Crosslinguistic Perspectives on Argument Structure

Implications for Learnability

Melissa Bowerman, Penelope Brown

• Publisher: Routledge
• Year: 2007
• Print ISBN: 9780805841947
• eBook ISBN: 9781136792540
• Topic: Languages & Linguistics
• Subtopic: Linguistics
• Index: Languages & Linguistics

This book offers a unique interdisciplinary perspective on argument structure and its role in language acquisition. Drawing on a broad range of crosslinguistic data, this volume shows that languages are much more diverse in their argument structure properties than has been realized.The volume is the outcome of an integrated research project and com

CHAPTER 1

Introduction

Melissa Bowerman and Penelope Brown Max Planck Institute for Psycholinguistics

Verbs are the glue that holds clauses together. As elements that encode events, verbs are associated with a core set of semantic participants that take part in the event. Some of a verb's semantic participants, although not necessarily all, are mapped to roles that are syntactically relevant in the clause, such as subject or direct object; these are the arguments of the verb. For example, in John kicked the ball, ‘John’ and ‘the ball’ are semantic participants of the verb kick, and they are also its core syntactic arguments—the subject and the direct object, respectively. Another semantic participant, ‘foot’, is also understood, but it is not an argument; rather, it is incorporated directly into the meaning of the verb. The array of participants associated with verbs and other predicates, and how these participants are mapped to syntax, are the focus of the study of ARGUMENT STRUCTURE.

At one time, argument structure meant little more than the number of arguments appearing with a verb, for example, one for an intransitive verb, two for a transitive verb. But argument structure has by now taken on a central theoretical position in the study of both language structure and language development. In linguistics, argument structure is seen as a critical interface between the lexical semantic properties of verbs and the morphosyntactic properties of the clauses in which they appear (e.g., Grimshaw, 1990; Goldberg, 1995; Hale & Keyser, 1993; Levin & Rappaport Hovav, 1995; Jackendoff, 1990). Increasingly, this interface is understood as being mediated by a rich representation of event structure based on causal dynamics, the internal temporal structure of events (aspect), or both (e.g., Croft, 1991, 1998, in press; Dowty, 1979; Erteschik-Shir & Rapoport, 2005; Pustejovky, 1991; Tenny, 1994; Tenny & Pustejovsky, 2000; Van Valin & LaPolla, 1997).

Although theorists differ in how they represent argument-structure-related properties of language, they often agree on the view that, across languages, there are strong consistencies in the number of arguments associated with verbs with certain kinds of meanings, and in the typical mapping of these arguments to syntactic roles (e.g., Keenan, 1976; Perlmutter & Rosen, 1984; see Pinker, 1989: 94–95 for discussion). Strong similarities across languages—universals or near-universals—demand explanation, and it has been a familiar step, since Chomsky's work of the 1960s and 1970s, to locate this explanation in the child's innate capacity for language acquisition. Universals, according to this way of thinking, reflect children's a priori expectations about the structure of language. Children come to the acquisition task with inborn knowledge of those abstract aspects of grammar that are universal, and this knowledge enables them to home in quickly on how these universals are instantiated in their particular language. Individual languages, for their part, are constrained to conform to the universal architecture, because if they did not, children would be unable to learn them.

Given this hypothesized link between linguistic universals and the capacity for language acquisition, proposals about universals of argument structure have caught the attention of language acquisition researchers. If children have innate expectations about argument structure—in particular, about the typical correspondences between the semantic and syntactic roles associated with verbs—they could draw on this information to solve a number of learning problems. This thought lies at the heart of several influential hypotheses, to be reviewed shortly, about how children set up their initial phrase structures, acquire the meanings of novel verbs, and figure out which verbs can occur in which syntactic frames.

Despite the emphasis on universals, most of the research on the proposed role of innate argument structure knowledge in language acquisition has revolved around English and closely related languages. It is not yet clear whether children could, in fact, use the hypothesized universals of argument structure to acquire the structures that confront them in a broad range of typologically distinct languages. The present volume attempts to address this concern. The research it presents came about as the result of a unique situation: a cooperation, within the framework of the Argument Structure Project at the Max Planck Institute for Psycholinguistics, between crosslinguistically minded language acquisition specialists and field linguists working on a diverse set of lesser-known languages. Working together over a period of several years, our group of linguists, psychologists, and anthropologists explored claims about argument structure universals, and used findings to jointly analyze and rethink the possible role of argument structure in children's language development. The work often challenges current proposals, especially in showing that there is more crosslinguistic diversity in the domain of argument structure than most specialists in language acquisition have realized. But it also suggests new directions for finding solutions, and it calls attention to argument structure acquisition puzzles that have so far been neglected.

In this chapter, we introduce some of the proposals, controversies, and problems that have inspired and motivated the authors of this volume, and we summarize the organization and contents of the book. Section 1 here overviews the so-called “bootstrapping” hypotheses, according to which innate knowledge of argument structure plays a critical role in language acquisition. Section 2 further sets the stage by reviewing some current major controversies surrounding these hypotheses. Finally, section 3 lays out the plan of the book, and highlights the key findings of the specific chapters.

1. THE PROPOSED ROLE OF ARGUMENT STRUCTURE UNIVERSALS IN LANGUAGE ACQUISITION

What are the language acquisition challenges that children might solve by drawing on inborn knowledge of argument structure? There have been two broad lines of theorizing. Both presuppose that children come to the language acquisition task with expectations about the linking between the semantic and syntactic roles associated with verbs, but they differ in their assumptions about the typical direction of learners’ inferences.

One scenario goes from meaning to syntax. In this approach, children are hypothesized to use their nonlinguistic cognitive understanding of the world to determine the structure of everyday events like ‘running’, ‘hitting’, and ‘giving’, including how many participants there are in the event and what their semantic function, or thematic role, is (e.g., agent, patient, recipient). Children then draw on their innate knowledge to predict how noun phrases representing these participants should be mapped to syntactic roles.

The second scenario reverses the direction of inferencing, going from syntax to meaning rather than meaning to syntax. In particular, this proposal is about how children learn the meaning of verbs. On encountering a new verb, children are hypothesized to use the syntactic frame(s) in which it appears to predict some basic aspects of its meaning. Proposals of the first type are lumped loosely under the rubric “semantic bootstrapping,” whereas those of the second type are termed “syntactic bootstrapping.”

1.1. Semantic Bootstrapping: Using Meaning to Predict Syntax

Cracking into Grammar. Following Chomsky's claim that universals of language are innate, many child language researchers have assumed that children have inborn knowledge of putatively universal syntactic categories and relationships such as “noun”, “verb”, “subject”, and “direct object”, and—importantly—of the abstract syntactic properties associated with these constructs. But this would buy children nothing unless they had some way to identify concrete instances of these constructs in the speech around them. To explain how this identification takes place, Pinker (1984, building on Grimshaw, 1981) proposed that children's inborn linguistic toolkit includes not only information about syntactic categories and relationships, but also some cognitively simple “semantic flags” by which these elements can be recognized. He termed this use of semantics to make good guesses about syntax “semantic bootstrapping.”

According to this hypothesis, for example, children will initially assume that if a word names a concrete object it is a noun, and if it names an action it is a verb. Of course, not all nouns pick out objects and not all verbs name actions. But this procedure could help children identify a starter set of nouns and verbs correctly; from here they can go on to notice the morphosyntactic elements associated with these forms, and so to identify instances of the categories that lack the default semantics. For instance, learners of English will notice that words that name objects—hence, nouns—typically occur in contexts like a/the/another/my____, or two____ and ____-s, and words that name actions—hence, verbs—occur in contexts like ____-ed and is / are____-ing. This morphosyntactic knowledge will enable children to identify bath as a noun and stay as a verb even though these words do not name an object or a dynamic action, respectively.

Children have to figure out not only the part-of-speech membership of the words in the sentences they hear, but also their syntactic function. According to the semantic bootstrapping hypothesis, “semantic flags” help with this problem, too (Pinker, 1984). For example, suppose that the syntactic constructs “subject” and “direct object”—by hypothesis inborn—are initially cued by the relational roles “agent” and “patient”. When a learner of English sees a dog biting a cat, her nonlinguistic understanding of the event tells who does what to whom. If she now hears the sentence The dog is biting the cat, she can infer that the dog (naming the agent) is the subject and the cat (naming the patient) is the direct object. Once she has identified a few subjects and direct objects like this, she will discover how these constituents are typically ordered with respect to the verb, whether and how they are case-marked, and so on. And this knowledge will now allow her to identify other constituents as subjects and direct objects even when they lack the default agent–patient semantics, as in Mary heard a noise or The costume frightened the dog.

With semantic bootstrapping, then, the child gets a toe in the door of grammar by an initial simple mapping between words and basic semantic types.

Constraining Argument Structure Overgeneralizations. The basic logic of semantic bootstrapping has been applied not only to the initial stages of grammar construction, but also—in a more elaborate form—to a thorny problem arising later in the course of language development: how children avoid overgeneralizations of argument structure alternations.

Many verbs of English and other languages appear in more than one syntactic frame, and whole groups of semantically related verbs often show similar patterns of frame alternations. Children become sensitive to these alternation patterns in the course of language acquisition, and sometimes apply them too liberally. Two kinds of alternations, and a few errors based on them, are shown in (1)–(2) (from Bowerman, 1982a, 1982b; see also Pinker, 1989).

1. Causative Alternation (cf. The stick broke/Harry broke the stick)

a. You staggered me. (After mother pulls on child's arm when child stumbles. 3;10 [age in years; months])

b. I saw a witch and she disappeared them. (Pretending a witch has made some blankets disappear. 4;8)

2. Locative Alternation (cf. Mary sprayed paint on the wall/Mary sprayed the wall with paint)

a. Can I fill some salt into the bear? (= fill the bear [a bear-shaped salt shaker] with salt. 5;0)

b. Pour, pour, pour. Mommy, I poured you (waving empty container near M. M: You poured me?) Yeah, with water. (= I poured water on you. 2;11)

It is widely acknowledged that children are rarely corrected for such errors. Why then do they stop making them—how do they end up with adult-like intuitions about which verbs do and don't undergo a certain alternation? This question has given rise to much debate (e.g., Bowerman, 1988; Braine & Brooks, 1995; Pinker, 1984, 1989; Randall, 1990).

According to one line of reasoning, the adult state is reached by gradual learning. Children discover an alternation pattern by a process of abstraction and schema formation after exposure to a sufficient number of exemplars of it, and they overgeneralize it because they do not yet know the relevant semantic, morphophonological, or idiosyncratic constraints. With increasing linguistic experience, they fine-tune the pattern to reflect these constraints, and errors fade out (Bowerman, 1982a, 1988; Brooks & Tomasello, 1999; Goldberg, 1995; Goldberg, Casenhiser, & Sethuraman, 2004).

According to a very different approach, children get a tremendous boost toward the adult system through aspects of their innate knowledge (see Pinker, 1989, for the theory, and Gropen, Pinker, Hollander, & Goldberg, 1991a, 1991b, for some applications).

In this approach Pinker updates his (1984) view of how verb meaning is related to argument structure. Rather than relying on simple semantics–syntax correspondences like “if agent, then subject” (as described under Cracking into Grammar, earlier), which requires reference to a fixed list of atomic thematic roles (e.g., agent, theme, location, source, goal…), with each noun phrase in a clause assigned to just one role, he adopts the decompositional approach to verb meaning found in, for example, Jackendoff (1987, 1990) and Levin and Rappaport Hovav (1995). Here, thematic roles are positions in a semantic representation of verb meaning that is structured around a set of primitive meaning elements such as “cause,” “go,” and “be,” with each thematic role triggering its own linking rule, for example, “first argument of ‘cause’ is subject,” “second argument of ‘cause’ [the ‘affected object’] is direct object.” A particular argument can bear more than one thematic role because it can participate in more than one semantic substructure in the verb's semantic representation; for example, it can be both the second argument of “cause” and the first argument of “go.”

As an example, consider the English verbs fill and pour. Both verbs specify the causation of a change of some sort, but they differ in which argument is specified to undergo the change (i.e., is the “affected object”): pour means roughly “cause X (a liquid or particulate mass) to go downward in a stream,” whereas fill means something like “cause Y (a container) to go to a state of being full by means of causing X to go...

Table of contents

• Front Cover
• Half Title
• Title Page
• Copyright
• Contents
• Preface
• 1 Introduction
• I. Verb Meaning and Verb Syntax: Crosslinguistic Puzzles for Language Learners
• II. Participants Present and Absent: Argument Ellipsis and Verb Learning
• III. Transitivity, Intransitivity, and their Associated Meanings: A Complex Work-Space for Learnability
• Author Index
• Language Index
• Subject Index