Second Language Sentence Processing

Alan Juffs, Guillermo A. Rodríguez

• Publisher: Routledge
• Year: 2014
• ISBN:
  9781136217203
• Edition: 1
• Topic: Filología
• Subtopic: Lingüística

This addition to the Cognitive Science and Second Language Acquisition series presents a comprehensive review of the latest research findings on sentence processing in second language acquisition. The book begins with a broad overview of the core issues of second language sentence processing research and then narrows its focus by dedicating individual chapters to each of these key areas. While a number of publications have discussed research findings on knowledge of formal syntactic principles as part of theories of second language acquisition, there are fewer resources dedicated to the role of second language sentence processing in this context. This volume will act as the first full-length literature review of the field on the market.

1
INTRODUCTION

This book is a summary of research on second language sentence processing from the perspective of formal theories of morpho-syntax and how such theories might shed light on second language development (Chomsky, 1981, 1995; White, 2003). The book is intended to describe and summarize research that has already been published as well as some of our previously unpublished work. Within a formal framework, we address how adult users of a second language process language at the level of the sentence (usually during reading tasks), focusing on whether and how they deploy their knowledge of grammar to build a representation of the morpho-syntax of the clause being processed. Our goal is not to present a complete comparative overview of competing first and second language theories of sentence processing and acquisition. However, because the field finds itself so sharply divided, where appropriate we do refer to work in other frameworks that the interested reader should consult. We refer the reader to the excellent introduction to the edited volume by van Gompel (2013b) and papers in that volume for additional discussion on current issues in the field of sentence processing in general.

One might ask why researchers should investigate second-language processing at the fine-grained level of detail that we will present in this book. One fundamental reason is that such research is consistent with an approach to language learning that sees theories of language structure, language acquisition, and language processing as inextricably linked. It is an approach that is consistent with views of first language development that see processing breakdown as a trigger for acquisition (e.g., Fodor, 1998). White (1987), in her paper maintaining that comprehensible input alone was not adequate for language acquisition, has also suggested that processing breakdown is one component of a transition theory for second language acquisition. Moreover, processing is the basis for VanPatten’s (1996) approach to reconciling the competition between form and meaning in classroom instruction that can lead to successful learning.

In spite of this link between processing and acquisition, the book is not intended as a deep examination of how processing breakdown is the one key to understanding second language acquisition. Indeed, we do not consider that SLA can be explained by any one set of procedures; on the contrary, an understanding of SLA requires complementary theories of different scope as implied by Long (2007, p. 27). Instructed SLA is even less amenable to single, simple theoretical accounts because language itself is complex. For example, the development of pragmatic inference (Bardovi-Harlig & Hartford, 2005) will require a different set of conceptual tools for an explanation to the development of constraints in syllable structure. Moreover, language learning in classrooms depends on a wide variety of input types and output practice in different cultural contexts. Hence, this is not a book about how sentence failure ‘causes’ SLA, but rather a review of the issues that second language researchers who work in formal approaches to processing have worked on. The book is, however, a useful background for researchers who may wish to include processing failure in a theory of learning.

Second, a more profound understanding of second language sentence processing may afford insight into one of the three components of Sharwood-Smith’s and Truscott’s (2005) requirements for an overall theory of language acquisition: a theory of X, a theory of the development of X, and how X was produced or comprehended. This view provides the framework for discussing how and when processing can account for non-native performance and whether second language performance proceeds according to routines suggested by linguistic theories that have been proposed for first language performance. In other words, studying second language performance is a goal in itself, independent of its role in the performance process.

Ultimately, the knowledge from the study of second language processing may be applicable to pedagogical interventions such as input enhancement Sharwood-Smith (1986, 1993) and processing instruction (VanPatten, 2007), but the goal of this book is not to recommend such direct applications. It is up to scholars in pedagogy to adapt teaching to incorporate insights from processing indirectly, given the complexities of second language classrooms.

Many studies of language processing involve processing of written texts rather than spoken input. Reading sentences or texts is of course one step removed from language processing based purely on sound-structure. Moreover, anyone who has learned to read a deep orthography like English knows the challenge of mastering the imperfect relationship between phonemes and graphemes, and so text-decoding ability should factor more into our understanding of processing experiments. In spite of these decoding effects, processing performance in reading does allow researchers to infer how a learner’s grammar is used in real time, thereby providing insight into the learner’s grammar itself. Thus, the goal is to better understand the cognitive representation of second language processing and grammar. Finally, from behavioral measures such as reading experiments, it is possible to lay the groundwork for additional future work in the rapidly developing field of cognitive neuroscience of second language knowledge.

Research in second language sentence processing requires the integration of several components of language study. In SLA research, White (1989, 2003) among others has largely convinced the field that a theory of what is being acquired (a property theory) is necessary if we are to understand adult SLA. Moreover, Gregg (1989, 2003a) has made it clear that a theory of how acquisition is achieved (a transition theory) is also necessary. Thus, a fine-grained theory of the syntactic representation of their second language that the learners have, or seek to have, is necessary. Indeed, some of the major questions in this area concern precisely the nature of the syntactic representations learners use when they comprehend sentences in the second language in real time. Research has focused on whether second language grammars are as complex and detailed as the grammars of native speakers of a language and whether such abstract representations are used to build syntactic structure online so that comprehension can be successful.

The theory of the target of learning, the property theory, in much SLA research has been that of Chomsky’s Principles and Parameters theory (Chomsky, 1981; 1986). This approach to language proposes that knowledge of language consists of universal constraints, a set of abstract features that may be realized in different languages in an arbitrary set of morpho-syntactic or morpho-phonological ways (e.g., Case and Agreement), a universal interpretive component (Logical Form, LF), a phonological component (Phonological Form, PF), and a lexicon.

Principles and Parameters theory developed from earlier versions of transformational formalisms of syntax that had already introduced the idea that superficially long-distance dependencies between elements in a clause or clauses were in fact ‘covertly’ local. The formalisms developed in the course of this research were shown to generalize beyond specific structures to a range of phenomena including anaphoric reference, wh-movement, and quantification. (For a useful summary see a textbook treatment such as Carnie, 2003). Part of the formalism for this covert locality included a set of general constraints on the association between two or more non-adjacent positions in syntactic representations. Such constraints referred to the features associated with functional projections (e.g., WH features or tense features) and the (relative) position of nodes bearing such features in a non-symmetric configuration. Second language acquisition researchers have used these formalisms to investigate knowledge of a second language since the mid 1980s. Hence, the research discussed in this book is couched in the well-known formalisms of Principles and Parameters theory. Each chapter includes a review of the basic syntactic phenomena that form the basis of the processing experiments.

Second, processing research needs a theory of how incoming linguistic information—either from speech or a text—is related to the existing mental grammar so that successful comprehension can occur. This requirement bears repeating: every time we read or hear language, the brain must use a system to compare that input to the existing system in order to understand it. If that system is deficient or makes the wrong analysis, then comprehension cannot occur. This question can be considered from two points of view: (i) how is incoming text processing related to the second language (L2) grammar? (ii) What is the influence of the first language when processing a second language? These are the key questions related to processing. One might ask whether there are totally separate ‘processing principles’ for parsing that are totally independent of the grammar. Such principles might be analogous to ‘learning principles’ such as the Principle of Contrast for word learning (Clark, 1987) or the Subset Principle for structure (e.g., Wexler & Manzini, 1987).

A third, and of course related, reason that processing is important relates to its role in acquisition. Any theory of (first) language development has to account for the relationship between the linguistic input that learners encounter (first through listening and later through reading) and the development of syntactic representations. In formal theories of language acquisition, processing is considered to be important because failures in processing have been said to ‘trigger’ grammatical development (Fodor, 1998). The idea here is that if the existing grammar cannot parse the input and therefore assign a meaning to it, then a conflict is set up between the grammar and the input. Gibson and Wexler (1994) considered how the appropriate feature values provided by Universal Grammar (UG) are learned for specific languages. They propose that properties of the target language are accessed through parsing the input, and that learning occurs when the grammar fails to parse the input with the existing grammar. The problem that Gibson and Wexler address is that some surface word orders present data that is ambiguous as to the underlying syntactic structure. German provides an interesting example because of its variable word order in main and embedded clauses. A simple German sentence is SVO, e.g., die Katze sucht die Maus, ‘the cat looks for the mouse,’ with the finite lexical verb before the object. But when a modal is introduced, e.g., will = ‘want,’ the lexical verb is sentence final, die Katze will die Maus suchen, ‘the cat wants to look for the mouse.’ Hence, the learner must determine on the basis of this conflicting evidence, whether to set the headedness of the verb phase to V initial [V [NP]] or V final [[NP] V]. Additional evidence for the verb final nature of the verb phrase in German comes from sentences where the lexical verb is in sentence final position with compound tenses, e.g., die Katze hat die Maus gesucht, ‘the cat has looked for the mouse,’ and embedded clauses die Frau glaubt daß die Katze die Maus sucht, ‘the woman thinks that the cat is looking for the mouse.’ Hence, the data available to both a first and a second language learner show that the verb and the object may appear either to the left or the right of the lexical verb. Crucially, in main clauses, the finite lexical verb always precedes the object.

Fodor (1998, p. 13) also addresses this problem and suggests that acquisition consists, in part, in avoiding ambiguous triggers, and relying only on unambiguous triggers. Thus, for German, the learner would focus on the majority of clauses that are verb final and avoid using the simple clause ‘the cat looks for the mouse’ as a definitive indicator of the headedness of VP. The key points from Gibson and Wexler (1994) and Fodor (1998) here for our argument are that successful L1 learning consists of (1) parsing the input; (2) avoiding ambiguous evidence; (3) never giving up on a superficially successful grammar, but only one that fails to license the input; (4) only shift to the new grammar if it successfully parses the input. Presumably, there is a tension between 3 and 4. For example, children learning German as a first language start off with a verb-final VP structure. The V-final parser might fail with an SVO sentence, but no shift to the new grammar will occur unless it is successful with other clauses too. The conflicting data will be an indication that a deeper property of German is at play—namely, that finite verbs in German ‘move’ in a simple finite main clause for tense reasons. Indeed, German children only start using consistently correct VO word orders with subjects when they start using subject-verb agreement productively (Clahsen, 1990, p. 381). Hence, learning deeper properties of language from surface input turns out to be quite easy as long as you focus on key properties of the language being processed and ignore potentially misleading data. Hence, learning the structure of the German VP requires learners to ignore potentially misleading data until they discover the relationship between tense and the position of finite verbs. It seems that German children do exactly this.

A reviewer points out that parsing failure as an engine of acquisition is paradoxical—that is, “how can a learner make use of unparsable sentences if the learner cannot process them in the first place?” One possible answer to this is that the learner comes equipped with some knowledge concerning what is possible and not possible—the standard UG argument from the poverty of the stimulus perspective (White, 1989). White (1987) illustrates this point with an example from passive. If a learner who does not know passive in English encounters a sentence such as ‘the rabbits were eaten,’ the fact that ‘eat’ is optionally intransitive might allow the learner to skip the morphology, ‘the rabbits [*] eat[*]’ and assume that the sentence means ‘the rabbits ate something,’ comparable to the grammatical sentence, ‘the rabbits eat at dusk.’ In contrast, ‘the rabbits were killed,’ does not permit an ‘end-run’ around the grammar by ignoring morphology. The verb ‘kill’ is obligatorily transitive in English and so an argument is missing, which is a violation of the theta criterion. In this case, the learner would be compelled to reassess the input and seek an alternative to the parse, incorporating the morphology that allows a grammatical function change of the Theme from a syntactic object in the VP to the subject of a passive. Such failure driven parsing requires the consultation of fundamental principles in grammar (in this case the theta criterion and NP movement). It should be noted that opponents of generative theory are unwilling to accept such UG-based parsing accounts of acquisition, e.g., see Kidd (2004) and other reviews of Crain and Thornton (2000).

Based on these considerations, some questions for a transition theory of adult SLA would thus seem to be: (a) Are learners sensitive to structural ambiguity? (b) Do they use their L1 grammar to parse L2 input at the inception of L2 learning or do they start from scratch? (c) Can they successfully shift parameters by parsing input with a grammar that fails, and then switching to one that does the job? (d) Do learners not parse the input using grammatical principles at all, thus showing that their knowledge and processing of the L2 is quite different from L1 grammar and processing systems? The same questions arise in third language acquisition where the input may be deficient in some ways, cause processing breakdown that triggers development, or is misprocessed leading to incorrect grammar development (Corder, 1967; VanPatten, Dvorzak, & Lee, 1987; White, 1987). This concept is already familiar to L2 researchers in the distinction that is made between input and intake, but the focus here is on how input is processed in order to become intake in real time. Moreover, in instructed SLA, processing failures based on L1 influence and too much reliance on semantic cues have both been thought to be a cause of errors in comprehension (VanPatten, 2007).

In sum, then, understanding the nature of processing in second language acquisition has the potential lead to insights into developmental processes and to the establishment of better instructional treatments. As mentioned already, however, this volume is not intended to be an account of processing-driven acquisition but a review of existing research that explores the relationship between the grammar and the parser. If it can be established that adults can use abstract principles, then the existence of such principles in an overall theory of acquisition will have to be accounted for. Let us consider these three issues in more detail.

A Theory of Grammar

In a series of articles, Gregg (1989, 2003a) and White (1987, 1989, 2003) among others have argued that a theory of second language acquisition requires two elements. First, a theory of the target of acquisition—a property theory—and second, a theory of how that grammar is acquired—a transition theo...

Table of contents

• Cover
• Half Title
• Title Page
• Copyright Page
• Dedication
• Table of Contents
• List of Tables
• List of Figures
• Series Editor’s Preface
• Preface
• Acknowledgments
• 1 Introduction
• 2 Grammars and Parsers
• 3 The Influence of VERB Argument Structure on Processing
• 4 Processing Long-Distance Wh-Dependencies in a Second Language
• 5 Processing Relative Clauses in a Second Language
• 6 Processing of Noun Phrases and Agreement in Noun PhraseS in a Second Language
• 7 Processing Ambiguous Relative Clause Attachment
• 8 Related Research Fields and Conclusions
• References
• Index