Central Laboratory for Parallel Information Processing,
Bulgarian Academy of Sciences
Cann, Ronnie, Claire Grover and Philip Miller, ed. (2000) Grammatical Interfaces in HPSG. CSLI Publications, paperback, xxii+292 pp., ISBN 1-57586-314-6, Studies in Constraint-Based Lexicalism 8
The book is an edited collection of papers, many of them presented at the Sixth HPSG Conference, held at the University of Edinburgh in August 1999. An introduction by the editors overviews the papers exploring interface phenomena between various grammatical components within HPSG. The papers deal with a whole range of languages and constructions of those languages. The book consists of 15 chapters, each one of them being an independent article (including the bibliography referred to in the text). There is also a list of the contributors and their affiliations, as well as a Subject Index and a Name Index.
Chapter 1 is entitled „Incorporation in Danish: Implications for Interfaces“ and is written by Ash Asudeh and Line Hove Mikkelsen. The authors explore the syntax-prosody interface in their attempt to provide a formal analysis of a linguistic phenomenon in Danish: syntactic noun incorporation. Their approach is compatible with the core HPSG notion of parallelism in information encoding and introduces a lexical account of the Danish syntactic noun incorporation that extends the HPSG phonology in a non-trivial way. The phenomenon is first linguistically analyzed on all four basic levels of description: phonology, syntax, semantics and pragmatics. The formal analysis shows how phonological phrasing can be built in parallel with syntactic combination and is based on three mechanisms: 1) the multiple-inheritance type hierarchy of Abeille and Godard (2000) which cross-classifies the type sign for weight and phrasality (lite/non-lite signs, words/phrases); 2) description level lexical rules (Meurers, 1999) stating the lexical relationship between normal transitive verbs and syntactic-noun-incorporation verbs; 3)phonological constituency stemming from the string-based approach of Bird and Klein (1994), and Monachesi (1999) for partition of the sort „phon“ and appropriate extensions to the PHONOLOGY feature in the HPSG feature structure descriptions.
The morphology-syntax interface is demonstrated in Chapter 2, „Incorporating Contracted Auxiliaries in English“ by Emily M. Bender and Ivan A. Sag, whose topic is the auxiliary contraction in English. A previous approach to the phenomenon is that of Sadler (1997) and Barron (1998) who propose an inflected-pronoun analysis where the tense inflection is treated as a morpheme attaching to a nominal that does not head the sentence. Bender and Sag suggest an alternative to the „inflected subject“ analysis: they propose to treat contracted auxiliary forms as verbs with an incorporated pronominal subject. In such a way, the fused lexical element is the head of the clause and this representation is consonant with the cross-linguistic nature of headed constructions. Argumentation is provided in favour of the „subject incorporation“ analysis, which is linguistically motivated for English, conforms to HPSG basic assumptions about headedness and locality, and makes use of the HPSG well-established conception of valence.
In Chapter 3, „Negation in Colloquial Welsh“ by Robert D. Borsley and Bob Morris Jones, an HPSG analysis of negative sentences (main clauses) in colloquial Welsh is proposed. A considerable amount of data is presented which illustrates the interplay between verb forms and the positioning of certain verbal particles and negative words (n-words) responsible for the negative or non-negative interpretation of the sentence. Types of negative verbs are defined and their syntactic relations with negative words (n-words) are postulated. The authors foresee two possible mechanisms for capturing generalizations over negative verb categories and their interaction with n-words: 1) lexical rules for providing the necessary negative verb categories; 2) formulation of constraints over certain basic underspecified verbal categories. The second approach is the preferred one due to two main reasons: a number of members of the HPSG community have argued against the use of lexical rules for capturing relations between lexical items; the first approach misses a generalization of the similarities between the four types of negative verbs that are distinguished. The HPSG analysis is based on the distinction of the selectional properties of the types of negative verbs compared to the affirmative ones. The authors demonstrate the appropriateness of their formal analysis and suppose that this analysis for Welsh is „potentially relevant to both English and French, and perhaps to other languages as well“ (p.49).
Chapter 4, „Argument Realization and Dutch R-Pronouns: Solving Bech’s Problem without Movement or Deletion“ by Gosse Bouma, as the title suggests, presents an analysis of R-pronouns in Dutch which accounts for their properties as extracted arguments of prepositions, as adverbs of the Middle Field of the Dutch sentence binding a gap originating in a prepositional phrase, and as subjects. Previous analyses of the phenomenon involve movement of R-pronouns in the phrase structure or, in the case of HPSG treatment, introduce an ad hoc phrase structure schema. G. Bouma’s analysis of the Dutch R-pronouns follows the formalization presented in the article of Bouma et al. (2001) which has already become a milestone in the development of HPSG: an auxiliary level of representation DEPS (dependency structure) is the mediator between the ARG-ST (argument structure) and VALENCE of lexical signs. Appropriate principles and constraints are formulated (like the Verbal Argument Structure Extension Principle, the Prepositional Argument Realization, the Bind R-Pronoun Principle, the Passive Argument Structure Extension, the Locative Subject Constraint), which interact with the principles and constraints that are at the core of Bouma et al.’s (2001) formalization (like Argument Realization and Slash Amalgamation).
The interface between morphology and syntax is also explored in Chapter 5, „Syntactic Transparency of Pronominal Affixes“ by Berthold Crysmann. Bound pronominals in European Portuguese are examined which challenge the dichotomy established in clitic analysis where clitics are formally described either at the morphological level as lexical affixes, or at the syntactic level in terms of linearization. Clitics in European Portuguese occupy a transitional place between morphology and syntax. From one side, they clearly manifest the morphological properties of affixes, and at the same time, their placement preverbal or postverbal is clearly conditioned by the presence and position of distinct classes of syntactic items: wh-phrases, nominal and adverbial quantifiers. B. Crysmann provides a linearization-based analysis triggered by a technically suitable representation of the clitic cluster and the verbal host as separate domain objects. Besides the strong argument of the existence of classes of triggers and non-triggers of proclisis, the analysis is favoured by another syntactic factor: the possible intervention of syntactic material between the clitic and its host. Thus a multi-dimensional HPSG grammatical representation is provided where the internal morphological structure of the clitic cluster is encoded in the lexical component in a way suitable for the syntactic treatment of clitic placement in the word-order component.
Chapter 6, „From Argument Raising to Dependent Raising“ by Kordula De Kurthy and W. Detmar Meurers addresses the partial constituent phenomenon in German. The phenomenon can be, so to say, split into two parts: 1) constituents are fronted leaving behind one or more complements: this is the partial constituency that is extensively discussed in the literature; 2) constituents are fronted leaving behind adjuncts: this sub-phenomenon has generally been neglected in the literature. In this paper the authors concentrate on the second sub-phenomenon. The assets of their analysis lie not only in the fact that they promote the much less discussed side of German partial constituent fronting, but, what is more, they provide an integrated HPSG analysis that captures the parallel character of the two kinds of partial constituent phenomena and derives both from the same underlying mechanism. In the HPSG tradition of Pollard and Sag (1994), adjuncts are not represented lexically but are inserted syntactically in a head-adjunct structure. The recently emerging and gaining grounds idea of lexicalizing the selection of adjuncts (the most recent development being the above mentioned Bouma et al. (2001)) provides the underlying mechanism for a uniform analysis of lexical dependents (both arguments and adjuncts). The authors extend their lexical argument-raising principle, so that partial constituents can result not only from raising of complements but also from raising of adjuncts.
Chapter 7, „The Nature of Pragmatic Information“ by Georgia M. Green discusses the character of pragmatic information revealed in its relation to the rest of the information carried by the signs of language. The underlying statement in this paper is that pragmatic information is about mental models (that is, the speaker’s and the addressee’s mental models of each other) and that neither the problem, nor the solution of context and background presupposition (the HPSG repositories of the information generally treated as stemming from the level of pragmatics) is strictly linguistic. Conversational implicature is a function of a general theory of action, in particular, a certain kind of intentional action and its interpretation. The idea suggested in the paper is demonstrated in discussion of sentence-final particles in languages like Chinese and Japanese, sentence-initial discourse markers and honorifics. The inference from the discussion is that it is not necessary to try to encode this part of the information, which turns out to be not of linguistic nature, in the CONTEXT and BACKGROUND component of a feature structure description of a sign. Thus the formal analysis correctly avoids the overload of the pragmatics zone of the sign description with rather complex and highly structured linguistic information, which, intended to convey facts about the utterance, will never be enough and sufficient to explain the utterance due to the fact that the pragmatic side of the utterance tends to be explained in terms of world knowledge rather than in an exhaustive description of the linguistic reality. General principles like the CONTEXT Inheritance Principle of Pollard and Sag (1994) or the Cooperative Principle (Grice, 1975) (together with Grice’s „Maxims“) would suffice to operate on a grossly but appropriately structured pragmatic component of the description of a language sign.
Chapter 8, „The Morphosyntax of Lai Relative Clauses“ by Andreas Kathol advocates a lexicalist HPSG analysis of the relative clause construction in Lai (Hakha Chin), a Tibeto-Burman language spoken in Western Burma (Chin State). Relative clauses in Lai are formed by means of an invariant relative marker following the verb and depend on the fact that Lai is a predominantly SOV and hence a head-final language. A. Kathol argues that a long-distance approach is not suitable for Lai relatives. This statement is supported by the fact that there is lack of evidence for unbounded dependencies of the filler-gap sort in other parts of the grammar of Lai. A. Kathol proposes a lexicalist analysis that triggers generalizations about the relationship between verbal morphology and a relativized grammatical function for two types of constructions: internally and externally headed relative clauses. The analysis follows the recent developments in HPSG in terms of the interface between the two levels of representation in the feature structure description of lexical signs: argument structure (the invariant syntactic structure) and valence (the overt syntactic realization of arguments in a given construction).
Relative clauses challenge the HPSG framework also in Chapter 9, „Grammatical Interfaces in Korean Relatives“ by Jong-Bok Kim and Byung-Soo Park. In this paper Korean relative clause constructions are analysed whose structural complexity (strong context-dependence and violation of the canonical, syntactic island constraints) requires grammatical interfaces among morphology, syntax, semantics and pragmatics. On the morphological level, the verbal form of the highest verb in the relative clause marks the realization of that clause. This modification information is distributed in sentences in accordance with syntactic constraints which further interact with lexical, semantic and pragmatic factors. The constraints ensure the account of relativization of both complements and adjuncts, based on Bouma et al.’s (1998) analysis involving a DEPENDENTS level of representation in a feature structure description. It should be noted that the authors employ information structure (topic/focus constructions) to provide analysis of the specific Korean phenomena. It is demonstrated that the complex, at first sight, interaction of different grammatical levels results in a simpler grammar of the whole variety of Korean relative constructions.
In Chapter 10, „Prosodic Constituency in HPSG“ by Ewan Klein, the relation between prosodic and syntactic structure is explored and it is demonstrated how prosodic constituency can be built recursively in parallel with the definition of syntactic constituency. The analysis is restricted to ‘core’ syntactic constructions including transitive and ditransitive VPs, modals, nominal complements and pronominal subjects. The standard HPSG approach to phonology is elaborated by modifying the value of the phonology feature in the description of signs which now can be of two types: word-forms and metrical trees. Metrical trees introduce prosodic domains (possibly nested) and a way of marking the most prominent element in a given domain. A prosodic relation between phonology values and metrical trees is defined and this relation is incorporated into prosodic constraints within a constructional hierarchy. Thus the analysis shows how HPSG techniques can license the expression of syntax-prosody mismatches, define independent prosodic constraints and relate prosodic structure to syntactic phrases by means of a multiple inheritance hierarchy in which the dimensions of clausality and headedness are augmented by the dimension of prosody.
Chapter 11, „A Head-Driven Account of Long-Distance Case Assignment“ by Robert Malouf presents an analysis of case assignment that is generalized enough to capture the more complex case systems of some languages where case is associated with a particular structural position rather than a particular lexical head. R. Malouf provides evidence from some Australian languages characterized by non-local case marking. The author provides an HPSG head-driven account of case stacking and long-distance case assignment by generalizing the value of the CASE feature so that it becomes a list of morphosyntactic cases. In this way, the core HPSG analysis of case assignment as a local relationship between a lexical head and its arguments is preserved and there is no addition of new mechanisms of a structural nature (contrary to some other approaches mentioned in the paper). A Case Concord Principle and a Case Realization Principle are formulated on the basis of Bouma et al.’s (2001) geometry of feature structure descriptions of lexical signs (three-level representation of the subcategorization properties: ARGUMENT STRUCTURE, DEPENDENTS and VALENCE). R. Malouf claims that the analysis presented in the paper sheds some light on a number of ‘case attraction phenomena’, in which a constituent bears the case of a larger constituent of which it is a part. Examples of Classical Armenian and of Gothic support this claim.
In Chapter 12, „German Particle Verbs and the Predicate Complex“, the author, Stefan Mueller, argues that preverbs or verbal particles (preferred recent terms for the traditionally called separable prefixes of German finite verbs) behave similarly to other elements of the predicate complex (an extended notion of verbal complex). Evidence for the above statement lies in the following linguistic facts: preverbs can be fronted in the same way as other parts of the predicate complex are; preverbs are serialized like verbal or predicative adjectival complements in the right periphery of a clause; resultative constructions and object predicatives share a lot of properties with particle verbs. The HPSG analysis that S. Mueller proposes treats preverbs as part of the verbal complex, hence they are selected by the same valence feature as other elements that form a complex with their head. This analysis accounts for the similarities that particle verbs share with resultative constructions and object predicatives. The advantage of the approach is in the uniform analysis of a range of constructions, which leads to generalization of existing mechanisms rather than introduction of new ones.
In Chapter 13, „A Constraint-Based Semantics for Tenses and Temporal Auxiliaries“ by Frank Van Eynde, the syntax-semantics interface is demonstrated by the development of a constraint-based semantic representation of a substantial part of the tense-aspect system of Dutch. The analysis is inspired by insights of Donald Davidson, Hans Reichenbach and Discourse Representation Theory (DRT) but has become an integrated part of the HPSG framework. The semantic representation of tensed verbs is facilitated by the introduction of verbal indices as a value of the CONTENT feature of verbal signs, which thus obtain the same format as nominal signs. By adding appropriate features to the structure of CONTENT and CONTEXT in the descriptions of verbal signs, information is encoded about the relation between utterance time, reference time and event time. The analysis of the different aspectual tenses in Dutch and the proper semantic interpretation of sentences are guaranteed by the introduction of a sortal distinction between substantive and nonsubstantive verbal signs and the application of a small set of implicational constraints. The syntax-semantics interface described in this paper is non-directional and that is advantageous for expressing the relation between the syntactic, semantic and pragmatic properties of signs.
Chapter 14, „Spanish Psychological Predicates“ by Carl Vogel and Begona Villada examines the issue of argument structure and subject-verb agreement in Spanish psychological predicates. In view of the conclusion about the problematic notion of sentence „subject“ and the idiosyncrasies of the argument structure and agreement control of predicates in Spanish, the authors revert to the first version of HPSG (Pollard and Sag, 1987) for handling the language phenomenon in terms of positioning of the arguments in the single SUBCATEGORIZATION list of the predicate. The Spanish psychological predicates are grouped in five valency patterns depending on the relation between argument position and agreement control, and between the semantic entities of cause and experiencer. The analysis is based on the introduction of positional features (like the head feature POS (position)) associated with lexical items. The subcategorization frame of a verb is the local domain in which the head occurs, and a lexical head declares its position within that domain as at the beginning, the end, or somewhere in the middle. Related to argument structure and agreement is the problem of raising and control in Spanish. The authors add the head feature AGRC (agreement controller) to the description of signs for capturing subject and object control, as well as subject raising (object raising is not permitted) in Spanish.
In Chapter 15, „A Bistratal Approach to the Prosody-Syntax Interface in Japanese“ by Kei Yoshimoto, an HPSG-based analysis integrates information from what is traditionally assumed to be the syntactic structure and independently represented prosodic information. The two types of information are encoded in the proposed P(ROSODIC)-STRUCTURE and C(ONSTUTUENT)-STRUCTURE of the feature structure descriptions of signs. Keeping the distinction between phonology and syntax intact, minimal interfaces are posited for capturing the bi-directional interdependencies between the two strata of grammatical representation. The definition of the interface from C-structure to P-structure is connected with the phenomenon of metrical boost, an effect of the syntactic structure on the pitch of the modifying phrase, where the information about the branching of the syntactic tree is percolated to the level of prosody. The interface from P-structure to C-structure is defined in terms of constraining the position of Japanese interjectional postpositions via the Prosodic-Hierarchy Principle operating on the structurally enriched HPSG stratum of phonology.
In sum, this collection of papers is a snapshot of the development of the theory and practice of HPSG, from which several tendencies are inferred. The feature structure geometry is continuously elaborated by means of structural expansion of the phonology, semantics and pragmatics components of the description of linguistic signs. In parallel, the interaction of the information encoded at all levels of grammatical representation is explored and the appropriate mechanisms for supporting this interaction are devised. The general tendency towards lexicalism in the linguistic analyses is confirmed and further deepened: the analysis of Lai relatives (Chapter 8), as well as the adoption of the integrated lexicalist approach to argument and adjunct selection (Bouma et al., 2001) as the grounds for the analysis in Chaper 6, Chapter 9, Chapter 10, and Chapter 11 are just the most prominent manifestations. Also the tendency of drawing generalizations over constructions, languages and representational devices is evident from this valuable collection of HPSG grammatical analyses.
Abeille, A. and D. Godard (2000). French word order and lexical weight. In R. D. Borsley, ed., The Nature and Function of Syntactic Categories, New York: Academic Press.
Barron, J. (1998) Have contraction: Explaining „trace effects“ in a theory without movement. Linguistics 36(2):223-251.
Bird, S. and E. Klein (1994) Phonological analysis in typed feature systems. Computational Linguistics 20:55-90.
Bouma, G., R. Malouf, and I. A. Sag (1998) A unified theory of complement, adjunct and subject extraction. Proceedings of the FHCG-98, pp. 83-97.
Bouma, G., R. Malouf, and I. A. Sag (2001) Satisfying constraints on extraction and adjunction. Natural Language and Linguistic Theory 19:1-65.
Grice, H. P. (1975) Logic and conversation. In P. Cole and J. L. Morgan, ed., Syntax and Semantics 3: Speech Acts, New York: Academic Press.
Meurers, W. D. (1999) Lexical generalizations in the Syntax of German Non-Finite Constructions, Ph.D. thesis, Seminar fuer Sprachwissenschaft, University of Tuebingen, Germany, Volume 145 in Arbeitspapiere des SFB 340.
Monachesi, P. (1999) A Lexical Approach to Italian Cliticization. CSLI Publications, Stanford.
Pollard, C. and I. A. Sag (1987) Information-Based Syntax and Semantics, Volume 1: Fundamentals. CSLI Publications, Stanford, CA.
Pollard, C. and I. A. Sag (1994) Head-Driven Phrase Structure Grammar. Chicago, IL: University of Chicago Press.
Sadler, L. (1997) English auxiliaries as tense inflections. To appear in the Special Issue of Essex Research Reports produced on the occasion of the retirement of Keith Brown.
The reviewer, is currently a team member of BulTreeBank (HPSG-based Syntactic Tree Bank of Bulgarian): a joint project between Linguistic Modelling Department, Bulgarian Academy of Sciences and Seminar für Sprachwissenschaft (SfS), Eberhard-Karls-Universität, Tübingen, Germany. Her research interests are in the application of the theory and formalism of HPSG to Bulgarian (argument structure and reflexivity of verbs), as well as in the standardization and automation of the production and processing of annotated electronic text corpora.