You are on page 1of 31

This article was downloaded by: [TÜBİTAK EKUAL]

On: 1 April 2011


Access details: Access Details: [subscription number 772815468]
Publisher Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-
41 Mortimer Street, London W1T 3JH, UK

Iranian Studies
Publication details, including instructions for authors and subscription information:
http://www.informaworld.com/smpp/title~content=t713427941

Syntax, Morphology, and Semantics of Ezafe


Parviz Parsafar

Online publication date: 14 December 2010

To cite this Article Parsafar, Parviz(2010) 'Syntax, Morphology, and Semantics of Ezafe', Iranian Studies, 43: 5, 637 — 666
To link to this Article: DOI: 10.1080/00210862.2010.518029
URL: http://dx.doi.org/10.1080/00210862.2010.518029

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf

This article may be used for research, teaching and private study purposes. Any substantial or
systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or
distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses
should be independently verified with primary sources. The publisher shall not be liable for any loss,
actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly
or indirectly in connection with or arising out of the use of this material.
Iranian Studies, volume 43, number 5, December 2010

Parviz Parsafar

Syntax, Morphology, and Semantics of Ezafe

Although ezafe has been studied by many scholars for many years, it does not yet have a
transparent grammatical status. Grammarians have regarded ezafe as a polysemous
“word” carrying over ten different “meanings/functions.” After a brief review of the
previous treatments of ezafe, this paper will present a syntactic analysis, followed by a
morphological description and a semantic analysis of this ubiquitous morpheme. It will
also compare the distributional properties of other relevant bound morphemes with those
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

of the ezafe. It will finally conclude that ezafe is a dummy clitic-like morpheme which is
semantically void, while syntactically it functions as an “associative marker” which
subordinates its [+N] host, on the left, to its following complements.

Ezafe, which literally means “annexation” or “addition” and is traditionally


known as a “Genitive” marker, is an indispensable element inside any noun
phrase comprising a head modified by at least one non-clausal modifier and/or
complement. That is, any study of Persian involving noun phrases (NPs) in
subject position or predicate position, whether followed by light verbs or
thematic verbs, is bound to encounter ezafe in numerous example sentences.
The purpose of this paper, which is a very condensed version of chapter one of
Parsafar’s doctoral dissertation,1 is to present a clear analysis of ezafe. The first
section is a brief review of previous works. The main part then starts with the
syntactic aspects of this morpheme and is followed by a morphological and
finally a semantic analysis.

Previous Treatments of Ezafe

This section presents a very brief description of some scholars’ accounts of ezafe.
Homayunfarrokh maintains that ezafe is a “case” which determines the semantic
relationship between two words or groups of words both phonologically and

Parviz Parsafar is Professor of ESL & Linguistics, Yuba College, Marysville, CA, USA
(pparsafa@ yccd.edu).
Note that the findings in this paper are part of research that took place between 1990 and 1996. I
am greatly indebted to my advisor, Professor Laurence Horn (Yale), and Professors Stanley Insler
(Yale) and Gernot Windfuhr (University of Michigan). However, all the errors and lapses are mine
alone.
1
Parviz Parsafar, “Spatial Prepositions in Modern Persian” (PhD diss., Yale University, 1996).
ISSN 0021-0862 print/ISSN 1475-4819 online/10/050637 –30
#2010 The International Society for Iranian Studies
DOI 10.1080/00210862.2010.518029
638 Parsafar

semantically. He further adds that “the sign of ezafe is an [e] which is added to the
first word” as in /kolah-e mæn/“my hat.”2
He considers ezafe a polysemous item and divides all the ezafe-bearing
constructions into six groups: possessive, particularizing, descriptive, metapho-
rical, definitional, and analogical.
Phillott’s classification of ezafe constructions consists of many “different
types,”3 of which those that have not been mentioned by Homayunfarrokh are
ezafe of profession, and territorial, epithetical, patronymic, partitive, and super-
lative ezafe.
Lazard maintains that “when a substantive is accompanied by a modifier, it is
followed by the ‘enclitic particle’ -e.” He further adds that “ezafe does not indicate
anything regarding the nature of the relation which holds between the modifier
and the modified nominal.” Then, he classifies all the “relations which are
represented by ezafe” into five major groups.4
Palmer’s analysis is different from those mentioned above. He derives all his
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

ezafe constructions from underlying (deep structure) relative clauses. There are
two “subsets” of ezafe constructions. Those in the first group, which do not
 “conjugated verb” in their deep structures, will carry either the
contain any
verb /da t-/ “have” or the verb /bud-/ “be” in their “sentence paraphrases.”5
The second subset of ezafe constructions consists of those containing infinitives.
The introduction of infinitives into the constructions is also accounted for by
transformational rules.
After criticizing the traditional treatments of ezafe, Sami’ian starts with a defi-
nition of ezafe and its functions. She maintains that ezafe “which literally means
“addition” refers to the unstressed morpheme [-e]” appearing between the
head and certain modifiers and complements in noun phrases, in adjectival
phrases, and in prepositional phrases.6 Furthermore, she believes that “the
Ezafe morpheme is not base generated but transformationally inserted before
each phrasal complement” inside the noun phrase. She then proposes that “all
Ezafe bearing phrasal complements are generated” under N  (N-bar).
Karimi refers to “the ezafe particle -e” very briefly and maintains that in the
ezafe constructions, this particle “structurally relates the embedded phrase to
the head noun.”7 She further adds that ezafe is not a case assigner, but it does
“transfer the case of the head noun to its complement(s).”

2
A. R. Homayunfarrokh, Dæstur-e Jame’-e Zæban-e Farsi [A Comprehensive Grammar of the
Persian Language], 2nd ed. (Tehran, 1339/1960).
3
D. C. Phillot, Higher Persian Grammar (Calcutta, 1919).
4
Gilbert Lazard, Grammaire du Persian Contemporain (Paris, 1957).
5
A. Palmer, “The Ezafe Construction in Modern Standard Persian” (PhD diss., University of
Michigan, 1971).
6
Vida Sami’ian, “Structure of Phrasal Categories in Persian: an X-Bar Analysis” (PhD diss.,
UCLA, 1983).
7
Simin Karimi, “Aspects of Persian Syntax, Specificity, and the Theory of Grammar” (PhD
diss., University of Washington, 1989), 83 –84 and 116.
Syntax, Morphology, and Semantics of Ezafe 639

The Syntax of Ezafe

The distribution of ezafe. This subsection briefly examines the distributional restric-
tions. Ezafe is unable to begin or end any tensed clause or to occur more than
once consecutively. In general, the construction in which ezafe occurs carries at
least two elements, each on one side of the ezafe. For instance, in (1) only the
last example is acceptable:

(1) a. / e-saye-ye deræxtan /
E-shade-E trees

b. / saye-ye deræxtan-e /
-E -E

c. / saye-ye e-deræxtan/
-E E-
d. / saye-ye deræxtan / “the shade(s) of the trees”
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

-E
Furthermore, ezafe cannot be immediately adjacent to a finite verb, as in (2 – 3):

(2) a. / dærs-e xand /
lesson-E read-[3sg]
b. ∗ / xand-e dærs /
Cf. c. / dærs xand / “He/she studied.”

(3) / mirævæd-e mædrese /
goes-E school

In infinitival constructions of periphrastic verbs (PVs), ezafe cannot precede the


verbal component which is the last element of the PV:

(4) a.∗ / jævab-e dadæn /


answer-E to give
Cf. b./jævab dadæn / “to (give an) answer”

However, simple infinitives can be used as modifiers, in which case ezafe


obligatorily precedes the infinitive (by attaching, as usual, to the head of the
NP). In (5), the subject NP is /sa’æt-e ræftæn/ “time of leaving” where the infi-
nitive / ræftæn/ is modifying /sa’æt/ “time.”

(5) [ sa’æt-e ræftæn ] næzdik æst


NP
time-E to go near is
“[The time of departure ] is approaching.”
640 Parsafar

That /ræftæn/ is a nominal is illustrated in (6) where it functions as the subject,


and in (7) where it is the object of the preposition.

(6 ) / ræftæn fayede-‘i nædaræd/ “It’s useless to go.”


to go use-a 
not-has
(7) /æz ræftæn xæste ode-æm/
from to go tired become-1sg
“I’m tired of going/leaving.”

What is significant to note is that there seems to be no syntactic restriction on


the post-infinitival occurrence of ezafe in either case, i.e. whether the infinitive is
the verbal element of a PV or a simple verb. In other words, ezafe can attach to an
infinitive on its left whether the infinitive is used as a modifier of a preceding con-
stituent or it is the verbal element in a PV, as in (8) and (9). Observe that there are
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

no syntactic/semantic differences between these sentences and their colloquial


versions.

(8) [ æz [jævab dadæn]-e ‘u ] ] xo æm næyamæd
PP NP
from answer to give-E him please-1sg not come-[3sg]
“His answering didn’t please me.”
“[lit.] From his answer-giving pleasure did not occur to me.”
(9) [ [ [sa’æt]-e ræftæn]-e ma] æst
NP NP
time-E to go-E us is
“It’s our departure time.”

As these examples suggest, Persian infinitives function only as nominals, hence


deverbal nouns. Furthermore, note that the head of Persian NPs is always the noun
(or the QP which functions as a noun) on the left of the ezafe. For example, in (5)
and (9), /sa’æt/ “time” is the head of the NP, and in (8) /jævab dadæn/ “answer-
giving” is the head. Thus far, then, it seems tentatively reasonable to conclude
that an implied corollary to (5– 9) is that adjacency to nominals and nominalized
elements is one possible environment for the occurrence of ezafe.
Categorical sensitivity of ezafe. Chomsky suggests the possibility “that the
categories noun, verb, adjective are the reflection of a deeper feature structure,
each being a combination of features of a more abstract sort.”8 He further
adds that in “this way, the various relations among these categories might be
expressible.”

8
N. Chomsky, “Remarks on Nominalization,” in Readings in English Transformational Grammar,
ed. by R. A. Jacobs and Peter S. Rosenbaum (Cambridge, MA, 1970), 184 –221.
Syntax, Morphology, and Semantics of Ezafe 641

Stowell mentions that Chomsky “proposed an explicit theory of syntactic


features from which the major lexical categories should be derived.”9 The fea-
tures list in (10) shows a summary of that system:

(10) [+N] (nouns, adjectives)


[-N] (verbs, prepositions)
[+V] (verbs, adjectives)
[-V] (prepositions, nouns)

Note that Chomsky’s feature system predicts that the “unnatural” syntactic
classes are (nouns, verbs) and (adjectives, prepositions). Stowell mentions that
Van Riemsdijk has raised some doubts about Chomsky’s natural classes. He
has pointed out that this feature system “implicitly assumes that no rule of
grammar should be able to refer to a class of three of the major categories to
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

the exclusion of a fourth.” Van Riemsdijk’s skepticism is based on his belief


that “there is a rule in the grammar of Dutch which incorporates adjectives,
nouns, and prepositions into the verbal complex, although it does not apply to
verbs themselves.”10
However, it will be shown that Chomsky’s feature system adequately covers
the functions of the Persian ezafe, since, contrary to Sami’ian’s arguments, ezafe
merely applies to [+N] lexical and phrasal categories. If this proves to hold,
then the existence of such specific parochial rules which do not apply to all the
categories, but to subsets of the given categories, will serve as further evidence
to support Chomsky’s feature system.

Ezafe in adjectival phrases. Ezafe can attach to adjectives and past participles used
as heads. In predicates with auxiliary verbs /budæn/ “to be” and / odæn/ “to
become,” ezafe can attach to the head of certain adjectival phrases as in (11):

(11) [mo taq-e didar-etan] hæstæm
AP
eager-E seeing-E-your am
“I’m eager to see you.”

There is, however, a restriction on such predicative adjectives. It seems that


the head (of these APs) obligatorily subcategorizes for an NP and the resulting
AP together with the auxiliary makes a periphrastic verb. Consequently, when
these APs are used in the attributive, rather than predicative, position, they
will result in ill-formed NPs as in the following:

9
Timothy A. Stowell, “Origins of Phrase Structure” (PhD diss., MIT, 1981), 21.
10
Stowell, “Origins of Phrase Structure,” 55. See also H. C. Van Riemsdijk, “The Case of
German Adjectives,” in Linguistic Categories 1, ed. by F. Heny and B. Richards (Dordrecht,
1983), 223 – 252.
642 Parsafar


(12) a.∗ / mærd-e mo taq-e didar-etan/
man-E eager-E seeing-E-your 
b.∗ / pesær-e motevæjje-ye hærekat-e xode /
boy-E attentive-E actions-E himself
Now in order to capture both the conclusion in the first part of this section and
the one in this subsection, we can, still tentatively, hold that ezafe can attach to
[+N] categories. Later, it will be illustrated that indeed ezafe attaches to the
preceding [+N] category both syntactically and phonologically.
As for past participles, they can be used both predicatively and attributively. In
(13), ezafe is attached to the head adjective /suxte/ “burned.” However, as will be
shown later, the second ezafe in (14) is attached to the NP /pænjere-ye ekæste/
“broken window,” rather than to the adjective / ekæste/ “broken”:
(13) [ [suxte]-ye aftab ] “weather-beaten”, “[lit.] burned by sunlight”
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

AP
burned-E sunshine 
(14) [ [ [pænjere]-ye ekæste]-ye ma]
NP NP
window-E broken-E we
“our broken window”

In any event, these examples, too, are indicative of the fact that ezafe attaches to
nominal or nominalized elements.
Ezafe in noun phrase specifiers. Of the Persian NP specifiers, i.e. demonstratives
(including interrogatives), quantifiers, superlative adjectives, numerals, and classi-
fiers, only superlatives, ordinals, and those quantifiers which occur in partitive
phrases take ezafe obligatorily.
Quantifiers. First, consider sentences (15–16) which show the use of quantifiers.
The (b) examples are ill-formed because the head quantifiers do not carry ezafe.

(15) a. /hæme-ye doxtærha / “all the girls”


all-E girls
b.∗ / hæme- doxtærha/
(16) a. /nesf-e çayi / “half of the tea”
half-E tea
b.∗ /nesf çayi/

Note that these quantifier phrases, i.e. the (a) examples, can be used as subjects
and objects, as in (17 –18), which means that they are NPs.
(17) a. /hæme-ye setareha be ma mixændænd/
All-E stars to us laugh-[3pl]
“All the stars laugh at us.”
Syntax, Morphology, and Semantics of Ezafe 643

b. /hæme-ye golhar-ra bu kærdæm /


all-E flowers-SM11 smell did-[1sg]
“I smelled all the flowers.”
(18) a. nesf-e mærdom xub budænd
half-E people nice were-[3pl]
“Half of the people were nice.”
b. nesf-e baqbanha-ra æz kar bi-kar kærdænd
half-E gardeners-SM from work without work did-[3pl]
“They fired half of the gardeners.”
Examples (19) and (20), in which /hæme/ “all” functions as the object and the
subject, respectively, further show that the initial elements, i.e. the quantifiers,
in (15 – 18) are also nominalized.
(19) /hæmæ-ro bebær /
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

all-SM take-[2sg]

“Take (them) all!”
(20) / hæmæ- rixte ru zæmin / “It is all spilled on the ground.”
all-it spilled on ground


Superlatives and numerals. Comparatives are made by the suffix /-tær/ added to
the simple adjectives, as in /qæ æng-tær/ “prettier” while superlatives
 are
formed by adding the suffix /-in/ to the comparative base, as in /qæ æng-tær-
in/ “prettiest.” From among simple, comparative, and superlative adjectives,
only the last ones are obligatorily used pre-nominally, as in (21) and (22). The
other two usually follow the noun they modify.

(21) /bozorg-tær-in deræxt / “the biggest tree”


biggest tree
(22) /bozorg-tær-in deræxtha / “the biggest trees”

What the specifier superlative in (21 – 22) has in common with demonstratives
and interrogatives (cf. 24–26) is that it can modify plural nouns whereas speci-
fiers such as cardinal numbers cannot.
Superlatives can also be used nominally as in (23), where the superlative is
employed as the head of the subject NP. Its being a nominal here is shown by
the fact that while the complement of the head is a plural noun, the verb
agrees with the singular head (of the NP).

11
SM stands for the Specificity Marker of Objects and Topics as argued for in Parviz Parsafar,
“The Persian /ra/” (Unpublished qualifying paper presented to the faculty, Department of Linguis-
tics, Yale University, New Haven, CT, 1990). This morpheme is represented by /-ra/ which has
two other allomorphs in the colloquial language: /-ro/ and /-o/.
644 Parsafar


(23) /bozorgtærin-e bæççe-ha vared od /
biggest-E children enter became-[3sg]
“The biggest of the children entered.”

Examples (24a–c) and (25a–c) show that when demonstratives and interrogatives
are used as determiners, hence obligatorily placed in a prenominal position, they
cannot be pluralized but are able to modify plural nouns. When used as pronouns,
however, demonstratives and interrogatives can be pluralized, as in (24d) and (25d):

(24) a. /”in doxtær / “this girl”


this girl
b. /’un doxtæra / “those girls”
that

Cf. c. / ‘una doxtæra /
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

those girls
d. /’una ‘umadæn / “They/those have arrived.”
those have come-[3pl]
(25) a. / kodum nevisænde / “which writer?”
which writer
b. / kodum jængæla / “which forests?”

Cf. c. /koduma jængala /
which-PL
d. /koduma-ro migi /
which-PL-SM talking-[2sg]
“Which ones are you talking about?”

On the other hand, (26) illustrates the inability of cardinals to pluralize or to


modify plural nouns. As S. Insler (p.c.) has pointed out, this is due to the fact
that cardinals are “lexical plurals.”

(26) a. / do çeraq / “two lamps”


two lamp

b. / doha çeraqa /
two-PL lamps

c. / do çeraqa /
two lamps

As for ordinals, there are three forms. The basic form is used as a post-modifier
in non-literary contexts as in (27a). Of the two derived forms, the one which is
formed by the addition of /-i/ is also used post-nominally, as in (27b). The
third one is derived by the suffix /-in/ added to the root and is ordinarily
placed before the noun it modifies, as in (27c):
Syntax, Morphology, and Semantics of Ezafe 645

(27) a. /mah-e sevvom / “the third month”


month-E third
b. /pesær-e sevvom-i / “the third boy”
c. / sevvom-in pesær / “the third boy”

The simple ordinal form, as in (27a), is adjectival. Accordingly it is used as a post


modifier preceded by the ezafe.
The suffix /-i/ which seems to be a [+specific] nominalizer (NOM), derives a
noun from the adjective base, e.g. /sevvom-i/ as used in (27b). That /sevvomi/
is a nominal can be illustrated by the fact that it can be used alone in subject or
object position whereas the base form, i.e. /sevvom/, cannot:

(28) a. sevvom-i/∗ sevvom çi od “What happened to the third one?”
Third-NOM what became-[3sg]
b. sevvom-i-ro/∗ sevvom-ro peyda
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

kærdi
third-NOM-SM find did-[2sg]
“Did you find the third one?”
The third type of ordinal which carries /-in/ depicts two similarities with the
superlative adjectives in that (a) its only position is prenominal as in (27c),
where it is used as an adjective, and (b) it can also be followed by ezafe and a
plural noun:

(29) /sevvomin-e anha doxtæri bud bes(i)yar ziba /


third-E they girl-a was very pretty
“The third one of them was an extremely beautiful girl.”

However, the difference between this kind of ordinal and the superlative adjec-
tives is that when the former is not followed by ezafe, it cannot modify a plural
noun (cf. 22). This is due to the fact that there is no number-agreement between
numerals and nouns, which is why (30) is ill-formed:

(30) / sevvomin deraxt-ha/
third trees

Now (21), /bozorg-tær-in deræxt/ “the biggest tree,” and (27c), /sevvom-in
pesær/ “the third boy” show that superlative adjectives and /in/-carrying ordi-
nals can be used prenominally as demonstratives. Using Jackendoff’s12 classifica-
tory terminology, such specifiers can be marked [+Art] in the lexicon.
On the other hand, (23) and (29) show that superlative adjectives and /in/-
bearing ordinals can also be used in partitive phrases, hence [+Partitive]. In

12
R. Jackendoff, X-bar Syntax: A Study of Phrase Structure, Linguistic Inquiry Monograph No. 2
(Cambridge, MA, 1977).
646 Parsafar

this function, just as the [+Partitive] quantifiers in (15-18), they are obligatorily
followed by ezafe and plural nouns.
What needs to be further clarified is whether the [+Partitive] superlatives
reveal a nominal function like other [+Partitive] elements mentioned above.
Since [+Partitive] superlatives, too, can be used only in NP positions of subjects
and objects, therefore they are indeed nouns, rather than adjectives:

(31) / bozorgtærin-e anha xers-e siyah-i bud/


biggest they bear-E black-a was
“The biggest (one) of them was a black bear.”

As can be readily seen, the NP status of the subject in (31) is further justified by
the subject-predicate agreement depicted by the singular verb /bud-[3sg]/
“was.” The conclusion that can be drawn here is that the ezafe-bearing quanti-
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

fiers, superlatives and ordinals can be employed as nominals, or [+N] constitu-


ents, which further confirms the previous conclusion. Note that “adjectives are
[+N] in any case, regardless of whether they are nominalized” (L.Horn, p.c.).
Ezafe in noun phrase complements. The complements that follow head nouns in ezafe
constructions are attributive adjectives, attributive nouns, and prepositional phrases.
Relative clauses can modify head nouns which carry ezafe only if the head has at least
one complement before the relative clause (cf. the morphology section).
As was explained above, Persian noun phrases are usually head initial. This
implies that, as in French, adjectives usually follow the nouns they modify.
And the occurrence of ezafe after the head is obligatory if the head is modified
by complements. In what follows the structure of attributives and prepositional
phrases will be closely examined. 
Attributive nouns and adjectives. In (32), /na enas/ “unknown” is an adjective
following the head noun it modifies.

(32) / adæm-e na enas/ “(an) unknown person”
person-E unknown

The attributive nouns also follow their head nouns, as in (33):


(33) /ranænde-ye ‘otobus / “(the) bus driver”
driver-E bus

If a noun is modified by more than one modifier, each modifier is obligatorily


followed by an ezafe:

(34) a. /kif-e siyah-e kohne-ye italiya’i-ye mæn/


briefcase-E black-E old-E Italian-E I
“my old, black Italian briefcase”
Syntax, Morphology, and Semantics of Ezafe 647

As for the scope of modification inside the noun phrase, each adjective or noun
modifies only the first NP to its left. What this implies for an NP such as (34a) is
that it has the structure (34b), with the phrase structure rules partially listed in
(34c), indicating that when a head is modified by a series of adjectives, each
ezafe morpheme will attach to the (whole) NP preceding it rather than to individ-
ual adjectives. This claim is based on the fact that adjectives cannot be modified
by other adjectives.13

(34) b. [ [ [ [ [kif]-e siyah]-e kohne]-ye italiya’i]-ye mæn]


NP N N N N Adj Adj Adj N
c. NP N – (NP)
N N – (AP)
N N – (NP)
AP Adj
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

In (35a) and (35c), the adjective /kuçæk/ “small” modifies the head /særbaz-
 “barrack,” which is a compound noun, whereas the same adjective modifies
xane/
/ æhr/ “city” in (35b) because this is the closest noun to the left of the adjective.

(35) a. [ [ [særbaz-xane]-ye kuçæk]-e æhr]
NP soldier-house E small-E city
“the city’s small barrack” 
b. [ [særbaz-xane]-ye [ [ æhr]-e kuçæk] ]
NP NP
“the small city’s barrack” 
c. [ [særbaz-xane]-ye kuçæk]-e [ [ æhr]-e bozorg] ]
big
“the big city’s small barrack”


As can be seen, in (35c), both the head of the complex NP, which is /særbaz-
xane/ “barrack,” and the head of the modifying NP, which is / æhr/ “city” are
modified by their own adjectives.
What needs to be reiterated here is that in any given complex ezafe-bearing
NP, ezafe attaches both to the head of the NP and to all constituents in the
complement position modifying the head, except the right-most constituent.
Now, since the head of the NP is ordinarily a noun, this is also theoretically
suggestive of a recursion of multiple NPs within ezafe constructions as in (35c).

13
The only apparent exception to this claim is that the color adjectives can be modified by a
restrictive group of adjectives, a phenomenon that also exists in English, as in /qermez-e ro æn /
“light red” and /abi-ye kæmræng / “pale blue.” However, as will be seen later, an adjective carrying
an ezafe is not an anomaly.
648 Parsafar

In (36a –d), as in their corresponding English sentences, changing the pos-


itions of the nominals /xahær/ “sister” or /dust/ “friend” will lead to different
meanings. Again it can be seen that each noun modifies the entire NP to its left.

(36) a. /dust-e xo kel-e færansævi-ye xahær-e jan/
friend-E pretty-E French-E sister-E John

“John’s sister’s beautiful French friend”
b. /dust-e xahær-e xo kel-e færansævi-ye jan /

“John’s beautiful French sister’s friend”
c. /xahær-e xo kel-e færansævi-ye dust-e jan /
 beautiful French sister”
“John’s friend’s
d. /xahær-e xo kel-e dust-e færansævi-ye jan /
“John’s French friend’s beautiful sister”
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

Recall that Sami’ian claims that ezafe refutes Chomsky’s feature theory in
applying to three “natural classes” to the exclusion of the fourth. Hitherto,
however, we have seen that in all the examples (32 – 36), ezafe occurs after
either a noun or an adjective, which further confirms our claim that ezafe
occurs after merely [+N] lexical and phrasal categories.
Here, it should also be mentioned that there are linguists whose views on ezafe
rest on the other end of the spectrum when compared to that of Sami’ian’s.14 For
example, Karimi and Brame, who consider ezafe a “suffix,” have proposed a
Strong Holistic Thesis (SHT) which claims that the “ezafe marker is suffixed
to nominals and only to nominals, and it is followed by nominals and only
nominals.”15
However, this paper has already illustrated that in addition to applying to
nominal heads, ezafe can also attach to adjectival heads in the predicate position.
Parsafar has provided further evidence that shows how Karimi and Brame’s
arguments are untenable and why their SHT does not hold in Persian. After
an extensive discussion, he concludes that, indeed, “ezafe occurs after [+N] cat-
egories which include adjectives and nouns.”16
Prepositions. Prepositions and prepositional phrases have been claimed to be
able to function both as the head and as the complement of the head. On the con-
trary, it will be illustrated that “true” prepositions cannot function as heads of
NPs, i.e. they cannot take ezafe. Since Parsafar has already done a comprehensive
analysis of Persian prepositions and prepositional phrases,17 it is neither
appropriate nor plausible for me to embark on that task here. However, in the

14
Sami’ian, “Structure of Phrasal Categories in Persian,” 37–38.
15
Simin Karimi and Michael Brame, “A Generalization Concerning the Ezafe Construction in Persian”
(paper presented at the annual meeting of the Western Conference of Linguistics, Canada, 1986).
16
For a more detailed discussion of this and other sections, see Parsafar, “Spatial Prepositions in
Modern Persian,” Ch.1.
17
Parsafar, “Spatial Prepositions in Modern Persian”.
Syntax, Morphology, and Semantics of Ezafe 649

following two subsections only a very brief discussion of prepositions will be


presented which will serve our purpose.
Prepositions as heads? This paper has shown that from among all the major and
minor categories discussed so far, ezafe is attached only to [+N] categories. If it
can also occur after prepositions which are [-N, -V], then that will refute Choms-
ky’s feature system which has been proved to hold in some languages. Note that
the English Fronting Rule may be taken as counter-evidence to such a feature
system since it “applies to NP, PP, AP, but not VP” (L. Horn, p.c.).
As has been mentioned in the literature by Rubinchik, Lazard, Sami’ian, and
Karimi and Brame, among others,18 there are some prepositions that take
ezafe, others that do not, and still others that take it optionally.
At this stage, all the prepositions will be classified into two groups: (a) those
that never take ezafe, such as /æz/ “from,” /ba/ “with,” and (b) all the other pre-
positions. For ease of reference, the first group will be referred to as True Prepo-
sitions (TP) and the second group as Pseudo-prepositions (SP), the terms that
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

were first proposed by Parsafar.19


The SPs display several nominal characteristics whereas the first group lacks
such properties. In (37a) the phrase /zir-e miz / “under the table,” which
begins with one of the SPs, is the subject NP of the sentence. The same NP is
the object of the true preposition /æz/ “from” in (37b). Note that the periphrastic
verb in this sentence obligatorily subcategorizes for a PP, whereas the periphras-
tic verb (with a similar meaning) in (37c) subcategorizes for an NP. Finally, in
(37c), the phrase /zir-e miz/ is the accusative NP:

(37) a. / [ zir-e miz] kæsif-e / “Under the table is dirty.”


NP
under-E table dirty-is 
b. / [æz [ zir-e miz]] xo æm nemiyad /
PP NP
from under-E table please-[1sg] not-come-[1sg]
“Under the table doesn’t please me.”
c. /[ zir-e miz]-ra dust nædaræm /
NP
under-E table-SM like not-have-1 [sg]
“I don’t like under the table”
Thus, since /zir-e miz/ can be used as subject, object of PP, and direct object,
it is definitely an NP. Now, internal to this NP, there is the noun /miz/ “table”
which is the complement of the head of the NP, i.e. /zir/ “under.” On the other

18
Yu. A. Rubinchik, The Modern Persian Language (Moscow, 1971); Lazard, Grammaire du Persian
Contemporain; Sami’ian, “Structure of Phrasal Categories in Persian”; Karimi and Brame, “A
Generalization Concerning the Ezafe Construction in Persian.”
19
Parsafar, “Spatial Prepositions in Modern Persian.”
650 Parsafar

hand, the head of an NP is expected to be a noun, too. Therefore, /zir/ “under”


which is “semantically” a preposition-like morpheme is syntactically functioning
as a noun. Parsafar has shown that all the other pseudo-prepositions display
nominal (or adjectival) functions whenever they are followed by ezafe.20
To sum up, again it can be concluded that ezafe can be attached only to [+N]
categories, i.e. adjectives and nouns. One question which might arise here is
whether ezafe is itself a nominalizer or it only follows nominals and already nomi-
nalized categories. The next section will provide an answer to this question.
Attributive prepositional phrases. Prepositional phrases can modify the heads of
the NPs. It will be shown very briefly here that although they are PPs in form,
they function merely as adjectivals rather than nominals or adverbials.
There are basically three types of prepositional phrases that can be used as
modifiers of heads in complement positions. The first type does not carry any
verbal element as in (38). That the seemingly prepositional phrase /ba sæfa /
“with pleasantness” is used as an adjective in (38a) can be illustrated by the
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

fact that it can take the comparative suffix /-tær/ as in (38b). This casts doubt
on the nature of prepositions in such constructions. They might in fact be deri-
vational prefixes, in this case a denominal adjectival prefix.

(38) a. / adæm-e ba sæfa/


person-E with pleasantness
“a pleasant/agreeable person”
b. /xosro æz mæmmæd ba sæfa-tær-e /
Khosrow from Mohammad with pleasant-er-is
“Khosrow is more agreeable than Mohammand.”

The second type of prepositional phrase used in the complement position


always consists of a past participle verb as in (39). It is clear that the PP /be
taraj ræfte/ “plundered” is used as an adjective in (39a). The adverb /taze/
“recently” in (39b) which modifies this prepositional phrase provides proof for
the adjectival function of the PP.

(39) a. [ [æmval]-e [be taraj ræfte] ]-ye ‘u


NP PP
belongings-E to plunder gone-E he
“his plundered possessions”
b. [ [æmval]-e [taze be taraj ræfte] ]-ye ‘u
recently
“his recently plundered possessions.”
The possibility of PPs functioning as modifiers, however, cannot refute the con-
clusion drawn so far, which claims that ezafe can attach only to [+N] constituents.
20
Parsafar, “Spatial Prepositions in Modern Persian”.
Syntax, Morphology, and Semantics of Ezafe 651

In (39), both the head /æmval/ and the entire embedded NP receive ezafe. In other
words, the second ezafe is not affixed to the PP but to the embedded NP.
The third type of PP that can be employed in the complement position enjoys
the fact that the use of ezafe (preceding the PP) is optional; therefore it is redun-
dant, as in (40a) and (41a). The major difference between this PP and the other
two is that this one ordinarily cannot be further followed by other constituents
as complements of the same head, as in (40b) and (41b):
(40) a. /enteqal(-e) be tehran /
transference(-E) to Tehran
“transference to Tehran”
b. ?? /enteqal-e be tehran-e ma/
transfer-E to Tehran-E we
“[lit.] transference to Tehran of ours”
c. /enteqal-e ma be tehran /
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

transference-E we to Tehran
“Our transfer to Tehran”
(41) a. / e’teraz(-e) be jæng / “objection to war”
objection(-E) to war
b.∗ /e’teraz(-e) be jæng-e anha/
objection(-E) to war-E they
c. /e’teraz-e anha be jæng /
objection-E they to war
“their objection to war”

Sentence (40b) does not sound acceptable to many speakers because the PP /be
tehran/ “to Tehran” and the head /enteqal/ “transfer” do not make a constituent
in order for them to be further modified by /ma/ “we.” To make it acceptable
and grammatical, /ma/ is moved to the position after /enteqal-e/ “transfer-
ence-E” with which it makes a constituent NP, as in (40c). The same analysis
holds true with (41b – c), as well.
Thus the final conclusion to this section is that the elements that can be used as
constituents preceding and taking ezafe are N, NP, and adjectives (only in the pre-
dicate position of certain auxiliaries). These can be considered as [+N] heads. On
the other hand, not only can [+N] constituents follow ezafe as complements but
also PPs with an adjectival function. This conclusion is not unusual, considering
the fact that the complements of a head inside any NP should all exhibit an adjec-
tival function since they are modifying a head N/NP.

The Morphology of Ezafe

What is the nature of ezafe morphologically? Is it a suffix or a clitic? In the former


case, is it derivational or inflectional? If it is the latter, how does it interact with
other clitics? To search for morphological replies to these questions, the behavior
652 Parsafar

of ezafe will be closely compared and contrasted with that of three other mor-
phemes, namely /-i/, the conjunct possessive pronouns (henceforth POSS),
and the adposition /-ra/.
This section will first provide evidence indicating that the ezafe bound mor-
pheme /-e/, the so-called “suffixes” /-i/ and POSS, and /-ra/ are not inflectional
or derivational suffixes for several reasons, the most immediate and prominent of
which is the fact that these are merely “extra-inflectional.”21 Then, it will be
argued that ezafe and POSS are clitic-like whereas /-i/ and /-ra/ are more
word-like.
Extra-inflectional morphemes. In his discussion of the structure of complex
words, Bloomfield refers to “an outer layer of inflectional constructions” and
“an inner layer of constructions of word-formation.”22 This major difference
between inflectional and derivational affixes has been further supported by
many linguists such as Aronoff, Zwicky and Pullum, Scalise, and Klavans,
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

among others.23
Parsafar shows that in plural NPs, the conjunct possessive pronouns, and
the “Indefinitizer /-i/” are used after the inflectional affixes /-ha/, /-an/, and
/-jat/,24 which are all plural markers, whereas none of the other inflectional or
derivational suffixes exhibits such a property.
Let us first study the morphemes under discussion when they are affixed to
singular nouns. Examples (43 – 45) show that the Indefinitizer, the plural
makers, and the possessive pronouns attach (only) to the right of nouns. Sen-
tences in (46) illustrate that the Indefinitizer appears after the derivational
(bound) suffixes (DA) such as /-ban/ “keeper.”

(42) /ketab/ “book”


(43) /ketab-i / “a book”
book-ID25
(44) /ketab-ha/ “books”
book-PL

21
See Judith L. Klavans, “The Independence of Syntax and Phonology in Cliticization,”
Language, 61, no. 1 (1985): 95 –120.
22
Leonard Bloomfield, Language (Chicago, 1933), 222.
23
Mark Aronoff, “Word Formation in Generative Grammar,” Linguistic Inquiry, 1 (1976);
Arnold M. Zwicky and Geoffrey Pullum, “Cliticization vs. Inflection: English n’t,” Language, 59,
no. 3 (1983): 502 –513; S. Scalise, Generative Morphology (Dordrecht, 1984); Klavans, “The Indepen-
dence of Syntax and Phonology in Cliticization.”
24
Parviz Parsafar, “The Morphology of Modern Persian Suffixes” (Unpublished qualifying
paper presented to the faculty, Department of Linguistics, Yale University, New Haven, CT,
1990), 49. He illustrates that “this /i/ is unspecified for definiteness in the sense that when it is
used in isolation, it is ‘unspecific indefinite’, but when used in context, it can be ‘specific indefinite’
or ‘definite’ .”
25
In what follows, PL stands for the plural markers, ID for Indefinitizer /-i/, and POSS for the
possessive pronominal suffixes (or, as will be shown, enclitics).
Syntax, Morphology, and Semantics of Ezafe 653

(45) /ketab-æm/ “my book”



book-my
(46) a. / otor-ban-i/ “a cameleer”

camel-DA-ID
b. ∗ / otor-i-ban/

(47) shows that the inflectional plural marker /ha/, too, occurs after the deri-
vational suffixes. The same final position holds for the POSS, which in (48) has
been attached to the derivational (diminutive) bound suffix /-çe/. Note that as all
these examples show, one common property of /-i/ and POSS is that neither one
can change the category of its base.

(47) a. / otor-ban-ha/ “cameleers”
 PL
b.∗ / otor-ha-ban/
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

(48) a. /baq-çe-æm / “my flowerbed”


garden-DA-POSS
b.∗ /baq-æm-çe/

As (49) illustrates, the possessive pronouns and /-i/ cannot co-occur simul-
taneously in the same NP. The reason is that /-i/ is an “unspecific indefinite”
marker, in isolation, and the possessive pronouns are inherently “definite.”
Since both of these are to be used post-nominally, their co-occurrence is seman-
tically implausible in Persian. Neither one is powerful enough, nor does it have
the motivation, to neutralize the effect of the other.

(49) a. ∗ / ketab-æm-i /
book-POSS-ID
b. ∗ / ketæb-i-æm/
book-ID-POSS

As for the plural nouns, either /-i/ or POSS can encompass the plural suffix, but
not vice versa:

(50) a. /ketab-ha-i / “some books”


book-PL-ID
b. ∗ / ketab-i-ha/
(51) a. /ketab-ha-yæm/ “my books”
book-PL-POSS
b. ∗ / ketab-yæm-ha/

Furthermore, as has been shown by Parsafar, the right-most nominal element


in Persian is the adposition /ra/ which is “the Specificity Marker of Objects and
654 Parsafar

Topics,”26 henceforth SM.27 This adposition, which does not change the cat-
egory of its base, occurs even after /-i/, or

(52) /ketab-ra xæridæm/ “I bought the book.”


book-SM bought-[1sg]
(53) /ketab-ha-i-ra aværd ke. . ./
book-PL-ID-SM brought-[3sg] that
“He/she brought the books that . . .”
(54) /ketab-ha-yæm-ra . . . / “my books”
book-PL-POSS-SM

Thus, the orders of these morphemes and the post-position /ra/ are fixed:
/-ha-i-ra/ and /-ha-POSS-ra/. Hitherto, it can be concluded that the ID /-i/,
the POSS, and the SM /-ra/ cannot be derivational suffixes for the following
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

reasons: First, they do not change the category of their bases. Parsafar shows
that the majority of Persian derivational morphemes are category changing.
The ones that are category preserving are basically evaluatives, which include
diminutives, derogatives, and hypocoristics.28 (Some English derivational
affixes such as un- are also category preserving.) Second, they occur to the
right of the inflectional morphemes. Furthermore, recall that, as has been
shown so far, ezafe shares these two properties, as well.

What kind of morpheme? The question to be dealt with in this section is whether the
ezafe /-e/, the ID /-i/, the POSS, and /-ra/ are clitics or inflectional affixes,29 or
some other kind of morpheme. Initially, a description of clitics is in order.

26
Parsafar, “The Morphology of Modern Persian Suffixes,” 40–48.
27
Assuming the principles of Chomsky’s (1981) GB theory, Karimi (“Aspects of Persian
Syntax”, 100) argues that /ra/ “follows a specific NP if the latter is not marked [+NOM] and is
not in the minimal government-projection of a (¼N, A, or P). Independently, and based on the
theory of Relational Grammar developed by David Perlmutter, “Relational Grammar,” Syntax
and Semantics, 13 (1980): 195 –227, Parsafar (“The Persian /ra/”) argues that /ra/ is “a Specificity
Marker of Objects and Topics.”
28
Parsafar, “The Morphology of Modern Persian Suffixes,” 23–24.
29
Analyzing certain Italian data, Scalise (Generative Morphology, 126) refers to Kiparsky’s mech-
anism of “inflection blocking” which states that “sequences of consecutive inflectional elements
are prohibited.” However, this “mechanism” has been illustrated to be ineffective in Persian by
Parsafar (“The Morphology of Modern Persian Suffixes”), as in the following:

kuçik ‘small’
kuçik-tær ‘smaller’
kuçik-tær-a ‘the little/younger/smaller ones’
Adj IA IA

It should also be mentioned that Parsafar (Ibid) considers the ID /-i/ as an inflectional suffix,
whereas the analysis in the present work will propose that /-i/ is probably more word-like.
Syntax, Morphology, and Semantics of Ezafe 655

Clitics defined. Giving examples from English and some other languages,
Aronoff briefly discusses the existence of some “grammatical morphological
phenomena which cannot be subsumed under inflection,”30 the best known
of which is cliticization. Zwicky elaborates on this and other borderline
cases.31
In these papers, Zwicky surveys a host of analytical problems related to a
Madurese “reduplicative morpheme.” In the course of his discussion, Zwicky
presents six principles, namely Ordering, Internal Sandhi, Rule Immunity,
Binding, Construction with Affixes, and Accent to distinguish affixes from
each other.
Then he describes three classes of exceptional cases (to the last three prin-
ciples). The first class consists of cases where “an unaccented bound form acts
as a variant of a stressed free form with the same cognitive meaning and with
similar phonological makeup.” These conjunct clitics often show “special
syntax.” For example, in French declarative sentences, conjunct clitics are
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

placed before the verb even though French is an SVO language.32


In Persian, which is an SOV language, the unstressed conjunct object
pronouns are obligatorily affixed to the right of the verb, as in (55b), while
their corresponding stressed free units, or disjuncts, are ordinarily placed before
verbs, as in (55a):

(55) a. /mæn bæhram-ra didæm/ “I saw Bahram.”



I Bahram-SM saw-[1sg]
b. /didæm-æ / “I saw him.”
saw-[1sg]-[3sg]

As for the possessives, the conjunct and the disjunct forms of possessive
pronouns are used post nominally. The latter, however, are preceded by ezafe
because they function as genitive nominals modifying the head of the ezafe,
whereas the former cannot be preceded by the ezafe morpheme:

(56) a. /ketab-e mæn / “my book”


book-E I

b. /ketab mæn/
(57) a. /ketab-æm/ “my book”
book-POSS-[1sg]

b. /ketab-e-æm/

30
Aronoff, “Word Formation in Generative Grammar,” 3–4.
31
Arnold M. Zwicky, On Clitics, Indiana University Linguistics Club publication (1977); Arnold
M. Zwicky, “On Clitics,” in Phonologica, ed. by Wolfgang U. Dresser and Oskar E. Pfeiffer
(Innsbruck, 1976). This is a shorter version of the 1977 paper.
32
His examples are “Je vois Jean” (“I see John”) and “Je le vois” (“I see him”).
656 Parsafar

The second class of cliticization is usually “associated with stylistic conditions,


as in the casual speech cliticization of object pronouns in English.” Zwicky
calls the first class “Special Clitics” and the second class “Simple Clitics.”
His third class is referred to as “Bound Words.” These are unaccented
morphemes which are usually placed at the very margins of their hosts, standing
even outside inflectional affixes. The English example is the possessive
morpheme.
Based on Zwicky’s arguments, then, it seems that the Persian Possessive (as
well as Objective) Pronouns are of the “Special Clitic” form while the Indefini-
tizer /-i/, the ezafe, and the SM /-ra/ might be of the “Bound Clitic” form.
The next section will further investigate this distinction.
Clitics, affixes, or words? It will be illustrated here that ezafe and conjunct
possessives are more clitic-like while the ID /-i/ and the SM /-ra/ share some
properties with “words” and some with “clitics.” Zwicky and Pullum present
six criteria distinguishing between English clitics and inflectional affixes. Four
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

of these criteria will be applied to the Persian data as follows:


1. “Clitics can exhibit a low degree of selection with respect to their
hosts, while affixes exhibit a high degree of selection with respect to their
stems.”33

The ezafe /-e/, the ID /-i/, and the POSS can attach virtually to any NP,
whether singular or plural. None of the inflectional (or derivational) suffixes
are as productive as these three. Nevertheless, these three cannot co-occur. On
the other hand, the SM /-ra/ exhibits a selective application in that it marks
only Specific Object and Topic NPs.

2. “Arbitrary gaps in the set of combinations are more characteristic of affixed


words than of clitic groups.”

Examples (58 – 61) show that even the plural markers /-ha/ and /-an/, which
are among the most productive inflectional suffixes, are faced with some arbitrary
gaps.34

(58) a. /sib-ha /
apple-PL

Cf. b. /sib-an/ “apples”
apple-PL

33
Zwicky and Pullum, “Cliticization vs. Inflection: English n’t.”
34
M. Mo’in, Mofrad-o Jam’ [Singular and Plural] (Tehran, 1340/1961), presents a nearly exhaus-
tive list of the types of bases for /-ha/ and /-an/ which need not be mentioned here. Parsafar (“The
Morphology of Modern Persian Suffixes,” 31) claims that “almost any base that can be pluralized by
/-an/ is also pluralizable by /-ha/, but not vice-versa.”
Syntax, Morphology, and Semantics of Ezafe 657

(59) a. /zæban-ha/
tongue-PL

Cf. b. /zæban-an/ “tongues”
tongue-PL
(60) a. ?/mæhru-ha/
beautiful-PL
Cf. b. /mæhru-yan/ “the beautiful (women)”
(61) /deræxt-ha/ ¼ /deræxt-an/ “trees”
tree-

Gaps do also frequently occur with derivational suffixes, for example /-e / and
/-id/ in (62) and (63). In contrast, the ID /-i/, the ezafe /-e/, the SM /-ra/, and
the POSS do not allow gaps.
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

   
(62) /ku -/ “try” 1 /-e / /ku -e / “attempt, try” n
present stem 
(63) /xær-/ “buy” ∗ /xær-e /
present stem
Cf. /xær/ +/-id/ /xær-id/ “purchase” n
3. “Morphological idiosyncrasies are more characteristic of affixed words than
of clitic groups.”
There are no cases where a particular host-clitic combination, with ID /-i/, or
E /-e/, SM /-ra/, or POSS, shows an unexpected phonological form. Hosts are
not affected by these four. However, they themselves have allomorphs, as
depicted in the following table:

(64) Clitics and affixes

On the other hand, idiosyncratic alternations are abundant among inflectional


(and derivational) suffixes. The past tense morphemes, for instance, are involved
with a great number of morphological idiosyncrasies:
658 Parsafar

(65) a. /ist/ pres. stem ‘stand’ /istad/


 past stem ‘stood’
b. /dar/ ¼ ¼ ‘have’ /da t/ ¼ ¼ ‘had’
c. /bær/ ¼ ¼ ‘carry’ /bord/ ¼ ¼ ‘carried’
d. /xah/ ¼ ¼ ‘want’ /xast/ ¼ ¼ ‘wanted’

4. “Clitics can attach to material already containing clitics, but affixes cannot.”

This criterion is only partially borne out for the three types of clitic-like
morphemes under scrutiny. As was discussed above, there are no derivational
or inflectional affixes that can attach to ID /-i/ or the conjunct possessives,
save for the Specificity Marker of Objects and Topics (SM) /-ra/. As for ezafe,
there is absolutely no morpheme, including /-ra/, that can be affixed to it.
(Recall, however, that ezafe cannot end an NP unless it is followed by another
word or constituent.)
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

Since these morphemes are not completely in accordance with the criterion
being considered here on the account of their relative inability to apply to each
other, it can be maintained that they have a mixed behavior between words
and clitics. Nevertheless, /-i/, /-ra/, and POSS are more clitic-like than ezafe
simply because /-ra/ can attach to /-i/ and POSS, but no morpheme can
attach to /-e/. Moreover, the fact that /-ra/ and /-e/ are each the last elements
attached to a given N/NP can be regarded as a major difference between these
and the other two.
Zwicky, who further elaborates on his earlier papers on clitics, provides a list
of tests for distinguishing “clitics from independent words.”35 Some of these
tests are paraphrases of the above-mentioned criteria. However, his “movement”
test can be of significance to our discussion here.

Movement: “in an X + Y combination, if either X or Y can be moved without


the other, then X and Y are words: neither of them is a clitic.”
It will be shown here that /-ra/ and /-i/, but not /-e/ and POSS, can be moved
in some contexts. In another paper, Zwicky maintains that the English possessive
clitic is “attached phonologically to the last word of the noun phrase,”36 even if it
is not the head of the NP:

(66) The Queen of England’s hat


(67) The woman I talked to’s arguments

The SM/-ra/ behaves similarly in that, in simple NPs and in ezafe construc-
tions, it is attached to the right-most element of the noun phrases and cannot
be moved inside them:
35
Arnold M. Zwicky, “Clitics and Particles,” Language, 61, no. 2 (1985): 283 – 305.
36
Zwicky, On Clitics, 7.
Syntax, Morphology, and Semantics of Ezafe 659

(68) a. /sa’æt-æm-ra/ “my watch”


watch-my-SM
b. [ sa’æt-e tæla-ye su’isi-yæm]-ra . . .
NP
watch-E gold-E Swiss-my-SM
“. . .my gold Swiss watch”

However, when the head is modified by a relative clause, /-ra/ can attach either
to the head or to the rightmost word of the relative clause, as in (69).

(69) a. [ [mærd-i]-ra ke diruz didæm] . . .


NP N
man-ID-SM that yesterday saw-[1sg]
“. . .the man I saw yesterday.”
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

b. [ [mærd-i] ke diruz didæm]-ra ‘. . .the man I saw yesterday.’


NP N
man-ID that yesterday saw-[1sg]-SM

In this case, the behavior of /-i/ seems to be the converse of that of /-ra/.
Although the /-i/ that is attached to the head of the relative clauses, as in
(69a– b), cannot be moved from its position, the /-i/ in (70a), where it is attached
to a head noun which is not modified by a tensed clause, can be moved to the end
of the object NP, as in (70b).
Notice that in (70b), the attachment of ezafe to the head noun is obligatory
because it makes the modifying function of the following PP possible. It is
only after the combination of [ [N]-e PP] that /-i/ and subsequently /-ra/ find
an appropriate NP to attach to.

(70) a. [ [særbaz]-i æz jæng bær-gæ te]-ra didæm ke
NP N
soldier-ID from war returned-SM saw-[1sg] that
 who. . .”
? “I saw a returned-from-the-war soldier
b. [ [ [særbaz]-e æz jæng bær-gæ te]-‘i]-ra didæm ke
NP NP N
soldier-E from war returned-ID-SM saw-[1sg] that
? “I saw a returned-from-the-war soldier who . . .”

Examples (69 –70), then, illustrate that /-ra/ and /-i/ can be categorized as
bound “words,” rather than “clitics.” Moreover, of these two, /-ra/ is more
word-like than /-i/ due to its excessive freedom in movement. On the other
hand, it was shown above that they can also be clitics. The “mixed status” of
these two may still seem more dubitable considering the fact that they are
stressless just as the clitics are. Zwicky calls those words which are “prosodically
660 Parsafar

subordinate to their neighbouring material” leaners. He further adds that some-


times leaders act as clitics “in certain circumstances.”37
Furthermore, Zwicky claims that there are two kinds of leaners; obligatory and
optional. Obligatory leaners, such as the English articles (71a) and coordinating
conjunctions (71b), cannot occur alone. Optional leaners, however, such as the
English prepositions (72a) and auxiliaries (72b), can occur without “a member
of the category on which” they depend.38

(71) a. “∗ Wilma said she was pointing at a lion, but I couldn’t see the (at all)”
b. “∗ It was Susan that I saw Terry and (in London)”
(72) a. “It was Wystan I sent the poem to (last week)”
b. “Margaret thinks Norman is a genius, but I don’t think he is (at all)”

Considering the mixed status of /-i/ and /-ra/, they can be conceived as
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

morphemes which share some properties with both “bound clitics” and
“obligatory leaners.” However, what these two morphemes do not share with
Zwicky’s “leaners” is that “leaners are disinclined to attach to other leaners.”39
As for /-i/ and /-e/, what they have in common is that they cannot attach to
tensed clauses. There are also two major differences between them; (a) ezafe’s
ability to have multiple attachments to all the constituents inside the NPs, and
(b) ezafe’s inability to move.
Finally, the POSS morphemes, too, attach only to the last element of any given
NP. Examples (74b) and (75b – c) are ill-formed because the possessive pronoun
is not affixed to the entire NP.

(73) /goldan-æm/ “my vase”


vase-POSS-[1sg]
(74) a. /goldan-e qermez-æm/ “my red vase”
vase-E red-POSS-[1sg]

b. /goldan-æm qermez /
vase-POSS-[1sg] red
(75) a. [ [ [goldan]-e qermez]-e ekæste]-æm]
NP NP
vase-E red-E broken-POSS-[1sg]
“my broken red vase”

37
Arnold M. Zwicky, “Stranded to and Phonological Phrasing in English,” Linguistics, 20 (1982):
3–57.
38
Arnold M. Zwicky, “Stranded to,” Ohio State University Working Papers in Linguistics 24 (1980):
166– 173.
39
Zwicky, “Stranded to,” 171.
Syntax, Morphology, and Semantics of Ezafe 661



b. /goldan-æm qermez-e ekæste /

c. [ goldan-e qermez-æm ekæste]
NP  NP
Cf. d. [ goldan-e qermez-æm] ekæste
NP NP
vase-E red-POSS-[1sg] is broken
‘My red vase is broken.’

Examples (76 a– b) show that POSS cannot be separated from the head if the
latter is modified by a relative clause:

(76) a. [ [goldan-æm] ke diruz ekæst]. . .
NP NP
vase-POSS that yesterday broke [3sg]

“My vase which broke yesterday”
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011


b. [ [goldan] ke diruz ekæst-æm ] . . .
NP NP
Recall that this inability to move is also shared by ezafe. Regarding Zwicky’s
“movement” test, then, it seems that the POSS morphemes and ezafe are more
clitic-like than /-i/ and /-ra/. Therefore, to conclude this section, in the
absence of further evidence, it appears reasonable enough to consider both
ezafe /-e/ and POSS as “clitics” and the other two, i.e. /-i/ and /-ra/, either
as “obligatory leaners” (à la Zwicky) or simply as “bound words.”
Furthermore, of all of these morphemes, the conjunct possessive pronouns
have shown to be the most reliable since they have passed all the tests successfully.
Henceforth, they will be referred to as the Enclitic Possessive Pronouns (EPPs).
Is ezafe enclitic or proclitic? If ezafe is a clitic, does it attach to its preceding words or its
following word? This subsection will illustrate that ezafe is phonologically enclitic
and syntactically, too, it attaches to its preceding element. Klavans presents
examples from Kwakwala, Tepecano, Nganhcara, and Greek,40 and Marantz
gives examples from French, Yagua, and Papago, all indicating that it is possible
for some clitics to have two separate hosts, syntactic and phonological.41
It seems that the Persian ezafe, however, does not enjoy this “dual citizen-
ship.”42 Sami’ian mentions only one piece of phonological evidence for ezafe.
The fact that ezafe forms a phonological unit with its preceding element can be
attested by displacing the pause in any given phrase. The phrase will be well-
formed only when the pause follows ezafe:43

40
Klavans, “The Independence of Syntax and Phonology in Cliticization.”
41
Alec Marantz, “Clitics and Phrase Structure,” in Alternative Conceptions of Phrase Structure, ed.
by Mark R. Baltin and Anthony S. Kroch (Chicago, 1989): 99 –116.
42
Klavans, “The Independence of Syntax and Phonology in Cliticization,” 104.
43
Sami’ian, Structure of Phrasal Categories in Persian, 39 and 78.
662 Parsafar

(77) a. [N]-e + {PAUSE} + N /çeraq-e mæn / ‘my lamp”


lamp-E I
∗ ∗
b. [N] + {PAUSE}+ e-[N] /çeraq e-mæn/
Here, I will present another piece of evidence for ezafe’s being phonologically
an enclitic morpheme. As mentioned above, whenever the preceding element, i.e.
the host, ends in a vowel, the palatal approximant /y-/ will be inserted before
ezafe. This insertion, however, will not take place if the following element
begins with a vowel whether the head noun ends in a vowel or not:

(78) /xane-ye jæm id / “Jamshid’s house”
house-E
(79) a. /xane-ye ema/ “Ema’s house”

b. /xane-ye y-ema/

c. /xane y-ema/
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

As for the syntactic evidence, hitherto it has been implicitly assumed that ezafe
attaches to its preceding element, which means it is enclitic. That ezafe is not syn-
tactically proclitic can be shown by a “constituency test” similar to Radford’s
“Shared Constituent Coordination.”44
In (80), since /otaq/ “room” can substitute, both syntactically and semanti-
 /pænjere-ye otaq/ “room’s window” after /ya/ “or,” then /otaq/
cally, for
and /a pæzxune/ “kitchen” are both syntactically free of ezafe. That is, ezafe is
not part of these two attributive nominals.
 
(80) pænjere-ye a pæzxune ekæste ya (pænjere-ye) otaq
window-E kitchen broken-is or (window-E) room
“Is the kitchen window broken or the room’s (window)?”

This, however, cannot be used as the conclusive evidence for arguing that the com-
 N or NP. As (81)
bination of the preceding noun and ezafe forms a single constituent
exhibits, door-E cannot be used as a constituent whereas/dære /“its door” can.

(81) pænjere-ye hæmum ekæsse ya
window-E  bathroom broken-is or
dære /∗ dær-e/∗ dær
door-its/∗ door-E/∗ door
“Is the bathroom window broken or its door?”

This test shows that /dære / “its door” is substituting for the whole subject NP
/dær-e hæmum/ “bathroom door.” As was mentioned before, the reason is that
the combination [NP +E] or [N + E] is not an independent constituent, per se,

44
Andrew Radford, Transformational Grammar (Cambridge, 1988): 89 –105.
Syntax, Morphology, and Semantics of Ezafe 663

because ezafe must obligatorily be followed by a complement modifying the host


of the ezafe.
Once again, this is indicative of the fact that the behavior of ezafe is very different
from the English enclitic possessive -’s since the latter can form a constituent with
its host:
(82) Did you break John’s pencil or Mary’s (pencil)?
Mary’s (pencil).
Klavans argues that “affixes subcategorize at the lexical level” while “clitics
subcategorize for phrasal hosts.”45 However, ezafe can attach to N(P) and Adj(P).
This may cast further doubt on the clitichood of ezafe and on Klavans’ argument.
As a final remark, it seems that instead of being introduced through a transfor-
mational rule as proposed by Sami’ian, ezafe can be (alternatively) described as a
 (N-bar). It will then filter down, pre-
feature46 that is initially assigned to the N
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

sumably like the accusative case in Latin or the dative case in German, to the head
 and, simultaneously or afterward, to all the constituents to the right of
of the N
the head, i.e. the complements of the head rather than its specifiers. For instance,
the structure of (83) can be partially shown to be that in (84):
(83) goldan-e siyah-e kuçæk-e otaq-e bæradær-e ‘u
vase-E black-E small-E room-E brother-E he
“the small black vase in his brother’s room”
(84) Partial structure for (83):

45
Klavans, “The Independence of Syntax and Phonology in Cliticization.”
46
I owe this analysis to Professor Laurence R. Horn, my teacher and adviser.
664 Parsafar

The Semantics of Ezafe

Is ezafe a genitive marker? If so, why is it used to show relations other than pos-
session? Is ezafe a polysemous morpheme or are there many semantically different
but homophonous ezafe’s?
This section will demonstrate that ezafe is merely a dummy Connective or
Associative morpheme devoid of any meaning. It only serves to connect [+N]
lexical and phrasal heads to complements modifying the heads. The type of
semantic relationship that the native speaker infers from each ezafe construction
is solely dependent on the syntax and semantics of the constituents associated with
each other by means of the dummy ezafe.
First, it is worth noting that regarding the genitive uses of this morpheme,
ezafe is comparable to the English possessive -’s. As Quirk et al. have mentioned,
the English genitive case can carry different meanings such as “possessive geni-
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

tive” (85), “subjective genitive” (86), “objective genitive” (87), “genitive of


origin” (88), “descriptive genitive” (89), “genitive of measure” (90), genitive
of attribute” (91), and “partitive genitive” (92):47

(85) “the earth’s gravity” “The earth has (a certain) gravity.”


(86) “the boy’s application” “The boy applied for . . .”
(87) “the boy’s release” “(. . .) released the boy”
(88) “the general’s letter” “The general wrote a letter.”
(89) “a woman’s college” “a college for women”
(90) “ten day’s absence” “The absence lasted ten days.”
(91) “the victim’s courage” a) “The victim had courage.”
b) “The victim was courageous.”
(92) “the baby’s eyes” “The baby has (blue) eyes.”

All these “meanings” are expressed by ezafe constructions in Persian. However,


as has already been clarified, ezafe can help express many other “relations” as well.
Second, the relation expressed in (93) is that of possession. By changing the
order of the two constituents, we get two different meanings. Ezafe’s function
seems to be that of subordinating the constituent to which it is attached, i.e.
its host, to its following constituent. Apart from this syntactic function, ezafe
does not carry any semantic load in (93):

(93) a. /bæradær-e zæn/ “wife’s brother”


brother-E wife
b. /zæn-e bæradær / “brother’s wife”
wife-E brother
47
R. Quirk, Sidney Greenbaum, Geoffrey Leech, and Jan Svartvik, A Comprehensive Grammar of
the English Language (London, 1985). See also Christopher Lyons, “The Syntax of English Genera-
tive Constructions,” Journal of Linguistics, 22 (1986): 123 –143.
Syntax, Morphology, and Semantics of Ezafe 665

Third, when the word /ævvæl/ is used adjectivally, it means “first,” as in (94a),
but when it is nominalized, probably through zero-derivation, as in (94b), it
means “beginning.” In isolation, its only semantics is the adjectival “first” and
it acquires its nominal meaning “beginning” when ezafe is cliticized to its right.

(94) a. /ketab-e ævvæl/ “the first (grade) book”


book-E first
b. /ævvæl-e ketab/ “the beginning of the book”
beginning-E book
Cf. c. /dastan-e ævvæl-e ketab /
story-E beginning-E book
“the story at the beginning of the book”
Probably one might argue that a corollary to this syntactically-triggered
semantic phenomenon is that ezafe is a nominalizer. Such an argument,
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

however, is untenable because, as was shown above, ezafe is not a derivational


morpheme.
Finally, the head of the NP /jæ’be/ “case” in (95a) is the same as that in (95b),
while the second element is the noun /tæla/ “gold” in the former, but the
denominal adjective /tæla’i/ “golden” in the latter. Thus, the meanings of the
two phrases are primarily based on the two complements.

(95) a. / jæ’be-ye tæla/ “a jewelry case”


case-E gold
b. / jæ’be-ye tæla’i / “a golden case/box”
case-E golden

Nevertheless, that the semantics of the host does also have some effect on the
meaning of the entire NP may be illustrated by the examples in (96). Unlike
/jæ’be/ “box,” /gonbæd/ “dome” may not be used as a container. Therefore,
both the NPs have the same meaning. What these observations entail again is
that ezafe is semantically empty.

(96) a. / gonbad-e tala / “a golden dome”


dome-E gold
b. / gonbad-e tala’i / “a golden dome”
golden

In brief, then, what ezafe communicates syntactically is that its left constituent
is modified or complemented by its right element. This partially resembles
“nominal compounds in English where the ‘meaning’ is basically that an AB is
‘a B that has something to do with A’” (L. Horn. P.c.). However, as has been
illustrated above, ezafe’s functions have a greater scope than that of the
English nominal compounds.
666 Parsafar

Conclusions

This paper has been dedicated to an analysis of ezafe, a ubiquitous and highly
essential morpheme inside Persian noun phrases. It was established that the
unstressed ezafe /-e/ is phonologically enclitic and that syntactically it attaches
to its left-element. It was also illustrated that ezafe is semantically void, while
syntactically it is an associative marker which communicates, to the speaker,
the existence of a certain kind of relation or connection between its left-located
host and the modifier and/or the complement on its right.
The host is always a [+N] constituent, which can be a noun, an adjective, a
nominalized category, or an NP which does not contain any tensed clause. On
the other hand, the modifiers and/or complements of the head of the ezafe can
be nouns, adjectives, or prepositional phrases. A major corollary to these con-
clusions is that ezafe can be used as a reliable diagnostic in distinguishing nom-
Downloaded By: [TÜBTAK EKUAL] At: 15:17 1 April 2011

inals and nominalized constituents from the majority of non-nominalized


adjectives and other constituents.
A theoretical analysis was also proposed that considers ezafe as a feature that is
first assigned to the given NP. It will then filter down to the head noun and all the
other (non-clausal) constituents to the right of the head. Determiners are ordina-
rily prenominal and do not receive ezafe. Only those determiners that can be used
as nominals, i.e. superlative adjectives, the /-in/ ordinals, and some quantifiers,
require the obligatory attachment of ezafe.
It was also illustrated that compared to ezafe, the conjunct possessive pronouns
behave like clitics, while the Indefinitizer /-i/ and the Specificity Marker of
Objects and Topics /-ra/ exhibit a “mixed status” between clitics and “bound
words.”

You might also like