LINGUIST List 6.1760

Mon Dec 18 1995

Review: Levin & Rappaport Hovav (1995)

Editor for this issue: T. Daniel Seely <dseelyemunix.emich.edu>

Directory

  • Daniel Seely, Review of Levin and Rappaport Hovav

    Message 1: Review of Levin and Rappaport Hovav

     Date: Mon, 18 Dec 1995 01:07:05 Review of Levin and Rappaport Hovav
    From: Daniel Seely <dseelyemunix.emich.edu>
    Subject: Review of Levin and Rappaport Hovav


    BOOK REVIEW

    Beth Levin and Malka Rappaport Hovav. Unaccusativity: At the Syntax-Lexical Semantics Interface. Linguistic Inquiry Monograph Twenty-Six. Cambridge, Massachusetts. London, England: The MIT Press, 1995. Pp. XIII + 336. ISBN 0-262-12185-9.

    Reviewed by Sebastian Shaumyan (Yale University) shaumyanminerva.cis.yale.edu

    The Unaccusative Hypothesis, first formulated by Perlmutter within the framework of Relational Grammar (RG) and later adopted by Burzio (1986) within the Government-Binding (GB) framework (Chomsky 1981), has been advanced to explain the puzzling syntactic behavior of intransitive predicates: it has been observed cross-linguistically that some intransitive predicates can never be passivized, while other intransitive predicates can. Under the Unaccusativity Hypothesis, these two observable classes of intransitive predicates represent two hypothetical classes of intransitive verbs: the unaccusative verbs and the unergative verbs, associated with different underlying syntactic configurations (an initial-stratum entity of RG or a deep structure entity of GB). In their underlying syntactic configurations, an unergative verb takes a subject but no object whereas an unaccusative verb takes an object but no subject. The Unaccusativity Hypothesis was introduced by Perlmutter in the context of the Universal Alignment Hypothesis, which suggests that the syntactic expression of arguments is always determinable on the basis of the meaning of the verb.

    It is interesting to note that the puzzling behavior of intransitive verbs was not perceived by linguists as a serious problem before the Unaccusativity Hypothesis was advanced. Only after this it became clear that the puzzling behavior of intransitive verbs is one of the most intriguing and fundamental facts of syntactic typology. The Unaccusativity Hypothesis has given impetus to an active research to test this hypothesis and its consequences. Actually, the split intransitivity, as I call this phenomenon, is a challenge for any linguistic theory. This is why there is already a large body of research produced by linguists that work not only in RG but in other linguistic theories as well. Since the Unaccusativity Hypothesis was introduced, a wide range of phenomena in various languages have been studied that concern a distinction between unaccusative and unergative verbs. The book under review is an important contribution towards the common goal that is being pursued presently by linguists espousing different theoretical views--the goal of explaining the split intransitivity.

    Beth Levin and Malka Rappaport Hovav address a fundamental problem that concerns the Unaccusative Hypothesis and is implied by the Universal Alignment Hypothesis: Is the syntactic distinction between the two classes of verbs--unaccusative and unergative--determined semantically? The tentative positive answer to this question has been suggested by the Universal Alignment Hypothesis. But it has been discovered that the relation between the syntactic behavior of verbs and their meaning is more complicated than it has been initially supposed. In reality, there exist phenomena (so-called unaccusative mismatches) that suggest that the classification of verbs as unaccusative or unergative cannot be completely determined semantically. In essence, this book is an attempt to meet the challenges that the mismatches present to the original Perlmutter's hypothesis that unaccusativity is both syntactically encoded and semantically predictable.

    The authors examine alternative approaches to the unaccusativity problem--the syntactic approach (represented by C.Rosen) and the semantic approach (represented by Van Valin). Rosen denies that the unaccusativity is semantically predictable, while Van Valin claims that the two classes of verbs can be differentiated on semantic grounds alone and that therefore there is no need to attribute to these different classes of verbs different syntactic representations.

    To support their central thesis that unaccusativity is both syntactically encoded and semantically predictable, the authors, in chapter 2 of the book, provide an extensive study of the unaccusative diagnostic, the English resultative construction, which is meant to give evidence for the syntactic encoding of unaccusativity. In chapters 3 and 4, the authors look closely at the lexical-semantics--syntax interface as it pertains to unaccusativity to provide evidence for semantic factors that determine the unaccusativity. In chapter 4, a set of linking rules is presented that classify the various semantic classes, examined by the authors, into unaccusative and unergative.

    Building on the results of chapter 4, the authors turn in chapter 5 to the variable behavior of verbs, which presents a problem for the semantic determination of unaccusativity. There is a set of verbs that show characteristics of both unaccusative and unergative verbs. The authors propose to solve this problem by stating that in most instances such verbs have two distinct meanings, one associated with an unaccusative analysis and the other with an unergative analysis and with the corresponding two distinct kinds of the syntactic behavior.

    In chapter 6, the authors turn to locative inversion. Some linguists, Bresnan and Kanerva, Hoekstra and Mulder, among others, claim that locative inversion is a valid unaccusative diagnostic. But the authors present arguments against this claim.

    The book has two appendices: 1) a list of major classes of intransitive verbs discussed in the book; 2) verbs found in the locative inversion constructions.

    Limitations of space preclude me from a detailed discussion of the book, rich in facts and ideas. I will only discuss very briefly some points which, I think, have a great methodological interest and significance.

    First, the authors have presented forceful arguments to substantiate their thesis that unaccusativity is semantically determined and syntactically encoded. Although this thesis is not new (it is implied by the Universal Alignment Hypothesis of Perlmutter), it was extremely difficult to test this thesis in detail by an analysis of conflicting properties of concrete words. The authors must be credited with having undertaken very complex research in the course of which they have unearthed many important facts leading to new insights. This work has methodological significance. We must consider a wider hypothesis variously formulated by many linguists that syntactic relations map semantic relations. A broad linguistic research testing this general hypothesis is very important. It seems that this hypothesis has its limits. Not every syntactic construction is determined semantically from a synchronic point of view. For example, why the Russian verb which stands for "thank" governs the accusative case but the German verb with the same meaning governs the dative case? By studying the history of these languages we may find the different semantic motivations for these different syntactic constructions, but today these constructions are not determined semantically; they are purely conventional. Similarly, it is highly improbable that every instance of unaccusativity is determined semantically. Some instances may be purely conventional, although they were determined semantically in the past. As a general principle we may state that syntactic constructions are determined partly by the lexical meaning of words and partly by convention. This, of course, does not undermine the significance of what the authors have done. On the contrary, it is very important that building upon the hypothesis that syntactic encoding is determined semantically the authors have pushed this hypothesis to the limit. What remains to be done is to find out whether certain instances of unaccusativity are motivated by syntactic convention rather than by the meaning of the intransitive verbs.

    A few words about the syntactic approach to the split intransitivity. Relational Grammar is an important linguistic theory. This theory has generated a considerable body of research which demonstrates that to develop a rich, flexible, homogeneous, fruitful theory of syntax as the system of grammatical relations, we must strictly distinguish between syntax and the lexicon, we must not confound syntax as the system of grammatical relations with the lexicon by smuggling lexical concepts such as thematic role descriptions into syntactic theory. I can understand the intransigence of Rosen who in defense of the homogeneous theory of syntax has overstated her case. An error stated clearly is better than a half truth. A clear statement, although erroneous, may be rectified; it leads us to the discovery of the truth. A half truth generates confusion and leads nowhere. Building upon a clear distinction between the domains of syntax and the lexicon, we can turn to the syntax-lexicon interface, which is a separate, a distinct domain of research. This is what the authors of the book have done. If the syntactic approach to the split intransitivity means the denial of the concept of the syntax-lexicon interface or the importance of its study, then it it is wrong; but if it means only that analyzing the problem within the framework of syntactic theory we must build on homogeneous syntactic concepts, then it is correct and has a fundamental importance.

    Let me turn to another point. On page 120 of the book, the authors, following the studies of Lyons, claim that verbs of existence are dyadic. The argumentation in support of this claim is convincing. This is a correct claim. But if we recognize this claim we cannot consider verbs of existence to be unaccusative verbs because unaccusative verbs are monadic by definition. To solve this difficulty, the authors propose that verbs of existence take two internal arguments rather than one external and one internal. In chapter 4, they present detailed arguments in support of their proposal.

    Granted that we agree with the authors' proposal that verbs of existence have two internal arguments, this does not solve our difficulties, because unaccusative verbs are intransitive verbs, and intransitive verbs are monadic as recognized by RG. Unless we arbitrarily stretch the concept of intransitive verbs, we cannot consider intransitive verbs to be dyadic. Let us distinguish clearly between empirical facts and hypotheses advanced to explain them. The fact, I call the split intransitivity, which needs an explanation is this: It has been observed cross-linguistically that some intransitive predicates can never be passivized, while other intransitive predicates can. The Unaccusative Hypothesis has been advanced to explain this fact. How does this hypothesis do this? By postulating two underlying classes of intransitive verbs: unergative and unaccusative. In their underlying syntactic configurations, an unergative verb takes a subject but no object whereas an unaccusative verb takes an object but no subject. In accepting the Unaccusative Hypothesis, we face a serious difficulty. The trouble is that direct object is, as recognized by many linguists, a syntactically marked term, while subject is a syntactically unmarked term. This means that in a sentence, object cannot occur without subject, while subject can occur without object. From all this it follows that, unless we stretch concepts arbitrarily or misinterpret empirical facts, we must recognize that objects can be combined only with transitive verbs, while subjects can occur both with intransitive and intransitive verbs.

    If we agree with the authors that verbs of existence are dyadic and if we do not stretch the concept of intransitive verbs arbitrarily, we must recognize that the correct interpretation of verbs of existence as dyadic provides a strong empirical argument against the concept of unaccusative verbs rather than in support of it.

    The above difficulties present a serious challenge to the Unaccusative Hypothesis. In view of difficulties of this kind and, in the first place, in view of the fundamental conceptual difficulties, discussed in my book A SEMIOTIC THEORY OF LANGUAGE, it is natural to look for alternative, more plausible hypotheses. In my book, a hypothesis, called the Syntactic Neutralization Hypothesis, is proposed. Under this hypothesis, subject is a syntactically unmarked term of the opposition SUBJECT:OBJECT, while object is a syntactically marked term. The syntactic context does not affect the function of object, but it affects subject. Object remains object in any syntactic context, but subject changes its function depending on the syntactic context. Thus, in the context of active constructions subject is subject proper but in the context of passive constructions subject has the meaning of object. The behavior of the members of the syntactic opposition SUBJECT:OBJECT has counterparts in phonology and the lexicon. Consider the opposition of consonants VOICELESS:VOICED in Russian where voiceless consonants are unmarked and voiced consonants are marked members of this opposition. At the end of a word only a voiceless consonant can occur, but it can function either as voiceless proper or as a replacement of a voiced consonant. Similarly, in the lexicon. Consider the opposition LION:LIONESS, where LION is the unmarked term of this opposition, and LIONESS is the marked term. As an unmarked term, LION is ambiguous, it may mean a male or a female, but LIONESS always means only female. Compare "I see a lion and a lioness" with "I see a lion". In the second sentence outside of the opposition with LIONESS, LION is ambiguous, it may refer either to a male or female.

    The Syntactic Neutralization Hypothesis means that the intransitive verbs can combine only with subjects which as unmarked terms of the opposition SUBJECT:OBJECT have a generic function, that is, depending on the context, a subject may function as subject proper or as object. Hence dual syntactic behavior of intransitive verbs: an intransitive predicate may be passivized when its subject functions as object, and it cannot be passivized when its subject does not function as object. The important thing to note is that the subject of an intransitive construction remains a subject no matter whether it functions as subject proper or as object. To explain the split intransitivity or any linguistic phenomenon for that matter, theoretical linguistics does not need to resort to obsolete concepts such as deep structure or its counterparts.

    Limitations of space do not allow further discussion of the rich facts and ideas presented in this book. To conclude, I must say that the book has a methodological significance, in the first place. The authors examined both a syntactic and a semantic approach to the problem of the split intransitivity in the domain of the syntax-lexicon interface. They presented convincing arguments against these approaches, showing their inadequacy. The authors advocate the inseparability of syntactic and semantic analysis in the study of the syntax-lexicon interface. They succeeded in demonstrating the validity of this approach. The book gives a lot and deserves a careful study.

    Sebastian Shaumyan Professor of Linguistics Emeritus Yale University