Jerrold M. Sadock

AUTOLEXICAL SYNTAX: A PROPOSAL FOR THE TREATMENT OF NOUN INCORPORATION AND SIMILAR PHENOMENA*

1. BACKGROUND According to the standard picture of the interface between syntax and morphology, the words of well-formed natural-language expressions are attached to the leaves of the syntactic phrase structure trees. Furthermore, each word may be represented by a structured tree whose leaves represent the constituent morphemes, and each morpheme is a (presumably structureless) list of segments. The representation of the structure of the expression is thus a smoothly branching tree with the highest syntactic category at the top, and individual segments at the bottom, as in diagram (1), where A-E represent syntactic categories with D a lexical category, a-d represent subword constituents where b, c and d are simple morphemes, and S1-Sm are phonological segments.

A machine learning approach to modeling scope preferences

This article describes a corpus-based investigation of quantifier scope preferences. Following recent work on multimodular grammar frameworks in theoretical linguistics and a long history of combining multiple information sources in natural language processing, scope is treated as a distinct module of grammar from syntax. This module incorporates multiple sources of evidence regarding the most likely scope reading for a sentence and is entirely data-driven. The experiments discussed in this article evaluate the performance of our models in predicting the most likely scope reading for a particular sentence, using Penn Treebank data both with and without syntactic annotation. We wish to focus attention on the issue of determining scope preferences, which has largely been ignored in theoretical linguistics, and to explore different models of the interaction between syntax and quantifier scope.

Preverb Climbing in Hungarian

This paper offers an account of the distributional possibilities of Hungarian preverbal particles in the autolexical framework proposed in Sadock 1985. We show how a complex range of facts can be predicted from the interaction of a few independently motivated assumptions once autonomous modules of the grammar are allowed to assign partially noncongruent representations to the same linguistic entity. In the course of the discussion the Incorporative Principle of Sadock 1985 is extended and turned into a general constraint on allowable discrepancies between representations contributed by various components of the grammar.*

Affixial verbs in syntax: A reply to Grimshaw and Mester

ConclusionsWe have pointed to a range of data which fulfill exactly the predictions G & M ascribe to a syntactic theory of Eskimo complex verb constructions. In particular we have found that complex verbs occur in different phrase structure configurations from ordinary verbs and trigger different kinds of case marking and that anaphoric processes must recognize clause boundaries within complex verbs; that the internal structure of certain complex verbs is unique in a way that argues for word-internal clause boundaries; that derivations by means of suffixes of the -kqu- class have the same sorts of properties as syntactic causatives; that certain suffixes distinguish between basic and derived stems, and that certain derived forms do not interact freely with other word formation processes.We therefore suggest that any adequate theory of Eskimo morphology must recognize the syntactic salience of certain word-internal formatives. An adequate theory must, in particular, provide a level of analysis at which there are clause boundaries that do not correspond to the word structure of the language.For one such theory which has the advantage of not requiring the interspersal of syntactic and morphological rules, see Sadock (1985). In all of this, we have been critical of an application of (a particular version of) the Lexicalist Hypothesis. But it is far from our intention to banish the hypothesis absolutely. Rather, we seek to clarify two logically separate sets of principles which in various formulations of the Lexicalist Hypothesis have tended to merge. First there is the recent notion that the formal word — with or without clitics and inflection — is the output of the lexicon and the sole formative in syntax (Aronoff, 1976; Lapointe, 1980; Anderson, 1982). This is the view G & M take. For Eskimo and other polysynthetic languages this isolates what Swadesh (1939) termed an “external syntax” (as opposed to an “internal syntax” within words). Such an approach does correctly isolate surface words in Eskimo which, though elaborate, manifest known phenomena and principles of formal morphology in quite an unexceptional way (see Sadock, 1980, 1985). On the other hand when taken by itself, Eskimo external syntax is aberrant and even incoherent given known principles of syntax in other languages such as Functional Uniqueness and Structure Preservation. Thus the formal surface word provides criteria that are only partly useful in Universal Grammar.The second set of principles has to do with complete distributional and semantic productivity and was proposed in Chomskys (1970) original Lexicalist Hypothesis. For English and other relatively analytic languages these principles yield syntactic formatives only slightly different from surface words: productively-formed destroying contains two formatives, while (slightly) irregular destruction contains one. For Eskimo languages however, these principles divide most words into productive bases and suffixes (such as those we have been considering in this paper).Like full words in less synthetic languages, these formatives themselves usually contain distributionally and semantically irregular formations, leading to a distinction between less-than-productive ‘real’ morphology and a totally productive ‘internal syntax’. Two excellent examples of Eskimo ‘real’ morphology are the causative transitivizer +te- in (24)–(26), and the irregular antipassives following -guma- discussed in reference to G & Ms prediction VI: the former is an element in the formation of productive bases, the later in the formation of productive suffixes. Our contention has been that when the syntax is allowed access to these productive elements, the aberrations of pure ‘external syntax’ will disappear while familiar phenomena and principles of syntax will reemerge. If we are correct, then this earlier set of lexicalist principles also belongs in Universal Grammar. The task then must be to determine how much and in what ways formal morphological principles and distributional syntactic principles interact and constrain each other in an explanatory Universal Grammar, how the roles of each can be restricted, and how syntactically analyzed strings can and cannot map into morphological analyses of the same strings. Research into the many polysynthetic languages of the Americas and elsewhere, undertaken with genuine openness to richness that may be found, must play a major role in this.Allen et al. (1984) on Southern Tiwa and Goddard (to appear) on Fox, in very different ways, point out the importance of this problem and the breadth of the North American data.

The Lexicon as Bridge between Phrase Structure Components

Context-free phrase structure grammar (henceforth PSG) is capable of describing infinite languages consisting of finite strings drawn from a finite vocabulary and associating with each string of the target language a division into immediate constituents that can be represented as a labeled tree. This much, it would seem, is the least that we can expect of any grammar that would pretend to adequacy as a scheme for describing languages of the kind spoken by human beings. PSG has several advantages over competing systems of grammatical description. In particular: PSG is a conceptually simple formalism. Because PSG describes only a single level of analysis, movement rules and the constraints on them and triggers for them are eliminated. The ramifications of a postulated PSG rule are ordinarily immediately obvious and immediately testable against fact. PSG is completely formalizable. A finite set of rules and a finite vocabulary fully defines a PSG. The formal properties of such systems can be studied, and in fact there is a considerable formal literature on the subject. (See, for example Partee, et al. 1990 and the references it contains.). There are numerous interesting results concerning the classes of string sets and the classes of tree sets that PSGs are capable of specifying. PSG is computationally tractable. Quite efficient algorithms exist for parsing expressions in terms of PSGs. Though the parsing time of the best of these is still an exponential function of the length of the string, it is less disastrously so than for other formal grammar types.

Remarks on a west greenlandic verbal paradigm

Abstract Eskimo languages, as is well known, are ergative-absolutive .with respect to the case marking of the nominal terms of the clause: there is a special morphological case that marks the agent of a transitive clause, while the object of the transitive and the subject of an intransitive are in a different case. The case of the transitive agent is traditionally called the relative in Greenlandic studies (following Kleinschmidt 1968 (1851)), but I will refer to it as the ergative, since that is the term that is now more generally used in typological studies. The case of the transitive object and intransitive subject is the absolutive.

Two Views of

internalizing the statements in the grammar and developing skills in analyzing the complex words of Nahuatl. The vocabulary, while still providing a display of the lexical structure of stems, would be much more useful for the beginning translator if it were complete for the corpus in the workbook, including the supplementary texts, and if it included inflectional affixes and function words. The course is a contribution to pedagogical grammars, in particular to those on Classical Nahuatl. However, one wonders about the economics involved, which prices the combined 724 pages at

Speech act distinctions in syntax

52.50. Was it because an unduly limited number of copies were printed? It could not be due to the need for a special type font, apart from perhaps the macron over the vowels. Even the paperback workbook with 222 pages is overpriced at

Ambiguity tests and how to fail them

14.95. If it really costs this much for a good English language course on Classical Nahuatl, it would be completely out of the question to publish grammars for the lesser known and less prestigious languages. The Summer Institute of Linguistics in Mexico is in the process of producing bilingual dictionaries and reference grammars for a number of Mexican Indian languages. The grammars are written in Spanish and are presented in the word-class model of description. The goal is to describe the Indian language in traditional school grammar terminology as much as possible, in order to make the descriptions accessible to the layman. There are, of course, linguistic differences between the Indian languages and Spanish, and in these cases, attention is given to contrast the two structures and to