Ana səhifə

Chapter 1 Introduction Issues and Objectives

Yüklə 1.08 Mb.
ölçüsü1.08 Mb.
  1   2   3   4   5   6   7   8   9   ...   17

Chapter 1
1.1. Issues and Objectives

The recent studies on Complex / Composite Predicates, Compound Verbs, or Light Verb Constructions across languages have raised some challenging questions for theories of linguistics (Cattell, 1984; Grimshaw and Mester, 1988; Rosen, 1989b; Masullo, 1992; Mohanan, 1994; Butt, 1995; etc.). Light Verb Constructions (LVCs), combinations of a so called ‘light’ or ‘bleached’ verb and another element as in (1-3), in Iranian languages (e.g. Persian and Kurdish) can create some interesting questions for linguistics theories; in particular Morphological and Syntactic theories. In (1-3), the LVCs are enclosed by square brackets, the Light Verbs (LVs) are presented in capital letters and the non-verbal elements are shown in boldface.

(1) a. Ali [?este?fa: DA:D / KARD]. (Persian)

Ali resignation gave / did

‘Ali resigned.’

b. Ali [juz KERD] (Kurdish)

Ali surrender did

‘Ali gave up. / Ali resigned.’

(2) a. ?in tim ?a:n tim-o [šekast DA:D]. (Persian)

this team that team-DOM defeat gave

‘This team defeated that team.’

b. ?e tim-a ?o tim-a [šekast DA:]. (Kurdish)

this team-the that team-the defeat gave

‘this team defeated that team.’

(3) a. ?a:n tim (?az ?in tim) [šekast XORD]. (Persian)

that team (from this team) defeat collide

‘That team suffered defeat (by this team). / That team was defeated (by this


b. ?o tim-a (la ?e tim-a) [šekast XWA:RD]. (Kurdish)

that team-the (from this team-the) defeat collide

‘That team suffered defeat (by this team). / That team was defeated (by this


Iranian LVCs behave like single lexical verbs in some respects. These structurally complex constructs predicate arguments at clausal level exactly like a single verb. They also function like a single verb as far as different types of nominalization (e.g. Gerundive Nominalization, Agentive Nominalization, Past Participle Adjective Formation) are concerned. In addition, these LVCs may undergo pure morphological rules to form derived adjectives and manner adverbials which definitely have a zero-level status. These facts might indicate that these LVCs are single lexical units. On the contrary, Iranian LVCs can be broken by certain morphological and syntactic elements and operations. Affixes, negation morpheme, direct object clitics, some auxiliaries, modals and Prepositional Phrases can intervene between the components of LVCs. Either of the components of an LVC might be gapped and the nominal elements can be co-ordinated. In some cases, the nominal components of LVCs may be relativised, modified by adjectives, scrambled, specified and function as direct objects, or focused on by interrogative Wh-words. These symptoms can incline one to believe that LVCs are phrasal constructs.

The verbal members of LVCs, Light Verbs (LVs), appear to be like auxiliaries in that they cannot independently predicate arguments as if they lack semantic force and argument structure of their own. On the other hand, LVs, eventually, function as main verbs, when they combine with different non-verbal items.

This dissertation will examine LVCs and their components in two Iranian languages; mainly standard Persian and in some cases one of the dialects of Kurdish. We attempt to describe and analyse a considerable body of data from these languages in order to answer all or some of the questions in (4) and other related descriptive and theoretical issues.
(4) a. Is an LV completely devoid of semantic load?

b. If so, how is an LV different from an auxiliary?

c. How is an LV different from a full lexical verb?

d. Why can’t an LV predicate arguments on its own?

e. Does an LV have a defective Lexical Conceptual Structure (LCS) and / or a

defective argument structure?

f. If so, how does it obtain or reconstruct its argument structure to function as a

main verb?

g. Can LVs and / or other elements of LVCs case-mark arguments individually

or as single units?

h. Are the NP components of LVCs case-marked?

i. What sort of items can combine with an LV and what role(s) do they play in

LVC formation?

j. How do the components of an LVC combine together in order to form

an LVC?

k. Where (at which level of representation) do LVCs form?

l. Is the internal structure of an LVC visible to morphological and / or syntactic


m. Does the study of LVCs suggest, support or undermine any particular theory

or theories of argument structure?

1.2. The Organisation of the Thesis
This thesis consists of five chapters. In the remainder of the present chapter, section {1.3}, the languages under study are introduced. Chapter {2} concentrates on argument structure, and the realisation of argument structure in LVCs. The discussion of the leading theorists’ views on argument structure and LCS, the relationship between LCS and argument structure and the type of lexical representation assumed in this work are dealt with in section {2.1}. The general characteristics of LVCs and the common properties of these constructs across languages are viewed in section {2.2}. In sections {2.3} and {2.4}, different theories proposed for the argument structure realisation of LVCs are presented and discussed.

Chapter three is entirely devoted to description and discussion of various aspects of Persian LVCs and their components. In section {3.1}, the lexical and syntactic properties of Persian LVCs are pointed out and explained. The previous studies on Iranian LVCs are summarised and critically approached in {3.2}. In {3.3} and {3.4}, the various members of LVCs are introduced. Then, LVs are compared with lexical full verbs and auxiliaries in this language. Similarities and differences between LVs, auxiliaries, causatives and full verbs are explained in detail. In {3.5-7}, lexico-semantic aspects of LVs are focused on and it is argued that LVs are not completely devoid of semantic load. Then, LVs are taken into consideration in connection with unaccusativity in {3.8}. The argumentation of non-verbal components of LVCs in regard to separability and inseparability of LVCs is taken over in {3.9}.

In chapter four, the manner and the place of LVC formation are taken into consideration. It is argued that LVC formation is the result of a process of argument structure formation. Then, the lexical and syntactic properties of LVCs in relation to visibility / invisibility of internal structure of LVCs are accounted for in detail. Finally, case-marking in LVCs is concentrated on.

The results of the analysis of Persian LVCs and the basic findings of this study are summarised in chapter five.

1.3. The Languages under Study
As mentioned, the languages under study are two Iranian languages. Iranian languages are known to be a subgroup of the Indo-Iranian language group which itself is a branch of the Indo-European family. Consider the schema (5) assumed by many Iranists.
(5) * Indo-European

* Indo-Iranian

* Indian * Iranian

Iranian languages are historically classified into three groups: Old languages, Middle languages and Modern languages. Old Persian and Avesta are two Old Iranian languages (? - 300 BC) from which a considerable bulk of written texts are available. Parthian, Middle Persian (Pahlavi), Middle Sogdian, Khwarazmi, etc. belong to Middle Iranian languages (300 BC - 900 AD). The preserved materials from these languages are often religious, literary and historical texts. Languages referred to as Modern Iranian languages (from 900 AD up to the present time), spoken in Iran or other neighbouring countries, are assumed to originate from one of the ancient Iranian languages.

Modern Persian1 or Farsi, the main focus of the present study, has been developed out of Middle Persian (Khanlari, 1987). Almost half of the population of Iran are native speakers of this language and it is also used as the official language of the country and as the lingua franca for Iranian communities with different languages. While standard written Persian, from which the majority of our data is taken, displays a restricted SOV order, spoken Persian exhibits a wide range of reorderings. Persian verbs are inflected for person and number. Although there are not overt case-markers, the postposition ra: is often considered as the realisation of the objective accusative case in this language.

Kurdish, the second language occasionally touched upon in this thesis, probably originated from Middle Parthian or one of its sister languages (Oranskij, 1979: 331-341). This modern Iranian language, spoken in different countries (Iran, Iraq, Turkey, Syria etc.) and being heavily influenced by languages such as Persian, Arabic, and Turkish, has developed into different dialects. By Kurdish in this study, we mean a dialect of this language prevalent in Ilam and Kermanshah; two provinces in the western part of Iran. This dialect has recently borrowed many words from Standard Persian. Therefore, it is the closest Kurdish dialect to Persian. This Kurdish dialect, unlike Standard Persian, does not follow a strict SOV pattern and it may display different free word orders. It does not have overt case-markers. Unlike Persian, this dialect of Kurdish has a morphologically realised definite article,-a(ga) ‘the’. As in Persian, its verbs are inflected for person and number in a slightly different way.

Chapter 2

Argument Structure and Light Verb Constructions

2.1. The Architecture of Argument Structure Realisation
The notion “argument structure” (henceforth a-structure) borrowed from propositional logic into linguistics is referred to as ‘Verb Valency’ on some occasions. Another term related to a-structure is subcategorization. Stowell (1992) believes that many of the facts that were explained in terms of subcategorization and selection are now attributed to the principles of theta-theory which relate a-structures to syntactic representation. Booij (1992:49) states that there is no need for a level of syntactic subcategorization in addition to a-structure, and Lexical Conceptual Structure (LCS). Rappaport et al (1993:53) write that a-structure is projected from a more fully elaborated lexical semantic representation. Though, unlike subcategorization frames, a-structure does not provide information about linear order, case and categorial selection, to the extent that this information is predictable. To Williams (1981), Rappaport and Levin (1988) and Grimshaw (1990) a-structure, in contrast to subcategorization, is both relevant to the syntax and semantics: to syntax - determining the arguments which project to syntactic relations, and to semantics- by being related to the thematic roles (see Shapiro and Levin (1990) and Shapiro et al (1987) for a comparison between a-structure and subcategorization).

A-structure is introduced in (slightly) different ways by linguists working in various theoretical frameworks or focusing on this notion from the perspective of different linguistics disciplines. In Government Binding (GB) theory, a-structure is the information provided by the lexicon that determines d-structure, the syntactic level containing semantic (thematic) relations between the constituents. A-structure is represented as a set of arguments identified by theta roles2, i.e. agent, patient, goal, etc. (Williams, 1981, Stowell, 1981; Williams and Di Sciullo, 1987; Belletti and Rizzi, 1988; among others).

To some, a-structure, containing the number and types of arguments of a predicate, represents an intermediate level of representation between the (lexical) semantics and the syntax. It is the lexico-syntactic information from which the essential syntactic structures are projected (Zubizarreta, 1987; Grimshaw, 1990; Booij, 1992; Rappaport et al, 1993; among others). It is believed that a-structure contains “those aspects of lexical meaning that are grammatically relevant” (Zubizarreta, 1987), or it is “that part of conceptual structure that is visible to syntax” (Jackendoff, 1987:405).

In Shapiro and Levin (1990) a-structure characterises a sentence in terms of participants involved in the action or state described by the verb. A-structure determines who does what to what or whom. It identifies the number of arguments and their relations to each other characterised by the verb in a sentence.

Alsina’s (1993) model, in Lexical Functional Grammar (LFG) framework, contains different simultaneous levels of representation. These levels encode different types of information and are related to each other by principles of correspondence. One of these levels is argument structure, providing information about the arguments of a predicate. The level of argument structure, derived from Lexical Semantic Representation (LSR), can be characterised as:
* The argument structure retains from LSR the information about the number of


* An ordering is imposed on these arguments according to the thematic hierarchy.

* Further information in argument structure will be derived from LSR, as with

the distinction between external and internal arguments.
The argument structure, in Alsina (1993), is put into correspondence with syntax, but it is not encoded in syntactic terms and the syntax cannot uniquely reflect the argument structure. Thus, the argument structure is autonomous from syntax and it can’t be read off the syntax.

However, in spite of the variety of systems and terminology employed by linguists to deal with this phenomenon, a trend has been developed and assumed by many researchers according to which the lexical entry of a predicate contains two distinct, but closely related, levels of lexical representation: Lexical Conceptual Structure (LCS) and a-structure (Hale and Keyser, 1986, 1987; Rappaport and Levin, 1988; Grimshaw, 1990; Booij, 1992; Rappaport et al, 1993; Tenny, 1994, etc.). We will first focus on LCS in the following section, then we will return to the notion a-structure and its relation to LCS later on.

2.1.1. Lexical Conceptual Structure (LCS)
LCS or lexical semantic representation, roughly equal to the dictionary meaning of a predicate, is that part of the lexical information encoded in the semantic specifications of a predicate which specifies the properties of the action or the state to which that given predicate refers. For example, (1) can be viewed as the LCS of the verb ‘give’ in English.
(1) LCS for English ‘give’:

An entity (x) initiates an action in which (x) causes another entity

(y) to be affected in a way that (y) goes from the possession of (x) into

the possession of another entity (z).

Hale and Keyser (1986: 9), assuming that lexical entry of a verb includes not only a-structure but also its LCS, consider (2) as the LCS of verb ‘cut’ which is in the spirit of Jackendoff’s (1976, 1983) views on Conceptual Structures.

(2) LCS for English ‘cut’:

[x produces linear separation in the material integrity of y, by

sharp edge coming into contact with y]

In the same line, Rappaport and Levin (1988) give (3) as the LCS of verb ‘put’.
(3) LCS for English ‘put’:

[x cause [y to be at z]]

(3) is supposed to indicate that ‘put’ refers to an action in which an agent (represented by the x variable in the LCS) brings about a change in the location of an entity (the y variable); the z variable indicates the goal of this change of location.

Jackendoff (1983, 1987, 1991, 1993, etc. ) introduces and develops his theory of Conceptual Semantics in which the issues of argument structure, theta roles and LCS are dealt with in a decompositional lexical semantics fashion. Because of the significance of Jackendoff’s machinery and views on LCS and a-structure in the present work, we will present his theory in this part in more detail.

Jackendoff believes that the concepts such as agent, theme, etc. are not primitives of linguistic theory and the issues of argument structure and theta-roles cannot be explained by providing a list of theta-roles. Jackendoff, to explain the notions of argument structure and theta-roles in the framework of his Conceptual Semantics, argues that the essential units of conceptual structure are conceptual constituents, each of which belongs to one of a small set of major conceptual categories (or semantic “parts of speech” ) such as Thing, Event, State, Action, Place, Path, Property and Amount. Each major syntactic constituent of a sentence (excluding expletives such as ‘it’ and ‘there’) corresponds to a conceptual constituent in the meaning of the sentence. For instance, in ‘Bill ran into the garden’, ‘Bill’ and ‘the garden’ correspond to Thing-constituent and the PP ‘into the garden’ corresponds to Path-constituent and the whole sentence is an Event-constituent. In this approach, each conceptual category has some realisations in which it is decomposed into function-argument structure; each argument in turn corresponds to conceptual constituent of some major category. A predicate is realised as a super-ordinate category, either as a STATE or an EVENT. For instance, in (4a), expressing a STATE, the arguments are Bill (Thing) and kind (property), and in (4b), an EVENT, the arguments are John (Thing) and the bike (Thing).
(4) a. Bill is kind.

b. John fixed the bike.

Jackendoff (1991: 24) remarks that “the conceptual structure of a lexical item is an entity with zero or more open argument places. The meanings of the syntactic complements of the lexical item fill in the values of the item’s argument places in the meaning of the sentence”. For example, BE in (4a) expresses a state-function whose arguments are realised in subject and object positions. FIX in (4b) has an Event-function whose arguments are the subject and the complement NP ‘the bike’.

To clarify the above postulates, let us consider (5) which illustrates some formation rules for the spatial domain taken from Jackendoff (1991: 43).

(5) a. [PLACE]  [ place PLACE-FUNCTION ([THING] )]





Path VIA

[ event GO ([THING], [PATH])

c. [EVENT]  [ event STAY ([THING], [PLACE])]

[State BE ( [ THING], [PLACE])

d. [STATE]  [State ORIENT ( [ THING ] , [PATH])

[State EXT ( [ THING ], [PATH] )


e. [EVENT]  Event CAUSE ( EVENT , [EVENT])

(5a) indicates that a conceptual constituent belonging to the category of place can be elaborated as a Place-Function plus an argument that belongs to the category Thing. (5b), similarly, elaborates a Path, or trajectory, as one of five functions that map a reference Thing or Place into a related trajectory. (5c) shows that a constituent of the category EVENT can be elaborated as either of the two Event-Functions GO or STAY; each of which takes two arguments. The arguments of GO, which denotes motion along a Path are the Thing in motion and its Path. This structure can be found in the sentences such as ‘Bill went to New York’. The arguments of STAY, which refer to stasis over a period of time, are the Thing standing still and its location, as in ‘Bill stayed in the kitchen’. (5d) denotes three state functions. The first, BE, refers to the location of objects ( e.g. ‘the dog is in the Park’), the second, ORIENTATION, specifies the direction of objects (e.g. ‘the sign points toward New York’) and the third, EXTENSION, refers to the spatial extension of linear objects along a Path (e.g. ‘the road goes from New York to San Francisco). Finally, (5e) elaborates an EVENT as the Event-Function Cause plus two arguments. The first argument will be an agent if it is a THING, while it will be a cause if it is an EVENT. The second argument, an EVENT, is the effect ( see Jackendoff (1983: chap. 9 for detail).

As mentioned, conceptual structure does not take theta-roles as primitives of the theory. Rather, the terms ‘agent’, ‘theme’, etc. are formally defined as structural positions in a conceptual structure. For example, Theme is a shorthand for the first argument of GO, STAY, or ORIENT; Agent is a shorthand for first argument of CAUSE or LET; Goal is a shorthand for arguments of TO or TOWARD. Thematic roles are part of conceptual structure, not part of syntax. Thematic roles are nothing but particular structural configurations in conceptual structure. Jackendoff assumes that the theta-roles are represented in two tiers: a thematic tier dealing with motion and location and an action tier dealing with Actor-Patient relations (Jackendoff, 1991, 1993). In this case, the argument positions in (6), as an example, can have the following analysis in Thematic and Action tiers:

(6) a. Sue hit Fred .

Theme Goal ( Thematic tier )

Actor Patient ( Action tier )

b. Peter threw the ball .

Source Theme ( Thematic tier )

Actor Patient ( Action tier)

c. Bill entered the room.

Theme Goal ( Thematic tier)

Actor --- ( Action tier )

d. Bill received a letter .

Goal Theme

As is shown in (6), in this machinery every argument is related to a theta-role on thematic tier, but it is not the case that it should always bear an actor or a patient role on the action tier.

In Jackendoff’s system the LCS for ‘give’, as an example, can be represented as in (7) where it is shown that ‘give’ is a V(erb) with two semantic tiers, Thematic Tier and Action Tier, presented in two separate lines.

(7) give


CAUSE ([], [GO poss ([ ]j [TO []] )])

AFF ([ ]i , [ ]k)
In the first line, the thematic tier, the first square bracket identified by the Greek letter () is an argument of CAUSE event, the second identified by the subscript (j) is an argument of the EVENT GO poss(ession) and the third with the Greek letter () is an argument of the PATH-FUNCTION TO. The first and the third arguments in the Thematic Tier are bound to the argument slots in the Action Tier by the superscripts () and () respectively. This means that the (i) and (k) arguments in Action Tier, in addition to their roles in Thematic Tier, are the arguments of the AFF(ect) function in Action Tier. The AFF function indicates the relationship between the arguments. Its first argument, (i) argument, plays an Actor role and its second argument, (k) argument, plays a beneficiary role. The first argument of AFF function affects its second argument. The idea is that participants in an event may be associated with Thematic Tier and / or Action Tier roles.

Jackendoff does not assume that there is a separate level of argument structure, but rather derives the syntactic phrase structure directly from the lexical conceptual structure. In this machinery, the argument positions of each argument bearing constituent are specified by a set of indices ( i, j, k) as in (7). It is supposed that these indices

determine the type and number of arguments in the structures. So, there is no need for a separate level on which to perform the manipulations of linking semantic arguments to syntactic ones. They can be carried out directly on conceptual structure. It is claimed that both idiosyncratic and productive alternations in argument structure can be formulated in his framework without loss of generality. A-structure consists of the set of indices (i.e. i, j and k) that relates the syntactic and conceptual arguments of a head (Jackendoff, 1983; 1987:393, 1990).

  1   2   3   4   5   6   7   8   9   ...   17

Verilənlər bazası müəlliflik hüququ ilə müdafiə olunur © 2016
rəhbərliyinə müraciət