Syntactic Principles Research Articles

Natural Language Computing: An English Generative Grammar in Prolog

This book's main goal is to show readers how to use the linguistic theory of Noam Chomsky, called Universal Grammar, to represent English, French, and German on a computer using the Prolog computer language. In so doing, it presents a follow-the-dots approach to language processing, linguistic theory, artificial intelligence, and expert systems. The basic idea is to introduce meaningful answers to significant problems involved in representing human language data on a computer. The book offers a hands-on approach to anyone who wishes to gain a perspective on natural language processing -- the computational analysis of human language data. All of the examples are illustrated using computer programs. The optimal way for a person to get started is to run these existing programs to gain an understanding of how they work. After gaining familiarity, readers can begin to modify the programs, and eventually write their own. The first six chapters take a reader who has never heard of non-procedural, backtracking, declarative languages like Prolog and, using 29 full page diagrams and 75 programs, detail how to represent a lexicon of English on a computer. A bibliography is programmed into a Prolog database to show how linguists can manipulate the symbols used in formal representations, including braces and brackets. The next three chapters use 74 full page diagrams and 38 programs to show how data structures (subcategorization, selection, phrase marker) and processes (top-down, bottom-up, parsing, recursion) crucial in Chomsky's theory can be explicitly formulated into a constraint-based grammar and implemented in Prolog. The Prolog interpreters provided with the book are basically identical to the high priced Prologs, but they lack the speed and memory capacities. They are ideal since anything learned about these Prologs carries over unmodified to C-Prolog and Quintas on the mainframes. Anyone who studies the prolog implementations of the lexicons and syntactic principles of combination should be able to use Prolog to represent their own linguistic data on the most complex Prolog computer available, whether their data derive from syntactic theory, semantics, sociolinguistics, bilingualism, language acquisition, language learning, or some related area in which the grammatical patterns of words and phrases are more crucial than concepts of quantity. The printed examples illustrate C-Prolog on an Ultrix Vax, a standard university configuration. The disk included with the book contains shareware version of Prolog-2 (IBM PC) and MacProlog (Macintosh) plus versions of the programs that run on C-Prolog, Quintas, Prolog-2, and MacProlog. Appendix II contains information about how to use the Internet, Gopher, CompuServe, and the free More BBS to download the latest copies of Prolog, programs, lexicons, and parsers. All figures (100+) in the book are available scaled to make full size transparencies for class lectures. Valuable special features of this volume include: * more than 100 full page diagrams illustrating the basic concepts of language processing, Prolog, and Chomsky's linguistic theories; * more than 100 programs -- illustrated in at least one script file -- showing how to encode the representations and derivations of generative grammar into Prolog; * more than 100 session files guiding readers through their own hands-on sessions with the programs illustrating Chomsky's theory; * a 3.5 disk (IBM Format) containing: 1. all programs in versions to run in C-Prolog or Quintas Prolog on an Ultrix Vax, and on an IBM PC and a Macintosh, 2. a shareware version of Prolog-2 for IBM PC clones which runs all programs in the book, 3. a shareware version of MacProlog for Macintosh which runs all programs in the book; * instructions on using Internet, CompuServe, and the free More BBS to download the latest copies of Prolog, programs, lexicons, and parsers; and * numerous references enabling interested students to pursue questions at greater depth by consulting the items in the extensive bibliography.

Attitude Markers and Conversational Implicatures in Turkana Speech Acts

Turkana has a set of particles expressing attitudes on the part of the speaker towards the propositional content of utterances in which such markers are used. Attitude markers in Turkana form a closed set whose distributional behavior partly follows from syntactic principles of the language. The absence of these attitude markers from certain syntactic positions follows from their lexical meaning and from pragmatic structure. Their current meaning is argued to have emerged through metonymic extension in certain lexical items, and through conventionalisation of their conversational implicatures. In addition some methodological issues are discussed concerning the interaction between grammar and culture-specific language use, by means of a comparison with similar markers in a number of other languages.

Syntactic codes and grammar refinement

We callsyntactic coding a technique which converts syntactic principles or constraints on representations into grammatical rules which can be implemented in any given rule grammar. In this paper we show that any principle or constraint on output trees formalizable in a certain fragment of dynamic logic over trees can be coded in this sense. This allows to reduce in a mechanical fashion most of the current theories of government and binding into GPSG-style grammars. This will be exemplified with Rizzi'sRelativized Minimality.

L2 acquisition of English synthetic compounding is not constrained by level-ordering (and neither, probably, is L1 )

This article investigates the acquisition of English synthetic compounding by native Spanish and native Chinese speakers. Data are presented which con tradict the claim by Gordon (1985), Clahsen (1991) and Clahsen et al. (1992) that morphological level-ordering is universally, innately available to lan guage learners to guide their acquisition of compounding constraints. Empirical arguments are given which show that compounding, at least, can not be universally subject to the particular inflectional constraints - namely, a restriction on plurals in compounds - imposed by the level-ordering mod els cited in the above acquisition studies. I also present additional experi mental results which demonstrate that L2 learners of English freely violate this restriction, and that such violations reflect particular L1 influence. I suggest an alternative approach to analysing the role of Universal Grammar in the acquisition of compounding which better accounts for both the L1 and L2 English data, by considering 1) the interaction of syntactic principles with lexical derivation; 2) the parametric differences between the L1 and L2; and 3) the language-specific nature of morphological affixation.

Syntactic codes and grammar refinement

We callsyntactic coding a technique which converts syntactic principles or constraints on representations into grammatical rules which can be implemented in any given rule grammar. In this paper we show that any principle or constraint on output trees formalizable in a certain fragment of dynamic logic over trees can be coded in this sense. This allows to reduce in a mechanical fashion most of the current theories of government and binding into GPSG-style grammars. This will be exemplified with Rizzi'sRelativized Minimality.

The Interaction of Referential Ambiguity and Argument Structure in the Parsing of Prepositional Phrases

This research addresses the question of whether the initial interpretation of structurally ambiguous sentences (such as He dropped the book on the chair) is made purely on the basis of syntactic principles or whether semantic processes also exert an influence. Two factors, referential ambiguity (e.g., whether there is more than one referent for the book), and verb argument structure (e.g., whether the verb takes on the chair as an optional or obligatory argument) were manipulated in two experiments using a subject-paced moving-window technique. These experiments found that the need to resolve an ambiguous noun phrase referent interacted with the obligatory/optional nature of verb arguments. Discourse information affected initial syntactic assignment of the prepositional phrase, but only for verbs taking optional goal arguments. For verbs taking obligatory goal arguments, the attachment decision was independent of whether the noun phrase was referentially ambiguous. A third experiment found that prepositional phrases are not initially interpreted as modifiers if a disambiguating adjective immediately precedes the noun, clearly implicating referential ambiguity as the mechanism affecting the interpretation of the prepositional phrase. These results support a limited influence of discourse semantics on the parsing of prepositional phrases and are inconsistent with models that propose no influence of semantics on syntax and models that propose complete interaction.

The balance between syntax and morphology: Dutch particles and resultatives

This paper focuses on Dutch verb-particle constructions and verb-resultative constructions. On the one hand, Dutch particles and resultatives share some properties; for instance, they mutually exclude each other. On the other hand, they show contrastive behavior with respect to, for example, movement. The similarities can be captured if to some extent, the two constructions receive the same analysis. It is argued that both particles and resultatives are base generated in a position adjoined to the verb. The differences between the constructions follow from the assumption that resultatives are adjoined to the verb at D-structure, while particles are adjoined to the verb in the morphological component. This analysis has several consequences for the syntaxmorphology interface: (i) there has to be a separate morphological component, (ii) the relation between this component and syntax is determined by generalizing metarules, and (ii) morphological structures are visible to syntactic principles such as the proposed constraint on the complexity of heads.

Perception, selection, and structural economy

In this essay I will explore the syntactic expression of the notion ‘clause’ by focusing on some syntactic and semantic properties of bare infinitive (BI) complements to perception verbs in English. I shall argue briefly that perception BI complements must be clausal, and then turn in more detail to the issue of what sort of clause the BI complement must be. It will be established that the categorical nature of the perception BI complement as IP or VP is contingent on selectional properties of the perception verb, properties which may be contextually influenced by modality and negation. More specifically, an interpretative distinction between ‘vision’ and ‘non-vision’see first suggested by Napoli (1988) will be shown, contrary to her proposal, to provide the key to the contrast in the selection of clausal types. The IP/VP distinction will be shown to correlate with the viability of interpretations for the BI clause that are ‘temporally independent’ of the matrix clause. Given the IP/VP distinction, it will be demonstrated further that the peculiarly limited distribution of the expletive formativethere in BI complements may be predicted from independently motivated syntactic principles that determine clausal structure.

Development of Principle B in Russian: Coindexation at LF and Coreference

This article presents evidence in the development of Russian for a theory that says that children at a terrain age know a syntactic principle (Principle B of the Binding Theory) that governs the distribution of pronouns but that they do not know a pragmatic or semantic principle (Principle P) that restricts the situations in which NPs may be contraindexed. The major result concerns Russian possessive pronouns that must be disjoint from subject antecedents. We show that this result follows from the raising of pronouns at Logical Form together with Principles B and P. The prediction thus is that children who do not know Principle P will accept this sentence. We experimentally confirm this result. We also predict that children will reject sentences with possessive pronouns that are bound by quantifiers. We also confirm this prediction. Along the way we replicate in Russian known results in English and other languages. The pattern of results provides strong support for the Principle P/modularity theory.

Negative polarity: A semantico-syntactic approach

Negative Polarity Items (NPIs) e.g. any, are licensed in negative sentences, questions, antecedents of conditionals, complements of adversative predicates, etc. On the other hand, Positive Polarity Items (PPIs), e.g. some, cannot appear in the scope of clausemate negation, but are licit elsewhere. I propose the following two (syntactic) principles and a (semantic) filter to account for the distribution of NPIs and PPIs: (1) NPIs are subject to Principle A of the Binding Theory. (2) PPIs are subject to Principle B of the Binding Theory. (3) UEFilter: *Polarity operator in an Upward-Entailing (UE) clause. (A potential binder for polarity items in either negation, or an empty polarity operator licensed in the Comp of non-UE clauses.) The binding principles (1) and (2) (see Progovac 1988, 1991) capture the following general facts: (4) NPIs (as anaphors) need a licenser. (5) The licenser for NPIs has to be local. (In languages with long-distance NPI licensing, such as English, NPIs raise at LF in order to be bound locally.) (6) PPIs (as pronominals) need no licenser. (7) PPIs cannot be bound by a local licenser. The role of the semantic filter (3) is to predict licensing of NPIs in exactly those environments in which they occur. An UE clause allows inferences from subsets to supersets, as in: (8) Mary ate kale. =) Mary ate a green vegetable. The switch to Upward Entailment can accommodate two problematic cases for Downward Entailment (DE) (cf. Ladusaw 1980, 1982, 1983): yes/no questions and on/y. Although neither DE nor UE, they license NPIs. While Ladusaw assumes that DE licenses NPIs directly, it is claimed here that a polarity operator is licensed in the Comp of a non-UE clause, and that the operator in turn licenses NPIs. This assumption can account naturally for a set of otherwise puzzling facts: (i) that non-negative

An application of pattern mapping to plane motion

A new, highly automated method for measuring plane motion, pattern mapping, has been developed for rigidbody motion and strain analysis. Pattern mapping employs image processing and syntactic pattern recognition principles to recognize a known pattern before and after motion. In this paper, the Lagrangian definition of motion was used to map points in the two images, and the map was used to determine plane motion. Full-image analysis is thoroughly demonstrated.

Children's Knowledge of Locality Conditions in Binding as Evidence for the Modularity of Syntax and Pragmatics

We report three experiments concerning English-speaking children's knowledge of locality conditions in the binding of reflexives and pronouns (Principles A and B). The children tested were between the ages of 2;6 and 6;6. By age 6, children know that a reflexive must be locally bound. At the same age, however, they appear to not know that a pronoun may not be locally bound. We suggest that children are missing a pragmatic principle, not the syntactic Principle B. This hypothesis predicts that children will not accept a local antecedent for a pronoun that is a bound variable. Experiment 4 confirms this prediction. We conclude that children know the grammatical principles of binding but do not know a relevant pragmatic principle. We suggest that such dissociation in children might be a useful tool in the study of linguistic theory.

The syntax of participial adjuncts in Eastern Bengali

In this article I will discuss the syntactic properties of participial adjuncts in Eastern Bengali. From a GB-point of view, these constructions are quite interesting, because they can contain a nominative which is apparently not assigned by AGR, and because they show a quite intricate pattern of possibilities for coreference and disjoint reference, with some seemingly arbitrary differences among the three types of adjuncts. In Section 2, I shall present the empirical data for these constructions. In Section 3 I will discuss the relevant general syntactic principles of Eastern Bengali, specifically those responsible for Case-assignment, word-order,pro-drop (including ergative verbs) and binding properties. In Section 4, I will show that the characteristics of the participial constructions, including the apparently haphazard binding properties, follow from the general syntactic principles of Eastern Bengali, if we assume one simple statement for each type of adjunct. No further construction-specific stipulations need be made. To the extent that the analysis I propose can be upheld, it will constitute indirect support for the GB-mechanisms that are crucially involved in it. Apart from various principles of configurationality and binding, I will make use of the idea that there is no abstract AGR, in these cases at least, and also of the analysis ofpro-drop as a silent clitic phenomenon, proposed in Safir (1985). It is of course a fact that the principles of grammar I appeal to still need to be investigated more carefully, and may have to be modified on the basis of data not yet taken into account or accurately analyzed. However, as they stand, the relevant principles of GB-theory appear to be able to account for the Eastern Bengali facts I discuss here. Apart from these more theoretical concerns, this paper naturally has an important descriptive component too, which I hope will make it also interesting to linguists working in a different theoretical framework and may stimulate linguists specifically concerned with Bengali to explore further the intricascies of this area of Bengali grammer.

Constituent attachment and thematic role assignment in sentence processing: Influences of content-based expectations

The experiments reported here use individual word reading times in a self-paced word-by-word reading task to examine the processing of prepositional phrase constituents in sentences like “The spy saw the cop with a revolver but the cop didn't see him”. In Experiment 1 we show that reading times for words immediately following the prepositional phrase (“with a revolver”) are predictable from the consistency of subjects' expectations for the attachment of such prepositional phrases with the attachment dictated by the content of the prepositional phrase itself. These expectations vary from sentence to sentence with the content of the material preceding the prepositional phrase and do not appear to reflect the syntactic principle of Minimal Attachment. Experiment 1 conflated violations of expectations for prepositional phrase attachment with violation of role and filler expectations; Experiment 2 examined the contribution of each of these three types of expectation violations to the slowing of reading times. Violations of filler expectations that did not change expected role or attachment produced a small but significant slowdown in processing the words just following the prepositional phrase. Violations of thematic role expectations and filler expectations produced a much larger slowdown, but violation of attachment expectations as well as filler and role expectations did not produce any additional slowing of processing. We interpret these results as supporting models of sentence processing in which thematic role expectations for upcoming constituents play a role in guiding the interpretation of these constituents as they are initially processed.

Ambiguity, parsing strategies, and computational models

Abstract A variety of computational models have been developed in recent years to model the behaviour of the Human Sentence Processing Mechanism (HSPM) when it encounters local syntactic ambiguities. The majority of these incorporate the assumption that the HSPM makes its initial choice of analysis according to a small number of exclusively syntactic principles. The arguments in favour of this structural approach range from the computational efficiency of parsers incorporating these structurally based parsing strategies to the empirical evidence which has been claimed to refute the alternative interactive account of ambiguity resolution, in which contextual information can be used in order to determine the initial choice of analysis. The present paper reviews some of these models, and argues that the assumptions concerning their underlying computational efficiency are flawed. Experimental evidence is presented which is suggestive of the alternative interactive account.

Structure Dependence in Grammar Formation

Structure Dependence in Grammar Formation

ANOTHER LOOK AT NEGATIVELY CONDITIONED SUBJECT‐OPERATOR INVERSION IN ENGLISH1

SummaryThe regularization of subject ‐ operator inversion as used after negatives, restrictives, and certain correlatives is a process that has been going on for centuries and may still be going on, for all we know. It started with full negatives like neither and was then extended to near‐negatives, restrictives, and correlatives (including so + adjective/adverb. that). Various attempts have been made to find the rationale underlying this process, but none of the concepts proposed so far can be raised to the status of an all‐embracing explanatory principle. Emphaticness is obviously one of the factors involved, but it is neither a necessary nor a sufficient condition for the use of inversion. Affect‐attraction, as defined by Klima, is broad enough to accommodate both interrogative and negative inversion but too narrow to account for inversion with so. Negative attraction, also invoked by Klima and by others before and after him, is a more likely candidate, but it has to be defined and explained in its turn. The conclusion arrived at in the present study is that negative inversion serves a twofold purpose: syntactically, it brings about a closer connection between negative and verb (negative attraction or, more generally, connectedness); semantically, it has the effect of making the clause non‐affirmative, thus eliminating the undesirable sequence negative opener + affirmative S‐V. In borderline cases, such as sentences with preposed not without reason or not for nothing, it is usually the syntactic principle that prevails, i. e. inversion is used in spite of the fact that the sentence is affirmative in meaning. The syntactic principle also accounts for the regular use of inversion after preposed correlatives, whether near‐negative like hardly or non‐negative like so.

Phonological Resolution of Syntactic Feature Conflict

Agreement and government can give rise to syntactic feature conflicts; e.g., a verb may have to agree in gender with a coordinate NP composed of NP's of differing genders. The resolution of such conflicts is to some extent determined by general semantic or syntactic principles ('principled resolution'). In addition, the existence of a governed or agreeing form that is neutral with respect to the features in question always permits a conflict to be resolved. There is idiolectal variation as to when principled resolution occurs, and also as to which neutral forms appear in the lexicon. But there are also cases (examples are provided from English, Xhosa, and German) in which resolution is permitted on the basis of a purely phonological resemblance; i.e., the fact that the conflicting morphosyntactic requirements lead accidentally to similar phonological consequences is sufficient to prevent the conflict from causing ungrammaticality. Four major claims are made here about the phenomenon of phonological resolution. First, it does not imply that there can be language-particular phonological conditions on syntactic rules; a universal statement is involved. Second, phonological identity cannot compensate for syntactic feature distinctness in general, even in those languages that clearly exemplify phonological resolution; the conflicting items must also differ only in syntactically imposed features. Third, it is not the underlying phonological form of an item that determines whether it can count as phonologically identical to another item in cases of true phonological resolution; these items must be identical at a more superficial level (conjectured to be identical to the 'lexical stratum' of lexical phonology). Fourth, a small sampling of data (from Icelandic and Finnish, as well as the languages already listed) suggests variation with respect to the number of syntactic feature differences that can be surmounted in phonological resolution.*

Logical Aspects of Combinatorial Duality

AbstractD. R. Woodall has introduced closely-related notions of Menger and König duals which can be applied to a broad range of combinatorial contexts. The present paper considers these two notions for finite ground sets in terms of syntactic duality principles. Specific graph-theoretic interpretations are cited.

Differentiation and Syntax in the Evolution of Behavioral Flexibility

Differentiation and Syntax in the Evolution of Behavioral Flexibility