Semantics‐Syntax Interface

Volume 1, Number 1 E‐ISSN: request pending

Spring 2014
 Semantics‐Syntax Interface
Volume 1, Number 1
Spring 2014

Editor‐in‐Chief
CONTENT
Negin Ilkhanipour
University of Tehran
Negin Ilkhanipour
Editorial Board
Laura Bortolotto
Editorial i
Ca’foscari University of Venice
Dustin Alfonso Chacon Original Articles
University of Maryland
Natalia Cichosz Barbara H. Partee
University College London
A Brief History of the Syntax‐Semantics
Despina Ikonomou
MIT Interface in Western Formal Linguistics 1
Moreno Mitrović
University of Cambridge Milada Walková
Ethan J. Poole
University of Massachusetts at Amherst Scalarity Determines Argument Structure in
Ayaka Sugawara English Particle Verbs and Slovak Prefixed
MIT
Anqi Zhang Verbs 22
University of Chicago

Associate Editorial Board

Pollet Samvelian and Pegah Faghiri
Octav Eugen DeLazero Persian Complex Predicates: How Composi‐
North‐Eastern Federal University
Ilaria Frana tional Are They? 43
University of Massachusetts at Amherst
Thomas Grano Short Contributions
University of Maryland
Valentine Hacquard
University of Maryland
Octav Eugen DeLazero and Parvane Geraee
Tim Hunter Ezafe Constructions as Relative Clauses:
University of Minnesota
Jungmin Kang Semantic Arguments 75
University of Connecticut
Jonathan E. MacDonald
University of Illinois at Urbana
Ramin Rahmany
University of Tehran

The editorial board is sincerely

honoured to be supported by
Rajesh Bhatt
University of Massachusetts at Amherst
Guglielmo Cinque
Ca’foscari University of Venice
Publisher
Kai von Fintel
MIT University of Tehran
Irene Heim
MIT Director‐in‐Chief
Sabine Iatridou
Gholamhosein Karimi‐Doostan
MIT
Angelika Kratzer University of Tehran
University of Massachusetts at Amherst
Howard Lasnik Cover Image
University of Maryland
Barbara H. Partee
Sara Shabanazad
University of Massachusetts at Amherst
Ian Roberts Website
University of Cambridge
[Link]
Editorial Assistant
Parinaz Dadras Email
University of Tehran semsyn@[Link]
Editorial

On behalf of the Editorial Board, it gives me great pleasure to welcome you to the
inaugural issue of Semantics‐Syntax Interface.
Edited by young scholars, Semantics‐Syntax Interface seeks to provide a
platform for the discussion of theoretical developments in the syntax/semantics
interface based on formal and experimental methods. The journal welcomes research
articles, short contributions (squibs and replies), and review articles mainly focused on
(but not limited to) mood and modality, quantification, argument structure, tense,
aspect, event structure, and negation.
The journal’s publication frequency is two issues per year, and a double‐blind
peer review process will be applied to all submissions. Semantics‐Syntax Interface is
an international journal, freely available online to all readers. The journal permits
unrestricted, non‐commercial use, distribution and reproduction of all articles in any
medium, provided that the work is properly cited.
In the development of Semantics‐Syntax Interface to date, we have been able to
enlist the support of our honorary editorial members who have been active in
promoting this online publication forum to their students. We would therefore like to
thank them for the support that they provided in this inaugural editorial. We are also
extremely grateful to Gholamhosein Karimi‐Doostan, Professor of Linguistics and
Dean of the Faculty of Letters and Humanities at the University of Tehran, for all his
support in launching this journal.

Negin Ilkhanipour
Editor‐in‐Chief

Semantics‐Syntax Interface, Volume 1, Number 1, Spring 2014

i
© 2014 by University of Tehran
[Link]
 A Brief History of the Syntax-Semantics
Interface in Western Formal Linguistics

Barbara H. Partee*
University of Massachusetts, Amherst, USA

Abstract: This essay describes and comments on some of the key

developments in the history of formal semantics and its relation to
syntax, focusing on the period from the beginnings of Chomskyan
generative grammar in the s up until the mature period of
formal semantics in the s, with only a few remarks about
earlier background and later developments. One crucial theme is
the challenge of developing a compositional semantic theory, and
how that challenge has taken on different forms with respect to
different syntactic and semantic theories.

Keywords: syntax-semantics interface, history of semantics,

compositionality, formal semantics

1. Introduction

The syntax-semantics interface is a relatively young topic in the history of linguistics.

In early Western linguistics, there was little syntax and essentially no semantics other
than some non-formal lexical semantics. Formal syntax came first, with Zellig Harris s
student Noam Chomsky revolutionizing the field of linguistics with his work on
syntax. Chomsky shared the field s general skepticism about the possibility of
semantics as a subject of scientific study, but as semantics began to be developed in the
context of generative grammar, the issue of the connection between syntax and
semantics came to the fore very quickly and soon became a major area of research.
Formal semantics arose a little later, with the logician Richard Montague as the key
figure in its origins, and linguists, logicians, and philosophers cooperated in its early
development. From both the syntactic and the semantic side, the emphasis was on the

* Email: partee@[Link]

Semantics-Syntax Interface, Volume 1, Number 1, Spring 2014

1–21
© 2014 by University of Tehran
[Link]
2 B.H. Partee

structure of sentences more than on the structure of words, and formal semantics has
always been in large part the semantics of syntax.

2. Before Syntactic Structures

In the 19th century linguistics existed within philology in Europe; in the U.S., it was
partly within philology (especially Indo-European historical linguistics) and partly
within anthropology, with fieldwork among native languages of the Americas a major
focus. In the 20th century, like so many other fields, linguistics emerged as a science.
Part of the Chomskyan revolution was to view linguistics as a branch of psychology
(cognitive science).
As for semantics, there was a mixture of negative attitudes and neglect in
American linguistics in the 20th century. The negative attitudes were influenced by
logical positivism and by behaviorism in psychology. The neglect of semantics in early
American linguistics was also because of the fieldwork tradition, where one generally
had to start with phonetics, then phonology, then morphology, then perhaps a little
syntax. Field linguists rarely got as far as semantics, other than in word lists and
analyses of certain semantic fields such as kinship terms.
In logic and philosophy, there was a great deal of progress in semantics in the
late 19th and early 20th centuries (Frege, Russell, Carnap, Tarski, and others), but that
was relatively unknown to most linguists.
So before Syntactic Structures (Chomsky 1957) there was no syntax-semantics
interface to think about in linguistics, and the logicians descriptions of the syntax and
semantics of their logical languages had no influence on linguistics.

3. In Syntactic Structures

Chomsky has always been ambivalent about semantics as a science and about the
relation between syntax and semantics. This is evident in Syntactic Structures and also
in his earlier-written The Logical Structure of Linguistic Theory (LSLT) (Chomsky 1975b).
In LSLT, Chomsky writes in the first chapter:
I will merely emphasize again that the legitimacy of semantics whatever that might
mean) has never, to my knowledge, been challenged, nor has there ever been any
question of the importance of incorporating a study of reference, meaning, language
use, and related topics within a full theory of language that will deal, in particular, with
the highly significant relations between formal structure and semantic interpretation.
The appeal to meaning must be clearly distinguished from the study of meaning. The
A brief history of the syntax-semantics interface 3

latter enterprise is unquestionably central to the general theory of language, and a

major goal of the SS-LSLT approach is to advance it by showing how a sufficiently rich
theory of linguistic form can provide structural descriptions that provide the basis for
the fruitful investigation of semantic questions. [emphasis added - BHP] (Chomsky
1975b:21)

From Syntactic Structures:

In proposing that syntactic structure can provide a certain insight into problems of
meaning and understanding we have entered onto dangerous ground. There is no
aspect of linguistic study more subject to confusion and more in need of clear and
careful formulation than that which deals with the points of connection between syntax
and semantics. The real question that should be asked is How are the syntactic
devices available in a given language put to work in the actual use of this language?
(Chomsky 1957:93)

Syntactic Structures concludes with the following paragraph, which nicely illustrates
Chomsky s ambivalence

More generally, it appears that the notion of understanding a sentence must be

partially analyzed in grammatical terms. To understand a sentence it is necessary
(though not, of course, sufficient) to reconstruct its representation on each level,
including the transformational level where the kernel sentences underlying a given
sentence can be thought of, in a sense, as the elementary content elements out of which
this sentence is constructed. In other words, one result of the formal study of
grammatical structure is that a syntactic framework is brought to light which can
support semantic analysis. Description of meaning can profitably refer to this
underlying syntactic framework, although systematic semantic considerations are
apparently not helpful in determining it in the first place. The notion of structural
meaning as opposed to lexical meaning , however, appears to be quite suspect, and it
is questionable that the grammatical devices available in language are used consistently
enough so that meaning can be assigned to them directly. Nevertheless, we do find
many important correlations, quite naturally, between syntactic structure and meaning;
or, to put it differently, we find that grammatical devices are used quite systematically.
These correlations could form part of the subject matter for a more general theory of
language concerned with syntax and semantics and their points of connection.
(Chomsky 1957:107–108)

Much of Chomsky s early emphasis on the autonomy of syntax is directed

against earlier attempts to base syntax on semantics or reduce much of syntax to
semantics, or to give semantic explanations for various syntactic phenomena and
restrictions.
We might loosely paraphrase Chomsky s attitude around the time of Syntactic
Structures thusly We don t really understand anything about semantics, which is a
4 B.H. Partee

subject not so far amenable to scientific investigation; but semantics is in principle

important, and deep structure reveals semantically relevant structure that is obscured
in surface structure.
The importance of deep structure for revealing semantically relevant structure is
illustrated in the famous pair easy to please vs. eager to please , which Chomsky
discusses in various works.

(1) a. John is easy to please < (for someone) to please John is easy

b. John is eager to please < John is eager (for John) to please (someone)

But sometimes transformations change meaning: Chomsky (1957:100–101) noted

that the following active-passive pair have different meanings, with the first quantifier
having wider scope in each case:

(2) a. Everyone in this room speaks two languages.

b. Two languages are spoken by everyone in this room.

In later years, those judgments about (2) came to be questioned; a number of

authors at different times claimed either that both (2a) and (2b) are ambiguous, or that
(2a) unambiguous but the passive (2b) is ambiguous. Chomsky himself has a long
footnote in Chomsky (1965) noting that it is possible that both sentences are
ambiguous, with some as yet unexplained factors making different readings preferred
for (2a) and (2b). We return to the topic of quantifier scope in Section 10.

4. From Syntactic Structures to Aspects: Katz, Fodor, Postal

‚t the beginning of the s, Jerrold Katz and Jerry Fodor, junior faculty at MIT,
started developing proposals for how a semantic theory could be developed in a
generative grammar framework. They were clearly concerned with what we now call
compositionality, and which they called the Projection Problem.
Because such things as negation and question-formation were treated via
transformations (T-NEG, T-Q) of affirmative declarative sentences, Katz and Fodor
(1963) figured that the meaning must depend on the entire transformational history.
The transformations are sketched in a very oversimplified way in (3a–b).

(3) a. [Mary [has [visited Moscow]]] T-NEG [Mary [has not [visited Moscow]]]
A brief history of the syntax-semantics interface 5

b. [Mary [has [visited Moscow]]] T-Q [has [Mary [visited Moscow]]]

Their idea of extending phrase-markers to T-markers showing the

transformational history of an expression, and computing the meaning on the basis of
the whole T-marker can be seen as aiming in the same direction as Montague s (later)
derivation trees.

(4) T-marker for (3a) = P-marker for the deep structure --- TNEG --- Taffix .

‛ut their semantics was very primitive. Katz and Fodor worked with semantic
features , and their semantic representations had no real structure, just bundles of
features – suitable at best for decompositions of one-place predicates. Later they added
some bits of structure to handle transitive verbs and their two arguments, but with no
attention at all to things like quantifiers.
And what they were trying to capture was restricted to things that could be
expressed in terms of readings – how many, and same or different. The three main
things to be captured were (i) ambiguity – having more than one reading; (ii) semantic
anomaly – having no reading; (iii) synonymy – sharing a reading (synonymy on a
reading), or the stronger version, having all the same readings. The examples of what
they could capture didn t seem very exciting, and the accounts were sometimes open
to easy counterexamples1.
Then Katz and Postal (1964) made the innovation of putting such morphemes as
Neg and a Question morpheme into the Deep Structure, as in (5), arguing that there
was independent syntactic motivation for doing so, and then the meaning could be
determined on the basis of Deep Structure alone.

(5) a. [NEG [Mary [has [visited Moscow]]]] T-NEG [Mary [has not [visited Moscow]]]

b. [Q [Mary [has [visited Moscow]]]] T-Q [has [Mary [visited Moscow]]]

On this view, Deep Structure would be the input to semantics, and Surface
Structure would then be the input to phonology.
The claim that transformations should be meaning-preserving was an interesting
and provocative one, and even without any real semantics at the foundation, it led to
interesting debates about apparent counterexamples. And the architecture of the

1 See Partee (2004, 2005, 2011) for more details on this and some of the other issues mentioned in this
paper.
6 B.H. Partee

theory (syntax in the middle, mediating between semantics on one end and phonology
on the other) seemed elegant and attractive.
Chomsky s thinking was evolving from Syntactic Structures (Chomsky 1957) to
Aspects (Chomsky 1965), and he tentatively accepted Katz and Postal s suggestion of a
systematic connection between syntax and semantics at the level of Deep Structure (His
continuing skepticism about semantics would still come out in such comments as
Well, I don t think anyone understands anything about semantics, but maybe what
Fodor and Katz and Postal are doing has some promise. .
During the brief period when Aspects held sway, there was a rosy optimism that
the form of syntactic theory was more or less understood and we could start trying to
figure out the substantive universals. I think of that period as the Garden of Eden
period: roughly the mid-to-late 19 s, before the linguistic wars broke out in full force,
a period when I think generative grammarians generally believed the Katz and Postal
hypothesis. That was also the period when people entertained the Universal ‛ase
Hypothesis , the conjecture that the grammars of all natural languages have the same
base rules (See brief discussion in Partee, Meulen and Wall (1990:556) and more in
Peters and Ritchie (1973)). The idea that meaning was determined at this deep level
was undoubtedly part of the appeal of the notion of Deep Structure beyond linguistics 2
and probably contributed to the aura surrounding the notion of language as a
window on the mind.
So around 1965, there was very widespread optimism about the Katz-Postal
hypothesis that semantic interpretation is determined by deep structure, and the
syntax-semantics interface was believed to be relatively straightforward (even without
having any really good ideas about the nature of semantics3).

5. Expulsion from the Garden of Eden: The Linguistic Wars

But at least among working linguists, that very appealing view fell apart when the
behavior of quantifiers was noticed. The many differences between quantificational
NPs and proper names (example (6) below) immediately created great conflicts
between syntax and semantics, and in a sense kicked all of us generative grammarians
out of our Garden of Eden.

2 See, for instance, the references to Deep Structure in Leonard ‛ernstein s 1973 Harvard Norton Lectures
(Bernstein 1976).
3 For more about debates between philosophers and linguists about the nature of semantics, see Partee
(2011).
A brief history of the syntax-semantics interface 7

5.1. Katz-Postal Hypothesis in Trouble

A surprising historical accident is that the behavior of quantifiers with respect to
transformational rules familiar from Syntactic Structures and Aspects was not really
noticed until the Katz-Postal hypothesis had for most linguists reached the status of a
necessary condition on writing rules. I believe that this historical accident was one of
the major causes of the linguistic wars between Generative Semantics and
Interpretive Semantics, which are described in Section 5.2 below. I give below a few
examples of derivations that would have been given in the Aspects theory, where the
examples given typically involved proper names like John and Mary, not quantifier
phrases, as in the (a) and (b) examples. We see that in the corresponding (c) and (d)
examples with quantifier phrases, the transformations radically change the meaning
(and it may change in different ways with different quantifiers). I suspect that the Katz-
Postal hypothesis would never have been suggested if such quantifier examples had
been noticed earlier.

(6) a. John voted for himself. FROM:

b. John voted for John.

c. Every man voted for himself. FROM:

d. Every man voted for every man.

(7) a. Mary wanted to win. FROM:

b. Mary wanted Mary to win.

c. Every candidate wanted to win. FROM:

d. Every candidate wanted every candidate to win.

(8) a. 3 is both even and odd. FROM:

b. 3 is even and 3 is odd.

c. No number is both even and odd. FROM:

d. No number is even and no number is odd.

(9) a. The pacifist who fought was inconsistent. FROM:

b. The pacifist fought. The pacifist was inconsistent.

c. All pacifists who fight are inconsistent. FROM:

d. All pacifists fight. All pacifists are inconsistent.

8 B.H. Partee

Example (6) shows that the transformational rule of reflexivization (Lees and
Klima 1963), which is apparently meaning-preserving with proper names4 is not
meaning-preserving with quantifier phrases. Example (7) shows the same thing for the
transformation of Equi-NP Deletion (Rosenbaum 1967), and (8) shows the same
thing for Conjunction Reduction (Chomsky 1957), where the transformation
preserves truth-value (False) for (a) and (b), but not for (c) and (d). As for example (9),
there were semantic problems lurking in the earlier derivations of relative clauses even
without quantifiers, but these did not come into focus until the post-Aspects period.

5.2. The Linguistic Wars

When the problematic relation between classic transformational derivations and
semantics came to light, there were two main classes of responses, characterized as
Generative Semantics and Interpretive Semantics (Newmeyer 1980; Harris 1993).
The Generative Semantics response (Lakoff, Ross, McCawley, Postal) was to
make the deep structures deeper , and try to find a semantically sound level of
underlying structure. The underlying structures of the Generative semanticists became
more and more abstract, often resembling first-order logic, and those structures plus
the syntactic rules to get from there to surface structure often seemed wild . In one
famous example, the underlying structure for Floyd broke the glass had clauses.5
The impact of philosophy and logic on semantics in linguistic work of the 50's
and 60's was limited; many linguists knew some first-order logic, aspects of which
began to be borrowed into linguists semantic representations, and there was
gradually increasing awareness of the work of some philosophers of language.
Generative semanticists in the late s and early s in particular started giving
serious attention to issues of logical form in relation to grammar, and to propose
ever more abstract underlying representations intended to serve simultaneously as
unambiguous semantic representations and as input to the transformational mapping
from meaning to surface form (see, for instance, Bach 1968; Fillmore 1968; Lakoff 1968,
1971, 1972; McCawley 1968b, 1970; Karttunen 1969). (But these semantic
representations were generally not suggested to be in need of further interpretation,
and truth-conditions and entailment relations were never explicitly mentioned as an

4 ‚ctually, phenomena with sloppy identity anaphora show that even with proper names, the reflexive
transformation preserves only truth value, not meaning (Keenan 1971); the reflexive pronoun is
semantically like a bound variable , not like a co-referential copy.
5 What seemed wild then might not now the shocking number of clauses or in Ross s deep
structure for Floyd broke the glass does not come close to the number of functional projections that now
intervene between various pairs of familiar syntactic categories in respected generative analyses such as
Cinque (1999).
A brief history of the syntax-semantics interface 9

object of study in the indigenously linguistic traditions that existed before formal
semantics came into linguistics in the 1970's.)
Interpretive semantics (Chomsky 1971; Jackendoff 1972) kept syntax close to
classical transformational grammar, and worked on figuring out which parts of
semantic interpretation should be based on deep structure, which on surface structure,
which on something else, allowing for a more heterogeneous syntax-semantics
relationship.

6. Then Montague Suggested a Different Garden of Eden

Richard Montague (1930–1971) was a logician and philosopher whose seminal works
on language (1970a, 1970b, 1973) founded the theory known after his death as
Montague grammar, one of the main starting points for the field of formal semantics.6
The Fregean principle of compositionality was central to Montague s theory and
remains central in formal semantics.

(10) The Principle of Compositionality:

The meaning of any complex expression is a function of the meanings of its parts
and of the way they are syntactically combined.

Montague developed a theory in his paper Universal grammar (Montague

1970b) in which the syntax is an algebra, the semantics is an algebra, and there must be
a homomorphism mapping the syntax onto the semantics. The nature of the elements
of both the syntactic and the semantic algebras is open to variation; what is constrained
by compositionality is the relation of the semantics to the syntax.
Details of Montague s own analyses of the semantics of English have in many
cases been superseded, but in overall impact, his last paper, The proper treatment of
quantification in ordinary English PTQ (Montague 1973) was as profound for
semantics as Chomsky s Syntactic Structures was for syntax. Bach (1989) summed up
their cumulative innovations thus Chomsky s Thesis was that English can be
described as a formal system; Montague's Thesis was that English can be described as
an interpreted formal system.
So what did the syntax-semantics interface look like in Montague s work, in
particular in the fragment of English in PTQ? Syntactic rules built up expressions from

6 For more on the antecedents to Montague s work within logic and philosophy, see Cocchiarella (1981),
Stokhof (2006) and Partee (2011). For more details about Montague s work, see Partee (1996, 2004, 2005)
and Partee with Hendriks (1997).
10 B.H. Partee

their parts; and for each syntactic rule there was a semantic rule that specified how the
meanings of the parts were to be combined to get the meaning of the whole. Syntax is
an algebra of forms , semantics is an algebra of meanings , and there is a
homomorphism mapping the syntactic algebra into the semantic algebra. The crucial
structure for syntax is the derivation tree , showing what parts have been combined
at each step, by what syntactic rule (See Dowty, Wall and Peters 1981).
In principle, a syntactic rule might involve a number of syntactic operations ,
e.g. concatenation of NP with VP plus agreement, to form a sentence. The Yes-no
question rule for English might involve Subject – Aux inversion plus do-support (Subj
– ‚ux inversion might be a Macro which English syntax uses in several different
Rules . Subject – ‚ux inversion would not itself be thought of as a Rule , and it has
no semantic interpretation).
Montague was not a syntactician. He had some interesting syntactic insights, but
he had no independent interest in syntax. For him, syntax was a necessary basis for
semantic structure syntax should provide the relevant part – whole structure for
compositionality to work.
‚ natural generalization of Montague s approach was suggested by Emmon ‛ach
(1979, 1984) think of a grammar as a simultaneous recursive definition of n-tuples
for larger expressions from the n-tuples for their parts, where the n-tuple consists of a
phonetic realization, a phonological representation, a morphological representation, a
syntactic representation, and a semantic interpretation. ‛ach and Wheeler s (1981)
Montague phonology showed what this would mean in phonology and
morphophonemics. This is a very pretty view of the architecture the main challenge
it faces is to account for mismatches between the alignment of structures on different
levels.

7. Towards a Synthesis of Montague and Chomsky

Montague s work was first introduced to linguists by Partee (1973, 1975b) and
Thomason (1974). Partee s goal was to find a way to combine Montague s semantics
(MG) with Chomskyan transformational grammar (TG).
An example of one major obstacle to integrating MG and TG was what to do
about deletion rules. In classical TG, as we saw above, (11a) was derived from a
structure something like (11b).

(11) a. Mary was eager to win.

b. [S Mary was eager for [S Mary to win]]

A brief history of the syntax-semantics interface 11

But given the principle of compositionality, and given the way MG works by
building up the meanings of constituents from the meanings of their subconstituents,
there is nothing that could correspond to deleting a piece of a meaning of an already
composed subpart.
Recall the consequences of the analysis in (11) for a sentence like that in (12a),
whose deep structure should classically have been something like (12b), which
would clearly give the wrong meaning.

(12) a. Everyone was eager to win.

b. [S everyone was eager for [S everyone Tns win]]

The MG-TG resolution suggested in Partee (1973, 1975b): what we want as

underlying subject in the embedded sentence was a variable which could get bound7;
I followed Montague s line and let it be bound by lambda abstraction to make a VP
type. (Others who believed in keeping an S type for the infinitive let the variable be
bound by the higher quantifier.)

(13) a. [[to win]] = λx win(x)]

b. alternatively everyone λx [x was eager for [x to win]])

In Chomskyan syntax, a corresponding change was eventually made, positing

the special null element PRO, interpreted as a bound variable, instead of the earlier
identical NP . Other syntactic theories, like Generalized Phrase Structure Grammar
(GPSG), Head-Driven Phrase Structure Grammar (HPSG), and Lexical-Functional
Grammar (LFG), and modern versions of Categorial Grammar (CG)8, were developed
after the quantifier issues had become well known, so they were designed from the
start not to run into the problems of the old Equi-NP Transformation.
One of the central questions dividing generative semantics and interpretive
semantics had been Do transformations ever change meaning? . The Katz-Postal
hypothesis had posited the answer No , and Chomsky had accepted that answer,
tentatively, in Chomsky (1965). ‛ut as the generative semanticists took that No
answer farther, Chomsky and the interpretive semanticists rejected it, and were left
without a clear picture of the nature of the syntax-semantics interface. Montague gave

7 Thanks to Paul Postal (p.c.) for reminding me that Jim McCawley was undoubtedly one of the first to
notice this point and its importance, at least as early as McCawley (1968a).
8 For GPSG, HPSG, LFG, and modern CG, see Gazdar, Klein, Pullum and Sag (1985), Pollard and Sag
(1994), Kaplan and Bresnan (1982), and Bach, Oehrle and Wheeler (1987), respectively.
12 B.H. Partee

a different kind of answer. For him, the syntactic Rules which combine parts into larger
wholes might compose phrase-structure-like operations and transformation-like
operations together. The corresponding semantic rules combine the meanings of the
parts in systematic ways to give the meaning of the whole. There could in principle be
a transformation that takes just one expression as input and changes it into some other
form. Within Montague grammar, such a rule could change the meaning as long as it
changed it systematically: as long as we could describe the meaning of the output as a
function of the meaning of the input. So, for example, in principle (probably not in
practice , there could be a negation transformation that changes an affirmative
sentence into its negative form , and the meaning of the output would just be the
negation of the meaning of the input. (In practice, natural language negation is never
that simple!)

8. Chomsky’s Resistance to Formal Semantics

My earliest work on formal semantics was concerned with how to integrate it with
Chomskyan transformational grammar (Partee 1973, 1975b). I was surprised when it
turned out that Chomsky was deeply skeptical of formal semantics and of the idea of
compositionality in any form. One of the clearest statements of his skepticism,
including a strong attack on compositionality, came in his 1974 Linguistic Institute
Golden Anniversary lecture (Chomsky 1975a), in which I was the commentator on his
paper (Partee 1975a). I have never been able to fully explain his skepticism; it has
seemed to me that it was partly a reaction to a perceived attack on the autonomy of
semantics, even though syntax is descriptively autonomous in Montague grammar. But
syntax is not explanatorily autonomous in Montague grammar, or in any formal
semantics, and I do not see any rational basis for believing that it should be. The child
learns syntax and semantics simultaneously, with undoubtedly a great deal of innate
knowledge guiding the acquisition of both. My metaphor: if syntax were
explanatorily autonomous , it should be just as easy to learn the syntax of a language
by exposure to the radio as by immersion in a real culture or by exposure to television.
I don t think there are any relevant experiments in nature I hope not! . ‛ut I would
be willing to place a bet that it is much harder, if not impossible, to learn the syntax of a
language without learning semantics simultaneously: choices of analysis in either one
affect choices in the other.
In any case, formal semantics spread and became mainstream semantics in the
US and Europe in spite of Chomsky s skepticism, and MIT hired its first formal
A brief history of the syntax-semantics interface 13

semanticist, Irene Heim, in 1989, and its second, Kai von Fintel, in 1994, and is now one
of the leading programs in formal semantics as well as syntax.

9. GPSG and the Possibility of Eliminating Transformations

One interesting possibility raised in Montague s work, a new idea for generative
linguists at the time, was that for sentences (11)–(12 , there doesn t have to be an
embedded full sentence in the syntax at all; an embedded VP is enough, and the
identity between the subject of be eager and the subject of win can be part of the
semantics of predicates like eager. Control can be considered a lexical semantic
property of verbs (Chierchia 1983). Lexical semantics can include the information that,
for instance, X promises Y to leave entails that X promises Y that X will leave; or that X
tries to win entails that X acts with the goal of bringing about a state of affairs in which
X wins.
Similarly for the relation between active and passive sentences: as Dowty (1978)
argued, all governed transformations , that is, transformations whose conditions of
application depend on the presence of a lexical item of an appropriate sort transitive
verb , approximately, in the case of passive , could be and arguably should be replaced
by lexical rules. That is, instead of a transformation mapping (14a) into the passive
(14b):

(14) a. Archaeologists have discovered a new fossil skeleton.

b. A new fossil skeleton has been discovered by archaeologists.

there should be a lexical rule as in (15):

(15) (simplified): If V is a transitive verb with syntactic frame NP1 ___ NP2 and with
meaning α, then V+ed is a (passive) verb with syntactic frame NP2 be ___ ( by
NP1), with a meaning which is the inverse of α.

In general, once we have a semantics that can do some real work, then syntax
doesn t have to try to solve problems that may be semantic in nature. The division of
labor and the nature of the interface become very interesting questions.
One of the methodological principles implicit in transformational grammar,
explicit in some (not all) versions that included the Katz-Postal hypothesis, and carried
to extremes in Generative Semantics, was the principle that sameness of meaning should
be reflected in sameness of deep structure. ‛ut with a real semantics, we don t need
14 B.H. Partee

sameness at any syntactic level, including LF , to capture sameness of meaning (cf.

Thomason 1976).
So when it was noticed that many arguments for syntactic relatedness that
motivated transformations were implicitly semantic, this led to the renewed interest in
the possibility of English as a context-free language first raised by Harman (1963),
Gazdar (1982) and subsequent work in GPSG. That is probably the principal reason for
the positive (although partial) correlation between preference for a
non-transformational syntax (GPSG, HPSG, versions of categorial grammar, etc.) with
work in formal semantics. This preference is far from absolute, however, and especially
since the s, with Heim and Kratzer among the leaders, many contemporary linguists
combine contemporary Chomskyan syntax with formal semantics.

10. Quantifier Scope Ambiguity: A Perennial Problem for the Syntax-Semantics Interface

In both transformational and non-transformational syntactic theories, the problem of

quantifier scope ambiguity has remained a difficult one: it is in fact a difficult problem
for every theory of the syntax-semantics interface. The basic problem is that if one
accepts the principle of compositionality, then an ambiguous sentence like (2b) or (16)
must have two different syntactic structures, even though there may be no
independent syntactic evidence of ambiguity.

(16) Every student read one book.

One can hardly count the number of different solutions that have been proposed
to the problem of quantifier scope ambiguity. A nice early comparative paper is
Cooper and Parsons (1976), comparing the way generative semantics, interpretive
semantics, and Montague grammar handled quantifier scope ambiguities and binding,
and providing an algorithm for converting among the three different approaches.
Here we give just a thumbnail sketch of six different approaches to quantifier
scope ambiguity from the s to the s, before the introduction of choice functions
and various non-quantificational analyses of indefinites. There are many more
approaches by now.

(i) Generative semantics (McCawley 1970; Lakoff 1971; and others): Underlying
structures were proposed that were similar to first-order logic structure, plus a
transformation of Quantifier Lowering. The perceived need to constrain derivations so
that scope corresponded to surface c-command led to Transderivational Constraints
(See debate in Lakoff 1970; Partee 1970).
A brief history of the syntax-semantics interface 15

(ii) Interpretive semantics (Chomsky 1971; Jackendoff 1972) proposed a separate

mechanism for quantifier scope; see Cooper and Parsons (1976) for a reformulation as
an indexing mechanism (Cooper and Parsons proved intertranslatibility among
approaches i, ii, and iii).

(iii) Montague s Rule-by-rule approach, with a rule of Quantifying In. Different

derivational order of quantifying in different quantifiers is automatically interpreted as
giving different scopes. Montague generates one syntactic structure via distinct
derivations; compositionality is homomorphism between syntactic and semantic
derivations.

(iv) Cooper storage: Cooper (1975) proposed a modified version of compositionality:

the semantics should compositionally derive a set of interpretations for each syntactic
structure. Working up the syntactic tree, when you reach a quantifier phrase, you can
optionally store it, then retrieve it from storage when you hit a suitable higher
node, like an S or a VP. (Scope islands represent points in the derivation where the
store must be empty.) Cooper storage was also used in GPSG.

(v) Quantifier Raising: May s (1977) Quantifier Raising rule is a syntactic rule, roughly
the inverse of the Generative Semantics rule of Quantifier Lowering. Quantifier Raising
(QR) produces a derived syntactic structure dubbed LF for logical form that s the
syntactic level that gets compositionally interpreted.

(vi) Type-shifting: On Hendriks s (1988) approach, there is no movement alternative

readings are derived via type-shifting of verbs and other functors so as to change
relative function-argument command relations.

11. Logical Form or LF in Later Generative Grammar

Generative semantics lost favor for a complex of reasons; but not because of any great
success of interpretive semantics . I will not discuss this part of linguistic history in
any detail (see Newmeyer 1980; Harris 1993), but I will state as my own opinion that
some of the semantics that was developed under the label of Logical Form or LF in
the Chomskyan school (first within the Revised Extended Standard Theory, then in
Government and Binding, then in Principles and Parameters) amounted to an
adaptation or rediscovery of various ideas from Generative Semantics but upside
down LF was not viewed by Chomskyans as a deep structure , but as a separate
syntactic level derived from Surface Structure. So where Generative Semantics had
Quantifier Lowering , May (1977) invented Quantifier Raising .
16 B.H. Partee

Chomsky himself has remained quite skeptical of whether real semantics

belongs in linguistics; he seems to have some respect for it, but he also seems inclined
to draw a line at some level such as LF: something which may have a close relation to
semantics, but it s still syntactic in form. Chomsky now seems to find much of the
compositional part of formal semantics quite reasonable, but prefers to view it as a
kind of syntax. In recent years Chomsky s deepest skepticism about formal semantics
appears to be more centered on issues of reference and on the possibility of grounding
lexical meaning in truth-conditions. Those issues, especially the problem of lexical
semantics, also remain foundational questions within formal semantics itself.

12. Later Synthesis: Heim, Kratzer Doing Formal Semantics on (Improved) LF

Both Generative Semantics and Logical Form in Chomskyan theories included

structures that were hard to make clear semantic sense of. Heim in her dissertation
(Heim 1982), and Heim and Kratzer together in their textbook (Heim and Kratzer
1998), made some simple but fundamental changes that made it possible to give a
compositional formal semantic interpretation of sentences represented at a suitable
level of Logical Form . The architecture of the syntax-semantics interface on the
model of the Heim and Kratzer textbook is that there are syntactic rules that derive a
Logical Form for each sentence, and then the compositional semantic interpretation
rules operate directly on that structure. To a classical Montague grammarian, the
unattractive feature of that approach is that it gives up on being able to
compositionally interpret the independently motivated syntactic structure of sentences.
The attractive feature is that there is a rather close match between the structure
represented by a Montagovian derivation tree and the structure of a Heim and
Kratzer LF tree.

13. Alternative Conceptions of Syntax and Semantics and Their Interface

There are many theories of syntax, and many theories of semantics, and the interface
questions look different for all of them. Jackendoff (2002) suggests a view on which
semantic structures and syntactic structures are independently generated, and the
interface conditions may be quite complex.
Among the theories which do use or could use a compositional formal semantics,
most are non-transformational. Transformational grammars, in spite of the many good
things about them and their central role in the development of modern linguistic
theory, were never computationally very tractable, nor formally elegant, nor easy to
A brief history of the syntax-semantics interface 17

work with for models of how we process language word by word. Non-
transformational grammars compatible with compositional semantics include GPSG,
HPSG, and their descendants, several modern versions of Categorial Grammar, Joshi s
Tree-‚djoining Grammar, and ‛resnan and Kaplan s Lexical Functional Grammar.
Such theories have been particularly popular within the computational linguistics
community, where great progress is being made, including much progress in
computational formal semantics.
More recently the mathematical linguist Ed Stabler has shown how one can
formalize current Minimalist Chomskyan syntax in such a way that it can be made
computationally tractable as well (Stabler 2011); that combined with compositional
semantics of the sort described by Heim and Kratzer makes that theory more
comparable to other more Montagovian theories than I had earlier thought possible.
The contemporary architectures that are probably closest to the vision of what
Montague s ideas might make possible are various kinds of Surface
Compositionality (Jacobson 1999; Barker and Jacobson 2007), employing syntactic
theories that generate surface structures directly, and interpreting them
compositionally. Bittner (2006), who is known for her work on Greenlandic Eskimo,
has done a great deal of work on formal semantics and typology, and is another
advocate of surface compositionality; her typological work has led her to propose a
number of innovations in the basic ontology of the model theory underpinning natural
language semantics, as well as a very dynamic architecture for the semantics.

14. Tectogrammatical Structure

One promising way of bringing different particular theories closer together are
approaches that invoke a tectogrammatical structure. The distinction between
tectogrammatical and phenogrammatical structure comes from Haskell Curry (1961).
Tectogrammatical structure is analogous to Montague s derivation trees it
reflects how a sentence is built up, abstracting away from language-particular
realization. David Dowty (1982) advocated linguistic recognition of the
tectogrammatical/phenogrammatical distinction, using a modified version of Montague s
analysis trees, eliminating the language-particular expressions at the nodes.
Montague s distinction between syntactic rules (such as transitive verb-plus-
object combination) and syntactic operations (such as concatenation and assignment of
accusative case) corresponds to the tecto-/pheno- distinction. Dowty suggested that the
rules, and hence the tectogrammatical structures, may well be universal. What varies
are the morpho-syntactic operations that are used in the rules.
18 B.H. Partee

Reinhard Muskens (2010) offers a version which gives a core tectogrammatical

structure for each sentence, and then with explicit mapping rules maps that common
structure homomorphically both onto a morphosyntactic structure (and a terminal
string) and a semantic interpretation. This structure in effect shows the common
algebraic structure of semantics and syntax for a language, and is compatible with
many different theories of what meanings are and of morpho-syntactic structure. As in
Montague: syntax is an algebra, semantics is an algebra, and there should be a
homomorphism between them. This is the same idea in cleaner and clearer form.

15. Concluding Remarks

Formal semantics is still a relatively young field. It has a small, but growing, presence
in fieldwork and in language typology, and a small but growing presence in
psycholinguistics and theories of on-line language processing. But it has already had
enough impact to convince linguists that it is important to study syntax and semantics
together.
In recent decades, which I mostly do not discuss here, work at the syntax-
semantics interface has been greatly enriched by the fact that it has become much more
common for linguists who want to work on either syntax or semantics to feel some
obligation to pay attention to both. Syntax remains formally autonomous on many
current approaches, but the goal of compositionality is now a widely shared goal, and
it is understood that both syntactic and semantic evidence can be important for finding
the best syntactic or semantic analysis of any phenomenon, and essential for figuring
out the best division of labor between syntax and semantics and the nature of the
interface between them.

References

Bach, Emmon. 1968. Nouns and noun phrases. In Universals in linguistic theory, ed. by Emmon Bach and
Robert T. Harms, 91–124. New York: Holt, Rinehart, and Winston.
Bach, Emmon. 1979. Montague grammar and classical transformational grammar. In Linguistics, philosophy,
and Montague grammar, ed. by Steven Davis and Marianne Mithun. Austin: University of Texas
Press.
Bach, Emmon. 1984. Some generalizations of categorial grammars. In Varieties of formal semantics, ed. by
Fred Landman and Frank Veltman, 1–24. Dordrecht: Foris.
Bach, Emmon. 1989. Informal lectures on formal semantics. New York: State University of New York Press.
A brief history of the syntax-semantics interface 19

Bach, Emmon, Richard Oehrle, and Deidre Wheeler (eds.). 1987. Categorial grammars and natural language
structures. Dordrecht: Reidel.
Bach, Emmon, and Deirdre Wheeler. 1981. Montague phonology: A first approximation. In University of
Massachusetts Occasional Papers (UMOP 7), ed. by Wynn Chao and Deirdre Wheeler, 27–45.
Amherst, MA: GLSA, University of Massachusetts.
Barker, Chris, and Pauline Jacobson. 2007. Introduction to direct compositionality. In Direct
compositionality, ed. by Chris Barker and Pauline Jacobson. Oxford: Oxford University Press.
Bernstein, Leonard. 1976. The unanswered question: Six talks at Harvard: The Charles Eliot Norton lectures
1973. Cambridge, MA: Harvard University Press.
Bittner, Maria. 2006. Ontology for human talk and thought (not robotics). Theoretical Linguistics 32:47–56.
Chierchia, Gennaro. 1983. Outline of a semantic theory of (obligatory) control. In WCCFL 2: Proceedings of
the Second West Coast Conference on Formal Linguistics, ed. by Michael Barlow, Daniel Flickinger
and Michael Wescoat, 19–31. Stanford, CA: Stanford Linguistic Association.
Chomsky, Noam. 1957. Syntactic structures. The Hague: Mouton.
Chomsky, Noam. 1965. Aspects of the theory of syntax. Cambridge, MA: MIT Press.
Chomsky, Noam. 1971. Deep structure, surface structure, and semantic interpretation. In Semantics: An
interdisciplinary reader in philosophy, linguistics, and psychology, ed. by Danny D. Steinberg and Leon
A. Jakobovits, 183–216. Cambridge: Cambridge University Press.
Chomsky, Noam. 1975a. Questions of form and interpretation. Linguistic Analysis 1:75–109.
Chomsky, Noam. 1975b. The logical structure of linguistic theory. New York: Plenum/Chicago University
Press.
Cinque, Guglielmo. 1999. Adverbs and functional heads: A cross-linguistic perspective. New York: Oxford
University Press.
Cocchiarella, Nino. 1981. Richard Montague and the logical analysis of language. In Contemporary
philosophy: A new survey, vol. 2, Philosophy of language/philosophical logic, ed. by Guttom Fløistad,
113–155. The Hague: Martinus Nijhoff.
Cooper, Robin. . Montague s semantic theory and transformational syntax. Doctoral dissertation,
University of Massachusetts at Amherst.
Cooper, Robin, and Terence Parsons. 1976. Montague grammar, generative semantics, and interpretive
semantics. In Montague grammar, ed. by Barbara H. Partee, 311–362. New York: Academic Press.
Curry, Haskell B. 1961. Some logical aspects of grammatical structure. In Structure of language and its
mathematical aspects, ed. by Roman Jakobson, 56–68. Providence: American Mathematical Society.
Dowty, David. 1978. Governed transformations as lexical rules in a Montague Grammar. Linguistic Inquiry
9:393–426.
Dowty, David. 1982. Grammatical relations and Montague grammar. In The nature of syntactic
representation, ed. by Pauline Jacobson and Geoffrey Pullum, 79–130. Dordrecht: Reidel.
Dowty, David, Robert E. Wall, and Stanley Peters, Jr. 1981. Introduction to Montague semantics. Dordrecht:
Reidel.
Fillmore, Charles. 1968. The case for Case. In Universals in Linguistic Theory, ed. by Emmon Bach and
Robert T. Harms, 1–88. New York: Holt, Rinehart and Winston.
Gazdar, Gerald. 1982. Phrase structure grammar. In The nature of syntactic representation, ed. by Pauline
Jacobson and Geoffrey Pullum, 131–186. Dordrecht: Reidel.
Gazdar, Gerald, Ewan Klein, Geoffrey K. Pullum, and Ivan A. Sag. 1985. Generalized phrase structure
grammar. Cambridge, MA: Harvard University Press.
Harman, Gilbert. 1963. Generative grammars without transformation rules: A defense of phrase structure.
Language 39:597–616.
Harris, Randy Allen. 1993. The linguistics wars. New York and Oxford: Oxford University Press.
20 B.H. Partee

Heim, Irene. 1982. The Semantics of definite and indefinite noun phrases. Doctoral dissertation, University
of Massachusetts, published 1989, New York: Garland.
Heim, Irene, and Angelika Kratzer. 1998. Semantics in generative grammar. Oxford: Blackwell.
Hendriks, Herman. 1988. Type change semantics: The scope of quantification and coordination. In
Categories, polymorphism and unification, ed. by Ewan Klein and Johan van Benthem, 96–119.
Edinburgh: Centre for Cognitive Science.
Jackendoff, Ray. 1972. Semantic interpretation in generative grammar. Cambridge, MA: MIT Press.
Jackendoff, Ray. 2002. Foundations of language: Brain, meaning, grammar, evolution. Oxford and New York:
Oxford University Press.
Jacobson, Pauline. 1999. Surface compositionality and variable-free semantics: Lecture notes of the 11th
European Summer School in Logic, Language and Information. Ms., Utrecht University, ESSLLI.
Kaplan, Ronald M., and Joan Bresnan. 1982. Lexical-functional grammar: A formal system for grammatical
representation. In The mental representation of grammatical relations, ed. by Joan Bresnan, 173–281.
Cambridge, MA: MIT Press.
Karttunen, Lauri. 1969. Pronouns and variables. In CLS 5, 108–116. Chicago: Chicago Linguistic Society,
University of Chicago.
Katz, Jerrold J., and Jerry A. Fodor. 1963. The structure of a semantic theory. Language 39:170–210.
Katz, Jerrold J., and Paul M. Postal. 1964. An integrated theory of linguistic descriptions. Cambridge, MA: MIT
Press.
Keenan, Edward L. 1971. Names, quantifiers, and a solution to the sloppy identity problem. Papers in
Linguistics 4(2):211–232.
Lakoff, George. 1968. Pronouns and reference. Parts I and II. Bloomington: Indiana University Linguistics
Club.
Lakoff, George. 1970. Repartee. Foundations of Language 6:389–422.
Lakoff, George. 1971. On generative semantics. In Semantics: An interdisciplinary reader in philosophy,
linguistics, and psychology, ed. by Danny D. Steinberg and Leon A. Jakobovits, 232–296.
Cambridge: Cambridge University Press.
Lakoff, George. 1972. Linguistics and natural logic. In Semantics of natural language, ed. by Donald
Davidson and Gilbert Harman, 545–665. Dordrecht: Reidel.
Lees, Robert B., and Edward S. Klima. 1963. Rules for English pronominalization. Language 39:17–28.
May, Robert. 1977. The grammar of quantification. Doctoral dissertation, MIT.
McCawley, James D. 1968a. Lexical insertion in a grammar without deep structure. In CLS 4: Papers from
the Fourth Meeting of the Chicago Linguistic Society, 71–80. Chicago: Chicago Linguistic Society.
McCawley, James D. 1968b. The role of semantics in a grammar. In Universals in linguistic theory, ed. by
Emmon Bach and Robert T. Harms, 124–169. New York: Holt, Rinehart, and Winston.
McCawley, James D. 1970. Where do noun phrases come from? In Readings in English transformational
grammar, ed. by Roderick A. Jacobs and Peter S. Rosenbaum, 166–183. Waltham: Ginn.
Montague, Richard. 1970a. English as a formal language. In ., Linguaggi nella Società e nella Tecnica, ed. by
Bruno Visentini et al., 189–224. Milan: Edizioni di Comunità.
Montague, Richard. 1970b. Universal grammar. Theoria 36:373–398.
Montague, Richard. 1973. The proper treatment of quantification in ordinary English. In Approaches to
natural language, ed. by K.J.J. Hintikka, J.M.E. Moravcsik and P. Suppes, 221–242. Dordrecht:
Reidel.
Muskens, Reinhard. 2010. New directions in type-theoretic grammars. Journal of Logic, Language and
Information 19:129–136.
Newmeyer, Frederick J. 1980. Linguistic theory in America: The first quarter century of transformational
generative grammar. New York: Academic Press.
A brief history of the syntax-semantics interface 21

Partee, Barbara H. 1970. Negation, conjunction, and quantification: Syntax vs. semantics. Foundations of
Language 6:153–165.
Partee, Barbara H. 1973. Some transformational extensions of Montague grammar. Journal of Philosophical
Logic 2:509–534.
Partee, Barbara H. 1975a. Comments on C.J. Fillmore's and N. Chomsky's papers. In The scope of American
linguistics, ed. by Robert Austerlitz, 197–209. Lisse: Peter de Ridder Press.
Partee, Barbara H. 1975b. Montague grammar and transformational grammar. Linguistic Inquiry 6:203–300.
Partee, Barbara H. 1996. The development of formal semantics in linguistic theory. In The handbook of
contemporary semantic theory, ed. by Shalom Lappin, 11–38. Oxford: Blackwell.
Partee, Barbara H. 2004. Reflections of a formal semanticist. In Compositionality in formal semantics: Selected
papers by Barbara H. Partee, 1–25. Oxford: Blackwell Publishing.
Partee, Barbara H. 2005. Reflections of a formal semanticist as of Feb 2005. Ms., University of
Masachusetts. [Link]
Partee, Barbara H. 2011. Formal semantics: Origins, issues, early impact. In Formal semantics and pragmatics.
Discourse, context, and models. The Baltic international yearbook of cognition, logic, and communication.
Vol. 6 (2010), ed. by Barbara H. Partee, Michael Glanzberg and Jurģis Šķilters, –52. Manhattan,
KS: New Prairie Press.
Partee, Barbara H. with Herman L.W. Hendriks. 1997. Montague grammar. In Handbook of logic and
language, ed. by Johan van Benthem and Alice ter Meulen, 5–91. Amsterdam/Cambridge, MA:
Elsevier/MIT Press.
Partee, Barbara H., Alice ter Meulen, and Robert Wall. 1990. Mathematical methods in linguistics. Dordrecht:
Kluwer.
Peters, P.S., Jr., and R.W. Ritchie. 1973. On the generative power of transformational grammars. Information
Sciences 6:49–83.
Pollard, Carl J., and Ivan A. Sag. 1994. Head-driven phrase structure grammar. Stanford and Chicago: CSLI
and University of Chicago Press.
Rosenbaum, Peter S. 1967. The grammar of English predicate complement constructions: Research monograph
47. Cambridge, MA: MIT Press.
Stabler, Edward. 2011. Computational perspectives on minimalism. Revised version. In The Oxford
handbook of linguistic minimalism, ed. by Cedric Boeckx, 617–642. Oxford: Oxford University Press.
Stokhof, Martin. 2006. The development of Montague grammar. In History of the language sciences, ed. by
Sylvain Auroux, E.F.K. Koerner, Hans-Josef Niederehe and Kees Versteegh, 2058–2073. Berlin -
New York: Walter de Gruyter.
Thomason, Richmond H. 1974. Introduction. In Formal philosophy: Selected papers of Richard Montague, ed. by
Richmond H. Thomason, 1–69. New Haven: Yale University Press.
Thomason, Richmond H. 1976. Some extensions of Montague grammar. In Montague grammar, ed. by
Barbara H. Partee, 77–118. New York: Academic Press.

To cite this article:

Partee, Barbara H. 2014. A brief history of the syntax-semantics interface in Western formal
linguistics. Semantics-Syntax Interface 1(1):1–21.
 Scalarity Determines Argument Structure
in English Particle Verbs and Slovak
Prefixed Verbs

Milada Walková*
Technical University of Ko:ice, Ko:ice, Slovakia

Abstract: The theory of situation-type aspect based on scales (e.g.

Tenny 1994; Jackendoff 1996; Rappaport Hovav 2008; Rappaport
Hovav and Levin 2010) is applied in the present study to aspectual
English verb particles and Slovak verbal prefixes. I argue that both
particles and prefixes come in two types, non-scalar and scalar,
depending on how they describe the change denoted in the verb
root. Non-scalar particles and prefixes do not alter argument
structure, e.g. The children walked (on) in silence. They appear in
predicates without an underlying scale, which are consequently
atelic. In contrast, scalar particles and prefixes appear in predicates
with an underlying scale, which are potentially telic. Scalar
prefixes and particles can alter argument structure because scales
require an obligatory realization of the measured participant, e.g.
Ally slept *(off) a headache. The study reviews manifestations of
argument-structure alternations as well as factors that determine
the effect of particles and prefixes on telicity.

Keywords: aspect, telicity, particle verbs, prefixed verbs, scalarity,

argument structure

* Email: [Link]@[Link]

The research presented in this paper was carried out during my stay at the University of Groningen. I
would like to thank Angeliek van Hout, Jack Hoeksema, two anonymous reviewers of a Dutch journal, as
well as two anonymous reviewers of this journal for their comments and suggestions on earlier versions of
this paper.

Semantics-Syntax Interface, Volume 1, Number 1, Spring 2014

22–42
© 2014 by University of Tehran
[Link]
Scalarity determines argument structure 23

1. Introduction

This paper proposes a unified account of the syntactic behaviour and semantics of
aspectual particles and prefixes in two typologically distinct languages, namely
particles in English, an analytical language, and prefixes in Slovak, an inflectional
Slavic language.1 This paper offers a new perspective to view known data rather than
attempting to provide new data. I argue that both particles and prefixes come in two
types, each with different semantic (aspectual) and syntactic (argument-structure)
effects. These two types I term non-scalar and scalar, building on the scalar theory of
aspect (e.g. Tenny 1994; Jackendoff 1996; Rappaport Hovav 2008; Rappaport Hovav
and Levin 2010; among many others).
For Slavic, the existence of two types of prefixes – internal/lexical and
external/superlexical prefixes – has been observed in Gehrke (2004), Milićević (2004),
Svenonius (2004), and di Sciullo and Slabakova (2005), for instance. In this literature,
the two types of prefixes are characterized as follows: external/superlexical prefixes,
such as the one in (1a), originate outside a verb phrase (VP), with Aspect head above V.
These prefixes have quantificational as well as aspectual meanings, marking temporal
boundedness (perfectivity), but not telicity. In contrast, internal/lexical prefixes like
(1b) originate inside a VP, with Result head below V. They cannot stack on top of other
prefixes. This type of prefixes can cause argument-structure alternations and mark the
aspectual notion of telicity. Besides aspectual meanings, they also have spatial and
idiosyncratic meanings.2

(1) a. Ivan po-čital (knigu). (Russian)

Ivan PREF-read (book)
Ivan read a book for a little while.
(from Babko-Malaya 2003, cited in Svenonius 2004:236)

1 The contribution of particles to the aspect of a predicate may depend on whether the particle has
primarily a spatial or aspectual meaning (see, for instance, McIntyre 2004 who shows that the transitivity
of certain particles depends on their spatial or aspectual use, and Giddings 2001 who shows that the
particles down and out change the aspect of predicates unless the verb denotes a directed motion event).
Therefore, in this paper, I focus on aspectual meanings only and do not address particles and prefixes in
their directional uses (in motion events), e.g. bring up fetch upwards , take in take inside , v-behnúť in-run
run into , od-letieť away-fly fly away , or non-compositional (idiomatic) uses, e.g. put off delay , pass away
die , u-tiecť PREF-flow run away , zá-vidieť PREF-see envy (see, for instance, Celce-Murcia and Larsen-
Freeman (1999:432–433) and Jackendoff (2002) for English particles).
2List of abbreviations in the glosses: ACC – accusative, IPFV – imperfective, NOM – nominative, PFV –
perfective, PREF – prefix, REFL – reflexive, SUFF – suffix.
24 M. Walková

b. Ivan na-pisal *(pisjmo). (Russian)

Ivan PREF-wrote *(letter)
Ivan wrote a letter.
(from Babko-Malaya 1999, cited in Svenonius 2004:217)

My account makes a parallel distinction, yet offers deeper insight into it by

arguing that the necessary condition for telicity marking and argument alternation
effects is underlying scalarity. Therefore, internal/lexical prefixes are scalar and
external/superlexical prefixes are non-scalar. In addition, the present account discusses
the dynamicity marking effect of prefixes on stative verbs. Competing accounts have
been proposed by Součková (2004) for Czech, Filip (2008) for Slavic, and Kagan (2012)
for Russian. I will discuss these accounts and point out the problems with them in
Section 6.
English aspectual particles have been standardly treated as marking telicity since
Brinton (1985), for example, atelic eat and telic eat up. Not all aspectual particles,
however, mark telicity. For example, the particles in drive on/along and work away (at the
problem) mark continuative or iterative aspect. In a more recent account, Cappelle and
Chauvin (2010) suggest that aspectual particles are either comparative (e.g. (anger) build
up), or resultative, (e.g. team up). However, their account suffers from the same
shortcoming as ‛rinton s it cannot be generalized over all aspectual particles, as
the particles in chatter on or sleep away, for instance, are neither comparative nor
resultative. I therefore propose that English aspectual particles also come in two types
with different effects on telicity and argument structure.
The rest of the paper is organized as follows. Section 2 introduces the scalar
theory of aspect and lays out the proposal and its predictions. Sections 3 to 5 discuss
relevant data, providing support for the predictions. Section 6 compares the proposed
account to previous accounts. And the last section concludes the paper.

2. Scalar Theory of Aspect

There are two kinds of aspectual distinctions in language (Smith 1997): grammatical
aspect or viewpoint aspect, on the one hand, and situation-type aspect or lexical aspect,
also known as Aktionsart, on the other. Grammatical aspect involves the distinction
between the perfective and the imperfective. While the perfective presents a situation3
as a whole, including its endpoint, the imperfective presents only part of a situation,

3 I use the term situation to cover both states and dynamic events.
Scalarity determines argument structure 25

not including its beginning or endpoint (Comrie 1976:4). Situation aspect, on the other
hand, involves the distinctions between dynamicity and stativity (depending on
whether a situation involves change or not, e.g. stative know, believe and dynamic play,
sing), durativity and punctuality (depending on whether a situation has duration or
not, e.g. durative walk, play and punctual enter, die), and atelicity and telicity (e.g. atelic
sing, walk and telic eat an apple, make a chair). Telicity has been described as a goal (from
Greek telos , an inherent endpoint beyond which the process cannot continue
(Comrie 1976:45), a culmination point, or a moment of temporal transition van Hout
1996:92). This paper discusses only particles and prefixes and their contribution to the
situation aspect of predicates.
To account for the aspectual contribution of English particles and Slavic prefixes,
I build on a theory according to which scalarity underlies situation aspect (more or less
explicit in Tenny 1994; Jackendoff 1996; Hay, Kennedy and Levin 1999; Smollett 2005;
Beavers 2008; Filip 2008; Rappaport Hovav 2008; Rappaport Hovav and Levin 2010, for
instance). Rappaport Hovav (2008), in particular, distinguishes two classes of dynamic
verbs, depending on the kind of change they denote. Non-scalar verbs describe a
change that is not ordered along a scale (e.g. play, laugh and exercise). Scalar verbs
denote a change ordered along a scale, which is an ordered set of degrees along a
particular dimension, such as temperature, cost, or depth. While non-scalar verbs
typically express a manner, scalar ones typically express a result (Rappaport Hovav
and Levin 2010). Scales can be open, as in lengthen, slow and ascend, or bounded, as in
flatten, enter and die, which has to do with whether the verb denotes an inherent
culmination endpoint serving as a bound on the scale or not.4 Therefore, only bounded
scales underlie telic predicates. Atelic predicates can denote either non-scalar change or
change on an open scale.
Furthermore, Rappaport Hovav (2008) distinguishes three types of scales
depending on the attribute that undergoes the scalar change (see Table 1). Path scales
are established by the theme s changing position along a path. They are lexicalized in
(i.e. lexically expressed by) various phrase or sentence constituents, such as a directed
motion verb (e.g. ascend or enter), a goal prepositional phrase (e.g. walk to the store), or a
measure phrase (e.g. walk a mile). Property scales express a change in a property of the
theme, such as brightness, temperature, or speed, and are typically lexicalized in
change of state verbs (e.g. brighten, warm, slow). Finally, volume/extent scales are
established by the volume or extent of the theme incrementally involved in the event,
realized as the direct object (e.g. build a house or eat a sandwich).

4
Rappaport Hovav (2008) also distinguishes between multi-point scales and two-point scales, giving rise to the
distinction between durative and punctual situations. Therefore, in such an account, punctual verbs like die are
also understood as scalar.
26 M. Walková

Table 1. Types of scales (based on Rappaport Hovav 2008).

TYPE OF SCALE ESTABLISHED DENOTES LEXICALIZED EXAMPLES
BY IN

PATH change of directed motion verb ascend

theme s position goal run to the line
along a path prepositional
phrase
measure phrase run a mile
PROPERTY change in change of state verb freeze
theme s property open
VOLUME/EXTENT theme s volume incremental direct object eat an apple
or extent theme read a book

It is important to stress that when the scale is lexicalized in a constituent other

than the verb, the verb itself is non-scalar and it is the goal prepositional phrase,
measure phrase, or the direct object that establishes a scale, e.g. walk, run, and eat (non-
scalar) compared respectively to walk to the beach, run a mile, and eat an apple (scalar).
This is connected to the require ment] that the participant whose property is
measured by scales be overtly realized Rappaport Hovav . ‚s a
consequence, scales can alter argument structure. For instance, intransitive eat or read
are always non-scalar; in order to become scalar, they require a scalar direct object, as
in eat an apple and read a book.
Another source of scalarity in English, besides the ones in Table 1, are complex
predicates, such as resultative phrases (2) and particle verbs (3).

(2) a. John ran the soles of his shoes bare.

b. Mary wiped the table clean.

(3) a. John ran the soles of his shoes off.

b. Mary wiped down the table.

While scalarity in resultative phrases has received some attention (e.g. Wechsler
2005), the present paper attempts to be a first step towards discussing scalarity in
particle verbs.
Taking the scalar theory of aspect as a framework, I propose that both English
aspectual particles and Slovak aspectual prefixes describe the change denoted in the
verb root as scalar or non-scalar and thus fall into two types (scalar particles/prefixes
and non-scalar particles/prefixes). The advantages of this theory relevant for an
Scalarity determines argument structure 27

account of particles and prefixes are the following: first, it exhaustively covers the data,
as it allows one to deal with both particles and prefixes that mark telicity as well as the
ones that do not (see Section 1). The proposed account not only covers both types of
particles and prefixes, but also offers an aspectual account for the existence of the two
types of prefixes.
Second, the scalar theory of situation aspect offers an explanation of argument-
structure alternations. According to Rappaport Hovav (2008:24 , scales require that
the participant whose property is measured by them be overtly realized . It follows
that a telic predicate, or scalar predicate in general, must express a measured
participant (e.g. eat and read (non-scalar) compared to eat an apple and read a book
(scalar)). Consequently, the presence of scalar particles and prefixes in a predicate
makes the realization of a measured participant obligatory.5 This participant is realized
as either the unaccusative subject (e.g. the lake froze up, the soup cooled down) or the direct
object (e.g. the nurse cooled down the soup, the child read a book through). Thus particles and
prefixes can alter the argument structure of (optionally) intransitive verbs, turning
them into obligatorily transitive particle verbs or prefixed verbs. As this also subsumes
complex predicates with unselected objects (e.g. dance the night away, sleep off a
headache), no separate treatment of these is necessary as far as aspect is concerned
(contra Jackendoff 1997).
Third, the scalar theory allows us to explain how the aspectual value of the
predicate can change with the addition (or change, in some cases) of sentence
constituents, such as a direct object or a measure phrase, as illustrated in (4)–(6).

(4) a. Jake read. (atelic)

b. Jake read a book. (atelic/telic)

(5) a. Jake ate. (atelic)

b. Jake ate an apple. (telic)

(6) a. The lake cooled. (atelic/telic)

b. The lake cooled four degrees. (telic)

In the scalar analysis, these added sentence constituents either introduce a scale
(such as a book and an apple in 4b and 5b) into the predicate, making it potentially telic,

5 Exceptions include dropping objects recoverable from context in Slovak and a few uses of some particle
verbs, typically in imperative, in English (e.g. Come on, eat up! Hey, wait up!). Interestingly, van Hout
(2000:fn. 1) gives similar Dutch examples of object drop in imperative (e.g. Eet eens op! Eat up! ).
28 M. Walková

or mark a bound on the scale already present in the predicate (as four degrees does on
the property scale lexicalized in cool in 6b), making the predicate unambiguously telic.
As particles and prefixes can change the aspectual value of predicates (e.g. atelic/telic
read a book and telic read a book through, and the Slovak counterparts atelic čítať knihu
and telic pre-čítať knihu), the scalar theory of situation aspect is a suitable framework of
analysis. Namely, scalar particles and prefixes can mark a bound on an open scale
denoted in a predicate and thus function as telicity markers (e.g. read a book can be
interpreted as involving an open scale, while through in read a book through marks a
bound on the scale). Otherwise, scalar particles and prefixes require that the predicate
be scalar, i.e. that a scale be present in the predicate, lexicalized in a suitable sentence
constituent (e.g. eat Ø/porridge/an apple but eat up *Ø/*porridge/an apple). It has to be
noted though that aspectual particles and prefixes do not lexicalize a scale themselves.
This paper treats situation aspect within the framework of lexical semantics in
which aspectual features are built up in a monotonic compositional fashion (e.g. Olsen
1994; Rappaport Hovav and Levin 1998). Thus sentence constituents can expand a
stative predicate into a dynamic one (e.g. stative John resembles his father to dynamic
John is resembling his father more and more), a non-scalar predicate into a scalar one (e.g.
non-scalar read to scalar read a book), and an atelic predicate into a telic one (e.g. atelic
run to telic run a mile). However, only certain constituents can do so, namely if they are
marked for a feature (e.g. atelic run, telic run a mile, and atelic run miles). I follow Olsen
(1994) in treating aspectual features as privative rather than equipollent. In an
equipollent binary opposition, a member is either marked or unmarked for a feature,
e.g. [+telic] and [-telic]. This markedness or unmarkedness is not cancellable. In
consequence, as Olsen points out, analyses that treat telicity as equipollent (e.g. Smith
1997:20) fail to account for the fact that a [-telic] predicate can be turned into a [+telic]
predicate upon the addition of certain sentence constituents, but not contrariwise. A
privative opposition is asymmetric; while one of the members is marked and
uncancellable, the other unmarked member can become marked by certain linguistic or
pragmatic contexts (e.g. Lee ran (unmarked for telicity) and Lee ran a mile/to the store
(marked for telicity)). Marked features ([+dynamic], [+scalar], [+telic]) determine the
overall aspect of a predicate and thus can override unmarked features ([0dynamic],
[0scalar], [0telic] . Therefore, a clash occurring between two or more constituents
with different aspectual values presents no problem (e.g. hear [0dynamic] + out
[+dynamic] = hear out [+dynamic]). Similarly, a doubled contribution of constituents is
not problematic (e.g. walk [+dynamic] + on [+dynamic] = walk on [+dynamic]). Notice
also that the doubled contribution goes beyond complex verbs (e.g. come [+telic] + to the
store [+telic] = come to the store [+telic]).
Scalarity determines argument structure 29

If particles and prefixes describe the change denoted in the predicate as either
scalar or non-scalar, as I propose, then, taking into account the monotonicity and
compositionality of aspectual features (dynamicity, scalarity, telicity), the following
predictions hold:

(7) If particles and prefixes describe the change denoted in the predicate, they cannot
appear in stative complex verbs, as states do not denote any change, either scalar or
non-scalar. Therefore, particle verbs and prefixed verbs are dynamic.

(8) Scalar particles and prefixes can alter argument structure, since scales require an
obligatory realization of the measured event participant. This effect only concerns verb
roots that are (optionally) intransitive. Non-scalar particles and prefixes do not alter
argument structure.

(9) Since the denotation of a telic predicate must contain a bounded scale, only scalar
particles and prefixes, but not non-scalar ones, can turn atelic predicates into telic ones.
Nevertheless, scalar particles and prefixes do not necessarily mark telicity, for instance,
when they refer to open scales.

The following sections support these predictions in turn with data from English
and Slovak. Similarities and differences between the two languages are pointed out.

3. Particles, Prefixes and Stativity

Since dynamicity is built up in a monotonic compositional way (see previous section), a

dynamic complex verb contains at least one constituent that is [+dynamic] (the other
constituent being [+dynamic] or [0dynamic]), while a stative complex verb contains
only constituents that are [0dynamic]. It follows that stative complex verbs can contain
stative but not dynamic verb roots, whereas stative verb roots can be part of stative or
dynamic complex verbs. However, English s tative verbs such as know, want, see, hear,
hope, resemble, etc. practically never combine with a particle , as Fraser
observes.6 The exceptions that result in dynamic particle verbs such as (10)–(11) are few
and, as pointed out to me by one of the reviewers, triggered by a change in theta roles.

6 This paper only considers prototypical stative verbs, such as verbs of perception or mental attitude, as
stative (cf. e.g. Novakov 2009). Therefore, I consider verb roots such as sleep and sit with an agentive
subject as dynamic.
30 M. Walková

(10) a. Did you hear that? (hear, stative)

b. Please, hear me out! (hear out listen to everything one wants to say , dynamic
example taken from The Free Dictionary online,
[Link]

(11) a. The garlic smells. (smell, stative)

b. The garlic smelled out the kitchen. (smell out fill with smell , dynamic

c. The dog smelled out the drugs. (smell out find by smelling , dynamic

In Slovak, prefixes freely combine with stative verbs. The resulting prefixed verbs
are dynamic with an inchoative meaning, as shown in (12)–(13).7

(12) a. Anna milovala Jozefa. (stative) (Slovak)

[Link] [Link] [Link]
‚nne loved Joseph.

b. Anna sa za-milovala do Jozefa. (dynamic) (Slovak)

[Link] REFL [Link] into [Link]
‚nne fell in love with Joseph.

(13) a. Chlapec videl mačku. (stative) (Slovak)

[Link] [Link] [Link]
The boy saw a cat.

b. Chlapec u-videl mačku. (dynamic) (Slovak)

[Link] [Link] [Link]
The boy spotted a cat.

Since complex verbs can be stative only if they contain a stative verb root and
prefixes and particles turn stative verb roots into dynamic complex verbs, I conclude
that the prediction postulated in (7) holds: particle verbs and prefixed verbs are
dynamic. This supports the hypothesis that particles and prefixes describe the change
denoted in the predicate.

7 This may be valid for other Slavic languages as well. Młynarczyk arrives at the same
conclusion for Polish prefixes.
Scalarity determines argument structure 31

4. Particles, Prefixes and Argument Structure

Particles and prefixes differ in their ability to alter argument structure, depending on
whether they are scalar or non-scalar. Non-scalar particles (e.g. in sleep away, walk on,
play around) and prefixes appear in predicates in which there is no scale established by
the theme s change of position, property, or volume/extent. They do not alter argument
structure, as shown in (14)–(15).

(14) a. John and Mary danced.

b. John and Mary danced away. John and Mary kept on dancing.

(15) a. Dieťa (si) spievalo. (Slovak)

[Link] (REFL) [Link]
The child sang for him/herself .

b. Dieťa (si) za-spievalo. (Slovak)

[Link] (REFL) [Link]
The child sang for him/herself for some time .

While non-scalar particles have a continuative meaning (e.g. Celce-Murcia and

Larsen-Freeman 1999:432), non-scalar prefixes have an attenuative a bit and/or
autobenefactive for oneself meaning. The latter meaning is additionally carried in
Slovak by the reflexive marker si, which combines with both prefixed and non-prefixed
verb roots. Si is optional in both cases, although it may be pragmatically (dis)preferred
in some contexts. I therefore do not consider (15) as a counterexample to prediction (8).
Scalar particles appear in predicates in which a scale is established by the theme s
change of position along a path, e.g. come up to someone and pri-behnúť come running ,
the theme s change in property, such as temperature, e.g. cool down and zo-hriať polievku
warm up the soup , or the theme s volume/extent incrementally involved in the event,
e.g. eat up an apple and po-staviť vežu build a tower . Scalar particles and prefixes in
these predicates express a result or goal.
Scalar particles can alter argument structure, turning an optionally transitive
verb root into an obligatorily transitive particle verb, as in (16). Particle verbs
containing scalar particles can also license a direct object unselected by the verb root, as
in (17).8 This effect depends on the transitivity of the verb root; scalar particles do not
alter argument structure of obligatorily transitive verb roots. This is shown in (18).

8 In this respect, both prefixed verbs and particle verbs behave like resultative constructions (e.g. run the
soles of your shoes *(bare)). ‚ccording to Jackendoff s analysis of the so-called time -away
32 M. Walková

(16) a. Lisa drank (her milk).

b. Lisa drank up *(her milk).

(17) a. Ally slept *a headache.

b. Ally slept off *(a headache).

(18) a. I sent *(a parcel) to my sister.

b. I sent off *(a parcel) to my sister.

Scalar prefixes have the same effect on argument structure, as shown in (19),
including unselected direct objects, as in (20), unless the verb root is obligatorily
transitive, as in (21) (see, for instance, Svenonius 2004; di Sciullo and Slabakova 2005).

(19) a. Líza pila (mlieko). (Slovak)

[Link] [Link] ([Link])
Lisa drank milk .

b. Líza vy-pila *(mlieko). (Slovak)

[Link] [Link] ([Link])
Lisa drank up *([a certain amount of] milk).

(20) a. Víly tancovali *noc. (Slovak)

[Link] [Link] [Link]
The fairies danced *the night.

b. Víly pre-tancovali *(noc). (Slovak)

[Link] [Link] ([Link])
The fairies danced * the night away.

(21) a. Gazda na poli pestoval *(zemiaky). (Slovak)

[Link] on field [Link] ([Link])
The farmer grew * potatoes on the field.

b. Gazda na poli vy-pestoval *(zemiaky). (Slovak)

[Link] on field [Link] ([Link])

construction, the unselected direct object is the direct object of the verb phrase rather than of the verb
itself. I suggest this holds not only for English scalar particles but also for Slovak scalar prefixes (see
below).
Scalarity determines argument structure 33

The farmer grew * potatoes on the field.

Slovak also has a special case of an argument structure effect that is not observed
in English. In (22b), the prefixed verb licenses an obligatory reflexive marker sa
unselected by the verb root.9 This reflexive marker refers to the overt subject argument
that measures out the event. It manifests the ability of scalar prefixes to alter argument
structure and is therefore different from the optional autobenefactive marker si that
combines with non-scalar prefixes, as in (15).

(22) a. Otec *sa jedol. (Slovak)

[Link] REFL [Link]
Father ate.

b. Otec *(sa) na-jedol. (Slovak)

[Link] (REFL) [Link]
Father ate himself full .

Such cases arise when the scale is lexicalized in the subject argument – the event
is delimited by the hunger/fullness of the father, realized as the subject argument in
(22b). Such a case is not accounted for in Rappaport Hovav (2008), which is based on
English data (cf. the English translation of (22b) by a resultative phrase). This type of
scale is difficult to classify in Rappaport Hovav s framework. While
volume/extent may seem the closest type for (22b), given the cognitive link between
volume and fullness as opposite to hunger, the same choice may be questioned for

9 Similar cases are found in German (i) and Dutch (ii). I owe (i) to Roman Sorger and (ii) to Jack Hoeksema
(pers. comm.). Notice that German, Dutch, and Slavic languages generally use reflexives where English
does not, e.g. to shave sich rasieren (German), zich scheren (Dutch), holiť sa (Slovak) but shave *oneself.

(i) a. Der Student hat *sich gelesen. (German)

the student has himself read
The student read.

b. Der Student hat *(sich) ver-lesen. (German)

the student has (himself) PREF-read
The student misread.

(ii) a. De student sliep *zich. (Dutch)

the student slept himself
The student slept.

b. De student ver-sliep *(zich). (Dutch)

the student PREF-slept (himself)
The student overslept.
34 M. Walková

Otec sa vy-spal Father slept enough . ‛oth carry a meaning component of to

satisfaction . Further examples of scalar verbs that are difficult to classify are do-jesť
finish eating , na-kúpiť to buy a lot , and na-fajčiť to make a lot of smoke by smoking .
These data suggest that Rappaport Hovav s classification of scales might need to be
adjusted for application to other languages. I tentatively assume the existence of a scale
that measures a property of the event itself, such as intensity or temporal contour. For
now, I leave the issue to future research.
An important disclaimer has to be made. Some forms, such as away, za-, and po-,
can function either as non-scalar in intransitive complex verbs or as scalar in their
transitive uses (e.g. dance away keep on dancing non-scalar)10 and dance away 5
kilos/my depression to make kilos/my depression go away by dancing scalar , za-
spievať si sing for some time non-scalar) and za-spievať si pesničku sing a song
(scalar), po-spať si sleep for some time non-scalar) and po-šliapať trávnik trample a
lawn scalar ). I therefore distinguish two uses of these forms, namely non-scalar
away1, po-1 and za-1, and scalar away2, po-2 and za-2. It also has to be noted that English
has a richer variety of non-scalar particles (about, along, (a)round, away1, on) than Slovak
(po-1 and za-1). In addition, Slovak lacks exclusively non-scalar prefixes since po- and za-
can also be used as scalar (see the examples above). In contrast, out of English non-
scalar particles, only away can also function as scalar.
To sum up, this section supports prediction (8): Scalar but not non-scalar particles
and prefixes alter argument structure.

5. Particles, Prefixes and Telicity

This section reviews the ability of particles and prefixes to mark telicity. In order to test
telicity, I use the standard for/in X time adverbial test from Dowty (1979:60) and its
Slovak counterpart, the use of preposition za. Atelic predicates are compatible with for
time adverbials, but not with in time adverbials. In Slovak, adverbials without
preposition za are compatible with atelic predicates while adverbials with za are not
(ingressive reading excluded in both languages). This is shown in (23).

(23) a. Mary ran for/*in an hour. (atelic)

b. Mária bežala (*za) hodinu. (atelic) (Slovak)

[Link] [Link] (in) hour
Mary ran for/*in an hour.

10
Note that the particle verbs with non-scalar away can occur in conative frames, e.g. Billy bashed away at the piano
(example from Jackendoff 1997:540).
Scalarity determines argument structure 35

Telic predicates behave the other way round: they are acceptable with in time
adverbials, but not for time adverbials, as shown in (24).

(24) a. John built a house *for/in a month. (telic)

b. Ján po-stavil dom *(za) mesiac. (telic) (Slovak)

[Link] [Link] [Link] (in) month
John built a house *for/in a month.

Nevertheless, some predicates in English are aspectually ambiguous, allowing

for both an atelic and a telic interpretation, as in (25).

(25) The child read a book for/in 10 minutes. (atelic/telic)

Let me now review what contribution particles make to telicity of predicates.

Non-scalar particles and prefixes do not mark telicity, as shown in (26)–(27).

(26) a. The children walked in silence for/*in a long time. (atelic)

b. The children walked on in silence for/*in a long time. (atelic)

(27) a. Ján (si) (*za) hodinu spal. (atelic) (Slovak)

[Link] (REFL) (in) hour [Link]
John slept for/*in an hour.

b. Ján (si) (*za) hodinu po-spal. (atelic) (Slovak)

[Link] (REFL) (in) hour [Link]
John slept for/*in an hour.

In Filip s –203) view, predicates with po-, such as (27b), are telic because
the prefix measures the event. For her, predicates that denote events bounded in time
(e.g. walk for an hour) are telic (see also Filip 2003:59–60). In contrast, I do not consider
perfective prefixed verbs with delimitative po-, as in (27b), telic since they do not
denote an event with an inherent culmination point, but merely an event with an
arbitrary endpoint, that is, a temporally bounded process. The atelicity of these verbs is
manifested by the acceptability of occurring with for time adverbials. With an in time
adverbial, the verb has only an ingressive reading, which points to its atelicity.
Scalar particles can mark telicity, as in (28), but they do not always do so, as in
(29b), which retains both an atelic and a telic interpretation despite the presence of up.
Notice, however, that (28) denotes a volume/extent type of scale, while (29) denotes a
36 M. Walková

property type of scale (see Table 1). Indeed, the telicity marking effect of scalar
particles depends partly on the type of scale lexicalized in the predicate, as I argue in
Walková (2013).

(28) a. Lisa drank her milk for/in five minutes. (atelic/telic)

b. Lisa drank up her milk *for/in five minutes. (telic)

(29) a. Grandma warmed the soup for/in five minutes. (atelic/telic)

b. Grandma warmed up the soup for/in five minutes. (atelic/telic)

Slovak scalar prefixes mark telicity, as in (30), but, unlike English particles, they
always do so regardless of the type of scale, as shown in (31).11,12

(30) a. Líza pila mlieko (*za) päť minút. (atelic) (Slovak)

[Link] [Link] [Link] (in) five minutes
Lisa drank milk for/*in five minutes.

b. Líza vy-pila mlieko *(za) päť minút. (telic) (Slovak)

[Link] [Link] [Link] (in) five minutes
Lisa drank up the milk *for/in five minutes.

(31) a. Babka hriala polievku (*za) päť minút. (atelic) (Slovak)

[Link] [Link] [Link] (in) five minutes
Grandma warmed the soup for/*in five minutes.

11 One can assume that this is true not only for Slovak but also for other Slavic languages, as Milićević
(2004:29) finds the same for Serbian.
12 At present, I have no explanation for why scalar prefixes, unlike scalar particles, always mark telicity.
This seems to be part of the parametric variation among languages in the encoding of telicity (see e.g. van
Hout 2008 and references therein). Telicity in English is a compositional property of the verb phrase rather
than only of the verb, as it depends not only on the verb root, but also the direct object, prepositional
phrase, measure phrase, and/or particles. In contrast, Slavic languages exploit a verb-marking strategy to
encode telicity, i.e. telicity in Slavic languages is dependent on the verb only and not on other sentence
constituents. Notice also that English, unlike Slovak, allows for aspectually ambiguous predicates.
Moreover, the phenomenon is not directly related to the absence of articles in Slavic languages (note that
articles play a major role in marking telicity via quantification in English, e.g. atelic eat apples compared to
telic eat an apple), as it is not restricted to Slavic languages that lack articles, but applies also to Slavic
languages with articles, as suggested by di Sciullo and Slabakova s study of ‛ulgarian. The fact led
Slabakova (2001) to propose a lower phrase structure position for particles in English and a higher one for
prefixes in Bulgarian.
Scalarity determines argument structure 37

b. Babka zo-hriala polievku *(za) päť minút. (telic) (Slovak)

[Link] [Link] [Link] (in) five minutes
Grandma warmed up the soup *for/in five minutes.

In addition, prefixes mark not only telicity (situation aspect) but also perfectivity
(see Section 2). Perfectivity marking concerns both types of prefixes: (30b), (31b) and
also (27b) are all perfective. The interaction between situation and grammatical aspect
has an important consequence: It has to be noted that testing telicity in Slavic
languages, such as in (30a) and (31a), is problematic due to the fact that the unprefixed
verbs are imperfective. What is crucial for determining telicity is whether the predicate
denotes an inherent culmination endpoint of a situation. Imperfective aspect, however,
views only part of a situation and thus does not include its endpoint (see Section 2).
The imperfective simply does not see the endpoint, and imperfective verbs always
appear as atelic (see Smith 1997:231–234). However, I assume (30a) and (31a) to be
uncontroversially atelic, with the verbs referring to the processes of milk drinking and
soup warming without a culmination point and roughly corresponding to engage in a
drinking process and cause to become warmer , respectively.
The same applies to the so-called secondary imperfectives, that is, imperfective
verbs composed of a perfective prefixed stem and an imperfectivizing suffix, which in
a way over-rules the original perfective value of the prefixed stem (e.g. zo-hrie-vať PREF-
warm-SUFF warm imperfective ). I propose that a scalar prefix in secondary
imperfectives marks not only perfectivity but also telicity, so the prefixed stem is
perfective and telic. However, the culmination point is not visible in the imperfective
(grammatical) aspect, so a secondary imperfective verb appears as atelic. Therefore, the
existence of secondary imperfectives does not run counter to the claim that scalar
prefixes always mark telicity. Instead, grammatical and situation aspect should be
treated separately (e.g. Smith 1997; de Swart 1998; Borik and Reinhart 2004), with the
former functioning as an aspectual operator over the latter.
This section has shown that scalar but not non-scalar particles and prefixes mark
telicity, providing support for prediction (9). While the telicity marking effect is
obligatory for prefixes, it is optional for particles.

6. Comparison with Previous Accounts

I have shown that the predictions postulated in (7)–(9) hold for both English particles
and Slavic prefixes: (i) Prefixed verbs and particle verbs are dynamic. There are two
kinds of particles and prefixes with different effects on argument structure and
38 M. Walková

telicity:13 scalar but not non-scalar particles and prefixes can (ii) alter argument
structure and (iii) mark telicity. This section compares my proposal to the proposals of
other researchers.
While there exists literature on Slavic prefixes that distinguishes between two
types of prefixes (see Section 1), such a systematic distinction in English particles has
been lacking. To the best of my knowledge, the only exception is McIntyre s ,
2004, 2007) distinction between intransitive and transitive (atransitive and non-
atransitive in McIntyre s terms uses of particles. His distinction roughly corresponds
to mine. At the same time, though, McIntyre (2001:136–137) rejects a connection
between transitivity and telicity because the two do not correlate; namely, non-scalar
particles block all kinds of direct objects, not only those that delimit an event (e.g. he
played (*his guitar) along/around/away/on). In addition, McIntyre (2001:137) tentatively
proposes that particles that block direct objects may do so by blocking case assignment
by the verb. However, McIntyre does not elaborate on the idea further because he
misses an independent motivation for the different syntactic behaviour of the two
types of particles. My proposal based on scalarity might serve as such an independent
motivation. For now, I leave the issue to future research.
There are also other accounts of Slavic prefixes based on an underlying scalar
structure (e.g. Součkov{ Filip Kagan 2012). In Filip s account,
prefixes play a role in specifying the scale that underlies telicity. For her, all prefixes
imply a scale, yet only perfective prefixed verbs are telic. However, as seen in (27b),
there are also perfective prefixed verbs that are atelic. In addition, Filip views
transitivity of most prefixed verbs as motivated by underlying scalarity and claims that
prefixed verbs are intransitive when the entity measured by the scale can be retrieved
from the context. In contrast, in my account, the argument structure of prefixed verbs is
seen as being determined by scalarity or its absence.
‚long the lines of Filip s earlier work, Součkov{ discusses Czech
superlexical prefix po- and argues that its different meanings can receive a unified
analysis when understood as containing an extensive measure function of events,
meaning a little . This prefix can either measure or delimit an available scale (and
consequently the event . This scale, in Součkov{ s view, can be a path scale, a property
scale, or a temporal scale.

13 See van Hout (1996) for similar observations in Dutch (e.g. door-schrijven write on and mee-lezen read
along intransitive, atelic compared respectively to af-schrijven finish writing and uit-lezen finish
reading transitive, telic ). In my interpretation, particles door on and mee along are non-scalar and
particles such as af off and uit out are scalar. Van Hout (1996: fn. 96) also notes that door is ambiguous
between the uses, just like English away and Slovak za- and po-.
Scalarity determines argument structure 39

Kagan (2012) discusses two Russian prefixes: lexical pod- and do-, which has
properties of both lexical and superlexical prefixes. Kagan (2012 proposes that a
verbal prefix imposes a relation between two degrees on a scale, one of which is a
degree associated with the event denoted by the verbal predicate, and the other, the
standard of comparison. Namely, for pod- the degree associated with the event is
lower than the standard of comparison (e.g. pod-rasti grow up a little bit ), and for do-
the degree associated with the event is the same as the standard of comparison (e.g. do-
čitať knigu finish reading a book ). The standard of comparison is determined either by
linguistic material in the sentence or contextually.
Overall, the difference between my account on one hand and Součkov{
and Kagan (2012) on the other is that the latter accounts treat my non-scalar prefixes as
scalar ones, the scale being one of time. This seems to be an ad hoc solution, as all events
occur over a time span and therefore can be said to progress along a temporal scale. In
addition, an approach that treats all prefixes uniformly as scalar fails to explain why
the two types of prefixes have different effects on argument structure and telicity.
However, these differences can be accounted for, as I hope to have shown in this paper,
by analysing internal/lexical prefixes as scalar and external/superlexical prefixes as
non-scalar.

7. Conclusion

This paper has shown that English aspectual particles and Slovak aspectual prefixes
both come in two types which exhibit systematic behaviour in terms of dynamicity,
telicity, and argument-structure effects. These phenomena can be successfully analyzed
in terms of scalarity underlying aspect. Particle verbs and prefixed verbs are always
dynamic. They describe the change denoted in the verb root as either scalar or non-
scalar. Scalar, but not non-scalar, particles and prefixes can alter argument structure
and mark telicity.
While the data demonstrate the predictive power of a theory of situation aspect
based on scales, they also reveal that the theory, when applied cross-linguistically, has
its weaknesses which still need to be addressed. In particular, future research should
aim at discovering whether there are also other types of scales beyond the basic three
(path, property, volume/extent) and whether there is further parametric variation
among languages with respect to scales.
40 M. Walková

References

Beavers, John. 2008. Scalar complexity and the structure of events. In Event structures in linguistic form and
interpretation, ed. by Johannes Dölling, Tatjana Heyde-Zybatow and Martin Schäfer, 245–265.
Berlin: Mouton de Gruyter.
Borik, Olga, and Tanya Reinhart. 2004. Telicity and perfectivity: Two independent systems. In Preliminary
Papers of LOLA 8, ed. by László Hunyadi, György Rakosi and Enikő Tóth. Debrecen, Hungary.
Available at [Link]
[Accessed Feb 6, 2013].
Brinton, Laurel J. 1985. Verb particles in English: Aspect or aktionsart? Studia Linguistica 39(2):157–168.
Cappelle, Bert, and Catherine Chauvin. 2010. Interprétations aspectuelles des verbes à particule en anglais:
Téliques, comparatifs, résultatifs. In Approches de la scalarité, ed. by Pascale Hadermann, Olga
Inkova, Michel Pierrard and Dan Van Raemdonck, 249–281. Genève: Droz.
Celce-Murcia, Marianne, and Diane Larsen-Freeman. 1999. The grammar book. An ESL/EFL teacher’s course.
Boston: Heinle and Heinle.
Comrie, Bernard. 1976. Aspect. Cambridge: Cambridge University Press.
Dowty, David R. 1979. Word meaning and Montague grammar. The semantics of verbs and times in generative
semantics and in Montague’s PTQ. Dodrecht: D. Reidel Publishing Company.
Filip, Hana. 1999. Aspect, eventuality types and nominal reference. New York: Routledge, Taylor and Francis
Group (Garland).
Filip, Hana. 2003. Prefixes and the delimitation of events. Journal of Slavic Linguistics 11(1):55–101.
Filip, Hana. 2008. Events and maximalization. In Theoretical and crosslinguistic approaches to the semantics of
aspect, ed. by Susan Rothstein, 217–256. Amsterdam and Philadelphia: John Benjamins.
Fraser, Bruce. 1976. The verb-particle combination in English. New York, San Francisco and London:
Academic Press.
Gehrke, Berit. 2004. Adverbial functions of Slavic prefixes. Paper presented at the Workshop on event
structure in linguistic form and interpretation, Leipzig. March 17–19, 2004. Available at
[Link] [Accessed March 10, 2013].
Giddings, Catherine. 2001. What it means to be down and out: The semantics of particles in English. In
Reading Working Papers in Linguistics 5, ed. by Michalis Georgiafentis, Paul Kerswill and
Spyridoula Varlokosta, 155–173. Reading: University of Reading.
Hay, Jennifer, Christopher Kennedy, and Beth Levin. . Scalar structure underlies telicity in degree
achievements . In Proceedings of SALT 9, ed. by Tanya Mathews and Devon Strolovitch, 127–144.
Ithaca, NY: Cornell University.
van Hout, Angeliek. 1996. Event semantics of verb frame alternations: A case study of Dutch and its acquisition.
Doctoral dissertation, Tilburg University. Published in 1998. New York and London: Garland
Publishing.
van Hout, Angeliek. 2000. Projection based on event structure. In Lexical specification and insertion, ed. by
Peter Coopmans, Martin Everaert and Jane B. Grimshaw, 397–422. Amsterdam and Philadelphia:
John Benjamins.
van Hout, Angeliek. 2008. Acquiring telicity cross-linguistically: On the acquisition of telicity entailments
associated with transitivity. In Crosslinguistic perspectives on argument structure: Implications for
learnability, ed. by Melissa Bowerman and Penelope Brown, 255–278. Mahwah, NJ: Lawrence
Erlbaum.
Jackendoff, Ray. 1996. The proper treatment of measuring out, telicity, and perhaps even quantification in
English. Natural Language and Linguistic Theory 14(2):305–354.
Jackendoff, Ray. 1997. Twisting the night away. Language 73(3):534–559.
Scalarity determines argument structure 41

Jackendoff, Ray. 2002. English particle constructions, the lexicon, and the autonomy of syntax. In Verb-
particle explorations, ed. by Nicole Dehé, Ray Jackendoff, Andrew McIntyre, and Silke Urban, 67–
94. Berlin and New York: Mouton de Gruyter.
Kagan, Olga. 2012. Degree semantics for Russian verbal prefixes: The case of pod- and do-. Oslo Studies in
Language 4(1):207–243.
McIntyre, Andrew. 2001. Argument blockages induced by verb particles in English and German: Event
modification and secondary predication. In Structural aspects of semantically complex verbs, ed. by
Nicole Dehé and Anja Wanner, 131–164. Berlin: Peter Lang.
McIntyre, Andrew. 2004. Event paths, conflation, argument structure and VP shells. Linguistics 42(3):523–
571.
McIntyre, Andrew. 2007. Particle verbs and argument structure. Language and Linguistics Compass 1(4):350–
397.
Milićević, Nata:a. . The lexical and superlexical verbal prefix iz- and its role in the stacking of prefixes.
Nordlyd 32(2):279–300.
Młynarczyk, ‚nna. . ‚spectual pairing in Polish. Doctoral dissertation, Utrecht University.
Novakov, Predrag. 2009. Dynamic-stative distinction in English verbs. Zbornik Matice srpske za filologiju i
lingvistiku 52(2):187–195.
Olsen, Mari Broman. 1994. The semantics and pragmatics of lexical aspect features. Studies in the Linguistic
Sciences 24:361–375.
Rappaport Hovav, Malka. 2008. Lexicalized meaning and the internal temporal structure of events. In
Theoretical and crosslinguistic approaches to the semantics of aspect, ed. by Susan Rothstein, 13–42.
Amsterdam and Philadelphia: John Benjamins.
Rappaport Hovav, Malka, and Beth Levin. 1998. Building verb meanings. In The projection of arguments:
Lexical and compositional factors, ed. by Miriam Butt and Wilhelm Geuder, 97–134. Stanford: CSLI
Press.
Rappaport Hovav, Malka, and Beth Levin. 2010. Reflections on manner/result complementarity. In Syntax,
lexical semantics, and event structure, ed. by Malka Rappaport Hovav, Edit Doron and Ivy Sichel,
21–38. Oxford: Oxford University Press.
di Sciullo, Anne M., and Roumyana Slabakova. 2005. Quantification and aspect. In Perspectives on aspect,
ed. by Henk Verkuyl, Henriette de Swart and Angeliek van Hout, 61–80. Dordrecht: Springer.
Slabakova, Roumyana. 2001. Telicity in the second language. Amsterdam and Philadelphia: John Benjamins.
Smith, Carlota S. 1997. The parameter of aspect. Dordrecht: Kluwer.
Smollett, Rebecca. . Quantized direct objects don t delimit after all. In Perspectives on aspect, ed. by
Henk Verkuyl, Henriette de Swart and Angeliek van Hout, 41–59. Dordrecht: Springer.
Součkov{, Kateřina. . There is only one po-. Nordlyd 32(2):403–419.
Svenonius, Peter. 2004. Slavic prefixes inside and outside VP. Nordlyd 32(2):194–242.
de Swart, Henriette. 1998. Aspect shift and coercion. Natural Language and Linguistic Theory 16(2):347–385.
Tenny, Carol. 1994. Aspectual roles and the syntax-semantics interface. Dordrecht: Kluwer.
Walková, Milada. 2013. The aspectual function of particles in phrasal verbs. Doctoral dissertation,
University of Groningen.
Wechsler, Stephen. 2005. Resultatives under the event-argument homomorphism model of telicity. In The
syntax of aspect—deriving thematic and aspectual interpretation, ed. by Nomi Erteschik-Shir and Tova
Rapoport, 255–273. Oxford: Oxford University Press.
42 M. Walková

To cite this article:

Walková, Milada. 2014. Scalarity determines argument structure in English particle verbs and
Slovak prefixed verbs. Semantics-Syntax Interface 1(1):22–42.
 Persian Complex Predicates: How
Compositional Are They?

Pollet Samveliana* and Pegah Faghirib

a,b Université Sorbonne Nouvelle & CNRS, Paris, France

Abstract: In this paper, we address the issue of compositionality of

Persian complex predicates. We argue against the “radical” or
“fully” compositional views prevailing in the studies within the
generative framework. We further show that any a priori
compositional account of Persian complex predicates is doomed to
failure, since the respective contribution of the verb and the
preverbal element cannot be determined a priori and irrespective of
the other member. We claim, nevertheless, that a compositional
account of these combinations is possible, provided
compositionality is defined a posteriori as in idiomatically
combining expressions. We then present a Construction‐based
account that allows to account for the productivity of Persian
complex predicates in a way which does justice to their
compositionality without overlooking their idiomaticity.

Keywords: Persian complex predicates, compositionality,

idiomaticity, Construction Grammar

1. Introduction

Although Persian complex predicates (CPr) have been a focus of interest for almost
thirty years now, most of theoretical studies have unfortunately explored a limited set
of data. As a consequence, their empirical underpinnings are not reliable and their

* Email: [Link]@univ‐[Link]

A Part of the work reported here was presented at the 39th Annual Meeting of the Berkeley Linguistics Society
(Berkeley, 2013). We thank the audience for their questions and suggestions. This work is related to the
work package LR4.1 of the Labex EFL, a public grant funded by the French National Research Agency
(ANR) as part of the “Investissements d’Avenir” program (reference: ANR‐10‐LABX‐0083).

Semantics‐Syntax Interface, Volume 1, Number 1, Spring 2014

43–74
© 2014 by University of Tehran
[Link]
44 P. Samvelian and P. Faghiri

generalizations turn out to be inaccurate when a larger set of data is taken into account.
The problem is particularly sharp when it comes to the issue of compositionality, one
of the most discussed questions in recent studies. Since Karimi‐Doostan’s (1997)
seminal work on light verb constructions in Persian, a “radical” or “fully”
compositional view of Persian CPrs has progressively become prevailing in the studies
within the generative framework. Despite substantial variations, these studies all rely
on the assumption that not only the verb and the non‐verbal element of a CPr have
each a precise contribution to its meaning but also that their contribution is defined a
priori and remains consistent through all their combinations to form a CPr. Divergences
only rise with respect to the “division of labor” between the components and between
the lexicon or on the syntax.
Although these studies do not deny the idiomaticity of Persian CPrs, they
nevertheless claim that at least the argument structure and the eventive/aspectual
properties of a CPr can be deduced on the basis of its component parts.
In this paper, after discussing one of the recent and most influential
“constructionist” analyses of Persian CPrs, namely Folli, Harley and Karimi (2005), we
show that none of the generalizations put forward by the authors hold a close scrutiny
and that contrary to their claim, the respective contribution of the verb and the
preverbal element of a CPr cannot be determined a priori and irrespective of the other
member. We further claim that any a priori compositional account faces similar
problems.
We argue, nevertheless, that a compositional account of the Persian CPrs is
possible, provided compositionality is defined a posteriori as in idiomatically combining
expressions in the sense of Nunberg, Sag and Wasow (1994). We then present a
Construction‐based compositional account of Persian CPrs developed by Samvelian
(2012) and Samvelian and Faghiri (in press) and implemented in the PersPred database
(Samvelian and Faghiri 2013; Samvelian, Faghiri and El Ayari 2014).

2. Folli, Harley and Karimi’s (2005) Account of Persian CPrs

The compositionality of Persian CPrs may seem strikingly obvious when examples
such as those in (1) and (2) are considered: 1

1 Glosses follow the Leipzig Glossing Rules ([Link]/lingua/resources/glossing‐[Link]). The

following non‐standard abbreviations are used for clarity: COP = copula, DOM = differential object marking,
EZ = Ezafe.
Persian complex predicates 45

(1) a. sili=e sārā be omid

slap=EZ Sara to Omid
‘Sara’s slap to Omid’

b. sārā be omid sili zad

Sara to Omid slap [Link].3SG
‘Sara slapped Omid.’

c. omid az sārā sili xor‐d

Omid from Sara slap collide‐PST.3SG
‘Omid was slapped by Sara.’

(2) a. maryam (hamiše) in ettefāq=rā be yād dāšt

Maryam always this event=DOM to memory [Link].3SG
‘Maryam (always) remembered this event.’ (durative reading)

b. maryam (nāgahān) in ettefāq=rā be yād āvar‐d

Maryam suddenly this event=DOM to memory bring‐PST.3SG
‘Maryam (suddenly) remembered this event.’ (punctual reading)

In these examples:

(a) The referential meaning of the CPr and the roles assigned to the event participants
are determined by the nominal element, since the semantic participants of the CPr, as
in (1b), are identical to those of the noun within the NP it projects, as in (1a).

(b) The verb, on the other hand, determines the argument mapping, since its
substitution by another verb entails a change in the mapping of the participants and
the grammatical functions, as in (1b) and (1c).

(c) The verb seems to determine also a part of the aspectual properties of the CPr, since
the verb alternation in (2), dāštan ‘to have’ vs. āvardan ‘to bring’, gives rise to an
aspectual contrast.

Building on data comparable to those in (1) and (2), Karimi‐Doostan (1997)

provides one of the first serious attempts to formalize the respective contribution of the
verb and the non‐verbal element in CPr formation. Based on But’s (1995) work on
argument structure, Karimi‐Doostan proposes a projectionist account in terms of
argument “fusion” or “reformation”. Following Grimshaw and Mester (1988), he
assumes that the light verb (LV) does not assign theta‐roles and thus does not have an
argument structure. However, it displays aspectual properties and assigns an aspectual
role. Being thematically defective, the LV must combine with another element, namely
the preverbal element of the predicate, to develop into a syntactically and semantically
46 P. Samvelian and P. Faghiri

complete verb. This combination gives rises to two kinds of CPrs, either compositional
or non‐compositional. The first kind results from the combination of the LV with a
predicative noun, that is, a noun displaying an argument structure, such as sili ‘slap’ in
(1). Non‐compositional CPrs are formed when the LV combines with a “thematically
opaque” noun, a noun that does not display an argument structure. yax zadan ‘to
freeze’ (Lit. ‘ice hit’), qofl kardan ‘to lock’ (Lit. ‘lock do’) and āb dādan ‘to water’ (Lit.
‘water give’) are examples of non‐compositional CPrs. CPr formation involves the
fusion of the information encoded in the respective lexical entries of the verb and the
noun.
LVs are divided into three categories with respect to their aspectual properties:
Initiatory, such as dādan ‘to give’, Transition, such as xordan ‘to collide, to strike’ and
Stative, such as dāštan ‘to have’. Some verbs may belong to more than one category, for
example kardan ‘to do’, which is either Initiatory or Transition, and thus has two lexical
entries. The aspectual category of the LV determines the aspectual type of the CPr.
Initiatory verbs form CPrs with at least one external argument, that is, either unergative
or transitive (or causative) CPrs. They are thus compatible with nouns having at least
one external argument, which refers to the initiator of the action denoted by the CPr
and is associated to the aspectual role Initiatory assigned by the verb. Transition verbs
form CPrs with a single internal argument, or, in other words, unaccusative predicates.
They are thus compatible with nouns having at least one (internal) argument. The
latter is mapped into the subject function and receives the Patient role. A mapping rule
ensures the correct association between an LV and a preverbal element. For instance, a
noun like šekast ‘defeat’, which assigns Agent and Patient thematic roles, can either
combine with an Initiatory verb or a Transition verb. In the first case, its external
argument (i.e. the Agent) is mapped into the subject function, as in (3a), while in the
second case, it is the internal argument that becomes the subject, as in (3b).

(3) a. ali sāsān=rā šekast dād

Ali Sasan=DOM defeat [Link].3SG
‘Ali defeated Sasan.’

b. sāsān az ali šekast xor‐d

Sasan from Ali defeat collide‐PST.3SG
‘Sasan was defeated by Ali.’

As mentioned previously, Karimi‐Doostan’s (1997) analysis postulates a

categorical distinction between compositional and non‐compositional CPrs. Non‐
compositional CPrs are disregarded, even though, a part of the regularities mentioned
above also hold for them.
Persian complex predicates 47

Megerdoomian (2001, 2012) and Folli et al. (2005) are representative examples of
constructionist approaches to Persian CPrs. Based on work by Hale and Keyser (1993,
2002) and Borer (1994), these studies claim that the syntactic and the semantic
properties of a CPr are derived from the syntactic construction in which the verb and
the preverbal element are inserted, and not from their respective lexical entries. A fully
compositional approach is thus maintained, but the burden shifts from the lexicon to
the syntax.
In Folli et al.’s (2005) approach, the verb in the CPr realizes the v head in Hale
and Keyser’s approach, as illustrated in (4)–(6), corresponding respectively to gerye
kardan ‘to cry’ (Lit. ‘cry do’), bidār šodan ‘to awake’ (Lit. ‘awaken become’) and bidār
kardan ‘to awaken’ (Lit. ‘awaken make’). Persian CPrs are thus the non‐incorporated
counterpart of verbal constructions suggested by Hale and Keyser.

(4) vP

DP v’

Kimea N v

gerye ‘cry’ kard ‘did’

‘Kimea cried.’
(Folli et al. 2005:1374, (15b))

(5) vP

AP v

DP A

Kimea bidār šod

‘awake’ ‘became’
‘Kimea awoke.’
(Folli et al. 2005:1375, (16b))
48 P. Samvelian and P. Faghiri

(6) vP

DP v’

Papar AP v

D A

Kimea‐ro bidār kard

‘awake’ ‘made’
‘Papar awakened Kimea.’
(Folli et al. 2005:1375, (17b))

In this approach, the thematic role of Agent/Cause is assigned by v to its external

argument (Kratzer 1996; Marantz 1997): kardan, in (4) and (6), but not šodan, in (5), form
an agentive predicate. In other words, in Folli et al.’s approach, the LV, being the
lexical realization of v, is responsible for the agentive properties of the CPr, while the
non‐verbal element plays no role:
“(…) the Agent‐selecting properties of any given light verb are consistent across
all Complex Predicates formed with a given LV.” (Folli et al. 2005:1376)
“(…) it is clear that agentivity is a property of the LV in the CPr, and never
depends on the nature of the NV element selected.” (Folli et al. 2005:1377)2
Note that Megerdoomian (2001) claims the same:
“(…) the choice of the light verb determines whether an external argument is
projected.” (Megerdoomian 2001:69)
This claim is supported by the fact that changing the verb in a CPr entails a
change in the mapping between the arguments and the grammatical functions, as
illustrated in (3).
Following Bashiri (1981), it is claimed that the verb determines also some
aspectual properties of the CPr, namely dynamic vs. stative and durative vs. punctual:
“In addition to determining whether the CPr is causative and its external
argument is agentive, the light verb distinguishes between eventive and stative CPrs.”
(Folli et al. 2005:1378)
“Normally the eventiveness of a complex predicate depends on the light verb
involved and not on the non‐verbal element.” (Folli et al. 2005:1379)

2 Folli et al. (2005) notice, however, an exception here: Predicates formed with movement verbs such as
āmadan ‘to come’ or gozaštan ‘to pass’ may be agentive or not, depending on the non‐verbal element. This is
due to the fact that in most languages, movement verbs may have both, an inchoative/unaccusative or an
agentive/unergative reading. The agentivity of the CPr depends on the selected reading of the verb.
Persian complex predicates 49

“Another property that partially depends on the LV is the duration of the CPr.”
(Folli et al. 2005:1379)
This explains the aspectual contrast between (2a) and (2b) above. Like Agent‐
selection, aspectual/eventive properties of a given LV are consistent through all its
combinations to form a CPr. For instance, dāštan ‘to have’ is always stative.
The non‐verbal element, the authors claim, determines the aktionsart properties of
the CPr, that is, its telicity, and its referential meaning:
“(…) the telicity of the CPr is dependent on the non verbal element involved, in a
very transparent fashion.” (Folli et al. 2005:1374)
More precisely, if the non‐verbal element of the CPr is a prepositional phrase, a
particle, an adjective or an eventive noun, the CPr is telic, otherwise, that is, if the non‐
verbal element is a non‐eventive noun, the CPr is atelic.

Table 1. Telic and atelic complex predicates (Adapted from Folli et al. 2005:1386, Table 3)
Telic Complex Predicates
PP + LV: be donyā āmadan ‘to be born’ (Lit. ‘to world come’), be ātaš
kešidan ‘to put on fire (Lit. ‘to fire pull’)
Particle + LV: kenār āmadan ‘to get along’ (Lit. ‘side come’)
Adjective + LV: derāz kešidan ‘to lay down’ (Lit. ‘long pull’)
Eventive noun + LV: šekast xordan ‘to be defeated’ (Lit. ‘defeat collide’),
šekast dādan ‘to defeat’ (Lit. ‘defeat give’)
Atelic Complex Predicates
Non‐eventive noun + LV: dast xordan ‘to get touched’ (Lit. ‘hand collide’)
kotak xordan ‘to get beaten’ (Lit. ‘beating hit’)
dād zadan ‘to yell’ (Lit. ‘scream hit’)
dast andāxtan ‘to mock’ (Lit. ‘hand throw’)

3. Why “Radical” Compositional Approaches Are Doomed to Failure

Notwithstanding their differences, “radical” compositional approaches all face the

same set of problems, since they build on the wrong assumption that the respective
contribution of the components participating in CPr formation is consistent through all
their combinations and can be defined a priori. As it has been extensively argued in
Samvelian (2012), the same verb can give rise to CPrs with different agentive and
eventive properties. Likewise, the non‐verbal element’s contribution can vary through
its combinations with different verbs. This section provides examples of inconsistencies
in the behavior of the verb and the non‐verbal element of the CPr (Section 3.1).
50 P. Samvelian and P. Faghiri

Two more issues remain with these approaches. The first one concerns the lexical
semantics of the CPr and the second the collocational or idiosyncratic association
between the LV and the non‐verbal element. As we will see in Section 3.2, in most
cases, the precise meaning of a CPr cannot be driven from its component parts, even in
the case of semantically transparent and fully compositional CPrs. Furthermore, the
association between the LV and the non‐verbal element is not always predictable on
the basis of the properties of each member (Section 3.3).

3.1. The Non‐consistent Contribution of the CPr Component Parts

In this section we show that neither the verb nor the non‐verbal element has a
consistent and invariable contribution through its combinations to form a CPr.

3.1.1. The Non‐Consistent Contribution of the Verb

Recall that in Folli et al.’s (2005) analysis, the verb is responsible for Agent‐selection
and some aspectual properties. At this point, it should be mentioned that the only
criterion to which Folli et al. (2005) resort is the compatibility of a CPr with the adverb
amdan ‘intentionally’. Agentive CPrs, are compatible with amdan, while non‐agentive
ones are not. Note that this criterion singles out only a part of agentive predicates,
namely those whose first argument is human, while, agentive predicates are defined by
their capacity to project an external argument (human or not). Thus, appealing to this
criterion alone, which wrongfully rules out causative predicates with a non‐human
first argument, is not satisfactory. Hence, in order to obtain a reliable diagnosis of
agentivity, we will use two additional criteria: a) the possibility for the nominal
element of the CPr to be rā‐marked, and b) the availability of a passive counterpart for
the CPr.

[Link]. rā‐Marking of the Nominal Element of the CPr

Persian displays Differential Object Marking (DOM), realized by the enclitic =rā and
triggered roughly by definiteness (Lazard 1982). Consequently, the possibility for an
argument of a given verb to be rā‐marked amounts to the capacity of the latter to
project an external argument (Samvelian 2006–2007).3
As for CPrs, several studies (Samiian 1983; Ghomeshi and Massam 1994;
Samvelian 2001, 2012; among others) have noted that the nominal element of some
CPrs can be modified, quantified, determined and eventually rā‐marked under some
conditions. The relevant fact for our discussion is that rā‐marking can involve the
nominal element of “(semantically) intransitive” (unergative) CPrs, such as harf zadan

3 This explains why the analysis of =rā as an accusative case‐marker is common in the studies with the
generative framework (Karimi 1996; Ghomeshi 1997; among others).
Persian complex predicates 51

‘to talk’ (Lit. ‘talk hit’), tohmat zadan ‘to accuse’ (Lit. ‘slander hit’) and aks gereftan ‘to
photography’ (Lit. ‘photo take’).4 Consider now the followings examples:

(7) a. maryam harf zad

Maryam speech [Link].3SG
‘Maryam talked.’

b. maryam in harf‐hā=rā zad

Maryam this speech‐PL=DOM [Link].3SG
‘Maryam said these.’

(8) a. maryam kār kard

Maryam work [Link].3SG
‘Maryam worked.’

b. maryam in kār‐hā=rā kard

Maryam thihs work‐PL=DOM [Link].3SG
‘Maryam did these works.’

(9) a. maryam az omid aks gereft

Maryam from Omid photo [Link].3SG
‘Maryam took a picture/pictures of Omid.’

b. maryam aks=e avval=rā az omid gereft

Maryam photo=EZ first=DOM from Omid [Link].3SG
‘Maryam took the first picture of Omid.’

Although the CPrs in these examples combine with only one direct argument,
namely the one realized as the subject, and as such are semantically “intransitive” (or
unergative), the nominal element can nevertheless be rā‐marked. This is possible
because these CPrs project an external argument, or, in other words, are not
unaccusative. Building on extensive data, Samvelian (2006–2007) shows that in
unaccusative CPrs, the nominal element can never be rā‐marked.

4 By “semantically intransitive” we mean CPrs that select only one semantic argument, the subject of the
CPr. Consequently, the LV combines with one nominal (non‐prepositional and non‐subject) element,
which is the nominal element of the CPr itself, as opposed to “transitive” CPrs, like rang zadan ‘to paint’
and āb dādan ‘to water’, in which the LV combines with two nominal (non‐prepositional and non‐subject)
elements. In the case of the latter, the nominal element of the CPr cannot be rā‐marked.
52 P. Samvelian and P. Faghiri

(10) a. āb yax bast

water ice [Link].3SG
‘The water froze.’

b. *āb in yax=rā bast

water this ice=DOM [Link].3SG

To sum up, the possibility for the nominal element of a CPr to be rā‐marked
constitutes a sufficient condition for considering the CPr as unergative or causative and
thus projecting an external argument.5

[Link]. Availability of a Passive CPr Counterpart

In Persian, the passive voice is realized by a periphrastic construction, whose
properties are parallel to those of copula‐predicative complement constructions
(Moyne 1974; Dabir‐Moghaddam 1982; Bonami and Samvelian, in press). This
construction involves the past participle of the lexical verb and the auxiliary šodan
‘become’.

(11) a. maryam in ketāb=rā xar‐id

Maryam this book=DOM buy‐PST.3SG
‘Maryam bought this book.’

b. in ketāb (tavassot=e maryam) xar‐id‐e šo‐d

this book by=EZ Maryam buy‐PST‐PTCP become‐PST.3SG
‘This book was bought by Maryam.’

In the case of CPrs, the valency alternation corresponding to the passive

alternation can be realized in two ways, either by changing the verb of the CPr, as in (3)
above, or by realizing the periphrastic passive form of the verb in the CPr.6 Note that
these two possibilities are not mutually exclusive.

5 Note that the reverse is not true, that is, the fact that the nominal element of a CPr cannot be rā‐marked
does not necessarily imply that the CPr is unaccusative, since rā‐marking depends on a variety of
conditions a part of which are semantic or pragmatic.
6 It must be mentioned that Folli et al. (2005) categorically reject this second option, claiming that verb
alternation is the only available option to form the passive variant of a CPr (pp. 1395–1396). Samvelian
(2012) gives however several atested counterexamples. Here are some additional examples from
Bijankhan Corpora ([Link]

(i) 22 nahang az marg=e goruhi nejāt dād‐e šo‐d‐and

22 whale from death=EZ collective rescue [Link]‐PTCP become‐PST‐3PL
‘22 whales were rescued from a collective death.’
Persian complex predicates 53

(12) a. maryam be omid tohmat zad

Maryam to Omid accusation [Link].3SG
‘Maryam accused Omid.’

b. be omid tohmat zad‐e šo‐d

to Omid accusation [Link]‐PTCP become‐PST.3SG
‘Omid was accused.’

c. be omid tohmat xor‐d

to Omid accusation collide‐PST.3SG
‘Omid was accused.’

As these examples show, even intransitive (unergative) CPrs can be passivized.

This is not surprising, since, as Samvelian (2012) extensively argues, the nominal
element of a CPr displays all properties of a syntactic complement, namely a direct
object. Unaccusative CPrs, on the contrary, cannot be passivized, (Samvelian 2006–
2007).
Now we have three criteria at our disposal in order to decide whether a CPr
projects an external argument or not. As for aspectual properties, we keep the criteria
used by Folli et al., which are mainly those of Vendler (1957). In what follows, we will
provide examples of inconsistency in the contribution of a few LVs among the most
productive in CPr formation. We should insist that these inconsistencies are not
exceptional, as extensively documented by Samvelian (2012).

● The verb zadan ‘to hit’

For Folli et al. (2005), zadan is consistently agentive and is used as an example to
support the consistency of LVs with respect to Agent‐selection. However, although a
great number of CPrs formed with zadan are indeed agentive, e.g. jāru zadan ‘to broom’
(Lit. ‘broom hit’) and tohmat zadan ‘to slander’ (Lit. ‘slander hit’), there are numerous
examples of unaccusative CPrs formed with zadan, as illustrated by the following
examples:

(13) a. in nān kapak zad‐e ast

this bread mold [Link]‐PTCP [Link].3SG

(ii) 525 hektār be kāšt=e sibzamini extesās dād‐e šo‐d

525 hectare to cultivation=EZ potato assigning [Link]‐PTCP become‐PST.3SG
‘525 hectares were assigned to the cultivation of potatoes.’

(iii) agar dar vaqt‐e ezāfi hič gol=i zad‐e na‐šav‐ad (…)
if in time=EZ extra no goal=INDF [Link]‐PTCP NEG‐[Link]‐3SG (…)
‘If no goals are scored in the extra time…’
54 P. Samvelian and P. Faghiri

‘This bread has gone moldy.’

b. in dar=e āhani zud zang zad

this door=EZ iron‐made early rust [Link].3SG
‘This iron‐made door quickly rusted.’

c. āb yax zad
water ice [Link].3SG
‘The water froze.’

The subject of the CPrs involved in these examples is clearly not an Agent, but a
Patient and the Locus of the process denoted by the CPr. The three criteria mentioned
previously uncontroversially single out these CPrs as unaccusative:

‐ amdan ‘on purpose’ is not compatible with these predicates:

(14) a. #in nān amdan kapak zad‐e ast

this bread on‐purpose mold [Link]‐PTCP [Link].3SG
‘#This bread has gone moldy on purpose.’

b. #in dar=e āhani amdan zud zang zad

this door=EZ iron‐made on‐purpose early rust [Link].3SG
‘#This iron‐made door quickly rusted on purpose.’

c. #āb amdan yax zad

water on‐purpose ice [Link].3SG
‘#The water froze on purpose.’

‐ The nominal element can never be rā‐marked:

(15) a. *nān kapak‐hā=ye bozorg=rā diruz zad

bread mold‐pl=EZ big=DOM yesterday [Link].3SG

b. *dar=e āhani in zang=rā zad

door=EZ iron‐made this rust=DOM [Link].3SG

c. *āb in yax=rā zad

water this ice=DOM [Link].3SG

‐ The passive construction with šodan ‘become’ is also excluded: *kapak zade šodan,
*zang zade šodan, *yax zade šodan.
Persian complex predicates 55

Consequently, not all CPrs formed with zadan are agentive.7 The same
inconsistency is observed with respect to aspectual properties. Remember that in Folli
et al.’s analysis, the verb is also responsible for durative vs. punctual and stative vs.
dynamic distinctions. Once again, several examples go against this generalization:

‐ Dynamic CPrs: jāru zadan ‘to broom’, sili zadan ‘to slap’, qadam zadan ‘to walk’.
‐ Stative CPrs: barq zadan ‘to be shiny’, laq zadan ‘to be unsteady’.
‐ Durative CPrs: jāru zadan ‘to broom’, qadam zadan ‘to walk’.
‐ Punctual CPrs: sili zadan ‘to slap’, hads zadan ‘to guess’.

 The verb kardan ‘to do’

Folli and al. (2005) consider kardan as the lexical counterpart of an abstract v, namely
DO or CAUSE, which consequently projects an external argument, e.g. kār kardan ‘to
work’ (Lit. ‘work do’), parvāz kardan ‘to fly’ (Lit. ‘fly do’) and rang kardan ‘to paint’ (Lit.
‘paint do’). However, as for zadan, there are non‐agentive CPrs formed with kardan, as
illustrated by gir kardan ‘to get caught’ (Lit. ‘blocking do’), ādat kardan ‘to get
accustomed’ (Lit. ‘habit do’), rošd kardan ‘to grow’ (Lit. ‘growing do’), boruz kardan ‘to
reveal oneself’ (Lit. ‘revelation do’):

(16) a. mu‐hā=ye maryam lā=ye zip gir kard

hair‐pl=EZ Maryam between=EZ zip blocking [Link].3SG
‘Maryam’s hair was caught in the zip.’

b. maryam kamkam be kār kardan ādat kard

Maryam little‐by‐llittle to work do habit [Link].3SG
‘Maryam got used to working little by little.’

c. derāxt‐hā zud rošd kard‐and

tree‐PL early growing [Link]‐3PL
‘The trees grew quickly.’

d. māhiat=e vāqeʔi=aš dar raftār=aš boruz kard

nature=EZ true=her/his in behavior=her/his revelation [Link].3SG
‘Her/His real nature revealed in her/his behavior.’

For sake of space, we apply the three criteria introduced above only to gir kardan.
Their application to the other CPrs gives similar results, singling out these CPrs as
clearly “non‐agentive” or unaccusative.

7
For more examples, see Samvelian (2012) and the PersPred Database (Samvelian and Faghiri 2013).
56 P. Samvelian and P. Faghiri

(17) a. #mu‐hā=ye maryam amdan lā=ye zip gir kard

hair‐PL=EZ Maryam on‐purpose between=EZ zip blocking [Link].3SG
‘#Maryam’s hair was caught in the zip on purpose.’

b. *mu‐hā=ye maryam lā=ye zip in gir=rā kard

hair‐PL=EZ Maryam between=EZ zip this blocking=DOM [Link].3SG

c. *lā=ye zip gir kard‐e šo‐d

between=EZ zip blocking [Link]‐PTCP become‐PST.3SG

● The verb gereftan ‘to take’

Likewise, gereftan forms agentive, as in (18), as well as non‐agentive or unaccusative
CPrs, as in (19):

(18) a. maryam otāq=rā andāze gereft

Maryam room=DOM measure [Link].3SG
‘Maryam measured the room.’

b. maryam duš gereft

Maryam shower [Link].3SG
‘Maryam took a shower.’

c. maryam tasmim gereft ke az irān be‐rav‐ad

Maryam decision [Link].3SG that from Iran SUBJ‐[Link]‐3SG
‘Maryam decided to leave Iran.’

(19) a. in kār (tavassot=e maryam) anjām gereft

this work by=EZ Maryam accomplishment [Link].3SG
‘This work was done by Maryam.’

b. xāne ātaš gereft

house fire [Link].3SG
‘The house caught fire.’

c. zaxm‐hā eltiām gereft‐and

wound‐PL healing [Link]‐3PL
‘The wound healed.’

As for aspectual properties gereftan can form various types of CPrs:

- Dynamic CPrs: aks gereftan ‘to take photos’, andāze gereftan ‘to measure’.
- Stative CPrs: ruze gereftan ‘to fast’, mātam gefertan ‘to be sorrowful’.
- Durative CPrs: ruze gereftan ‘to fast’, āftāb gereftan ‘to sunbathe’.
Persian complex predicates 57

- Punctual CPrs: pāyān gereftan ‘to end’, bārān gereftan ‘to begin to rain’.

● The verb dāštan ‘to have’

Contrary to verbs like zadan and kardan, dāštan ‘to have’ is claimed by Folli et al. to form
only stative CPrs. This claim is also contradicted by the following examples:

(20) a. maryam name=rā be omid ersāl dāšt

Maryam letter=DOM to Omid sending [Link].3SG
‘Maryam sent the letter to Omid.’

b. dowlat pāyān=e jang=rā eʔlām dāšt

government end=EZ war=DOM announce [Link].3SG
‘The government announced the end of the war.’

c. maryam romān=aš=rā be mādar=aš taqdim dāšt

Maryam novel=her=DOM to mother=her dedication [Link].3SG
‘Maryam dedicated her novel to her mother.’

In all these examples, dāštan alternates with kardan without changing the
meaning of the CPr. The only notable difference between the two variants is the
language register, the variant with dāštan is used in the formal register. Note also that
contrary to Folli et al.’s claim, here the eventive properties of the non‐verbal element
are responsible for the dynamic interpretation of the CPr.

● The verb bastan ‘to close, to tie’

Bastan also can form agentive CPrs, as in (21), as well as unaccusative CPrs, as in (22).

(21) a. maryam bā omid qarārdād bast

Maryam with Omid agreement [Link].3SG
‘Maryam made an agreement with Omid.’

b. polis tazāhor‐konande‐gān=rā be golule bast

police demonstrator‐PL=DOM to bullet [Link].3SG
‘The police shot at the demonstrators.’

c. moštari‐hā jelo=ye maqāze saf bast‐and

customer‐PL front=EZ shop line [Link]‐3PL
‘The customers lined up in front of the shop.’

(22) a. dast‐hā=yaš pine bast‐e=and

hand‐PL=her/his callosity [Link]‐PTCP=COP.3PL
‘Her/his hands have become callous.’
58 P. Samvelian and P. Faghiri

b. divār kamkam xaze bast

wall little‐by‐little moss [Link].3SG
‘The wall was little by little covered by moss.’
c. āb yax bast
water ice [Link].3SG
‘The water froze.’

From the data examined in this section, it can be concluded that the contribution
of the LV is not consistent through its combination with various non‐verbal elements to
form a CPr neither with respect to agentivity nor to aspectual properties. A
compositional approach to Persian CPrs that determines the contribution of the verb a
priori and regardless of the other member of the CPr is thus doomed to failure, unless
all the above‐mentioned counterexamples are overlooked.

3.1.2. The Non‐Consistent Contribution of the Non‐Verbal Element

The contribution of the non‐verbal element of the CPr is also assumed to be consistent
by compositional approaches. In Folli et al.’s (2005) analysis, for instance, the non‐
verbal element is responsible for aktionsart. Adjectives and prepositional phrases form a
Small Clause with their complement and thus always give rise to telic CPrs
(accomplishment or achievement), while CPrs formed with a nominal element may
either be telic or atelic, depending on the noun: eventive nouns form telic and non‐
eventive nouns atelic CPrs.
In this section we will provide examples against Folli et al.’s assumption.

[Link]. The Non‐Consistency of Adjectives

The examples below show that CPrs containing an adjective are not systematically
telic.

(23) a. maryam xabar=rā az omid penhān dāšt

Maryam news=DOM from Omid hidden [Link].3SG
‘Maryam hid the news from Omid.’

b. maryam xabar=rā se sāl az omid penhān dāšt

Maryam news=DOM three year from Omid hidden [Link].3SG
‘Maryam hid the news from Omid for three years.’

c. #maryam xabar=rā dar yek sāʔat az omid penhān dāšt

Maryam news=DOM in one hour from Omid hidden [Link].3SG
(Lit.) ‘#Maryam hid the news from Omid in one hour.’
Persian complex predicates 59

(24) a. maryam omid=rā az daqdaqe dur dāšt

Maryam Omid=DOM from worry far [Link].3SG
‘Maryam kept Omid far from worries.’

b. maryam omid=rā sāl‐hā az daqdaqe dur dāšt

Maryam Omid=DOM year‐PL from worry far [Link].3SG
‘Maryam kept Omid far from worries for years.’

c. #maryam omid=rā dar panj daqiqe az daqdaqe dur dāšt

Maryam Omid=DOM in five minute from worry far [Link].3SG
(Lit.) ‘#Maryam kept Omid far from worries in five minutes.’

(25) a. maryam pul lāzem dāšt

Maryam money necessary [Link].3SG
‘Maryam needed money.’

b. maryam sāl‐hā pul lāzem dāšt

Maryam year‐PL money necessary [Link].3SG
‘Maryam needed money for years.’

c. #maryam dar panj daqiqe pul lāzem dāšt

Maryam in five minute money necessary [Link].3SG
(Lit.) ‘#Maryam needed money in five minutes.’

The CPrs used in these examples, namely penhān dāštan ‘to keep hidden’ (Lit.
‘hidden have’), dur dāštan ‘to keep far away’ (Lit. ‘far have’) and lāzem dāštan ‘to need’
(Lit. ‘necessary have’) are formed with an adjective and nevertheless are not telic.

[Link]. The Non‐Consistency of Prepositional Phrases

Like adjectives, PPs are claimed by Folli et al. (2005) to form only telic predicates. Once
again, the existence of several atelic CPrs containing a PP goes against this
generalization. In the following examples, be masxare gereftan ‘to mock’ (Lit. ‘to
mocking take’) and be ohde dāštan ‘to have the responsibility of, to be in charge of’ are
two examples of atelic CPrs formed with a PP.

(26) a. maryam omid=rā be masxare gereft

Maryam Omid=DOM to mocking [Link].3SG
‘Maryam mocked Omid.’

b. maryam omid=rā sāʔat‐hā be masxare gereft

Maryam Omid=DOM hour‐PL to mocking [Link].3SG
‘Maryam mocked Omid for hours.’
60 P. Samvelian and P. Faghiri

c. #maryam omid=rā dar yek sāʔat be masxare gereft

Maryam Omid=DOM in one hour to mocking [Link].3SG
(Lit.) ‘#Maryam mocked Omid in one hour.’

(27) a. maryam esteqbāl az dānešju‐yān=rā be ohde dār‐ad

Maryam welcome of student‐PL=DOM to charge [Link]‐3SG
‘Maryam is in charge of welcoming the students.’

b. maryam sāl‐hā esteqbāl az dānešju‐yān=rā be ohde dāšt

Maryam year‐PL welcome of student‐PL=DOM to charge [Link].3SG
‘Maryam was in charge of welcoming the students for years.’

c. #maryam dar yek sāʔat esteqbāl az dānešju‐yān=rā be ohde dāšt

Maryam in one hour welcome of student‐PL=DOM to charge [Link].3SG
(Lit.) ‘#Maryam was in charge of welcoming the students in one hour.’

[Link]. The Non‐Consistency of Nouns

The following examples show that non‐eventive nouns, contrary to Folli et al’s (2005)
claim, can form telic CPrs:

(28) a. maryam otāq=rā andāze gereft

Maryam room=DOM measure [Link].3SG
‘Maryam measured the room.’
b. #maryam otāq=rā panj daqiqe andāze gereft
Maryam room=DOM five minute measure [Link].3SG
(Lit.) #Maryam measured the room for five minutes.’
c. maryam otāq=rā dar čand sāniye andāze gereft
Maryam room=DOM in few second measure [Link].3SG
‘Maryam measured the room in a few seconds

(29) a. nān kapak zad

bread mold [Link].3SG
‘This bread went moldy.’

b. #nān se ruz kapak zad

bread three day mold [Link].3SG
(Lit.) ‘#This bread went moldy for three days.’

c. nān ye‐ruze kapak zad

bread in‐one‐day mold [Link].3SG
‘This bread went moldy in one day.’
Persian complex predicates 61

(30) a. mār pust andāxt

snake skin [Link].3SG
‘The snake sloughed off.’

b. #mār do ruz pust andāxt

snake two day skin [Link].3SG
(Lit.) ‘#The snake sloughed off for two days.’

c. mār ye‐ruze pust andāxt

snake in‐one‐day skin [Link].3SG
‘The snake sloughed off in one day.’

To conclude, almost none of the empirical generalizations on the respective

contribution of the verb and the non‐verbal element stated by Folli et al. (2005) hold
when a larger range of data is taken into account.

3.2. The Non‐Predictable Semantic Content of the CPr

The non‐predictability of the meaning of the CPr is another significant impediment to
fully compositional approaches. In order for the latter to work, the meaning of the CPr
must be derivable on the basis of the meaning of its components. However, as
mentioned in several studies (Goldberg 1996; Karimi‐Doostan 1997; Family 2006;
Bonami and Samvelian 2010; Samvelian 2012; Samvelian and Faghiri 2013; among
others), numerous Persian CPrs are semantically opaque. Moreover, as shown by
Samvelian (2012) and Bonami and Samvelian (2010), even when a CPr is semantically
transparent, it is barely ever the case that its meaning is fully predictable from the
meaning of its component parts. CPrs are subject to various levels of lexicalization, in a
way that closely parallels what is seen with lexemes formed by morphological means.8

3.2.1. Specialization
In some cases, the meaning of the CPr is a specialization of the predictable meaning of
the combination: čāqu zadan (Lit. ‘knife hit’), is not only to hit somebody with a knife
but to stab them; dast dādan (Lit. ‘hand give’) does not only imply that you put your
hand in somebody’s else hand, but that you shake hands with them; āb dādan (Lit.
‘water give’), like to water in English, is not just pouring water on something, and
finally šir dādan (Lit. ‘milk give’) is not just the action of giving milk to somebody, but
to breast‐feed them.

8 There are of course instances of CPrs with a fully predictable meaning. These are mostly CPrs formed by
a predicative noun, which conveys the entire lexical meaning of the CPr, e.g. anjām dādan ‘to realize’ (Lit.
‘accomplishment give’), tekān xordan ‘to move’ (Lit. ‘movement collide’), faryād zadan ‘to shout, to cry’ (Lit.
‘shout hit’) and entexāb kardan ‘to choose’ (Lit. ‘choice do’).
62 P. Samvelian and P. Faghiri

3.2.2. Semantic Drift

In other examples semantic drift has taken place, either by metaphor or by metonymy.
The link between the compositional meaning and the lexicalized meaning is sometimes
still recoverable synchronically. For instance, the lexicalized meaning of guš kardan ‘to
listen’ (Lit. ‘ear do’) can be recovered via metonymy: The CPr designates the
prototypical action accomplished by ears. In zanjir zadan, the elliptical element of the
meaning, sine ‘chest’, can also be recovered: the CPr comes from bā zanjir (be) sine zadan
‘to hit one’s chest with chains’ or zanjir be sine zadan ‘to hit chains against one’s chest’.
However, in numerous other cases the initial link is no longer perceivable by the
speakers. ru gereftan ‘to become cheeky’ (Lit. ‘face take’) and dast andāxtan ‘to mock’
(Lit. ‘hand throw’), for instance constitute opaque sequences in synchrony.
Folli et al. (2005) do not propose any explicit treatment for the issue raised by
these examples, namely the fact that argument structure and eventive properties does
not exhaust the meaning of a CPr and even if one assumes that they can be determined
on a constructionist basis, still they do not allow to determine the exact meaning of
most of CPrs, which can be more or less idiomatic.

3.3. The Collocational Association between the CPr Components

Another problem faced by fully compositional accounts is the collocational association
between the verb and the non‐verbal element of the CPr. Thus, it may happen that two
synonymous nouns combine with two different verbs to form synonymous CPrs, as it
is the case with rašk and hesādat, both meaning ‘envy’ or ‘jealousy’, which respectively
combine with bordan ‘to take’ and kardan ‘to do’ to form rašk bordan and hesādat kardan,
both meaning ‘to envy’ or ‘to be jealous’. Likewise, āroq ‘burp’ combines with zadan ‘to
hit’, āroq zadan ‘to burp’, while sorfe ‘cough’, which is semantically close to āroq,
combines with kardan, sorfe kardan ‘to cough’. However, such facts have either been
overlooked or minimized in the studies on Persian CPrs. Folli et al. (2005), for instance,
acknowledge this collocational association but do not address it thoroughly and state
hasty generalizations on the basis of a restricted set of examples:

(31) a. miz=o tamiz kard‐am (Folli et al. 2005:1389, (60))

table=DOM clean [Link]‐1SG
‘I cleaned the table.’

b. miz tamiz šo‐d

table clean become‐PST.3SG
‘The table got/became cleaned.’
Persian complex predicates 63

(32) a. bačča=ro hamum kard‐am (Folli et al. 2005:1389, (62))

child=DOM bath [Link]‐1SG
‘I bathed the child.’
b. *bačče hamum šo‐d
child bath become‐PST.3SG
(intended) ‘The child was given a bath.’

These examples show that while tamiz ‘clean’ and hamum ‘bath’ both combine
with kardan ‘to do’ to form agentive CPrs, only tamiz ‘clean’ combines with šodan ‘to
become’ to form an unaccusative CPr. According to Folli et al. (2005:1389), such
restrictions are not idiosyncrasies and can be accounted for on the basis of the syntactic
properties of šodan, which “selects for a predictive small clause complement, while
kardan can select for either a nominal complement (when it gets a ‘do’ meaning) or a
small clause complement (when it gets a ‘make’ reading).” This explains the success of
a kardan/šodan alternation in CPrs with a predicative non‐verbal element, and its failure
in CPrs with a nominal element. Let us now consider the following examples:

(33) a. polis maryam=rā jarime kard

police Maryam=DOM fine [Link].3SG
‘The policeman fined Maryam.’

b. maryam jarime šo‐d

Maryam fine become‐PST.3SG
‘Maryam was fined.’

(34) a. divār‐hā=rā pārsāl rang kard‐im

wall‐PL=DOM last‐year paint [Link]‐1PL
‘[We] painted the walls last year.’

b. divār‐hā=rā pārsāl rang šo‐d‐and

wall‐PL=DOM last‐year paint become‐PST‐3PL
‘The walls were painted last year.’

(35) a. maryam dar=rā qofl kard

Maryam door=DOM lock [Link].3SG
‘Maryam locked the door.’

b. dar qofl šo‐d

door lock become‐PST.3SG
‘The door was locked.’
64 P. Samvelian and P. Faghiri

These examples show that there is no ban on the combination of nominal non‐
predicative nouns with šodan to form unaccusative CPrs, which implies in turn that the
restrictions exemplified by Folli et al. in examples (31) and (32) cannot receive a
syntactic explanation, but are idiosyncratic in nature.
The same problem rises for some other restrictions noted by Folli et al., namely
dād zadan ‘to shout’ (Lit. ‘shout hit’) vs. *dād xordan (Lit. ‘shout collide’); kār kardan ‘to
work’ (Lit. ‘work do’) vs. *kār xordan (Lit. ‘work collide’). Here is the explanation
provided by the authors:
“(...) certain NV elements may not be combined with the unaccusative xordan
‘collide’, because the events that they denote can only be caused agentively – they are,
in essence, inherently unergative.”(Folli et al. 2005:1390)
No criterion is provided for the identification of inherently unergative nouns, but
let us admit that dād and kār are indeed such nouns. We note however that although kār
does not combine with xordan to form an unaccusative CPr, it nevertheless combines
with šodan ‘to become’:

(36) darbāre=ye in matlab besyār kār šo‐d‐e ast

about=EZ this subject very work become‐PST‐PTCP COP.3SG
‘A great deal of work has been done on this topic.’

Moreover, other “inherently unergative” nouns like sili ‘slap’ or kotak ‘beating’
combine also with xordan to form unaccusative CPrs:

(37) omid az sārā sili xor‐d

Omid from Sara slap collide‐pst.3sg
‘Omid was slapped by Sara.’

Consequently, the fact that dād does not combine with xordan cannot be
accounted for on the basis of its syntactic or semantic properties. This restriction is
idiosyncratic and must be stipulated.
To conclude, in this section we showed that none of the generalizations stated by
Folli et al. (2005) withstand a close scrutiny.

4. Compositionality Revisited

Despite all the problems mentioned in the previous section, we claim that a
compositional account of Persian CPrs is still possible, provided compositionality is
Persian complex predicates 65

defined a posteriori, like in idiomatically combining expressions, in the sense of

Nunberg et al. (1994). This view of Persian CPrs can be developed into a Construction‐
based approach (Goldberg 1995), where:
1) Each CPr corresponds to a Construction.
2) CPrs can be grouped in classes according to their semantic and syntactic
properties and each class can be represented by a partially fixed Construction.
3) Constructions can be structured in networks, thus accounting for different
semantic and syntactic relations between CPrs.

4.1. Persian CPrs as Idiomatically Combining Expressions

With respect to their compositionality, Persian CPrs are comparable to Idiomatically
Combining Expressions, that is, “idioms whose parts carry identifiable parts of their
idiomatic meanings” (Nunberg et al. 1994:496). This means that the verb and the non‐
verbal element of a CPr can be assigned a meaning in the context of their combination.
Thus, the CPr is compositional, in the sense that its meaning can be distributed among
its components, and yet it is idiomatic, in the sense that the contribution of each
member cannot be determined out of the context of its combination with the other one.
This is the line of argumentation developed by Nunberg et al. (1994) to support a
compositional view of expressions such as spill the beans. Although the meaning of the
sequence is conventional or idiomatic, in that it must be learned and cannot be driven
from its component parts, the sequence is still compositional, in that its meaning can be
distributed among its components. In the context of its association with beans, spill
means ‘to divulge’, and in the context of its association with spill, bean means ‘secret’.
This view of Persian CPrs is defended in recent studies by Müller (2010), Samvelian
(2012) and Samvelian and Faghiri (in press). Table 2 illustrates this point for a set of
CPrs formed with zadan. Each line contains examples of CPrs where the verb can be
assigned a meaning comparable to that of a lexical verb in English, in combination with
a set of nouns.
An important advantage of this approach is that not only it provides a fine‐
grained and empirically valid compositional analysis of Persian CPrs, but it also allows
to account for their productivity. Given the meaning assigned to zadan and the
meaning of the CPr as a whole, new combinations can be produced and interpreted.
For instance, tag zadan ‘to tag’ (Lit. ‘tag hit’), formed with the loanword tag, is created
on the basis of barčasb zadan ‘to label’ (Lit. ‘label hit’), tambr zadan ‘to stamp’ (Lit. ‘stamp
hit’), etc.
Samvelian (2012) provides a thorough study of CPrs formed with zadan ‘to hit’.
On the basis of a detailed syntactic and semantic classification, the author establishes a
66 P. Samvelian and P. Faghiri

list of abstract Constructions, each associated to a syntactic structure, argument

structure and an abstract meaning.

Table 2. Meanings of zadan in the context of its CPrs

CPrs formed with zadan (N + zadan) The meaning of zadan

kare zadan ‘to butter’, lāk zadan ‘to varnish’, rang zadan ‘to paint’, ‘to apply’
šāmpu zadan ‘to shampoo’, vāks zadan ‘to polish’, ...
adviye zadan ‘to put spice’, felfel zadan ‘to pepper’, namak zadan ‘to add’, ‘to incorporate’
‘to salt’, ...
barčasb zadan ‘to label’, dastband zadan ‘to handcuff’, lejām ‘to put’
zadan ‘to bridle’, mangane zadan ‘to staple’, tambr zadan ‘to
stamp’, ...
eynak zadan ‘to wear glasses’, kerāvāt zadan ‘to wear a tie’, māsk ‘to wear’
zadan ‘to wear a mask’, ...
javāne zadan ‘to bud’, juš zadan ‘to sprout’, kapak zadan ‘to go ‘to develop’, ‘to form’
mouldy’, šabnam zadan ‘to dew’, šokufe zadan ‘to bloom’, tabxāl
zadan ‘to develop coldsore’, tāval zadan ‘to blister’, zang zadan
‘to rust’, pine zadan ‘to become calloused’, ...
pol zadan ‘to build a bridge’, jādde zadan ‘to build a road’, sad ‘to build’
zadan ‘to build a dam’, saqf zadan ‘to build a ceiling’, ...
čādor zadan ‘to set up a tent’, dārbast zadan ‘to erect a ‘to set up’
scaffolding’, ordu zadan ‘to set a camp’, pašeband zadan ‘to put
up a fly sheet’, ...
arbade zadan ‘to yell’, dād zadan ‘to shout’, faryād zadan ‘to ‘to emit (a cry)’
shout’, jiq zadan ‘to scream’, ...
ar zadan ‘to bray’, čahčahe zadan ‘to sing (bird)’, jik zadan ‘to ‘to emit (an animal cry)’
chirp’, šeyhe zadan ‘to neigh’, ...

4.2. Persian CPrs as Constructions

A Construction, in the sense of Goldberg (1995) and Kay and Fillmore (1999), is a
conventional association between a form and a meaning. Given that Persian CPrs have
a conventional meaning, they each correspond to a Construction, and are thus
comparable to lexemes. Constructions can be of various levels of abstractness and can
be organized hierarchically, going from the most specific ones to the more abstract
ones. Samvelian (2012) applies this approach to a set of CPrs formed with the verb
zadan ‘to hit’ and a nominal element. The latter are grouped in semantic classes, with
various degrees of coherence. Each class corresponds to a partially fixed Construction.
Here are a few examples of these Constructions:
Persian complex predicates 67

(38) Spreading‐zadan Construction

N0 (be)N1 N2 V
Agent Ground Figure
‘N0 applies N2 on N1’ or ‘N0 covers the surface of N1 with N2’

(39) Instrument‐zadan Construction

N0 (be) N1 N2 V
Agent Patient Instrument
‘N0 accomplishes the typical action for which N2 is used (on N1)’

(40) Slandering‐zadan Construction

N0 be N1 N2 V
Slanderer Slanderee Slander
‘N0 accuses N1 of N2’

(41) Forming‐zadan Construction

N0 N1 V
Location Theme
‘N1 is formed on N0’

In each Construction notation:

(a) The first line gives the label of the Construction. The fact that zadan is
specified in the label implies that the Construction is partially fixed and the
verbal lexeme in the construction is one of the forms of zadan.

(b) The second line gives the valency (subcategorization frame) and the syntactic
construction of the minimal sentence including the CP. The formalism is inspired
by Gross’s (1975) Lexicon‐Grammar notation: N0 stands for the subject, N1 for
the first complement (object), N2 for the second complement (object), etc.

(c) The third line provides the mapping between the semantic roles and the
grammatical functions.

(d) The last line indicates the abstract meaning assigned to the Construction as a
whole.

Here are examples of CPrs associated with each Construction:

 Spreading‐zadan Cons: āb – ‘to wet’, āhār – ‘to starch’, kare – ‘to butter’, rang – ‘to
paint’, vāks – ‘to polish’...
68 P. Samvelian and P. Faghiri

 Instrument‐zadan Cons: jāru – ‘to broom’, mesvāk – ‘to brush one’s teeth’, otu –
‘to iron’, šāne – ‘to comb’...
 Forming‐zadan Cons: javāne – ‘to blossom’, juš – ‘to sprout’, kapak – ‘to mouldy’,
tāval – ‘to blister’, zang – ‘to go rusty’...
 Slandering‐zadan Cons: ang –, bohtān –, ettehām –, tohmat – ‘to slander’...

Note that in this approach, all lexicalized CPrs are listed in the lexicon, contrary
to Folli et al.’s approach, which claims that Persian CPrs, despite their idiomaticity,
need not to be listed.9

4.3. Constructions in Networks

We adopt furthermore a type hierarchy or an inheritance hierarchy of Constructions of
varying levels of specificity (Michaelis, to appear). Constructions are thus structured in
networks, which represent different relationships such as hyponymy/hyperonymy
(subtypes vs. supertypes), synonymy and valency alternations. Each Construction
belongs to several different Construction types. Consequently, a given Construction
type can inherit properties from multiple dominating types.

4.3.1. Subclasses and Superclasses

Some semantic classes can be grouped together into a more abstract class. In this case,
the Construction associated with them is the subtype of a less specific Construction.
For instance, the Spreading‐zadan Construction can be considered as a subtype of
Locatum Construction. Locatum verbs (Clark and Clark 1979), e.g. to paint, to salt,
incorporate a Figure (i.e. the noun to which the verb is morphologically related) and
have a Ground argument realized as an NP or a PP: ‘to paint sth’ = ‘to put paint
(Figure) on sth (Ground). As for Persian Locatum CPrs, the Figure is the nominal
element of the CPr:

(42) Locatum‐zadan Construction

N0 (be) N1 N2 V
Agent Ground Figure
‘N0 puts/incorporates N2 on/into N1’

9 Note however that this does not imply that the whole sequence is listed as a lexical entry in the lexicon.
For an implementation of this approach see Müller (2010).
Persian complex predicates 69

Figure 1. Subtypes of Locatum‐zadan Construction

The subtypes of the Locatum‐zadan Construction, that is, the Spreading‐zadan

Construction, the Incorporation‐zadan Construction and the Putting‐zadan Construction
(see Figure 1), all have an Agent and a Ground argument; the nominal element of the
CPr is the Figure. They diverge in the manner that the Figure is placed on or inside the
Ground argument. In the predicates realizing the Spreading‐zadan Construction, the
Figure is spread or applied on the surface of the Ground. The verb zadan in these CPrs
can generally alternate with the lexical verb mālidan ‘to apply’, ‘to spread’. In the CPrs
corresponding to the Incorporation‐zadan Construction, the Figure is incorporated into
the ground. The verb zadan is then synonymous to rixtan ‘to pour’. Finally, in Putting‐
zadan Construction, the Figure is placed on the Ground. These differences may entail
further ontological distinctions. For instance, at the end of the event denoted by a CPr,
which is an instance of the Incorporation‐zadan Construction, the Figure can be
confounded with the Ground or absorbed by it.

4.3.2. Synonymous Constructions

The same Construction can be realized by different verbs, e.g. kardan ‘to do’ and kešidan
‘to pull’ also form Instrumental predicates, e.g. jāru kardan and jāru kešidan ‘to broom’.
So, along with the Instrument‐zadan Construction, there are also the Instrument‐kešidan
Construction and the Instrument‐kardan Construction. These three partially fixed
Constructions are sub‐types of a more abstract Construction, with no lexically fixed
element, that is, the Instrument Construction. Synonymy rises when the same noun
occurs in the same Construction realized by different verbs.
70 P. Samvelian and P. Faghiri

Figure 2. Instrument (active) Construction

4.3.3. Valency Alternating Constructions

The same Construction can display valency alternations. For instance, in an Instrument
Construction, the Agent can be mapped to the grammatical subject and the Patient to
the grammatical object, which gives rise to an “active” Instrument Construction, or the
Patient can be mapped to the grammatical subject, forming a “passive” or
“unaccusative” Instrument Construction. Valency alternations in CPrs are often
realized by the choice of the verb: otu zadan ‘to iron’ vs. otu xordan ‘to be ironed (Lit.
‘iron collide’), ātaš zadan ‘to set fire’ vs. ātaš gereftan ‘to take fire’ (Lit. ‘fire take’).
All of the relations described above can be structured in a network using
multiple inheritance hierarchies. Figure 3 provides a partial hierarchy of Persian CPrs.

Figure 3. Partial hierarchy of Persian CPrs

Persian complex predicates 71

4.4. Class Coherence, Size and Productivity

In our approach, the productivity of the Persian CPrs is accounted for via the
analogical extension of the existing classes. It can be compositionality‐based or not. In
the first case, new combinations are created on the basis of the meaning assigned to the
Construction as a whole and to its components (see Table 2). However, we want to
defend the idea that productivity is not always compositionality‐based and that non‐
compositional Constructions (or classes) can also be productive. The Communicating‐
zadan Construction, e.g. telefon zadan ‘to phone’ (Lit. ‘phone hit’), telegrāf zadan ‘to send
a telegraph’ (Lit. ‘telegraph hit’), bisim zadan ‘to walkie‐talkie’, ‘to communicate by
means of a walkie‐talkie’ (Lit. ‘walkie‐talkie hit’), illustrates this situation. While it is
impossible to assign a meaning to zadan in these combinations, recent combinations
such as imeyl zadan ‘to email’ or esemes zadan ‘to text, to sms’ are nevertheless created by
analogical extension. We furthermore hypothesize that the productivity of Persian
CPrs is also related to other parameters such as the coherence of the classes and their
size.
The Construction associated to each class is an abstraction based on the most
salient and regularly shared properties of its members. Classes vary in their coherence
and the meaning associated to the Construction can be more or less abstract or specific.
In some classes, the meaning of each CPr can be straightforwardly derived from the
meaning associated to the Construction. This is the case in the Spreading‐zadan
Construction: For each member the paraphrase ‘N0 applies N2 on N1’ provides its
meaning and its syntactic properties. This situation yields a fully compositional class,
where the verb is comparable to a semi‐lexical or lexical verb and the relation between
the verb and the nominal element is comparable to the semantic selection of an
argument by a verb. One could even consider the creation of a lexical entry for the verb
zadan with the meaning of ‘to apply’, especially with respect to the fact that zadan
alternates with mālidan ‘to apply’ in these combinations. Consequently, the class is
highly productive and listing all potential combinations, apart from lexicalized ones, is
impossible.
Some other groupings are based on more abstract properties. For instance, CPrs
corresponding to the Forming Construction denote heterogeneous processes implying
the eruption, the forming or the transformation of an entity, e.g. kapak zadan ‘to go
mouldy’ (Lit. ‘mould hit’), tāval zadan ‘to blister’ (Lit. ‘blister hit’), yax zadan ‘to freeze’
(Lit. ‘ice hit’), zang zadan ‘to go rusty’ (Lit. ‘rust hit’). Contrary to the previous case, the
nominal element is not exclusively selected via its conceptual properties and the link
between the components is more collocational: javāne zadan ‘to blossom’ (Lit. ‘blossom
hit’) vs. gol dādan ‘to flower’ (Lit. ‘flower give’) and not gol zadan. Consequently, one
72 P. Samvelian and P. Faghiri

can assume that the productivity of this class is not comparable to that of the previous
one.
Classes also vary with respect to their size. It is expected that fully compositional
classes should be large, since new combinations can regularly enrich the class. This is
indeed the case with the Spreading‐zadan Construction, for instance, which constitutes
one of the largest classes of CPrs formed with zadan. However, some compositional
classes are restricted due to the small number of nouns that could fit in the
Construction. For instance, the Scream Emission Construction corresponds to only
about twenty CPrs, which is not surprising, given the number of nouns denoting a cry
in Persian.
Finally, the issue of the productivity cannot be investigated without integrating
the concurrent possibilities, that is, the competing verbs to form a CPr: various verbs
can be used with the same non‐verbal element to realize the same “verbal concept”,
thus giving rise to the existence of synonymous CPrs. This situation is reminiscent of
the choice between competing affixes to form a new word in construction morphology.

5. Conclusion

The “fully” compositional view of Persian have become so prevailing that it now seems
to be taken for granted. In this paper we advocated for a view going against this
widespread consensus. We showed that not only the lexical meaning of Persian CPrs is
barely ever fully predictable from the meaning of their component parts, but also that
even more abstract properties, such as argument and event structure, cannot be
determined a priori, on the basis of solely one component of the CPr regardless of the
other one.
In line with Goldberg (1996) and Family (2006), we argued that Persian CPrs
must be listed since they correspond to conventional pairings of forms and meanings.
However, we defended the idea that there is no contradiction between storage and a
compositional account of these combinations, provided compositionality is defined a
posteriori, in the sense of Nunberg et al. (1994) for idiomatically combining expressions.
The Construction‐based approach we outlined is based on the assumption that despite
their idiomaticity, Persian CPrs can be grouped upon their syntactic and semantic
similarities. Productivity results in this account from the possibility to analogically
extend the existing classes and can be compositionality‐based or not.
To finish, we want to insist on the importance of considering a large set of data
before drawing generalizations on CPr formation in Persian. Like morphological
lexeme formation processes, CPr formation in Persian displays a great amount of
Persian complex predicates 73

idiosyncrasies. Regularities apply within small and medium‐sized sets and rarely
concern the whole set of combinations formed with a given verb. A careful
investigation of data is necessary if sound and reliable generalizations are intended.

References

Bashiri, Iraj. 1981. Persian Syntax. Minneapolis: Burgess Publishing Company.

Bonami, Olivier, and Pollet Samvelian. 2010. Persian complex predicates: Lexeme formation by itself.
Paper presented at Septièmes Decembrettes Morphology Conference, Toulouse.
Bonami, Olivier, and Pollet Samvelian. in press. The diversity of inflectional periphrasis in Persian. Journal
of Linguistics.
Borer, Hagit. 1994. The projection of arguments. In University of Massachusetts Occasional Papers in
Linguistics 17, Syntax and Semantics, ed. by Elena Benedicto and Jeff Runner. GLSA, University of
Massachusetts, Amherst.
Butt, Miriam. 1995. The structure of complex predicates. Stanford: CSLI Publications.
Clark, Eve V., and Herbert H. Clark. 1979. When nouns surface as verbs. Language 55(4):767–811.
Dabir‐Moghaddam, Mohammad. 1982. Passive in Persian. Studies in the Linguistic Sciences 12(1):63–90.
Family, Niloufar. 2006. Explorations of semantic space: The case of light verb constructions in Persian. Doctoral
dissertation, EHESS.
Folli, Raffaella, Heidi Harley, and Simin Karimi. 2005. Determinants of event type in Persian complex
predicates. Lingua 115:1365–1401.
Ghomeshi, Jila. 1997. Topics in Persian VPs. Lingua 102:133–167.
Ghomeshi, Jila, and Diane Massam. 1994. Lexical/syntactic relations without projections. Linguistic Analysis
23(3/4):75–217.
Goldberg, Adele E. 1995. A Construction Grammar approach to argument structure. Chicago: University of
Chicago Press.
Goldberg, Adele E. 1996. Words by default: Optimizing constraints and the Persian complex predicate. In
Annual Proceedings of the Berkeley Linguistic Society 22, 132–146. Berkeley.
Grimshaw, Jane, and Armin Mester. 1988. Light verbs and theta‐marking. Linguistic Inquiry 19:205–232.
Gross, Maurice. 1975. Méthodes en syntaxe: Régime des constructions complétives. Paris: Hermann.
Hale, Kenneth L., and Samuel J. Keyser. 1993. On argument structure and the lexical expression of
syntactic relations. In The View from Building 20, ed. by Kenneth L. Hale and Samuel J. Keyser, 53–
109. Cambridge, MA: MIT Press.
Hale, Kenneth. L., and Samuel J. Keyser. 2002. Prolegomenon to a theory of argument structure. Cambridge,
MA: MIT Press.
Karimi, Simin. 1996. Case and specificity: Persian râ revisited. Linguistic Analysis 26:74–194.
Karimi‐Doostan, Gholamhossein. 1997. Light verb constructions in Persian. Doctoral dissertation,
University of Essex.
Kay, Paul, and Charles J. Fillmore. 1999. Grammatical constructions and linguistic generalizations: The
What’s X doing Y? construction. Language 75:1–33.
Kratzer, Angelika. 1996. Severing the external argument from its verb. In Phrase structure and the lexicon, ed.
by Johan Rooryck and Laurie Zaring, 109–137. Dordrecht: Kluwer Academic Publishers.
74 P. Samvelian and P. Faghiri

Lazard, Gilbert. 1982. Le morphème rā en persan et les relations actancielles. Bulletin de la Société de
Linguistique de Paris 77(1):177–207.
Marantz, Alec. 1997. No escape from syntax: Don’t try morphological analysis in the privacy of your own
lexicon. University of Pennsylvania Working Papers in Linguistics 4(2):201–225.
Megerdoomian, Karine. 2001. Event structure and complex predicates in Persian. Canadian Journal of
Linguistics 46(1/2):97–125.
Megerdoomian, Karine. 2012. The status of the nominal in Persian complex predicates. Natural Language
and Linguistic Theory 30(1):179–216.
Michaelis, Laura A. to appear. Sign‐based Construction Grammar. In The Oxford Handbook of Linguistic
Analysis, ed. by Bernd Heine and Heiko Narrog. Oxford: Oxford University Press.
Moyne, John A. 1974. The so‐called passive in Persian. Foundations of Language 12:249–67.
Müller, Stefan. 2010. Persian complex predicates and the limits of inheritance‐based analyses. Journal of
Linguistics 46(2):601–655.
Nunberg, Geoffrey, Ivan A. Sag, and Thomas Wasow. 1994. Idioms. Language 70:491–538.
Samiian, Vida. 1983. Origins of phrasal categories in Persian: An X‐bar analysis. Doctoral dissertation,
UCLA.
Samvelian, Pollet. 2001. Le statut syntaxique des objets nus en persan. Bulletin de la Société de Linguistique de
Paris 96(1):349–388.
Samvelian, Pollet. 2006–2007. Le sujet, l’objet et l’inaccusativité dans les prédicats complexes nom‐verbe en
persan. Cahiers de Linguistique de l’INALCO 6:155–190.
Samvelian, Pollet. 2012. Grammaire des prédicats complexes. Les constructions nom‐verbe. Lavoisier.
Samvelian, Pollet, and Pegah Faghiri. 2013. Introducing PersPred, a syntactic and semantic database for
Persian complex predicates. In Proceedings of the Workshop on Multiword Expressions. Atlanta,
Georgia, USA.
Samvelian, Pollet, and Pegah Faghiri. in press. Rethinking compositionality in Persian complex predicates.
In Proceedings of the 39th Annual Meeting of the Berkeley Linguistics Society.
Samvelian, Pollet, Pegah Faghiri, and Sarra El Ayari. 2014. Extending the coverage of a MWE database for
Persian CPs exploiting valency alternations. In Proceedings of the Ninth International Conference on
Language Resources and Evaluation (LREC’14), 4023–4026.
Vendler, Zeno. 1957. Verbs and times. Philosophical Review 66:143–160.

To cite this article:

Samvelian, Pollet, and Pegah Faghiri. 2014. Persian complex predicates: How compositional are
they?. Semantics‐Syntax Interface 1(1):43–74.
 SHORT CONTRIBUTION

Ezafe Constructions as Relative Clauses:

Semantic Arguments

Octav Eugen DeLazeroa and Parvane Geraeeb*

aNorth‐Eastern Federal University, Yakutsk, Russia
b University of Tehran, Tehran, Iran

1. Persian Ezafe Constructions: An Introduction

Persian Ezafe constructions have been extensively discussed in the literature (Samiian
1994; Ghomeshi 1997; Abolghasemi 2002; Larson and Yamakido 2005; among many
others). Samiian (1994) defines Persian Ezafe as the unstressed morpheme ‐e (‐ye when
following a vowel segment) which appears before some right‐branching phrasal
complements and modifiers in nonverbal phrasal categories. Even though this
morpheme is phonologically adjacent and connected to the head noun, its occurrence is
syntactically motivated by the right‐branching phrasal complement. Proposing the
surface distribution of Ezafe before right‐branching phrasal complements and
modifiers under the first bar level in nonverbal phrasal categories, she concludes that
in Persian, in accordance with the principles of UG, the case assigning categories are
the [‐N] categories V and P1 (e.g. ba ‘with’ and bi ‘without’), and Ezafe is a dummy case
assigner which occurs in phrases headed by non‐case‐assigning categories distinct
from [‐N] categories: N, A, and P2 (e.g. zir‐(e) ‘under’, ru‐(ye) ‘on’). She also argues that
Persian Ezafe is a case marker, inserted before complements of [+N] categories,
including Ns, As, and some Ps (cf. Larson and Yamakido 2005).
Ghomeshi (1997), on the other hand, proposes that Ezafe refers to an unstressed
vowel ‐e that links together elements belonging to a single constituent. The Ezafe vowel
appears between an adjective and its complement, between a preposition and its
complement, and more generally between any two items that have some sort of
connection.
The historical origin of Modern Persian ‐(y)e in Ezafe constructions as a relative
pronoun going back to Old Persian haya (Abolghasemi 2002:32) suggests at least the

* Email: [Link]@[Link]

The authors would like to thank two anonymous reviewers of SSI for their helpful remarks.

Semantics‐Syntax Interface, Volume 1, Number 1, Spring 2014

75–81
© 2014 by University of Tehran
[Link]
76 O.E. DeLazero and P. Geraee

possibility that Modern Persian adjectives in Ezafe may have preserved the original
syntactic structure, requiring a treatment as relative clauses with ‐(y)e as a relative
pronoun and a phonologically null copula. In this squib, we argue that the selectional
restrictions of modal adjectives in Modern Persian (i.e. their combinability with nouns)
support a syntactic analysis of adjectives in Ezafe Constructions as relative clauses.

2. Inherited Syntax in Ezafe Constructions

The Modern Persian ‐(y)e in Ezafe constructions has been regarded as continuing the
Old Persian masculine singular nominative relative pronoun haya formed by the
agglutination of the demonstrative base *so‐ to the relative base *yo‐ in Proto‐Indo‐
European *so‐yo‐ (Abolghasemi 2002:32; de Vaan and Lubotsky 2009:167).
This etymology suggests the possibility that Modern Persian adjectives in Ezafe
may have preserved the original syntactic structure. If this is the case, they would
require a treatment as relative clauses, that is as a CP with ‐(y)e originally base
generated in the specifier of a small clause (SC). The Ezafe marker ‐(y)e then moves to
the specifier of CP as a relative pronoun.

(1) a. ʔabi [CP ei [C’ [SC ti særd][C0 Ø]]]

water EZ cold
‘cold water’ – literally ‘water which is cold’

b. NP

N

N CP

N0 ei C
ʔabi
SC C0
[ti særd] Ø

(1b) is head‐final, consistent with the prevailing word‐order in Old and Modern
Persian, with the heavy CP constituent to the right. This analysis also accounts for the
apparent noun‐adjective order in Modern Persian, as opposed to the adjective‐noun
order, which is much more common in head‐final languages: if adjectives in Ezafe are
Ezafe constructions as relative clauses 77

relative clauses underlyingly, they will be heavy constituents, and as such placed to the
right.1

3. Modal Adjectives in Modern Persian

Noun‐adjective combinations: The analysis in (1) is supported by the selectional

properties of modal adjectives, e.g. momken ‘possible’ and ʔehtemali ‘probable’. In
Modern Persian, these adjectives display a more limited combinability in attributive
position than they do in English. This behavior is most conspicuous in combinations
with role nouns, i.e. nouns which describe entities in context; for instance, a winner is a
person winning a certain contest. With role nouns, not all combinations that are
allowed in English are acceptable in Modern Persian – see the contrast between (2) and
(3) (Ilkhanipour and DeLazero 2013):

(2) bærænde‐ye ʔehtemali

winner‐EZ probable
‘probable winner’

(3) #bærænde‐ye momken

winner‐EZ possible
‘possible winner’

On the other hand, these adjective are disallowed in predicative position in both
languages:

(4) #bærænde ʔehtemali / momken ʔast

winner probable / possible be.3SG
‘#The winner is probable/possible’

1 A reviewer asked if the same analysis can be applied for Ezafe element that comes between adjectives or
prepositions and their complements. For the vowel e that comes between some adjectives and their
complements, as in ʔælaqemænd‐e be sinæma ‘interested in cinema’, we consider ‐e not as Ezafe but as a
vowel inserted for phonetic reasons; also note that the occurrence of this vowel, unlike Ezafe, is optional.
With regard to the Ezafe marker that appears in prepositional phrases, as in zir‐e miz ‘under the table’, it is
worth mentioning that Persian complex prepositions have been studied as “Ezafe nouns” and thus can be
analyzed as noun‐noun combinations (Negin Ilkhanipour, pers. comm.).
78 O.E. DeLazero and P. Geraee

Selectional restrictions based in syntax: An analysis of (3) as involving a relative

clause accounts for its unacceptability, since the intensional adjective will be in a
predicative position, where it is disallowed (5b). The ungrammaticality in (5a) follows
from the uninterpretability of the ‘small clause’ SC in (5b).

(5) a. #bærændei [CP yei [C’ [SC ti momken][C0 Ø]]]

winner EZ possible
‘possible winner’ – literally ‘#winner who is possible’

b. NP

N

N CP

N0 yei C
bærændei
SC C0
#[ti momken] Ø

Exceptions: On the other hand, acceptable examples of combinations of role

nouns with modal adjectives like (2) could be accounted for by language contact: as it
was the case with several other languages, including English (DeLazero 2013:188–190,
192–195), Modern Persian borrowed combinations of atributive intensional adjectives
plus role nouns (here, from Western European languages) in a piecemeal fashion,
without generalizing their use. Also as in other languages, intensional adjectives in
Modern Persian are originally confined to learned registers of the language.

Syntax of acceptable combinations: For the cases of role nouns with modal
adjectives which are acceptable, we propose an analysis involving a PRO element in
Spec,NP, whose reference is given by the discourse. This PRO will denote the context
(situation) in which the entity described by the role noun performs that role. For
instance, in (2) and (3), the entity described as a ‘winner’ will be the winning party in a
certain competition, contest, game, battle, war, etc., i.e. in a certain context. This context
will be the referent of PRO.
Ezafe constructions as relative clauses 79

(6) NP

PROi N

N CP

N0 yei C
bærænde
SC C0
[ti ʔehtemali] Ø

Here, SC is interpretable, because the relative ye refers not to the entity described
by the role noun bærænde ‘winner’, but to the context (situation) referred to by the PRO.
Indeed, nouns describing situations can be used as subjects in sentences where modal
adjectives are predicates – in Modern Persian as well as in English (see Ilkhanipour and
DeLazero 2013):

(7) piruzi‐ye tim ʔehtemali bud

victory‐EZ team probable [Link].3SG
‘The victory of the team was probable.’

The PRO in Spec,NP is analogous to the possessor in Spec,NP in Semitic

construct states (Riter 1988): the context where the entity plays a certain role is
analogous to the context of possession given by the possessor. In both cases, the
context can be said to ‘contain’ the entity described by the noun.

Semantics of acceptable combinations: In the acceptable example in (2)

interpreted as ‘winner in a probable context’, what is described as ‘probable’ is not the
winner’s existence, but the occurrence of the context in which the respective individual
will emerge as a winner. A role noun must combine not only with an entity x (semantic
type e), but also with the context c (semantic type s) where that entity will perform the
respective role (see the semantics of role nouns in DeLazero 2013):

(8) [[bærænde]] = λcsλxe[winner(c)(x)] = λcsλxe[x is a winner in c]

The denotation of the Ezafe construction will be as in (9), and it will combine
with the noun by predicate modification (10):

(9) [[‐ye ʔehtemali]] = λcs[probable(c)] = λcs[c is probable]

80 O.E. DeLazero and P. Geraee

(10) [[bærænde‐ye ʔehtemali]] = λcsλxe[winner(c)(x) & probable(c)] =

λcsλxe[x is a winner in the probable context c]

4. A Cross‐Linguistic Comparison

A situation similar to the one discussed here can be observed in Chinese, where
constructions like (2) and (3) are not acceptable, as in (11), and adjectives are linked to
the nouns by the particle de which also introduces relative clauses and links
possessions to possessors. As such, all adjective‐noun combinations in Chinese that
contain de can be regarded as relative clauses, as illustrated in (12), with de as a
complementizer, and the small clause containing a PRO co‐indexed with the noun
(there is no relative pronoun, because Chinese has no wh‐movement).

(11) #kěnéng de huòshèngzhě

possible/probable PART winner
‘possible/probable winner’

(12) #[CP [C’ [SC PROi kěnéng][C0 de]]] huòshèngzhěi

The role of Chinese de is similar to the role of Modern Persian ‐(y)e in that it links
adjectives to nouns and possessions to possessors, although their syntactic categories
are different (Chinese de is a complementizer in C0, while Modern Persian ‐(y)e is a
relative pronoun in Spec,CP). As regards adjectives, they appear to be relative clauses
underlyingly in both Modern Persian and Chinese, this analysis being supported by
the similar selectional restrictions of modal adjectives in both languages.

5. Conclusion

We have argued that an analysis of Ezafe constructions as involving a relative clause

accounts for the distribution of modal adjectives in the nominal domain, thereby
suggesting the preservation of syntactic structure throughout the evolution from Old
to Modern Persian. The parallel structure and behavior of nominal modal constructions
in Chinese point to the possibility of cross‐linguistic comparison as a means to verify
the correlation between relative clause syntax and selectional properties of ad‐
nominals.
Ezafe constructions as relative clauses 81

References

Abolghasemi, Mohsen. 2002. tarix‐e moxtæsær‐e zæban‐e farsi [The brief history of Persian]. Tehran: Tahoori.
DeLazero, Octav.E. 2013. Modal adjectives in the history of English. In Periphrasis, replacement and renewal.
Studies in English historical linguistics, ed. by Irén Hegedűs and Dóra Pődör, 187–211. Newcastle
upon Tyne: Cambridge Scholars.
Ghomeshi, Jila. 1997. Non‐projecting nouns and the Ezafe construction in Persian. Natural Language and
Linguistic Theory 15:729–788.
Ilkhanipour, Negin, and Octav E. DeLazero. 2013. Selectional properties of Persian modal adjectives. Talk
presented at ICIL5, Bamberg, Germany.
Karimi‐Doostan, Gholamhosein, and Razieh Bagheri. 2011. saxt‐e ezafe dær zæban‐ha‐ye irani [Ezafe
construction in Iranian languages]. Tehran: Sokhan.
Larson, Richard, and Hiroko Yamakido. 2005. Ezafe and the deep position of nominal modifiers. Barcelona
Workshop on Adjectives and Adverbs.
Riter, Elizabeth. 1988. A head movement approach to construct state noun phrases. Linguistic Inquiry
26:909–929.
Samiian, Vida. 1994. The Ezafe construction: Some implications for the theory of X‐bar syntax. In Persian
studies in North America, ed. by Mehdi Marashi, 17–41. Betheda, MD: Iranbooks.
de Vaan, Michel, and Alexander Lubotsky. 2009. Altpersisch. In Sprachen aus der Welt des Alten Testaments,
ed. by Gzella Holger, 160–174. Darmstadt: Wissenschaftliche Buchgesellschaft.

To cite this article:

DeLazero, Octav E., and Parvane Geraee. 2014. Ezafe constructions as relative clauses: Semantic
arguments. Semantics‐Syntax Interface 1(1):75–81.
Semantics‐Syntax Interface
Volume 1, Number 1
Spring 2014

CONTENT

Negin Ilkhanipour
Editorial i

Original Articles

Barbara H. Partee
A Brief History of the Syntax‐Semantics Interface in Western Formal
Linguistics 1

Milada Walková
Scalarity Determines Argument Structure in English Particle Verbs and
Slovak Prefixed Verbs 22

Pollet Samvelian and Pegah Faghiri

Persian Complex Predicates: How Compositional Are They? 43

Short Contributions

Octav Eugen DeLazero and Parvane Geraee

Ezafe Constructions as Relative Clauses: Semantic Arguments 75
 ‫ﺷﻨﺎﺳﻨﺎﻣﺔ ﻧﺸﺮﻳﻪ ‪Semantics‐Syntax Interface‬‬

‫ﺻﺎﺣﺐ اﻣﺘﻴﺎز‪ :‬داﻧﺸﻜﺪة ادﺑﻴ‪‬ﺎت و ﻋﻠﻮم اﻧﺴﺎﻧﻲ داﻧﺸﮕﺎه ﺗﻬﺮان‬

‫ﻣﺪﻳﺮ ﻣﺴﺌﻮل‪ :‬ﻏﻼﻣﺤﺴﻴﻦ ﻛﺮﻳﻤﻲدوﺳﺘﺎن )اﺳﺘﺎد زﺑﺎنﺷﻨﺎﺳﻲ‪ ،‬داﻧﺸﮕﺎه ﺗﻬﺮان(‬

‫ﺳﺮدﺑﻴﺮ‪ :‬ﻧﮕﻴﻦ اﻳﻠﺨﺎﻧﻲﭘﻮر )داﻧﺸﺠﻮي دﻛﺘﺮي ﺗﺨﺼ‪‬ﺼﻲ زﺑﺎنﺷﻨﺎﺳﻲ‪ ،‬داﻧﺸﮕﺎه ﺗﻬﺮان(‬

‫داﻧﺸﮕﺎه ﺻﺎدرﻛﻨﻨﺪة ﻣﺠﻮ‪‬ز‪ :‬داﻧﺸﮕﺎه ﺗﻬﺮان‬

‫زﻣﻴﻨﺔ اﻧﺘﺸﺎر‪ :‬ﻋﻠﻤﻲ‬

‫ﺗﺮﺗﻴﺐ اﻧﺘﺸﺎر‪ :‬دوﻓﺼﻠﻨﺎﻣﻪ‬

‫ﮔﺴﺘﺮة ﺗﻮزﻳﻊ‪ :‬داﻧﺸﮕﺎه ﺗﻬﺮان‬

‫ﺗﺎرﻳﺦ اﻧﺘﺸﺎر‪ :‬ﺑﻬﺎر ‪1393‬‬