntax of Adjuncts. Cambridge University Press, Cambridge.Ghaniabadi, S., 2010. The Empty Noun Construction in Persian. (Ph.D. Dissertation). University of Manitoba.Ghomeshi, J., 1996. Projection and Inflection: A Study of Persian Phrase Structure. (Ph.D. Dissertation). University of Toronto.Ghomeshi, J., 1997. Non-projecting nouns and the Ezafe construction in Persian. Nat. Lang. Linguist. Theory 15, 729--788.Ghomeshi, J., 2003. Plural marking, indefiniteness, and the noun phrase. Stud. Linguist. 57 (2), 47--74.Ghomeshi, J., 2008. Markedness and bare nouns in Persian. In: Karimi, S., Samiian, V., Stilo, D. (Eds.), Aspects of Iranian Linguistics. Cambridge

Scholars Publishing, Newcastle upon Tyne, pp. 85--112.Halle, M., Marantz, A., 1993. Distributed morphology and the pieces of inflection. In: Hale, K., Keyser, S.J. (Eds.), The View from Building 20. MIT

Press, Cambridge, pp. 111--176.Hamedani, L., 2011. The Function of Number in Persian. (Ph.D. Dissertation). University of Ottawa.Harley, H., 2008. Compounding in Distributed Morphology (lingBuzz/000602).

70 While primary stress is on the definite marker, there still seems to be noticeable secondary stress on the last syllable of the noun (andadjectives) which make up this construction. In the Ezafe construction, main stress still falls on the last syllable of the last word in the wholephrase. Meanwhile, the stress on the final syllable of the head noun and intermediate adjectives is noticeably greater than their counterparts in theMWE construction, pointing to a clear difference between these constructions and their syntactic structure. These claims need to be furtherevaluated with phonetic experimentation, an issue I leave for future research.

http://refhub.elsevier.com/S0024-3841(14)00166-1/sbref0005




























Holmberg, A., Odden, D., 2005. The Izafe and NP structure in Hawrami. Durham Working Papers in Linguistics 10, pp. 77--94.Kahnemuyipour, A., 2000. Persian Ezafe construction revisited: evidence for modifier phrase. In: John, T., Jensen, J.T., van Herk, G. (Eds.),

Cahiers Linguistique d’Ottawa, Proceedings of the 2000 Annual Conference of the Canadian Linguistic Association, pp. 173--185.Kahnemuyipour, A., 2001. On Wh-questions in Persian. Can. J. Linguist. 46 (1/2), 41--61.Kahnemuyipour, A., Massam, D., 2006. Patterns of phrasal movement: the Niuean DP. In: Gaertner, H.-M., Law, P., Sabel, J. (Eds.), Clause

Structure and Adjuncts in Austronesian Languages. Mouton de Gruyter, pp. 125--150.Karimi, S., Brame, M., 1986. A generalization concerning the Ezafe construction in Persian. In: Paper Presented at the Annual Conference of the

Western Conference of Linguistics, Canada.Karimi, S., Brame, M., 2013. A generalization concerning the EZAFE construction in Persian. Linguist. Anal. 38, 111--143.Kayne, R.S., 1994. The Antisymmetry of Syntax. MIT Press, Cambridge.Kayne, R.S., 2000. A note on prepositions, complementizers and word order universals. In: Kayne, R.S. (Ed.), Parameters and Universals. Oxford

University Press, New York, pp. 314--326.Kayne, R.S., 2004. Prepositions as probes. In: Belletti, A. (Ed.), Structure and Beyond: The Cartography of Syntactic Structures, vol. 3. Oxford

University Press, New York, pp. 192--212.Larson, R., 2009. Chinese as a Reverse Ezafe Language. Yuyanxue Luncong (Journal of Linguistics) 39. Peking University, Beijing, pp. 30--85.Larson, R., Yamakido, H., 2006. Zazaki ‘‘Double Ezafe’’ as double case-marking. In: Paper Presented at the Linguistics Society of America Annual

Meeting, Aluquerque, NM.Larson, R., Yamakido, H., 2008. Ezafe and the deep position of nominal modifiers. In: McNally, L., Kennedy, C. (Eds.), Adjectives and Adverbs:

Syntax, Semantics and Discourse. Oxford University Press, Oxford, pp. 43--70.Lazard, G., 1992. A Grammar of Contemporary Persian. Mazda Publishers, Costa Mesa, CA.Marantz, A., 1997. No escape from syntax: don’t try morphological analysis in the privacy of your own lexicon. In: Dimitriadis, A., Siegel, L., Surek-

Clark, C., Williams, A. (Eds.), Proceedings of the 21st Annual Penn Linguistics Colloquium U Penn Working Papers in Linguistics 4.2. PennLinguistics Club, Philadelphia, pp. 201--225.

Matushansky, O., 2008. On the attributive nature of superlatives. Syntax 11, 26--90.McGinnis, M., 2013. Irreversible binding. In: Kan, S., Moore-Cantwell, C., Staubs, R. (Eds.), Proceedings of NELS 40 November 2009 GSLA

UMass Amherst, vol. 2. pp. 81--92.Merchant, J., 2001. The Syntax of Silence: Sluicing, Islands and the Theory of Ellipsis. Oxford University Press, Oxford.Moyne, J., Carden, G., 1974. Subject reduplication in Persian. Linguist. Inq. 5 (2), 205--249.Pearce, E., 2002. DP structure and DP movement in Maori. In: Paper Presented at COOL5. University of Canberra, Australia.Rackowski, A., Travis, L., 2000. V-initial languages: X or XP movement and adverb placement. In: Carnie, A., Guilfoyle, E. (Eds.), The Syntax of

Verb Initial Languages. Oxford University Press, Oxford, pp. 117--142.Ritter, E., 1991. Two functional categories in noun phrases: evidence from Modern Hebrew. Syntax and Semantics 25, 37--62.Ritter, E., 1992. Cross-linguistic evidence for number phrase. Can. J. Linguist. 37 (2), 197--218.Rizzi, L., 2003. The Structure of CP and IP -- The Cartography of Syntactic Structures, vol. 2. Oxford University Press, Oxford.Samiian, V., 1983. Origins of Phrasal Categories in Persian, an X-bar Analysis. (Ph.D. Dissertation). UCLA.Samiian, V., 1994. The Ezafe construction: some implications for the theory of X-bar syntax. In: Marashi, M. (Ed.), Persian Studies in

North America. Iranbooks, Betheda, MD, pp. 17--41.Samvelian, P., 2007. A (phrasal) affix analysis of the Persian Ezafe. J. Linguist. 43, 605--645.Samvelian, P., 2008. The Ezafe as a head-marking inflectional affix: evidence fromPersian and Kurmanji Kurdish. In: Karimi, S., Samiian, V., Stilo,

D. (Eds.), Aspects of Iranian Linguistics. Cambridge Scholars Publishing, Newcastle upon Tyne, pp. 339--361.Shlonsky, U., 2004. The form of Semitic noun phrases. Lingua 114 (12), 1465--1526.Shlonsky, U., 2010. The cartographic enterprise in syntax. Lang. Linguist. Compass 4/6, 417--429.Skjærvø, P.O., 2009. Middle West Iranian. In: Windfuhr, G. (Ed.), The Iranian Languages. Routledge, New York, pp. 196--278.Travis, L., 1984. Parameters and Effects of Word Order Variation. (Ph.D. Dissertation). MIT.














































revisiting the persian ezafe construction: a roll-up movement analysis

Documents