Linguistik
Refine
Year of publication
- 2001 (155) (remove)
Document Type
- Part of a Book (66)
- Conference Proceeding (31)
- Article (25)
- Working Paper (9)
- Preprint (8)
- Book (6)
- Review (6)
- Report (2)
- Diploma Thesis (1)
- Periodical (1)
Language
- English (121)
- German (29)
- Portuguese (4)
- Polish (1)
Is part of the Bibliography
- no (155)
Keywords
- Syntax (32)
- Semantik (20)
- Prädikat (18)
- Deutsch (15)
- Englisch (13)
- Informationsstruktur (10)
- Satzakzent (8)
- Lexikologie (6)
- Polnisch (6)
- Russisch (6)
Institute
In this paper, I argue (i) that Japanese has constructions that are almost the exact mirror images of the right-node raising constructions in English, and (ii) that the properties of those constructions, which I refer to as left-node raising constructions, can be captured straightforwardly if and only if the CONTENT values of domain objects, not those of signs, are assumed to be the principal locus of meaning assembly. In the theory proposed, it is claimed that semantic composition (including "quantifier retrieval") takes place not when some signs are syntactically combined to produce a new, larger sign but when some domain objects (which are essentially prosodic constituents) are merged (by the total or partial compaction operation) to produce a new domain object (i.e. a new, larger prosodic constituent).
Since the introduction of the X-bar principles it is commonly assumed that prepositions are heads of PPs, in the same way as nouns and pronouns are heads of NPs. However, while this is well motivated for a large majority of the pronouns and the prepositions in many languages, there are also exceptions. More specifically, Van Eynde (1999) argues that the reduced or minor pronouns of Dutch — as opposed to their full or tonic counterparts — cannot head an NP, and the purpose of this paper is to demonstrate that there are also prepositions which cannot head a PP. The first section introduces the distinction between major and minor categories. The second shows how it can be applied to the prepositions and presents a way of treating minor prepositions in HPSG. The third singles out the Dutch te(to) as a plausible candidate for a minor preposition treatment, and the fourth provides criteria for the identification of other minor prepositions. The concluding section points out the wider significance of these findings.
One kind of relative clause in Modern Hebrew is formed with a gap, as in (1a). However, in certain situations, the gap can be replaced by a resumptive pronoun, as in (1b):
(1a) ha-yeled she ra'iti
the-boy that saw-1.SG
(b) ha-yeled she ra'iti 'oto;
the-boy that saw-1.SG him;
the boy that I saw
Some previous approaches, such as (Borer 1984) and (Sells 1984), have treated gaps and resumptives with different mechanisms. This paper examines several properties that Hebrew resumptive pronouns share with gaps, motivating a more unified treatment in HPSG using non-local feature propagation for both. This machinery is then used in the analysis a variety of Hebrew relative clause phenomena, including in situ resumptive pronouns, fronted resumptive pronouns, relative clauses lacking a complementizer, bare gap relatives, and subject-verb inversion.
References
Borer, Hagit (1984). "Restrictive relative clauses in Modern Hebrew." Natural Language and Linguistic Theory 2:219-260.
Sells, Peter (1984). Syntax and Semantics of Resumptive Pronouns. Ph.D. Thesis, UMass Amherst.
This paper presents a dynamic semantic approach to the licensing of Polarity Sensitive Items (PSIs) and n-words of Negative Concord. We propose that PSIs are unified by the semantic scale property, which is responsible for their sensitivity to the context; we develop a semantic licensing analysis based on Fauconnier's (1975) scales and Ladusaw's (1979) notion of entailment. The first part of the paper concludes with a formalization of semantic licensing in the sense of Ladusaw (1979) within HPSG (see, e.g., Pollard & Sag (1994)) which allows for a uniform treatment of the licensing of PSIs and n-words of Negative Concord and accounts for the disambiguating nature of PSIs in scopally ambiguous sentences.
The second part of the paper is concerned with the limitations of semantic licensing, which, we claim, needs to be sensitive to the context. We present the discussions of, e.g., Heim (1984) and Israel (1996) with respect to the importance of the context in particular licensing constellations, and then turn to linearity constraints on licensing. We present data from German which may not be accounted for by linearity constraints and sketch an analysis for this data which supports the necessity of context-sensitive semantic licensing.
One of the major controversies in present-day HPSG is whether the information about a word's argument structure should also be available on this word's phrasal projections. Some works assume that ARG-ST is present on words only; this is the claim of, e.g., Pollard and Sag 1994, Miller and Sag 1997, Abeillé, Godard and Sag 1998, and Bouma, Malouf and Sag (to appear). The reason for this assumption is that it leads to more restrictive grammars: with this restriction, words cannot select their arguments on the basis of the argument structure of these arguments' heads (e.g., there seems to be no language in which a verb selects exactly VPs with an NP[dat] argument). On the other hand, various other works assume the presence of the complete information about a word's argument structure on this word's phrasal projections. This is the stance of, e.g., Grover 1995 (to formulate a fully nonconfigurational binding theory), Frank 1994 (to deal with verb second in German), Frank and Reyle 1995 (to account for the interactions between scope and word order in German), Calcagno and Pollard 1997 and Abeillé and Godard 2000 (to analyze French causatives), Baxter 1999 (in an account of purpose infinitives in English), and Meurers 1999 (to deal with case assignment in German verb clusters).
In this paper, I endeavor to make two kinds of linguistic contribution.
On the theory-internal side, I argue that the issue whether ARG-ST or any such attribute should be present at the level of possibly saturated phrases, in addition to its presence on words, is not an "all or nothing" issue. Although I show that there are some environments in Polish which do seem to require the presence of ARG-ST on phrases, I also link this presence to the common feature of such environments, namely, to their semantic vacuity (understood in the sense of Pollard and Yoo 1998). Although no formal proof can be given that this is the only possible analysis, I try to proceed carefully by examining a variety of possible alternatives and showing that all of them fail in one way or another.
Since semantically vacuous environments are extremely rare, the resulting grammar is not less restrictive than, say, a grammar which allows a verb to subcategorize for a lexical argument (and, hence, have access to this argument's ARG-ST), a possibility often taken advantage of in HPSG analyses of complex predicates in various languages.
On the empirical side, I look at two rarely considered and ill-understood constructions in Polish, namely, at "long raising" across a preposition, and at case agreement with predicative phrases. Neither of these constructions has been successfully analyzed so far. Although the analyses proposed in this paper may be perceived as less than satisfactory on the aesthetical side, they constitute the first formal and uniform account of these phenomena.
In this paper I show that object to subject raising approaches as suggested by Pollard (1994) and Müller (1999) are problematic since they cannot account for adjective formation in a satisfying way. The approach by Heinz and Matiasek (1994), which is a formalization of Haider's (1986) ideas, cannot account for modal infinitives and control.
I develop a lexical rule based approach and it will be shown that this approach also extends to tricky cases of remote passive.
Directional Serial Verb Constructions (Directional SVCs), which are a subset of Serial Verb Constructions (SVCs) in Thai and involve motion-related verbs, are studied in this paper. According to two syntactic tests, two phrase structural schemata are involved in Thai Directional SVCs, including a recursive VP-over-VP structure and a complementation structure. Thai Directional SVCs also exhibit a dissociation between constituent structure and linear order. With this distinctive syntactic structure, Thai Directional SVCs are not reduceable to previously described SVCs. Nevertheless, within Head-Driven Phrase Structure Grammar, the rich featural specifications of heads and the mechanisms available for the percolation of specific head properties in a default interpretation of the Head Feature Principle allow for a straightforward model of Thai Directional SVCs.
The complement structure of tough constructions containing VP complements with gap sites linked to the tough predicate subject has been subject to considerable discussion in the syntactic literature, with an apparent consensus that in "John is easy for us to please, for us" is a PP constituent which controls the subject specification of the following infinitival constituent. I reexamine the classical arguments for this position, including Bresnan's seminal 1971 paper which first argued for this control structure analysis, and argue that none of these arguments are empirically tenable. In all cases, data exist which convincingly undermine central claims or assumptions, and hence there turns out to be no convincing reason to prefer the control structure over the clausal analysis, introduced in Postal's 1971 monograph on crossover and defended in the Gazdar et al. monograph on generalized phrase structure grammar, in which for us to please is a clausal complement to easy. I then offer a number of arguments for the superiority of the clausal analysis, appealing to data from comparatives, parasitic gap constructions and extraposition. My claim that tough complementation of the kind alluded to is clausal must, if sound, be compatible with standardly assumed semantics for these constructions, in which the subject of the complement clause must also serve as an argument of the tough predicate — a conclusion seemingly at odds with a clausal complement syntax. The difficulty is that a constituent whose denotation is one of the terms in the relation denoted by the tough predicate must be retrieved from with a clause, where it is presumably inaccessible under normal Montegovian compositional assumptions. I offer further cross-linguistic evidence based on Guyanese Creole that such an apparent conflict between syntax and semantics is unavoidable, and then offer a syntactic solution, based on work by Detmar Meurers which posits a HEAD feature for verbs structure-shared with their SUBJspecification. This device, which also can be argued for in English on the basis of the Richard construction and several other phenomena, offers a way for information about the subject to be accessible to specifications of the selecting head in a way which compromises locality to the minimal extent possible.
This paper focuses on the semantic properties and the syntactic behaviour of Modern Greek (hence MG) Experiencer-Subject Psych Verb Constructions (hence ESPVCs).
MG ESPVCs include verbs like miso (hate), agapo (love), or latrevo (adore), which feature a nominative experiencer in agreement with the verb and an accusative theme (see examples (1)-(3)). MG ESPVCs include also predicates like fovame (fear), which feature an experiencer-subject in agreement with the verb and either an accusative theme (example (4)), or a theme as the object of a prepositional phrase (example (5)). We should underline here that examples (4) and (5) below convey the same meaning. That is, they do NOT differ semantically.
O Gianis misi to sholio.
the Gianis.N hate.3S the school.A
"John hates school."
O Gianis agapa tin Maria.
the Gianis.N loves.3S the Maria.A
"John loves Mary."
O Gianis latrevi tin musiki.
the Gianis.N adore.3S the music.A
"John adores music."
I Maria fovate tis kategides.
the Maria.N fear.3S the storms.A
"Maria is afraid of the storms."
I Maria fovate me tis kategides.
the Maria.N fear.3S with the storms.A
"Maria is afraid of the storms."
The challenge that constructions like the ones in (4) and (5) pose lies on the split syntactic realization of the "experienced" (hence EXPD) semantic role (i.e., the theme), which in constructions like (4) is syntactically realized as the object of the sentence, while in constructions like (5) it is syntactically realized as the object of a prepositional phrase.
Our aim is to propose a unified linking account of the MG ESPVCs. This unified account
1) is based on the assumption that the individual denoted by the object NP (or PP) of the MG ESPVCs is entailed to be semantically underspecified, and
2) makes use of Wechsler's (1995) Notion Rule, Davis and Koenig's (2000) linking theory, as well as Markantonatou and Sadler's (1996) proposal for the linking of indirect arguments.
References
Davis, A.R. and J.-P. Koenig (2000). Linking as constraints on word classes in a hierarchical lexicon. Language 76, 56-91.
Markantonatou, S. and L. Sadler (1996). Linking Indirect Arguments. Essex Research Reports in Linguistics 9, 24-63.
Wechsler, S. (1995). The Semantic Basis of Argument Structure. Stanford: CSLI Publications. Series: Dissertations in Linguistics, Joan Bresnan, Sharon Inkelas, William J. Poser, and Peter Sells (eds.).
This paper discusses the behavior of picture NP reflexives in German and English. Taking the analysis of Pollard/Sag (1994) as a starting point, we show that their main conclusion for English, viz. that picture NP reflexives are exempt from Principle A, does not apply to German. As a first step, we present an alternative formulation of Principle A for German. But the principles proposed for German and English do not offer any explanation for the universal behavior of anaphors if they cannot be related to each other. We thusn propose a more general Principle A to hold universally. Individual, language-specific instantiations of this Principle A are derived from determining certain parameter settings.
The so-called 'double' or 'multiple' nominative constructions (henceforth DNC or MNC) have been one of the puzzling phenomena in topic-prominent languages like Korean, Japanese, and Chinese. One intriguing property of the MNCs is that there is no conceptual limit to the number of nominative nominals as long as the two consecutive nominatives are in a certain semantic relation. This paper provides a 'head-driven' and 'constraint' based analysis in the sense that the lexical head and the tight interaction among declarative constraints play a crucial role in the formation of puzzling DNCs/MNCs. This analysis captures the gramamtical/functional properties of each nominative phrase in these constructions in a precise way, while describing the close specifier-head relation between the two consecutive nominatives without violating the locality principle. The analysis eventually allows us an explicit grammar for generating DNCs/MNCs in Korean. This is achieved through the familiar mechanism of argument composition and constraints on the lexical heads such as noun and verb.
The so-called was-w-construction in German has received a fair amount of attention in recent syntactic theorizing. Most of the discussion has focused on the properties of was.
One line of research maintains that was is a scope marker that indicates the semantic scope of the wh-phrase in the embedded interrogative clause. The alternative view, usually referred to as the indirect analysis, was first developed with respect to Hindi (Dayal 1994) and then generalized to German (Dayal 1996). It holds that the was of the was-w-construction is associated not with the embedded wh-phrase, but rather with the embedded clause as a whole.
Hinrichs and Nakazawa present some novel evidence in favor of an indirect analysis of the was-w construction. However, the main focus of their research is on two questions that by comparison have received little attention, namely:
1. what is the set of matrix predicates that can enter into this construction, and
2. how can one account for the curious fact that predicates that ordinarily do not license wh-complements allow such complements in the was-w-construction?
On the basis of Ginzburg and Sag's verb classification (Ginzburg and Sag, in preparation) Hinrichs and Nakazawa identify a natural class of predicates that license this construction and utilize the notion of type coercion to account for the apparent mismatch between the syntactic form of the embedded interrogative and its semantic function.
This presentation is essentially a "guided tour of interesting sites" of the Norwegian language: passive, presentational constructions, anaphora and V2 patterns. The data is related to issues concerning Argument Structure and whether the analysis of root clauses in Norwegian should include a node "C" hosting the finite verb. The paper points to areas of Norwegian grammar which constitutes possible challenges to central proposals made in the HPSG literature, but, in addition, it sketches possible analyses within the HPSG framework.
This paper examines the distribution of English self-pronouns (himself, herself, etc.) in contrast with personal pronouns. We confirm that there are factors affecting the acceptability of self-pronouns at the syntactic, semantic, and discourse levels, and devise a Principle A with three distinct clauses to account for all of them. Each clause is crucially ordered with respect to the others, with syntax taking precedence over semantics, which in turn takes precedence over discourse constraints.
We discuss evidence in Halkomelem, a Coast Salish language of British Columbia, which supports the hypothesis put forward by Manning and Sag (1999) that a universal passive argument structure (ARG-ST) is complex and has two a-subjects. We argue that morphological and syntactic control phenomena in Halkomelem are best described by saying that an a-subject is accessible, where an a-subject is the first argument on an argument structure list.
ARG-ST <bi <a, Proi, ...>>
The Halkomelem passive data show that two notions of subject are essential for capturing control phenomena. One set of constructions-motion auxiliaries, desideratives, and reflexive causatives-involve linking to the internal a-subject. One construction-the control construction–links to either the highest a-subject or the internal a-subject. Similar conclusions have been drawn for data from Russian (Perlmutter 1984), Philippine languages (Schachter 1984), and other languages of the world. As Manning and Sag (1998) point out, one does not have to draw the conclusion that passive must be given a multilevel syntactic analysis from such data. Rather, their analysis of passive, which posits a complex argument structure, easily accounts for Halkomelem. Control facts in Halkomelem, with examples drawn from both morphological and syntactic constructions, can be added to the catalog of phenomenon that support this view of the passive.
A number of the languages of Polynesia, including Tongan and Samoan, display a process whereby a pronominal argument of the main predicate of a clause appears to be realized as a preverbal 'second position' (2P) pronoun. All other arguments, if overt, are realized postverbally, the languages being rigidly predicate-initial. This paper examines the characteristics of these pronouns in Tongan arguing that in most cases they are best treated as distinct words in their own right (though often phonologically deficient) while in a handful of cases they are affixal material composed morphologically with a preceding preverbal Tense/Aspect Marker (TAM). Despite the fact that Tongan preverbal pronouns clearly do not appear in a typical argument position, standard approaches to 2P pronominal elements (e.g. 'clitic climbing' and 'prosodic inversion') do not seem naturally applicable to the Tongan data. The relation-based analysis provided here exploits a natural consequence of various potential definitions of 'subjecthood' within HPSG, treating the preverbal pronouns as the (unique) instantation of the valence feature SUBJ and correctly blocking the possibility of the pronoun appearing in true second position above the TAM when a clause-initial conjunction is present, except in particular specified circumstances. Thus the Tongan pronouns are not strict '2P' elements despite the fact that they most often appear in second position in a clause.
In this paper, we present a surface-based analysis of a specific type of French parenthetical adjunct clauses introduced by the adverb comme (similar to as in English). The construction we focus on belongs to the domain of reported speech, and we call it reportive-comme clause (RCC). The set of data we consider exhibits a large amount of notable properties that can only be fully explained under the assumption of constructional constraints. Therefore, following Sag (1997) and Abeillé et al. (1998), we base our approach on the central notion of "construction". We claim that RCCs are adverbial extraction contexts. We integrate them in a cross-classified typed hierarchy as a subtype of relative clauses, and a subtype of head-adjunct and head-filler phrases. Semantic specifications of RCCs are expressed with constraints on different levels. We draw a general distinction between head-modifier adjuncts and parenthetical adjuncts in order to account for the fact that parenthetical adjuncts do not contribute the referential content of the head phrase they selected for. We posit two subtypes of RCCs determined by a Direct speech (and quotative) vs. Indirect speech distribution of properties. The two sets of defining constraints allow to characterize the restricted classes of verbs possible in the different RCCs, the syntactic realization (gap or pronominal affix) of their object argument and its anaphoric semantics. This treatment constitutes a more general proposal for direct speech or quoted argument selection, which is known as a puzzling problem of the syntax-semantic interface. It innovates in presenting a formalized account of reported speech phenomena and present a typed-based classification of the semantic relations of reported speech predicates.
Formalized as a systematic interaction between a tier of co-arguments and a tier of co-dependents, the concept of diathesis offers a considerable theoretical advantage in stating linguistic generalizations. Based on Slavic data, this paper argues for the general notion of dependents in HPSG, in addition to arguments and subcategorized elements (valence). It attempts to provide a systematic inventory of ARG-ST / DEPS mappings which results in a diathetic paradigm. The approach offers an insightful cross-linguistic and cross-constructional perspective.
In French, Italian, and Romanian, forms inherited from the Latin paradigm 'esse' are used for the copula, the passive auxiliary and tense or perfective auxiliaries. We show that the copula and the passive auxiliary should be identified, while the tense or perfective auxiliaries are different lexemes. Moreover, the copula has the same description across all the Romance languages. While they all are argument inheritance verbs, the copula and the tense or perfective auxiliaries differ with respect to their complement structure: (i) the second only have one complement structure (the complement participle is complement unsaturated), the copula has two complement structures (the predicate is either complement saturated or unsaturated); (ii) French and Italian tense auxiliaries are the head of a flat VP, where the participle is sister to its subcategorized complements, the Romanian perfective auxiliary is the head of a verbal complex; (iii) when the complement predicate is unsaturated, the Romance copula is the head of a flat VP.
Two consequences of the analysis are worth mentioning. First, the copula (and the passive auxiliary) should not be taken into account when stating the environments for auxiliary selection in French and Italian. Second, argument inheritance and complement structure are different phenomena, and argue in favor of theories which systematically distinguish between valence features and constituent structures.
A contrast to a trace
(2001)
For movement, such as quantifier raising, the three different structures illustrated in (1) are discussed in the recent literature.
(1) A girl danced with every boy
a. [every boy]x a girl danced with x (copy + replace)
b. [every boy]x a girl danced with [every boy] (copy)
c. [every boy]x a girl danced with [thex boy] (copy + modify)
In this paper, I'll call the proposal illustrated by (1a) the copy+replace theory since the movement is analyzed as first copying the moving phrase followed by replacing the moving phrase with a trace in the base position of movement. Chomsky (1993) and Fox (1999) argue against the copy+replace theory (1a) on the basis of Condition C data that show that moved material can behave as if it occupied the base position of movement. This behavior would, for example, be expected on the copy theory of movement illustrated by (1b), which also seems conceptually simpler than the copy+replace theory since it involves only copying without replacement. This conceptual advantage, however, is probably only apparent since a theory of the interpretation of structures like (1b) would probably be more complicated than for (1a). Standard assumptions about interpretation, at least, don't predict the right meaning when applied to (1b). For this reason, Chomsky and Fox propose what I'll call the copy+modify-theory illustrated in (1c). This proposes that copying is followed by a trace modification operation that replaces the determiner of the moved DP with something else. I assume that this is an indexed definite determiner, the interpretation of which is to be clarified below.