Marina Pantcheva

The Nominal Element in Complex Predicates

In this talk, I propose a syntax-based approach to Persian Complex Predicates (CPs) based on Ramchand's (2008) the decomposition of the verb. The main problem I focus on is the status of the non-verbal element (NV) in CPs of the form noun+light verb (LV).

In Ramchand's system, the specifier position of each subevent head hosts the thematic participants in the event, i.e. the arguments of the verb. The full decomposition of the VP is shown below.
     [initP INITIATOR [init [procP UNDERGOER [ proc [resP RESULTEE [res XP] ] ] ] ] ]
The external argument of a verb is obligatorily an INITIATOR. The internal argument is UNDERGOER or RESULTEE, or both. Subevent heads can have as their complement also nonverbal material. Such nonverbal complements are called RHEMES (e.g. the XP above). RHEMES are not arguments of the verb but part of the description of the predicate and together with the verb they build one joint predication.

I suggest that LVs lexicalize subevent heads just like heavy verbs. The difference between light and heavy verbs is that the former do not have conceptual-intentional content. The content comes from the NV which lexicalizes the RHEME and builds with the LV one joint predication. Then I go on to explore CPs of the form noun+z�æd�æn. I show that the NVs in some of them are distinct from direct objects, while in others they are not. However, even when the noun NV behaves like a direct object, the semantics of the noun+LV complex is the one of the CP, as if the NV still forms one predicate with the LV, i.e as if the noun occupies the RHEME. I suggest that a noun NV can move to a specifier of a verbal head (e.g. RESULTEE or UNDERGOER) from inside the RHEME position. This accounts for the cases where a noun NV functions as the internal argument of a LV, but is still semantically interpreted as an element of the CP.

Annahita Farudi and Maziar Toosarvandani

Ergativity and the structure of complex predicates in Dari

Much of the debate on the structure of complex predicates in Iranian has focussed on the syntactic status of nominal nonverbal elements (NVEs). Some authors have treated nominal NVEs as nonspecific direct object DPs (Mohammad and Karimi 1992, Ghomeshi and Massam 2001, Farudi 2005), while others have argued that the two should be kept structurally distinct (Megerdoomian 2002, Folli et al. 2005). Here, we explore the interaction between ergativity and complex predicates in Dari (Northwest Iranian, Central Plateau: Yazd, Iran) and its consequences for this debate. Assuming the core of Bittner & Hale's (1996) theory of ergativity, we argue that each of these approaches is correct, but only for a subset of complex predicates.

Keivan Zahedi

Does Incorporation Offer a Minimalist Account for Complex Predication in Modern Persian?

The present article questions the necessity of incorporation in the formation of complex predicates in Modern Persian (Farsi) within the Minimalist framework. First, it will be demonstrated that (1) combined complex predicates in Modern Persian may take NP objects without any functional markers, which will pose a problem for the incorporation analysis of such verbal compounds, and (2) not all functional markers need to be lost, which in turn will cast a serious doubt on the incorporative nature of such constructions. It will then be argued that (i) contrary to Dabir-Moghaddam (1997), there is no lexical process of incorporation involved in complex predication in Farsi; the lexical verbs do not take Ns but NPs, and as NPs they may take certain grammatical markers and are not obliged o remain adjacent to the verb, (ii) unlike Ghomeshi (1996) and Mohammad and Karimi, (1992) such NPs are not arguments since only DPs constitute true arguments (iii) Karimi's (1997) LF-incorporation analysis may not be a minimalist account for the combined complex predicates and (iv) the so-called 'incorporated' and 'combined' complex predicates differ in their phase-hood status. The results not only corroborate recent Chomskyan minimalist conjectures but also have significant bearings on a number of related phenomena. It will be shown that the occurrence of 'ra' is in fact limited to DPs, and not NPs, and as suggested by Zahedi (2008), 'ra' as a postposition will form a DP-shell. Moreover, it will be argued that predicative nominals forming complex predicates in combination with light verbs are 'bare' NPs, i.e., with no DP on top. In the end, the author will deal with how 'bare NPs' are checked for case.

Karine Megerdoomian

Preverbal Nominals and Telicity in Persian Complex Predicates

Since Vendler's (1967) seminal work on verbal aspectual classes, it has often been argued that verb phrase aspect is not determined in the lexicon but rather compositionally in the syntax. Harley (1999) extends this notion of compositionality to the internal structure of the denominal verbs in English and argues that the boundedness of the incorporating noun affects the telicity of the verb. Applying this argument to Persian and providing evidence from Locatum verbs, Folli, Harley and Karimi (2005) propose that the preverbal nominal is the sole determiner of telicity in the complex predicate construction in which it appears. I will show, however, that the boundedness of the preverbal nouns such as jaru 'broom' or shune 'comb' does not affect the telicity of the corresponding Instrumental predicates jaru/shune z�æd�æn 'sweep/comb', constituting clear evidence that the structure of locatum verbs and verbs of instrument is more complex than proposed. In addition, verbs such as jaru z�æd�æn 'sweep' and locatum verbs as in �æfsar z�æd�æn 'harness' display distinct modification possibilities in the presence of number implicature on the preverb.

In determining verb phrase aspect, Goals and quantitatively determinate Themes delimit the event denoted by the verb phrase, whereas instruments do not. I argue that a similar effect seems to be present at the level of complex predicates in Persian. I propose two distinct structures where the Instrument preverb occupies an adjoined position while the Locatum preverb occupies an argument position. This proposal provides support for the notion of syntactic decomposition of verbs, and argues for a compositional construal of telicity in complex predicates that considers the syntactic relation of the preverbal noun and the light verb element, and does not simply determine verbal aspect from the inherent properties of the preverbal noun.

Session 4: Historical perspective

Ludwig Paul

Persian Complex Verb Constructions in a Historic Perspective

Persian Complex Verb Constructions (CVC) have been studied intensively, using various general linguistic approaches. What is still missing, is a historical-linguistic approach investigating how the usage of CVCs has developed during the history of Persian. Such an approach should compare CVCs to other verb-building strategies that compete with them, esp. non-compound (inherited) verbs that are presumedly replaced by CVCs. The replacement of verbs like paziroftan through qabul kardan (both "accept") should be documented both in statistical terms, and in terms of possible specific forms or contexts in which each verbal construction occurs.

A glance will be taken at a select list of classical Persian texts. In the medical text Hidayat al-Muta'allimun (composed ca. 970 A.D.), a large number of CVCs with Arabic nouns is already found (tamam kardan "complete" etc.). In the national epic Shahname (early 11th cent. A.D.), however, non-compound verbs still predominate and the CVCs that occur, are mostly built with an (etymologically) Persian nominal complement (bar kardan "load"). The Siyasatname (composed late 11th cent.) still contains many non-compound verbs, and many CVCs, both with Persian and Arabic nominal complements. Interestingly, it also shows a couple of synonyms from both groups within one and the same chapter, e.g. paziroftan / qabul kardan "accept", goshadan / baz kardan "open".

In the Nuzhat al-Qolub, a geographic work from the early 14th century, the number of "simple" verbs has definitively diminished, and so has the number of CVCs with Persian complement as against the number of Arabic ones. In addition, the latter occur in a greater variety of syntactic contexts than the former, e.g. in passive (hesab karde shod "I appreciated the possessions").

It is assumed that investigating the diachronic development of CVCs will greatly enhance also the understanding of its synchronic functioning in Persian.

Soha Safa�ï

A diachronic and synchronic analysis of two complex spatial predicates in Persian

In the scope of the great interest that linguists have taken in complex predication in general and in Iranian languages more particularly, complex spatial predicates have passed under the radar. This issue has long been studied in Germanic languages, in Persian however it has yet to be raised. We offer hereby to examine the "complement of place + verb" in Persian (an SOV language) from a diachronic (Avestan, Old Persian, Middle Persian) as well as from a synchronic point of view, focalizing solely on two verbs: "to come" and "to go".

In Avestan and Old Persian, we have the two stems gam- and i-, both meaning "to go"; they can be binded with preverbals that are mainly very old adverbs. In Middle Persian, madann "to come" comes from gam-, i- vanishes and the verb raftan "to go" appears; they are used in Modern Persian too.

In the old languages, we won't be able to make any test in order to see how much the preverbal and the verb are binded ; but we will study the occurrences of the forms, see the context, and observe a lot of things from the texts that we have, for example from the conjugation, the way of writing, the negative sentences, the cases that follow the verbs; on another hand, in Modern Persian, we will be able to do other tests, like changing the place of the preverbal, the object, the construction of the sentence...
The question is whether we can still talk about a preverbal, like in the origin, and if not, what is the status of the first member of the complex spatial predicates in Persian.

Agnes Korn

Voice and Transitivity in Complex Predicates

New Persian complex predicates often come in pairs, with one member (with kardan/zadan) functioning as active, the other one (with shodan/x^wordan) as its passive or intransitive counterpart. So in addition to the inflection, the light verb also carries the voice property of the complex predicate. This paper will investigate how this use of light verbs may have come about. I will argue that the relevant verbs started out as markers of aktionsart and were later on reinterpreted as light verbs implying the voice property.

Constructions composed of a nominal and finite verb are found already in Old Iranian. One of these, meaning "can do something" ("potentialis"), is composed of the perfect passive participle (later on the past stem) of the main verb plus a finite form of "make". This pattern is attested in Old Persian; in several Middle Iranian languages and is still in use in several New Ir. languages. A periphrasis with "stand" functions as durative aktionsart in various Ir. languages, the earliest examples being found in Avestan. Several other verbs ("remain", "hold" etc.) are used in similar light verb function in Ir. languages.

In Middle Iranian, some verbs are grammaticalised as light verbs, transforming the aktionsart function into a transitivity feature. For instance, the active potentialis may be transformed a passive / intransitive "can be done" by substituting "become" for "make" in several languages.

The light verbs expressing intransitivity also fill the gap left by the loss of the Old Ir. middle voice that occurred in the Middle Iranian period. In this sense, complex verbs are a periphrastic construction expressing voice and transitivity. This development ranges itself within the broad trend in Iranian of substitution of synthetic constructions by analytical ones

Session 5: Formation of complex predicates

Stefan M�üller

A Lexical Analysis of Persian Complex Predicates

The dual nature of complex predicates in Persian was a much discussed topic during the past decades. On the one hand complex predicates behave like words since they undergo morphological derivations and on the other hand complex predicates behave like syntactic objects. For instance their parts can be separated by scrambling or by clitics or morphological material like the imperfective prefix or negation (Mohammad and Karimi, 1992). Often proposals were made that treat the complex predicates entirely in syntax or entirely in the morphology component. I will demonstrate that a lexicalist analysis can account for the Persian data. There is a common misunderstanding of the term lexicalist and of what it means to be licensed in the lexicon. For instance Embick (2004, p. 389) assumes that lexical analyses of resultative predicates imply that the predicate `hammer flat' is formed in the lexicon component and rejects such analyses on the basis of syntactically more complex examples. An alternative lexical analysis would assume that resultative constructions are licensed by a lexical rule that licenses a lexical item that selects for a result predicate (see Simpson, 1983; Wunderlich, 1992, p. 45; Verspoor, 1997; Wechsler, 1997; Wechsler and Noh, 2001; M�üller, 2002 for analyses of English, German, and Korean resultative constructions).

I will show that a similar analysis is possible for Persian Complex Predicates, that an argument composition approach can explain the differences in argument realizations in active and passive-like structures. The selection-based approach has no problems in accounting for the fronting of the preverb, the separation of preverb and verb by negation, the imperfective prefix, the future auxiliary, and clitics. The light verb contains the information about the presence of the preverbal element and can take its semantic contribution into account. For instance negation can then scope over the semantic contribution of the whole complex predicate. I show that the morphological bracketing-paradox (Vahedi-Langroudi, 1996, p. 6, p. 202--203, 211 and Karimi-Doostan 1997) disappears if a selection-based approach is assumed.

Ekaterina Lyutikova and Sergei Tatevosov

Causative-inchoative alternation: Evidence from complex predicates in Ossetian

We examine complex predicates (CPs) consisting of an adjectival stem and the light verb kenin 'do, make' in Ossetian. The crucial property of this type of CP is that it yields the causative-inchoative alternation. The problem is that the inchoative variant of this CP is the only intransitive configuration where kenin 'make' occurs: kenin can only be transitive if used as a lexical verb, and combining with unaccusatives it derives a causative construction that lacks an inchoative counterpart. We propose that the range of uses of 'make' mostly derives from more general assumptions, required independently, about syntax of basic verb classes in Ossetian - alternating transitives, non-alternating transitives, and unaccusatives. Two specific assumptions about 'make' we make are: (a) there are two lexical items, the verb of creation make_LEX, and the causative morpheme make_CAUS , and (b) the latter can merge either as V or v, provided that its complement possesses a matching logical type. When merged as V, 'make' creates VP which is subsequently merged with either of the two 'flavors' of v, causative or inchoative. It is in this way, we argue, that the causative-inchoative alternation associated with CPs in question is derived.

Kim Gerdes and Pollet Samvelian

Automatic extraction and partial annotation of Persian complex predicates

This talk proposes a statistical method for the extraction of Persian light verb constructions applied to a corpus of 748 966 sentences taken from the Kayhan newspaper. Persian light verb constructions have been extensively discussed in the literature, because of their abundance, productivity and particularities, but all quantitative measures have been hampered by the lack of sufficiently large and sufficiently annotated corpora for this language. The present work is done with a specialized web crawler for linguistic corpus research, named "Grosmoteur". The goal of Grosmoteur is to give access to corpora without prior linguistic data and moreover, to provide accessible interfaces for the common linguist without special computer training. We spidered the complete webpage of Kayhan with their online archive resulting in nearly 6000 different web pages containing 748 966 sentences, 21 266 372 tokens, nearly 800 MB of disk usage including the MySQL indices. For our implementation, we developed a very effective computation of the cumulative hypergeometric distribution, with configurable thresholds (cutting off computation if the probability drops below, for example, 1e-10). The proposed methods of corpus construction are language independent and the results on a relatively small corpus of 20 million word are promising since they confirm the power of association measures based on the hypergeometric distribution. The resulting lists show a graduation of lexicalization and the semantic homogeneity of some light verb subcategorization schemes which could be the reason for their wide usage.