Deconstructing the English Passive 9783110199215, 9783110196320

Table of contents :
Frontmatter
Contents
Chapter 1. Introduction
Chapter 2. The English passive and linguistic theory
Chapter 3. Ingredients of the English passive
Chapter 4. The implicit agent in English passives
Chapter 5. The use of the passive in academic discourse: A case study
Chapter 6. Conclusion
Backmatter

Citation preview

Deconstructing the English Passive

≥

Topics in English Linguistics 41

Editors

Bernd Kortmann Elizabeth Closs Traugott

Mouton de Gruyter Berlin · New York

Deconstructing the English Passive

by

Anja Wanner

Mouton de Gruyter Berlin · New York

Mouton de Gruyter (formerly Mouton, The Hague) is a Division of Walter de Gruyter GmbH & Co. KG, Berlin.

앝 Printed on acid-free paper which falls within the guidelines 앪 of the ANSI to ensure permanence and durability.

Library of Congress Cataloging-in-Publication Data Wanner, Anja. Deconstructing the English passive / by Anja Wanner. p. cm. ⫺ (Topics in English linguistics ; 41) Includes bibliographical references and index. ISBN 978-3-11-019632-0 (alk. paper) 1. English language ⫺ Passive voice. 2. Grammar, Comparative and general ⫺ Voice. 3. Grammar, Comparative and general ⫺ Verb phrase. 4. English language ⫺ Discourse analysis. I. Title. PE1285.W36 2009 425⫺dc22 2009020627

ISBN 978-3-11-019632-0 ISSN 1434-3452 Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” Copyright 2009 by Walter de Gruyter GmbH & Co. KG, 10785 Berlin All rights reserved, including those of translation into foreign languages. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Cover design: Christopher Schneider, Laufen. Typesetting: PTP-Berlin Protago-TEX-Production GmbH, Berlin. Printed in Germany.

Acknowledgments This book has been long in the making. When I joined the University of Wisconsin-Madison in 2000, with a background in formal grammar and the mission to teach syntax courses in an applied English linguistics program, it was clear to me that I wanted to show that insights from generative grammar can provide answers to questions associated with language use. Less clear, however, was how exactly I would go about making that argument. It was during the summer institute of the Linguistic Society of America at the University of California, Santa Barbara, in particular through conversations and classes with Adele Goldberg, Howard Lasnik, and Masayoshi Shibatani, who all aproach transitivity alternations from very different viewpoints, that I began to develop the perspective for this book. The University of Wisconsin-Madison has funded this project generously. My thanks go to the English Department, the Graduate School of the College of Letters and Science, and the Vilas Foundation for providing support through research leaves, travel grants, and project assistantships. I thank my colleagues in the English Department as well as anonymous external reviewers, who read work-in-progress versions of this book as part of the tenure process. I’ll never forget the day when I was welcomed home with a sign that read “unanimous.” Specifically, I thank Joan Houston Hall, chief editor of the Dictionary of American Regional English (DARE), for illuminating sociolinguistic aspects of the get-passive and for making the transcripts of the DARE questionnaires available to me. My thanks also go to Brad Hughes, Director of the Writing Center at the University of Wisconsin-Madison, for sharing his expertise in teaching college-level writing in the United States. And let not be unmentioned the colleague in literary studies whose assertion that “the passive is not allowed in my classes” inspired parts of the chapter on the passive in academic writing. That chapter also could not have been written without the insightfulness and generosity of Heidrun Dorgeloh (University of Dusseldorf), ¨ with whom I have collaborated on the subject of academic writing for a number of years. We are both greatly indebted to Christian Mair (University of Freiburg), who gave us access to the ARCHER corpus and to a tagged version of the FROWN corpus before it became publicly available. Katie Gilbert’s superb editorial assistance smoothed out many rough stylistic edges. I very warmly thank the editors of the series Topics in English Linguistics, Elizabeth Traugott and Bernd Kortmann, as well as my editor at Mou-

vi

Acknowledgments

ton de Gruyter, Birgit Sievert, for their substantial input and support during the various stages of preparing the manuscript. Above all, I am deeply grateful to my family, Ralph Grunewald and our daughter Claire, who had to live on opposite sides of the Atlantic while this book was being finished. We are now all back together in Madison, and thanks to you, this is the best time of my life. Clairchen, no baby should have to collect as many frequent flier miles as you did in your first year. This one’s for you.

Contents Acknowledgments

v

Chapter 1 Introduction

1

Chapter 2 The English passive and linguistic theory 1. Introduction 2. The status of the passive in linguistic theory 3. What is a passive? 3.1. Towards a morpheme-based definition 3.2. Passive as an instantiation of “voice” 3.2.1. Definitions of voice 3.2.2. Voice and transitivity 4. Functions of the passive – a preliminary sketch 4.1. Demotion of the agent 4.2. Promotion of the theme 5. The passive in language acquisition theory 6. Summary

7 7 7 12 14 29 29 35 37 38 40 42 46

Chapter 3 Ingredients of the English passive 1. Introduction 2. Verb classes and passivizablity 2.1. Transitive verbs and NP-Movement 2.2. Ditransitive verbs and recipient passives 2.3. Intransitive verbs and impersonal passives 2.3.1. It’s all about “be” 2.3.2. It’s all about Case 2.3.3. It’s all about “it” 2.4. Prepositional verbs and prepositional passives 3. The passive participle 3.1. On the relationship between the participle and the adjective 3.2. Passive meaning without passive participle 4. The by-phrase 5. The subject 6. The auxiliary verb

49 49 52 52 56 58 59 60 62 66 70 72 75 76 80 83

viii

Contents

A re-evaluation of the get-passive Locating the get-passive in the spectrum of get-constructions 7.2. Ergative and causative get 7.3. Verbal versus adjectival get-passives 7.4. Get-passives in the FROWN corpus 7.4.1. Ergative get-passives 7.4.2. Causative get-passives 7.4.3. Adjectival get-passives 7.5. Revisiting the responsibility factor 7.6. Revisiting the adversity effect 8. Summary 7. 7.1.

Chapter 4 The implicit agent in English passives 1. Introduction 2. Evidence from syntax 2.1. Purpose clauses and control 2.2. By-phrases 2.3. Anaphors and binding 3. Linking theory considerations 4. Evidence from psycholinguistics 5. Evidence from language use 5.1. Sentences that require an implicit agent 5.2. Unaccusative passives 5.2.1. Unaccusative passives used by native speakers 5.2.2. Unaccusative passives in second language acquisition 6. Representation of the implicit argument 6.1. Syntactic approaches 6.1.1. The absorption analysis 6.1.2. The pro analysis 6.2. Lexical approaches 6.2.1. Argument suppression at predicate argument structure (PAS) 6.2.2. Lexical binding at lexical-conceptual structure (LCS) 7. Summary

85 88 91 95 99 99 101 102 103 107 109 113 113 114 115 120 122 124 127 130 131 133 133 138 140 141 141 144 146 148 150 152

Contents

ix

Chapter 5 The use of the passive in academic discourse: A case study 1. Introduction 2. Characterization of the abstract as a genre 3. The tradition of the passive in scholarly writing 4. Synchronic case study 4.1. Corpus Information 4.2. Distribution of the passive: Preliminary findings 4.3. The distinction between reporting and reported events 4.4. The use of the passive in reporting events 4.4.1. The role of reporting: Differences across the disciplines 4.4.2. The passive as a reporting strategy 4.4.3. The representation of agenthood in reporting events 4.4.4. Reporting events and politeness considerations 4.5. The use of the passive in reported events 5. Support from a diachronic study 6. Summary

155 155 156 160 165 165 166 172 175 176 178 180 183 186 189 191

Chapter 6 Conclusion

197

Appendix: Exemplary abstracts References Sources of examples Subject and author index

203 207 225 227

Chapter 1 Introduction The passive is a construction that is right at the intersection of different approaches to the study of language: It addresses the issues of how semantic information is mapped onto syntax, how morphology affects syntactic structure, and how speakers employ syntactic options to organize information within a clause. As a construction that is found in many languages of the world and that comes with specific affixes in the classical languages, it firmly holds its place in traditional grammars of English as well as in comparative studies. As a sentence pattern that motivates the use of transformational rules, it has been one of the main subjects of inquiry in early generative grammar. As a stylistic device that rearranges the major constituents in a sentence, it is of great interest to functionalists and discourse analysts, who examine how speakers make use of the options that grammar gives them and how a construction is shaped by its communicative function. Also, no style manual would be complete without giving advice about the (non-)appropriateness of the passive, and the passive construction “Mistakes were made” is probably one of the most recognizable (and most ridiculed) sentences in political commentary.1 The syntax of the passive in English has been examined from many different angles. Every reference grammar has a section on the passive (e. g., Jespersen 1927; Quirk, Greenbaum, Leech and Svartvik 1985; Huddleston and Pullum 2002), often with an emphasis on the relationship between the syntactic class of the underlying verb and the grammaticality of passivization. There are specialized studies on impersonal passives (Perlmutter 1978), prepositional passives (Couper-Kuhlen 1979), the get-passive (Haegeman 1985; Herold 1986; Gívon and Yang 1994; Collins 1996), passives in first language acquisition (Maratsos, Fox, Becker and Chalkley 1985; Pinker, Lebeaux and Frost 1987; Budwig 1990; Verrips 2000; Meints 2003), passives in second language acquisition (Zobl 1989; Ju 2000; Oshita 2000), passives and Case Theory (Jaeggli 1986; Baker, Johnson and Roberts 1989; Goodall 1993), passives as an instantiation of voice (Svartvik 1966; Shibatani 1985a; Comrie 1988), and comparative aspects of the passive (Azvedo 1980; Åfarli 1989; Cornelis 1996; Xiao, McEnery and Qian 2006). 1. It even has its own Wikipedia entry: http://en.wikipedia.org/wiki/Mistakes were made (last visited on July 29, 2008).

2

Chapter 1 Introduction

The functions of the passive have also been discussed extensively. The passive may be used because the agent of an action is not known or should not be mentioned. There are passives whose main function is to allow for the expression of a non-agentive sentence topic, passives that are employed to create cohesion in a text, and passives that are chosen for politeness reasons, to name just a few of those functions (Stein 1979; Thompson 1987; Biber, Johansson, Leech, Conrad and Finegan 1999). Without a doubt, the English passive is a well-known construction with a long history of analysis in almost any linguistic framework. What could possibly be added to an already substantial body of research? This study sets out to bridge the gap between formalist2 and functionalist approaches in the spirit of Newmeyer (1998), who argues that the grammar of a language can have an autonomous syntax and can yet be motivated by grammar-external factors, such as the communicative function of language: “There is nothing in the program of external explanation of typological facts that is incompatible with the existence of an autonomous structural system. And there is nothing in the generative program that demands that all typological facts be attributed to the setting of innately specified parameters” (Newmeyer 1998: 364). Generative linguists are concerned with making the grammar that underlies the competence of the native speaker visible (Chomsky 1986). The maximal unit of analysis is the sentence – there are no principles of Universal Grammar that go beyond the sentence boundary. Why a language affords something like a passive is not something in which most generative linguists are particularly interested, nor is the question of how it is used and which constructions it competes with in a specific discourse environment. This is not to say that generative grammar does not recognize the relevance of pragmatics and its effect on word order and grammaticality. For example, it is now textbook material to integrate Focus Phrases and Topic Phrases into the X-bar format (Zubizaretta 1998; Haegeman and Gu´eron 1999). However, there is still a focus on the sentence as the unit of linguistic research. In discourse-based functionalist approaches, on the other hand, the point of studying the structure of sentences is to learn about the ways speakers employ language to convey meaning in interaction. While both approaches are not compatible on the level of technical analysis (generative grammar postulates abstract levels of representation, movement operations, 2. Newmeyer (1998: 8) points out that the term “formal” is ambiguous – it can mean “‘pertaining to (grammatical) form,’ as opposed to meanings and uses” or “stated in a mathematically precise vocabulary.” To avoid this ambiguity I will follow him in referring to formalist approaches as “generative grammar.”

Chapter 1 Introduction

3

and invisible elements like traces or empty categories), their insights are not incompatible, and this book attempts to make them visible to each other. In this study, I will not attempt a full-scale analysis of the form and function of the passive in English. Rather, I will focus on a specific element that is an integral component in all passives: the implicit argument (most often an implicit agent). Generative grammar, with its repertoire of non-overt categories, provides the tools to detect the implicit argument and to categorize it as a structural component of the passive. It will be shown that the implicit argument is not just a conceptually evoked event participant, rather, it is built into the structure – and thus the interpretation – of the passive construction. Building on this theoretical foundation, I will then look at uses of the passive and semantically related constructions, particularly with regard to the expression of agenthood. Throughout this monograph, I will assume that there is something like a formal core of the passive, but it cannot be described in terms of a specific syntactic configuration (like NP-be-Ven-by NP) because, depending on the syntactic features of the passivized verb, there are many different surface forms the passive in English can take. This means that I will not use the term construction in the holistic sense of Construction Grammar (Goldberg 1995; Croft 2001), as a pairing of meaning and form that does not depend on syntactic relations, but rather as a reminder of the compositional nature of the form and meaning of a sentence. As a syntactic reference model I will use the Principles & Parameters approach (Chomsky 1981; Chomsky and Lasnik 1993), which is centered around universal principles with languagespecific, but not construction-specific, settings. To outline the book: In chapter 2 I will give an overview of the status of the passive in linguistic theory, combining findings from formal and functional approaches. I will argue that there is reason to use the label passive for sentences with very different surface structures and will settle on a definition of the passive that depends on the presence of the passive morpheme and the implicit argument reading. Deconstructing the passive means taking a construction apart and looking at its components and their contribution to the syntax and the semantics of the construction as a whole. Chapter 3 is concerned with showing that the various surface forms of the passive are a result of the interaction of the verb and the passive morpheme with general principles of structure, such as the Theta Criterion and the Case Filter. One normally thinks of the passive as a construction that involves a subject that is semantically the object of the verb, a form of auxiliary be (or, alternatively, of get), a passive participle and, optionally, an agentive by-phrase. However, the passive

4

Chapter 1 Introduction

has many different surface forms in English. Not only are there passives with and without by-phrases, there are passives with and without thematic subjects, with and without auxiliaries, and with and without postverbal objects. What they all have in common is the passive participle and the implicit agent reading.3 Chapter 3 also includes a corpus-based study of a specific type of passive: The get-passive illustrates that the passive as such does not have a ring of formality about it. When combined with get, a verb that is characteristic of colloquial English, the result is characteristic of spoken language and the verbs that are passivized are verbs not associated with formal registers (such as get knocked over). To further explore the compositional character of the passive, I will analyze the get-passive in the context of other constructions that are used as complements of get. I will show that characteristics often associated with the get-passive, such as dynamicity, responsibility of the non-agent subject for the occurrence of the event, and a tendency for expressing events that are of an adversarial nature, can either be derived from the syntactic characteristics of get, or are simply tendencies arising from the characteristics of spoken discourse. There is no indication that they are built into the construction, and the get-passive thus provides further evidence against an account that emphasizes the construction over its components. In chapter 4 I will review research on the syntactic reality and representation of the implicit argument in the passive construction. Based on data from syntax, semantics, and psycholinguistics, I will argue that the implicit argument status is directly related to the process of passivization itself and that a number of characteristics of the passive can be derived from satisfying an open argument position at the level of argument structure. Finally, in chapter 5 I will turn towards the use of the passive in what some would consider its natural habitat, the language of academia. I will first discuss rhetorical changes that led to a strong movement against the use of the passive in academic writing, particularly in the United States, and towards forms that express agenthood more overtly. I will analyze whether recommendations found in style sheets and writing manuals have any effect on the use of the passive. In particular, I will discuss which constructions the passive competes with and how they compare with regard to the expression of agenthood. The chapter is based on data from the FROWN corpus of American English, a small-sized corpus of abstracts from peer-reviewed re3. For the sake of simplicity, I will use the term “implicit agent”, even though there are passives in which the implicit agent does not have an agent role (for example, passives formed on the basis of psych verbs, such as fear).

Chapter 1 Introduction

5

search journals, as well as on findings from the historical ARCHER corpus. Overall, the goal is to contribute to the understanding of the passive as a construction whose syntax and semantics are compositional and whose use cannot not be fully explained without reference to the notion of an implicit argument.

Chapter 2 The English passive and linguistic theory 1. Introduction In this chapter, I will give an overview of research on the English passive, with an emphasis on questions that center around the form of the passive, particularly in the generative framework (section 2), and on placing the passive within a larger context of constructions that have no agent subjects (section 3). Section 4 is a sketch of two of the main functions of the passive, and section 5 presents data from language acquisition that show that children as young as three easily form and understand passives, even those based on non-existing verbs, suggesting that whatever constitutes the process of passivization, it is acquired well before the age of five.

2. The status of the passive in linguistic theory Perhaps no single construction has received more attention throughout the history of generative linguistics. (Baker 1988a: 305)

Baker’s statement is certainly true for early generative grammar, which had a strong focus on transformation rules. Which construction would lend itself more readily to a theory that was all about rearranging syntactic constituents? In order for a transformation rule to be applicable, the string on which it operates had to be described. In the case of the passive, that string seems to be an ordinary transitive sentence NP – V – NP, and all the specific information about what happens in a passive – word order change, insertion of an auxiliary, adding a suffix to the verb stem, inserting a byphrase – had to be put into the passive transformation rule: “Thus, the passive transformation applies to strings of the form NP – Aux – V – NP and has the effect of interchanging the two noun phrases, adding by before the final noun phrase, and adding be + en to Aux” (Chomsky 1957: 61). The Principles and Parameters approach (Chomsky 1981; Chomsky and Lasnik 1993), however, more or less eliminated construction-specific rules. It is characterized by “the effort to decompose such processes as ‘passive,’ ‘relativization,’ etc., into more fundamental ‘abstract features’: the Case Filter, the binding principles, Move-α and the principles of bounding etc.” (Chom-

8

Chapter 2 The English passive and linguistic theory

sky 1981: 121). The idea is that the grammaticality status of a sentence is negotiated on the basis of principles of Universal Grammar and their parameter settings in a given language. Movement in the passive, for instance, is motivated by a principle called the “Case Filter.” Every NP4 needs to be in a specific structural relationship with a Case-assigner, and passive participles, unlike verbs, cannot assign Case.5 Therefore, they cannot be followed by a direct object and the original object has to move out of this position to a position in which it is assigned Case. This kind of movement is referred to as “A-movement,” “NP-Movement,” or “DP-Movement,” and it is at work in a variety of constructions in English, not just in the passive (Haegeman and Gu´eron 1999; Carnie 2007). The next step in the analysis of the passive was an answer to the question of why exactly the passive participle is not able to assign accusative Case to an NP (Jaeggli 1986; Roberts 1987; Åfarli 1989; Baker, Johnson and Roberts 1989). The context for this kind of research was a new focus on the grammatical properties of affixes, exemplified by the application of X-bar Theory to the level of words (Selkirk 1982) and by Baker’s (1988a) incorporation theory, according to which affixes are syntactically active elements that attach to stems via movement. In his seminal paper on the passive, Jaeggli (1986) argued that the syntax of the passive can, to a large extent, be explained as a response to the lexical features of the passive morpheme. Attachment of the affix to the verb stem is like the first domino in a chain that can take very different shapes but that is still predictable. The core of Jaeggli’s analysis – affixes can behave like arguments – has been absorbed into more recent developments of generative grammar, such as the Minimalist Program (Chomsky 1995), even though the whole mechanism of Case assignment underwent a major revision, going from an asymmetrical relationship between assigner (e. g., the verb) and assignee (the NP) to a symmetrical relationship of checking off Case features shared by the verb or a functional category and a nominal argument. Quite strikingly, the passive did not get a lot of attention in the discussion of argument alternations in the nineties (Fagan 1992; Jackendoff 1992; Hale and Keyser 1993; Levin 1993; Tenny 1994; Ritter and Rosen 2000). Most of this research was concerned with linking theory – the relationship between lexical entries and the grammatical behavior of verbs. The focus was on which verbs allow for a specific alternation and which do not, and 4. For the discussion at hand, it does not matter if one refers to NP (noun phrase) or DP (determiner phrase). 5. “Case,” with a capital C, refers to the abstract notion of Case as developed in Case Theory (Chomsky 1981).

The status of the passive in linguistic theory

9

on how an alternation affects the meaning of a sentence. For instance, in the so-called “conative alternation” an event goes from being telic to atelic through the intervention of a preposition between verb and direct object (e. g., Celia ate the apple/Celia ate at the apple), while the spray/load alternation (e. g., She sprayed the paint onto the wall/She sprayed the wall with paint) illustrates that the progression of an event can be “measured out” (to use Tenny’s term) through incremental changes in the direct object. The passive, being very productive and not having any influence on the aspectual structure of the event, was apparently not very interesting from this perspective. Nor was it a particular interest in early Construction Grammar. In Goldberg’s approach to argument alternations (Goldberg 1995), the passive is more or less omitted as well, while datives, resultatives (e. g., She sang herself hoarse), and even the rather specific way construction (e. g., She elbowed her way into the room) get their own chapters. The very fact that the passive does not seem to change the verb’s meaning makes it a primary candidate for functionalist analyses. Passives mean choice: Speakers may choose the passive over the less marked active, and this change in word order is accompanied by a change of perspective. The passive is much more than an order variation with an SVA [Subject Verb Adverbial]. It involves a structural reorganization of the clause, and can be described as a systematic means of choosing a participant other than the agent as the starting point for a message, without departing form the normal subjectinitial word order. . . . Passive and active constructions are by no means equivalent, and their use varies widely upon the type of text. (Biber et al. 1999: 154)

Active and passive clauses may be equivalent in propositional meaning, but they present an event from different points of view. To put it in terms of Lambrecht (1994), as “pragmatically structured propositions” they are clearly very different. The information structure of the passive and the active clause is not the same: Passives never have agents as their topics, in the sense of “the thing which the proposition expressed by the sentence is ABOUT” (Lambrecht, 1994: 118, his emphasis), while actives usually do, to name just one difference. In English, a subject-oriented language (Li 1976) with grammatical word order (Thompson 1978), subjecthood and topichood often coincide. The mapping of the semantic object onto the position of the subject in the passive can therefore be seen as a side effect of its being positioned according to its informational status as the topic in the clause. Within a syntactic approach to the passive, however, one would argue that the semantic object undergoes a syntactic operation (movement) that is motivated by syntactic principles: The Extended Projection Principle (EPP) re-

10

Chapter 2 The English passive and linguistic theory

quires that the subject position be filled, and the Case Filter requires that the object be assigned Case (for details see chapter 3). Movement to the subject position satisfies both principles: The subject position is filled, and the moved NP receives Case from the inflectional head of the clause (INFL). Support for this approach comes from the fact that once a verb is changed into a passive participle, the object cannot remain in its original position in the passive: it has to be moved. Also, passivized verbs that are not followed by NPs (such as verbs with object clauses) do not induce movement – where there is no NP, there is no Case Filter violation. In these cases, the EPP is satisfied by the insertion of a semantically empty subject (It was expected that they would be late). Regarding the information structure of the sentence, however, it is not so much the structure of the passive clause that has to be explained, but the choice between the active and the passive that has to be motivated. Because English is a language with strict word order and hardly any case markings, constituents cannot be arranged freely according to their information status, they are positioned according to their syntactic function (Hawkins 1986). Since themes are usually realized as internal arguments and are projected as direct objects, and since objects follow the verb in English, themes cannot easily be the topic in a sentence (given that topics tend to occur at the beginning of the clause, i. e. in the subject position). Choosing the passive over the active allows for the theme of the action to be realized in the subject position. Other theme-promoting constructions are the middle construction (e. g., This shirt irons easily) and the tough-construction (e. g., John is easy to convince). Unlike the passive, these two constructions express qualities rather than events and do not allow for the realization of individual agents, a hallmark of the passive construction (see chapter 4). Thus, one motivation for using the passive is to map a non-agent argument to the position of a topic (taking the route of changing its grammatical function from object to subject). Another way to look at it is that the association of agent and subject is broken up so that the agent is not in the position of the topic anymore. This may be desirable, for example, if the agent is introduced as new information, or if the agent is simply considered to be not relevant or topic worthy, or if its identity is not to be revealed. Another factor to take into account is the principle of end-weight, “the tendency for long and complex elements to be placed towards the end of a clause” (Biber et al. 1999: 898). It is difficult to say when exactly a phrase becomes “heavy,” but there is a correlation between weight and information status: “heavy constituents are more likely to be new than old” (Huddleston and Pullum 2002: 1371). Functionalists have been exploring these dimen-

The status of the passive in linguistic theory

11

sions of the passive within the domain of the clause (G´ıvon 1993) as well as larger chunks of discourse (Thompson 1987). One question addressed in functional analyses is that of whether the passive is more about backgrounding the agent or about promoting the theme. One possible answer is given by Thompson (1987), who argues that passives with and without byphrases have to be looked at separately. The passive without a by-phrase, commonly referred to as the “short passive,” is used to minimize reference to the agent, and the long passive is used in order to create more cohesion between sentences (because a non-agent argument has “higher thematicity” than the agent, for example). Functionalist approaches are also often guided by frequency analyses. The underlying assumption is that “‘Grammar’ itself and associated theoretical postulates like ‘syntax’ and ‘phonology’ have no autonomous existence beyond local storage and real-time processing” (Bybee and Hopper 2001: 2–3) and that the structure of language is driven by the communicative needs of its speakers. Corpus studies tell us, for instance, that in English the overwhelming majority of passives are used without a by-phrase (Stein 1979; Biber et al. 1999). This means that if there is anything like a unifying function of the passive, it will not be that of bringing the agent into a focus position at the end of the clause, as this is the case only in a minority of passives. It does not mean, however, that the by-phrase, or rather the possibility to express the agent in a by-phrase, is not an integral part of the passive construction. If frequency matters, one needs material to count and one will often turn to linguistic corpora.6 Developments in corpus linguistics, like standardized systems of annotations and tagging, have led to fine-tuned methods of collecting and analyzing data, enabling linguists to contextualize their analyses and to address questions like why a construction may be more common in a specific genre or register and how it evolved over time. The BROWN Corpus, compiled at Brown University, dating from the 1960s, for instance, has 2000-word samples from different types of written American English; the HELSINKI Corpus has samples from early periods of English; and the FROWN (Freiburg, Brown) Corpus follows the structure of the BROWN 6. Meyer (2002:1) notes that “many generative grammarians have shown an increasing concern for the data upon which their theories are based,” but at the same time he claims that data collection “remains at best a marginal concern in modern generative theory.” This is, of course, unfair: Generative linguists are mainly interested in describing the language system, and the fact that a construction does not occur in a given corpus does not allow any conclusions about whether or not it is part of the grammar of a language.

12

Chapter 2 The English passive and linguistic theory

Corpus for American English, but with more recent data (Meyer 2002). It is well known that the passive is a construction that is much more prominent in written than in spoken English. For example, Biber et al. (1999: 476) report for the lswe corpus of spoken and written English (about 40 million words of British and American English) that the percentage of finite verbs in the passive is 2 % in conversations, 15 % in news articles, and 25 % in academic prose. Therefore, I will concentrate on written English here. Another strand of research focuses on the formal description of information structure. Lambrecht (1994), for instance, recognizes that to a certain extent grammatical structures are autonomous and are not driven by communicative requirements of discourse: “There can only be a mapping from types of situations to preestablished formal types. Speakers do not create new structures to express new meanings. They make creative use of existing structures in accordance with their communicative intentions” (Lambrecht 1994: 26). A case in point is the passivization of intransitive verbs: English does not have impersonal passives of intransitive verbs, but German and Dutch do (see chapter 3). This does not necessarily mean that in German or Dutch the passive has a function that it does not have in English, as claimed by Cornelis (1996). It could simply mean that in German and Dutch there are different requirements for filling an empty subject position than in English.7

3. What is a passive? In her crosslinguistic study of the passive Siewierska (1984) gives a very good reason for reexamining the tacit understanding that everybody has of what constitutes a passive: “The analysis of the various constructions referred to in the literature as passive leads to the conclusion that there is not even one single property which all these constructions have in common” (Siewierska 1984: 1). Any analysis of the syntax and the function of the passive will have to define what exactly is regarded as a passive and what is not. Most definitions of the passive are based on the relationship between the active and the passive and compare the two with respect to morphology (passives are usually marked with affixes), syntax (passivization leads to 7. Impersonal passives require dummy subjects to fill the empty subject position (since there is no object that could be moved to fill the position). English does have dummy subjects, but they seem to be limited to rather specific syntactic and semantic configurations, see the discussion in chapter 3.

What is a passive?

13

changes in word order), and semantics (the propositional content of the active and the passive is the same).8 These characteristics, however, do not always coincide, as Chomsky reminds us: “Even within a single language that has syntactic passives with movement and passive morphology, we may find passive morphology without movement, movement with the sense of passive but without passive morphology and the passive sense with neither passive morphology nor movement” (Chomsky 1981: 122). Haspelmath (1990: 27) gives the following definition of the passive: “A construction is called passive if: (i) the active subject corresponds either to a non-obligatory oblique phrase or to nothing; and (ii) the active direct object (if any) corresponds to the subject of the passive; and (iii) the construction is somehow restricted vis-`a-vis another unrestricted construction (the active), e. g., less frequent, functionally specialized, not fully productive.” These characteristics can all be observed in (1), but what about the sentences in (2) – should (2b) also be considered a passive sentence? (1)

a. b.

The butler murdered the detective. The detective was murdered (by the butler). (from Quirk et al. 1985)

(2)

a. b.

The butler broke the glass. The glass broke.

We can see why the passive in English is best regarded as a category that involves “two grammatical levels” (Quirk et al. 1985): It does not only affect the organization of the verb’s arguments in a sentence, but also the morphological form of the lexical verb. In English, the passivized verb takes the form of the past participle and auxiliary be is inserted to encode tense and agreement. I will argue that the syntactic structure of a passive clause is a direct consequence of an operation on the verb’s argument structure (chapter 4). This will lead us to a morpheme-based definition of the passive. Throughout this book I will demonstrate that the passive is not a monolithic construction and that it can have many different looks and multiple functions – not only across languages, but also in English. In the tradition of generative grammar, I will assume that restrictions on the English passive 8. Passivization may also change the semantics of a sentence. This is most evident when the scope of quantifiers is affected. Quirk et al. (1985: 165) give the following example: While the sentence Every schoolboy knows one joke at least favors the reading in which each schoolboy knows at least some joke or other, the passive One joke at least is known by every schoolboy favors the reading that there is one particular joke that every schoolboy knows.

14

Chapter 2 The English passive and linguistic theory

may be arbitrary and purely syntactic (as a reflection of the autonomy of syntax, see Newmeyer 1998: 25–55) and that grammar provides an array of options from which speakers can choose. In this approach, the distinction between grammatical and ungrammatical sentences is highly relevant. The fact that some verbs do not form a passive does not necessarily indicate that a passive based on these verbs would not serve any of the functions one normally associates with the passive. As mentioned above, the fact that there is no impersonal passive in English (*It was danced all night) most likely results from grammar-internal, rather than cognitive or functional, factors (for details see chapter 3). In order to better understand the choices that speakers make, we need to understand the options from which they choose. Since the passive does not normally change the propositional content of a sentence, there must be some extra value in expressing an event in a way that is doubly marked: It is more complex morphologically, and the expected linking of semantic role and syntactic position (agents9 are normally subjects, themes are direct objects) is broken up, which makes the passive more difficult to process than the corresponding active (Pinker 1989). In order to describe the menu that speakers can choose from I will rely on grammaticality judgments rather than corpus data in this section. If a certain type of passive is rare in recorded data, it does not mean that it is a peripheral phenomenon in English grammar (in the Chomskyan sense of periphery as the domain of marked and exceptional elements). 3.1. Towards a morpheme-based definition One of the reasons to regard the passive “as first and foremost a verbal morphological category whose meaning implies certain changes in the clause structure” (Haspelmath 1990: 25) is that “in general passive constructions without passive morphology do not exist” (Haspelmath 1990: 27). All the other ingredients of the passive are not essential: There are passives without theme subjects (It was believed that . . . ), passives without by-phrases, passives without be or any other auxiliary (in reduced relative clauses, for example, see chapter 3), but there are no passives without the passive mor9. Throughout this chapter I will use the term “agent” in a rather broad way, comprising all three components of agenthood identified by Jackendoff (1990: 129): “doer of action” (the instigator of an event), “volitional Actor,” and “extrinsic instigator” (roughly the causer of some sort of change, not necessarily a human being).

What is a passive?

15

pheme. In English, this means that there is no passive without a passive participle. In choosing a morpheme-based definition, one puts the verb center stage in the discussion of the passive. There are several reasons for this: First, as pointed out by Haspelmath, there is a causal relationship between adding the passive morpheme and changes in the word order of the sentence. The former induces the latter. An affix can be regarded as the instantiation of a lexical rule that affects the properties of a verb (Aronoff 1976; Williams 1981a) or, alternatively, affixes can get lexical entries that specify how they will affect the lexical properties of the base they attach to (Selkirk 1982; Grimshaw 1990). Nominalization suffixes like –er or –ion, for instance, may be the visible instantiation of a rule of nominalization, which changes the category of a lexical item from verb to noun (V → N). Alternatively, they can be regarded as lexical items with categorial features of their own. The category of the complex word will be determined by percolation of the categorial features of the head of the word, which is the rightmost element specified for these features (relativized head). In the case of -er, the complex word will therefore be a noun. If the lexical properties of a verb are changed, changes in the syntactic behavior of the item will follow. For instance, the verb drive has two arguments, the driver (agent) and the entity that is driven (theme), but the noun driver does not have an agent argument anymore – one can be a driver of a car but not a *driver by a teenager (Williams 1981a; Booij 1992). In the tradition of generative grammar, I will argue that passivization affects the lexical properties of the verb in at least two ways: The verb cannot assign accusative Case anymore (and does therefore not license a direct object), and the external argument is “suppressed,” which means that it will not be realized in an argument position (subject or object position) in the passive. It can, however, be realized inside a by-phrase. The introduction of an auxiliary can also be attributed to the morphology of the English passive: A finite verb need not be accompanied by an auxiliary, but a participle cannot normally stand on its own. The introduction of an auxiliary is dependent on this category change. Note that in a sentence that does not require a finite verb the passive participle need not be accompanied by an auxiliary. (3)

a. b. c.

Harry ran the race. Harry was running the race./*Harry running the race. The race was run by Harry./*The race run by Harry.

Dryer (1982) points out that this way of forming a passive is rather unusual and that it is far more common for a language to have a passive affix as

16

Chapter 2 The English passive and linguistic theory

part of the finite verb, inside tense and agreement inflection. In that case, no auxiliary is needed to pick up tense and agreement affixes. Syntactic heterogeneity within passives does not constitute a problem for the morpheme-based approach. Not every passive can be characterized by movement of a direct object. Even in English, a language in which intransitive verbs cannot be passivized, there are passives with impersonal subjects, in particular passives based on verbs with clausal complements. The morpheme-based approach allows us to explain why there is no movement in this case: If a verb through passivization loses its capability to license a direct object, we can explain why passives with verbs that have direct objects look different from verbs that have clausal complements: (4)

a. a’ b. b’

They explained the problem. The problem was explained by them./ *It was explained the problem (by them). They explained why we should leave early. It was explained (by them) why we should leave early.

In (4a’) the passive version with the dummy subject, it, is ungrammatical because the verb (now a participle) is still followed by a direct object (the problem). If, however, the verb does not have a nominal object but a clausal complement instead, as in (4b), passivization does not seem to have any effect on the position of this complement and the subject position can be filled with it as a placeholder element (see chapter 3 for details). Traditional sentence-based definitions of the passive do usually not include passives like (4b’): “[A]t the clausal level, changing from the active to the passive involves rearrangement of two clause elements, and one addition . . . : (a) The active subject becomes the passive AGENT; (b) the active object becomes the passive subject; and (c) the preposition by is introduced before the agent” (Quirk et al. 1985: 159). It has just been demonstrated that there are passives without thematic subjects, which is basically a way of saying that there are passives with and without movement. It should be added that there are also passives in which the moved element does not correspond to the direct object in an active clause. The moved NP can also be the subject in an embedded non-finite clause, as in (5a’), also called a raising construction, or the object of a preposition, as in (5b’), sometimes referred to as a “pseudo-passive,” or an indirect object, as in (5c’), also known as the recipient passive. Both the recipient passive and the prepositional passive became available in English only after the merging of accusative and dative case, which were separate morpholog-

What is a passive?

17

ical cases in Old English (Jespersen 1927; Fischer, Kemenade, Koopman and Wurff 2000: 78–79). There is really no reason to consider data like the following as anything but regular passives in English. (5)

a. a’ b. b’ c. c’

I expected [him to leave]. He was expected [ to leave] (by me). Queen Victoria slept in this bed. This bed has been slept in by Queen Victoria. I gave him the letter. the letter by me. He was given

Let us now look at the next ingredient of the passive, auxiliary be. What about the sentences in (6) – do they qualify as passives or not? (6)

a. b.

My license got taken away (by the police). Do you know anything about the poem written by Chaucer?

In (6a) we are dealing with an instance of the so-called “get-passive.” Get can be substituted for was here without any significant change in meaning. The sentence has a theme subject, a past participle, and an optional byphrase with agent interpretation. However, it does not have an auxiliary – we can easily determine that get does not behave like an auxiliary verb at all. For instance, it cannot be fronted to the initial position (the Comp position) for purposes of forming a question, and do needs to be inserted for negation: (7)

a. b.

When was/*got your license taken away? My license was/*got not taken away.

If one adopts a morpheme-based definition of the passive, there is no question that sentences like (6a) qualify as passives: The verb is in its participle form, it cannot be followed by a direct object (*It/There got my driver’s license taken away) and the agent of the verb is suppressed and can optionally be realized inside a by-phrase. The fact that there is no auxiliary is something that has to do with the syntax of get. In chapter 3 it will be argued that get is a main verb and takes a predicative small clause complement here (Haegeman 1985). Small clauses do not have syntactic projections of finiteness and that is the reason why the passive does not have an auxiliary here. (6b) illustrates another form of a “truncated passive” (Stein 1979) or “bare passive” (Huddleston and Pullum 2002). Again, the fact that an auxiliary is missing has nothing to do with the passive as such. We are simply dealing with a participial relative clause, or, as Huddleston and Pullum (2002: 1265) call it, a “gerund-participial.” They consider examples like (6a) instances

18

Chapter 2 The English passive and linguistic theory

of the passive because they allow a by-phrase. “Passive gerund-participials contrast with the past-participials in aspectuality as progressive vs. nonprogressive, but with actives the progressive vs. non-progressive distinction is lost,” as illustrated in the following example. (8)

Do you know anything about the student living in Boston?

Again, one can see that there may be syntactic configurations in which a passive occurs without an auxiliary. This is not to say that the auxiliary is generally optional in the passive. It has a syntactic function in most sentences, but there are constructions in English in which participles can be used without an auxiliary, and they need not be passives. It is safe to say, then, that the meaning of the passive does not depend on the auxiliary per se. The next element to look at is the by-phrase. The semantics of the byphrase corresponds directly to the semantics of the subject in an active sentence. (9) illustrates the compositional nature of the semantic role (also known as theta- or j-role) assigned to the external argument: Throwing a baseball is something one does with one’s arms and Fernando is the one who is doing it, in (a) as well as in (b). Throwing a party, on the other hand, is something that one does with one’s checkbook, and the department is the one who is paying, in (a) as well as in (b). Note that the agent reading (someone is throwing the ball/the party) is present also when the by-phrase itself is missing: “the implicit argument or the NP in the by-phrase fulfills exactly those th[eta]-roles which correspond to those of the active sentences: in other words, compositionality of the external th-role is preserved under passive” (Wyngaerd 1988: 159–160). Sentences like these also illustrate that any selectional restrictions that hold for the external argument (subject) in the active also hold for the by-phrase, a correlation that supports the assumption that the two are structurally related (either through transformation or through a common source, such as an element in the lexical entry of the verb). (9)

a. b.

Fernando threw a baseball./A baseball was thrown Fernando. The department threw a party./A party was thrown department.

by by the

In (10) we can see that the external argument of a verb need not always be an agent (see in particular (10b)), and that the moved element need not always be a classic theme – in (10c), for instance, the subject of the passive is a goal. The semantic role itself does not seem to matter, but whatever it is, the role

What is a passive?

19

of the by-phrase clearly corresponds to the role of the subject in the active construction.10 (10)

a. b. c.

The client admired the picture./The picture was admired by the client. The sudden noise frightened the children./The children were frightened by the sudden noise. They sent me a letter./I was sent a letter by them.

Although the by-phrase itself is optional and although most passives are “short passives,” i. e. passives without a by-phrase, the importance of the by-phrase should not be underestimated, as the presence of the by-phrase can be related directly to the lexical operation on the argument structure of the verb that the passive morpheme represents. The technicalities of this analysis will be discussed in detail in chapter 4; suffice it to say here that the (optional) by-phrase is an integral part of the passive construction, even though it is not expressed in the majority of passives. Adding a by-phrase is also a simple test for keeping true passives apart from so-called adjectival passives. The term “adjectival passive” is somewhat misleading – what looks like a passive is actually a copula construction with an adjectival predicate. It is well known that sentences like (11) have two readings: They either refer to states, with be as a main verb followed by an AP, or they refer to events, with be as an auxiliary followed by a participle (Wasow 1977; Chomsky 1981: 54–55; Levin and Rappaport 1986). This ambiguity does not occur in languages like German, Dutch, or Italian, which do not use the equivalent of be to form the passive. The passive auxiliary in German is werden (become). In German, then, sentences like (11a) have two translations, and one has to make up one’s mind as to whether the sentence is construed as a copula construction or a passive. (11)

a. b.

The wall is painted. The window was broken.

(12)

a.

Die Wand ist/wird gestrichen. the wall is/becomes painted ‘The wall is painted’

10. Examples like (10c) are problematic because the passive participle is followed by an NP (something ruled out by the assumption that the passive participle does not license a direct object that needs Case). For a more detailed discussion of ditransitive verbs see chapter 3.

20

Chapter 2 The English passive and linguistic theory

b.

Das Fenster war/wurde zerbrochen the window was/became broken ‘The window was broken’

In English there are several ways to test if one is dealing with an adjectival or verbal passive (or, to be more precise, with an adjective or a participle). As pointed out by Hust (1977) and Wasow (1977), adjectives can be coordinated with other adjectives, but they cannot be coordinated with participles. In (13a) unloved is an adjective (it cannot be a participle, as there is no verb unlove)11 and it can be coordinated with noisy, another adjective. The ungrammaticality of (13b) indicates that spanked is of a different class than unloved. (13)

a. The child was (both) noisy and unloved. (from Hust 1977) b. *The child was (both) spanked and unloved. c. *The child was (both) noisy and spanked.

Furthermore, adjectives do not normally license an agentive by-phrase.12 They can be the complement of verbs that are subcategorized for adjectival predicates, like seem, remain, or look, and, like other adjectives, they may be specified by degree adverbs, e. g., very. However, since not every adjective expresses a gradable quality, the ability to take a degree adverb is not a necessary condition for an adjectival passive (Huddleston and Pullum 2002: 1436–1440). Overall, one can generalize: “The adjectival passive has simply the distribution of an adjective” (Williams 1981a: 96). Like other adjectives, it can also occur after dynamic verbs like get or become. The whole sentence will have a resultative reading, as in (14d), with the adjectival passive embedded. (14)

a. b. c. d.

Harry seems happy/scared/ *kissed by Hermione. Harry was very happy/very scared/*very kissed by Hermione. The wall is being painted/*unpainted by the decorator herself. The substance became solid/frozen.

11. Prefixation with un- is often used to disambiguate between an adjectival and a participial reading. It should be noted, though, that verbs can also be prefixed with un- (unload, unzip), but un- here does not simply add negation to the word’s meaning (as in unkind, unhappy, unworthy), it expresses reversal of an accomplishment. 12. Note, however, examples like He remained unnoticed by most passengers. This is not a passive, since there is no verb to unnotice. The generalization is that every passive can take a by-phrase, but that not every by-phrase with agentive meaning occurs in a passive (a symphony by Mozart).

What is a passive?

21

The morphological relationship between the adjectival passive (which should more aptly be named “passival adjective”) and the verbal passive will be examined more closely in chapter 3. I will take the view that the passive is a construction that results from a morpheme-based lexical operation on the argument structure of the verb. Changes in the realization of the verb’s arguments are a consequence of this operation. All of this is, of course, a matter of definition. For instance, examples like (15b) have been referred to as “passive nominals” (Chomsky 1970; Grimshaw 1990). The label “passive” is applied because the agent of the underlying verb is realized inside a by-phrase and the original theme occupies the NP-internal “subject” position (the specifier of the NP). (15)

a. b.

the architect’s construction of the concert hall the concert hall’s construction (by the architect)

Although this construction seems to share some similarities with the passive on a superficial level (change of word order, optionality of the agent, the preposition by), it is first and foremost an instance of nominalization. It is not a passive: Neither does it involve passive morphology (there is no participle, nor, for that matter, any other verb), nor does it affect the whole clause (only an NP), and it underlies restrictions that are different from those for passivizing verbs. One of the restrictions mentioned with regard to constructions like (15b) is the “affectedness constraint” (Anderson 1978), the requirement that the fronted NP be “affected” (moved, changed, or altered in status) by the event. However, in the passive, there is no such requirement. For example, the verb admire does not express a change of state in the object. Nevertheless, it can be passivized, and in the course of passivization, the object is moved to the subject position (The building was admired by the architect). In the nominalized version, however, the theme cannot be preposed inside an NP due to the affectedness constraint (*the building’s admiration by the architect). The affectedness constraint also seems to hold for some verbal argument alternations, among them the middle construction, illustrated in (16c). Both (16b) and (16c) are constructions that share some of the syntactic and semantic characteristics of the passive, but they do not involve passive morphology and therefore do not constitute passives themselves.13 (16)

a.

This fact is known by anybody.

13. Because they do not have passive morphology, they also do not show implicit argument effects, a hallmark of the passive (discussed in detail in chapter 4).

22

Chapter 2 The English passive and linguistic theory

b. c.

the knowledge of the fact/ *the fact’s knowledge (Grimshaw, 1990) This shirt irons easily./*This fact knows easily.

Let us now look at some sentences that may or may not be considered passives, depending on the definition one uses. They all contain an element that looks like a passive participle. In (17) we are dealing with passives in which the subject, he, does not originate as a direct object of a verb. (17)

a. b. c.

The sniper made 5 phone calls and was hung up on five times. (from a report on CNN, Oct. 23, 2002) . . . the day I got broken up with by a Post-it. (from the TV-series Sex and the City, episode conclusion, Aug. 2003) Along his journey, he has been snowed upon, sluiced by slash- ing rains, sliced apart by the sun, attacked by insects, nearly done in by hunger. (from an article in The New York Times, Feb. 1, 2003)

In (17a) the subject NP The sniper is moved out of the prepositional phrase headed by on. This PP can be considered a complement of the particle verb hang up. While the stranding of a preposition through movement is common in English (Biber et al. 1999), preposition stranding in passives challenges the assumption that movement in the passive is motivated by the verb’s loss of the ability to assign Case. In prepositional passives the moved element follows a preposition. Prepositions can always assign Case, and they are not affected by passive morphology – so why does the NP move? This problem has been widely recognized in syntactic theory and will be addressed in more detail in chapter 3. One could, of course, say that the sentence does not fit the theory and should therefore not be counted as a passive, but this would be like running away from the data: The sentence has the syntax, the morphology, and the semantics of a passive (it includes an implicit agent and a by-phrase could be added), even if it constitutes a challenge to the theory of movement. In (17b) the passive occurs inside a relative clause and the moved element, I, originates in the position after with. Again, the PP is a complement of a particle verb, break up. This sentence also combines an instance of the get-passive and a non-agentive by-phrase. (17c) is a very intriguing example. Here we find an enumeration of five passive verb phrases that involve different syntactic classes of verbs, the prepositional verbs snow upon, the transitive particle verbs slice apart and do in, and the regular transitive verbs sluice and attack, all verbs that express some sort of forceful impact. There

What is a passive?

23

is no reason to consider only the verb phrase in which the trace directly follows the participle of a transitive verb a passive. The repetitive effect of multiple passives of verbs of contact adds to the interpretation of the subject as someone who truly was exposed to hardship. The following examples are classified as passives by Zandvoort (1957) and Stein (1979). Stein points out that Zandvoort uses a “wide” definition of the passive: Passives are usually formed with be, but be may be exchanged for another verb. From what has been discussed so far, it is clear that the sentences in (18) are not true passives: They do not have an active correlate and they do not license a by-phrase, they are just “adjectival passives.” (18)

a. b. c.

The door remained locked. He felt thoroughly disappointed. The village looked quite deserted.

Cases like (19) are definitely not to be considered passives. They are fossilized remnants of a time when English was a language with two different auxiliaries for forming the perfect, very much like modern German, Italian or French.14 They do not have the syntax of a passive nor the implicit agent reading. (19)

She was gone.

The next example is a variation of sentences like (2b). It is from William Safire’s weekly column “On Language” in The New York Times Sunday Magazine (Nov. 2, 2003), and he refers to constructions like these as the “mysterious passive.” (20)

To describe how the times they are a changin’ – revealing them- selves in magical ways – today’s bureaucratic elite say that events unfold.

The clause in question (“events unfold” – with italics in the original) is, of course, not a passive. It does not have passive morphology, it does not have the syntax of the passive (one cannot add a by-phrase), and it does not have the semantics of the passive either: A crucial characteristic of the passive is the implicit agent reading, but there is no hidden agent in the sentence. While it may be the case that events in the real world are always caused by something, the linguistic representation of events can be non14. In a study of auxiliary selection in English from the 16th to the 19th century, Lipson (2000) found that the percentage of be-perfects for go went down from 64 % to 6 %. Even more impressive are the figures for the be-perfect for come: The rate went down from 82 % to 0 %.

Chapter 2 The English passive and linguistic theory

24

causative. That is exactly what Safire’s criticism is about: An event that is caused by something or somebody is linguistically represented as an event that just “happened” or “unfolded” or “occurred.” These are all verbs that do not have an external argument – which is exactly why they cannot be passivized. Note the difference between (20) and the passive in (21). If there is a structure with passive morphology (were made), an implicit argument (agent) reading is assigned. A famous example is Ronald Reagan’s concession, “mistakes were made,” when he had to confront the public about secret weapon deals with Iran: (21)

While we are still seeking all the facts, it’s obvious that the execution of these policies was flawed, and mistakes were made. (Ronald Reagan, Radio Address to the Nation, Dec. 6, 1986)

Far from erasing agenthood and representing an event as uncaused, the passive draws attention to the fact that somebody made mistakes. What many people found disturbing was that President Reagan did not make the agent explicit, thereby avoiding to take responsibility (or to assign it to someone else). Other famous users of this “past exonerative,” a phrase coined by political analyst William Schneider (Broder 2007), include Richard Nixon, Bill Clinton, and Alberto Gonzales. In an essay with the subtitle “Mistakes Were Made,” the novelist Charles Baxter comments on the tenure of George Bush and Ronald Reagan as follows: Their administrations put the passive voice, politically, on the rhetorical map. In their efforts to attain deniability on the arms-for-hostages deal with Iran, their administrations managed to achieve considerable notoriety for self-righteousness, public befuddlement about facts, forgetfulness under oath, and constant disavowals of political error and criminality, culminating in the quasi-confessional passive-voice-mode sentence, “Mistakes were made.” Contrast this with Robert E. Lee’s statement the third day after the battle of Gettysburg and the calamity of Pickett’s Charge: “All this has been my fault,” Lee said. “I asked more of men than should have been asked of them.” (Baxter 1998: 5)

Note that even Lee uses a passive without a by-phrase here (should have been asked of them), but the syntactic context (I asked more . . . ) makes it clear that he himself is the one who asked what should not have been asked. What we can see from these examples is: (a) the passive does not erase agenthood as such, but it allows the agent – or rather: the external argument of the verb – to be implicit; and (b) one cannot label someone as evading and vague just because he or she uses a passive construction. Agenthood can be expressed in the passive, and even when the speaker chooses a passive with-

What is a passive?

25

out a by-phrase, the identity of the agent may be clear from the linguistic context. This was forcefully demonstrated in another presidential address: Today I say to you that the challenges we face are real, they are serious and they are many. They will not be met easily or in a short span of time. But know this America: They will be met. (Barack Obama, Inaugural Address, January 20, 2009)

The line They will be met was followed by a huge round of applause and figured prominently in the media coverage of the Inauguration. It was generally considered an expression of leadership and determination. English also has verb alternations that seem to erase the agent altogether. (22)

a. b.

They sank the ship./The ship sank. He ironed the shirt./This shirt irons easily.

(22a) is an example of the causative alternation (Levin and Rappaport Hovav 1995), and (22b) illustrates the middle construction (Fagan 1988; Ackema and Schoorlemmer 1995). Both resemble the passive in that the object of the transitive verb is in the subject position and that the external argument of the verb is not explicitly realized. But that is as far as the resemblance goes – both constructions are considerably more restricted than the passive, they are not morphologically marked, and they do not license an agentive adjunct, as illustrated in (23). (23)

a. *The ship sank by the fishermen. (vs. The ship was sunk by the fishermen.) b. *This shirt irons easily by me. (vs. This shirt was ironed by me.)

Throughout this study I will assume that one of the defining characteristics of the passive is that the external argument of the verb is an implicit argument in the passive. I will argue that implicit arguments are not projected onto argument positions, but can be made explicit and specific in an adjunct position (inside a PP). In chapter 4 I will discuss in detail how exactly these effects can be related to the presence of the passive morpheme and what makes the implicit argument in the passive different from a conceptually implied argument. At this point, it suffices to say that decausativization and middle formation share some of the characteristics of the passive, but are not passives themselves. The next set of examples are classified as get-passives by Biber et al. (1999: 481). It should be noted that the authors use this term to cover adjectival as well as true passives. It has already been shown that in many instances a sentence with a form of be followed by what looks like a pas-

26

Chapter 2 The English passive and linguistic theory

sive participle allows for two readings: a passive reading with an implicit argument (verbal passive), or a state reading with a copula and an adjectival predicate (adjectival construction). (24a) seems to be of the adjectival construction kind: It would only be a passive if the involvement in local society had been caused by somebody (implicit agent reading) and if that somebody could be made explicit in a by-phrase. If, however, one interprets the sentence more or less as “we became involved,” then involved would be an adjective. As such, it could be graded with a degree adverb like very, something that seems quite likely in this case (e. g., we started to get very involved . . . ). The same holds for all other examples. These sentences are passives only if they are interpreted as events caused by an agent, which may or may not be the most natural interpretation. (24)

a. b. c. d.

And then we started to get involved in local society. My head got stuck up there. It’s about those people who got left behind in Vietnam. She got married when she was eighteen.

Finally, let us consider sentences that look like passives but that do not have active correlates or do not allow by-phrases (examples from Huddleston and Pullum 2002): (25)

a. Pat is reputed/rumored to be very rich. a’ *They repute/rumor Pat to be very rich. b. I was born in Boston. b’ *He was born by a Greek architect in Boston.

Sentences like (25a) easily license a by-phrase and thus seem to be true passives. Why is there no active correlate? According to Biber et al. (1999: 479) this is just the idiosyncratic behavior of some verbs. Huddleston and Pullum (2002: 1435) call verbs like repute and rumor “morphologically defective,” indicating that they do not have the full inflectional paradigm that verbs in English normally have. This might take things a step too far – the Oxford English Dictionary still lists them as transitive verbs (see the examples in (26)), but also notes that most transitive uses and meanings have become obsolete. (26)

a. b.

Why reputest thou the dedes of my merytes to be unworthy? (1483) Conill Lord of Connaght honourably reputed him, and with all his people was converted. (1571)

What is a passive?

c.

27

Reputeing what was don to his great friend to be don to him self. (1711)

However, Internet searches bring up more recent examples of repute and rumor as case-assigning verbs with tensed forms:15 (27)

a.

b.

Ostensibly he practiced the law, though men close to him say that he did not care for it much, and did it only for the money, of which some reputed him overfond. (Davis 1994, p. 18) However, the morning they arrived to tell the three Campbell children that their parents had drowned, they both looked less confident than reputation rumored them to be. (Bacon 2002, p. 53)

Obviously, the transitive use of these verbs has become very limited, but the forms still exist. Since the sentences in (25a) have passive syntax and morphology as well as the passive reading (somebody is spreading the rumor), I will consider them passives here, rather than adjectives (Pinker 1989: 90). The case of born is different and indeed idiosyncratic. According to the Oxford English Dictionary, until the late eighteenth century, the past participle of bear was either bore or born, with no difference in meaning. Around 1775 bore was abandoned and borne became the ordinary form, while born was restricted to the sense of bear as in “To bring forth, produce, give birth to (offspring)” and “as such occurs only in the passive, if not followed by by and the mother” (Simpson and Weiner 1989). With no enforcement through the by-phrase, there is no passive reading: it has rather a neuter signification, . . . “come into existence, sprung” without explicit reference to maternal action; hence it is the form used adjectively, and figuratively. Cf. ‘She had borne several children, the children borne to him by this woman, born of the Virgin Mary, born in a stable, her first-born son, a lady born, new-born zeal, a flower born to blush unseen. (Simpson and Weiner 1989)

This leaves us with the conclusion that in Modern English, sentences like (25b) should not be considered true passives, since they do not seem to have passive morphology (the participle would be borne) and do not have the passive reading.

15. I used the “Search inside the Book” tool, launched by Amazon.com on Oct. 23, 2003, to bring up these data. At the time, this tool allowed for simple word searches in 120,000 books. Throughout this book, bold font is added to highlight certain passages in the examples.

28

Chapter 2 The English passive and linguistic theory

For the following discussion, let us sum up the definition of the passive used throughout this book: – Morphology: Passives in English are morphologically marked. The core of the passive is the passive participle. For discussion as to whether this is the same form as, or a form homophonous with, the past participle, see chapter 3. If there is no participle (as in The glass broke or They were very worried), one is not dealing with a passive. So-called adjectival passives may be derived from participles, but they are not participles themselves. Therefore, they are not passives. – Syntax: The morphological marking has consequences for the syntax of the passive. Participles are not finite verbs, therefore another verb, usually be, is needed to encode tense and agreement in finite clauses. Another effect of the morphological marking is that the participle cannot assign Case in the passive. As a consequence, if there is an NP that depends on the verb to be assigned Case, it will have to move to another position. According to Burzio’s Generalization (Burzio 1986), the subject position of an unaccusative construction is empty and can therefore be filled by movement. But there may be instances where movement is not motivated. “Passive syntax” means that the external argument of the verb is not in the subject position; it does not necessarily mean that another argument is in the subject position, as the position can also be filled with a non-thematic element (this holds for passives like It was explained that . . . ). – Semantics: The propositional content of passives is generally the same as in the active. Unlike decausativization, passivization does not change the aspectual structure of a clause. If the active has an agent, the passive will also have an agent. In the passive, the external argument is either realized inside a by-phrase or it becomes an implicit argument (see chapter 4 for details). The “passive reading” is an interpretation of the clause in which none of the arguments of the verbs is missing. In this section, I looked at a number of passives that do not fit the description of a prototypical passive construction (based on dynamic transitive verbs). It would be difficult to reduce the passive to exactly one syntactic pattern, which is one reason why a Construction Grammar approach does not seem to be the ideal approach to the passive. The basic assumption of Construction Grammar is that “[p]articular semantic structures together with their associated formal expression must be recognized as constructions independent of the lexical items which instantiate them” (Goldberg 1995:

What is a passive?

29

1). Although I also refer to the passive as a “construction,” I do not use the term in the way it is used by Goldberg and other proponents of Construction Grammar. In the Chomskyan tradition I will consider the syntax of the passive construction as arising from interaction of general principles (such as the Case Filter and the Extended Projection Principle) and the syntax and semantics of individual verbs and their arguments, thus accounting for many different surface looks of the passive in English and across languages. This is essentially a “reductionist model” of syntax, to use Croft’s (2001) term. In his version of Construction Grammar, the category “verb” does not exist outside of the constructions in which it occurs. Establishing an independent category “verb” from which the syntactic properties of the construction can be derived is not an option in his framework. In the following section I will discuss the function of the passive as such (not of specific passive constructions), assuming that it is a systematic way a language provides to present an event and its participants. The discussion will be subsumed under the label “voice.” 3.2.

Passive as an instantiation of “voice”

3.2.1. Definitions of voice Traditionally, the passive is considered as an instantiation of voice, a category (like tense and aspect) that affects the organization of the whole clause: “Voice is a grammatical category which makes it possible to view the action of a sentence in either of two ways, without change in the facts reported” (Quirk et al. 1985: 159). This definition presents voice as an inflectional category and stresses that passivization does not affect the propositional content of a clause. Shibatani (1988: 2) offers a more abstract view of voice: “Voice is to be understood as a mechanism that selects a grammatically prominent syntactic constituent – subject – from the underlying semantic functions (case or thematic roles) of a clause.” In essence, he refers to voice as the system of linking rules between semantic and syntactic information. The basic “voice strategy” in English “is to select an agent as a subject, and the active voice refers to the form resulting from this choice of agent as a subject” (Shibatani 1985b: 2). By comparison, “in the prototypical passive form a patient functions as a subject and an agent is syntactically unencoded (or, even if it is encoded, its grammatical prominence is marginal).” While Quirk et al. (1985) emphasize the relationship between active and passive, Shibatani’s definition leaves room for other instantiations of voice. One of them would be the ergative voice strategy, a linking system in which

30

Chapter 2 The English passive and linguistic theory

the patient argument of transitive verbs and the agent argument of intransitive verbs are selected as grammatically prominent. In a way, ergative voice resembles the passive, which has led to the adoption of the term “ergative” for certain theme-subject constructions in nominative-accusative languages such as English or Italian (Burzio 1986), but in languages with ergative case ergative voice is the basic strategy to map semantic information onto syntax. The examples in (28) – taken from Comrie (1988: 14) – illustrate that in Warlpiri the agent argument of the one-argument verb shout has the same case (absolutive) as the theme argument of the two-argument verb see, while the agent/experiencer argument of see bears a different case (ergative). (28)

a.

Ngatju ka-rna purla-mi I-abs aux-1sg-nom shout-nonpast ‘I shout’

b.

Ngatujulu-rlu ka-rna-ngku nyuntu nya-nyi. I-erg aux-1sg-nom-2sg-acc you-abs see-nonpast ‘I see you’

c.

Nyuntulu-rlu ka-npa-tju ngatju nya-nyi You-erg aux-2sg-nom-1-sg-acc I-abs see-nonpast ‘You see me’

For Comrie (1988: 9), too, voice is a system of grammatical encoding of event participants. In his article on differences between ergative and passive voice he points out that while passive voice and ergative voice are alike in assigning subject properties to non-agents, they are different in that ergative voice “involves greater integration of the agent phrase into the syntax of the clause” and that ergative voice is generally unmarked (in ergative languages), while the passive is always marked, or derived. He bases his notion of “markedness” on criteria like frequency (the active is much more frequent than the passive), formal complexity (the passive involves more morphemes than the active), productivity (while all verbs have an active form in English, not every verb can be used in the passive), and discourse distribution (in English the active is preferred unless there are good discourse reasons to use the passive). By these criteria, Dyirbal, an Australian Aboriginal language, is a language with ergative, not passive, voice: In Dyirbal it is the unmarked case that the patient of a transitive verb and the agent of an intransitive verb show subject-like properties (the similarities go beyond

What is a passive?

31

absolutive case), while in English this is only possible in passives and other constructions that involve NP-movement.16 (29)

a. b. c.

The man followed the boy and the man ran away. The man followed the boy and ran away. The boy was followed by the man and ran away.

(29b) is only grammatical if the understood subject for ran away is the man. It cannot be construed as meaning “The man followed the boy and the boy ran away.” But that is exactly the reading that would be grammatical in Dyirbal (Comrie 1988: 10). In English, one would have to use the passive to achieve that reading, as illustrated in (29c). But while the passive in English is a marked voice, the ergative in Dyirbal is not. In ergative languages, one needs a marked construction (the antipassive) to achieve a distribution of arguments that is similar to the English active. Note that the behavior that is shared by patients of transitive verbs and agents of intransitive verbs in ergative languages can be limited to case marking (absolutive case, as opposed to ergative case, which is assigned to the agent argument of transitive verbs), but can go well beyond that, as in Dyirbal. The English passive is like ergative voice in Dyirbal in this respect,17 but it is crucially a marked voice. So while ergative linking systems are not inherently less natural than nominative/accusative active linking systems, in English a structure in which the object of a transitive verb shows the grammatical properties of a subject is often considered derived from a structure in which the semantic object is mapped onto the position of the direct object. In generative analyses of the passive the more basic status of active voice in English is reflected in a movement analysis for the passive. The subject in the passive starts out as an internal argument of the verb, in the position of a direct object, and then undergoes A-Movement (movement into an argument position). (30)

a. b.

The little girl cut the flowers. [The flowers]i were cut ti by the little girl.

16. Following Burzio (1986) the term “ergative” has come to be accepted for unaccusative constructions. If we assume that in a sentence like The water boiled, the subject has been moved from the object position, the sentence shows ergative properties, as the subject of the intransitive verb is generated in the same D-Structure position as the object of the transitive verb. 17. The parallels are striking: In Dyirbal, agent phrases are freely omissible, but sentences without them will give rise to an implicit agent reading, just like the English passive (Comrie 1988: 18).

32

Chapter 2 The English passive and linguistic theory

This analysis is carried over to constructions that would also qualify as “passive voice” in the definition given above, but which do not show passive morphology. Going back to Burzio (1986) and his analysis for patientsubject verbs in Italian, the intransitive use of causative verbs as in (31a) is commonly referred to as “ergative.” Although the NP The glass is undeniably the subject of the clause, it also has object-like qualities (in addition to its semantic status as an affected object). Burzio demonstrated this for Italian, making use of the clitic ne, which can only attach to direct objects, but there are also tests for (underlying) syntactic objecthood in English. One of them is the formation of an adjectival passive: Only the direct object of the verb can be the external argument of an adjective based on the past participle (broken glass versus *broken boy – in the relevant reading that the boy broke the glass), see also Levin and Rappaport (1986), Grimshaw (1986), and Wanner (1999). (31)

a. b. c.

The glass broke. → [The glass]i broke ti This shirt irons easily. → [This shirt]i irons ti easily The flower wilted. → [The flower]i wilted ti

Sentences like (31b) are known as “middle constructions,” for which a movement analysis has been suggested by Keyser and Roeper (1984) and Stroik (1992), among others. Middles express states and therefore do not allow resultative adjectives (an ironed shirt is fine, but it does not have the middle reading). And finally, (31c) is an example of a basic unaccusative construction, as described by Perlmutter (1978). As an entity undergoing a change of state, the flower is semantically an object and also shows properties of syntactic objects (the wilted flower). Again, a movement analysis reconciles the surface structure with the active linking system in English. The common name given to these constructions is that they are all “unaccusative” (Perlmutter 1978): The subject position is empty and is filled by movement. The moved element needs to be moved into this position because it is not assigned Case in its original position. Movement, then, is caused by a particular instantiation of voice in the broad sense. A very different understanding of the role of voice and the relationship between active voice and the presence of an agent is presented in recent work on the representation of the verb’s arguments in X-bar Theory. Inflectional categories are treated as syntactic heads projecting their own phrase. For example, the clause is considered to be an IP (Inflection Phrase). The agent argument, traditionally classified as the “external argument” of the verb (Williams 1981a), is separated from the verb and is considered an argu-

What is a passive?

33

ment of a functional head. The organization of argument structure and the kind of information it should encode has been widely discussed (Williams 1981b; Rappaport and Levin 1988; Grimshaw 1990; Hale and Keyser 1993). One recurring topic in this discussion has to do with the status of the argument that will eventually end up as the subject, the external argument. What makes the external argument special? First, it is the unique position of the subject: Since there can only be one syntactic subject, there can only be one external argument (while verbs can have two complements). Second, it is the compositional nature of the interpretation of the external argument that makes it different from other arguments (see the examples discussed in (9)). The interpretation of the argument in the subject position is not just determined by the verb, it is influenced by the choice of the internal argument. That is why Chomsky (1981) refers to the theta-role assigned to the external argument as an “indirect” role, assigned by the VP rather than the verb. In later work, the external argument is cut off completely from the lexical verb. It has been analyzed as the argument of a non-overt causative verb (Hale and Keyser 1993) or of an abstract do (Speas 1990), two suggestions that go back to Larson’s (1988) analysis for double object constructions, which involves an additional layer of a “small vp”. As an alternative, Kratzer (1996: 120) suggests that the external argument be treated as an argument of the inflectional head “Voice,” and that it should be generated in the specifier position of a VoiceP (voice phrase), which is positioned between the TP (tense phrase) and the VP (verb phrase). The diagram in (32) only shows the relevant part of the syntactic representation for Mittie fed the dog: (32)

VoiceP NP Mittie

Voice’ Voice [active]

VP NP

V’

the dog

V feed

In Kratzer’s model, there are two kinds of specifications of voice: active and non-active. If Voice is specified as non-active, then no external argument is licensed (and consequently no agent interpretation arises). In her analysis, the syntactic head, Voice, does not correlate with morphological voice,

34

Chapter 2 The English passive and linguistic theory

but rather with aspect: “action verbs get agent arguments, stative verbs get holder arguments, all other verbs start out with non-active voice markers, and hence without an external argument” (Kratzer 1996: 123). The external argument is added through the operation of event identification, which is possible only if the two predicates that are conjoined are compatible with respect to lexical aspect (Kratzer 1996: 122). The compositional nature of the interpretation of the external argument can also be dealt with in a lexical framework like Grimshaw’s (1990): She assumes that the external argument is not entirely different from any other argument, it is simply the “most prominent” argument in the lexical representation of the verb. Since theta-marking always proceeds from the least to the most prominent argument, it follows that the external argument is theta-marked last, i. e. after the internal argument (Grimshaw 1990: 35), which accounts for data like Marantz’s. If the external argument is truly an argument of the verb (rather than being introduced by an abstract head), one can explain more easily why some verbs have external arguments and others don’t – it depends on their lexical semantic representation. If, on the other hand, external arguments are introduced by some functional head, one needs to explain why this functional head cannot be present in every clause. I will adopt the broad definition of voice as a linking system here. Let us assume that when it comes to mapping semantic roles onto syntax, there is something like a proto-agent and a proto-patient, as described by Dowty (1991). His Argument-Selection Principle for English states that in “predicates with grammatical subject and object, the argument for which the predicate entails the greatest number of proto-agent properties will be lexicalized as the subject of the predicate; the argument having the greatest number of proto-patient entailments will be lexicalized as direct object” (Dowty 1991: 576). Among those properties are volitional involvement and causation for the proto-agent and undergoing a change of state or being causally affected for the proto-patient. In a prototypical event, then, an agent acts on a patient. Verbs with just one argument may be unergatives (with an argument closer to a proto-agent) or unaccusatives (with an argument closer to a proto-patient). The mechanism of movement in syntax allows for the marked realization of arguments (patients in subject positions) in languages in which the underlying selection principle is that of active voice. In languages with ergative voice, the basic selection principle is obviously different from the one stated above.

What is a passive?

35

3.2.2. Voice and transitivity In a narrow technical sense, transitivity refers to the ability of a verb to take a direct object or clause as its complement (compare Quirk et al. 1985: 53– 54). In generative syntax, this translates to the ability to assign Case to an NP. Unaccusative verbs take a complement at D-Structure, i. e. their single argument is an underlying object. Since unaccusative verbs cannot assign Case, the complement has to move. Movement will go to the subject position, which is always empty for unaccusative verbs. Movement also leaves a trace in its original position, which explains why unaccusative verbs cannot take cognate objects (they have an underlying object already, cf. Felser and Wanner 2001). In a framework that makes use of movement, syntactic transitivity cannot be defined via subcategorization for an object only, it also has to be defined via Case. Otherwise, unaccusative verbs, such as wilt in sentences like (33) would have to be considered transitive, as they have an underlying object. Unergative verbs, single-argument verbs with an underlying subject, should also be excluded from the class of transitive verbs. Though most unergative verbs can be used transitively (as in (33b)), the postverbal NP is not a true argument of the verb. True transitive verbs must have at least two arguments: an external and an internal argument. (33)

a. b.

The floweri wilted ti (*a slow death). The visitor smiled (a helpless smile).

In a wider sense, transitivity is a syntactic reflection of the asymmetrical organization of event participants in a clause. Structurally, the subject (NP1) is clearly “higher” than the object position (NP2), or, to put it in more technical terms, it c-commands the object:18 (34)

IP NP1

Infl’ Infl

VP Spec V’ V NP2

18. An element x c-commands y, if the first branching node that dominates x also dominates y. In this diagram, NP1 c-commands NP2 , but not vice versa.

Chapter 2 The English passive and linguistic theory

36

In English as an SVO language, the subject always precedes the object. Since there is a tendency for given information to come first in a clause, the subject tends to be the topic, the entity the clause is about. However, in English the subject position has to be filled even if it is not topical (It is assumed that . . . ). There is no such requirement for the object position. While a verb can have more than one object (one direct, one indirect), there is only one subject position in the clause. And while there are clauses without objects, there is no clause without a subject in English. If the verb does not select for a subject argument, a dummy subject will fill the position. A more comprehensive view of transitivity is offered by Hopper and Thompson (1980). They argue that Transitivity (spelled with a capital T) is a system of structuring information in terms of foregrounding and backgrounding and they identify the following parameters as dimensions along which transitivity can be measured (Hopper and Thompson 1980: 252): (1) Adicity: clauses with two or more arguments score higher than clauses with one (2) Dynamicity (“Kinesis”): dynamic events score higher than states (3) Aktionsart (“Aspect”): telic events score higher than atelic events (4) Punctuality: punctual events score higher than non-punctual events (5) Affirmation: affirmation scores higher than negation (6) Mode: realis scores higher than irrealis (7) Agency: higher scores for subjects “high in potency” (8) Affectedness: higher scores for affected objects (9) Individuation: higher scores for individuated objects

A maximally transitive clause is a clause in which a human agent causes a definite change of state in a definite object in a single moment, something like The hunter shot Bambi. We could restate this as: A maximally transitive clause is a clause in which the participants are maximally asymmetrical in terms of the features listed above. Like other argument alternation constructions, passivization changes the asymmetry between the participants in an event. Verbs that express events that are symmetric in nature (such as resemble or weigh) form awkward passives, but if they are used in a semantically more transitive way, i. e. with more asymmetry between the participants, passivization is fine. (35)

a. b.

The dog weighs 20 pounds./ *Twenty pounds are weighed by dog. The handler weighed the dog./The dog was weighed by the handler.

Functions of the passive – a preliminary sketch

37

While also transitive in the traditional sense, a clause like The dog does not fear rats would score much lower on the overall Transitivity scale (it is neither dynamic nor telic nor affirmative nor agentive, and the object is not individuated). If all the properties listed above are really just facets of a complex notion of Transitivity, it can be expected that they co-vary systematically. This is captured in the Transitivity Hypothesis, which states that if two clauses differ in that clause A is higher in Transitivity according to any of the features, then it will not be lower in Transitivity in any other of the features, provided that they are marked at all (Hopper and Thompson 1980: 255). To give an example: If a language marks telicity and sentence A is telic and sentence B is atelic, and if in that language individuated objects are marked in some way, then it should not be the case that sentence A has the telicity marking, while the non-telic sentence B has the marking for individuated objects. However, it may also be the case that the properties above are related in some other way, not necessarily by virtue of expressing a facet of a more general notion of Transitivity. Take dynamicity and aktionsart: A clause that is non-dynamic (i. e. a state) cannot be telic. There is no inherent endpoint without movement or change (compare Vendler 1967; Dowty 1979). In more recent work on aspect the role of the direct object has been stressed (Verkuyl 1993; Tenny 1994): Tenny argues that the direct object “measures out” an event. If there is no measurable direct object, the event is not telic. Obviously, Hopper/Thompson’s notion of “Transitivity” bears close resemblance to the prototypical event as described by Dowty (1991). Passivization decreases the structural, semantic, and aspectual asymmetry between the participants in a clause – and thus decreases transitivity.

4. Functions of the passive – a preliminary sketch Choosing the passive over the active means breaking up the standard alignment of arguments and syntactic positions associated with active voice. Passives are costly with respect to language comprehension, as can be seen from language acquisition data. Children tend to interpret subjects as agents, which can lead to problems with interpreting passive, in particular reversible passives, such as The cow was kicked by the horse (Slobin 1966). On the other hand, children productively passivize novel verbs under appropriate discourse conditions (Pinker, Lebeaux and Frost 1987), which indicates their awareness of the pragmatic conditions under which one would choose a pas-

38

Chapter 2 The English passive and linguistic theory

sive over a less costly (in terms of sentence processing) active. What is the benefit of making a sentence more difficult to parse, or, to put it in Lambrecht’s terms: “Why should grammar provide so many different syntactic and prosodic structures for expressing one and the same propositional content?” (Lambrecht 1994: 9). This is a point on which generative grammar, which is mainly concerned with explaining the structure of passive sentences, remains silent. Since passivization affects different elements in the sentence and since it is a process that verbs of many different semantic subclasses can undergo (Pinker 1989), it is not to be expected that there will be one single core function of the passive. One potential answer to Lambrecht’s question lies outside the domain of the individual sentence, as pointed out by Biber et al. (1999: 935): “Primarily the passive serves the discourse functions of cohesion and contextual fit through ordering of information, omission of information (especially short passive), weight management (especially long management).” In the following, I will briefly outline how the syntax of the passive is interpreted in terms of arranging important and unimportant, given and new information.

4.1. Demotion of the agent Demotion of the agent means that the agent, the semantically most prominent argument (Grimshaw 1990), is not realized in the syntactically most prominent position; in fact, it need not be realized overtly at all. The passive therefore is a good choice if the identity of the agent is not known, not considered important, or if it is not to be revealed (Jespersen 1927: 167; Thompson 1987: 497). Corpus results from Biber et al. (1999: 938–941) confirm that the majority of passives are indeed short passives, i. e. they do not have by-phrases. For example, in (36a) the identity of the person who listens to the concerns of the generic subject, people, does not matter, therefore, the agent need not be expressed. In (36b) the situation is different: This sentence is taken from an abstract that precedes an article in a scholarly journal. In this genre, it is important to link an idea to a specific author. The impersonal construction It is argued . . . seems to be a violation of that requirement, but in the context of the abstract it is clear that the agents of argue are the authors of the paper. Compare this to the short passive in the subordinate clause: Here, the adverb standardly is added to rule out the interpretation that the implicit agent of ascribed is constructed as coreferent with the authors of the article. The authors put forward a specific argument and point out the newness of the argument. One can see that the motivations for keep-

Functions of the passive – a preliminary sketch

39

ing the agent implicit can be very different: In (36a) the exact identity of the agent does not matter, in the first passive in (36b) the identity of the agent can be inferred through familiarity with genre conventions, and in the second passive in (36b) the generic character of the agent is expressed through another linguistic element, the adverb standardly. (36)

a.

b.

“People want to be heard,” Mr. Gray says, “and when they realize they are being heard, they are being empowered and will be more willing to find solutions.” (FROWN; A13) This paper summarises his account of authority in educational relationships. . . . Next, the paper explores the rather neglected topic of parental authority. . . . It is argued that the authority of the parent is in important respects different from the authority standardly ascribed to the teacher. (JME 21)19

By contrast, in the following examples the identity of the agent is important and it cannot be inferred from context. (37)

a.

b.

Even Boris Yeltsin himself has yet to cooperate with giving up the president’s special files. These archives were created by dictator Joseph Stalin who assiduously gathered compromising material on his opposition and colleagues. (FROWN; A08) Not that the images don’t have real effects: a homemade video of a black motorist being beaten by police succeeded in burning down a sizable part of Los Angeles. (FROWN; A12)

What led to outrage and riots in Los Angeles in 1992 was the fact that the “black motorist,” Rodney King, was beaten not just by anybody, but by police officers. Similarly, in (37a), the fact that the archive in question was created by Joseph Stalin himself is what makes the files interesting enough for Boris Yeltsin to insist on keeping control over them. The identity of the archive creator is also truly non-derivable information, which makes the byphrase the focus of the sentence (Lambrecht 1994: 207). According to Biber et al. (1999: 941) more than 90 % of agent phrases in the passive bring in new information. The structural demotion of the agent from subject to adjunct can thus also be reframed as an informational upgrade to the focus of the sentence. The right periphery of the sentence is not just a common position to accommodate new information, it is also the preferred position for “heavy” 19. JME = Journal of Moral Education. More data from abstracts will be discussed in chapter 5.

40

Chapter 2 The English passive and linguistic theory

phrases (principle of end-weight). In (37a) the by-phrase includes a nonrestrictive relative clause, while the subject, the NP the archives, only consists of a determiner and the head noun. Findings from large-scale corpus studies confirm that there is a “clear tendency” for the by-phrase to be longer (and more complex) than the subject (Biber et al. 1999: 940–941). Again, the structural demotion of an argument goes hand in hand with an “upgrade” on another level (structural complexity) 4.2. Promotion of the theme A consequence of the agent’s demotion to an adjunct or to an implicit argument is that another argument can take the position of the subject. This has an effect on the informational structure of the sentence: In English, the grammatical subject is typically the topic of the clause (Quirk et al. 1985: 726),20 in the sense of “the entity that the sentence is about” (Huddleston and Pullum 2002: 1371). It is part of the “given” or “old information,” which Lambrecht (1994: 50), in his speaker-oriented concept of topichood, defines as “the sum of ‘knowledge’ . . . evoked in a sentence which a speaker assumes to be already available in the hearer’s mind at the time of utterance.”21 The linking rules in English are such that normally an agent will end up in the subject position (Wanner 1999; Levin and Rappaport Hovav 2005). If a non-agent is to be the topic of the sentence, one needs to rearrange the sentence structure. English has a number of constructions that allow for a non-agent to be the topic of a clause or, to speak more generally, to occur in a position preceding the agent. For example, a constituent can be moved into a non-argument position at the beginning of the sentence (topicalization),22 as in (38a), or a constituent can be explicitly marked as a topic, as in (38b), or the verb’s argument structure can be manipulated in such a way that the agent need not be expressed explicitly anymore, which is what happens in the passive, as in (38c). By allowing for a non-agent to be realized in the unmarked topic position, the passive “may facilitate the connexion of one sentence with another” (Jespersen 1927: 168). 20. Quirk et al. actually prefer the term “theme,” but in order to avoid confusion with the semantic role, I will use the term “topic” here. 21. Lambrecht uses the term “pragmatic presupposition” for this kind of knowledge assumed by the speaker. 22. Huddleston and Pullum (2002: 1374, Fn. 5) reject the term “topicalization,” because they do not accept the implication “that the (characteristic) function of the construction is to accord topic status to the preposed complement.”

Functions of the passive – a preliminary sketch

(38)

a.

b. c.

41

A: Did you buy a whole new wardrobe for school? B: No, I have lots of clothes. Most of my stuff my mom gets at Alexander’s. (from Huddleston and Pullum, 2002) As for external funding, Smith has a grant application pending. (from Huddleston and Pullum, 2002) I need your essays by Friday. Late submissions will not be accepted.

Note that it is the syntactic category of the complement of the verb that determines whether or not movement has to take place. Movement in the passive is obligatory if the complement is an NP. If the verb has a clausal complement, the subject in the passive can be filled with a placeholder element. In terms of generative grammar, (39b) is ungrammatical because the NP following the passive participle does not receive Case. By contrast, clauses are not subject to the Case Filter, which means that a clausal complement is not affected by the passivization of the verb – there can be passivization without movement, as in (39a). (39)

a. It was explained why they had left early. b. *It was explained the reason for their early departure.

Since topics are not morphologically marked in English, it can be difficult to pinpoint what exactly makes something the topic of a sentence. Minimally, the topic does not receive focal stress, as the focus of the sentence is “the portion of a proposition which cannot be taken for granted at the time of speech” (Lambrecht 1994: 207). Syntactically, since topics express given information, they can be definite and are often realized as pronouns. Although complex subjects are not uncommon in passives, Biber et al. (1999: 940) found that “there is a clear tendency for the subject to be shorter than the agent phrase” in passives with by-phrases. This is in line with the assumption that the NP in the subject position expresses given information. Results from quantitative studies also confirm that for long passives (i. e. passives with by-phrases) there is a clear tendency for the subject to have a “higher level of givenness” (Biber et al. 1999: 941) than the agent phrase. A more structural effect of moving the object to the subject position is that the moved phrase can undergo syntactic processes that target the subject position. Shibatani (1985: 830) refers to the passive a “a syntactic pivot” for constructions that depend on co-reference between two subjects:

42

(40)

Chapter 2 The English passive and linguistic theory

John came in and was seen by Mary. (Dixon and Aikhenvald 1997) b. *John came in and Mary saw.

a.

Here we are dealing with the coordination of two VPs, came in and was seen by Mary. Both clauses share the subject John. Without passivization, coordination would not be possible. One cannot interpret (40b) as John coming in and Mary seeing him, because the shared constituent would be the subject of the first clause and the object of the second. Similarly, in (41a) the subject of the matrix clause, you, is co-referent with the subject of the infinitive, which allows for a control construction with a non-overt PRO subject (as in (41a’)). Without passivization, the two clauses cannot be connected in this way. (41b) is ungrammatical in the given reading that the subject of the matrix clause, you, is co-referent with the object of the preposition to in the subordinate clause.23 (41)

I have been talking incessantly all night, and with nothing to say. But with you, Fanny, there may be peace. You will not want to be talked to. (from Jane Austen, Mansfield Park) a’ Youi will not want [PROi to be talked to ti ] b. *Youi will not want [someone to talk to PROi ]

a.

This is a clear difference (among others) between passivization and topicalization: While the latter goes to a position that precedes the subject and is entirely optional, movement in the passive is more solidly anchored syntactically – in most instances, it is obligatory, and it always goes to a position that has to be filled, the subject position.

5. The passive in language acquisition theory According to the Principles and Parameters model of grammar, language acquisition can be described as a process of selecting between options provided by Universal Grammar (Guasti 2002), based on positive evidence, i. e. utterances that children are exposed to. Data from language acquisition are an important factor in evaluating hypotheses about linguistic structures: If 23. PRO is a non-overt element that can only occur in ungoverned positions (PROTheorem, Chomsky 1981). The subject position of an infinitive clause is not governed, but the object of a preposition is governed by the preposition. Therefore, there can be no PRO in the position after the preposition in (41b).

The passive in language acquisition theory

43

children’s acquisition of syntax is guided by UG principles, their emerging grammars should be rule-governed systems. In early language acquisition research in the generative framework, the passive received attention mainly because acquisition of the passive could indicate when children acquire NP-movement, thereby demonstrating knowledge of the principles that motivate NP-movement. More recently, the interest has shifted more towards semantic properties of the passive. Do data from language acquisition support the implicit argument hypothesis in any way? The acquisition of the passive is gradual. Early studies of the acquisition of the passive show that children struggle with passives in general and in particular with passives based on reversible actives (Turner and Rommetveit 1967). (42a) is reversible in that the roles of subject and object could be reversed (the horse could also have kicked the cow), (42b), on the other hand, is not reversible (apples do not eat boys). (42)

a. b.

The horse was kicked by the cow. The apple was eaten by the boy.

Turner and Rommetveit (1967: 657) found that young children tend to interpret sentences like (42a) as the horse kicking the cow. They conclude that children essentially process passives as actives, “treating the passive markers is, -ed, and by as if they are signs of uncommon tense.” Others argued that it is the aspectual class of the passivized verb that mattered: Horgan (1978) carried out a study in which she tried to elicit passives by asking children to describe a picture in which one participant (not the agent) was highlighted. The assumption was that the highlighted entity would become the topic of the sentence. The situations depicted were a mix of situations resulting from reversible and irreversible events, as well as of states. It turned out that children younger than five did not use passives with by-phrases at all and did not generally put action verbs in the passive. Sentences like Tree is broken were used as descriptions of states, i. e. as adjectival passives. This was interpreted as a sign that children acquire the adjectival passive before the verbal passive and that children do not acquire the verbal passive before the age of five. However, the preference for the adjectival passive may be due to the setup of the experiment, since children were asked to describe pictures, not movies or scenes that were enacted. To reconcile findings like these with the idea of Universal Grammar, Borer and Wexler (1987) proposed the “Maturational Hypothesis”: Syntactic principles of Universal Grammar “mature” and become available only after certain cognitive de-

44

Chapter 2 The English passive and linguistic theory

velopments have taken place. Specifically, they assumed that children under a certain age cannot not form A-chains, which result from movement of an argument, yet and hence cannot comprehend or produce syntactic passives. By contrast, they can form adjectival passives such as Tree is broken, since this construction does not require syntactic movement (and consequently no A-chain). There are, however, empirical and theoretical problems with this approach. The theoretical problem is that in more recent versions of syntactic theory in the Chomskyan framework, it is assumed that, more or less, all sentences involve A-movement of an NP to the subject position (Haegeman 2006). According to the VP-internal subject hypothesis, the subject NP is generated in the specifier of the verb phrase (VP), as an argument of the verb, and then moves to Spec, IP, the canonical subject position (specifier of inflection phrase), in order to fulfill the Case Filter and the Extended Projection Principle (see chapter 3). A child that does not have an understanding of A-chains should not be able to form sentences in which the subject is in the Spec, IP position. However, children at the age of five clearly produce sentences with subjects in the correct position. In a revision of the Maturational Hypothesis, Borer and Wexler (1992) argue that not all A-chains are a problem for children, but only those that connect two positions that are potential theta-positions, essentially the subject and object position inside the VP. In the case of the passive, an NP moves from the object position (where it receives its theta role) to Spec,VP (a theta position in principle, but not a theta position in the passive, due to the absorption of the agent theta role) and then to Spec, IP (as a specifier of a functional projection, not a theta position under any circumstance). It is the first step that makes the passive different from A-movement in regular active clauses. If this analysis is correct, children should also struggle with other instances of A-Movement from object to subject position, however, they do not. Guasti (2002: 257–260) discusses data from the CHILDES database that show that children under the age of five produce adult-like sentences with unaccusative verbs,24 i. e. they move the object to the subject position via the Spec,VP position, thus creating an A-chain between argument positions. Data from Italian, a language in which unaccusative and intransitive verbs select different auxiliaries to form the perfect (Burzio 1986), confirm that children at the age of three have no trouble handling unaccusative verbs in a targetlike manner (Guasti 24. However, they also produce clauses in which the internal argument of unaccusative verbs occurs in postverbal position. This is generally regarded as evidence for a syntactic approach to unaccusativity.

The passive in language acquisition theory

45

2002). A second objection against the Maturational Hypothesis is based on data from language production. Pinker et al. (1987) show that the generalization that early passives are not syntactic passives with NP-Movement is not correct. From the age of three, children produce spontaneous passives, often using incorrect participle forms or non-existent transitive verbs, which indicates that they are not just repeating what somebody said to them. The passives below may not be adult-like in their form, but they are clearly syntactic passives, with a noun phrase that is not an agent (and also not necessarily a prototypical patient, see (43c)) and an implicit agent argument. Note also that the first example is a get-passive, which indicates that be is not treated as an obligatory passive marker. (43)

a. b. c.

I don’t want the bird to get eated. (Adam, 3;3) How could it go up if its not . . . if it’s not flyed? (Adam 3;10) Both of these things can be put things in. (CF, 3;3)

These data show that children are not conservative learners when it comes to the passive. Just as they use verbs transitively that cannot be used transitively in English (He’s gonna die you, David!), they use passives that would be ungrammatical in adult English. In elicited production and comprehension experiments, children demonstrated that they easily form and understand passives based on transitive nonce-verbs, even if the verbs in question were not action verbs (Pinker, Lebeaux and Frost 1987). In the learning phase of the first experiment, four one-syllable nonce-verbs, two of them agent-patient verbs and two of them experiencer-theme verbs (specifically, “to rub the back of the neck of,” “to jump on top of and then over,” “to see through a binoculars-like instrument,” and “to hear through an eartrumpet-like instrument”) were taught to 16 four-year-olds through the description of a scene that involved toys. The four verbs were used either in the active or the passive (The dog is pelling the elephant or The elephant is being pelled by the dog). In the testing phase, children were put in a discourse context that facilitated the use of the passive (“What is happening to the kitten?”) or the active (“What is the kitten doing?”). The most important result was that children freely produced passives of verbs that they had only heard in the active, at a rate of 59 % for the action verbs and of 69 % for state verbs, thus showing that they had a productive rule for forming passives. In a comprehension experiment, the experimenter showed the children two toys and asked them to act out a scene that was described either in the active or in the passive (“Can you make it so that the doggy is being kaled by the elephant?”). The comprehension rate was very high: 94 %

46

Chapter 2 The English passive and linguistic theory

for the action verbs that children had been taught in the active, 100 % for the state verbs children had been taught in the active (Pinker, Lebeaux and Frost 1987: 215). The authors conclude that four-year-old children possess a productive rule of passivization that is not restricted to verbs that have prototypical agent and patient arguments.25 This does not mean that children’s use of the passive is not restricted at all. In another experiment, children aged five to eight were exposed to “non-canonical verbs,” i. e. verbs whose argument realization patterns were a violation of argument mapping principles (linking rules) in English. For example, The dog floosed the giraffe meant that a giraffe leapfrogged a dog. Normally, if a verb has an agent argument, it will be projected as an external argument and hence as the subject of the sentence (Grimshaw 1990), but in these non-canonical realizations the agent ends up as the object. These are not naturally occurring verbs in English. It was not easy to elicit these verbs in the passive when they had been taught in the active. The proportion of successful elicitations for a group of sixteen five- to eight-year olds was 28 % (compared to a success rate of 75 % if the verb had been taught in the passive). Pinker et al. conclude that “we have reliable evidence that children’s rule of passivization is sensitive to the verb’s assignment of grammatical functions to thematic roles: the by-object must be more agentive than the subject” (Pinker, Lebeaux and Frost 1987: 228), as described in Jackendoff ’s Thematic Hierarchy Condition (Jackendoff 1972). If this is correct, it cannot be claimed that children’s rule for forming passives lacks a role for the by-phrase (and consequently, for the implicit argument); quite the opposite: “the presence of by is the most important surface cue used by children to give a string a passive interpretation” (Pinker, Lebeaux and Frost 1987: 262).

6. Summary In this chapter, I presented insights from syntactic, functional, and cognitive approaches to the passive, and in the tradition of generative grammar set25. Tomasello (2003: 128) points out that “the same is not true for younger children.” He refers to an experiment in which two- to three-year-old children were exposed to novel intransitive action verbs (such as “tam” for an action like “roll”). The children were then presented with a situation in which the verb would have to be used transitively to describe the action (such as a dog making a toy car roll over). However, children did not productively transitivize these novel verbs (while they did not have any problems using novel transitive verbs correctly).

Summary

47

tled for a definition of the passive that depends on the characteristics of the passive morpheme. The common denominator for all passives is that they include the passive participle, whose syntactic features are systematically different from that of the base verb. While passivization has effects on word order and the realization of arguments, it does not change the propositional content of the clause. This means that all arguments are still represented, if not necessarily overtly. Chapter 4 will discuss the syntactic representation of the implicit agent in the passive in detail. Throughout this book I refer to the passive as a “construction” to highlight the idea that the syntax of the passive is compositional in nature. The syntax of the passive and the implicit agent reading largely follow from the changes caused by the attachment of the passive morpheme to the lexical verb, which is like the first stone in a domino chain that can take different shapes. While there is a common morphosyntactic denominator for the passive, the overall look of the passive in English can be very different – there are passives with and without movement, with and without auxiliaries, and with and without overt agent phrases. “Deconstructing” the passive means taking the construction apart and looking at its components, which do not always co-occur in the passive and some of which also occur in other argument alternations. This will bring out both the specific characteristics of the passive and its place in the larger context of “voice” as a system of mapping event participants to syntactic positions. My view of the passive is that its form is shaped, to a large extent, by autonomous syntactic principles, but that in order to understand its place in grammar one needs to look beyond the domain of the individual sentence. This approach includes considerations of the communicative purpose of the passive as well as its place in a spectrum of transitivity alternations in a language with an active-voice linking system. Because of the multicomponent nature of the passive, it is not possible to single out a single function of the passive as its core function. We saw that passives can be used to background an agent (by making it implicit) as well as to focus an agent (by realizing it as a heavy phrase at the right periphery of sentence), and to make a nonagent the topic of the clause. Passives may be costly in terms of sentence comprehension, but these costs are balanced out by a better fit between informational or discourse status and syntactic structure. Children as young as three produce passives spontaneously, though not very often. The low frequency of passives does not come as a surprise, since passives are quite rare in informal speech and hence in the linguistic input of children. In addition, it seems that if there is an agent in an event, children have a strong preference for making the agent the topic of the sentence, even

48

Chapter 2 The English passive and linguistic theory

if their attention is steered subtly towards another participant. However, a number of experimental studies have shown that children can produce and interpret passives well before the age of five. “[T]here is no absolute restriction in child speech against producing and comprehending nonactional passives, against producing the by-phrase, or against using passives with an eventive interpretation” (O’Grady 1997: 265). Children do best with prototypical transitive verbs that they have heard before in the passive, but they also productively passivize verbs they have never heard in the passive before, including state verbs. They are less successful at passivizing verbs that would not naturally occur in English, specifically verbs whose distribution of arguments is already passive-like (with an affected object in the subject position and an agent in a postverbal position). This all indicates that very early on children have an understanding of the detransitivizing function of the passive, they may just be more interested in talking about agents than patients or experiencers. In order to understand the choices that speakers make, we need to understand the options that are available to them. Chapter 3 will discuss the ingredients of the passive in English in detail. An element that is not normally considered at the center of the passive construction is the implicit agent. Its syntactic representation will be the subject of chapter 4. I will argue that the implicit agent is not just inferred from context and that its realization lies at the core of the passive construction. Chapter 5 presents a case study on the use of the passive and related constructions in a corpus of scientific writing, in which I examine the passive and its competitors with regard to the expression of agenthood.

Chapter 3 Ingredients of the English passive

1. Introduction It has been argued in chapter 2 that the passive is a construction whose syntactic and semantic properties can to a large extent be predicted from those of the verb, in conjunction with general principles of syntactic structure, such as the Case Filter and the Extended Projection Principle. One of the reasons against a Construction Grammar approach for the passive is that the passive has so many different looks that it would be difficult to come up with a surface generalization. As pointed out in the previous chapter, there are passives with and without thematic subjects, with and without by-phrases, and with and without auxiliary verbs. One could, of course, say that these are all different constructions, but then one would miss important generalizations about the passive, in particular the presence of an implicit argument. In this chapter, I will show that the various surface forms of the passive are a result of the interaction of the ingredients of the passive with general principles of syntactic structure. In Construction Grammar, the categories of a construction are defined by the construction itself. It is not assumed that one and the same verb can occur in the active or the passive, but rather, that the passive sentence contains a form of a lexical verb and that the regular transitive construction contains a form of a lexical verb, and these are not necessarily two versions of the same category – “they could be named ‘Rosencrantz’ and ‘Guildenstern’ as far as syntactic theory is concerned” (Croft 2001: 55). Establishing a category, “verb,” would constitute creating an “external meronomic link” (Croft 2001: 54) between two constructions, and this kind of link is rejected in Radical Construction Grammar. By contrast, in mainstream generative grammar the assumption that one and the same verb can be used in different constructions is still upheld. One construction may be derived from another (via movement, for example), or two constructions may be derived from one source (the lexical entry of a verb, for example). The assumption is that properties of the construction can be predicted from properties of the verb. This is captured in the term “argument alternation” (Levin 1993). In the non-reductionist model of Radical Construction Grammar, no independent category “verb” exists. Even individual words are not independent

50

Chapter 3 Ingredients of the English passive

of constructions, since word meaning may vary depending on the construction within which they occur. In her overview of verb classes and verb alternations, Levin (1993) discusses dozens of alternations that involve the arguments of a verb, among them the passive, the dative alternation, and the cognate object construction. Goldberg (2002) criticizes the alternation approach and argues that it is not appropriate to relate two constructions that are at best rough paraphrases of each other in such a way that they are derived from each other or from a common source. A case in point is the dative alternation (Goldberg uses the term “ditransitive construction”), as illustrated in the following examples, modeled after Goldberg (2002). (44)

a. b. c. d.

Mina bought a book/it for Mel. Mina bought Mel a book/??it. Mina baked a cake/it for Mel. Mina baked Mel a cake/??it.

A verb-based approach would put (44a) and (44b) together, since they are both ways to realize the arguments of the verb buy. The assumption is that the syntactic behavior of a verb (and hence its ability to participate in argument alternations) is determined by the verb’s semantics. The challenge is to represent the verb’s meaning in such a way that the relevant meaning components can be isolated: “If the hypothesis that syntactic properties are semantically determined is taken seriously, then the task is to determine, first to what extent the meaning of a verb determines its syntactic behavior, and second, to the extent that syntactic behavior is predictable, what components of verb meaning figure in the relevant generalizations” (Levin 1993: 14). By contrast, in Construction Grammar, one would group (a) with (c) and (b) with (d), since there are subtle meaning as well as distributional differences between (a) and (b) and between (c) and (d). For example, the direct object can be a pronoun in (a) and (c), but not in (b) and (d). There are also semantic differences between (a) and (b): While (a) can mean that Mina did Mel a favor by buying a book for an unknown recipient, (b) can only mean that Mina bought a book with the intention of giving it to Mel (Goldberg, 2002: 331). “There is little empirical motivation to decree that ditransitives must be derived from prepositional paraphrases. . . . The robust generalizations are surface generalizations” (Goldberg 2002: 333). In a verb-based alternation approach, one would not necessarily derive (b) from (a) and (d) from (c) – although there are many analyses of this

Introduction

51

kind (cf. Larson 1988) – but instead, (a) and (b) would be traced back to a common source. In the first part of her compendium on English verbs, Levin (1993) takes alternation patterns as her starting point and lists the verbs that allow them. In the second part of the book, she focuses on (semantically based) verb classes and discusses which alternations verbs that belong to a particular class allow. “Verbs of cutting,” for example, can occur in the conative alternation (Carol cut the bread/Carol cut at the bread). The underlying assumption is that the verb as a syntactically relevant category exists and that predictions about its participation in certain constructions can be made based on the syntax and semantics of the verb. This is the position that I will take here: The core of the passive is the verb in its participle form, and the syntax of the passive construction is determined, to a large extent, by the syntactic and semantic properties of the verb. This chapter will discuss the role of the verb in the passive construction and the contributions of other elements that are characteristic of the construction (but that do not necessarily occur in every instance of the construction), including the subject, the auxiliary, and the by-phrase. I will first examine which verbs can occur in the passive and will then look at each ingredient in detail. To further explore the compositional character of the passive, I will then reexamine the get-passive, which is often considered a stylistic variant of the be-passive. Data will come mainly from the FROWN (Freiburg-Brown) corpus of written American English, which contains 500 text samples from 15 different genres of written American English, e. g. newspaper reporting, scientific writing, and general fiction and for which data were collected in the early 1990s.26

26. Unless mentioned otherwise, FROWN data in this chapter are based on 88 text samples (each about 500 words) from the categories A-C (press: reportage, editorials, and reviews), 51 text samples from the categories D-E (prose: religion, and skills and hobbies), and 80 text samples from the category J (science writing). These categories comprise the genres in which the passive occurs with some frequency. They also had the very practical advantage of being syntactically tagged at the point of my writing this chapter. The FROWN data in the study of the get-passive are based on all categories of the corpus (forms of a specific verb, like get, can easily be extracted without syntactic tagging).

52

Chapter 3 Ingredients of the English passive

2. Verb classes and passivizablity Not every active sentence has a corresponding passive. Although every verb forms a past participle, not every verb can occur in a passive construction. Obviously, there are grammatical constraints on passivization that not every verb meets. However, this is not the full picture. If one considers voice, in a broader sense, as a system of encoding events syntactically, one will also expect to find restrictions on the passive that have to do with the function of the passive construction and that cannot be reduced to the syntactic properties of the verb. In this section, I will give an overview of the constraints on passivization in English related to the verb. 2.1. Transitive verbs and NP-Movement Transitive verbs seem to be suited best to illustrate how passivization affects the syntactic properties of a verb. Quirk et al. (1985) use a diagram like (45) to indicate the crosswise realization of arguments that is typical of the passive. (45)

John

admired

ACTIVE SUBJECT

ACTIVE VERB

Mary ACTIVE OBJECT

PASSIVE SUBJECT PASSIVE VERB AGENT BY-PHRASE

Mary

was admired

by John

In generative grammar, the double nature of the passive subject – semantically an object, but syntactically clearly a subject – is reflected in a movement analysis. The subject NP, Mary, is generated as the complement of the verb and moves to the subject position, leaving a trace in its original position, by which it can still be identified as the internal argument of the verb. The trace occupies the complement position and cannot be filled with another lexical element (*Mary was admired the statue by John).

Verb classes and passivizablity

(46)

53

IP NP

Infl’ InflVP Spec

V’ V VP be Spec

V’ V’

PP by John

V NP admired Mary Movement ensures that Mary, as the direct object in the active, can have the same interpretation as the subject in the passive. Admire is a transitive verb, but in the passive, its direct object leaves its original position. Two grammatical principles call for this kind of movement. The first, the Extended Projection Principle, states that clauses must have subjects. If the object were not moved, the subject position would remain empty and the principle would be violated. The second principle is the Case Filter, which states that each NP must be licensed by abstract Case (see chapter 2).27 Case identifies an NP phonologically, just as a theta-role identifies it semantically. In order to get Case, the NP has to be in a specific relationship with a Caseassigner. This is referred to as “structural Case.” In classic Case Theory (Chomsky 1981), the Case assigners in English are finite INFL (the head of the clausal projection), V, and P. While transitive verbs are Case assigners, passive participles are not. Verbal elements that do not have a thematic subject typically cannot assign structural Case (and vice versa). This is known as “Burzio’s Generalization” (Burzio 1986, 2000). (47)

a. [-js] ↔ – Acc

b. [+js] ↔ + Acc

A second kind of Case is “inherent Case.” It can only be assigned to an argument with a specific theta-role (Chomsky 1986, 1995). Whether or not it can be assigned by passive participles is a matter of debate (see below). 27. For a discussion of which is the stronger force in movement (the EPP or the Case Filter) see, for example, Reuland (2000).

54

Chapter 3 Ingredients of the English passive

If the postverbal NP cannot get Case from the verb, it needs to move to another position, normally the subject position, where it receives nominative Case from finite INFL, the head of the clausal projection. This analysis is supported by passives based on so-called Exceptional Case Marking verbs and other passives involving clausal complements. ECM verbs are verbs that assign accusative Case to the subject of their complement (rather than to their own direct object).28 In (48a) her is not an argument of believe, it is assigned its theta-role by the verb win in the embedded clause. However, it receives its Case (accusative) from the verb believe. (48)

a. They believed [her to have won the competition] b. *It was believed [her to have won the competition] c. Shei was believed [ti to have won the competition]

If an ECM verb is passivized, it cannot assign accusative Case anymore and that is what requires movement of the postverbal NP to the subject position of the main clause. Note that we are dealing with movement from a subject position to a subject position here. If a verb like believe takes a finite complement clause, the subject in the complement clause does not depend on the matrix verb for Case. Consequently, passivization does not enforce movement in cases like (49a); movement even leads to ungrammaticality, as illustrated in (49b). (49)

a. It was believed [that she had won the competition] b. *Shei was believed ti had won the competition.

Case Theory helps us to understand why not all passives have subjects that correspond to the direct object of the active verb. Movement is not an integral part of the passive and it only occurs when an NP depends on the passivized verb for Case. This also means that having a direct object is not a necessary condition for passivization, at least not in English. Just as not every subject in a passive construction corresponds to a direct object in an active construction, not every direct object of a transitive verb can become the subject in a passive construction. Verbs that are syntactically, but not semantically transitive, do not easily allow for the passive in English. A semantically transitive verb would be a verb whose two arguments are semantically specified in such a way that one is clearly closer to a 28. I will follow the classic ECM analysis here, in which the accusative NP following the verb is – and remains – the subject of the subordinate clause. For an alternative analysis (subject-to-object raising), see, for example, Davies and Dubinsky (2004).

Verb classes and passivizablity

55

proto-agent and the other closer to a proto-patient (see chapter 2). Among the proto-agent properties listed by Dowty (1991) are volitional involvement in an event or state, sentience, causation, and existence independent of the event. Properties contributing to a proto-patient include a change of state and incremental themehood. A prototypical transitive event, then, is an event in which a sentient agent volitionally causes a change of state in a patient, with incremental changes in the patient reflecting the progression of the event. One could also say that a prototypically transitive event is a maximally asymmetrical event. A less asymmetrical event would be one in which a non-sentient entity causes a change of state in a sentient entity. Not all transitive verbs express events in which an agent acts on a patient. In the example in (45) the direct object is not affected by the event and the event is not dynamic, but the arguments of the verb are still semantically and aspectually asymmetrical, as the subject argument is deliberately involved in the event, while the object argument is not. There are, however, verbs which are transitive in the narrow technical, or syntactic, sense, but which do not allow for the passive, among them verbs of measurement and verbs expressing symmetric relations: (50)

a. You resemble your brother. b. *Your brother is resembled by you.

For passivization to be felicitous and meaningful, semantic criteria have to be considered. Passivization makes the syntactic expression of a state or event less asymmetrical (by demoting the more agent-like argument to an implicit argument and, if syntactically required, elevating the more patientlike argument to subject status), and it only seems to work if there is some asymmetry to begin with. While state verbs in general can be passivized, verbs that express symmetrical relationships cannot. This fits with the observations that: (i) young children produce and comprehend passives based on actions better than passives based on states (Maratsos et al. 1985), even though, under appropriate discourse conditions, they do form passives based on state verbs, including novel ones; and that (ii) they have difficulty passivizing verbs in which the realization of agent and patient is noncanonical, i. e. less asymmetrical than expected (Pinker, Lebeaux and Frost 1987, see also chapter 2).

56

Chapter 3 Ingredients of the English passive

2.2. Ditransitive verbs and recipient passives Ditransitive verbs are verbs that can take two objects: a theme (the direct object) and a goal or beneficiary (the indirect object). They include verbs of giving and sending, future having (offer, promise), communicating (tell, ask, write), obtaining and creation (bake, make, get, buy), cf. Pinker (1989). If both objects are realized as bare NPs, this is known as a double object construction (Iwakura 1987; Larson 1988) or dative construction (Oehrle 1976; Czepluch 1982) or ditransitive construction (Goldberg 1995). (51)

a. b.

I offered him a drink./I offered a drink to him. I baked him a cake./I baked a cake for him.

The double object construction, in which both objects are realized as bare NPs, is more restricted than the V-NP-PP pattern, both with respect to the verbs and the NPs that can occur in the constructions. Goldberg (1995) points out that the two are not equivalent in meaning either and therefore considers them two different constructions. For instance, while I baked a cake for him can mean that I did it as a favor (so that he would not have to do it), I baked him a cake is restricted to the meaning that the subject made a cake with the intention to give it to another person (expressed as indirect object). In Goldberg’s framework, intended transfer is part of the meaning of the ditransitive construction (V-NP-NP). In generative grammar the two argument realization patterns are usually considered to be related to each other. The focus of the attention is once more on Case: If a verb can assign Case only once, how can a verb have two accusative complements? A number of suggestions have been made – either the verb can assign one additional Case, a so-called inherent Case (Chomsky 1981), which is tied to an argument with a specific theta-role, or there is an underlying preposition that assigns regular structural Case to the indirect object (Czepluch 1982), or there is an abstract verb involved, which projects a second VP (or rather, vp), allowing for one more Case position for an NP (Larson 1988). Larson’s shell analysis, in which he derives both word order patterns from one underlying syntactic structure via movement, has become the standard for the double object construction in the generative paradigm. Neglecting the details here, let us just note that both NPs are assigned structural Case. If there are two objects and two Cases, one might expect that each of these Cases can be absorbed in the passive, resulting in two different passives, one with the theme and one with the goal or beneficiary in the subject position. This expectation is borne out, see the data

Verb classes and passivizablity

57

in (52a) and (b). However, if the direct object is moved, in most dialects of English, the indirect object cannot stay behind as a bare NP, see (52c). (52)

a. b. c.

John was offered a drink. A drink was offered to John. ??A drink was offered John.

The recipient or indirect passive in (52a) is a more recent construction in English than the passive in (52b). The first “reasonably clear” recipient passives of verbs like give and tell occurred in the late fourteenth century and the construction remained rare until the late fifteenth century (Denison 1993: 110). One of the developments that led to the grammaticality of sentences like (52a) was the loss of the dative in English (Jespersen 1927). If the two complements of a three-argument verb are not distinguished by morphological case, both of them may be eligible to be the subject in a passive construction. Or, to put it in terms of generative grammar, if each of the two objects bears objective Case, each of these Cases can be absorbed in the passive, and the NP whose Case has been absorbed has to move to the subject position in order to get nominative Case. However, the comparison of (52a) and (c) shows that the two objects do not behave totally alike. First, they come in a particular order: In the active V-NP-NP pattern, the indirect object precedes the direct object (*They offered a drink John). Second, as illustrated above, if the direct object is moved, the indirect object cannot stay behind as a bare NP in most dialects of English. Jaeggli (1986: 596) argues that the two Cases assigned to the two objects are actually different: While they may both be morphologically identical (i. e. accusative case), one may be a structural Case and the other an inherent Case. Inherent Case is assigned to arguments with a specific theta-role and can also be assigned by unaccusatives, including passive participles (Belletti 1988). Somewhat counterintuitively, if one assumes that the direct object is assigned inherent Case and the indirect object is assigned structural Case, the word order regularities in passives based on ditransitive verbs can be explained. If structural Case is absorbed through passivization, only the theme argument (the one with inherent Case) can occur after the participle. The goal or beneficiary argument depends on structural Case and is not licensed after an unaccusative verb. For alternatives to this approach see, for example, Larson (1988).

58

Chapter 3 Ingredients of the English passive

2.3. Intransitive verbs and impersonal passives Intransitive verbs do not form passives in English. As a result, English does not have true impersonal passives (passives in which no argument of the verb is projected), unlike Dutch or German. (53)

Es wurde gelacht, getanzt, gesungen. It became laughed, danced, sung. ’There was laughing, dancing, singing’ b. *It was laughed, danced, sung.

a.

Impersonal passives are usually taken as evidence that passivization is more about the syntactic “demotion” of the external argument than about the “promotion” of the internal argument: Obviously, there are passives in which no theme is “promoted” to the position of the subject, while there are no passives in which the agent is not “demoted” from its status as the subject of the sentence. In chapter 4, it is argued that a process affecting the agent is indeed at the core of the passive construction. However, if the passive is all about “demotion” of an agent, why are sentences like (53b) ungrammatical? Closest to the impersonal passive are constructions like the following: (54)

a. b.

There was laughing, dancing, singing. It was explained why we could not leave early.

Of those two, only (54b) is a passive construction – clearly the sentence in (54a) does not show passive morphology and one could not insert adverbs like deliberately. (54b) is a passive construction without movement and with a non-thematic subject. Movement is not necessary because there is no NP that depends on Case from explained, as the verb occurs with a finite clause as its complement here. In other words, English has passives that illustrate that the passive is not always about the “promotion” of an argument. It does not really matter if one calls (54b) an impersonal passive (because of the dummy subject) or not – the question that one has to deal with is: If the subject position in passives can be filled with a “dummy element” (Perlmutter 1978), why are sentences like (53b) ungrammatical? There is a comparatively easy answer for a subtype of intransitive verbs, unaccusative verbs, such as arrive or wilt, and a more complicated answer for the one-argument verbs in (53b), unergative verbs.

Verb classes and passivizablity

(55)

59

a. *It was arrived early by all guests. b. *It was wilted quickly.

First, the easy answer: By definition, unaccusative verbs do not have an external argument (Burzio 1986; Levin and Rappaport Hovav 1995). Their subject argument starts out as an underlying object (semantically, it is closer to a proto-patient than to a proto-agent) and is moved to the subject position in order to be assigned Case. Since unaccusative verbs do not have an external argument, they cannot be the input for a lexical rule that specifies that an external argument be suppressed, nor could they, in an alternative approch, fulfill the lexical requirement of the passive morpheme that it be assigned an external theta-role. If suppression of an external argument is the defining property of passivization, as argued in chapter 4, it is only to be expected that unaccusative verbs cannot be passivized, neither in English nor in languages that have impersonal passives, a generalization first put forward by Perlmutter (1978). There is nothing special about unaccusative verbs not forming a passive in English. Therefore, if one is talking about the non-existence of the impersonal passive in English, one really refers to intransitive verbs of the unergative type, such as smile or dance. These verbs do have an external argument and they can be passivized in languages closely related to English, such as German or Dutch (Zaenen 1993), but in English they can only be passivized if they are used transitively, for example with a cognate object. The result of this kind of passivization is not really an impersonal passive, of course, but it shows that the constraint does not lie in the semantics of the verbs in question. (56)

Smiles were smiled, dances were danced, winks were winked and a good time was had by all.

In the following pages, I will review several proposals that attempt to explain why intransitive verbs of the unergative type cannot be passivized in English. 2.3.1. It’s all about “be” Using the framework of cognitive grammar, Cornelis (1996) argues that English does not have impersonal passives because of properties of the auxiliary, be. Her argument runs like this: In English, the passive is ambiguous between the adjectival and the verbal reading. Because of this ambiguity, the agent (Cornelis refers to it as the “causer”) is not always part of the “ground” (the representation of the speech event and its participants,

60

Chapter 3 Ingredients of the English passive

including the “conceptualizer,” usually the speaker), while the “trajector” (the participant whose state or location is asserted or followed through) is. Since the causer is not necessarily evoked in all passives (due to the possibility of an adjectival passive reading), the meaning of the passive in English seems to be tied more strongly to the trajector: “the passive designates that the patient’s point-of-view should be taken, which means there has to be a patient/trajector” (Cornelis 1996:). Unergative verbs have no patient arguments and hence cannot form a passive. By contrast, in Dutch, the passive auxiliary, worden (become), is a processual verb, which means that the agent (causer) is always evoked. In Dutch, therefore, the meaning of the passive construction is not necessarily tied to the presence of a trajector, hence impersonal passives are possible. Cornelis does not take syntactic considerations into account. If the passive is all about the viewpoint of the patient, one wonders why unaccusative verbs cannot be passivized – after all, they have patient arguments. Furthermore, since there are passives without thematic subjects in English (It was explained . . . ) – Cornelis (1996: 262) refers to them as passives with “very meagre patients” – the statement that the passive in English is always about the viewpoint of the trajector seems to be too far-reaching. A thematic/referential subject is not obligatory in English, neither in the active, nor in the passive. 2.3.2. It’s all about Case Jaeggli (1986) attempts a Case-oriented approach to the question. In his view, it is not at all surprising that a language does not allow intransitive verbs to form passives – they do not have objects, therefore they do not assign Case, and if there is no Case, then no Case can be absorbed by the passive morpheme. If they are used transitively, they do assign Case, which can then be absorbed in the passive, resulting in sentences like (56). What he has to explain, then, is that some languages do have impersonal passives. After all, it is the defining property of intransitive (unergative) verbs that they do not have objects. He stipulates that in languages like German and Dutch unergative verbs are structural Case assigners, but in English they are not. Since the passive morpheme in his analysis has to “absorb” (be assigned) structural Case, it can only attach to a verb that has such a Case to assign. That rules out passives of unergatives in English, but in a rather arbitrary way. It has been shown that unergative verbs in English assign Case if it is needed, as is evidenced by cognate objects or resultative constructions (He smiled a big smile/He smiled his way into the committee). So why can they not

Verb classes and passivizablity

61

assign Case when it is needed by the passive morpheme? Also, one would expect that the same argument would hold for transitive verbs in English: If they were Case assigners only when they occur with an NP complement, one would expect that these verbs could not be passivized when they occur with a clausal complement, a hypothesis that is easily refuted by data such as (54b). The ability to assign Case does obviously not depend on the actual presence of a postverbal NP. If Burzio’s Generalization is correct, then every verb that has an external argument should be able to assign structural Case to an element that needs it, be it an NP or the passive morpheme. Another Case-related analysis is presented by Roberts (1985). Like Jaeggli, he considers the situation in English to be the default case: The verb has no object, hence the verb has no Case to assign, hence no Case can be absorbed, hence no passive can be formed. He differs with regard to his answer to the question of what is different in German and Dutch and builds his answer on passives based on transitive verbs that assign dative case to their complement. As in Old English (Denison 1993: 104), dative case in German is retained under passivization, which raises the question of which case is absorbed by the passive morpheme. It cannot be accusative Case, since the participle does not assign this Case. (57) Ihnen wurde geholfen. They-DAT became (3rd Sg. Past) helped ‘They were helped’ Roberts argues that the dative is an inherent Case in German. Inherent Case is assigned by specific verbs to arguments with specific theta-roles (Chomsky 1986). And indeed, the arguments that are marked with dative case, generally fall in the category of recipients, goals, and beneficiaries. If inherent Case cannot be absorbed by the passive morpheme in German, the morpheme must absorb another, a structural Case. This Case is the nominative, which Roberts (1987: 282 ) argues can be assigned inside the VP (and hence absorbed by the passive morpheme) in German. He gives the example in (58), in which the NP marked with nominative case (semantically, the theme or source of emotion) seems to be in the position of the direct object. (58) . . . dass meinem Bruder deine Musik nicht gefällt. . . . that my brother (DAT) your music (NOM) not likes29 ‘. . . that my brother doesn’t like your music’ 29. This is the glossing that Roberts (1987: 283, Ex. 49) gives. Actually, gefallen corresponds to please rather than to like: It is one of those psych verbs in which the theme gets nominative Case. The equivalent of like is mögen, and the experiencer argument receives nominative Case (My brother-NOM likes your music-ACC).

62

Chapter 3 Ingredients of the English passive

According to Roberts’ argument, then, the situation in German is different from the one in English in two ways: Nominative Case can be assigned inside the VP (via “chain government”) and the requirement for the passive morpheme is that it has to absorb some kind of structural Case (nominative or accusative). In English, by contrast, there is no nominative Case inside the VP and the passive morpheme has to absorb accusative Case. Intransitive verbs do not assign Case – for English, that means that no Case is available for absorption and the passive is ruled out; for German, it means that the passive morpheme can still absorb nominative Case, and passivization is possible. Again, this analysis does not take into consideration that intransitive verbs in English can assign Case if it is needed, as in a cognate object construction. It is not immediately clear why the same mechanism would not hold for the passive construction. 2.3.3. It’s all about “it” A different syntactic approach is to look into the restrictions for impersonal subjects. In English the subject position in a clause must be filled (Extended Projection Principle). Normally, it is filled with the external argument of the verb, but if there is none, or if this element is not projected syntactically (as in the passive), the position will have to be filled either by movement (not an option in the case of unergative verbs) or with a dummy subject, i. e. a non-referential placeholder element. Dummy subjects in English are it and there: (59)

a. b.

A unicorn is in the garden./There is a unicorn in the garden. John seems to have learned a lesson./It seems that John has learned a lesson.

There-constructions are highly restricted with respect to the verbs and the NPs with which they can occur. For example, the NP has to be indefinite and the verb has to allow for a presentational or existential reading (*There wilted the flowers). The focus will be on it here, since it is an element that can be the subject of passive constructions in principle, as was shown in (54b). The following examples, adapted from Huddleston and Pullum (2002), illustrate the various non-pronominal uses of it as a subject. (60)

a. b. c. d.

It has been raining all day. It seems that she is a good candidate. It is obvious that he has been wrong. It cannot be doubted that she did everything in her power to help.

Verb classes and passivizablity

63

In (60a), it is used as the subject of a weather verb, generally treated as separate from other instances of it. (60b) is a raising construction. Raising verbs have clausal complements and do not have an external argument. If the complement is non-finite, the subject of the complement clause can be moved to the subject position of the matrix clause (He seems to be a good candidate), otherwise the position is filled with it it. Huddleston and Pullum (2002: 1481) call this the “impersonal it.” Note that the complement clause itself cannot be moved to the subject position (*That she is a good candidate seems). By contrast, the fronted clause option is available for (60c) and (60d): (61)

a. b.

That he has been wrong is obvious. That she did everything in her power to help cannot be doubted.

The construction in (60c) and (d) is often referred to as extraposition and the postverbal that-clause is perceived as an extraposed subject (Quirk et al. 1985: 1049). Consequently, the it-subject in constructions like these is known as “extrapositional it.” However, this terminology may be misleading. Semantically, the “extraposed” that-clause is the internal argument of the verb (or adjective) and as such the postverbal position would be its original position. As a transitive verb, doubt can assign structural Case. This Case is absorbed by the passive morpheme. Since no NP depends on Case from the verb, movement is not obligatory, and the empty subject position can be filled with it. The that-clause can remain in its postverbal position – it is not moved or “extraposed.” Another argument against the assumption that the subject position is the original position of the that-clause is that the preverbal that-clause in sentences like (61b) does not behave like an ordinary subject. For example, it cannot be the subject in an embedded clause and it does not seem to occur in questions: (62)

a. *He asked if [that she had done everything . . . ] was doubted by anybody. b. *Is [that she did everything . . . ] doubted by anybody?

Haegeman and Gu´eron (1999: 118) therefore suggest that the preverbal that-clause is not in the subject position, but is an adjunct (adjoined to IP or CP). They argue that the subject position itself is filled with a non-specified empty category, coindexed with the adjoined clause. “Extrapositional” it, on the other hand, behaves like a real subject (Can it be doubted that . . . ),

Chapter 3 Ingredients of the English passive

64

very much like the “impersonal” it in raising constructions.30 What both versions of it have in common, apart from being inserted in the subject position of a finite clause in order to fulfill the EPP, is that they both occur in constructions in which the verb is followed by a that-complement clause. Although Huddleston and Pullum treat it as a non-referential pronoun, they concede that “there is clearly some resemblance here to the use of it in anticipatory anaphora” (2002: 1482), as in (20a): (63)

a. b.

It’s ridiculous! They’ve given the job to Pat. It’s ridiculous that they have given the job to Pat.

In (63a), there is no structural relationship between it and the clause that it could represent, but what about (20b)? Quirk et al. (1985: 89) argue that the role of the anticipatory pronoun it . . . is essentially a structural one in the sense that it carries virtually no information in itself, but merely supplies the structural requirement for an initial subject. (Its semantic function, in so far as it has one, is merely to signal that the content of the subject is expressed in a later position in the same sentence.) A somewhat parallel role is performed by the introductory word there in existential sentences.

The following examples show that the relationship between it and the thatclause seems to underlie locality principles similar to those between an antecedent and its trace (Subjacency Condition, see, for example, Haegeman 1994: 402–405). (64)

a. It was believed that John was here. b. *It was believed NP [the claim CP [that John was here]] (from Burzio 2000) c. *Whoi did IP [you believe NP [the claim CP [ti IP [that she had dated ti ]]]]

30. Huddleston and Pullum (2002: 961) point out that the two types of that-clauses (after raising verbs and after passivized verbs) cannot be coordinated, “suggesting that the content clause does not have the same function in the two cases.” They consider sentences like the following as ungrammatical: *It seemed and was later confirmed that he was trying to hide his true identity. However, a Google search brings up similar examples (e. g., it . . . was often other than it seemed, and was believed, to be).

Verb classes and passivizablity

65

Data like these may indicate that there is some sort of syntactic relationship between it and the that-clause.31 Such a relationship is referred to as an “expletive chain” (Roberts 1987). The question is, with what, exactly, can it form a chain? The answer seems to be that it can only form a chain with a clause. This restriction would explain why intransitive verbs in English don’t form passives: The subject position has to be filled and neither the requirements for it nor for there are met. The situation must be different in languages that do allow impersonal passives. In German, impersonal passives do not necessarily contain es, the German equivalent of it. In main clauses, there is one position preceding the finite verb, the so-called “Vorfeld” (pre-field). This position can be filled with es, but also with PPs or adverbials; see (65a) and (b). It is less clear (to me) if sentences comparable to (64c) are grammatical in German, compare (65c). Still, if there are locality restrictions between es and a finite complement clause, they cannot be the same as in English. (65)

Hier wird nicht geflüstert. here becomes not whispered ‘There is no whispering in here’ b. Hinter dem Vorhang wurde gelacht. behind the curtain became laughed ‘There was laughing behind the curtain’ c. ??Es wurde die Prophezeiung geglaubt, dass . . . it became the prophecy believed that . . . ‘The prophecy was believed that . . . ’

a.

Clearly, the relationship between es and the embedded dass-clause is different from that between it and a that-clause in English. Of the three approaches to impersonal passives discussed here, the one that attributes the non-existence of passives of unergative verbs to the not-quite-dumminess of it is the most satisfactory. Adopting this explanation, we can conclude that 31. According to Burzio (2000: 206), the key to this relationship is Case: “In an A-chain, all positions must receive the same Case by the same assigner” (Case Uniformity: They share the same j-role, they share the same Case). If it and the clause are in an A-chain, they must share nominative Case (assigned by finite Inflection). This means that the clause should share nominative Case with it, but it cannot, because it is embedded inside an NP and therefore cannot “connect” with Inflection. The Case Uniformity Principle thus accounts for movement-like locality conditions between expletive and clause. A simpler Case-related explanation is that the NP following the passive participle cannot be assigned Case (because passive participles are unaccusatives).

66

Chapter 3 Ingredients of the English passive

whether or not a one-argument verb can form a passive depends on: (a) the verb’s argument structure (unaccusatives do not form passives no matter what), and (b) the requirements for filling the empty subject position in a given language. 2.4. Prepositional verbs and prepositional passives English is one of the rare languages in which an NP can be moved out of a PP, leaving the preposition behind (preposition stranding), as illustrated in (66b). One can also move the whole PP (pied piping), leaving behind a trace after the verb. Prepositional passives result from movement of an NP out of a PP in a passive construction, leaving a trace after the preposition. The prepositional passive developed in Middle English; preposition stranding as such is also documented in Old English (Denison 1993). (66)

a. b. c.

I used to sit on this chair when I was little. Which chair did you sit on? On which chair did you sit?

In more traditional approaches, passives like the ones in (66a) and (66b) are sometimes called “pseudopassives” (compare Siewierska 1984) because they are not based on transitive verbs. Since the definition of the passive developed here does not depend on the presence of a direct object, there is no prima facie reason why verbs with prepositional objects should not form passives. Like other passives, prepositional passives are based on the past participle and have an implicit argument reading, and a by-phrase could be added to (67a) and (67b).32 (67)

a. b.

The house was broken into on Sunday. This floor should not be lain on.

According to Biber et al. (1999: 105), in English wh-questions, preposition stranding is clearly preferred over pied piping. Prepositions can also be stranded in other movement contexts, for instance in relative clauses (the chair that we sat on) and in passives (This chair has been sat on by Queen Victoria). So what is remarkable about the prepositional passive? Recall that in a syntactic analysis of the passive, the motivation for movement in the passive is related to the moved NPs need to get Case. Passive participles cannot assign structural Case and may therefore not be followed 32. Most of the examples in this section are adapted from Couper-Kuhlen (1979).

Verb classes and passivizablity

67

by a direct object. The situation is very different with prepositional verbs: In a configuration like the one in (68) NP2 gets its Case from P, not from V. Both V and P are in a c-command relationship with the NP, but V cannot assign Case to the NP because of the intervening PP (see Chomsky 1981). (68)

VP NP1

V’ V

PP

Spec

P’ P

NP2

If NP2 gets its Case from P, it should not matter at all whether V can assign Case or not, since the NP depends on the preposition for Case assignment. But in the passive, NP2 has to move, just like a direct object, see (69). Actually, movement should not only not be obligatory, it should be ruled out: If NP2 gets Case from P and if it is moved to the subject position, it should get a second Case (nominative), which would be a violation of the Case Filter. What is more, (69c) clearly shows that the moved NP bears nominative Case. This means that it cannot have been assigned Case in its original position as a complement of the preposition – a major challenge for Case Theory. (69)

a. *It/There can be relied on him. b. Hei can be relied on ti. c. *Himi can be relied on ti.

The main questions to ask are (1) why prepositional passives are possible at all, and (2) why they are not possible with all V-PP sequences. To address these questions, one first needs to clarify that prepositional verbs are different from particle verbs, also known as phrasal verbs. Unlike prepositions, particles can follow their NP complement and generally have to do so if the NP is a pronoun: (70)

a. b.

She brought up six children./ She brought them up. She insisted on an apology./*She insisted it on.

One of the crucial syntactic differences between particles and prepositions is that particles cannot assign Case (Haegeman and Gu´eron 1999). Thus, there must be a mechanism to ensure that the NP following them receives

68

Chapter 3 Ingredients of the English passive

Case from the verb. This mechanism is movement: Either the particle incorporates into the verb via head-to-head movement, or the NP in question moves to a position closer to the verb in an instance of NP-movement (for details see Haegeman and Gu´eron 1999: 257–264). If particles cannot assign Case, it is clear why movement is obligatory in the passive: The NP depends on Case from the verb, and if that Case is absorbed, it cannot be assigned anymore. The NP has to move to the subject position to get Case – just like a transitive verb. However, since prepositions do assign Case, neither of these options is available for prepositional passives. It does not really make sense to assume that passivization absorbs the Case that is assigned by the preposition (as suggested by Fabb 1984: 51–52), since it is clearly the verb, not the preposition, that is affected by passivization. This approach would also suggest that the grammaticality of passivization only depended on the preposition, which is clearly not the case, as illustrated in the following exmaples. (71)

a. This bedi has been slept in ti by Queen Victoria. b. *A hurryi was left in ti by him.

The classic analysis for this kind of problem is Hornstein and Weinberg’s (1981) rule of reanalysis, which states that V and P can optionally be reanalyzed as a complex verb if they are adjacent and if the result is a “semantically possible word” (Hornstein and Weinberg 1981: 67). In other words: Reanalysis allows for prepositions to behave like particles and become part of a complex verb. Reanalyzed verb-preposition complexes can also be the input for conversion to adjectival passives, as can be seen in (72c) – however, this seems to be somewhat marginal; compare the examples in (d): (72)

a. b. c. d.

They [referred to] the solution The solutioni was [referred to] ti a referred to solution ?? an insisted on solution, ?? a looked at person

So when is reanalysis possible? A syntactic condition named by Hornstein and Weinberg is adjacency: If the verb and the preposition are not adjacent, they cannot be reanalyzed as a complex verb. This correctly rules out examples like (73c), in which the preposition is separated from the verb by the direct object. (73)

a. b.

John put the book on the table. The book was put on the table.

Verb classes and passivizablity

69

c. *[The table]i was put the book on ti The second condition is more lexical: V and P can be reanalyzed if the result could be expressed by a single word. This criterion of potential wordhood is something that seems difficult to check. One might expect that prepositions which are not directly selected by the verb may not be reanalyzed, as they are semantically more distanced from the verb. But the prepositional passive does not only occur with complement PPs, but also with adjunct PPs, as illustrated in (71a). Here, the preposition in clearly has its own meaning and it is not selected by the verb; one can also sleep on a bed, under a bed, or next to a bed, as pointed out by Huddleston and Pullum (2002). In (71b), there is also movement of an adjunct, but this time the result is ungrammatical. The situation is made even more complicated by examples like (74). Here, the verb and the preposition do show rather strong semantic cohesion – potential replacement words would be see or date – and yet passivization is not possible. With a look at idioms, Couper-Kuhlen (1979: 98) points out that “too much cohesion will block the passive”; see, for example, (74b’). (74)

a. a’. b. b’.

He met with an old friend at the dinner party. *An old friend was met with at the dinner party. John beat about the bush. *The bush was beaten about by John.

The verb and the preposition must form a semantic unit, but the unit must not be formulaic, idiom-like. It is very difficult to draw the line. Hence, Huddleston and Pullum conclude, rather pessimistically: “Whether a specified preposition can be stranded in a prepositional passive is not something that can be predicted by a general rule: it has to be recorded in the dictionary as a particular property of the verb or idiom concerned” (Huddleston and Pullum 2002: 1434). It seems that the more semantically opaque the idiom, the less felicitous the passive. A formulaic NP does not make a good topic in a passive construction, not even if the verb is transitive (*The bucket was kicked). Attempts have also been made to explain prepositional passives purely in terms of semantics: If the NP is perceived as an affected object, it can be construed as a direct object (Siewierska 1984); compare the following examples: (75)

a. *England was lived in by John. a’. This house has never been lived in. b. *This bed was slept near.

70

Chapter 3 Ingredients of the English passive

b’. This bed was slept in. A house is more affected than a country by being the residence of somebody. In a similar vein, a bed looks different if someone slept in it, but not if somebody just slept near it. However, these restrictions do not hold for preposition stranding in other syntactic contexts. The restrictions found on stranding in the passive are stricter than those that hold for stranding in general. For prepositional stranding in wh-questions (which does not involve movement to an argument position), for example, affectedness does not seem to be an issue. The fact that adjuncts can be turned into affected objects tells us something about the flexibility with which events can be construed. There are several other ways in which an intransitive verb like sleep can be used transitively, for instance in the resultative construction (sleep oneself sober). Transitive verbs can be used resultatively, and prepositional objects and even adjuncts can be reanalyzed as direct objects. Though it may be difficult to predict under which cirumstances this can be done, the rule and its syntactic constraints are quite clear. Other factors have been pointed out by, among others, Couper-Kuhlen (1979): Sometimes a prepositional passive is possible only in the literal, not the metaphorical meaning of a verb (see (76a) and (b)), infrequent V-P combinations are odd in the passive, and the prepositional passive is also more restricted with respect to modality. (76)

She sat on the egg for three weeks./The egg was sat on for three weeks. a’. She sat on the committee for three weeks./*The committee was sat on for three weeks.

a.

Again, one could say that in the metaphorical reading the NP following the preposition is less affected by the event than in the literal reading. Prepositional passives thus are an important indicator of the semantic component of passivization. The more the object of a preposition resembles the prototypical object of a verb, the more felicitous is passivization.

3. The passive participle It has been argued that the only element that occurs in all passive constructions is the passive morpheme. A problem for the morpheme-based

The passive participle

71

approach is that there is no single morpheme in the passive that does not occur elsewhere. The verb be is also used to express progressive aspects, the preposition by can also occur with nouns (a symphony by Mozart) and the participle occurs in at least three different syntactic environments. As “passive participle” it is combined with be and licenses an implicit agent reading, as “past participle” it is combined with have and does not have any effect on the argument structure of the verb, and as the core of an “adjectival passive” it can occur in positions that are characteristic of adjectives: as predicate in copula constructions (he seems excited) or as a modifier or attribute inside an NP (the excited audience). Many linguists postulate a specific passive suffix – Jaeggli (1986) refers to it as “-en” –, but one cannot overlook the fact that the participle used in the passive has the same word form as the participle used to express perfective aspect. Unless one wants to suggest that the two forms just happen to look identical (which would be a great coincidence, considering all the allomorphs of the morpheme in question; just think of taken, begun, painted), one has to address the question of how they are related to each other. Since the form used for the passive and the perfect seems to be the same, Huddleston and Pullum conclude that they are not two different participles: “there are no verbs where the form used in the passive is different from that used in the perfect. For this reason we take the perfect and passive constructions to involve different uses of the same inflectional form, not different forms” (Huddleston and Pullum 2002: 77–78). The problem is that the participle in the passive does not seem to have the same syntactic properties as the participle in the perfect. In the passive, the participle of a monotransitive verb cannot be followed by a direct object, but in the perfect it can be. (77)

a. The verb has already assigned Case to the NP. b. Case has already been assigned by the verb. c. *It has already been assigned Case by the verb.

If the two participles are actually just one and the same form, how can we explain that their syntactic behavior is so different? In addition, one needs to account for the fact that every verb can form perfect tense, but not every verb can occur in a passive construction. (78)

a. b.

The leaves have/*were fallen from the tree. Your argument has/*was never seemed very convincing.

If one takes the position that fall has a past participle, but not a passive participle, one still runs into problems with verbs that only allow passives under

Chapter 3 Ingredients of the English passive

72

certain circumstances. Intransitive verbs, for instance, generally cannot be passivized in English, but if they are used transitively,33 they can be: (79)

a. She waved/smiled (a thankyou) at the crowd. b. *It was waved/smiled at the crowd. c. Smiles were smiled, goodbyes were waved, and a good time was had by all.

Either one would have to say that transitive smile is a completely different verb than intransitive smile, or one could conclude that passivization involves a lexical rule that changes the syntactic and semantic properties of the past participle. The second option is clearly more economical and will be explored in more detail in chapter 4. Another possibility is discussed by Roberts (1987). He argues that the syntactic behavior of the participle (which is the same form in the passive as in the perfect) depends on the auxiliary it occurs with. He considers have to be a transitive auxiliary, i. e. it can transmit a theta-role and can assign Case, while be is an unaccusative auxiliary, which can do neither. 3.1. On the relationship between the participle and the adjective It has already been pointed out that adjectival passives are not true passives: Neither do they show any implicit argument effects, nor do they license a by-phrase, they can be coordinated with adjectives and can be prefixed by affixes that are subcategorized for adjectives, like un-. In his classic paper on the differences between the adjectival and the verbal passive, Wasow (1977) suggested that adjectives and participles should be different forms, one generated in the lexicon, the other in syntax. But that would lead to the same problem of inflating the lexicon : such an analysis would be costly: in addition to postulating two -ed affixes for regularly formed participles, we would have to list each irregular allomorph twice, once as an adjective and once as a verb. Such a system . . . has another disadvantage: it counts as accidental the fact that, for any given verb, the allomorph listed as a verb and the one listed as an adjective are always identical (that is, it does not explain why write hat the form written as both its adjectival participle and its verbal participle). (Lieber 1983: 276) 33. One could, of course, say that these are two different verbs, but attempts have been made to generalize that all unergative verbs are in fact underlyingly transitive (Felser and Wanner 2001).

The passive participle

73

If the adjective is to be considered a derived category, we need to ask what it is a derivation of and how exactly it is derived. Lieber (1983) opts for attachment of a zero affix, while Levin and Rappaport (1986) suggest a rule of conversion from passive participle to adjective. The reason why they base the adjective on the passive participle (rather than the other way around) is that the passive participle does not have an external argument. The externalization of an internal argument of the verb is a consequence of the conversion process (since adjectives have an external argument) and need not be stated separately. Not every verb forms an adjectival passive. Participles that are converted to adjectives characterize nouns with respect to verbal actions. This holds for the –ing participle as well as for the –ed participle or, as Haspelmath (1990: 40) calls it, the “resultative participle.” According to him, the adjectival passive can only be used to characterize the patient of an action because it is the only argument affected by the action. Williams (1981a) captures this in his rule of creating adjectival passives as “E (Th)” – a rule that specifies that the theme argument of the underlying verb be externalized. However, Levin and Rappaport (1986) demonstrate that the externalized argument need not always be the patient or theme of the verb. Compare the following examples (from Levin and Rappaport 1986), which involve externalized goals. In the case of teach and feed the result is grammatical, in the case of hand it is not. We need to teach TH[these skills] GOAL [to all children] well-taught children untaught skills

(80)

a. b. c.

(81)

a. You need to feed GOAL [the baby] TH [more cereals]. b. the unfed baby c. *unfed cereals

(82)

a. They handed TH [a knife] GOAL [to the gardener]. b. *the handed knife c. *the handed gardener

Levin and Rappaport argue that the constraint underlying the rule of Adjectival Passive Formation is not thematic, but syntactic: Only an argument that can stand as the “sole complement” of the verb (Levin and Rappaport 1986: 258) can be externalized. As illustrated in (83), in the case of teach, either the theme or the goal can be the sole complement of the verb, hence the verb forms two adjectival passives. In the case of feed, only the goal can be the sole complement, and in the case of hand, neither theme nor

74

Chapter 3 Ingredients of the English passive

goal can be the sole complement, hence there is no grammatical adjectival passive. (83)

a. We need to teach these skills./We need to teach these children. b. You need to feed the baby./*You need to feed the cereals. c. *They handed the knife./*They handed the gardener.

This observation has interesting consequences for the syntactic representation of intransitive verbs, of which some allow adjectival passives, while others don’t: (84)

a. b.

The flower wilted → the wilted flower The children giggled → *those giggled children

The standard interpretation of these data is that the flower must be a complement of wilt at some level of representation, while the children is not a complement of giggle at any point in the derivation. In other words: Wilt is an unaccusative verb with an underlying object, giggle is an unergative verb, with a base-generated subject. The reason why the meaning of adjectival passives seems to be based “on the passive rather than the active verb” (Bresnan 1982: 23) lies in the syntax of the construction: The adjective is based on the participle. (85)

a. b.

[The flower]i wilted ti [The children] giggled.

Not everybody buys into the syntactic reality of unaccusativity, but there is a long tradition of research into unaccusativity tests in English and other languages (Burzio 1986; Grimshaw 1986; Levin and Rappaport Hovav 1995; Wanner 1999; Hoekstra 2000). Adjectival Passive Formation is considered a good test for the presence of an NP complement or trace at the level of D-Structure, while cognate objects and reflexive resultatives are constructions in which unergative verbs can be found (the children giggled themselves silly/*the flower wilted itself onto the desk). Perlmutter (1978) noted that in languages that have impersonal passives not all intransitive verbs can form a passive. He assumed that the verbs which can form impersonal passive have an underlying subject (i. e. they are unergative), while the verbs that cannot form impersonal passives have an underlying object (i. e. they are unaccusative). The rule of passivization affects the external argument (argument suppression, see chapter 4), while the rule of adjectival passive formation affects the internal argument (argument externalization). We can see that they are in fact complementary rules – they target different elements,

The passive participle

75

they have different effects, and there is no reason to consider them two sides of the same coin “passive.” However, there remains one problem with the externalization approach. If one regards the passive participle as the input for the rule that creates adjectival passives, how can one explain adjectival passives of verbs that do not have a passive participle to begin with (verbs like wilt)? In a footnote, Levin and Rappaport concede that the adjectival passive (at least of unaccusative verbs) may in fact involve “the affix with the features of the perfect and not the passive participle” (Levin and Rappaport 1986: 269, fn. 16). If the rule of Adjectival Passive Formation essentially is a rule of conversion from verb to adjective with the externalization of an element that can function as the “sole complement” of the verb, it also follows that the adjectival passive will not show any implicit argument effects: Unlike in the passive, the external argument of the verb is deleted and another argument takes its place. Unergative verbs will not form adjectival passives unless they are used transitively (a half-smiled smile), because otherwise there is no argument that could be externalized. 3.2. Passive meaning without passive participle According to Huddleston and Pullum (2002: 1429), the participle “is almost, but not quite, an invariant feature of the passive.” They regard gerundial constructions, like (86a), which can come after verbs like need, require, deserve, want, and bear, as “concealed passive(s),” because they can be paraphrased as regular passives, as in (86b).34 (86)

a. b.

This draft needs checking carefully (by the editor). This draft needs to be checked carefully (by the editor).

They point out that this construction is different from what they call the “hollow active” (Huddleston and Pullum 2002: 1199), which may not be convertible into a passive and which doesn’t allow a by-phrase: (87)

a. The article is worth having a careful look at (*by the editor). b. *The article was had a careful look at.

Jespersen (1927: 112) notes that “gerunds were originally indifferent to the distinction between active and passive meaning; accordingly in some con34. The OED states that in some dialects of British and American English the construction also works with a passive participle (Does my hair need combed?).

76

Chapter 3 Ingredients of the English passive

texts they are still understood passively.” He cites examples that even combine an active and a passive reading, as in the following examples, taken from Jespersen (1927). (88)

a. b.

Harriet only wanted drawing out and receiving a few hints. (from Jane Austen, Emma) She deserves punishing for punishing me. (from Emily Bront¨e, Wuthering Heights)

The actual gerundial passive (They are not fond of being teased) is a more recent development. Jespersen dates its beginnings around 1600. With the emergence of this construction, ambiguity between an active and a passive reading could be avoided. The “concealed passive” will not be considered a true passive here, as it seems to be a remnant of the ambiguity found in earlier gerundial constructions. Furthermore, the spectrum of participles in the gerundial construction is more restricted than in the passive. For example, neither the direct nor the indirect object of a ditransitive verb can be the subject in the gerundial construction: (89)

a. *John needs giving the book. vs. John needs to be given the book. b. *The book needs giving to John. vs. The book needs to be given to John.

The fact that sentences such as (86a) can be paraphrased as passives alone does not make them passives.

4. The by-phrase The preposition by occurs in the passive regardless of the thematic role of the argument embedded in the PP. The theta-role of the by-phrase is identical to that of the external argument in the active (Roberts 1987; Wyngaerd 1988), often, but not always, an agent. This is one of the strongest arguments for relating passives and their corresponding active clauses by movement: If there is no relationship between the two sentences, it is difficult to account for the fact that the subject in construction A has exactly the same interpretation (and underlies the same selectional restrictions) as an adjunct in construction B. According to the linking rules that map arguments from semantic to syntactic structure (Pinker 1989; Grimshaw 1990; Levin and Rappaport Hovav

The by-phrase

77

2005) the external argument, although not necessarily an agent, is always higher on the Thematic Hierarchy than any other argument of the verb. This means that in the passive the by-phrase is “more agentive” (Pinker, Lebeaux and Frost 1987: 228) than the subject phrase (Thematic Hierarchy Condition, Jackendoff 1972). Pinker, Lebeaux and Frost demonstrate that although most passives occur without by-phrases, children’s rules of passivization do not lack a role for the by-phrase (see chapter 2). Although by is one of the most prominent markers of the passive construction (and one of the most important cues for children to recognize a passive), it has not been one of the core ingredients of the passive from the very beginning. In Old English, the prepositions associated with the passive were of and fram. The OED dates the first use of by in the passive at 1400. By was originally an adverb and became a locative preposition indicating nearness or direction, as in He lives by the river.35 What makes the status of the by-phrase so interesting is that across all registers, most passives (80 % or more) occur without a by-phrase. Again and again, corpus-based research has shown that the vast majority of passives are “short” passives (Svartvik 1966; G´ıvon 1979; Thompson 1987; Biber et al. 1999) – a fact that, incidentally, seems to be something that style manuals which consider the passive unnecessarily wordy, completely neglect. Due to the lack of an explicit agent phrase, many passives are actually shorter than the corresponding active. And yet the by-phrase, or rather the potential to take a by-phrase, can be regarded as a hallmark of the passive construction. It will be argued in chapter 4 that the by-phrase is licensed by the implicit argument in the passive. Syntactically, it behaves like an adjunct, but semantically, it is an argument of the verb. Since passives with by-phrases (long passives) express exactly the same propositional content as the active, “it is . . . reasonable to expect that the choice [between active and passive] may be influenced by the same types of factors as affect pure word-order variations” (Biber et al. 1999: 940). The two most important factors that Biber et al. list are information flow (discourse-given versus new information) and weight (length and complexity of a phrase). There is a tendency for given information to occur in the subject position (hence the higher number of subject pronouns than object pronouns) and for long and complex phrases to occur at the end of the sentence (principle of end-weight, compare chapter 2). Biber et al. (1999) find 35. By-phrases that are lexically selected by adjectives (surrounded by trees) or that have locative or instrumental meaning (live by the sea, write a letter by hand) will not be considered here.

78

Chapter 3 Ingredients of the English passive

that to a certain extent this is also the case in passives with by-phrases. In their corpus (finite sentences only), “about 90 % of agent phrases [in passives] bring in new information” (Biber et al. 1999: 941) and there is also “a clear tendency for the subject to be shorter than the agent phrase in long passives” (Biber et al. 1999: 40). However, they also recognize that there are many passives in blatant violation of the principle of end-weight. In (90) the subject is “heavy” both in terms of length (13 words) and structure (finite clause, not just an NP), while the by-phrase is short (two words) and not very complex. On the other hand, the sentence is in line with the information principle: The subject sentence relates to information that has already been mentioned (as indicated by the use of similar and the demonstrative these) while the agent is introduced as new information. Obviously, choosing the passive over the active here cannot be related to the length or complexity of the agent phrase. (90)

That similar relationships occur with these two species under field conditions in Saskatchewan was suggested by Pickford (1960, 1966a). (Biber et al., 1999)

Actually, the mere fact that most passives occur without a by-phrase is an indicator that the length of the phrase that expresses the agent is not always a major concern in choosing the passive over the active. Many linguists, among them Thompson (1978) and Biber (1999), have argued that the short passive and the long passive are essentially different constructions and need to be looked at quite separately. Thompson (1978: 501) considers passives without by-phrases “obligatory under certain conditions, these conditions having less to do with the structure of the discourse than with the encoder’s intentions and assessments of the decoder’s inferential capabilities.” The short passive is used because the agent is not to be mentioned explicitly, either because its identity is unknown, unimportant, or evident from the context. An example (from Thompson, 1987, originally from G´ıvon, 1979) is given below: (91)

I left under circumstances of considerable honor. I was given a farewell luncheon by half the staff of the law firm, meaning the lawyers themselves. I was asked to make a speech and I was much applauded.

In this example, it can safely be inferred that the people who host the luncheon are the people who ask the narrator to give a speech. Choosing the passive allows for a chain of three sentences with the same topic. Note that the passive is not “agentless” (a term used by Thompson), the agent is sim-

The by-phrase

79

ply implicit. In this regard, the passive is quite similar to the use of nonanaphoric they in active sentences: (92)

I left my job under circumstances of considerable honor. They gave me a farewell luncheon and asked me to make a speech.

Here, they refers to the external argument, but its exact identity is not revealed. What is different here is that non-anaphoric they occupies the subject position, i. e. the position associated with the topic of the sentence. In passives, short or long, this position is never taken up by the agent argument. Hence, it is not quite the case that passives with implicit agents are not really interesting with regard to information structure and discourse organization. The short passive allows for a non-agent to take the position of the topic in the sentence and for minimizing the visibility of the agent. But there may be other reasons for choosing a passive without a byphrase. In a case study of 160 abstracts preceding scholarly articles in journals from different academic disciplines, discussed in greater detail in chapter 5, 77 out of 378 passives (21.5 %) occurred without a by-phrase. Even though the corpus is small (21,804 words), the frequency of by-phrases is very much in line with results from larger corpora. Due to the nature of the abstract as a genre (scientists and scholars summing up their papers), in many instances the agent of the event that is reported is the author of the paper. What is striking is that not a single by-phrase in the corpus refers to the authors of the papers themselves. This is certainly not because the identity of the agent is not important (in scientific writing, attributing claims and findings correctly is essential), it is because in the sciences there is still an understanding that the author should detach himself or herself from the work to be reported (Hansen 1998). If this is a concern, the passive may be preferred over the active, even if this results in sentences with awkwardly complex subjects, which, as in (93b), are neither short nor given. (93)

a. b.

Previous studies of the get-passive are reviewed and their focus on contextual and interpersonal meanings is noted. (ELL 13) Standardised gross and net fixed capital stock estimates for the 1950–94 period are presented for seven Latin American countries. (CJE 3)

The passive is chosen because it allows for the agent to remain implicit, i. e. the authors do not have to refer to themselves explicitly. The point is not so much that a long passive would have been more appropriate here, the point is that the passive allows for a very specific realization (implicit argu-

80

Chapter 3 Ingredients of the English passive

ment) of a very specific agent (first person). Even though there is a strong “Avoid Passive” movement in academic writing, in particular in the United States (Zwicky 2006), the impersonal style of reporting methods and results in scholarly article is still very much alive, especially in the experimentoriented sciences (see chapter 5 for a more detailed analysis).

5. The subject While only a minority of passives occur with a by-phrase, the subject is a component of every passive – simply because the passive is a clausal construction, and in English, every clause has a subject. Technically, therefore, the subject is less of a marker of the passive than other ingredients are. Beyond the fact that they occupy a specific syntactic position in a specific syntactic constructions, subjects in passive clauses are not really a very coherent class. As has been discussed, the subject in the passive is not always an argument of the passivized verb. It can be the raised subject of a lower clause (John was expected to leave), the object of a preposition (This bed has been slept in), or an anticipatory filler subject (It was suggested that . . . ). Which of these options occurs depends on the syntactic class of the passivized verb. It has also been noted that the subject in the passive is not always an affected entity, although this may be a necessary condition for passives that require the reanalysis of an argument, e. g. prepositional passives, to be felicitous. And it has been shown that children as young as four can comprehend and produce passives that do not involve prototypical agents and patients (Pinker, Lebeaux and Frost 1987). Passives that occur in infinitives can have invisible subjects (PRO), just like other infinitive clauses (He claims PRO to have been invited). Semantically, the subject in the passive can have a variety of theta-roles, with the exception of the agent role (the reason is that an agent would always end up as an external argument and would therefore become an implicit argument in the passive). Due to the movement operation at work in most passives, the strong association of the subject position with an animate entity is often broken up. Hence, the passive is typical of registers that show a preference for inanimate subjects, such as academic discourse. Biber et al. (1999: 378) report that in academic prose over 60 % of all causative verbs, over 30 % of all activity verbs, and about 20 % of all communication verbs occur with inanimate subjects (these numbers do not include passives).

The subject

81

It has already been mentioned that in passives with by-phrases the subject is usually shorter than the by-phrase (Biber et al. 1999: 940). Biber et al. argue that the syntactic complexity of the agent argument is one of the main reasons for choosing the passive over the active. There is a preference for placing complex phrases towards the right periphery of the clause (principle of end-weight), because this makes it easier for the listener/reader to process the information. This is why that-clauses that depend on verbs or adjectives usually occur in post-predicative position (traditionally known as extraposition), as in (94a), rather than in subject position, as in (94b). In the passive, the subject position is empty at D-Structure, but there is usually no choice between filling it through movement or with a dummy element. The NP following the passive participle needs Case and therefore needs to be moved to the subject position, whether or not this leads to a long and complex subject. Choosing the passive often means accepting a complex subject phrase. (94)

It is vital [that leaking water is avoided]. (from Biber et al., 1999) b. [That leaking water is avoided] is vital. c. [The implications for education and social policy] are discussed. d. *It is/There are discussed the implications for education and social policy.

a.

The following examples show rather drastic violations of the principle of end-weight. A passive is chosen in spite of the complexity of the resulting subject. (The examples come from the abstract corpus discussed in detail in chapter 5.) In all four cases, the subject of an agentless passive has more than five words and includes some sort of coordination. The unnamed agent in all four examples is the author of the abstract. (95)

a. b. c. d.

. . . some fairly plausible scenarios involving overfull employment and inflation are presented. (AJES 13) Satellite repeats were absent, but tandem arrays of long repeats and many transposons were found. (Cell 16) The implications for education and social policy are discussed. (JME 3) Previous studies of the get-passive are reviewed and their focus on contextual and interpersonal meanings is noted. (ELL 13)

Another important factor in determining the choice of the subject (and thus, indirectly, determining the choice between active and passive) is the information status of the subject versus the agent in the passive. Generally,

82

Chapter 3 Ingredients of the English passive

given information is placed at the beginning of the clause and new information at the end: “The affected participant is chosen as subject if the context makes it a more natural starting-point than the agent, especially if it is given in the context and is less informative than the agent” (Biber et al. 1999: 943). There is a connection between this information principle and the principle of end-weight: Given information can be expressed by shorter phrases (notably pronouns), while new information often requires longer phrases, i. e. the two principles often coincide, although this is obviously not the case in the examples above. If somebody reports his or her research, it is not immediately clear why the agent would not make a good “natural starting-point.” However, the conventions of scientific writing being what they are, there is still a tendency to avoid the first person. The passive is one way of doing so. Other ways are discussed in chapter 5. As to the information status of the subject in the passive in the examples above, it qualifies as given, or at least as not discourse-new, in all cases. In (95a), plausible scenarios refers to a previously mentioned phenomenon; in (b), the subject is contrastive (note that the passive also allows for the superficial stylistic parallelism of were absent versus were found); in (c), the implications refers to a finding mentioned previously; and in (d), their focus refers back to the subject of the previous clause, previous studies. However, in each of these examples the agent, too, is an entity that has previously been introduced to the discourse. Abstracts are generally preceded by the name of the author(s) and the title, i. e. both agent and theme are evoked. A typical opening sentence for an abstract refers to the text, the study, and the agent (In this study, we examined X). Choosing the passive over the active under these circumstances does not necessarily result in a sentence that is more in line with the principles of information flow from given to new and from short to complex. A specific case of creating cohesion can be seen in the passive that occurs in a participial clauses, as illustrated in the following examples. The passive here is part of a participial relative clause. Due to the nature of the construction, the auxiliary be is not necessary and the subject position in the embedded clause is filled with an empty category.

(96)

a.

b.

The signal [received at the target] is of course several orders stronger than the echo signal [received back at the radar transmitter]. (FROWN; J80) Like Sheffrin’s study, ours puts this debate in an international context. We include several countries [not studied by Sheffrin]

The auxiliary verb

c.

83

notably Australia, Canada, and Japan, and introduce new data for Sweden. (FROWN; J46) Wigmore’s ideas for reforming the law of evidence were part of the emergence of legal progressivism, or sociological jurisprudence, [led by Roscoe Pound]. (FROWN; J44)

In these examples, choosing the passive has the effect that no overt pronoun or NP intervenes between the modified NP in the main clause (the signal, the echo signal, several countries) and the non-overt subject in the modifying relative clause. The distance between the two co-referential NPs therefore is minimal. Since scientific discourse is often concerned with describing properties of non-animate entities, it comes as no surprise that this construction, the participial relative clause, is popular in that specific register.

6. The auxiliary verb In most descriptions of the passive, the auxiliary be is recognized as an integral component of the construction. Its use reaches back to Old English. In Old English, the passive could also be constructed with a form of weordan (“become”), as is still the case in Modern Dutch and German, but be “has always been the principal auxiliary verb used in passives” (Denison 1993: 418). According to Traugott (1972: 83) the grammaticalized use of be in the passive is a result of the restructuring of adjectival constructions. However, there are also passives without be or, for that matter, get, so-called “bare passives” (Huddleston and Pullum 2005: 1430). Unlike get-passives, bare passives are not so much a stylistic alternative to the be-passive, they are passives that occur in a syntactic environment that does not require the use of a finite verb to pick up tense and agreement features. These positions include modifiers or adjuncts (examples adapted from Huddleston and Pullum (2002)), as well as complements of verbs. Depending on the position of the bare passive, its subject can be overt or non-overt. (97)

a. b. c.

I came across a letter [written by my great-grandfather]. [Looked at from an economic perspective], the proposal doesn’t make any sense. Two-year-old Constantin will have [his cleft palate repaired].

84

Chapter 3 Ingredients of the English passive

As an alternative to Biber et al. (1999), who classify constructions like the bracketed clauses in (97) morphologically (as “ed-clauses”), I will classify them according to their function and syntactic position. (97a) is a participial relative clause, (b) is a participial adverbial clause, and (c) is a participial complement clause. What makes the clauses passives is that the participle is a passive participle (with all the syntactic and semantic consequences, first and foremost an implicit agent reading). With regard to Case-assignment and NP-Movement, the syntax of these passives is the same as that of their finite counterparts, as indicated in (98); for the use of a non-overt relativizer that binds a wh-trace, see Haegeman and Gu´eron 1999: 189–191. (98)

a. b.

NP[a N[letter CP[OPi ti NP[a

written ti by my grandfather]]] [letter CP[whichi ti was written ti by my grandfather]]]

N

Of these constructions, the participial relative clause36 is of specific interest. Postmodifiers are “generally rare in conversation” but “extremely common in academic prose” (Biber et al. 1999: 606). It is well-known that in English relative clauses with non-overt pronouns are much more frequent than relative clauses introduced by wh-pronouns or that (e. g., Biber et al. 1999). However, the non-overt pronoun is not an option for relative clauses with a subject gap (*I came across a letter was written by my great-grandfather), which means that the participial relative clause is the only way to use a relative clause in this context without using an overt relativizer. It is to be expected that the participial relative clause is particularly popular in genres that favor postnominal modifiers and compact expressions, such as some genres of academic writing (see chapter 5). However, many corpus analyses of the passive only count a construction in which be is followed by the passive participle as a passive (e. g., Seoane 2006; Xiao, McEnery and Qian 2006). Passives like the ones in (96) and (97) are often not included in corpus-based analyses. The reason for excluding them is probably more pragmatic than principled: Passives that are defined via a combination of auxiliary be and the past participle are much easier to extract from a corpus than these “bare passives.” In the examples in (96), the passive occurs in a clause that modifies a nonanimate noun. In the corpus of abstracts already mentioned, this is the case 36. I will focus on that kind of participial relative clause that contains a passive participle here. Of course, there are also participial relative clauses that contain present participles (The man [making the bogus collections] was described as middle aged, from Biber et al. 1999). For distributional differences between the two kinds of participial clauses see Biber et al. (1999: 199–200).

A re-evaluation of the get-passive

85

for the majority of participial passives. Another version of the participial passive is the adverbial clause, illustrated below. (99)

a. b.

Viewed collectively, these findings seem to suggest . . . (APL 18) . . . which would be the first . . . evidence for an effect of . . . gravity as predicted by general relativity (Sci 3)

The participial passive is the most compact way to integrate the adverbial or modifying clause into the matrix clause. Since non-animate topics and postnominal modifiers are rather frequent in academic discourse and since compact expressions are highly valued, it does not come as a surprise that the participial relative clause is not a marginal phenomenon in that particular register. “Bare passives” are in line with the generalization that subjects in the passive are generally given. The realization of the external argument of the verb is also no surprise. If the identity of the agent matters, the agent occurs in a by-phrase. If the identity of the agent can be inferred from the context, it may be realized as an implicit argument. In chapter 5 I will discuss why the second option is the preferred choice if the authors of the article themselves are the agents of the reported event. Even though comparing the frequency of be-passives across different genres or periods of time will lead to important insights about the increase or decrease of the passive (Seoane 2006), if one is interested in the function of the passive as such, one should not write off bare passives as marginal (or simply as too difficult to find in a corpus that is not syntactically parsed). The discussion in chapter 5 will show that in certain genres up to 20–25 % of passives may occur without be.

7. A re-evaluation of the get-passive Passives in English are formed periphrastically: The lexical verb occurs in its non-finite past participle form, and the auxiliary be encodes voice, tense and agreement. There are some exceptions: If the lexical verb is part of a participle small clause, i. e. a clause without a tense projection, no auxiliary is needed and the result is a passive without an auxiliary, as has just been discussed. The second exception from the generalization that all passives involve a form of be is, of course, the get-passive. Unlike the passive without auxiliaries, which is dependent on a specific syntactic configuration, the getpassive seems to be more of a stylistic variant of the be-passive. The verb

86

Chapter 3 Ingredients of the English passive

get seems to replace the form of be and is followed by what looks like a participle, as in (100): (100)

a. b.

John was/got hurt in an accident. After the disastrous meeting, Harry was/got fired.

Corpus studies have shown that the get-passive is generally rare, even in spoken English. Johansson and Oksefjell (1996) report a ratio of get-passives to be-passives of 1:30 for the London-Lund Corpus (500,000 words from spoken British English). Among the claims that have been made about how the get-passive is more restricted than the be-passive are the following: – Register: The get-passive is far more common in spoken than in written English (Biber 1988; Johansson and Oksefjell 1996), but even in conversational English, the get-passive is much rarer than the be-passive (Biber et al. 1999: 476). – Regional variation: The get-passive is used more often in American than in British English (Sussex 1982) and is also acquired earlier in American English than in British English (Meints 2003). – Social variation: The get-passive started “in the speech of less educated people” (G´ıvon and Yang 1994: 138) and is more common in the speech of working-class speakers (Herold 1986). – Semantic and aspectual restrictions: The get-passive is only common with a very limited set of verbs (Biber et al. 1999: 481), the get-passive receives a more dynamic interpretation than the be-passive (Quirk et al. 1985) and cannot occur with stative verbs (Taranto 2002). – Adversity: The get-passive often has adversative – and sometimes beneficial – interpretation (Chappell 1980; Carter and McCarthy 1999), and reflects the speaker’s attitude towards the event (Lakoff 1971). This characteristic of the get-passive has been traced back to its very beginning in the seventeenth century (G´ıvon and Yang 1994). – Subject responsibility: The subject in the get-passive is generally animate and, although underlyingly the object of the verb, is interpreted as somehow responsible for the action (Arce-Arenales, Axelrod and Fox 1994; G´ıvon and Yang 1994; Huddleston and Pullum 2002). In a sentence like Mary got shot on purpose, Mary is not the agent of shot, but she is considered to be somehow responsible for the event (having brought the event about through some action of hers). This is also known as the “secondary agent” reading (Roeper 1987). – History: While the be-passive can be traced back to Old English, the getpassive is a development of early Present Day English (first recorded in

A re-evaluation of the get-passive

87

the seventeenth century); it may be considered a late successor to the Old English passive formed with weordan (G´ıvon and Yang 1994). Most of the studies cited above compare the get-passive to the be-passive and try to single out factors that work in favor of either get or be. For example, get may be preferred over be to avoid the stative reading that can be associated with be and that can make a passive ambiguous between an adjectival and a verbal reading (the wall was/got painted). The restrictions on the get-passive can also be reframed as a description of the get-passive as a Construction (in the sense of Construction Grammar): Characteristics like the adversative reading and the responsibility of the subject for the event would then directly be associated with the Construction rather than be construed as predictable from its components (see Goldberg 1995: 4). What is problematic about this approach is that there are many get-passives that do not show the characteristics listed above. In this section I will take a different approach, starting with a discussion of the syntax of get and positioning the get-passive within the spectrum of get-constructions. One of the reasons for looking at the get-passive as a getconstruction lies in the fact that syntactically, get in the get-passive behaves just like the lexical verb get, not like an auxiliary. It does not show any of the NICE features (Negation, Inversion, Code, Emphasis) typically associated with an auxiliary verb. Just like the lexical verb get, it requires do support in contexts of negation and question formation. Furthermore, it cannot be stressed and cannot be stranded in VP-deletion contexts (Haegeman 1985). If get behaves like a lexical verb, it should be relatable to get in other contexts. (101)

a. b. c. d.

He got/was promoted recently. He did not get/*got not/was not promoted recently. Did he get promoted recently?/*Got he promoted recently? John got killed in an accident and Bill *got/did too.

In the following, I will first focus on the question of how the syntax of the get-passive is related to other uses of get and whether or not the properties of the get-passive can be explained through the properties of get. A key factor will be the differentiation between causative and ergative (i. e. unaccusative) get, as suggested by Haegeman (1985). I will then examine constructions that have been classified as get-passives in the literature, applying the criteria for classifying a construction as passive established in chapter 2. Specifically, I will show that a number of sentences that involve adjectival passives have been misclassified as get-passives. Since the get-passive is con-

88

Chapter 3 Ingredients of the English passive

sidered to be generally rare and common only with a limited number of verbs, it is very important that the analysis not be blurred by the inclusion of data that are in fact not passives at all. The claims that have been made about the get-passive, in particular the responsibility reading assigned to the subject, will be looked into on the basis of data from the FROWN corpus of contemporary written American English. It will be shown that some of the generalizations listed above do not hold and that the characteristics of the get-passive are not inherent to the construction. I will argue that the high frequency of adversative get-passives has to be seen in the context of the register in which the get-passive is used. Negative get-passives often involve colloquial expressions (such as get whacked), while get-passives in more formal contexts are often combined with verbs that have no negative associations (get elected). Regarding the subject involvement factor, I will argue that the existence of causative and reflexive get-passives influences the reading of the subject as responsible for the action. The analysis will not only shed light on what is peculiar about the get-passive, it will also add to the general discussion of the role of the representation of agenthood in a clause.

7.1. Locating the get-passive in the spectrum of get-constructions In Present Day English (PDE), get is an extremely versatile verb. With more than 80 meanings listed by the OED, it does not come as a surprise that get is one of the most frequent verbs in English. Biber et al. (1999) list it as the second most frequent lexical verb in English, only surpassed by say, and as the most frequent verb (by far) in conversations (Biber et al. 1999: 373–375). It is rarer in written registers “because many of its uses have strong casual overtones which are avoided by more careful writers of informational prose” (Biber et al. 1999: 376). This is something get also brings to the passive construction – while the passive in general is more frequent in formal registers, the get-passive (though overall less frequent than the be-passive) is more typical of conversations. Mair and Leech (2006: 332) point out that in spite of a trend in written language to adopt features associated with spoken language (hence a general decrease of the passive, see Seoane 2006), there is no indication that the be-passive is replaced by the more colloquial get-passive in formal registers. Even though there is a general increase in the use of the get-passive in written English, this increase is “infinitesimal in terms of absolute figures and cannot compensate for the drop in be-passives” (Mair and Leech 2006: 332).

A re-evaluation of the get-passive

89

But the similarities between get in the get-passive and in other constructions go beyond a ring of informality. It is precisely the rather unspecific, context-dependent meaning of get that leads linguists like Gronemeyer (1999) to a unified analysis of get. In her study based on 1,482 occurrences of get in the BROWN corpus, Gronemeyer (1999: 3) lists uses that “range from lexical to grammatical meanings” and that include meanings like possession, causation, movement, obligation, permission, passive, inchoative, and ingressive aspect. Even though get in the get-passive looks very much like an auxiliary (it does not seem to contribute any propositional meaning and can usually be replaced by a form of be), it still behaves like a full lexical verb, as illustrated above. Gronemeyer (1999) characterizes get as a “semigrammaticalized” verb, whose meaning is largely determined by its syntactic context. For instance, get has causative meaning when it occurs in the double object construction (I got him a bike). Haegeman (1985) assumes that there are two basic variants of get, one with an underlying subject and one without, and that the classic get-passive is an instance of the non-agentive, “ergative” (or unaccusative) get. Diachrony provides a second bridge between the get-passive and causative uses of get. Givon/Yang ´ (1993) attempt to explain the peculiarities of the get-passive (like the secondary agent reading that is often ascribed to its subject) through its relationship with the uses of get from which the getpassive arose. In its earliest documented uses, get was an agentive verb with the meaning “to obtain” or “to grab” (OED). From there, get developed into a verb that could take more complex complements, as illustrated below (examples taken from G´ıvon and Yang 1994). (102)

a. b. c. d.

V+NP+PP: Er that he myghte get his wyf to shipe? (Chaucer, Miller’s Tale) V+Clause: Our youth got me to play the woman’s part. (Shakespeare, Two Gentlemen of Verona) V+Participle: Ay, marry, will we; therefore get ye gone. (Shakespeare, Henry VI, Part 2) V+Clause in passive: To get me be admitted for a nun. (Marlowe, Jew of Malta)

G´ıvon and Yang (1994: 130) argue that the two crucial precursors of the getpassive were both causative: (1) the transitive/causative construction with the participle verbal complement gone, as in (103a), and (2) the causative construction with a be-passive, as in (103b).

90

(103)

Chapter 3 Ingredients of the English passive

a. b.

So get thee gone, that I may know my grief. (Shakespeare, Henry VI) Or by what means got thou to be released? (Shakespeare, Henry VI)

According to G´ıvon and Yang, the first step in the development towards the get-passive was an increase in semantic and syntactic complexity through the addition of a beneficiary (get something for somebody). The next step was that the additional phrase did not have to be a beneficiary, it could also have locative meaning (get something/someone somewhere). The causative sense of get was then extended to include changes that were non-locational. Syntactically, this resulted in the possibility of verbal complements, first of verbs associated with locations (get thee gone) and from there to verbs in general (get someone to leave), including passivized verbs (get someone to be admitted). The second line of development led to ergative uses of get. G´ıvon and Yang (1994: 143) argue that “a two-step process of detransitivization suggests itself.” First, the causative locative use of get became reflexive (get oneself into the house) and then the reflexive was dropped, resulting in a non-causative use (get into the house). The combination of the two, the possibility of verbal complements and the non-causative use of get, is what led to the get-passive (get killed). The adjectival get-passive (get mad) is seen as a further development. This argument is supported by the fact that many of the adjectives in the adjectival get-passive are “de-verbal perfect-participle forms,” i. e. adjectival get-passives. Although G´ıvon and Yang want to avoid the impression that the development of the get-passive can be described in a neat linear sequence, they are very clear on the order of the three types of get-passives:37 First came the passive after causative get, then the passive after non-causative get, and then the adjectival passive. This is also the direction of derivation usually assumed for the two members of a verb that allows the causative alternation (Levin 1993; Levin and Rappaport Hovav 1995). Ergative get is derived from causative get, presumably by a process of argument deletion. It therefore makes sense, both synchronically and diachronically, not to exclude causative uses of get from an analysis of the get-passive. Gronemeyer points out that get is more general and more frequent than semantically related verbs like obtain or acquire, and thus lends itself more easily to grammaticalization processes (Traugott and Heine 1991). Overall, 37. The adjectival passive is not really a passive construction (since it involves an adjective, not a participle), but since the name is so well-established, I will continue to use it.

A re-evaluation of the get-passive

91

the grammaticalization of verbs of giving and receiving is a phenomenon that goes beyond English (and beyond the passive). German, for example, has a passive formed with bekommen (“receive”) or kriegen (“get”), the socalled “dative passive” (Wegener 1985), which allows for the absorption of dative case. In the standard passive – constructed with a form of werden (“become”) – only accusative case can be absorbed and the direct object becomes the nominative subject. If a monotransitive verb assigns dative case, this case is not absorbed, i. e. there is no reason for the object to move to the subject position and to appear in nominative case.38 (104)

a.

b.

c.

Sie halfen ihm. They helped him-DAT. ‘They helped him’ Ihm wurde von ihnen geholfen. Him-DAT became by them helped ‘He was helped by them’ Er bekam von ihnen geholfen. He-NOM got by them helped ‘He was helped by them’

Similar cases of polysemy can be observed in Swedish, where the equivalent of get, f˚a, can express possession as well as obligation, causation and ingressive aspect (Gronemeyer 1999), and in Dutch, where the equivalent of get, krijgen, can also be used to form a recipient passive. 7.2. Ergative and causative get The verb get is extremely versatile and occurs with a number of different subcategorization frames. It can be used transitively, ditransitively, as a copula verb, or as a phrasal verb, and it can have causative or inchoative meaning; the subject can be interpreted as an agent (I got him a present), a goal (I got a present from Mary), or as affected entity (I got wet in the rain). Based on Quirk et al.’s (1985) classification, Johansson and Oksefjell (1996) list 17 different complement structures from the LOB corpus of written British English and the London-Lund corpus of spoken British English, among them the following: 38. This is similar to the situation in Old English: Helpan (“to help”) assigned dative case to its object and this case was not absorbed in the passive; i. e. it stayed with the internal argument, and there was no nominative noun phrase in the clause (Fischer et al. 2000).

92

(105)

Chapter 3 Ingredients of the English passive

a. b. c. d. e. f. g.

get +NP: I got this card from Africa. get + NP + XP: We really ought to get this thing straight. get + PP: I’m going to get into this. get + Particle: I’d just ring up and ask how she’s getting on get + Passive: I think they got paid for this. get + NP + NP: OK and just I’ll get you some. get + Infinitive: . . . when we got to know each other.

While this classification focuses on the category of the complement, Haegeman (1985) suggests an analysis that also takes into consideration the interpretation of the subject. Leaving aside idiomatic uses of get, she assumes that there are two (related) syntactic forms of get: transitive/causative get, illustrated in (106), and ergative/inchoative get, illustrated in (107). (The bracketed representations do not include the VP-internal subject position, but this could easily be integrated.) (106)

a. b. c.

John got IP[[his students] [to work on another topic]] John got SC[NP[his feet] AP[very wet]] John got SC[NP[his girlfriend]i VP[invited ti ] (by Susan’s friends)]

(107)

a. b. c.

Maryi got IP[ti to work on another topic] Maryi got SC[NP[ti ] PP[into trouble]] / AP[bored] Maryi got SC[NP[ti ] VP[invited ti (by Susan’s friends)]

As a transitive verb, get has two arguments, one of them external (see (108a)). The two arguments are the subject, interpreted as the agent or causer of the event, and a clause, which is the event that is caused. The clause may be a full finite clause or a small clause, a clausal structure lacking a projection of tense. The NP that follows the verb is not an argument of the verb, but the subject in a clause. Ergative get, on the other hand, has only one internal argument (small clause or infinitive) and it cannot assign structural Case. As a result, the subject of the small clause has to move to the subject position of the main verb get, where it receives nominative Case from tensed inflection. This movement leaves a trace, indicated in the examples in (107). Essentially, Haegeman treats get like other causative verbs that can undergo detransitivization, such as break or sink.39

39. Haegeman argues that her analysis should also be applicable to possessive get, but does not give a full account.

A re-evaluation of the get-passive

(108)

a. b.

93

causative (transitive) get: [j1 j2 ] ergative (unaccusative) get: [j2 ]

Both ergative and causative get can occur with a passive participle; compare (106c) and (107c). Should both be classified as get-passives? As pointed out by Huddleston and Pullum (2002: 1443), since get is a not an auxiliary or a passive marker, but rather a catenative verb, the passive construction is not the whole clause, but only the predicative construction following get: “the get-clauses here are not themselves passive, but merely contain passive clauses as complements.” Note that this holds for causative and ergative get alike. When one refers to the subject of the get-passive, one needs to clarify if one is talking about the subject of the passive construction per se, i. e. the subject of the small clause following get (the NP his girlfriend in (106c) and the NP-trace of Mary in (107c)), or about the subject of get, which is John in (106c) and Mary in (107c). The passivized verb itself is located inside a small clause in both cases. (109)

a. b.

ergative get-passive: IP [Maryi VP [ti got SC [ti invited ti ]]] causative get-passive: IP [Johnk VP [tk got SC [Maryi invited ti ]]]

The difference between (a) and (b) lies in the syntactic and semantic properties of get. In (a), get is an unaccusative (or “ergative”) verb. It does not have an external argument and cannot assign Case. Therefore, the internal argument of invite, the NP Mary, which cannot be Case-marked in its original position (due to Case absorption by the passive morpheme) has to move up to Spec, IP, moving through the subject position of the small clause. In (b), get is a transitive verb. It has an external argument (John) and can assign Case to the subject of the small clause, the NP Mary. The small clause itself contains a “bare” passive construction, as in (a). The NP Mary gets its theta-role as the direct argument of the verb invite, which has been passivized and cannot assign Case, just as in (a). As a consequence, the object has to undergo NP-Movement until it reaches a position in which it can be Case-marked. That position is the subject of the small clause. One can see that the difference between the two structures does not lie in the embedded passive, but solely in the argument structure and case assigning abilities of the matrix verb (not an auxiliary) get. However, in the literature, only constructions like (107c), but not their causative equivalents, are usually considered in a discussion of get-passives. For example, in their corpus linguistic study of the passive in English and Chinese, Xiao, McEnery and Qian (2006: 113) set the parameters of their search software in such a way that only immediate sequences of get and a past participle are included. They

94

Chapter 3 Ingredients of the English passive

accept the fact that undisputed get-passives such as got violently knocked over escape this filter, because it is more important to them to rule out that passives after causative get (“a structure conveying a causative rather than a passive meaning”) will come up in their word search. However, a sentence with two verbs can easily convey causative and passive meaning. Following Svartvik’s (1966) and Quirk et al.’s (1985) gradient approach to what should count as a passive, Collins (1996: 45) presents a list of getpassives that reaches from prototypical to peripheral. Prototypical get-passives have a by-phrase and can easily be turned into actives, see (110a); less prototypical get-passives do not have a by-phrase, but have an implicit agent (not Collins’s terminology) and could also be turned into actives, as in (110b); more peripheral get-passives do not really allow for an implicit agent reading, as in (110c), and at the very periphery are constructions in which the predicate inside the small clause is clearly an adjective, not a participle, as illustrated in (110d); all examples are from Collins (1996).40 (110)

a. b. c. d.

If I get phoned by a woman friend to come to dinner . . . With soaps you get labelled as your character . . . And sometimes I find when I start reading I get motivated. And everybody got awfully drunk.

Collins does not include passives after causative get – he only mentions those in which the NP following get is an anaphor, as in (111a), and points out that they are extremely rare (four tokens per one million words).41 He does not consider them passives because get in this case clearly functions as a main verb with a non-finite complement (Collins, 1996: 49). However, with respect to these characteristics, there is no difference between ergative and causative get. The only difference is that causative get cannot be replaced by be (because there is no causative version of be), but since be is not an integral part of the passive (see section 6 in this chapter), and since ergative and causative get are syntactically related, there is no reason to exclude passives after causative get from the discussion of passives after get. There is also no reason, really, to consider reflexive get-passives as something spe-

40. Examples quoted from Collins (1996) are from the Australian Corpus of English, the Australian ICE Corpus (the Australian version of the BROWN corpus), the Lund-London Corpus of British English, and the LOB Corpus of British English. 41. In the FROWN corpus, there is only one reflexive get-passive (. . . one of his men got himself kilt [sic]).

A re-evaluation of the get-passive

95

cial. The syntactic structure of (111a) – from Collins (1996) – and (111b) is exactly the same: they are both instances of passives after causative get.42 (111)

a. b.

But you also got yourself promoted to lieutenant. But you also got him promoted to lieutenant.

Overall, there are three reasons to include passives after causative get in the discussion of get-passives. First, they show the core characteristics of the passive: They have a participle, which cannot assign Case and thus induces movement within the passive construction, and they allow a by-phrase that spells out the implicit agent of the passive. Second, ergative get is not an auxiliary verb or a passive marker, it is a lexical verb derived from causative get (Haegeman 1985). Whatever properties ergative get has, they have to be seen in relation to the syntactic and semantic properties of causative get. This puts causative get in a position that is very different from other causative verbs which allow for passive complements, as, for example, have in I had my hair cut. A third reason comes from the history of the get-passive. According to G´ıvon and Yang (1994) the causative get-passive was a necessary step in the development of the ergative get-passive and passives after get first occurred as causative get-passives. 7.3. Verbal versus adjectival get-passives In the definition of the passive developed here, be is not considered an integral part of the passive construction, while the passive participle is. In section 3 of this chapter I discussed how and why to differentiate between verbal passives and the so-called adjectival passive, which is a copular construction and not really a passive. Since get can take adjectives, or rather small clauses with AP predicates, as its complement, it does not come as a surprise that there are also adjectival get-passives, such as get involved or get stuck. Although this is generally recognized – for example, Quirk et al. (1985: 161) suggest the term “pseudo-passive” – adjectival get-passives are still counted as instances of the get-passive, while passives after causative get, though clearly passive participles, are not (Xiao, McEnery and Qian 2006). The distinction between adjectival and verbal passive does not seem 42. Collins (1996) limits himself to counting only those instances of causative get that are reflexive (get oneself elected). That explains the rather small number of tokens in his corpus. If one counts all instances of passives after causative get, the number is, of course, much higher. For a frequency analysis based on the FROWN corpus see section 7.4.

96

Chapter 3 Ingredients of the English passive

to matter as much with respect to get. Why should that be the case? One answer to this question lies in the syntactic status of get: No matter which type of complement get is followed by, it is always a catenative verb, while be is a copula verb when it is followed by an adjective and an auxiliary verb in a passive construction.43 Another difference is aspectual: Sentences with get always express dynamic events, no matter whether one is dealing with an adjectival passive or a verbal passive. Sentences with be, on the other hand, express states when be is used with an adjective and are dynamic when be is followed by a verbal passive (based on a non-stative verb). This is due to the fact that get is never a stative verb, unlike copular be. Although get can be combined with adjectives, it cannot be combined with verbal passives that do not express dynamic events, as can be seen in the following examples (from Huddleston and Pullum 2002: 1442): (112)

a. b.

It was/*got believed that the letter was a forgery. Obviously, the manager is/*gets feared by most of the staff.

In (113), (a) is a complex resultative event, but that does not make the complement of get anything but an adjective (or, to be more precise, a small clause with an AP predicate). The same holds for (113b) and (c): They may look like passives, but they are not passives, as the sentences do not lend themselves easily to the inclusion of a by-phrase, an agent-oriented adverb, or any other indicator of an implicit agent reading. Also, degree adverbs, which typically modify adjectives, are possible (very frightened, so hung up). (113)

a. b. c.

We got very angry. My sister and I were alone for so long I began to get frightened. Don’t get hung up on a word!

While a dynamic reading with be is an indicator for the verbal passive, a dynamic reading with get is not an indicator of anything. However, although there is no ambiguity between state and event readings, participle-like forms after ergative get can still oscillate between an adjectival and a verbal reading, as in the following example: (114)

John got involved in the project very quickly.

The passive reading includes an implicit agent, i. e. someone who actively involved John in the project quickly; as an adjective, involved would describe 43. However, it has been suggested that both auxiliary and copula be take a small clause as their complement (Haegeman and Gu´eron 1999).

A re-evaluation of the get-passive

97

a state and states do not license agents, the whole sentence would still be dynamic, though not causative, and would mean that John became involved in the project quickly. Often the context will tell which reading is more likely. Huddleston and Pullum (2002: 1441) point out that “the clearest case of adjectival passives with get involve gradable adjectives,” such as frightened, drunk, or involved. (115)

a. b.

ergative get + adj. passive: IP [Maryi got SC [ti very scared]]. causative get + adj. passive: IP [We got SC [[the kids] involved]]]

According to the definition of the passive developed in chapter 2, neither of these constructions qualifies as a passive, get-passive or otherwise. They were not counted as get-passives in the analysis of the FROWN corpus here. Surprisingly, in many accounts of the get-passive, adjectival passives are not filtered out. Biber et al. (1999: 481), for instance, list examples like (116a) as get-passives, although they concede that they do not allow by-phrases and “are used as stative passives.”44 Collins (1996) calls sentences like (116b) “peripheral,” but still considers them get-passives, and the OED gives examples like (116c) through (e), which are all clearly adjectival, as no implicit agent reading is available, nor can the sentence easily be converted into active voice. The complex sentence expresses a dynamic event, but the embedded predicative structure (acquainted, entangled, used) does not. (116)

a. b. c. d. e.

And then we start to get involved in local society. (Biber et al., 1999) And everybody got awfully drunk. (Collins, 1996) A certain Spanish pretending alchymist . . . got acquainted with foure rich Spanish merchants. (OED, from 1652) We got entangled among a quantity of heavy drift-ice. (OED, from 1823) ‘The taste, I suppose, is peculiar’ . . . ‘Just at first,’ answered Campbell; ‘but one soon gets used to it.’ (OED, from 1848)

44. As Siewierska (1984: 139) points out: “Stative passives qualify as passive only . . . if the passive is defined either in terms of the semantic role of the subject (and then it would be impossible to distinguish passive clauses from certain actives) or if it is defined in terms of the presence of an auxiliary verb and a deverbal adjective.”

98

Chapter 3 Ingredients of the English passive

In sum, the verb get can be followed by verbal or adjectival predicates, just like be. It is important to keep adjectival structures separate because their syntactic, semantic and aspectual structure is different from that of passives. Within the passive analysis developed here, the most important difference is that adjectives do not have a suppressed external agent argument. Corpus analyses have shown that the get-passive occurs with some frequency only with a limited set of verbs. On the other hand, a survey of the use of get in the transcripts of the questionnaires the Dictionary of American Regional English (DARE) is based on shows that the ergative get-passive is actually quite productive. One hundred and eighty-four ergative get-passives in the past tense (i. e. got-passives) are spread out over 68 different verbs, many of them occurring only once.45 Among the more common combinations, Biber et al. (1999) list get married, get hit, get involved, get left and get stuck.46 Judging from the examples they give, quite a number of these occurrences are actually adjectival constructions. This is something to take into account when interpreting the observation that “most of the verbs that are moderately common with get passives have negative connotations, conveying that the action of the verb is difficult or to the disadvantage of the subject.” (Biber et al. 1999: 481). If the construction one looks at is not a passive to begin with, the generalizations about the kind of verbs that are typical have to be reexamined.

45. The DARE questionnaires contain 1847 questions about categories such as weather, food, beliefs, and relationships. 1002 questionnaires were completed across the United States in 1964. The responses form the basis of the Dictionary of American Regional English and its unique mapping system (explained in Cassidy 1985). 46. The questions that elicited the most responses involving get-passives were questions about humorous ways to refer to someone entering the state of marriage. Responses included more than 50 different expressions, among them got branded, got chained, got choked, got cornered, got hitched, got hooked, got knocked up, got lassoed, got murdered, got roped in, got saddled, got sentenced, got shorn, got smashed, got splashed, got spliced, got stung, got sucked in, got tangled, got tied, got trapped, got took, and got yoked. This may show as much about the informants’ sense of humor as it does about the nature of the get-passive. It should be noted that the actual questions (“What joking ways do you have around here of saying that peope got married?”) also included a form of get (Cassidy 1985). Also note that most of the expressions listed clearly have an implicit agent reading (unlike get married itself).

A re-evaluation of the get-passive

99

7.4. Get-passives in the FROWN corpus The following data are based on an analysis of the FROWN corpus of written American English (all text categories). Although the get-passive is known to occur more often in speech than in writing, a corpus of written American English was chosen to ensure comparability between get and be data. All instances of get and its inflected forms (got, gotten, gotta, gets, getting) were collected and manually classified, depending on their syntactic and semantic characteristics. As one would expect, the frequency of get in a corpus that is based on written English is lower than in a corpus that mixes written and spoken English. Overall, there were 1,670 tokens of get and its inflected forms, confirming the status of get as a high frequency verb. Of these, 193 (11.6 %) were potential get-passives, i. e. a form of get followed by what might be a passive participle (get translated, get inspected, get interested, get played out, get transferred).47 These occurrences were sorted into four classes, as illustrated below (all examples are from FROWN): (117)

a.

b.

c. d.

ergative get followed by adjectival passive: My sister and I were along for so long I began to get fright ened. (L22) ergative get followed by verbal passive: [P]ublic inequities of gender, race, or class get transferred into private relations. (F39) causative get followed by adjectival passive: We can get everybody engaged. (A05) causative get followed by verbal passive: “I’ll get it taken care of,” Lily offered. (N03)

Of the remaining 1,477 occurrences of get, i. e. the ones not involving a V-en form (a passive participle or an adjective derived from it), most (1,286 or 77 %) were instances of ergative get, such as get wet, get into trouble, and getting a dose of reality. Causative uses of get (get them out of a shell, get the children to school, get somebody to petition something, get the economy moving) occurred much less frequently (176 tokens or 10.5 % of all occurrences of get). 7.4.1. Ergative get-passives The ergative get-passive is the construction with which most analyses of the get-passive are concerned. Unaccusative (ergative) get is followed directly 47. I will refer to these forms as “V-en forms.”

100

Chapter 3 Ingredients of the English passive

by a passive participle. The subject of get receives its theta-role from the passivized verb and is moved up to Spec, IP in order to receive nominative Case, leaving an NP-trace after the participle, as illustrated in (109a) above. Of the 1670 tokens of get, 57 were ergative get-passives. They occurred mostly in informal writing, for example in passages of dialog in a novel. Only two of them included a by-phrase. (118)

a. b. c.

Not surprisingly, as Nikki notes, the dancers “get propositioned all the time.” (A29) Much of the savings gets passed along to consumers in the form of lower prices. (B10) Getting drafted must be much like what a frog feels, being taken into a larger organization – say, a heron. (G01)

The subject of the ergative passive is not always animate or a potential secondary agent, as can be seen in (118b). Of the 57 ergative get-passives, only 33 had animate/human subjects48 (this includes PRO-subjects in non-finite clauses as in (118c)). A tendency for verbs that express events which have negative consequences for the subject can be observed. More than half of the ergative get-passives express events of this kind (get cheapened, get hit, get ushered out, get blown off, get killed, get run over, get beat up, get shot, get fixed up, get squeezed, get caught, get sucked into), but there are also sentences that are perfectly neutral: a dish gets served, a house gets inspected, somebody’s chances to get re-elected are examined, and a topic gets written prominently into the political agenda. Animacy of the subject is an important factor here: If the subject is animate, usually a pronoun, there is a strong tendency for the get-passive to express an event with undesirable consequences. G´ıvon and Yang (1994: 139) consider the adversity effect for human subjects to be a consequence of the “adversative nature” of the getpassive. Since we only think in terms of adversity or advantage for humans, this reading will only be available if the subject is human. However, if the adversative interpretation is to be considered an integral part of the getpassive as a construction, it is quite surprising that more than a third of all get-passives do not fall into this category. The adversative character of the get-passive will be discussed in more detail in section 7.6.

48. Non-human subjects, such as the NP fish in fish getting grilled, can normally not be construed as responsible for the event.

A re-evaluation of the get-passive

101

7.4.2. Causative get-passives Causative get-passives are normally not considered in the discussion of the passive. They are subsumed under causative constructions. However, if causative get and ergative get are related and one can be derived from the other (as claimed by Haegeman (1985)), and if, furthermore, the complement of both types of get is essentially the same in both passives (a small clause), there is no reason not to include passives after causative get in a survey of get-passives. The examples in (119) illustrate the variety of passives after causative get. (119)

a. b. c.

d.

This gave her time to zip by Oasis to get a nail repaired. (K13) I’ll get it taken care of. (N03) I’m not much lookin’ forward to tellin’ the Colonel that one of his men got himself kilt [sic], and I didn’t do nothin’ about it. (N03) . . . you could get your clock cleaned by a Boy Scout if you started chasing him incautiously. (L12)

In these examples, the predicate after get is clearly verbal and it has an implicit agent reading. Unlike in the ergative get-passive, there are two agents here: the agent of causative get and the implicit agent of the passivized verb. The latter can also be realized visibly, as a by-phrase, but (119d) is the only example of its kind in the corpus. In the FROWN corpus, passives after causative get were less frequent than ergative get-passives (36 versus 57 occurrences). This number includes a single instance of a reflexive get-passive, listed in (119c). The passivized verbs did not belong to a specific semantic class, but they tended to be verbs that express the successful completion of an event or chore: Somebody has to get sneakers made, a bill passed, jobs done, houses inspected, a fingernail repaired and a clock cleaned. The same tendency can be observed in the DARE corpus: The questionnaires elicited 52 causative got-passives49 (compared to 184 ergative got-passives) and they mostly occurred as responses to the question of how one refers to the hardest part of a task being finished (of course, there were also many answers that did not involve a form of get). Answers included got the bad part done, got it made, got the worst part done. It does not come as a surprise that the external argument of get, i. e. the causer of the event, was generally animate. Of 36 causative get-passives, 33 49. In the DARE corpus, I only looked at get-passives that involved the form got, which was the most frequent inflectional form of get to occur with a participle.

102

Chapter 3 Ingredients of the English passive

had an animate (human) subject, usually a pronoun or PRO. There were only three inanimate causers, and they were all abstract nouns (tradition, our institution, and worries). What is more interesting to look at is the semantics of the NP that follows causative get. This NP, the subject in the passive small clause, receives its theta-role from the passivized verb, just as the subject of the ergative get-passive. The postverbal NP in the causative getpassive was generally inanimate. There were only three tokens of animate subjects in the passive small clause following causative get: get the mayor recalled, get Smoke Jensen riled, getting Bush re-elected. That is 8.3 % compared to 57.9 % in the ergative get-passive. This result is, in part, due to the high frequency of variations of get the job done (get things done, get zero done, get your work done, get it done). Of 36 causative get-passives, 14 are variations of get the job done.50 The high frequency of this expression also indicates that there is no preference for verbs that express unpleasant events or results which would be unfavorable for either the subject of the passive or the causer of get. If anything, there is a preference for positive events, e. g. verbs that express an urgency that an action be completed. 7.4.3. Adjectival get-passives Both ergative and causative get can be followed by adjectival passives. As with be, these constructions are not really passives, since the predicate is an adjective, not a participle, and there is no implicit agent reading. Quirk et al. (1985: 161–162) link get in this “pseudo-passive construction” to other copula verbs, including become, grow and seem. They stress that data like those in (120) should not be considered passives, but also point out that the adjectival get-passive and the ergative get-passive have in common a focus on the subject referent’s condition, “usually an unfavourable condition.” The following examples illustrate adjectival passives after ergative get. In (120b), stoned is not the participle of the verb to stone, it is clearly used in its adjectival meaning “drunk, extremely intoxicated” (OED). Obviously, this is not a get-passive. (120)

a. b.

. . . an explanation for why the essential human questions get lost in all discussions on the Holocaust. (G07) Mary just wanted to get stoned. (K26)

50. Some of these were difficult to classify as the context did not make it very clear whether an adjectival or a verbal reading was intended. I decided to count them all as verbal passives, as a by-phrase was generally possible and as they fitted the semantic pattern found most frequently in the causative get-passive.

A re-evaluation of the get-passive

103

Like ergative get, causative get can also be followed by a small clause with an adjectival predicate. This is illustrated in (121) – there is no implicit reading involved, engaged and involved are simply adjectives. (121)

a. b.

We can get everybody engaged in high-tech jobs . . . (A05) . . . he would like to get people involved . . . (A27)

In the FROWN corpus, there were 77 adjectival passives following ergative get. Combinations that occurred more than twice included get married (usually clearly in an adjectival reading), get involved, get lost, get drunk. Syntactically, these examples are just like regular adjectives (get ready, get angry, get hot). Due to the semantics of the adjective, the subject that the adjective is predicated of was usually animate, as illustrated in the following examples: (122)

a. b. c.

I never really did get accustomed to all the bowing. (R03) “Oh, you know,” she said, stroking Whiskers’ back,“sometimes people get tired of their dogs. (R08) She really gets upset at tractors on the road. (P24)

Adjectival passives after causative get, of which there were only 15 in the corpus, also primarily expressed states that are predicated of animate subjects (get someone interested/excited/involved). 7.5. Revisiting the responsibility factor It is often claimed that in the get-passive the surface subject is generally animate, allowing for an involvement reading of the subject (the subject is not just undergoing the event, but also actively pushing the event forward). G´ıvon and Yang (1994: 120) illustrate this point with the following contrastive pair: (123)

a. Mary was shot on purpose, the bastards! b. *Mary got shot on purpose, the bastards!

They argue that (123b) is infelicitous because in the get-passive the subject is construed as somehow responsible for the event (a reading that is not generally available for the be-passive). The agent of the passivized verb is implicit, i. e. Mary is not the one who fired the shoots, but the animate subject of the matrix clause is perceived as somehow responsible for the whole event having taken place (secondary agent reading). If the respon-

104

Chapter 3 Ingredients of the English passive

sibility factor is an integral part of the get-passive, one would expect that in the get-passive the subject is usually animate and that it occurs without a by-phrase, as by-phrases are explicit agent arguments and as such would interfere with the notion that the subject is responsible for the event. The OED, for example, lists 10 examples for get-passives, none of them with a byphrase, and in 8 cases the subject is animate.51 Quirk et al. (1985: 161) also state that “get tends to be limited to constructions without an expressed animate agent. . . . Get with an animate agent is not, however, unknown.” ArceArenales, Manuel and Fox (1994:12) analyzed 69 get-constructions from a non-specified “large body of transcribed American English conversations” and identified 32 passives (from their description it is clear that they refer to ergative get-passives), of which 28, or 88 %, had human subjects. They compared these numbers to that of be-passives in the same corpus (of which they identified 97) and found that the animacy rate for subjects in be-passives was considerably lower (46 out of 97, or 47 %). This pattern cannot be confirmed by data from the FROWN corpus. Only 33 out of 57 instances of the ergative get-passive (58 %) had animate subjects. This number is more in the region of what Arce-Arenales, Manuel and Fox found for the be-passive in their own study. The examples in (124) illustrate that ergative get-passives with inanimate subjects do not sound strange or inappropriate in the least, which indicates that the responsibility factor, if it exists, is not built into the get-passive as a construction, but may be a reading that becomes available under certain circumstances. (124)

a. b. c.

the serious problems of urban poverty will get written prominently into the political agenda. (J45) Much of the savings gets passed along to consumers in the form of lower prices. (B10) She is, in fact, a rare species in politics: a woman who knows how a familiar story will get played out. (F15)

The following examples show that an animate subject will not necessarily be interpreted as responsible for the event – there is no reason to assume that the bugs in (125b) are knowingly causing their own destruction, nor does the sentence in (125a) imply that the Reverend is manipulating the election.

51. However, of the 10 get-passives listed in the OED, 3 are clearly adjectival, compare the first example, dating from 1652: A certain Spanish pretending alchymist . . . got acquainted with foure rich Spanish merchants.

A re-evaluation of the get-passive

(125)

a. b.

105

And Rev Tillis will get elected. (L12) Behind us, the bug-zapper was working overtime: I can still hear the pop and sizzle of bugs getting fried. (K26)

The “involvement of the superficial subject” (Lakoff 1971: 55) in the getpassive does not seem to be an integral part of the construction, it is “merely a possible implicature” (Huddleston and Pullum 2002: 1443). This becomes evident when ergative get is contrasted with causative get. In (126a) the reading that Mary arranged the event may be a possibility, but in (126b) it is a certainty – this is an instance of causative get, and Mary is the agent of get, while herself gets a theta-role as the object of shot. The subject in (126a) may be construed as a secondary agent, but it is not an argument of the verb that has been passivized, i. e. it is not a true implicit agent. (126)

a. b.

Mary got shot. Mary got herself shot.

G´ıvon and Yang try to show that the subject of the get-passive can license a purpose clause, very much like an implicit agent in the passive. This would mean that in sentences like (127) the controller of the PRO subject in the purpose clause could be the subject of the get-passive (rather than the implicit agent), as indicated in (127a). (127)

a. b.

Radicalsi must get/?be arrested-IMPk [PROi to prove their machismo]. Radicalsi must get/?be arrested-IMPk [PROk to uphold the rule of the law].

However, according to Jaeggli (1986) and Roeper (1987), this secondary agent reading is also available, albeit marginally, for be-passives. Jaeggli (1986) claims that the passive in (128a) allows for a reading in which Mary is the one who wants to prove a point and therefore arranged for her arrest. Roeper points out that the secondary agent reading is even available in ergative constructions (which do not have a primary agent), as in (128c), and that purpose clauses can also be combined with states, as in (128d). He concludes that secondary agents and implicit agents cannot be the same kind of element and that there may simply be a tendency to construct the subject as responsible for an event wherever it seems to make sense. (128)

a. b. c.

Maryi was arrested-IMPk [PROk/?i to prove a point]. Theyk arrested Maryi [PROk/∗i to prove a point]. The boy fell to deceive his mother. (from Roeper 1987)

106

Chapter 3 Ingredients of the English passive

d.

Flamingoes are pink to attract the opposite sex. (from Roeper 1987)

While the secondary agent interpretation is not exactly prevalent in the passive, it is certainly excluded in the active. The reason is that a secondary agent has to be in the subject position, and in the active, in most cases, that position is already filled with the primary agent. Structural objects are not interpreted as secondary agents. There is a strong association between the structural position of a subject and the semantic interpretation of an agent, especially if the NP in question is animate (Oosten 1986). This connection can also be observed in language acquisition: When children acquire the passive, they have to learn to dissociate the subject from the agent interpretation. This is easier for them if the passive is irreversible, i. e. if knowledge of the world tells them that the subject cannot be the agent (Pinker, Lebeaux and Frost 1987), as in The apple was eaten by the horse. In a reversible passive, e. g., The cow was kicked by the horse, the child has to rely on syntactic knowledge only to assign the correct meaning of the sentence, and if the passive rule has not been acquired yet, the child will follow the default alignment of subject and agent and will point to a picture of a cow kicking a horse. The fact that the get-passive is associated with a secondary agent reading more easily than is the be-passive can be connected to the existence of causative get. The subject of be is never a causer, the subject of get can be. Historically, the link between the two versions of get is the reflexive use of get (get thee gone), which syntactically is an instance of the causative passive, but which in semantic terms is considered to be an instance of detransitivization, as it does not introduce a second participant upon which that the agent acts. According to G´ıvon and Yang (1994: 144), the development went from get oneself somewhere, essentially a causative use of get, to get somewhere, essentially an ergative use of get. While the causative-reflexive use of get is still grammatical, it is not very frequent. There is only one example in the FROWN corpus. So far, the responsibility factor has been discussed for the ergative getpassive. The situation is different for the causative get-passive. For causative get-passives, one has have to differentiate between the subject of get, which is a causer and is most often animate (in 50 out of 57 instances, or 88 % of all occurrences in the FROWN corpus), and the subject of the embedded passive, which is the NP that is following the verb get. In the latter case, only 18 % of subjects are animate. What is interesting about this is that the NP following causative get has the exact same D-Structure posi-

A re-evaluation of the get-passive

107

tion as the NP that ends up as the subject in the ergative get-passive. In the ergative get-passive, this NP is typically animate (as in (129b)), while in the causative get-passive it is inanimate in the majority of cases (as in (129a)). (129)

a.

b.

The plea bargain angered Little’s family, who said they didn’t understand how Lilly could get a first-degree murder charge reduced to attempted murder. (A30) “We did get ushered out rather firmly,” she agreed. (L13)

Based on this observation, one could argue that there is really no point in comparing the ergative to the causative get-passive – the verbs that occur are different, the subjects of the passive are different, and the adversity factor can only be observed in one of the two passives. Another way to look at this is to take into consideration the strong association that exists between subjects and agents. In the ergative get-passive the object of the verb ends up as the subject of get, while in the causative get-passive it is only moved to the subject of the small clause, i. e. to a postverbal position. In the causative get-passive there are two nominal arguments (plus the implicit agent in the passive). This is a situation which lends itself to an asymmetrical construction of an event, with an agentive proto-agent and an affected proto-theme (Dowty 1991). In the ergative get-passive, on the other hand, there is only one visible argument. It is an underlying object, but it has been moved to the subject position. Since get in principle is a verb that can be associated with agentive subjects (unlike be), a quasi-agentive reading may be available for animate subjects, but it is not really built into the construction. 7.6. Revisiting the adversity effect The high “involvement” (Lakoff 1971) of the subject in the get-passive is also discussed in terms of the effect that the event has on the subject. It has been observed that the get-passive often has “negative connotations, conveying that the action of the verb is difficult or to the disadvantage of the subject” (Biber et al. 1999: 481)52 and that the subject in the ergative getpassive is “somehow responsible for his own misfortune” (G´ıvon and Yang 52. This generalization does not quite seem to match their own data, since the front runner in their corpus is the combination get married, which occurs more than four times as often as any other get-passive (more than 20 occurrences per one million words). However, it is safe to guess that many of these instances are adjectival passives.

108

Chapter 3 Ingredients of the English passive

1994: 120). G´ıvon and Yang’s argument, that the get-passive has been adversative from the beginning on, is based on their corpus of five Shakespeare plays, one play by Christopher Marlowe, one novel by Laurence Sterne, an autobiography by Benjamin Franklin, one novel each by Charles Dickens, Mark Twain, and Larry McMurtry, and one play by Sam Shepard. The first get-passive they find is from Sterne and it is not adversative (he had well got announced), and it is really only in the colloquial speech depicted in Huckleberry Finn that the get-passive expresses, more often than not, adversative (11 out of 13 tokens) events. Early examples listed in the OED are also not uniformly adversative (for example, I got supplied with bread, cheese and a pint of wine, 1814). Involvement need not be negative involvement. Pullum and Huddleston (2002: 1442) note that “[g]et-passives are characteristically used in clauses involving adversity or benefit” (emphasis added) and quote examples like the following: (130)

a. b.

Kim got promoted. My letter got published.

At first sight, data from the FROWN corpus seem to confirm what has been said about the adversative character of the get-passive. Of the 57 ergative get-passives, about 2 in 3 have negative connotations (get caught, get overrun, get overlooked, get cheapened, get squeezed, get zapped, get blown to pieces, get sucked into, get pressured into, get killed, get fired, get struck). However, the remaining ergative get-passives are perfectly neutral. What is more, the causative get-passive does not show any preference for adversative interpretations whatsoever. Judging from data from the FROWN corpus, it seems to be the case that it is the informal, colloquial character of the getpassive, rather than the construction itself, which favors colloquial expressions of something being strongly affected. This assumption is supported by the stylistic match of verb and subject in the get-passive, as illustrated in (131). A formal subject, like calcium, is paired with a semantically neutral formal-register verb, like absorbed, while a subject typical of informal registers, like a couple of buddies, is paired with the decidedly colloquial expression get blown to pieces. (131)

a.

b.

Spinach contains a great deal of calcium, but has very low bioavailability – only about 3 percent of its calcium gets absorbed into the body. (G03) . . . to get off their self-righteous asses and learn about war first hand. . . . To watch a couple of buddies get blown to pieces. (C02)

Summary

109

The get-passive often occurs in informal contexts and its most frequent type of subject is a personal pronoun. Taking into consideration the richness of colloquial expressions that a language usually has for expressing things going wrong, it is plausible to assume that it is the informality of the getpassive that favors its use in personal accounts of events which are perceived as not going very smoothly.53 Data from the DARE questionnaires support this assumption. The purpose of the questionnaire the dictionary is based on was to elicit immediate, spontaneous responses, “as close to normal local usage as an investigator can come to hope” (Cassidy 1985: xii). An event such as having to pay money for something is not really all that harmful, but speakers may emphasize its unpleasantness, and English provides many colorful expressions to do so. Answers to the question of how one could rephrase the statement I paid ten dollars for it (note that there is no indication that this price is too high) included I got drownded [sic], got gyped, got hooked, got nicked, got robbed, got soaked, got stuck for. Even a joyous event, like getting married, is described, if jokingly, in terms of violence and betrayal, and the subject takes the role of a victim rather than a secondary agent (got branded, got chained, got choked, got cornered, got roped in, got spliced, got trapped). The get-passive is part of the English vernacular, and within this register, colorful expressions referring to impactful bodily contact are abundant.

8. Summary In this chapter, I looked at the structural ingredients commonly found in the English passive. Special emphasis was placed on the verb, as the structural and semantic core of the passive construction. While there are passives without subjects that correspond to an underlying object, passives without by-phrases, passives without be, and passives without an auxiliary at all, there is no passive without a verb, or rather a participle form. What all passives have in common in terms of semantics is that the external argument of the verb need not be expressed explicitly but remains an integral component of the semantics of the construction, a so-called implicit argument. This invariant semantic feature of the passive will be related to the invariant structural ingredient of the participle in chapter 4.

53. This fits with a result from Herold (1986), cited in G´ıvon and Yang (1994), who found that the get-passive is more frequent and more clearly adversative in the speech of working-class speakers.

110

Chapter 3 Ingredients of the English passive

If the implicit argument is expressed explicitly, it takes the form of an adjunct that is placed at the end of the clause. However, this option is usually not employed. Roughly four out of five passives occur without by-phrases, which indicates that the main function of the passive is not to shift a structurally complex NP to the end of the clause, in line with the principle of end-weight (although this can certainly be the point of choosing the passive over the active in individual cases), but rather to make use of the option to express an agent or causer as an implicit argument. Thus, both in structural and functional terms, the implicit argument seems to be at the center of the passive. The by-phrase, on the other hand, is a derivative characteristic of the passive. The passive is not the only construction in English that allows for the demotion of agent expressions. Chapter 5 will discuss some of the competitors of the passive in a specific register. Another side-effect of changing the status of the external argument of the verb to that of an implicit argument is that in the passive the subject position, normally associated with the external argument, is freed up for another NP. Whether this NP is a direct object, an indirect object, the object of a preposition, the subject of a lower clause, or an expletive element, depends on the syntactic class of the verb. Again, it is the verb whose syntactic features determine the structure of the passive to a large extent. In English, the subject position is usually linked to the topic of a clause, and the passive is a way of making an argument that is not the agent the topic. A construction that deserves specific attention is the passive in participle clauses, notably relative clauses (a letter written by my great-grandfather). A participle clause does not require a finite verb and a relative clause in English does not require an overt relativizer. As a consequence, the passive here is a compact way to create cohesion by tying in information from the modifying clause with the head noun (letter) without any overt intervening elements. I pointed out that corpus studies of the passive often overlook this construction because it does not include the auxiliary be (and is thus difficult to track automatically), but another way of looking at it is that be is not really a marker of the passive, it is an element that may be structurally required in many, but not all contexts in which the passive occurs. Be does not make any semantic contribution to the passive and does not make the passive any less dynamic than the corresponding active. Since I consider the lexical verb the structural core of the passive, I discussed which verbs form passives and which verbs do not. An uncontroversial point is that verbs that do not have an external argument, i. e. unaccusative verbs, cannot be passivized. However, this does not mean that every verb that has an external argument can form a passive. Unergative

Summary

111

verbs, for example, do not passivize in English unless they are used transitively (in which case they are not, technically speaking, really unergatives anymore). Since they readily form passives in languages that are closely related to English (such as Dutch or German), it is unlikely that the reason for the ungrammaticality of impersonal passives in English lies in the semantic properties of these verbs or in a restriction that could be connected to the passive as a construction. After reviewing several approaches to the ungrammaticality of the passive in English, I sided with those who argue that the ungrammaticality of impersonal passives is related to syntactic properties of the so-called dummy subject it, which seems to require the formation of a chain with a clausal complement. Hence, it can fill the subject position if the verb is followed by a clause (It was expected that . . . ), but it cannot fill the subject position if the verb has no complement (*It was smiled). While English is quite restricted with regard to passives formed on the basis of one-argument verbs, as a prepositional stranding language it allows a type of passive that is rather rare across languages, the prepositional passive. These passives posit a problem for Case Theory, since an NP that is the complement of a preposition should be Case-marked by the preposition and should not be able to move to the subject position and pick up a second Case (nominative). Prepositional passives require the rule of reanalysis (Hornstein and Weinberg 1981), which essentially demotes the preposition to an element that incorporates into the verb, very much like a particle. What is notable here is that prepositional passives work best if the moved NP is truly “affected” (Anderson 1978) by the action expressed by the verb, i. e. if it behaves semantically like a direct object of the verb. The affectedness factor also comes into play when one looks at ungrammatical passives formed on the basis of transitive verbs that express symmetrical states or events (*I am resembled by my brother). These observations are reminiscent of the discussion of the acquisition of the passive in chapter 2: In young children’s grammars, the first productive passives are those in which the passivized event has a prototypical agent acting willfully on a prototypical patient. The less asymmetrical the event, the more children struggle with passives. While states can be passivized in adults’ grammars, there needs to be some sort of asymmetry to begin with. The passive then decreases this asymmetry by making the agent less and the theme more prominent. In this chapter, I also revisited some of the claims that have been made about the get-passive, claims that seem to indicate that the semantics of this construction is different from that of the be-passive and that some of its characteristics (adversative interpretation, subject responsibility) come from the construction itself rather than from its structural ingredients. The

112

Chapter 3 Ingredients of the English passive

get-passive is also interesting to look at because it shows that the passive, per se, is not necessarily associated with formal registers in English, as the getpassive mainly occurs in spoken, informal English and often with verbs that express bodily contact. I argued against a Construction Grammar approach in which the semantic and syntactic characteristics of the get-passive would be ascribed to the Construction rather than to the properties of get and the passivized verb. Instead, I proposed that the properties of the get-passive have to be seen in relation to those of the lexical verb get and the regular passive, an approach that led to the inclusion of passives after causative get. A review of the literature showed that while causative get is hardly ever discussed in the context of the get-passive, quite a number of adjectival constructions, specifically adjectives that can only be attributed to sentient beings, such as annoyed and worried, are included in the discussion of the getpassive, which may have contributed to the generalization that the subject of the get-passive is generally animate and is construed as somehow responsible for the event. Inanimate subjects simply don’t make good secondary agents. Data from the FROWN corpus showed that, if one rules out adjectival constructions, the animacy rate of the subject is not quite as high as has been claimed and that get-passives are also used to refer to perfectly neutral or even positively connotated events (get elected). The responsibility reading of the subject seems to be purely contextual and very different from that of an implicit agent, which is obligatory and represented in the verb’s argument structure. I suggested that the secondary agent reading may be influenced by the existence of causative get (as in get the job done). A major difference between get and be is that the subject of be never qualifies as an agent, while get can be used as an agentive verb. Hence, the subject of get is more easily interpreted as a secondary agent. The trend for the get-passive to express negative or even violent events has to be seen in the context of the register in which the construction occurs. Data from the corpus used for the Dictionary of American Regional English showed that there are dozens of expressions involving get that refer to someone entering the state of marriage – many of them sound quite violent, but speakers use them jokingly. Overall, the discussion of the get-passive supports the assumption that the passive can best be understood by looking closely at the verb semantics, the representation of agenthood, the match between subject and verb, and the use of a construction in a specific context and register. The ingredients that are needed to understand the construction are right there.

Chapter 4 The implicit agent in English passives 1. Introduction There is a general consensus that the passive is an agent-demoting construction (in terms of syntactic and semantic prominence) or agent-backgrounding construction (in terms of information structure). The point of this chapter is to support a more specific claim: In the passive construction, the external argument is demoted to a particular syntactic status, which will be referred to here as that of an “implicit argument.” The purpose of this chapter is to demonstrate the reality of an implicit argument in the passive and to give an account of its syntactic potential and representation. It will be shown how the implicit argument status is directly related to the process of passivization itself and that a number of characteristics of the passive can be derived from this operation. As a starting point, let us use “implicit,” rather untechnically, for an entity that is grammatically represented but does not correspond to any overt category in syntax, and “argument” for an entity that is selected by the verb as an integral part of the propositional content of the sentence and that is not just inferred from context.54 The effect of the implicit argument in passives is illustrated in (132). Unlike ergatives or adjectival passives, verbal passives are compatible with agent-oriented adverbs like (un)willingly, deliberately, and voluntarily, even if there is no explicit agent in the sentence: (132)

a. The government decreased the price unwillingly. b. The price was decreased unwillingly. c. *The price decreased unwillingly.

Data like these show that the function of the passive cannot be to maximally background the external argument, but to background it to a certain extent – to an element with specific syntactic and semantic characteristics, the “implicit argument.”

54. Roeper (1984: 37) makes the distinction between an “implied” and an “inferred” theta-role. The former relates to a syntactically represented entity and the latter to information that can be inferred from context. The claim here is that the implicit argument is “implied,” not just “inferred.”

114

Chapter 4 The implicit agent in English passives

In this chapter, I will first present data from the original research on implicit argument effects from syntactic theory (Jaeggli 1986; Roberts 1987; Baker, Johnson and Roberts 1989), complemented by data from psycholinguistics (Mauner, Tanenhaus and Carlson 1995; Mauner and Koenig 1999), which both point in the same direction: The implicit argument in the passive is not just a conceptual construct, it is an integral part of the construction and consequently of native speakers’ grammar of English. In order to illustrate the last point, I will add some observations on the use of passives by native speakers versus that of second language learners. There is a type of ungrammatical passive construction that is used deliberately by native speakers, presumably with the purpose to add an implicit agent reading to an otherwise unagentive event. If we follow the generative analysis of the passive, the representation of an external argument as an implicit argument is an integral part of the passive construction in English. It is contingent on lexical properties of the passive morpheme, a so-called “thematic affix” (Roeper 1987), or of a lexical rule of passivization. The core of Jaeggli’s (1986) widely accepted analysis of the passive is that the passive morpheme may behave like an ordinary argument in that it can be (or needs to be) assigned a theta-role and abstract Case in order to be interpretable. But since it is a bound morpheme, it will not actually behave like a full argument, it will only be an “implicit argument” with restricted syntactic potential. According to this analysis, the distinction between “short passives” (passives without a by-phrase) and “long passives” (passives with a by-phrase), which is very important in functional approaches to the passive (Thompson 1987; Biber et al. 1999), is not fundamental – the external argument will be represented anyway, no matter if there is a by-phrase or not. The second part of this chapter will be concerned with different ways to represent the implicit argument. I will discuss two major approaches, broadly characterized as syntactic and lexical, and will argue that the implicit argument reading arises from saturating an argument position in the lexical entry of the verb (rather than from assignment of a theta-role to an affix, as suggested by the syntactic approach).

2. Evidence from syntax If the implicit argument is not just a conceptual construct but a syntactically active element, it should underlie principles of grammar and there should be certain effects of its presence. These effects will be explored in the following sections.

Evidence from syntax

115

2.1. Purpose clauses and control Among the things that an implicit argument in a passive construction can actually do in syntax is control a PRO subject in an adjunct purpose clause. The term “control” is used to describe “a coreference relation between two arguments that bear independent theta-roles” (Felser 1999).55 Unlike a trace, PRO is a base-generated category, hence the individual theta-role (traces share a theta-role with their antecedent). It is an argument of the verb and it fills the subject position in infinitives or adjunct small clauses. As an NP without phonetic representation it does not underlie the Case Filter. PRO can be controlled by the internal or the external argument of the verb in the matrix clause; see (133a) and (b), or it may be uncontrolled and have arbitrary interpretation; see (133c). (133)

a. Shei promised himk [PROi to take care of herselfi /*himselfk ] b. Shei persuaded himk [PROk to take care of *herselfi /himselfk ] c. It was unclear [how PROarb to behave oneselfarb /*himselfk ]

Traditional grammar also uses the term “control,” though in a slightly different way. In a grammar without empty categories, the infinitive in (134a) can only be a VP, and the NP Liz figures as the experiencer of hope and the agent of convince at the same time: “we will . . . say that the missing subject of the non-finite clause has a controlled interpretation” (Huddleston and Pullum 2002: 1194). By contrast, in generative grammar, rather than saying that there is a “missing subject” one would posit a non-overt subject, PRO, which is co-referent with the subject in the matrix clause. The Theta Criterion (Chomsky 1981) rules out that a single argument can have more than one theta-role. As a consequence, the agent role of promise and the agent role of take have to be assigned to two different NPs. One of them is the (overt) subject in the matrix clause, the other one is PRO, the subject in the embedded clause. (134b) illustrates that PRO is co-referent with the subject of the matrix clause, but it is a separate NP.

55. Chomsky (1981: 76) assumes that “the choice of controller is determined by jroles or other semantic properties of the verb, or perhaps pragmatic conditions” and specifically rules out structural requirements like c-command as a relevant condition for the control relationship. His opinion that the control relationship is determined thematically is shared by Jones (1988) and Williams (1987). Others, however, have argued that the choice of the controller is determined syntactically (Manzini 1983; Larson 1991).

116

(134)

Chapter 4 The implicit agent in English passives

a. b.

Liz hoped VP [to convince them] Lizi hoped CP [IP [PROi to convince them]]

The crucial differences between (134a) and (b) are (1) the status of the infinitive (a full clause in (b)), and (2) the 1:1 relation between theta-roles and syntactic constituents in (b). Even though the subject of the matrix clause and the subject of the infinitive are co-referent, they are two separate syntactic constituents (Hornstein 2000). It seems that the implicit argument in the passive can function as the controller for PRO, at least in purpose clauses. In (135a) the external argument of the matrix verb, sell, obligatorily controls PRO in the purpose clause: The referent of they (the people who sell the book) is identical with the referent of PRO (the people who want to make a profit). In (b) the external argument of sell is not in an argument position anymore, yet one can still get the same controlled interpretation for PRO. And in (c) there is not even a by-phrase – if PRO is still controlled, one needs to assume that the external argument of sell is present implicitly.56 An instance of control through an implicit argument that does not involve a purpose clause is given in (135d). (135)

a. b. c. d.

Theyi sold the books [PROi to make profit] The books were sold by themi [PROi to make money] The books were sold IMPi [PROi to make money] The game was played IMPi [PROi wearing no shoes] (from Roeper 1984)

Other agent-demoting constructions, such as middles and ergatives, do not license a purpose clause – presumably because there is no implicit agent in the matrix clause: no controller, no control. (136)

a. b.

The door opened (*[PRO to let in some fresh air]) Crime stories sell easily (*[PRO to make money])

Jaeggli (1986: 612) concludes: an Agent theta-role is present in passives even in the absence of a by-phrase, whereas this is not the case in middles. The lexical rule that forms middle verbs . . . must involve the deletion of the external j-role of the verb . . . In passives the external theta-role is not deleted.

56. “IMP” represents the implicit argument in a non-technical way, leaving open its exact syntactic status.

Evidence from syntax

117

It should be noted that not every passive can occur with a purpose clause – clauses that do not express dynamic events or do not have an agent argument (the one who acts purposefully) do not license a purpose adjunct.57 (137)

a. The books stood on the shelf (*PRO to amuse me) b. *Peteri received the books [PROi to impress Mary] c. *The books were received (by Peteri ) [PROi to impress Mary]

(137a) is ungrammatical because the NP The books is construed as the controller of PRO, but this NP does not qualify as an entity that would purposefully do anything – no agent, no purpose clause. The same holds for (137b): Though the NP Peter might qualify as a potential agent in another context, there is no agent reading available for the subject of receive, in other words: A purpose clause is again not licensed. Since passivization certainly does not add any agent-like qualities, the passivized version of (137b) is equally ungrammatical. If the implicit agent in the passive is somehow related to the passive morpheme, as argued for by Jaeggli, it is not unreasonable to check if there are other morphemes that would license implicit agents as well, as suggested by Roeper’s (1987) thematic affix analysis. The data in (138) seem to confirm that the presence of a specific affix may be a relevant factor indeed: While gerunds and nouns ending on -ion, which can have an eventive reading (Grimshaw 1990), may occur with a purpose clause, nouns ending on -er, which denote objects or persons, may not. (138)

Meat-eating-IMPi [PROi to gain weight] is not healthy. The destruction-IMPi of the bombs [PROi to avoid a war] seemed unlikely. c. *The destroyer-IMPi of the cages [PROi to let out the animals] was caught.

a. b.

However, the control abilities of implicit arguments are severely limited. For a start, the implicit argument is not obligatorily the controller of a PRO in a purpose clause. Chomsky (1981: 77), for instance, assumes that in sentences like (135c), PRO gets arbitrary interpretation. Secondly, and even more importantly, subject control verbs cannot be passivized (Visser’s Generalization), which seems to indicate that the implicit argument cannot control a 57. Roberts (1987) assumes that the basis for a stative reading is the non-coindexing of the verb and INFL. In a middle construction, this kind of coindexing is not available anymore. If there is no action, there can be no agent.

118

Chapter 4 The implicit agent in English passives

PRO subject in a complement clause. Note that object control is not affected by passivization, as the controller is still explicitly expressed; see (139c). (139)

a. Ii promised Johnk [PROi to wash the car] b. *Johnk was promised-IMPi tk [PROi to wash the car] c. Johnk was persuaded-IMPi tk [PROk to wash the car]

In an analysis without an implicit argument, it is easy to explain the ungrammaticality of sentences like (139b): Promise is a so-called subject control verb, which means that the external argument of promise obligatorily controls the PRO subject in the non-finite complement clause. “Under this view, the reason why subject control verbs fail to passivise is that passivisation removes the controlling argument so that PRO remains unidentified” (Felser 1999: 151). However, once we assume that the obligatory controller for PRO is not removed in the passive (as in the analysis of Borer (1998); see section 6 in this chapter), the grammaticality contrast in (139) cannot easily be accounted for. Jaeggli (1986: 615) addresses the problem by making a distinction between “argument control” (control by full arguments) and “thematic control” (control by implicit arguments). The former is required for control of a PRO subject in a complement clause, while the latter suffices for control of a PRO in an adjunct clause (e. g., a purpose clause). The following examples are adapted from Jaeggli (1986). (140)

a. Maryi promised Billk [PROi to leave Harryj ] b. *Billk was promised-IMPi tk [PROi to leave Harry j ] c. Johnk was promised-IMPi tk [that Mary would leave Harry] [PROi to get himk PROk to stop crying] d. The gifts were brought-IMPi [PROi to impress the grandparents] e. *[The gifts]k were brought-IMPi [PROk to be admired tk (by the grandparents)] f. Billi tried [PROi to be introduced ti to Mary]

According to Jaeggli’s proposal, (140b) is ungrammatical because PRO is the subject of a complement clause and as such needs to be argument controlled by the external argument of promise. Since the external argument is implicit, argument control is not possible. (140c), on the other hand, is grammatical because the infinitive is an adjunct here and consequently, PRO only needs to be thematically controlled (in the given reading the controlling element is the implicit agent of promise). (140d) is grammatical, as

Evidence from syntax

119

we are dealing with an adjunct infinitive here, which also only requires thematic control. Puzzling, however, is the ungrammaticality of (140e): Here, the adjunct clause is a passive construction itself, with PRO originating as the object of admire (and then undergoing movement to the subject position, to fulfill the PRO-Theorem, the requirement that PRO not be governed). As a subject in an adjunct clause, PRO should only require thematic control, but the sentence is clearly ungrammatical, and it is not immediately obvious why (see Jaeggli (1986) and Roeper (1987) for a c-command related explanation). (140f) illustrates argument control: The infinitive here is in the complement position, and its subject is controlled by the subject of the matrix clause. Another limitation of control by an implicit argument is illustrated by the following set of data, which involve predicative adjuncts. These are generally analyzed as small clauses with PRO subjects (Aarts 1992). (141a), for instance, means that they sang the song while they were drunk (not while some unspecified person was drunk). According to the Theta Criterion, the NP they cannot be the external argument of sing and of drunk at the same time. Therefore, the predicate drunk must have a subject of its own, which is PRO, controlled by they. Since we are dealing with an adjunct structure here, one would expect that this kind of PRO can also be controlled by an implicit argument. This is confirmed by (141b). However, if the implicit argument could freely control the PRO subject in any adjunct small clause, one would expect (141c) to be grammatical – but it is not. What is more, the grammaticality of (141d) indicates that the PRO subject can be controlled by the internal argument of the verb (which becomes the structural subject in the passive) as well. It seems, then, that thematic control is not quite as predictable as one would wish. (141)

a. Theyi sang the song [PROi drunk] b. [This song]i must not be sung-IMPk [PROk drunk] c. *[The room]i was left-IMPk [PROk drunk] d. [The room]i was left-IMPk [PROi empty]

The examples in (142) also show that a PRO subject in an adjunct small clause can be controlled by different elements in the dominating clause: either the subject (PROi ) or the object (the roomk ). Whatever licenses the small clause and its PRO subject, it does not seem to be a clear-cut structural requirement. (142)

a. b.

Theyi promised [PROi to leave [the room]k [PROk empty]] Theyi promised [PROi to leave [the room]k [PROi sober]]

120

Chapter 4 The implicit agent in English passives

To sum up: The control data show that the agent argument is still somehow accessible in the passive construction. However, since the structural prerequisites of control are far from clear and since the controlling potential of the implicit argument is limited, we do not know the exact syntactic status of this element yet. 2.2. By-phrases In contrast to other detransitivized constructions, the external argument of the verb can be realized overtly in the passive. This results in a “long passive,” a passive with a by-phrase. The NP inside the PP has exactly the same interpretation and underlies the same selectional restrictions as the subject in the active, experiencer in (143a), agent in (143b). It seems, then, that the by-phrase gets its interpretation directly from the lexical verb, making it an argument of the verb. On the other hand, though, the by-phrase is generally optional. Syntactically, therefore, it seems to behave like an adjunct (see also Zubizaretta 1985). Grimshaw (1990) uses the term “a-adjunct” (argument adjunct) to capture this seeming contradiction. (143)

a. b.

(144)

Everybody enjoyed the party./The party was enjoyed by everybody. The little boy broke the glass./The glass was broken by the little boy.

a. *The glass broke by the little boy. (meaning: The boy broke the glass) b. *This kind of glass breaks easily by little boys.

Jaeggli (1986: 599) points out that this kind of by-phrase is only possible in passive constructions. While there may also be by-phrases in active clauses, “active sentences may contain by-phrases only as locatives or instrumentals,” as illustrated in (145).58 In these examples, the content of the by-phrase could also be expressed by another PP, something like “on a train.” (145)

a. b.

John and Mary met by the river. John is traveling by train.

58. This claim seems to be reaching too far. Some adjectival passives do license byphrases that do not have instrumental or locative interpretation: We remained unnoticed by the guests (from Huddleston and Pullum 2002), and there are also nominal expressions like a chair by Breuer. Note, however, that a purpose clause is not possible here: * a chair by Breuer to use new materials.

Evidence from syntax

121

At first glance, the possibility of a by-phrase looks like strong support for the implicit argument analysis. It seems to be an overt version of the implicit argument, similar to a pronominal subject in a pro-drop language, which may or may not be overtly realized.59 However, the optionality of the by-phrase poses a serious problem for Jaeggli’s analysis of implicit arguments: If the external theta-role of the verb is already assigned to the passive morpheme, resulting in an implicit argument, how can it still be available for assignment to an NP inside the by-phrase? There are several ways out of this dilemma: Jaeggli himself suggests a process of “theta-role transmission” – the theta-role that is absorbed by the passive morpheme is “transmitted” from the morpheme to the NP inside the by-phrase through a multistep percolation mechanism, which is possible because the by-phrase is subcategorized by the passive morpheme. [T]he argument structures of both the verbal head and the passive suffix percolate to the branching node dominating them. The external theta-role is then assigned to the passive by-phrase. Assuming that it is assigned to the PP, it percolates to the head of the PP, the preposition by, and from there it is assigned to the object of by. (Jaeggli 1986: 600)

Roberts (1987: 5859) proposes that the by-phrase and the passive morpheme should be treated as an instance of clitic doubling. In languages like Spanish, clitics and their non-clitic double can co-occur, provided that the non-clitic has its own Case assigner (Kayne-Jaeggli Generalization). (146)

(Le) puso comida al canario. to-him put-3.sg food to-the canary ‘S/he gave food to the canary‘ (Halpers 1998: 107)

What is problematic about this approach is that the passive morpheme does not really seem to behave like a clitic, as it does not have any recognizable features of a pronoun (specification for person and number, for instance). The clitic approach thus totally depends on Jaeggli’s (1986) absorption analysis: He assumes that the passive morpheme is assigned the theta-role that would otherwise be assigned to the external argument. It is through this process that the passive morpheme acquires argument-like (and clitic-like) features – there is no independent evidence for its pronominal status.

59. Examples of pro drop languages are Latin, Italian and Spanish: In Italian, for instance, Parlo and Io parlo both mean “I speak,” and under normal circumstances the overt pronominal subject would be “dropped.”

122

Chapter 4 The implicit agent in English passives

An altogether different approach to the by-phrase problem is taken up by Booij (1992), Grimshaw (1990), Levin (1995), and others, who assume that an implicit argument is an entity in the verb’s lexical representation, which may or may not be realized in syntax. Both accounts will be compared in more detail in section 6 of this chapter. 2.3. Anaphors and binding Binding Theory offers a framework to target the syntactic potential of implicit arguments in terms of reference and structure. While the structural constraints on control are not quite clear, the principles that determine the interpretation of overt pronominal elements are very explicit. Anaphors, for instance, have to be locally bound by an argument; see (147a). If the implicit argument in the passive can bind an anaphor, we could conclude that it c-commands the anaphor and that it is in an argument position. Sentences like (147b) seem to show that the implicit argument can indeed bind an anaphor, at least one with arbitrary reference (oneself). (147)

a. b. c.

Theyi kept the privilege to themselvesi /*oneselfi Such a privilege should be kept-IMPi to oneselfi /*themselvesi This privilege should not be kept to himselfi /*oneself by Billi

If we assume that the anaphor has to be bound in its binding domain, in line with Binding Principle A (Chomsky 1981),60 the implicit argument has to be the binder in (147b), as there is no other nominal category that could be co-indexed with oneself. (148b) is ungrammatical because oneself is not bound, resulting in a violation of Binding Principle A. (148)

a. It was unclear [where PROi to wash oneselfi ] b. *It was unclear [that oneself needed to leave early]

Since binding means that there is a c-commanding antecedent that the anaphor can be co-indexed with, it can be concluded that the implicit argument 60. Principle A of the Binding Theory states that anaphors (reflexives and reciprocals) have to be A-bound in their binding domain (the smallest NP or clause that contains the anaphor, its governor, and a subject). An element is A-bound if it is co-indexed with another element in an argument position and c-commanded by this element (Chomsky 1981: 184–185). Precedence alone does not suffice to license an anaphor, it has to be c-commanded by the antecedent: *Tim’s sister invited himself to the party. The antecedent need not be an overt category.

Evidence from syntax

123

in (147b) is in a structurally high position (from which it can c-command the anaphor). Proponents of the syntactic approach to implicit arguments have argued that the implicit argument is located in the INFL position and that it may bind the anaphor from there (Baker 1988a). Note that in a passive with a by-phrase, the anaphor has to match the features of the NP inside the by-phrase (see (147c)). This is a problem for the clitic doubling analysis – it seems that when there is a by-phrase present, there is no indefinite implicit argument anymore (at least not in a position from which it could bind an anaphor). All in all, even Roberts (1987: 161), who argues for the structural presence of an implicit argument in passives, concedes that this kind of binding relationship is “at best awkward . . . and in many cases quite impossible.” He notices that plural anaphors are more readily bound than singular ones (with the exception of oneself) and each other more readily than themselves. As a borderline example he gives (149) – the intended reading is that the individuals who fold the blankets put them on each other. Even though there should be an implicit argument that could bind the reciprocal, the sentence is at best marginally acceptable in the intended reading. (149)

??The blankets were put on top of each other. (Roberts, 1987)

What further complicates the situation is that there are instances of oneself that do not seem to follow Principle A of the Binding Theory in the first place. In (150a) one does not necessarily get the interpretation that the implicit argument of consider is co-referent with oneself – the sentence can also mean that there are some people who considered it a sin for anybody to feel pride in himself or herself. In this reading, the implicit argument of consider does not bind oneself. This is even clearer in (150b) – this is a sentence without a potential structural binder for the anaphor. Neither be nor pride has an argument that could be co-indexed with oneself – the anaphor is used as a logophor.61 According to Ackema and Schoorlemmer (1995: 177), Binding Principle A only holds for anaphors in argument positions. Anaphors “that do not themselves constitute arguments, but that for instance are embedded in an argument,” as in (150) are not arguments of the verb and can be used logophorically.

61. Logophors typically occur in West African languages and do not have structural antecedents, but refer to the person the statement is about or to the person whose perspective is given (Bußmann 2002)

124

(150)

Chapter 4 The implicit agent in English passives

a. b. c.

Pride in oneself was considered a deadly sin. (Quirk et al. 1985) Pride in oneself is a deadly sin. Letters to oneself compose quickly. (Stroik 1992)

While Stroik (1992) takes sentences like (150c) as evidence that there must be an implicit agent in the middle construction, Ackema and Schoorlemmer argue against it: If oneself does not fall under Binding Principle A (since it is not in an argument position), the grammaticality of the sentence does not constitute evidence that there is a c-commanding antecedent. This brings up the question of whether the situation in the passive is any different. In example (150a) the anaphor is inside another phrase (the PP) – does this mean it is not an argument of the verb and can therefore be used logophorically? At any rate, it seems that the binding potential of the implicit argument is severely limited and that it is not very strong evidence for a specific syntactic position of the implicit argument.

3. Linking theory considerations An altogether different line of argumentation for the presence of an implicit argument comes from linking theory. Generally speaking, linking theory is concerned with the question of “how to map flexible cognition onto rigid syntax” (Pinker 1989: 102). Within the generative framework, there are two main branches of linking theory: Either information about the number of event participants, aspectual, and thematic roles is projected from the lexicon (Grimshaw 1990; Jackendoff 1990; Pustejovsky 1991; Tenny 1994; Levin and Rappaport Hovav 1995), or this information is composed in syntax and encoded in functional projections (Borer 1994; Rosen 2003). In this study, I will essentially follow the first approach, assuming, in the tradition of Levin and Rappaport, that there are two levels of representation, lexical-semantic structure (LCS) and predicate-argument structure (PAS). LCS specifies the aspectual structure of the event denoted by the verb, while PAS only lists the number of arguments that have to be mapped to syntax, singling out an external argument, if applicable (Booij 1992; Levin and Rappaport Hovav 1998; Rappaport Hovav and Levin 2001). Lexical rules may manipulate the information on either level of representation, for example by deleting or adding an argument. Among the issues that have to be addressed by lexical linking theories is the question of how many arguments a verb will have and what their semantic interpretations will be. It has been argued that aspectually com-

Linking theory considerations

125

plex events, like accomplishments (caused changes of state, expressed by verbs like kill), have more arguments than verbs used to describe simpler events, like activities (expressed by verbs like smile). The underlying idea is that complex events can be broken down into sub-events and that each sub-event needs to be identified by (at least) one argument (Rappaport Hovav and Levin 1998; Wanner 1999; Rappaport Hovav and Levin 2001) or another syntactic entity (Grimshaw and Vikner 1991; 1993). In the case of accomplishments, there are two sub-events: a causing sub-event, in which the causer does something (event type: activity), and a caused sub-event, in which another entity undergoes a change of state (event type: achievement).62 The first sub-event is identified by the external argument of the verb (in (151b), the cat), which is interpreted as the causer of the change of state; the second sub-event is usually identified by the element that undergoes the change of state (in (151b), the bird). (151)

a. b.

[x do-sth] cause [y become (at) z] The catx killed the birdy

The question now is how the causing sub-event is identified in a passive construction. Is there any evidence that the implicit argument can identify a sub-event? It has already been discussed that the passive does not change the aspectual type of the sentence. Adverbs like almost have been used to illustrate the aspectual complexity of sentences like (152a). Almost can have scope over the whole event or over one of the embedded sub-events, resulting in different readings for the sentence (Dowty 1979). In the wide-scope reading, (152a) means that the causing sub-event has not been initiated, i. e. nothing happened to the wall at all; in the narrow-scope reading the end state, in which the wall is completely covered with paint, has not been reached, but some painting action has taken place.63 Interestingly, the verbal passive retains this ambiguity, with or without a by-phrase. (152c) can mean that no action was taken at all or that the wall was only covered partly in paint. 62. It has also been suggested that the second sub-event is a state (Grimshaw and Vikner 1991, Pustejovsky 1991), but I will follow the decomposition outlined above. For the purposes of this argument, nothing hinges on the aspectual class of the second sub-event. 63. According to Tenny (1994), the internal argument “measures out” the event. This is to say that the event is finished when the wall is completely covered with paint. If almost has scope over the second sub-event, this end state is not reached.

126

(152)

Chapter 4 The implicit agent in English passives

a. b. c.

She almost painted the wall. Accomplishment[Activity [x do-sth] cause Achievement[y become painted]] The wall was almost painted (by her).

Apparently, in the passive the implicit argument can identify the causing sub-event. However, the question arises as to where this argument would be located. There is no argument position available, as the subject position is taken by the moved object of the verb. Grimshaw and Vikner (1991: 10) therefore propose that a sub-event can be identified not only by an argument (as suggested by Rappaport Hovav and Levin 1998), but by “any expression that specifies manner, time, duration, place, or purpose of process, or reason for the process.” Their aim is to account for passives with obligatory by-phrases. According to their argument, a by-phrase can be obligatory in the passive if it is the only way to identify a sub-event. This is the case with “constructive accomplishments,” events that result in the creation of an object. In (153a), for example, at the time the house-building process starts, there is no house yet. The NP this house can therefore not possibly identify the causing sub-event. That is why another phrase is needed, either a by-phrase or some other kind of adjunct: (153)

a. b.

This house was built *(by Frank Lloyd Wright). This house was built in 1990/as a present for . . . .

If the implicit argument could identify the causing sub-event, there would be no need for a by-phrase or any other identifying element in these examples. But if sub-events can be identified by almost anything – adjuncts, stress, negation, adverbs – what is the point of the sub-event identification principle, as put forward by Grimshaw and Vikner (1991; 1993) or Rappaport and Levin (2001)? This question is justly raised by Goldberg and Ackerman (2001), who argue for a pragmatic approach to obligatory adjuncts instead: An utterance must be informative – asserting something that is presupposed anyway is infelicitous. Adjuncts are just one way to provide a successful focus (with information that is not presupposed) in a clause, focal stress provides another option. (154)

a. b. c.

This house WAS built. The HOUSE was built (not the garage). The house was BUILT (not just designed).

Evidence from psycholinguistics

127

They also note that less complex events may require the presence of an adjunct, too. (155b) is much better than (155a) – and this is not a complex event with two different sub-events. Believing is simply an activity, no causer or change of state is involved. (155)

a. *The claim was believed. b. The claim was believed by many/in the seventh century/in the South.

According to Goldberg and Ackermann (2001), the fact that some passives require by-phrases in order to be grammatical has got nothing to do with the aspectual structure of the event: “Crucially, this infelicity is quite independent of assumptions about event structure and argument identifications for constructive predicates” (Goldberg and Ackerman 2001: 805). Their data are quite convincing and one has to conclude, again, that the implicit argument may be part of the aspectual interpretation of the sentence, but also that one cannot derive a specific syntactic status from this possibility.

4. Evidence from psycholinguistics Psycholinguistic experiments seem to confirm the structural reality of the implicit argument in passive constructions. Although this has also been claimed for the grammar of young children (Verrips 2000), I will focus on data that involve adult speakers here. In a series of experiments on the interpretation of purpose clauses as in (156), Gail Mauner and her research partners found that purpose clauses after passives were judged as just as grammatical and were processed just as easily as purpose clauses after agentive actives. By contrast, purpose clauses after ergatives (as in (156b)) were rejected or caused significant response delays, indicating that an agentive reading was not readily available. Of the four sentences in (156) only (156b) created significant processing difficulties. Another result from the study was that that short passives and long passives, i. e. passsives with by-phrases, were equally felicitous with purpose clauses. This, again, indicates that passives have implicit arguments, while ergatives do not (Mauner, Tanenhaus and Carlson 1995). (156)

a.

The ship was sunk to collect a settlement from the insurance company.

128

Chapter 4 The implicit agent in English passives

b. c. d.

The ship sank to collect a settlement from the insurance company. The owner sank the ship to collect a settlement from the insurance company. The ship was sunk by the owner to collect a settlement from the insurance company.

Comparable results were found in a cross-modal study. Melinger and Mauner (1999) presented participants with auditory stimuli that included purpose clauses and asked participants to read out visual (i. e. written) continuations of the sentences. For example, the auditory stimulus was something like To win lots of cash and a chance at the grand prize, the game show’s wheel . . . and its visual continuation could be either a passive (. . . was spun) or an ergative construction (ungrammatical) ergatives (. . . spun). As in the earlier experiment, reading times were considerably longer for the combination of purpose clause and ergative construction, which the authors interpreted as confirmtion for the hypothesis that the passive has an implicit argument and the ergative does not. However, these results do not rule out the possibility that the implicit argument in passive constructions is simply inferred from context or “derived from conceptual sources” (Mauner and Koenig 1999: 179). In a follow-up experiment, Mauner and Koenig (1999) therefore aimed at finding evidence for the grammatical reality of the implicit argument. In a self-paced grammaticality judgment task, modeled after Mauner, Tanenhaus, and Carlson (1995), they presented sentences like (157) on a word-by-word basis on a computer screen. (157)

a. b.

The antique vase was sold immediately to raise some money for the charity. The antique vase had sold immediately to raise some money for the charity.

After each word, participants had to press a button (“yes” or “no”), indicating whether or not they found the sentence acceptable so far (“make-sense judgment task”). The result was that sentences like (157b) elicited a higher number of “no” judgments along with significantly longer response times. Overall, participants’ response patterns were very similar to those in the prior study. The crucial difference between the verbs used is that the matrix verbs in the new experiment all expressed events that conceptually always require an agent, while the matrix verbs in first study expressed events that can happen spontaneously, without any interference from an agent. Conceptu-

Evidence from psycholinguistics

129

ally, every act of selling involves a seller, but not every act of sinking involves a sinker. If the implicit argument in the passive were of the conceptual kind, one would expect no significant differences in participants’ responses to the sentences in (157a) and (b). However, in this follow-up experiment, the short passive again patterned with constructions in which the agent was explicitly expressed, while the ergative did not. Mauner and Koenig concluded that the implicit argument in the passive must be present in a way that is different from a merely conceptually construable agent. In another experiment, an eye-movement study, they tried to find out at which point in the sentence the implicit agent interpretation becomes accessible. They used two different syntactic contexts: One involved a preposed purpose clause, followed by a main clause without any overt agent argument, (see (158a) and (b)), and in the second version the clause without the overt agent was a complement of a matrix verb like tell (see (159a) and (b)). (158)

a. b.

(159)

a. b.

To raise money for the charity, the antique vase was sold immediately to a collector. To raise money for the charity, the antique vase had sold immediately to a collector. The detective told the museum director that the antique vase was sold immediately to a collector. The detective told the museum director that the antique vase had sold immediately to a collector.

Participants’ eye movements were recorded for three regions of the second clause (the one with the verb sold): the subject, the verbal region (auxiliary plus lexical verb or participle), and the post-verbal region, containing the AdvP (immediately) and the PP (to a collector). The results: Sentences like (158b) elicited longer first-pass reading times (the time first spent reading a region) for the post-verbal region than sentences like (158a), and they elicited longer overall reading times, with more readers re-reading the ergative sentences than re-reading the passives (Mauner and Koenig 1999: 182–183). The temporal dynamics of executing eye movement are such that a reaction to the information taken up in the VP region can only be measured in the post-VP region. The “anomaly effects” that occurred with ergatives (but not with passives) indicate that the argument structure of the verb is accessed as the verb is read and that there is a mismatch between the expectation of an agent argument (as required by the fronted purpose clause) and the actual argument structure provided by the lexical verb sold in the case of ergatives.

130

Chapter 4 The implicit agent in English passives

No such differences occur in sentences like (159) – here, it did not matter whether or not the second clause had an agent since there was no purpose clause whose interpretation depended on it. As the sentence unfolds, the reader integrates the verb and its argument structure into the sentence interpretation. When the argument structure does not match, as in the case of ergative verbs following purpose clauses, longer reading times follow the recognition of the verb. Participants in the study also selectively revisited ergative verbs after purpose clauses, but they never went back to the passive sentences once they had finished reading them (Mauner and Koenig 1999: 182). All in all, Mauner’s experiments allow the conclusions that (i) the implicit agent is an integral part of the interpretation of the passive, (ii) it is also an integral part of the structure of the passive, and (iii) structurally, it is most likely linked to the passive participle, since its interpretation becomes available as soon as the passive participle is parsed. However, it does not become quite clear which role is played by the auxiliary verb, since it is grouped in one region with the passive participle. Since two verbal elements are grouped together, we cannot be sure where exactly the implicit agent is encoded; it could be encoded as a syntactically represented element inside the VP or IP or as part of the lexical representation of the verb.

5. Evidence from language use So far, I have looked at evidence from syntax, semantics, and psycholinguistics that confirm that there is an implicit agent in the passive construction (although it is not quite clear yet where it is located). Let us now turn to data from language use which indicate that speakers are very much aware of the implicit agent in the passive. Since passivization serves many different functions, it is somewhat challenging to find data which show that a passive is used to realize the agent specifically as an implicit argument. It is quite safe to say, though, that passivization is not usually about moving the agent to a focus position at the end of the clause. If that were the case, most passives would have by-phrases, but it is well known that passives with by-phrases are much rarer than short passives across all registers and genres (Svartvik 1966; Thompson 1987; Biber et al. 1999: 938).64 While functional approaches may recognize that the passive “evokes (implicitly or explicitly) the idea of an external responsible entity” (Cornelis 1996: 247), the focus 64. One of the reasons is that by-phrases tend to convey new information and are normally not used over longer stretches of text (which deal with the same agent).

Evidence from language use

131

is on the fact that the agent need not be mentioned or expressed overtly.65 The claim made here is more specific: The agent is an integral part of the representation of the passive – that is the reason why verbs which express actions that can only be carried purposefully (like murder) allow for the passive. In this section I will briefly discuss some examples that illustrate how speakers make use of an implicit agent. In the second part of this section, I will turn to passives that are the result of a speaker adding an implicit agent to an otherwise agentless verb (which usually results in ungrammaticality, but it may be a case of calculated ungrammaticality). 5.1. Sentences that require an implicit agent It has been shown that the syntactic potential of the implicit agent is limited, but that the implicit agent is nevertheless essential to the meaning of the passive. The following sentences are from a corpus study on scientific writing, which is discussed in more detail in chapter 5. They illustrate that the passive can be combined with a purpose clause and that the non-overt subject of the purpose clause (PRO) is construed as co-referent with the agent of the passive: In (160a), for example, the agent of study is co-referent with the agent of determine. These sentences are taken from abstracts that precede published articles in scholarly journals – the agents in the passive can remain implicit because their identity is known through contextual information and knowledge of genre-specific conventions: If an experiment is reported in detail in the abstract of an article, it is usually an experiment that was carried out by the authors of that article. Even though style manuals encourage the use of the first person, authors of scholarly articles, particularly in the natural sciences, continue to use constructions that enable them to avoid referring explicitly to themselves as the agents of an event. The passive is one of these constructions; others include non-animate subjects 65. Inconsistent use of the term “agentless” can obscure the exact position some-one takes towards the implicit agent. For instance, Thompson (1987: 498) calls adjectival passives “agentless” and does not consider them real passives, precisely because they cannot occur with by-phrases, but she also refers to verbal passives that do not occur with by-phrases (but could occur with them) as “agentless passives.” In the first case, “agentless” means that an agent is not part of the interpretation of the sentence; in the second case, “agentless” means that the agent is simply “not expressed,” but is “either inferable from the context or has a referent whose exact identity is not important” (Thompson 1978: 499).

132

Chapter 4 The implicit agent in English passives

as in This paper argues that . . . . If the identity of the agent is not clear, a by-phrase can be used to clarify its reference in the passive, as in (160d). (160)

a.

b.

c.

d.

e.

f.

g.

Private speech (PS) – or speech for the self – was studied in preschool-age children to determine how widely and with what characteristics it occurs when examined in the familiar home setting. (APL 8)66 Two aspects of grammar – binding and control – were assessed to determine whether poor readers had syntactic deficits. (APL 18) Society established in Madagascar in 1872 is used to examine how these categories were sought, made, and remade through discipline and social control. (JWH 16) The type of synapse used by each interneuron to influence its neighbors follows three functional organizing principles. (Science 7) Here, the theory of electromagnetic wave propagation in diffusive media is combined with information theory to show how interference affects the information transmission rate between antenna arrays. (Science 11) Mass spectrometry-generated peptide sequence information was used to clone a Vicia faba complementary DNA. (Science 15) Recombinant DNA technology was used to improve its nutritional value in this respect. (Science 16)

Data like these show that the reality of the implicit agent is employed to construct sentences that syntactically require an agent, but which occur in a context in which explicit reference to an agent is to be avoided. In the passive, the agent can be non-overt because it can be realized implicitly, and in the purpose clause the agent can be non-overt because it occurs as the subject in a non-finite clause. Chapter 5 will explore motivations for using the passive in academic writing in more detail. At this point, we can simply acknowledge that the short passive is used in constructions that depend on the existence of an agent argument.

66. Acronyms refer to the names of scientific journals and are explained in more detail in chapter 5.

Evidence from language use

133

5.2. Unaccusative passives In his classic paper on unaccusativity, Perlmutter (1978) made the generalization that unaccusative verbs (one-argument verbs with an underlying object) should not form passives. By and large, his generalization is still considered valid, although the explanations for it have changed.67 In generative grammar, it is generally assumed that unaccusative verbs cannot form a passive because they do not have an external theta-role that could be absorbed by the passive morpheme (Jaeggli 1986). Unaccusative passives, therefore, should not exist, and passivization is considered to be a standard test to differentiate between unergative and unaccusative verbs (Grimshaw 1986). If there is a passivized verb that, based on its semantics, is expected to behave like an unaccusative verb, this is considered to be a real problem (Abraham 2003; Aue-Apaikul 2006). In this section, I will not discuss these problem cases and their implications for a theory of linking. Rather, I will present data that show that speakers deliberately overextend the rule of passivization to verbs that do not have an external argument, thereby adding an implicit argument to an otherwise agentless verb. I will also discuss data from second language acquisition that show ungrammatical passives of a different kind: Learners may produce sentences that superficially look like passives but that have no implicit argument and are not intended to be passives. The point is to demonstrate that ungrammatical passives involving intransitive verbs come in at least two varieties: Those that indicate that a speaker is aware of the implicit agent in the passive and uses it effectively and pointedly, and those that indicate that the speaker is not aware of the implicit argument and is simply overgeneralizing a rule. 5.2.1. Unaccusative passives used by native speakers Noam Chomsky built the theory of generative grammar on the “fundamental distinction between competence (the speaker-hearer’s knowledge of his language) and performance (the actual use of language in concrete situations)” (Chomsky 1965: 4). At the core of his theory of grammar is the assumption that speakers have implicit knowledge of the grammar of their native language and that they will judge sentences that are not part of that grammar as not well-formed. Sometimes, however, speakers show knowl67. Perlmutter (1978) argued that there could be only one advancement to the 1Position (roughly, the subject position) in the sentence. Both unaccusative verbs and passivized verbs require movement to this position, i. e. there would be two advancements to the same position.

134

Chapter 4 The implicit agent in English passives

edge of principles by violating – rather than following – them for a certain effect, e. g., to manipulate the semantics of a verb through passivization in order to enforce an implicit agent reading. The result will be sentences that are ungrammatical (in the sense of: violating the rules of grammar) but acceptable, “perfectly natural and immediately comprehensible without paper-and-pencil analysis, and in no way bizarre or outlandish” (Chomsky 1965: 10). There are a number of verbs that cannot be passivized in English, among them all intransitive and copular verbs and some transitive verbs of measuring and having: (161)

a. b. c. d. e.

*Six pounds were weighed by the baby. *You are not fitted by the coat (from Quirk et al. 1985) *It was appeared on the horizon by a ship. *It was run all day. *The measles are had by almost every child.

Note that the passivizability of these verbs depends on how they are used. If, for instance, an intransitive verb is used transitively (e. g., in a resultative construction), it can be passivized, as in (162a). Similarly, if verbs like weigh and fit are used agentively, i. e. in a way in which the arguments are not perfectly symmetric, they allow for the passive, as in (162b). (162)

a. b.

These sneakers have been run threadbare by Madonna. Babies should be weighed at least once a week.

I will focus on intransitive verbs here. Ever since Perlmutter’s (1978) classic paper on unaccusativity it has been generally accepted in generative grammar that there are two sub-classes of intransitive verbs: unergatives and unaccusatives, also known as ergatives (Haegeman 1985; Burzio 1986). Following Perlmutter and other proponents of the syntactic approach to unaccusativity, I will assume that unergatives have an external argument (roughly, a base-generated subject), while unaccusatives/ergatives have an internal argument (a base-generated object), which has to move to the subject position because the verb cannot assign accusative Case (“Burzio’s Generalization”).68 Smile is an unergative verb, which means that its single argument, Harry, is an external argument. As such it starts out in the specifier of the 68. For a review of syntactic tests that help distinguishing unergatives from unaccusatives in English, see Wanner (1999: 44–68). Unergatives, for instance, allow cognate objects (Harry smiled an awkward smile), while unaccusatives do not (*She appeared a strange appearance).

Evidence from language use

135

VP (the VP-internal subject position). It then moves to the specifier of the IP to satisfy the Extended Projection Principle, which demands that subject positions be filled in English. This is also the position for the assignment of nominative Case (or, in the framework of the Minimalist Program, for checking off nominative Case features, see Reuland 2000). Wilt, on the other hand, is an unaccusative verb, which means it only has an internal argument. This argument, the NP the flower, starts out in the position of a direct object, to the right of the verb. Movement to the subject position happens for the same reasons, with an extra landing in the unfilled specifier of VP. (163)

(a) unergative pattern smiled]]

IP [HarryiVP [ti

(b) unaccusative pattern floweri [VP [ti wilted ti ]]

IP [The

Perlmutter showed that in languages in which single-argument verbs can be passivized, e. g., Dutch and German, it is only a sub-class of these verbs, the unergative verbs (verbs like smile), that allow for the passive. The following examples are from Perlmutter (1978). (164)

a.

Er wordt geblaft/gehinnikt/gekrast/gemiauwd. It becomes barked/crowed/meowed. ‘There was barking/crowing/meowing.’

136

Chapter 4 The implicit agent in English passives

b. *Er werd door het water binnenn een kwartier verdampt. It becomes by the water within a quarter hour evaporated ‘The water evaporates within a quarter of an hour.’ (165)

Es wurde die ganze Nacht lang gelacht und getanzt. It became the whole night laughed and danced ‘There was laughing and dancing all night long.’ b. *Es wurde vom Wasser schnell verdampft. It became by the water quickly evaporated ‘The water evaporated quickly.’

a.

In terms of the analysis presented here, the reason for the non-passivizability of unaccusatives is that they do not have an external argument – if they do not assign an external theta-role, then there is no role that could be absorbed by the passive morpheme. Or, to put it in more functional terms, if there is no external argument, it cannot be demoted to an implicit argument. In contrast, unergatives have an external argument and can therefore be passivized in general (see chapter 3 for a discussion of why English, unlike German or Dutch, does not allow truly impersonal passives). Interestingly, sometimes unaccusative passives are used quite deliberately, as in the following examples: (166)

a. b.

c. d.

It didn’t happen to me, it was happened upon me! (from Helen Fielding, Bridget Jones’ Diary) I wasn’t snogging someone. I was being happened upon against my will! (from Helen Fielding, Bridget Jones – The Edge of Reason) And then Minimalism arrived – or, if you want to put it this way, it was arrived. (a linguist, in a conversation) Thirty thousand people – a whole generation – were “disappeared” in seven years of military rule. (from an online article on a human rights website).

Note that there is no by-phrase and that in (166a) through (c) the unaccusative passive is contrasted with a grammatical active. The point of using a passive here is clearly to add an implicit agent to the clause. If something did not just “happen to me” but “was happened upon me,” there must be somebody who made it happen to me. The situation is the same in (166c): What the sentence means is that “Minimalism” (the current mainstream framework of generative grammar) did not simply “arrive” on the scene, but that somebody enforced it as a new standard. Pinker (1989: 159) jok-

Evidence from language use

137

ingly refers to sentences in which speakers overgeneralize an argument alternation rule (The experience grew me up in a hurry) as “Haigspeak.” He considers them as evidence that speakers can (and will) expand very specific alternation rules (“narrow-range rules”) to “broad-range rules.” The passive, however, is already a broad-range rule (the verbs it operates on can be defined broadly, without recurrence to individual semantic verb classes) and this process of expansion is therefore not to be expected. Passivizing verbs that do not fit the rule will result in straightforward ungrammaticality and is therefore expected to occur much more rarely than, say, violations of the narrow dative rule (as in I returned her the books). The case of disappear has a history of its own – in the way it is used in (166c), it is a direct translation from Spanish los desaparecidos (“the disappeared”), referring to the political situation in South America in the 1980s, when opponents of the military government were abducted and tortured by the police. Though this use of the disappeared has become rather common,69 users know that it is not a standard form, as indicated by the use of quotation marks in (166c).70 Note that the use of disappear in a passive may transcend its original context, as illustrated in the following example, in which, again, the verb is put in quotation marks: (167)

And now we have “The Choice,” a book about this year’s riveting cliffhanger of an election. (One might note in passing that Woodward himself makes this count four, not five. The Quayle book, of which more later, has been “disappeared” from the flyleaf directory of Woodward’s key oeuvre as proudly presented in the latest volume.) (Christopher Hitchins, salon.com)

The same phenomenon can be observed in other languages. While some speakers reject unaccusative passives hands-down, other speakers, notably people who delight in wordplay, deliberately use passives based on unaccusatives to draw attention to the fact that an event that is presented as agentless actually has an agent. The following example from the memoirs of Viktor Klemperer (1881–1960), a German philologist of Jewish descent, is a bitter wordplay on the semantics of sterben (“to die”), which, just like its 69. The Shorter Oxford English Dictionary (2002) notes as “colloquial” the transitive use of disappear: “Abduct and secretly kill or detain (someone), esp. for political purposes”. 70. It should be noted that formal transitive uses of disappear, though rare, have also been attested. The Oxford English Dictionary (1989) lists the following example from a chemistry journal in 1897: We progressively disappear the faces of the dodecahedron.

138

Chapter 4 The implicit agent in English passives

English equivalent, cannot easily be used transitively and cannot normally be passivized. (168)

Eine Frau . . . kommt . . . ins KZ, wo sie gestorben wird. (from Viktor Klemperer, Ich will Zeugnis ablegen bis zum letzten) A woman . . . comes . . . into the concentration camp, where she becomes died. ‘A woman is deported to a concentration camp, where she is killed.’

By passivizing the verb die rather than using the causative verb töten (“kill”) or ermorden (“murder”), the author draws attention to the discrepancy between what is claimed (that people “just die” of natural causes in concentration camps) and what actually happens (that they are brutally murdered). Data like these, as marginal as they may be in published writing, clearly show that speakers are aware of the implicit agent in the passive. The implicit argument reading is part of their grammar – some speakers deliberately force it on verbs that do not have an agent to start with. 5.2.2. Unaccusative passives in second language acquisition With regard to their form, the passives discussed above are reminiscent of ungrammatical passives produced by second-language learners, see the sentences below from Zobl’s (1989) analysis of interlanguage data: (169)

a. b. c.

The most memorable experience of my life was happened 15 years ago. [L1 = Arabic] My mother was died when I was just a baby [L1 = Thai] Most people are fallen in love and marry with somebody. [L1 = Japanese]

Here we also deal with unaccusative verbs, i. e. verbs that do not have an external argument and can therefore not be passivized. The examples are from essay assignments and there is no indication that an implicit agent reading is intended (for instance, (169b) is not supposed to mean that the mother was killed by someone). We can also see that these examples are not contrasted with grammatical passives of the same verb, nor are there quotation marks to indicate the writer’s awareness of the non-standardness of sentences like these. Zobl (1989) makes the plausible argument that these sentences are not intended as passives, but instead incorrect productions of perfect tense – if the auxiliary be were replaced by a form of have, the sentences would be just

Evidence from language use

139

fine (but they would not be passive constructions, of course). This leaves us with the question of why learners do not use the correct auxiliary. The selection of be as a perfective auxiliary for unaccusative verbs in English has also been found in in more controlled linguistic environments. In a questionnaire study (Balcom 1997) second language learners were asked to give grammaticality judgments and to fill in blanks in sentences like “The great fire (take place) in St. John’s on July 8, 1892.” Balcom reports that in both tasks the acceptance or use of inappropriate combinations of the auxiliary be and the past participle was significantly higher with unaccusative verbs (including stative verbs and middles) than with other verbs. In Zobl’s data, a transfer from L1 can be ruled out as an explanation, since the first languages of the learners did not provide different auxiliaries for unergative and unaccusative verbs (Oshita 2000: 297). According to Oshita, these sentences are overgeneralizations from true passives: Learners tend to mark NP-movement syntactically by the choice of the auxiliary that occurs in the passive.71 The problem with this analysis is that in the VP-internal subject format every sentence involves Case-motivated NP-movement: The subject is generated inside the VP and moves to the subject position in order to get (or check) Case. Why would a learner who has mastered nominative subjects in finite clauses associate NP-movement with the passive and the auxiliary be? Of course, there are many different types of errors relating to the mastery of the passive construction in English, especially for learners whose first language is topic-oriented and does not use syntactic processes that target the subject position to introduce non-agent topics. The converse type of the ungrammatical passives in (169) is illustrated by the following examples (from Han 2000): Here the sentences seem to convey passive meaning, but there is neither a passive auxiliary nor a passive participle. (170)

a.

b.

Thank you for your paper which you sent me on 22/02/96. Since you did not address the postcode (SW7 2BY), and so it arrived just this morning. The letter about graphics file has not received. Hello, Dear X, Your letter of 17/04/96 has just received. Thak [sic] you very much. I guess that the letter may be enough for the visa use. This letter will be delivered to Australia Embassy tomorrow.

71. It would be interesting to see if the same mistake occurs in other unaccusative structures, such as raising constructions.

140

Chapter 4 The implicit agent in English passives

c.

They told me that the attractive offer will be sent to me a bit later since what I sent to them have not received.

The data are from e-mails written by very advanced Chinese learners of English residing in the United Kingdom. Han (2000) analyzes these errors as a transfer effect resulting from discourse-based word order in Chinese. In (170a) the NP the letter about graphics file is introduced as a new topic and since topics often coincide with the grammatical position of the subject in English, the non-agent NP is realized in the position of the subject, although this is not in line with the argument structure of the non-passivized verb. A different kind of analysis is put forward by Huang (1984), who argues, on the basis of similar examples, that the non-agent topic NP is not in the position of the subject at all. Rather, it is in the topic position (a position to the left of the subject), and the subject position is filled with an empty argument (an arbitrary pro), as indicated in (171b). It is this non-overt subject that is assigned the verb’s agent theta-role. (171)

a. b.

Rational emotions must use for judging. [Rational emotions]i proarb must use ti for judging

Looking at the examples in (170a) and (b), one can see that the auxiliary is inflected for third-person singular, which is compatible with both an empty subject pronoun and with the topic NP the letter, and that in (c), the auxiliary have agrees with neither the clausal topic what I sent to them nor with a potential non-topical pro subject. Weighing between the two analyses is an issue that cannot be explored in depth here. What matters in the context of the questions discussed in this chapter is that the implicit argument reading is key in deciding whether or not a given sentence should be analyzed as an attempted passive. The sentences discussed are not meant to be passives and do not show any implicit argument effects. This is in sharp contrast to the data in (166) through (168).

6. Representation of the implicit argument I discussed data that indicatet that the implicit argument in passive constructions is not merely a conceptual construct or inferred from context. Just what exactly it is, however, is still unclear. We saw that the implicit argument does not behave like a full-fledged argument: It can control PRO, but only in very restricted circumstances; it can bind oneself, but appar-

Representation of the implicit argument

141

ently it doesn’t have to; it can be instantiated as a by-phrase, but with the syntactic status of an adjunct; and its potential to identify a sub-event may not be such an important clue, if the theory of sub-event identification (one argument per sub-event) can be refuted. What we do know is that the implicit argument cannot be one of the empty categories attested in English. It cannot be PRO, as PRO may only occur in non-finite clauses, and passive constructions can be finite. It cannot be an NP- or wh-trace either, because there is no position available that would be appropriate for a trace. Furthermore, a trace would have to be bound by an antecedent, and it is not clear where that antecedent should be located. In the following, I will discuss two approaches to the representation of the implicit argument in the passive: Most proponents of the syntactic approach (Fabb 1984; Jaeggli 1986; Roberts 1987; Baker 1988b; Wyngaerd 1988; Baker, Johnson and Roberts 1989) assume that the passive morpheme itself is the implicit argument and that it is assigned the theta-role in syntax. A variation of this absorption analysis is Borer’s (1998) suggestion that the implicit argument be treated as an empty pronoun (pro). Proponents of the lexical approach, on the other hand, argue that implicit arguments are not projected onto syntax at all. They are arguments that are “suppressed” at the level of argument structure and do not correspond to any specific syntactic element (Jackendoff 1987; Booij 1992). I will argue in favor of the latter approach. 6.1. Syntactic approaches The syntactic approach to the implicit argument in the passive assumes that the argument structure of the verb is not affected at all by passivization. Any differences in argument realization in comparison to the active will stem from syntactic operations, such as movement and percolation. 6.1.1. The absorption analysis At the core of Jaeggli’s (1986) absorption analysis of the passive, which is widely accepted in generative grammar, is the assumption that the passive morpheme, while being an inflectional morpheme, is assigned a theta-role by the lexical verb: “the passive suffix ‘absorbs’ the external j-role of a predicate simply by being assigned that j-role. Nothing more is involved” (Jaeggli 1986: 591). As a consequence, the theta-role in question cannot be assigned outside the passive participle, which means that the subject position

142

Chapter 4 The implicit agent in English passives

in the passive is not in an argument position anymore. Since the passive morpheme is also assigned Case by the verb, the passive participle as a whole cannot assign Case anymore. In other words: the passive participle behaves just like an unaccusative verb. As simple and elegant as this analysis seems, it is not immediately obvious why an affix can be assigned a theta-role. Jaeggli argues that the external argument is not associated with a subcategorized element and can therefore, in principle, be linked to anything. It is assigned to the passive morpheme, because it is the “defining morphological/syntactic property” (Jaeggli 1986: 592) of this element that it be assigned an external theta-role. Verbs that do not have an external argument cannot be passivized because the lexical requirements of the passive morpheme (to be assigned an external theta-role) would not be fulfilled. [+V, −N]

(172)

[+V, −N]

en

paint

js

jd [

NP]

Taking up Jaeggli’s absorption analysis, Roberts (1987), Baker (1988), and Baker, Johnson & Roberts (1989) discuss where the passive morpheme should be generated. Its double nature as an inflectional morpheme that has to attach to the verb, on the one hand, and as an argument, on the other hand, is reflected in their treatment of the morpheme as a clitic: “-en is syntactically a clitic but phonologically an affix” (Baker, Johnson and Roberts 1989: 223). This is debatable, as the passive morpheme does not really behave like a pronominal element – it is positioned to the right of the verb and does not show agreement. As a clitic, however, the morpheme may be associated with a full argument in a clitic chain. This chain would account for the simultaneous presence of the implicit argument and the by-phrase. Also, as a clitic, the morpheme has to govern the argument associated with it (Borer 1984). This means that the by-phrase starts out in Spec,IP (governed by the morpheme under INFL72) and has to move to a position inside the 72. After Larson’s (1988) influential analysis of double object construction in English, many syntacticians adopted the concept of a “small vP” with a functional head. The external argument is considered an argument of the functional head, not of the lexical verb. Kratzer (1996), for instance, assumes that there is a Voice

Representation of the implicit argument

143

VP, as the morpheme moves to V to attach to the verb. This kind of movement is problematic because it does not seem to leave a trace – after all, the subject position has to be available as a landing site for the Case-motivated movement of the object. Jaeggli himself takes a different route out of the by-phrase-plus-implicitargument dilemma. He argues that the by-phrase is an optional argument of the passive morpheme. In languages other than English, the morpheme might not subcategorize for such an argument – those would be languages without long passives.73 According to this analysis, the by-phrase is not an argument of the verb, but of the morpheme. Again, this sounds simple, but the consequences are not: If the by-phrase is not an argument of the verb, how can it be explained that the NP inside the by-phrase gets exactly the same interpretation that the external argument of the verb would get in an active clause? Jaeggli assumes that this is brought about by “j-role transmission” – “j-role assignment from the suffix to the by-phrase” through a percolation mechanism (Jaeggli 1986: 600). In contrast to the clitic doubling analysis, the external theta-role either stays with the passive morpheme (implicit argument reading) or it is transmitted to the by-phrase, in which case the passive morpheme does not have argument status anymore. The NP in question does not receive its theta-role from the predicate or from a projection of the predicate; rather it receives it from the preposition by. If anything, the NP is an argument of by. The whole PP might be considered an argument of -en; but once again, it is not an argument of the verbal predicate. (Jaeggli 1986: 603–604)

The transmission process itself is not very convincing, since theta-roles, once assigned, do not normally travel on to another element. One would also expect that one can observe different control properties depending on whether the passive has an implicit argument (only “thematic control” is possible, see section 2) or a by-phrase (which is not an implicit argument and should therefore be able to act as a controller like a full argument). But data like those in the sentences in (173) show that this is not the case. The by-phrase’s

Phrase, under which the participle is embedded. The internal argument of the verb is generated inside the VP, and the external argument is generated as a clitic. 73. Among the languages that have passives but that do not allow for the overt expression of an agent, Siewierska (1984: 46) lists Urdu, Classical Arabic, and Latvian. She also mentions languages in which passive clauses always include an overt agent (Kota, Indonesian), although this seems to be rarer and related to a requirement that the verb agree with the subject and the agent.

144

Chapter 4 The implicit agent in English passives

control properties do not seem to be any different from those of the implicit argument. (173)

a. *It was tried-IMPi [PROi to go to the zoo] b. *It was tried by Harryi [PROi to go to the zoo]

Clearly, the by-phrase cannot do more than the implicit argument can do. 6.1.2. The pro analysis A different proposal is put forward by Borer, who also generates the passive morpheme and the verb in separate places, but who does not suggest that the morpheme itself is the implicit argument. In her analysis, the implicit argument is a full-scale argument, located inside the VP. It is realized either as the non-overt pronoun pro (which does not need Case) or as an overt NP (DP), with Case coming from the dummy Case assigner by. Let us look at the basics of her analysis. As a proponent of the syntactic approach to aspectual classes, Borer encodes aktionsart distinctions in syntax (within an Aspect Phrase, or AspP). According to her, “‘passivization’ is the result of a reversal in the relationship between functional projections in the clause” (Borer 1998: 76). While in the active the order of projections is Aspect Phrase > Tense Phrase, in the passive they are in reverse order. This hierarchical change is brought about by lexical properties of the passive morpheme (it is projected in Asp and is co-indexed with T). If the passive morpheme is projected as the head of the AspP, the specifier of that phrase must be filled. One of the verb’s arguments has to move out of the VP in order to fill the position. The argument that moves into the AspP will then be interpreted as an affected object or “subject of result” (SOR). The AspP accounts for the eventive reading of the passive.74 In Borer’s framework, NP arguments are not ordered inside the VP since they usually move out of the VP anyway, as accusative Case is not available inside the VP (it is assigned in a higher functional projection). What looks like two internal arguments in (174b) is simply an unstructured listing of the two arguments a transitive verb, like paint, takes. One of the NPs, in this case the wall, moves out of the VP, through the specifier of the AspP (where it gets the interpretation of an affected argument) and into the specifier of the TP, where it is assigned 74. A problem of the analysis is that only quantified NPs should fill the specifier position, otherwise there cannot be a telic reading, since English does not have a telicity affix. But in English, bare plurals are fine as the subject in a passive: Dogs should not be bathed in the winter.

Representation of the implicit argument

145

nominative Case.75 The second NP remains in the VP – which raises the question of how it is Case-marked. There are two options: Either the NP is realized non-overtly as pro, thus escaping the Case Filter; see (174c),76 or the dummy preposition by is inserted as a Case-assigner, similar to of-insertion in complex NPs. (174)

a. b. c.

The wall was painted. TP [Spec VP [Spec be AspP [Spec –en VP [paint, DP1 , DP2 ]]]]  TP [The walli wask VP [ti tk AspP [ti tx VP [paint-edx ti ,pro]]]]

In other words, in a short passive, the implicit argument is caseless pro, “an indefinite, non-specific pro, in essence, a null bare-plural” (Borer 1998: 82). Small pro is an option that so-called “pro drop” languages allow: The subject in a finite clause need not be realized overtly, as illustrated in (175). In the structural representation of the clause, the subject position is filled with a non-overt pronoun, pro. (175)

a.

b.

Compro un giornale. (from Haegeman and Gu´eron 1999) buy-1sg a newspaper ‘I buy a newspaper.’ Lei compra un giornale. she buy-3sg a newspaper ‘She buys a newspaper.’

Languages that allow pro drop are either morphologically rich and the identification of the subject takes place via an agreement suffix on the verb (as in Italian), or they do not have any agreement morphology at all and pro is identified through specific principles of discourse (this holds for languages like Chinese or Korean), see Jaeggli and Safir (1989). English is not a pro drop language, as attested by the examples in (176). So how could the implicit argument in passives be pro? (176)

a. * bathed the dog. b. *In the winter, shouldn’t bathe the dog. c. * is a liar.

75. To Borer, it does not really matter which NP moves to the subject position. She considers sentences like The girls were painted by the wall to be fully grammatical. They just do not fit most people’s experience of the world. 76. This is the sentence after movement: The affix moves down to the verb, the auxiliary moves up to the head of the Tense Phrase, and one of the arguments of the verb moves up to the Specifier of the Aspect Phrase first and then on to the Specifier of the Tense Phrase, where it is assigned Nominative Case.

146

Chapter 4 The implicit agent in English passives

Borer argues that English does not allow pro subjects because in English the verb does not raise to T: A functional projection, TP, is licensed only when it has phonological content (Borer 1998: 87). Since the verb does not move up to T, some other position inside the TP has to be filled with an overt element. This other position is the subject position. An empirical problem of this analysis is that auxiliary verbs (as well as copular be) do move up to T, which means that in a passive construction there should be no reason for the object to move. If one follows Borer’s argument, one would expect that these verbs allow pro subjects, as the T position would give the TP phonological content. However, sentences like (176c) are ungrammatical in English. Another problem of the pro analysis lies in the limited syntactic potential of the implicit argument. Like its counterpart in non-finite clauses, big PRO, pro is a full-scale argument. As such, it should be able to control the PRO subject in a complement clause (“argument control,” according to Jaeggli 1986). But it has been shown that this is not the case; recall example (139), repeated as (177a), and its representation, according to Borer in (177b). If IMP were pro, it should behave like a full argument, but it does not. (177)

a. *Johnk was promised-IMPi tk [PROi to go to the zoo] b. *Johnk was promised tk proi [PROi to go to the zoo]

Borer herself claims that pro can control PRO, quoting the example in (178a), but if this were generally true, (b) and (c) should be grammatical as well. (178)

It was decided proi [PROi to bomb Lebanon] (Borer, 1998: 84) b. *It was tried proi [PROi to bomb Lebanon] c. *It was promised proi [PROi not to bomb Lebanon]

a.

Overall, this analysis is too strong for what the implicit argument in the passive can actually do. The implicit argument does not have the syntactic potential of a true argument and consequently should not be treated as such. 6.2. Lexical approaches While the syntactic approach is based on properties of the passive morpheme as a lexical item in its own right, the lexical approach looks at the passive as an operation that has effects on the lexical representation of the

Representation of the implicit argument

147

underlying verb. Depending on one’s view of the lexicon – a “prison” for “the lawless” (DiSciullo and Williams 1987: 3) or a grammar component with “a rich system of rules” (Stowell and Wehrli 1992: 2) – and on the levels of representation one assumes – one layer of conceptual structure (Jackendoff 1983, 1990) or two levels, lexical-conceptual structure and predicateargument structure (Booij 1992; Wunderlich 1997; Levin and Rappaport Hovav 1998) – lexical analyses of morphological operations can differ greatly from each other. I will follow a two-level model of lexical representation here: The meaning of the verb and its aspectual structure are encoded on the level of lexical-conceptual structure (LCS), and the number of arguments the verb takes is encoded at predicate-argument structure (PAS). Variables in LCS are mapped onto argument positions in PAS according to languagespecific linking rules or mapping principles. (179) shows the respective LCS and PAS entries for the causative verb dry (modeled after Levin and Rappaport Hovav 1998). (179)

a. b.

LCS: PAS: [[[x act] cause [y become (at) z]] , z = DRY

As in Jackendoff ’s model, theta-roles are “just convenient mnemonics for particularly prominent configurations” (Jackendoff 1987: 378). For instance, the theta-role “agent” could be translated into “first argument of act,” or a sub-type thereof (only volitional actors). As a consequence, it is not possible for lexical operations to target specific theta-roles directly (contrary Williams 1981a), as they are not part of the lexical entry. A morphological operation can affect the lexical entry either at the LCS or at the PAS level. What is the difference, and how does one know which level is appropriate? It is assumed that a process takes place at LCS, if the elements that are affected by this process can be described in terms of LCS (Booij 1992). For instance, if a process affects only actors, it should be an LCS process. If a process affects only external arguments, it should be a PAS process. Another difference between the two levels is the potential outcome of a lexical rule: At PAS, an argument can be deleted, added, or bound (assigned some sort of interpretation). It is the level for valency-changing operations, while LCS is typically the level for operations that change the meaning of a verb or relate multiple meanings of a single verb. The spray/load alternation is an example for a rule that seems to work at the LCS level: In spray the paint onto the wall the event is measured out by the amount of paint that is used up (the event is completed when all the paint is used up), while in spray the wall with paint the event is measured out by the size of the

148

Chapter 4 The implicit agent in English passives

wall (Levin and Rappaport Hovav 1998). Both uses of spray are transitive, and both require an external argument, i. e. the verb’s argument structure does not seem to be affected. In the following, I will discuss suggestions for treating passivization as a process that affects LCS or PAS. 6.2.1. Argument suppression at predicate argument structure (PAS) Since only verbs with external arguments can be passivized, it seems fairly obvious that passivization should happen at the level of PAS. That’s exactly the argument made by Booij (1992), who compares the passive to middle formation in Dutch and finds that the verbs that undergo the middle rule have to be action verbs and that the verbs that undergo passivization have to have an external argument.77 For English, at least the passive part of his findings seems to be implicitly accepted. Passivization has long been used as a test for the presence of an external argument (see Wanner 1999: 78–82). Grimshaw (1990), for example, argues that psych verbs of the type worry, if they are used with a non-human subject, do not form verbal passives, which leads her to the conclusion that they do not have an external argument in sentences like (180a). What seems like a passive in (180b) is really a copula verb with an adjectival predicate – one could replace is with seems, for instance, or one could include an adverb that typically modifies adjectives, such as very worried. Verbs like depress or terrify can have causative meaning, but they cannot always be passivized. If that means that they do not always have an external argument, then it seems that passivization cares more about the organization of PAS than LCS. (180)

a. The situation worries/annoys/preoccupies Fred. b. Fred is worried/annoyed/preoccupied by the situation. c. *Mary was being depressed by the situation. d. People are being terrified by the government.

This leaves us with the question of what exactly happens at the level of PAS. According to Booij (1992: 62) the external argument is “present, but ‘suppressed,’ i. e. it is not realized syntactically, although it can still function as a controller for PRO.”78 In the lexical approach, then, the equivalent of theta77. However, Fagan (1988), argues that the rule which creates middles must be applied at PAS, as it affects external arguments. Though Booij and Fagan differ with respect to the middle construction, the underlying assumption is the same: The level of a lexical operation is determined by the class of verbs that are affected by it. 78. For Booij, control relations are stated at the level of argument structure.

Representation of the implicit argument

149

role absorption is argument suppression. However, suppressing an argument means different things to different people. Grimshaw defines the suppression of an argument as saturating an open argument position at the level of PAS: “Suppressed positions are represented in the argument structure but are not available for purposes of theta-marking. They are, in effect, ‘implicit arguments”’ (Grimshaw 1990: 108). Grimshaw uses a diacritic mark to indicate a suppressed, or implicit, argument. In her framework, the argument that is surrounded by only one set of brackets (in this case x) is the external argument. In (181b) the external argument is marked as suppressed in the argument structure. (181)

a. b.

The enemy destroyed the city. The city was destroyed by the enemy.

(x (y)) (x-ø (y))

What this means is that an open position is closed, or bound, at the level of PAS, thereby preventing the expression of that argument in syntax, but without deleting the argument from the verb’s meaning. Basically, when an open argument position is closed the verb will have one argument less to project onto syntax. However, suppressed or implicit arguments will always license an “a-adjunct,” i. e. an adjunct phrase that picks up the interpretation of the suppressed argument – that is what makes them different from arguments that are bound at LCS, which will be more of a conceptual, rather than a syntactic, argument. If an argument is bound before it reaches PAS, it will not be visible to syntax. This is exactly how Levin and Rappaport Hovav (1995: 108–109) analyze decausativization: The causer argument is lexically bound and does not reach PAS. Consequently, no argument adjunct is licensed; see (182b). (182)

a. b.

They broke the glass. The glass broke./*The glass broke by them.

It is a general property of suppressed arguments that they license what Grimshaw calls an “a-adjunct,” an adjunct with argument-like characteristics, in this case the by-phrase: “They resemble arguments in their mode of licensing. Yet unlike arguments they are not theta-marked, and they do not satisfy a-structure positions” (Grimshaw 1990: 109). If by-phrases were true adjuncts, they should not be selected by the verb (i. e. the by-phrase should not correspond exactly to the external argument in the active); if they were licensed by some sort of abstract event position, they should also occur with ergatives, or, for that matter, with any dynamic verb in the active,

150

Chapter 4 The implicit agent in English passives

but they do not, see (183). Due to their status as “a-adjuncts,” by-phrases are always optional. (183)

*The window opened by Harry. (meaning: Harry opened the window)

Grimshaw discusses other instances of argument suppression that also license a-adjuncts. A case in point is the nominalization of verbs that results in an eventive reading, as in (184a). In the nominal system, the a-adjunct can also be a possessive, as in (184b). (184)

a. b.

the destruction of the city (by the Romans) the Romans’ destruction of the city

Other nominalizations may have different effects. For instance, the affix -er has the effect such that the external argument of the verb is bound by R, the reference argument of the derived noun. A teacher is the person who teaches – the external argument of teach (the agent) is identical with the reference argument R of teacher, the person to whom the noun teacher refers. In sum, the suppression analysis accounts for the fact that the external argument in the passive is syntactically active without being projected onto syntax in an argument position. Argument structure is the part of the lexical representation of verbs that is visible to syntax, and we saw from the discussion in section 4 in this chapter that the implicit argument reading becomes available as soon as the passive participle is processed. This is in line with the assumption that the implicit argument is located at PAS. 6.2.2. Lexical binding at lexical-conceptual structure (LCS) So far the focus of the discussion has been on implicit arguments as arguments that are created through a morphological process, but there are many instances where speakers interpret some sort of implicit or implied argument without any morphological process having taken place. Eat, for instance, is a transitive verb and its internal argument can be realized as almost any kind of NP (while it may be unusual to eat a handful of sand, one can certainly do so). An object deletion construction, however, as in Harry is eating, is interpreted as “Harry is eating something” with “something” being a food-related category, not just anything. Should we say, then, that food is an implicit argument, rather than an optional complement, of eat? This question is addressed in the work of Jackendoff, who defines an implicit argument as “a conceptual argument that is neither expressed syntactically

Representation of the implicit argument

151

nor bound to an argument that is expressed syntactically” (1987: 409). It is bound “purely in conceptual structure, where all arguments must be present in order to express meaning” (1987: 408). The question of whether there is an implicit argument after intransitive eat is related to the status of incorporated arguments like butter in the meaning of the verb butter, as in (185c). If you butter something, it always means that you put butter on the object, although no constituent usually corresponds to the theme in syntax. Contrary to eat, with verbs like butter the theme does not just underlie selectional restrictions, it is completely specified. If one spreads cream cheese on a bagel, one cannot refer to that event as “buttering the bagel.” (185)

a. b. c. d.

Harry ate a banana/a shoe/a handful of sand. Harry is eating. (meaning: Harry is eating some sort of food.) Harry buttered the bread. (meaning: Harry puts butter on the bread.) Harry buttered the bread generously with his favorite brand of Irish butter.

Should we say that the noun butter is an implicit argument of the verb butter? Note that a fully specified theme can be spelled out in syntax, but it has to be informative (more specific than the inherently specified theme), and it can occur in a prepositional phrase, see (185d), which reminds us of the status of the by-phrase in the passive. Jackendoff addresses these questions at the level of conceptual structure. In his framework, there is no separate argument structure – arguments will be those conceptual entities that are indexed for “argument fusion” with syntactic constituents. An incorporated argument will not have an argument index. This means that the argument is an integral part of the semantics of the verb, but it is not visible to syntax (Jackendoff 1987: 405). (186)

a. b.

Harry buttered the bread. [Eventcause ([Thing]i , [Eventgo ([Thingbutter], [Path to ([Place on ([Thing]j )])])])]

The first argument of go (the entity that is spread onto something) is fully specified (“butter”) and does not have an index, meaning that it will not be an argument in syntax. This is a fully incorporated argument, or, in terms of Levin and Rappaport Hovav (1998), a “constant.” In the case of eat, the first argument of go (the entity that is consumed) will also be specified (something like “[Thing food]j”), but it is still indexed. This means that eat is

152

Chapter 4 The implicit agent in English passives

generally a transitive verb and there has to be some sort of rule that derives its intransitive use (Booij 1992: 149). Building on Jackendoff ’s work, Iwata (1999) argues that passivization also affects the indexing of a conceptual argument. However, the index is not deleted, as in the case of middles (the focus of Iwata’s analysis), but is changed to the status of a “chômeur,” indicated by the subscript “C,” as in (187c).79 The second line gives the conceptual structure (CS). (187)

a. transitive verb CS: F([X]i , [Y]j )

b. middle c. passive CS: F([X], [Y]j ) CS: F([X]C , [Y]j )

If we look at the mechanics of this proposal, it bears a striking resemblance to Grimshaw’s analysis – one argument is singled out and its projection status is neither that of a full, indexed argument nor of an incorporated, non-indexed argument. It is less clear to me than to Iwata how this analysis is “capable of accommodating an implicit argument with a specific value, as well as properly stating ‘suppression’ of an argument, without invoking ad hoc machinery” (Iwata 1999: 545). Assigning the value C to a lexical variable is about as ad hoc as marking it with a “Ø”. The advantage of Grimshaw’s proposal is that it generalizes over more than just the passive and that it correctly predicts that only verbs with external arguments can be passivized. It also has the advantage of using one mechanism, lexical binding, to account for what happens to the external argument in the passive and in the causative alternation: In the passive, the external argument is bound at the level of PAS; in the causative alternation, it is bound at the level of LCS. Only in the former case will the argument still be visible to syntax.

7. Summary The purpose of this chapter was to provide a syntactic analysis of what exactly happens to a verb that is passivized. Throughout this book I argue that the passive in English may have many different surface structures, but that it has a syntactic core, which is not movement of the object, but rather the suppression of the external argument, resulting in an implicit argument. It is only to be expected that what lies at the core of the passive will also play 79. “Chômeur” (literally, “unemployed person”) is a term used in Relational Grammar for a constituent that no longer has any grammatical relation with the predicate.

Summary

153

a role in the use of the passive and in understanding what separates it from related constructions. Therefore, this chapter explored the syntactic status of the implicit argument in some detail. I argued that the external argument of the verb is an integral part of the syntax and semantics of passive constructions and that it has the status of an “implicit argument.” An implicit argument is selected by the verb (hence “argument”), it is part of the semantics of a sentence, visible to syntax in a restricted way, but it is not realized in an argument position. The implicit argument can act as the controller for a PRO subject in an adjunct clause and it licenses, or alternates with, a by-phrase on a regular basis. Psycholinguistic experiments show that the implicit argument becomes available as an event participant in the interpretation of the sentence as soon as the passivized verb is parsed. Data from language use indicate that speakers are aware of the presence of this element in the passive and may choose to use the passive for exactly that reason. With respect to the syntactic analysis of the implicit argument, I rejected the “passive morpheme as clitic” analysis, because there is no convincing independent evidence for the pronominal status of the affix. The “passive morpheme as pro” analysis is even more problematic, as it treats the implicit argument as a full argument, which is not in line with its limited syntactic potential, a potential that seems to be restricted to establishing co-reference relations (control, binding of an anaphor) that do not depend on a specific syntactic configuration (like c-command or government). I also discussed lexical approaches to the implicit argument and compared the implicit argument in the passive to other kinds of understood arguments, such as the implied object in object deletion (John is eating) or fully incorporated themes or goals, as in to butter or to pocket. They are different in that they do not get the interpretation of an existentially bound variable (John is eating does not mean John is eating something) and they do not license a controlled PRO (*John is eating-IMPi PROi raw). In the passive, the external argument is visible to syntax. Since passivization affects only external arguments, I argued that the “suppressed” or “implicit” argument is created at the level of argument structure (PAS). This is in line with observations from psycholinguistics that also locate the implicit argument in the verb’s argument structure. There are no implicit argument effects in ergatives or with incorporated arguments, as they involve the binding, or deletion, of an argument at LCS. If an argument does not make it to PAS, it will not be visible to syntax. The existence of elements that are syntactically active but not syntactically projected may be “conceptually problematic” (Bhatt and Pancheva 2001: 24), but declaring an element a syntactic

154

Chapter 4 The implicit agent in English passives

argument when it does not behave like one would be equally problematic – and not in line with the data. The absorption analysis is certainly elegant, in that explains why the passive participle can assign neither an external theta-role nor structural Case – both are absorbed by the passive morpheme. As a morpheme-centered analysis it is also perfectly compatible with the observation that passives across the world’s languages are generally morphologically marked (Haspelmath 1990). In the analysis suggested here, there is nothing that prevents a passive without a morpheme or a passive participle that can assign Case. But then, there seems to be a lot of variation with respect to Case absorption – in Ukrainian, for instance, the morpheme seems to absorb nominative Case (Baker 1988b) or, optionally, no Case at all (Goodall 1993). Baker, Johnson and Roberts (1989) therefore suggest that the Case absorption properties of the passive morpheme be treated as a parameter with different settings. A different proposal, in the Minimalist framework, comes from Reinhart (2000): Accusative Case is a feature of the verb that is normally checked off by the internal argument. Basically, this feature is a valency or “arity” marker and since the verb still has two arguments (although one is saturated at argument structure), its arity value should not be affected. However, since there is only one argument to be projected and since this argument has to satisfy the EPP, there must be another element – the passive morpheme – to check off the accusative Case feature. Since (passive) saturation cannot cancel accusative case, it is entailed that all languages should mark passive morphologically somehow, which appears to be the case. Even the morphologically poor English, which, as we shall see, does not mark reduction operations, marks its passive operation (Reinhart 2000: 50).

This analysis keeps absorption of the theta-role and absorption of accusative Case separate: The former happens at the level of argument structure (i. e. in the lexicon), the latter happens in syntax. Thus, one can account for the unaccusative structure of the passive without resorting to declaring the passive morpheme itself the implicit argument.

Chapter 5 The use of the passive in academic discourse: A case study 1. Introduction In this chapter I will examine the occurrence of the passive and the realization of agenthood in a discourse type known to be quite sympathetic towards the use of the passive: academic discourse.80 The data analysis will mainly be based on a small corpus of 160 abstracts from different academic disciplines. The point of the study is to find out about the role of the passive in a genre that is highly formalized (a plus for the use of the passive) and in which we are confronted with the paradoxical situation that the individual agent is very important (publishing articles is all about presenting one’s new work and individual claims), while the linguistic expression of agenthood, especially if the agent is identical with the writer, is conventionally minimized (Biber 1988; Atkinson 1992; Orasan 2001). Abstracts have been chosen because they are self-contained texts and present one of the densest forms of scientific writing. Abstracts add one more layer to the long process of research, writing, and polishing that goes into transforming a manuscript into a published research article, and the language resulting from this process has been described as highly integrated, abstract, and impersonal (Biber 1988; Swales 1990; Biber et al. 1999). Results from the abstract study will be complemented with more representative data from the category “Science” of the FROWN corpus, which is comprised of 80 text excerpts from scientific articles and manuals, covering disciplines like physics, biology, and chemistry, and with diachronic data from the ARCHER corpus. To outline this chapter: After a characterization of the abstract as a genre (section 2) and an overview over the tradition of the passive in scientific writing (section 3), I will focus on the linguistic realization of two kinds of events that are part of any published research: (a) material events that refer to a study that has been carried out (these will be referred to as “reported events”); and (b) the construction of the argument on the basis of 80. Part of this chapter is based on joint work with Heidrun Dorgeloh (Dorgeloh and Wanner 2003, in press) and is reprinted with kind permission from John Benjamins Publishing Company, Amsterdam/Philadelphia.

156

Chapter 5 The use of the passive in academic discourse: A case study

the data that have been collected or will be reinterpreted (referred to as “reporting events”). In both types of events, the agent of the verb that expresses the event usually corresponds to the author of the text. I will examine how agenthood is expressed in these two different types of events (section 4) and if there are noticeable differences among the disciplines, especially with regard to the use of the passive. Data from a historical corpus will complement the analysis (section 5), and the use of the passive will be put into a context of the rhetoric of scientific discourse. The overall goal is to gain further insight into discourse conditions that favor or disfavor the use of the passive and to explore which constructions the passive competes with.

2. Characterization of the abstract as a genre Scientific discourse is considered to be a genre in which the expression of agenthood is minimized (Biber 1988; Atkinson 1992; Orasan 2001). Syntactically, this is reflected by a high number of impersonal and agentless constructions, such as nominalizations (a regression standardization revealed that . . . ), gerunds (after separating the molecules . . . ), or short passives (it was found that . . . ). Among the reasons for the prominence of these constructions is the genre-specific need for objectification, which means that the individual author (who is usually reporting his or her own research) is generally not profiled. However, scientific genres are not purely informative. Biber et al. (1999: 16), for example, characterize academic discourse as a genre whose main communicative purpose includes argumentation (for or against a specific analysis). The persuasive function of scientific texts is also recognized in academic writing textbooks (Alley 1996; Hansen 1998; Penrose and Katz 1998). While the research article itself, the “master academic narrative of the last fifty years” (Swales 2001: 50), goes back to the establishment of the Philosophical Transactions of the Royal Society of London in 1665 (Bazerman 1988; Swales 1990; Atkinson 1996), the abstract is a more recent genre. According to Gross, Harmon, and Reidy (2002: 176), abstracts were only used occasionally in nineteenth century and early twentieth century articles and evolved as a replacement for long, descriptive titles and lengthy summaries of early research articles. Abstracts became a more regular phenomenon in the second half of the twentieth century81 and they are now a 81. In the first quarter of the twentieth century only 14 % of the journals studied by Gross, Harmon, and Reidy (2002) have abstracts preceding the articles, but 81 %

Characterization of the abstract as a genre

157

standard requirement in many scholarly journals, particularly in the experiment-based sciences. With the advent of online publishing the role and visibility of abstracts have increased steadily. The influential manual of the American Psychological Association (APA) advises writers to prepare an abstract with care, as it can be “the most important paragraph in your article” (APA 2001: 12) and points out that an abstract, once printed in a journal, will have a life of its own as part of electronic abstract databases, such as PubMed, a database of citations and abstracts of research published in the life sciences. Many print journals are also available online and they often give non-subscribers free online access to abstracts of current articles, while access to the articles themselves is restricted to subscribers or can be purchased for individual articles. In this context, an abstract does not necessarily occur in conjunction with the article that it summarizes, and the decision about whether or not a potential reader will pay for access to the full article may well depend on how clearly the abstract states what the article is about and which argument will be made. Another criterion for buying access to an individual article may be the buzz a particular article has created in the scientific community. Cell, for example, a leading journal in biochemistry and molecular biology, keeps a list of its “Top 20” on its website – these are the articles that have been downloaded most frequently over the last 30 days. Swales (1990: 58) defines genres as communicative events that have the same communicative purpose and a comparable rhetorical structure. As a “genre of distillation” (Swales 1990: 179) the abstract usually “considers the article as a whole and then makes a representation of it” (Bazerman 1988: 220). Consequently, the textual structure of an abstract often reflects the structure of the research article it precedes. In the experimental sciences this structure has been characterized as Introduction-MethodResult-Discussion (IMRD), see Swales (1990) and Hansen (1998). One obvious difference between the abstract and the article itself is that with the (implicit or explicit) constraints on length, only the most essential information can be given in an abstract. Journals may specify a word limit in their guidelines for submissions, but these limits are not always kept. For example, Applied Psycholinguistics sets a limit of 120 words, but among the 20 abstracts from APL examined here some are substantially longer (up of the articles in the third quarter and 95 % of the articles in the last quarter of the twentieth century have abstracts preceding the articles (Gross, Harmon and Reidy 2002: 176). This is in line with the data for the journals analyzed in this chapter: Science started printing abstracts in 1977 and AJES printed summaries as early as 1946 (Vol. 5), but started calling them “abstracts” much later (in 1964).

158

Chapter 5 The use of the passive in academic discourse: A case study

to 283 words). The APA manual strongly recommends that an author stay within the 120-word limit for a very practical reason: Some abstracting services will simply truncate the abstract to fit their databases, in which case only part of the abstract would be stored (APA 2001: 13). The example in (188) is a good illustration of an abstract that follows the IMRD structure. The first sentence establishes the general topic, the second sentence places the article in the context of existing research, the third sentence states the method of analysis and the core finding, and the last sentence points to a larger context again. (188)

Studies on the acquisition of a third language (L3) in a bilingual context have shown that literacy in two languages facilitates the acquisition of a third (Cenoz & Valencia, 1994; Swain, Lapkin, Rowen, & Hart, 1990). The present study seeks to contribute to this line of research by comparing the acquisition of English as an L3 by Catalan/ Spanish bilingual high school students in an immersion program with the acquisition of English by Spanish monolinguals. Data from 201 participants were submitted to a hierarchical multiple regression analysis, rendering results that show that bilingualism indeed has a positive effect on the acquisition of an L3. The evidence is discussed from a cognitive perspective. (APL7) 82

Note that although agentive verbs are used (seek, submit, show, discuss), they occur in constructions without the explicit expression of an agent. In most of the cases, the “true” agent of the action expressed by the verb would be the author(s) of the paper – they are the ones who are seeking to contribute to studies on the acquisition of a third language, they are the ones who intend to show that bilingualism has a positive effect on the acquisition of a third language, and they are the ones who discuss the data, but there is not a single explicit reference to the authors, neither in first person nor in third person. This is in line with the observation that scientific discourse shows a high proportion of non-animate subjects, even with verbs expressing activities and acts of communication or cognition.83 The passive is one 82. Throughout this chapter, abstracts are identified by the journal they appeared in (here: Applied Psycholinguistics) and by their consecutive number in the corpus. 83. Biber et al. (1999: 378) found that in the Longman Corpus of Written and Spoken English (LWSE) the category of academic prose stood out in its tendency to use abstract subjects even with verbs whose core meaning normally requires an agent. They report that 10 % of mental verbs (such as find or prove), 20 % of communication verbs (such as explain or suggest), and overall 30 % of all activity verbs (such as make or give) occur with inanimate subjects.

Characterization of the abstract as a genre

159

construction that can be employed to avoid agentive subjects. Section 4 will look in more detail at the way agenthood is expressed in abstracts and at the role of the passive in comparison to other constructions that allow for non-agentive subjects. While abstracts in conference brochures sometimes only describe the research question without giving any answers (Yakhontova 2002), abstracts that precede published articles generally state the main findings or thesis of the article - after all, the full article has already been written. In some journals, however, abstracts regularly end with the formulaic announcement that results and implications will be discussed, without any specific information about the direction of that discussion. Although there is considerable variation in length and structure of abstracts, an abstract that is really only a list of direct questions, without as much as a hint at the answers to these questions, like the one in (189), stands out as not quite fitting this textual and syntactic pattern: (189)

What missions have these institutions chosen? Which models are these institutions following? What challenges do they face? How are they coping with their relationship to existing public universities? How are they dealing with tuition and budgetary issues? What sources of funding are they tapping into? How are they dealing with issues of equity, excellence, and access? In the composition of their faculty, student body, and curriculum, are these institutions sufficiently international? How are the funding institutions and universities in the West helping these fledgling universities? Are these trends expected to accelerate in the future? We attempt to answer these questions within the context of a few case studies that reflect the breadth and diversity of this revolutionary growth in post-secondary higher education. (AJES19)

While there may be variation with respect to the structure and the function of an abstract, two components that are part of every abstract are: (i) information about the topic and the data that are to be interpreted; and (ii) a summary of the argument that is made in the article. Both of these components and the role of the passive in realizing them will be discussed in more detail in sections 4.3. and 4.4. in this chapter.

160

Chapter 5 The use of the passive in academic discourse: A case study

3. The tradition of the passive in scholarly writing The passive is known to be a construction that is characteristic of scholarly writing and “using it appropriately is part of learning how to speak as a professional scientist” (Penrose and Katz 1998: 170). Why, then, did an English biologist who wrote a letter to the editor of Nature in 1996 stating his preference for the passive meet with passionate criticism from his colleagues? The reason is that the passive is not just a syntactic construction; it is also considered the linguistic embodiment of a particular conception of the work of a scientist. According to the letter writer the passive is preferable over the active because it allows an “unbiased viewpoint,” supporting “disciplined writing,” in which “cases must agree” and “tenses must be used correctly” (Leather 1996). By contrast, using the active voice is portrayed as “an easy option” with “no need to discipline one’s thoughts,” encouraging “carelessness” and “partisanship.” Leather argues against switching between the active and the passive, as it results in “a paper full of inconsistencies and, of course, a general mixture of styles.” He even goes as far as to suggest that use of the active voice, with its perceived carelessness, ultimately paves the way to fraud. In follow-up letters to the editor Leather’s views are criticized as “superstitions” (Kirkman 1996) and “outrageous displays of scientific hypocrisy” (Perlman 1996). From Leather’s vague references to tense and case agreement one can see that he is not really concerned with the syntax of the passive, his comment is all about ideas associated with the passive as a marker of a specific style of academic writing. Others have quite different associations. In his handbook on clear and graceful writing, Williams (1999) links a preference for the passive to a not so graceful state of mind: If your cast of mind is impersonal and aloof, you will typically combine passives with nominalizations. . . . When writers in the professions combine nominalizations with passives, they create what readers variously call sociologicalese, educationalese, legalese, bureaucratese – a kind of prose written by those who confuse authority and objectivity with polysyllabic abstraction and remote impersonality. (Williams 1999: 79)

In general, it is recognized that the “‘scientific passive’ has a long and venerable tradition” (Penrose and Katz 1998: 45), but the ideas about what is regarded as good scientific writing have changed, and the passive seems to embody a tradition of writing that is not desirable anymore. It is characterized as a linguistic construction to express “the detached persona,” which “[f]or at least a century” was considered “the hallmark of good scientific writing” (Hansen 1998: 437). The detached persona reflects the belief that

The tradition of the passive in scholarly writing

161

science is all about the “dispassionate description of truth” (Penrose and Katz 1998: 169), and that a distinction can be made between those who uncover “facts” (the scientists) and those who decide how to interpret and use them (the audience). It gives the impression that the writer is objective and neutral, an observer and recorder of facts rather than someone who is actively involved in creating and interpreting data. Typical of the detached persona are the avoidance of the first person and the use of constructions with non-agent subjects, such as the passive. Gross et al. (2002: 19) describe how scientific discourse changed from being based on the “confidence that the data will speak for themselves” to a genre in which a finding has only the scientific significance that scientists “bestow upon it by the power of their argument” (Gross, Harmon and Reidy 2002: 26–27). While scientists have always aimed at presenting empirical evidence, making a convincing argument for a specific analysis in the context of existing research has become more and more important. This development has led to an increasing visibility of the scientist as someone who interprets findings and constructs a persuasive argument. To this purpose, style manuals for academic writing give blanket advice to “use the active rather than the passive voice, calling the active voice more “vigorous” (APA 2001: 41) and “straightforward” (Alley 1996). The underlying idea is that the active is maximally clear in terms of who did what to which effect. The passive, with its less asymmetrical organization of event participants and its potential ambiguity between a state and an event reading, is perceived as a construction that “saps . . . energy and leaves dead words on the page.” (Alley 1996: 108). These sweeping generalizations about one voice being better than the other illustrate that the passive is often regarded as a marker of a particular style of writing and that active voice is associated with more visible agents. Let us look at the linguistic reality behind those strong views: A specific point of criticism is that in the passive the agent or cause of an action is not explicitly expressed. It has already been discussed that the passive may be chosen because the identity of the agent is not known or not important. In contrast to other constructions with nonagentive subjects the passive, however, includes an implicit agent, i. e. the interpretation of an agent is part of the sentence, even if its exact identity is not revealed. In scientific writing, there is an additional reason for not mentioning the agent: Agents in scientific writing are often the authors themselves, who report their own doings and conclusions. They are the ones who set up an experiment or design a survey, who record and interpret data, who draw conclusions and make suggestions. Even though modern style manuals en-

162

Chapter 5 The use of the passive in academic discourse: A case study

courage the use of the active in the context of these verbs, many scientists still consider the use of the first person inappropriate in scientific writing (Alley 1996: 107). Using the passive allows the author to avoid the first person without recurring to an awkward third person (The authors show . . . ). Although there may be various reasons why an author would rather not use the first person, the use of the passive is often construed as an attempt to report events from a “neutral” perspective. However, the traditional view of the scientist as a neutral observer has undergone drastic changes (Bazerman 1988), and the detached persona is now something that is cautioned against, as it can “engender a false faith in science” (Penrose and Katz 1998: 171). Guidelines published by professional organizations such as the American Psychological Association and editorial boards as well as textbooks on scientific writing promote a more constructional style of presenting an argument. They stress that “knowledge is something humans create as much as discover, and that the style of scientific language ought to openly acknowledge the role humans play in their own research” (Hansen 1998: 439). This holds in the discussion and findings section of an article in particular, where the author can present a result as something that is self-evident or as an interpretation by himself or herself. Again, from a syntactic view it is rather striking that the passive – rather than truly agentless constructions – is blamed as a construction that presents events as happening without interference from an agent. But not all use of the passive is regarded as inappropriate. The passive is considered acceptable in expository writing “and when you want to focus on the object or recipient of the action rather than on the actor” (APA 2001: 42). The examples that are given include situations where the actors in question are the researchers themselves, for example when an experiment is carried out. The need to express the involvement of the author explicitly is not as pressing in this context as it is in the discussion section of an article. For example, if the description of an experiment says that a stimulus is shown on a screen, the identity of the person who makes the picture appear on the screen is not all that important. However, passives in the argumentation section of an article are less acceptable – not because the identity of the agent could not be recovered from context (who should argue but the author?), but because it is common practice to tie an argument to the author’s name. In addition to being “detached,” the passive is also considered to be less “dynamic” than the active, which, again, does not conform to the ideal of modern scientific writing. However, since passivization does not have any effect on the aspectual structure of the sentence (activities remain activi-

The tradition of the passive in scholarly writing

163

ties, states remain states, accomplishments remain accomplishments), the notion of dynamicity underlying this criticism cannot really be aspectual. In addition, when one looks at the data that are presented to illustrate this point, one realizes that the criticized constructions in question are not necessarily passives at all. The APA manual, for example, lists (190a) as an example of an unnecessary passive and suggests (190b) as an improvement (APA 2001: 42). (190)

a. b.

The participants were seated in comfortable chairs . . . Participants sat in comfortable chairs . . .

However, the participants were presumably not seated in their chairs by some unknown agent like teddy bears on a shelf, they were simply in a sedentary position (if not, (190b) would not be equivalent in meaning to (190a)). No change of state is implied in (190a), which is why the verb can be replaced with the state verb sit as in (190b). Sentence (b) may be “better” than sentence (a), but the choice is not between active and passive; it is between a one-word state verb (sit) and a light verb construction (be seated). The situation is exactly the same in (191): (a) is criticized as a needless and inefficient passive (Alley 1996: 105), and (b) is presented as the improved nonpassive version. Again, the criticized example is not a passive, but a copula construction, this time involving the adjective composed. What is more, the “improved” version actually changes the meaning of the original sentence. While (a) implies that a sapphire optical fiber is the sole, or at least the main, ingredient of the feedthrough, it could be just one of many ingredients in sentence (b). (191)

a. b.

The feedthrough was composed of a sapphire optical fiber, which . . . The feedthrough contained a sapphire optical fiber, which . . . .

Alley (1996: 105) does not seem to distinguish between the semantics of copula be and auxiliary be when he claims that be in the passive “acts as an equal sign.” If that were the case, all passives would be states, but crucially, passives belong to the same aspectual class as their corresponding actives, and if the active has an agent, so will the passive. Misconceptions about what constitutes a passive add to the generalization that passives are emblematic of a particular style in scientific writing. Another example of misconceptions about the passive is exemplified in the style sheet for Science. It advises authors to use the active voice “when suitable, particularly when necessary for correct syntax” (Science 2007). Now, of course an un-

164

Chapter 5 The use of the passive in academic discourse: A case study

grammatical passive should be avoided, as much as any other ungrammatical construction, but the data that are given to illustrate this point, given in (5a) and (b), are ill chosen. (192)

a. b. c. d.

To address this possibility, we constructed a lambda Zap library . . . To address this possibility, a lambda Zap library was constructed . . . Two aspects of grammar – binding and control – were assessed to determine . . . (APL 17) To rectify these weaknesses, child-generated normative prosocial behaviours were used to generate peer nomination items for the purpose of rating . . . (JME8)

The incriminated sentence, (192b), is perfectly grammatical. It includes a preposed purpose clause, the subject of which is controlled by the implicit agent of the passive, just as in the data in (c) and (d), which are from published abstracts. In fact, sentences like these are often used in linguistic theory to make the point that the passive has an implicit agent, which can control the PRO subject of a purpose clause (see chapter 4). If the passive is to be avoided here, it is not because it is ungrammatical. A trend against the use of the passive can also be observed in related genres. Authors who write for the ubiquitous self-help book series “For Dummies” (Digital Photography for Dummies and Chinese Cooking for Dummies are just two of many titles), for example, are instructed explicitly not to use the passive. If they do, the passive will be edited out. This may lead to distortions of the intended meaning, as reported by the endocrinologist Dr. Alan Rubin, the author of one of the most successful volumes in the series, Diabetes for Dummies: Sometimes I’ll write something like ‘the patient was comatose and was given thyroid hormone,’ and they’ll change that to ‘the patient was comatose and took thyroid hormone.’ . . . I have to tell them these are extremely sick patients, they can’t take care of themselves, they have to be passive whether Wiley [one of the editors, AW] likes it or not. (Donadio 2006)

One cannot escape the conclusion that more often than not style manuals are rather robotic in their approach to the passive. Avoiding the passive is often a higher priority than preserving the semantic integrity of a sentence. This is also the impression one gets from Microsoft Word’s grammar check, which operates on the assumption that “all passives are bad juju” (Pullum 2008). Overall, it seems that a strong sense of whether or not a

Synchronic case study

165

passive should be used is not necessarily based on a clear understanding of the construction. Specifically, the implicit agent reading of the passive is not considered, which is rather like saying that one should avoid infinitives because they do not have subjects.

4. Synchronic case study In the following I will analyze the use of the passive in a genre that may be described as one of the most condensed forms of scientific writing. I will put particular emphasis on the realization of agenthood in events in which conclusions and claims are presented. 4.1. Corpus Information The analysis is based on 160 abstracts from 8 different peer-reviewed journals (20 from each journal), which were published in 1999 and 2000 by academic publishers in the UK or the US: The American Journal of Economics and Sociology (AJES), Applied Psycholinguistics (APL), the Cambridge Journal of Economics (CJE), Cell, a journal in biochemistry, English Language and Linguistics (ELL), the Journal of Moral Education (JME), the Journal of Women’s History (JWH), and Science. The journals, most of which appear quarterly, were selected to represent different academic disciplines. 84 The abstracts are generally written by the authors of the articles, even if they refer to the authors in the third person (as in Bucur examines how . . . ), something that only occurred with some regularity in JWH. In the natural sciences, most articles have multiple authors and it is impossible to know if the abstract is written by a single person or if it is the result of a collaborative effort. According to Wood’s (2001) estimations, about 45 % of research articles in an international journal like Science (and, consequently, abstracts that precede these articles) are written by non-native speakers of English. 84. At the time of data collection, Science had four issues per month, Cell appeared biweekly, CJE bimonthly, and ELL semi-annually. Science, which bills itself as a “magazine,” does not just stand out in terms of publishing frequency. It is also thematically less specific than the other journals, it has the shortest abstracts (see Figure 1), and the fewest reporting events in the abstracts. It is included in the corpus because it is consistently the highest-ranked journal in the natural sciences (Institute for Scientific Information 2007).

166

Chapter 5 The use of the passive in academic discourse: A case study

Table 1. Length of abstracts (does not include abstract title) Sci

CJE

Cell

AJES JWH JME ELL

APL

Words per 20 abstracts

1924 2124

2433

2890

3410

Mean length of abstract

96.2

Overall

160 texts, 21804 words

2899

106.2 121.7 144.5 145

2930

3194

146.5 159.7 170.5

This is an issue that cannot be explored here. I will simply assume that all published material that has been peer-reviewed is within the limits of international scientific English (Crystal 1998). There are no formal instructions on how to write an abstract for a specific journal, with the exception of constraints on length (for instance, APL sets a limit of 120 words – the length recommended in the APA manual – and JME asks for “about 150 words”), but these limits are not always respected (the variation of length in abstracts from APL, for example, reaches from 93 to 283 words), see Table 1. The abstracts examined here are all abstracts that directly precede full articles. This includes articles that are responses to other articles or articles in a special edition of a journal (for example, Volume 29 of the Journal of Moral Education is devoted to the discussion of the contributions of John Wilson to the field). In some journals short squibs and book reviews are also preceded by abstracts, but these were not considered, as they are very short and different in structure. A corpus of this size and uniformity is very manageable, however, it is difficult to generalize any of the results from this study. In order to check if the distribution of the passive is in any way out of the ordinary, results from the abstract corpus will be compared to data from a larger, more representative corpus. The category “Science” of the FROWN corpus consists of 80 text excerpts of about 2000 words each. It was chosen because as a syntactically tagged corpus it allows for the extraction of all passive constructions. 4.2. Distribution of the passive: Preliminary findings Since abstracts are one of the densest forms of scientific writing, and since scientific writing is known for its high proportion of non-agentive subjects, the number of passives in this particular genre can be expected to be rather

Synchronic case study

167

high. This expectation is borne out: There are 371 tokens of passives, or 17 passives per 1000 words, none of them a get-passive. This is in the same range as the results presented in Biber et al. (1999).85 In their corpus of 1 million words of academic prose there are about 17,500 tokens, or 17.5 tokens per 1000 words, of passives with “dynamic verbs” (this excludes adjectival passives or resultative readings). For comparisons of data from different corpora it is standard procedure in corpus linguistics to normalize the frequency of a construction to the number of words in the corpus (Biber, Conrad and Reppen 1998; Meyer 2002). The frequency of the passive in fictional texts is compared to the frequency of the passive in academic prose by counting the occurrences and dividing them by the number of words in the corpus. Applying this procedure, Biber et al. (1999) arrive at the result that the frequency of the passive in academic prose is more than twice as high as in general fiction (about 17.5 vs. 6.5 tokens per 1000 words). However, there may be another way to measure, or at least to describe, the frequency of the passive. Biber et al. (1999: 939) point out that it is remarkable that the “passive frequency is lowest in conversation in fiction, although these are the registers with the highest frequency of lexical verbs.”86 Lexical verbs are more frequent in conversations than in writing because clauses – a syntactic unit that requires a verb – tend to be shorter in conversations (Biber et al. 1999: 66). In fiction, lexical verbs are also relatively frequent because a story may contain extended passages of dialogue. Academic prose, on the other hand, is known to have a rather low proportion of lexical verbs, but among them a high proportion of passives. This is confirmed by data from the abstract corpus: There are 1471 lexical verbs, or 67.5 lexical verbs per 1000 words (in the LSWE corpus, on average, every tenth word is a verb; see Biber et al. 1999: 65). If one considers the verb phrases projected by these verbs as potential passives – pretending that each verb can be passivized and that using the passive is a choice that is made once the rest of the sentence is settled on – one can see that the proportion of passives among the verbal constructions is indeed very high: approximately 1 out of 4 lexical verbs (371 out of 1471, or 25.2 %) occurs in the passive. A comparison to data from the FROWN corpus confirms 85. These numbers are calculated on the basis of the information stated in Biber et al. (1999: 938, Table 11.9). No absolute numbers are given - the authors prefer to report rounded frequencies at a level of precision that they judge to be replicable (Biber et al., 1999: 39–40). 86. A diagram illustrates that the number of lexical verbs in fiction and conversation is about 50 % higher than in academic prose (Biber et al., 1999: 65, Figure 2.2.). No exact numbers are given.

168

Chapter 5 The use of the passive in academic discourse: A case study

that the frequency of the passive in the abstract corpus is remarkably high. Of 160,914 words in the category “Science” of the FROWN corpus, about 11.3 % (18,176) are tagged as lexical verbs. This number includes instances of be, have, and do as main verbs (tagged as VVBN, VVDN, VVHN). However, a manual check of the words tagged as VVN, i. e. as past participle of a lexical verb (n=3890) resulted in a reclassification of almost 10 % of these tokens (n=353) as adjectives. Compare the use of related in (193). In (a) it describes a state and in (a’) it expresses an event. Consequently, only (a’) was counted as a passive. (193)

a. a’

The quantity b is a measure of clustering and is related to gravitational correlations. (FROWN; J01)87 The story is related that Randall went back to the original experiments of Hertz and . . . extended the Hertz resonant ring into a cylinder . . . (FROWN; J80)

More examples of constructions that were tagged as participles but that were manually reclassified as adjectival passives are given in (194). Data like these were not included in the passive count. (194)

a. b.

c.

The tree-and-branch approach is well suited to the characteristics of distributed video services. (FROWN; J74) Parenthood teaches both the parents and the children moderation. Since children’s reasoning powers are not developed, parents must find the mean between arguments and force . . . (FROWN; J62) Polymorphonuclear leukocytes . . . filled the endometrial gland lumens, and were scattered throughout the superficial stroma. (FROWN; J11)

The following table compares the frequency of the passive in the abstract corpus with larger corpora. Of the 371 passives attested in the abstract corpus, fewer than 1 in 5 (70 or 18.9 %) are long passives, i. e. passives with a by-phrase.88 This number, 87. If not noted otherwise, bold font in the data has been added to highlight certain words or constructions. 88. This is the only parameter in which data from the abstract corpus differ clearly from data in the LSWE corpus: In the abstract corpus (21,804 words), there are 77 long passives. In a 200,000 word sub-corpus of LSWE, i. e. a corpus that is about ten times as big as the abstract corpus, there are only 242 long passives (Biber et al., 1999: 940, Fn. 2 and 5 – numbers for short passives are not given).

Synchronic case study

169

Table 2. Proportion of passives Abstract corpus

FROWN corpus, category “Science”

LSWE corpus, “Academic Prose”, Biber et al. (1999)

Corpus size (words)

21,804

160,914

ca. 1,000,000

Number of lexical verbs

1,471

17,823

no exact numbers given

Number of verbal passives

371

2698

about 17,500

Passives per 1000 words

17

16.8

ca. 17.5

Passives per 100 lexical verbs

25.2

15.1

cannot be calculated

too, is in line with findings from other corpora, see Thompson (1978). The by-phrases are generally informationally and syntactically “heavy,” expressing new information and containing modifiers, often relative clauses. Again, this number confirms findings from studies carried out on the basis of larger corpora. Biber et al. (1999) report a clear tendency for the by-phrase to be longer than the subject and to express new information. This is due to a preference for identifying the subject position with given information (often expressed through pronouns, i. e. one-word noun phrases) and positioning new information and complex phrases at the end of the sentence (principle of end-weight). In the abstract corpus, by-phrases are typically not animate (51 out of 70, or 72.9 %) and they are often heavy, as illustrated in the following examples. The subject of the passive is rarely a pronoun in the abstract corpus. (195)

a. b. c. d.

The unique referential role of else is then highlighted by contrasting it with its closest synonym . . . (ELL 6) . . . from the vantage point of someone whose sense of these things has been shaped by the Aristotelian tradition (JME 19) . . . while vinculin, zyzin, VASP, and Mena are recruited to adhesion zippers by a mechanism that requires . . . (Cell 2) . . . focus on structural relationships between/among social groups can be, in the first instance, occluded by an ethical per-

170

Chapter 5 The use of the passive in academic discourse: A case study

e. f.

spective that centres on the welfare of discrete individuals (JME 7) . . . and the performance of both groups was influenced by the complexity and semantic ambiguity of the sentences (APL 15) . . . that these were reinforced by macroeconomic disturbances at the demand side (CJE 10)

While there are a number of passives that refer to actions carried out by the authors of the paper, there is not a single case in which a by-phrase refers to the author(s) of the paper. All passives in which the agent is the author of the paper (either as the agent in carrying out an experiment or as the agent in constructing an argument) are short passives, as in (196): (196)

a. b. c.

It is shown that such an approach does not yield adequate results. (ELL 5) Three distinct intermediates were identified . . . (Cell 6) Several methodological issues are discussed. (CJE 3)

Of the 371 attested passives, 78 (21 %) occurred in participle clauses, most of them (63 out of 78) in participial relative clauses. That means that one in five passives occurred without an auxiliary, a factor that needs to be taken into consideration when searching for passives in a large corpus using software tools like WordSmith. For example, Xiao, McEnery and Qian (2006) state in their corpus-based analysis of the passive that they looked for constructions in which be was directly followed by the past/passive participle. This does not just rule out passive sentences with adverbs, as in (197a), or negation, as in (10b), or parentheses, as illustrated in (195d), or coordination structures, as in (10c), it also completely ignores passive constructions that occur without auxiliaries, as in (198), even though the authors recognize that “the passive meaning is essentially expressed by past participles” (Xiao, McEnery and Qian 2006: 111). The omission of passives without auxiliaries is particularly significant, as more than 1 in 5 passives in the abstract corpus and about 1 in 4 passives in the category “Science” in the FROWN corpus occurred without auxiliaries.89 Passives without auxiliaries include particip89. The passives that were extracted from the FROWN corpus include passives without auxiliaries. Of 2,698 passives, 656 occurred without a form of be (or, for that matter, get) in the verb phrase. This number does not include any adjectival passives. The WordSmith parameters were set in such a way that past participles which occurred without an instance of the perfective auxiliary were scanned. The vast majority of passives without auxiliaries occurred in participial relative clauses.

Synchronic case study

171

ial relative clauses, as illustrated in (198a–c), as well as participial adverbial clauses, as illustrated in (198d–e). (197)

a. b. c.

(198)

a. b.

c.

d. e.

The Galatry line shape . . . is usually used to model the profiles of absorption lines (FROWN; J02) Patterns of weaning could not be assessed because only adults were studied. (FROWN; J13) Equivalent amounts of protein/lane were analyzed by electrophoresis . . . , transferred to nitrocellulose . . . , and immunoblotted with a mAb . . . (FROWN; J09) . . . they include several earlier references overlooked in the previous compilations (FROWN; J02) . . . to remain in a state of arrested development within the relatively protected environment provided by their hosts (FROWN; J08) A British mission led by Tizard (Sir Henry at that date) departed on 30 August for Washington bearing, inter alia, the multi-cavity magnetron, described in 1946 by one enthusiastic American scientist as “the most valuable cargo ever brought to our shores (FROWN; J80) As pointed out in the previous section, melt fracture did not appear to play a role . . . . (FROWN; J02) As shown, wives and husbands perform significantly different total amounts of labor and different types of labor. (FROWN; J26)

Another example of a corpus-based analysis that does not consider all passives in a corpus is Seoane’s study on the use of the passive in scientific British and American English based on the FROWN, FLOB, BROWN, and LOB corpora (Seoane 2006).90 She argues that that in the course of thirty years (the LOB and BROWN corpora are from 1961, while FLOB and FROWN are from the early 1990s) there has been a “drastic drop in the use of passives,” particularly in American English (Seoane 2006: 199). One of her main finding is that in the earlier BROWN corpus 34 % of all transitive verbs (i. e. verbs that can potentially occur in the passive) occurred in active voice. This number rose to 58 % in the later FROWN corpus (the 90. Seoane (2006: 192) worked with a 30,000 word sample of the FROWN corpus (all texts are from the category “Science”), for which she reports 480 passives (all be-passives in finite verb groups), compared to 663 finite transitive verbs in active voice.

172

Chapter 5 The use of the passive in academic discourse: A case study

comparable numbers for British English are 32.9 % in LOB and 43.2 % in FLOB). Seoane interprets this decrease of the passive in both corpora as a reflection of a shift towards a writing style that praises the use of the active and considers passives “functionally empty” (Seoane 2006: 204). What is problematic about her analysis is that she only considers “‘central’ bepassives occurring in finite verb groups” (Seoane 2006: 192). This means that she, too, misses out on all passives that occur in participle clauses. As discussed in chapter 3, there is really no principled reason to exclude bare passives from a discussion of the form and function of the passive construction, especially in a type of discourse that values compact expressions, such as scientific writing. The proportion of by-phrases among these bare passives is slightly higher than in the abstract corpus as a whole (29.6 % vs. 18.9 %). Again, byphrases are generally non-animate and they never refer to the authors of the paper, as illustrated in (199). (199)

a. b.

. . . explained all the variance contributed by kindergarten writing (APL 6) . . . OAZ contains a BMP signaling module formed by two clusters of fingers (Cell 4)

The generalizations reported so far by and large confirm results from studies based on larger corpora, with the exception of a comparatively low animacy rate for by-phrases. In the next section, I will explore in detail the role and the visibility of the author with respect to functionally defined verb classes. 4.3. The distinction between reporting and reported events In abstracts that precede full-scale articles one typically finds factual information about what has been done (why the topic is interesting, which kind of research has been carried out, which methods have been used in order to gain data) as well as an evaluation of these facts and the direction of the argument. I will refer to the first type of information as the “reported events” and to the second as “reporting events” (see Dorgeloh and Wanner 2003). Reported events are often, though not always, characterized by the use of the past tense, as in the following examples: (200)

a.

To test this hypothesis, we studied leptin-deficient and leptin receptor-deficient mice that are obese and hypogonadic. (Cell 1)

Synchronic case study

b.

173

We also investigated the role of alphabetic skills and socioliteracy variables . . . . (APL 6)

The nature of reported events and the verbs expressing them depend completely on the topic of the article. Transitive verbs like store, retrieve, reproduce, diffuse, and initiate are typical of the natural sciences (due to the central role of experiments and the challenge of duplicating results), while in the humanities more cognitive verbs like frame, shape, conceptualize, deconstruct, and engage can be found. However, these are only tendencies, reported events can be just about anything – monasteries in Siberia, the role of forgiveness in modern society, or chemical processes in the brain. Reported events may or may not involve the author of the abstract. For example, in (201a) an experiment is reported, and the agent is an implicit argument, but one can infer from the context that the writer of the article was part of the team that carried out the experiment. In (201b) the situation to be analyzed involves events that were not controlled or initiated by the writer of the article, i. e. the author is not part of the reported events. (201)

a.

b.

Information was stored as quantum phase in an N-state Rydberg atom data register. One or more flipped states stored in an eight-state atomic wave packet could be retrieved in a single operation, in agreement with a recent proposal by Grover. (Sci 27) As a result, male peasants in Wuxi migrated more frequently to urban locales in search of work, and women and children who were left behind became the poorest people residing in Wuxi villages. (JWH 4)

Unlike reported events, reporting events – also called “commentary” (Swales 1990) – are by definition invariably linked to the authors of the paper. They establish the identity of the researcher as someone who makes a persuasive argument and who integrates his or her argument into an existing body of research. The role of reporting and the linguistic means by which reporting events are realized are similar across the disciplines. They are realized by verbs with specific semantics and predictable syntactic behavior (Levin 1993). The ones that matter in the context of scientific argumentation are the “verbs of transfer of a message,” such as argue, demonstrate, explain, explicate, show, teach, tell, and the “say verbs,” such as say, note, declare,

174

Chapter 5 The use of the passive in academic discourse: A case study

suggest, reveal, state and claim.91 These verbs are usually realized in present tense (sometimes in future tense). Both types of reporting verbs can take sentential complements and both allow an additional to-PP indicating the addressee or goal of communication, but only the verbs of transfer of a message allow the double object construction (e. g., tell him a story). Verbs of communication that do not matter here include verbs that have incorporated an instrument of communication (e. g. cable, e-mail, phone, and fax), and verbs of manner of speaking (e. g. babble, chirp, growl, murmur, and whisper). Examples of reporting events are given in (202): (202)

a. b.

In what follows, I posit that moral educators can learn not only learn from . . . (JME 15) This paper argues that, based on the ontological insights of critical realism, epistemological guidelines can be established (CJE 11)

Swales (1990) pays great attention to the linguistic realization of rhetorical moves. One of these moves in abstracts is captured by reporting events: This is where factual material is explicitly shaped into an argument. In abstracts without reporting events - which are very rare - it is difficult put one’s finger on the exact argument that is being made. Without a reporting verb, the presence of the agent argument will not be evoked. The abstract in (203) illustrates the effect caused by the absence of reporting events: An abstract without any reporting verb does not only maximally background the writer, it is also a maximal backgrounding of the hedging function of the reporting event. This is not a result of the use of the passive or any other detransitivization strategy (there is only one passive in this abstract, designed), but a result of the lack of reporting events. In a reporting event in the passive the agent may be only an implicit argument, but in an abstract without reporting events, the role of the researcher as someone who shapes and interprets facts will be not represented at all. (203)

Marital status structured poor women’s interactions with the protowelfare state of the early-twentieth-century United States. Deserted

91. Reporting verbs found in the abstract corpus include analyze, argue, challenge, claim, contradict, conclude, demonstrate, discuss, examine, explain, explore, illuminate, illustrate, indicate, propose, reveal, show, and suggest. Of these, show and suggest were the two most common verbs. This is in line with numbers in Biber et al. (1999: 668–669), who also found that the most common verbs followed by that-clauses in academic prose are show and suggest. Outside of academic prose, say and think occurred more frequently.

Synchronic case study

175

women – “husbandless wives” – occupied a liminal category between widows and unmarried women. In New York City, Progressive Era reformers created an antidesertion system that consisted of the family law of the poor, domestic relations courts, and welfare policies designed to defer and deny deserted applicants’ requests for assistance. Reformers’ fiscal concerns and gender assumptions both converged and competed in the new system. The goal of enforcing male breadwinning in order to contain dependent women and children’s poverty within families conflicted with legal provisions that prioritized family “independence,” even if it meant occasionally compelling a wife to support her husband. As evidence from three hundred legal aid case files indicates, most deserted clients engaged in time-consuming, repetitive, and generally fruitless prosecution of their husbands. Ultimately, the antidesertion system disciplined women more than the men who were its ostensible targets. (JWH 3, 155 words) The passive is said to compete with the active, but when it comes to reporting events, the term “the active” is too vague, as could be seen from the abstract in (203). Although there is passive in this abstract, there is no indication of the author’s voice – there are simply no reporting events. What is more, the subjects of reported events are often inanimate, as in the last sentence (the antidesertion system disciplined women . . . ), which adds to the impersonal, non-agentive reading of this abstract. In her study of the FROWN corpus, Seoane (2006: 202) recognizes that “impersonalization . . . does not depend on voice exclusively.” She found that in roughly 3 out of 4 cases, the subject of a transitive verb used in active voice is non-human and gathers “that the increase in actives at the expense of passives does not imply a parallel increase in the frequency of human agents. . . . The hypothesis that passives decrease because the discourse becomes more personal and author-centered, therefore, does not hold” (Seoane 2006: 202).

4.4. The use of the passive in reporting events In the following I will discuss the role of reporting events in different journals and in particular the syntactic realization of the agent argument. I will demonstrate that the main competitors of the passive in the phrasing of an argument are indeed constructions that employ active voice, but not necessarily constructions in which the agent is any more visible than in the passive.

176

Chapter 5 The use of the passive in academic discourse: A case study

4.4.1. The role of reporting: Differences across the disciplines Reporting events are an integral part of research abstracts, but their role varies across the disciplines. For the purpose of this analysis, only reporting events realized by reporting verbs were considered. Reporting events realized through nominalizations (Our central proposition is that . . . ) are rare in the abstract corpus and were not counted. The number of reporting events is not quite identical with the number of reporting verbs that occur in the abstracts. Reporting by somebody else, as in (204), becomes part of the reported material and is not to be considered part of the commentary. (204)

Goswami and Bryant (1990) proposed a theory of reading development based on three causal connections. . . . To explain this connection, they suggested that young readers of English used analogies based on rimes [sic] as one means of deciphering the alphabetic code. (APL9)

Figure 1 shows the absolute and relative number of reporting events in each of the eight journals.

Figure 1. Number of reporting events per 100 words/absolute number

The absolute frequency of reporting events is lowest in Science and Cell, the two journals from the natural sciences. If these numbers are adjusted for abstract length, the picture does not change much. Science, the journal with the shortest abstracts, also clearly has the fewest reporting events proportionally. Relative frequency for Cell is the third lowest. JME and ELL,

Synchronic case study

177

two journals from the humanities, appear at the other end of the spectrum, with 2.6 and 3.0 reporting events per 100 words, or 76 and 96 tokens, respectively (compared to 15 and 31 tokens in Science and Cell, respectively). The two journals from the economic and social sciences, AJES and CJE, also score comparatively high on the reporting event scale (2 and 2.6 reporting events per 100 words92). The middle field is formed by APL and JWH. APL can be grouped with Science and Cell: Articles in these journals are usually experiment-based and it is to be expected that a sizable portion of the abstract is taken up with the description of the experiment. However, abstracts in APL tend to be considerably longer than the ones in Science and Cell, (which explains why the absolute number of reporting events in APL is higher than in Cell (although the relative number of reporting events is actually lower than in Cell). The journal JWH is different, and somewhat atypical, in several respects: The 20 abstracts from JWH show the highest intrajournal variation with respect to the number of reporting events. The mean number of reporting events in JWH (1.48/100 words) results from variation between 0 and 8 reporting events in the individual abstracts. JWH is the only journal in which abstracts of considerable length occur without any reporting event (as illustrated in (203)). As a result, some of these abstracts sound more like excerpts from an encyclopedia (a mere listing of facts) than a condensed version of a scientific argument. To sum up the results presented in Figure 1, it seems that, in the tradition of the “discovery” approach to gaining and shaping knowledge (Rudwick 1985), the interpreter role of the researcher is not as prominent in the abstracts printed in journals from the natural sciences as it is in the journals from the humanities and social sciences. For the latter, the construction of an argument (through reporting events) takes up a considerable portion of the abstract. This difference is underlined by the very low number of reporting events in Science, which cannot be attributed to space restrictions only. Abstracts in Science are shorter than most other abstracts, many have fewer than 100 words, and some have less than 50. They could easily have been longer without any danger for the authors of going over the word limit that is appropriate for an abstract. The authors’ voluntary restriction to the bare-bones essentials of scientific work makes it obvious that apparently no need is felt to focus on the author as someone who is shaping an argument, at least not in the abstract.

92. Numbers are rounded to the nearest tenth.

178

Chapter 5 The use of the passive in academic discourse: A case study

In the next section, I will discuss further the function of reporting events in different academic disciplines as it is reflected by the choice of linguistic strategies which are used to realize them. 4.4.2. The passive as a reporting strategy As illustrated above, style manuals for composition and academic writing promote the agent construction and praise it as straightforward, natural, and clear, as it makes the role of the author in constructing a conclusion most visible. Reporting events involve agent-oriented verbs, and from a viewpoint of the verbs’ semantics, the most natural choice to express a reporting event is the “agent construction”: Verbs like argue have two arguments, and the external argument should be realized as an agent in the position of the subject, the internal argument can be an NP or a clause. This is what is found in the following examples: (205)

a.

b.

Based on the observation that the omission of embedded subjects coincides with pronoun omission of reflexives, we tentatively suggest that in specific registers pronoun ellipsis is licensed by a specifier–head relation with a head carrying agreement features. (ELL16) We conclude that instruction in which students learn nouns in the context of distinctive lexical associates could profitably be supplemented by explicit instruction in genderending regularities. (APL2)

Most reporting verbs are transitive verbs and can be passivized. The passive as a reporting construction is illustrated in (206). Since many reporting verbs are used with clausal complements, there is often no movement of the direct object to the subject position and the subject of the passive is filled with a placeholder element, as in (206a). (206)

a.

b.

It is argued that methodological and instructional factors may be very important for the conceptual interpretation of studies. (APL9) H3 hyperacetylation is proposed as a molecular mechanism coupling enhancer activity to accessibility for V(D)J recombination. (Science38)

Examples like (206a), i. e. passives with non-thematic subjects, indicate that one of the major functions of the passive, movement of a non-agent into the

Synchronic case study

179

topic position of the sentence, is not what motivates the use of the passive in reporting events.93 This becomes even clearer when one looks at other strategies that result in non-agentive subjects. Reporting verbs also often occur with two types of inanimate subjects: In what I will call the “fact construction,” illustrated in (207), the subject position is taken by a noun phrase like these facts or these results, giving the impression that the numbers and data can speak for themselves, without any intervening or interpreting agent. The reporting event is in active voice, but no involvement of the agent is expressed. (207)

a.

b.

The structural data demonstrate how GDIs serve as negative regulators of small GTP-binding proteins and how the isoprenoid moeity is utilized in this critical regulatory interaction. (Cell13) These results contradict the notion that metal bioavailability in sediments is controlled by geochemical equilibration of metals between porewater and reactive sulfides . . . (Science9)

It will be shown that this construction is particularly popular in the natural sciences and that it is well documented historically (see section 5). A related, more recent reporting construction is the “paper construction” (Dorgeloh and Wanner, in press), a construction in which a textual category takes the position of the agent: (208)

a. b.

This article suggests that the new institutionalism contains ambiguous and contradictory notions of change. (AJES10) The paper rejects the standard view according to which every tensed verb in English agrees with its subject in person and number. (ELL17)

Semantically, the difference between the fact and the paper construction is that, unlike the fact construction, the paper construction presents an argument as something that is not self-evident but as something that has to be constructed through language. Both strategies exclude the expression of an agent argument (in a by-phrase or otherwise). In a way, they are less agent-oriented and even more “detached” than the passive. The paper construction may evoke the presence of an author as a shaper of ideas through the production of a text, but, crucially, the agent is not an implicit argu93. This is a clear difference in the use of the passive in reporting and in reported events (see sections 4.4. and 4.5. in this chapter).

180

Chapter 5 The use of the passive in academic discourse: A case study

ment, i. e. it is not an integral part of the syntax and semantics of the construction. The passive allows for a by-phrase, but passive reporting events in the abstract corpus are all short passives, i. e. the agent remains an explicit argument. According to Thompson (1978), the short passive is chosen over the long passive if the agent is not to be mentioned, e. g., if the agent is generic, or if its identity is unknown or irrelevant. These criteria are not met in reporting events: The agent is specific, its identity is important, and it can easily be inferred from the context. There must be other reasons at work for not making use of the possibility of expressing the agent in a byphrase (see section 4.4.2. in this chapter). As discussed in chapter 4, an implicit agent is always part of the representation (and interpretation) of the passive. This means that in (206) it is part of the verb’s meaning that someone is arguing and proposing – it is not information that is simply inferred from the context (the context is needed to provide the most salient reading for the reference of the agent). The situation is different with the fact and the paper construction: One knows that facts and experiments don’t argue (the APA manual criticizes these constructions as inappropriate anthropomorphisms), but crucially, an implicit agent is not part of the syntactic representation of the sentence, as the argument slot is taken up by the non-animate subject. Thus, one may not equate active morphology with more, or more explicit, agenthood. The fact construction and the paper construction are very much alike in this respect, but unlike the fact construction, the paper construction draws attention to the fact that an argument is not something that exists outside of academic discourse and just has to be uncovered. 4.4.3. The representation of agenthood in reporting events Myers (1992) has argued that in the natural sciences reporting normally involves the reporting of new facts, while in the humanities the new material to be reported can be a new interpretation of facts that may already have been known. This can be illustrated by contrasting how reporting is handled in two different journals. In the Journal of Moral Education (JME), for example, articles deal with questions like how to explain unethical behavior or what should be considered the role of poetry for human growth. Articles are usually not based on experiments and often discuss new answers to questions that have been asked before. They usually have a single author. Articles in Cell, on the other hand, discuss topics like the molecular structure of prions and genetic sequencing and almost always involve some kind

Synchronic case study

181

of experimental work, carried out by a team of researchers. The pattern of reporting strategies illustrated in Figures 2a. and 2b. is in line with the hypothesis that the visibility of the agent is more important in a modern, more constructional approach to presenting scholarly arguments and results. It has already been shown that abstracts in Cell are about 20 % shorter than in JME (121.7 vs. 146.5 words), but the differences in the use of reporting events are even more striking: For Cell the overall number of reporting events is 31 (1.3 reporting events per 100 words), for JME it is about twice as high, both in absolute and relative numbers (76 reporting events altogether, 2.6 per 100 words). Surprisingly, neither the active nor the passive construction is the favored strategy in either of the journals: The paper construction is most common in JME (36.7 %) and the least common one in Cell (6.5 %), while the complementary picture arises for the fact construction, which is the most frequent reporting strategy in Cell (45.2 %) and by far the least preferred strategy in JME (5.3 %). This pattern goes beyond these two journals, as illustrated in Figure 3.94 Neither the active nor the passive is the most frequent reporting strategy in any of the six journals analyzed here. Considering that so much of the discussion in style manuals focuses on these two strategies, it is quite surprising to see that there are actually two more popular alternatives. 4 18

26

28

Agent

Paper

11

14

4

Passive

2

Fact

Figure 2a. Reporting strategies in JME (Journal of Moral Education) 76 tokens, 2.6 reporting events per 100 words

Figure 2b. Reporting strategies in Cell

31 tokens, 1.27 reporting events per 100 words

94. I focus on 6 of the 8 journals here. Science, which is similar to Cell, but has much shorter abstracts than any of the other journals, and JWH, which shows considerable intrajournal variation, are not considered here.

182

Chapter 5 The use of the passive in academic discourse: A case study

Figure 3. Overview – distribution of reporting strategies in six selected journals (in percent of reporting events).

In abstracts from the experiment-based journals, Cell and Applied Psycholinguistics (APL), a preference for the fact construction can be observed. Not only is it the most frequent construction in both Cell and APL (used in 45.2 % and 38.9 % of all reporting events), the absolute number of tokens of the fact construction is higher in Cell and APL than in any of the other journals (14 tokens compared to only 5 in English Language and Linguistics). If one considers the fact that ELL has three times as many reporting events as Cell (96 vs. 31), this result becomes even more striking. The preferred construction for realizing reporting events in abstracts from the nonexperiment based journals is the paper construction. It is used in more than 30 % of all reporting events in ELL, JME, AJES, and CJE, while its proportion in Cell and APL is much lower (19.4 % in APL and 6.5 % in Cell). The use of the passive varies greatly, from 10.2 % in AJES to 36.4 % in ELL, both journals with a high number of reporting events. If one puts all active reporting strategies together, one would get the (incorrect) impression that Cell scores high on choosing an agentive style of reporting (with only about 1 in 8 reporting events in the passive), while ELL scores low (with more than 1 in 3 reporting events in the passive). However, from a structural viewpoint, the passive is actually more agentive than the fact construction and

Synchronic case study

183

the paper construction. The agent of the event is an implicit argument, but as such it is an integral part of the meaning of the sentence. In the fact and the paper construction, the agent slot is filled with a non-agent argument, which gives less visibility to the actual agent than the passive does. While experiment-based abstracts are not necessarily shorter than abstracts in the humanities, they tend to have fewer reporting events (Figure 1) and also less linguistic variation within them. Although the active and the passive construction seem to be the most natural choices for expressing reporting events from the viewpoint of lexical semantics, neither of these two is the most frequent construction in the abstracts examined here. Journals with a high proportion of reporting events show a preference for the paper construction, while journals with a lower frequency of reporting events prefer the fact construction. In light of the attention that the active and the passive receive in style manuals (the active as a desirable construction, the passive as a construction that is considered overused), these results are quite surprising and highlight the flexibility in filling the subject position with non-agent arguments that English grammar is known for (Levin 1993). 4.4.4. Reporting events and politeness considerations In scientific discourse, it is extremely important to tie an argument to the person who is making it. The author of an article takes a very active role in shaping a new argument, and as part of this process he or she has to situate the new argument in the context of previous research. In establishing the niche for the topic of an article one often find direct references to the authors of previously published research on the subject. However, while there is good reason for linking authors to their work, there is also reason for minimizing references to authors of previous research and to oneself. Placing a new argument in context can be construed as an infringement on a settled order. By their very nature, reporting events put emphasis on the role of the researcher as someone who interprets the data the article is based on, i. e. reporting events add to the visibility of the author. It was shown that the agent is often not explicitly expressed, even though this would be in line with the principle of maximal clarity. Thompson (1987: 501) points out that “users of English are content to code the agent as subject unless broadly thematic or more local cross-causal considerations require an alternative coding.” The use of the fact and the paper constructions cannot be explained by a need to make the affected entity the topic of the sentence (one of the main reasons to use the passive in reported events). Since the agent in a reporting event is neither irrelevant nor unknown and since the emphasis in the

184

Chapter 5 The use of the passive in academic discourse: A case study

event is not entirely on the theme argument (ruled out by the sheer number of reporting events), we are left with another, discourse-oriented motivation for using an agentless passive: “The speaker wishes to leave the identity of the agent vague, for such reasons as politeness or expediency, or, sometimes, to reduce the assertion of responsibility for the agent” (van Oosten 1986: 8–9). From a pragmatic viewpoint, and specifically within politeness theory (Brown and Levinson 1978; Brown and Levinson 1987), scientific articles constitute face threatening acts (FTAs): They consist of “presentations of claims . . . that the authors do not attribute to anyone else . . . . In most cases, these claims imply a criticism, a reformulation or an elaboration of former theories, beliefs etc. put forth by other researchers” (Garc´esConejos & Sa´ nchez-Macarro 1998: 176). Not surprisingly, one of the mitigating linguistic devices produced by these conditions is the use of the passive voice, which “is perhaps the means par excellence in English to avoid reference to persons involved in FTAs” (Brown and Levinson 1978: 199). Other strategies found are the use of impersonal constructions, indefinite or generalized subjects, or nominalizations (Garc´es-Conejos & Sa´ nchezMacarro 1998: 184–187). In the following examples one can see varying degrees of explicitness regarding the identity of the person who made an established argument and the person who challenges this argument (i. e. the author of the article/abstract). In (209a), the force of the FTA is mitigated by the use of summary references (Moral reasoning theorists . . . ) and the use of the passive (an argument is presented . . . ). The passive is clearly used for politeness reasons here, it serves to “reduce the assertion of responsibility for the agent” (Oosten 1986: 8–9). Compare this to the directness of the contradiction in (210). The author whose analysis is challenged is named explicitly, the analysis itself is criticized in the strongest possible way (“seriously flawed”) and the authors of the article refer to themselves in the first person. This kind of explicit challenging of an existing claim is rare, particularly in abstracts, which normally situate an article in the context of previous research in more general terms. (209)

Moral reasoning theorists working in the constructivist tradition have tended to explain unethical behaviour by assuming that a breakdown occurs in the link between a person’s moral judgement within a particular situation and his ultimate behaviour in that situation. . . . An argument is presented that unethical behaviour, especially that involving relatively minor moral breaches, is often not the result of a moral judgement-behaviour hiatus but rather a corruption of the construal process. (JME1)

Synchronic case study

(210)

185

In the 1998 article . . . in this journal, Emanuel D. Thorne advocates increased exhortation and advises against the adoption of markets in cadaveric organ procurement. In support of this view, Thorne offers analysis [sic] in which he purports to show that . . . We argue that Thorne’s analysis is seriously flawed and that his conclusions are suspect, if not entirely illogical. (AJES14)

Politeness issues in scientific communication thus suggest that the events in scientific writing are presented as impersonal, which means that strategies are preferred that make the agent less recoverable (Marin-Arrese, MartinezCaro and Becerril 2001: 370). The passive makes the agent less recoverable by turning it into an implicit argument. It could be spelled out in a byphrase, but while by-phrases occur with some regularity in the corpus, reporting events in the passive are always short passives. It is important to note that this is not a restriction on the verbs that are involved. Passivized reporting verbs that are used to report somebody else’s research are part of the reported material and freely occur with by-phrases. Example (211) illustrates such a case. The critique that is advanced is an argument made by somebody whose identity is important and cannot be inferred from the context. The critique mentioned in this sentence is not an argument that is made in this particular article, it is a description of something that happened in the past and that has to be integrated into the argument presented in the abstract, just like other facts and observations. (211)

Goswami and Bryant (1990) proposed a theory of reading development based on three causal connections. . . . To explain this connection, they suggested that young readers of English used analogies based on rimes [sic] as one means of deciphering the alphabetic code. This proposal has recently become the subject of some debate. The most serious critique has been advanced by Seymour and his colleagues (Duncan, Seymour, & Hill, 1997; Seymour & Duncan, 1997; Seymour & Evans, 1994). These authors reported a series of studies . . . . (APL9)

Like the passive, the paper construction as well as the fact construction result from a need for negative politeness: Both strategies assure the readers “that the writers do not intend to infringe on their wants” (Myers 1989: 12). Negative politeness is achieved by combining an impersonalized subject (this finding, this paper) with the hedging provided by a reporting matrix verb (someone or something suggests that a is b rather than just a is b, as in (203) above). The result is that scientific claims receive “some imper-

186

Chapter 5 The use of the passive in academic discourse: A case study

sonal agency” (Myers 1989: 17). However, unlike in the paper and the fact construction, in the passive the shaper of the ideas that are presented is structurally integrated into the meaning of the sentence. 4.5. The use of the passive in reported events In contrast to reporting events, reported events cannot be reduced to a specific semantic class of verbs. In the abstract corpus, reported material includes the evaluation of previous claims (as in (212a)), establishing a niche for the topic of the paper (as in (212b)), and a description of the data that will be discussed in the article (see (212c) and (d)). Reported events may or may not be agentive, they may or may not deal with mental events, and they may or may not involve the authors of the paper. (212)

a.

b. c. e.

Although it is commonly assumed within schools that drama has a place within moral education, there is very little theory or analysis to support the assumption. (JME 4) This raises the possibility that bone mass, body weight, and gonadal function are regulated by common pathways. (Cell 1) Empirical estimates of the coefficient are presented for the UK from 1972 to 1995. (CJE 12) Direct experimental evidence for the presence of dynamical heterogeneities in a dense liquid was obtained. (Science 12)

Due to these variables, there is no point in directly comparing passive frequencies in reported events across academic disciplines. What I will focus on here is the motivation for using the passive in reported events, which is different from that for using the passive in reporting events. Table 3 shows that reporting events account for 15.4 % to 40.7 % of all passives. Considering the small class of verbs that falls into this category, this rate is actually quite high, but it also means that most passives do not occur in reporting events. The choice of the passive in reported events is usually motivated by textual needs. According to Biber et al. (1999: 935) the passive primarily “serves the discourse functions of cohesion and contextual fit through ordering of information, omission of information (especially short passive), weight management (especially long management).” The passive is often chosen if a non-agent is part of the discourse topic, in the sense of a recurrent and persistent discourse entity (Givon ´ 1993: 202). For example, the passive in (213) is the first sentence in an abstract that deals with capital stock estimates (the

Synchronic case study

187

Table 3. Number of passives per 20 abstracts AJES

APL

Cell

CJE

ELL

JME

passives per 20 abstracts/per 100 words

39/1.35

43/1.26

32/1.32

46/1.89

86/2.69

55/1.88

passives in reporting events

6 (15.4 %)

11 (25.6 %)

4 (12.5 %)

18 (39.1 %)

35 (40.7 %)

18 (32.7 %)

title of the paper is “Standardised capital stock estimates in Latin America: a 1950–94 update”). The non-agent is clearly more topical than the agent in this case. In a similar vein, in (214) maturation has been introduced in the previous sentence (. . . mature into larger, stabler particles) and is thus more topical than the instrumental subject, proteolysis. This time, a long passive is chosen as it is important to point out what exactly is causing the maturation process. By contrast, in (213) there is a generic agent (. . . are used extensively), which may remain implicit, and in (215) the exact identity of the agent does not matter or is unknown. (213)

Capital stock estimates are used extensively in many areas of economic research, in spite of both theoretical and practical difficulties with respect to their use, estimation and meaning. (CJE 3)

(214)

Typical of DNA bacteriophages and herpesviruses, HK97 assembles in two stages: polymerization and maturation. First, capsid protein polymerizes into closed shells; then, these precursors mature into larger, stabler particles. Maturation is initiated by proteolysis, producing a metastable particle primed for expansion – the major structural transition . . . (Cell 6)

(215)

In particular, it is necessary to consider whether a given phonological awareness task requires the recognition of shared phonological segments (“epilinguistic” processing) or the identification and production of shared phonological segments (metalinguistic processing). It is also important to take into account the nature of the literacy instruction being implemented in participating schools. (ELL 9)

The following example illustrates the cohesive function of the passive. Choosing the passive in the second clause allows for a shared subject construction. Without the passive, the noun phrase Antarctic ice could not be

188

Chapter 5 The use of the passive in academic discourse: A case study

shared by the predicate in the first and the second clause and it would have to be realized twice. (216)

Antarctic ice occurred rapidly in the earliest Oligocene (34 million years ago) and was not accompanied by a decrease in deep-sea temperatures. (Sci 6)

Finally, (217) illustrates a use of the passive in a reported event in which the author of the paper is the agent of the verb (analyse). Again, the non-agent is topical and connects with the previous sentence. The agent of analyse remains implicit – not because it is not known or it is generic, but rather because it refers to the author. The abstract in (218) illustrates that in a description of an experiment the passive may be maintained through an extended passage. It has been noted that this is typical of the short passive, but not of the long passive (Biber et al. 1999), as the long passive would not be used if the agent remains the same throughout. (218) also illustrates a point made in chapter 3: The implicit argument in the passive (was studied) can control a PRO subject in a purpose clause (PRO to determine), which indicates that the implicit agent is actually part of the syntax of the passive. (217)

The connection between punishment, guilt and shame is analysed and a number of ideas on punishment are set against the back drop of moral development. (JME 9)

(218)

Private speech (PS) – or speech for the self – was studied in preschool-age children to determine how widely and with what characteristics it occurs when examined in the familiar home setting. Activities were selected that required several steps and that were intended to engage working memory or longer term recall. Both quantitative (numbers of words and utterances produced) and qualitative (utterance type) analyses were conducted on the children’s PS. (APL 8)

We can see that in reported events the backgrounding of the agent may be one of the motivations for using the passive, but unlike in reporting events it is usually accompanied by the function of allowing for the realization of a non-agent in the canonical topic position. This is different from reporting events, which, if passivized, often result in impersonal passive constructions. Reporting verbs usually have a clausal complement, and clauses do not underlie the Case Filter (compare chapter 2), which means that movement will not take place and the subject position will be filled with non-topical it (It will be argued that . . . ).

Support from a diachronic study

189

5. Support from a diachronic study It has been shown that the passive, though popular, is not the construction used most frequently in abstracts in which reporting events figure prominently. If this has to do with the orientation of scientific discourse towards making the constructional nature of the argument visible, the hypothesis is that in earlier stages of scientific discourse, characterized more appropriately by a “discovery approach” (Rudwick 1985), the passive would have occurred more frequently. In order to test this hypothesis, a second study was carried out which looked at the scientific register historically. The study is based on data from the ARCHER corpus (A Representative Corpus of Historical English Registers), which has text and speech samples from 7 historical periods, beginning with Early Modern English (Biber, Finegan and Atkinson 1994). The category “Science” in the ARCHER corpus consists of 7 sub-corpora from 1650 to 1990 of about 20,000 words each. Text samples in this corpus come from longer articles and monographs, which means that the proportion of reporting events is expected to be lower than in the abstract corpus. Due to the characteristics of the abstract as a genre, the number of reporting verbs should be higher in abstracts than in more narrative texts.95 Using a concordance program, ARCHER was scanned for the use of eight agent-oriented verbs commonly used for reporting in English: argue, demonstrate, indicate, show, suggest, examine, explore, and find. All of these verbs allow the agent and the passive construction, and most of them also allow the paper and the fact construction. Figure 4 summarizes the results. One can observe a steady increase in the use of these reporting verbs, which is in line with the assumption that the role of constructing an argument has grown over time. A second result is that the use of the passive among these reporting events also increased steadily until the most recent time segment of the corpus. It remains to be seen if the lower numbers of passives in reporting events are really a trend that specifically affects the passive, as suggested by Seoane’s studies of the overall use of the passive in scientific writing (Seoane 2006). The ongoing popularity of the fact construction makes it difficult to argue that the decrease of the passive is a consequence of a general trend towards expressing more “involvement” in scientific writing. 95. Text samples in the category “Science” are all from the Philosophical Transactions of the Royal Society (founded in 1600), which means that they are all more or less from the natural sciences. At the time of the foundation of the Royal Society, there was no comparable institution or journal for the humanities.

190

Chapter 5 The use of the passive in academic discourse: A case study

90 78

80 70 60

54 40

40

40

40

35

30

22

20

37

26

25 19

16

20 10

51 45

47

50

13

11 0 2

0 2

16501699

17001749

6 0

0

0

1

4

0

Agent

17501799 Paper

18001849 Fact

18501899

19001949

19501990

Passive

Figure 4. Realization of selected reporting verbs (argue, demonstrage, examine, explore, find, indicate, show, suggest) in ARCHER (tokens per 20000 words)

Thirdly, there is also a significant development among the three strategies that have been found to be competitors of the passive in reporting events: The agent construction starts out as the most popular reporting strategy in the earliest time segments, which is not a surprise, since professional scientific writing has its roots in narratives of personal encounters with nature (Atkinson 1992). In the ARCHER texts, the use of the agent construction decreases over time, but it never appears to become altogether unusual. By contrast, the fact construction rises steadily, from an almost negligible usage until the end of eighteenth century to a more frequent us-

Summary

191

age in the nineteenth century (13–16 tokens) and to even greater popularity in the twentieth century (47–51 tokens). As in the abstract corpus, the fact construction can be described as the winner of the reporting event competition. Not only is it the most popular strategy in the most recent time segment, it grows even although it is absolutely not in line with the rhetorical shift towards more involved, agent-oriented writing. It remains to be seen if the sharp decrease of the use of the passive that can be observed in the latest time segment really is a trend, but if it is, it is even more remarkable that the fact construction, which is allows the agent to be even more detached than the passive, is actually growing in popularity at the same time that the use of the passive is decreasing (presumably because unlike the passive it is not morphologically marked and thus not as often targeted in style manuals). The paper construction is, as expected, a fairly recent development. Its rise follows the establishment of the fact construction as a well-accepted form of constructing an argument in the sciences and runs parallel to the emergence of the scientific abstract as a genre. Judging from the popularity of the paper construction in the abstract corpus, it is safe to predict that the rise of this construction will continue.

6. Summary In this chapter I examined the use of the passive in one of the most condensed forms of scientific writing – abstracts that precede published research articles from different academic fields. The goal was to analyze the reasons for choosing the passive, to find out which constructions compete with the passive, and to explore the role of representing agenthood in scientific discourse. The use of the passive in scientific writing has a long tradition, but its evaluation has undergone drastic changes. What used to be considered as appropriately distanced is now often criticized as undynamic, pseudo-objective, and evasive. There is a general recognition that scientific genres are not purely informative and that facts cannot be reported without interpretation. For the research article, in particular, the focus has shifted from reporting a case or an experiment to building a persuasive argument and to offering sufficient explanations for it (Bazerman 1988; Swales 1990; Gross, Harmon and Reidy 2002: 231). Writing an abstract adds another step of interpretation and condensation to the long process of writing and revising that goes into the publication of research results.

192

Chapter 5 The use of the passive in academic discourse: A case study

In a study based on 160 abstracts from 8 scholarly journals from different academic disciplines, I looked at the use of the passive in this process of constructing arguments. I made a distinction between reporting events (events in which the author explicitly constructs an argument, using verbs like argue, suggest, and demonstrate) and reported events (the material on which the argument is based). The reason for making this distinction was that the constructions with which the passive competes in describing reporting events are very different from the construction with which the passive competes in describing reported events. Also, reporting verbs constitute a coherent class that is similar across the disciplines, i. e. one can compare their realization in journals that belong to different research traditions. A comparison of the use of verbal reporting strategies showed that, with regard to the abstract corpus, the number of reporting events was lower in abstracts from the experimental sciences than in abstracts from the humanities.96 Experiment-based abstracts are not necessarily shorter than abstracts from the humanities, but they tend to have fewer reporting events and also less linguistic variation within them. On the basis of these data I argued that reporting events play a less prominent role in journals that focus on experiment-based facts and observations than they play in the humanities and social sciences, where the production of newsworthy scientific knowledge often involves the evaluation or reevaluation of facts that are known. A second result from the analysis of the abstract corpus was that the passive is more popular as a reporting strategy in the experiment-based sciences than in the humanities. In abstracts from the humanities, the agent construction turned out to be a more popular construction used to formulate an argument. However, the most frequently used reporting strategies overall were the fact construction (These data show . . . ) in journals with a low proportion of reporting events and the paper construction (This paper argues . . . ) in journals with a high proportion of reporting events. Both constructions replace the agent with a non-animate entity the argument is based on, i. e. there is no true agent in the sentence, neither explicit (as in the active) nor implicit (as in the passive). In light of the attention that the agentive active and the passive receive in style manuals (the agent construction as a desirable construction, the passive as a construction that should be avoided), these results are quite surprising and highlight the flexibility in 96. The labels “humanities” and “experimental sciences” are only used with regard to the abstracts analyzed here. It is not claimed that the results from this small and homogeneous corpus can be projected to a larger scale in a straightforward manner.

Summary

193

filling the subject position with non-agent arguments that English grammar is known for. As pointed out by Bazerman (1988: 79), the modern scientific community is essentially an argumentative community: The individual author cannot rely on the power of “the self-evident truth of events,” instead, he or she has to build a convincing argument: “[T]he claim or conclusion – Popper’s third world – becomes the central item to be constructed within the article, to be supported by empirical evidence from the first world of proper method and reasoning from the second world” (Bazerman 1988: 78). In a case study of two biologists who revised their article a number of times in hopes of having it published in Science or Nature, Myers (1985) discusses the process of fine-tuning in putting forward an argument that is relevant and new. If the claim that is made is risky and far-reaching, the article will challenge a number of established claims, which requires a lot of argumentative work. If, on the other hand, the claim that is to be made only confirms what has been said before, it might not be interesting enough to be published. (The two biologists did not succeed in their efforts and had to settle for publication in another journal). Claims put forward by the author of an article often imply criticism of theories put forward by other researchers (Garces-Conejos and Sanchez-Maccaro 1998: 176) and can be regarded as face-threatening acts (Myers 1989). For reporting events, a motivation for using the passive lies in the face-threatening force of positing a new claim. The passive can be seen as a mitigating strategy, allowing for the expression of an agent-oriented event (like arguing or suggesting) without the explicit expression of the agent. This is matched by the fact that all reporting events in the corpus are short passives. The role of argumentation in scientific discourse has changed considerably over time: In his diachronic study of how knowledge is constructed in academic writing, Bazerman (1988) describes the changes that the reporting of experimental evidence has undergone in scientific texts. Experiments moved “from any made or done thing, to an intentional investigation, to a test of a theory, to finally a proof of, or evidence for, a claim” (Bazerman 1988: 65–66). In conjunction with this development, the role of the individual author in presenting his or her research has changed from that of someone who is “reporting a self-evident truth of events” to that of someone who has to present a “persuasive story” in which a claim is made that transcends the particulars of the experiment or investigation at hand (Bazerman 1988: 78). The “focus on facts over argument” (Gross et al., 2002: 19) that characterized early scientific discourse relied on a confidence that data would speak for themselves. By contrast, in modern scientific discourse a

194

Chapter 5 The use of the passive in academic discourse: A case study

finding only has the significance that the researchers “bestow upon it by the power of their argument” (Gross et al., 2002: 26–27). I hypothesized that this would lead to an increase in verbs of communication and persuasion (used to express reporting events) in modern scientific texts. Therefore, in a second study, based on data from the historical ARCHER corpus (mainly data from the natural sciences), I looked at the use of selected agent-oriented reporting verbs diachronically. One of the results was that there was a steady increase in the use of the passive for these verbs until the most recent time segment in the corpus. Accepting implicit arguments in reporting has a long tradition in scientific writing. However, it could be seen that the fact construction also steadily increased over time and that the paper construction, a construction more in line with modern, more text-oriented approaches to presenting and interpreting facts, is indeed a fairly recent phenomenon. Of the two, the fact construction minimizes any expression of agenthood in expressing an argument, and it does not come as a surprise that this is the construction most often employed in the natural sciences. The paper construction may conceptually evoke the presence of an agent (as the one who writes the paper), but, unlike the passive, it does not have a true implicit agent built into the interpretation of the sentence. The analysis has shown that it is simply not appropriate to equate the use of active voice with more visible agents. The success of the paper construction seems to be related to the stigmatization of the passive and other morphologically marked constructions that downplay the role of the individual agent. According to Baron (1989: 19) “by the 1940s the passive, with its deletable agent . . . became associated not simply with the mildly distasteful traits of wordiness and confusion, but with the even more negative practice of conscious deception by deliberately hiding the doer of the action.” The same could be said about the fact construction, except that it has not received quite the same attention. This may be due to the fact that, unlike the passive, it is not morphologically marked and is therefore more difficult to detect. More than other constructions, the passive is singled out as the embodiment of a style that is to be avoided. Excerpts from style manuals presented in this chapter have shown a bias against the passive so strong that it is carried over to sentences that only look like passives (in that they combine a form of be with what looks like a participle but may really be an adjective) and that sentence alternatives are suggested that don’t convey the meaning of the original sentence but are still considered an improvement over the passive, simply because they are not passives. There are, of course, limitations to the explanatory power of the results from the case studies compared in this chapter. First, the abstract corpus

Summary

195

is very small and only covers one specific text type of scientific discourse. However, a comparison with data from larger corpora showed that the distribution of passives in the abstract corpus is very much like that in those reference corpora. I argued that some of the results from studies carried out on the basis of larger corpora may actually not be all that representative because more often than not, passives that cannot easily be identified with corpus linguistic tools, such as passives in participle clauses, are not included in the analysis. A second problem lies in the comparability of the texts that the studies presented in this chapter are based on. The abstract corpus consists of 160 short texts of the same kind, but from different disciplines, while the ARCHER corpus is based on excerpts from a variety of texts, but all from the natural sciences. Among the factors that remained unexamined here is the linguistic background of the authors. It has been estimated (Wood 2001) that in journals like Science the number of articles written by non-native speakers is as high as 45 %. These authors may come from countries in which the use of the passive or, more generally, the attempt to minimize the linguistic expression of the involvement of the researcher, is not criticized as much as in the teaching of academic writing in the United States. Another factor that could not be looked at is the role of editors and style sheets, as well as some writers’ very noticeable efforts at linguistic variation. For example, in English Language and Linguistics, 8 of the 20 abstracts examined begin with the noun phrase This paper. While this beginning is rather formulaic, at the same time authors seem to strive consciously for stylistic variation with regard to the expression of reporting events. For example, a reporting event in the passive may be followed by one employing the paper strategy and then by an agentive active. By contrast, in journals like Science it is not unusual at all to find that all reporting events are of the same type. Repetition is even more obvious when it comes to the expression of reported events, in particular the description of an experiment, as in (219). It is difficult to say if this repetition is deliberate or if it is the result of a lack of stylistic efforts or linguistic skills of the author(s). (219)

Jurkat cells . . . were preincubated in phosphate-free RPMI . . . ; orthophosphate was added (1.0 mCi/ml), and the cells were incubated an additional 50 to 60 min . . . . After washing twice in ice-cold 0.9 % NaCl, the cells were resuspended in RPMI 1640, stimulated as described in the figure legends, pelleted, and lysed. (FROWN; J08)

The evaluation of agentless structures and in particular of the passive has changed vastly over time. What used to be a marker of professional distance

196

Chapter 5 The use of the passive in academic discourse: A case study

is now often considered a device to produce fake objectivity and authority, in particular when it comes to expressing an argument. Choosing the passive is associated deliberately hiding the instigator of an event. This is, of course, incorrect. Unlike the paper and the fact construction, the passive includes the agent as an implicit argument, but in the evaluation of the passive the focus is usually on the fact that the exact identity of the agent is not expressed. Although Baron assures his readers in a chapter entitled “The Passive Voice Can Be Your Friend” that the use of the passive is acceptable, he notes that “Today’s bias, however, even in business and in technical writing guides, is towards the active voice” and that it has become “increasingly popular in scientific writing . . . [to] advise technical students to pepper their work with active sentences” because “the active appears as a sign not of hubris but humility, a call for other researchers to validate the writer’s experiment or theory” (Baron 1989: 21). This chapter has shown that statements of this kind are oversimplified, that the proficient academic writer’s repertoire of expressing agenthood has more than two options, and that new structures evolve (or are evolved?) as they are needed.

Chapter 6 Conclusion The English passive has been studied from many different linguistic angles. It is a construction that may be costly in terms of sentence comprehension, but these costs are balanced out by a better fit between informational status, discourse needs, and syntactic structure. In this book I presented my view of the passive as a construction whose form is shaped, to a large extent, by autonomous syntactic principles, but whose place in grammar cannot be understood without looking beyond the domain of the sentence. I suggested that one needs to compare the passive with similar constructions (not just the corresponding active) to tease out what exactly is unique to the passive. This comparison should include considerations of the communicative purpose of the passive as well as an exploration of the passive in a spectrum of transitivity alternations in a language that normally maps agents to subject positions. To this purpose, I combined a syntactic analysis of the passive in the generative framework with corpus-based studies that investigated specific aspects of the use of the passive, in particular the notion of an “implicit agent.” Throughout the book I referred to the passive as a “construction” in order to highlight the idea that the syntax and semantics of the passive are compositional in nature. Following this hypothesis, I looked separately at each of the components of the passive construction in English and at their contribution to the construction as a whole. The focus of my interest was on a non-overt component of the construction, the implicit argument (most often an implicit agent). In chapter 2 I argued that the implicit agent reading lies at the core of the passive construction and that it is the implicit argument that sets the passive apart from other verbal argument alternations (such as the ergative construction) and also from pragmatically motivated word order changes that have no effect on the realization of the verb’s arguments (such as topicalization). Regarding the function of the passive, I did not explore in detail the effects of passivization on the distribution of given and new information of the sentence. Rather, I focused on describing the passive as a way to make a sentence less transitive (in the sense of Hopper and Thompson 1980), and I related the passive to other constructions that map the arguments of the verb onto syntax in a non-prototypical way. The syntactic and semantic asymmetry among the arguments of the passivized verbs is also an important factor in determining which verbs, or rather VPs,

198

Chapter 6 Conclusion

can be passivized and which cannot. Syntactically, a verb needs to have an external argument (and thus be able to assign structural Case) in order to form a passive. In addition, passivization works best if the event is not just transitive in the structural sense of the term, but if it is also semantically transitive (with a willful agent acting on an entity that is affected by the action). Verbs that express stative symmetrical relations (such as weigh or resemble) do not form passives very well. Mapping a thematic object to the subject position is unexpected and will slow down the comprehension of the clause. Children who have not acquired the passive yet will interpret the NP in the subject position as the agent in the clause. Next to the by-phrase, a subject that is a prototypical patient is an important clue for them to recognize the passive construction. Children’s first productive passives are those in which the passivized event has a prototypical agent acting willfully on a prototypical patient, i. e. an event that is semantically transitive. While there is a common morphosyntactic denominator for the passive, the surface structure of the construction can take different shapes. I settled on a definition of the passive focused on the passive participle and the implicit argument reading. The analysis in chapter 3 showed that all other components typically associated with the passive (the theme subject, the auxiliary be, and the by-phrase) depend on the syntactic and semantic characteristics of the verb and on the way the passive participle is embedded in its syntactic environment; they are not inherent to the construction. The discussion in chapter 3 also has ramifications for corpus-based studies of the passive. While it is a well-known fact that most passives (about 80 %) occur without a by-phrase, there is generally little awareness that a form of be is not part of every passive either. Bare passives, i. e. passives without auxiliaries, are often excluded from corpus-based analyses of the passive – not for any principled reason, but because they cannot easily be detected with automatic scanning tools, since in English the participle in the passive looks exactly the same as the participle used to form the present perfect. Analyses that do not consider these passives will miss a sizable number of passives in the corpus (as many as one in five in the category “Science” in the FROWN corpus). What is more, passives without be do not just occur randomly, they are clustered in specific syntactic environments. In the corpora considered here, most of the be-less passives occurred in participial relative clauses (such as products [provided by workstation vendors]), i. e. in an environment in which the use of the passive has the effect of creating seamless cohesion between a noun and a modifying clause. This is just one of the many functions the passive can have. Others include introducing a non-agent topic and reducing the processing load of parsing a sentence

Chapter 6 Conclusion

199

by positioning a syntactically complex phrase at the end of the clause (in a by-phrase). One purpose of this study was to show that a syntactic analysis in the generative framework can provide insights that help us understand aspects of the use of the passive construction. The implicit argument in the passive is a good candidate for this kind of approach, since generative grammar recognizes non-overt categories as structural entities and provides the tools to differentiate between various types of non-overt arguments. In chapter 4, a review of data from syntactic analysis, language acquisition, and psycholinguistics showed that, in terms of syntactic visibility, the implicit agent lies somewhere between a full-fledged, non-overt argument, such as the PRO subject in infinitives, and a purely conceptual argument, such as the instrument incorporated in the meaning of verbs like cut or brush. Unlike PRO, the implicit argument does not project a full syntactic noun phrases and cannot, for example, bind an anaphor or control a PRO subject in a complement clause. Unlike a conceptual or an inferred argument, the implicit argument can license agent-oriented adverbs and can control the PRO subject in a purpose clause (Two aspects of grammar . . . were assessed [PRO to determine whether . . . ]). I sided with those approaches that locate the implicit argument at the level of argument structure, which is in line with results from psycholinguistic experiments that show that the implicit argument reading becomes available as soon as the passive participle is processed. It is not available in other constructions with non-agentive subjects, such as the ergative construction, even though, conceptually, a sentence like The antique vase finally sold evokes the presence of a seller, i. e. an agent. A specific candidate for an analysis of the passive that does not treat the construction independently of its components is the get-passive. Oftentimes, the get-passive is treated as a stylistic variant of the be-passive, and syntactic and semantic differences between the two are thought to arise from the constructions themselves rather than from their structural ingredients. Properties ascribed to the get-passive include a responsibility reading for the nonagent subject, which means that the subject in John got fired is perceived as more responsible for the event than in the corresponding be-passive, John was fired. Another property associated with the get-passive is a preference for events that are non-beneficial for the subject, often expressed by verbs of bodily impact (get whacked, get beaten up). In chapter 3 I proposed that these semantic properties of the get-passive should be related to those of the verb get, an approach that led to the inclusion of passives after causative get (as in get oneself fired). A careful review of data from the FROWN corpus and from questionnaires filled out for the Dictionary of American Regional

200

Chapter 6 Conclusion

English showed that the responsibility reading, unlike the implicit agent in the passive itself, is purely contextual; it is not built into the construction. Regarding the perceived adversarial nature of the get-passive, it could be shown, that this feature, too, is not inherent to the construction. The preference for verbs of bodily impact, if it exists, can be related to the occurrence of the get-passive in colloquial spoken English, itself a feature that can be related to the verb get. Overall, the get-passive provides further evidence for the hypothesis that the passive can best be understood by looking closely at verb semantics and the use of the construction in a specific syntactic context. In chapter 5, I turned towards the use of the passive in a register that may be considered a natural habitat of the construction, academic writing. My goal was to explore the role of representing agenthood in scientific discourse, to analyze the reasons for choosing the passive, to find out which constructions compete with the passive. I first discussed how changes in the rhetoric of scientific discourse – from a paradigm of discovery of facts to one of shaping ideas through the use of language (Bazerman 1988) – have affected the perception of the passive. Recent corpus studies have shown a general decline of the passive in academic genres, in particular in American English (Seoane 2006). Once held in high regard as a construction that expressed an appropriately impersonal perspective, the passive is now criticized as undynamic, evasive, and pseudo-objective. Particularly in American English, the bias against the passive is so strong that style manuals and editorial style sheets sometimes go overboard in their efforts to suggest alternatives to the passive. One example I cited was that in a book on diabetes, a physician wrote about comatose patients who were given some sort of medicine. Since passives were not allowed in the book, the sentence was changed to active voice, with the result that in the “improved” version the patients, still comatose, were taking their medicine – a major semantic difference and not at all what the author had intended. I pursued the question of whether the strong criticism of the passive has had an influence on how the passive is used in the way scholars put forward an argument or a conclusion. The reason for focusing on that segment of academic writing was that it is the part of a text in which one can expect high visibility of the writer/researcher as an agent. In a study based on 160 abstracts from 8 scholarly journals from different academic disciplines I showed that, when it comes to the wording of a conclusion or scholarly argument, the main competitor of the passive was not, as one might think, the agentive active. Rather, the most popular competitors were constructions that combined active voice with non-agentive subjects. These constructions included what I called the fact construction (These data show . . . )

Chapter 6 Conclusion

201

and the paper construction (as in This paper argues . . . ). The former turned out to be the most frequent strategy of argumentation in abstracts from the experiment-based sciences, the latter was characteristic of the more textoriented humanities. Both constructions avoid the expression of the true agent (the author of the paper). The paper construction may conceptually evoke the presence of an agent (as the one who writes the paper), but, unlike the passive, it does not have a true implicit agent built into the interpretation of the sentence. The fact construction has no implicit agent either, and it presents an argument as something that arises naturally from the data, without the intervention of a human agent. Much more than the passive, it is a construction that embodies the traditional concept of the scientist as a discoverer rather than a shaper of ideas. A second study, based on selected data (eight verbs of argumentation) from the historical ARCHER corpus, confirmed the long history of the passive in scientific writing. By contrast, the paper construction is a fairly new development. It could only be found in the most recent time segment of the corpus (1950–1990) – the segment in which the frequency of the passive decreased (at least for the verbs under observation). This complementary pattern gives rise to the interpretation that the increase of the paper construction is, to some extent, motivated by an effort to avoid the passive. Like the passive, the fact construction has a long history in scientific writing, and it is quite surprising that it has escaped the negative attention that the passive has received and continues to receive. I speculated that the passive is criticized more often because it is noted more easily, due to its morpho-syntactic marking. Overall, the analysis showed that it is simply not appropriate to equate the use of active voice with more visible agents and that corpus analyses which only look at the distinction between active and passive voice give a rather pixely picture of the representation of agenthood and involvement in scientific writing. If mistakes were made, it is at least clear that they were made by someone, even if the identity of the agent is not revealed; it is when mistakes happen that one presents the world as one in which people do not act and therefore do not have to take responsibility for events that unfold (or are unfolded) around them. Recent studies from corpus linguistics have found a general trend towards the use of less formal language in writing (Mair and Leech 2006), and the development of communication through electronic media has broken up the pairing of written language and formal style for good. It remains to be seen how the passive, a construction for the longest time associated with formal writing, will be affected by this development. After all, the getpassive is testament to the fact that the passive and the use of implicit argu-

202

Chapter 6 Conclusion

ments are not inherently tied to formal registers. It also remains to be seen which effects the strong criticism of the use of the passive in academic writing, particularly in the teaching of writing in the United States, will have on the use of the passive in that register. In many academic fields, articles are published by teams rather than individual authors, and the writers may be of very different linguistic backgrounds. The study reported here has shown that users may avoid a construction that comes with a built-in red flag, only to create constructions that are, in principle, just as flaggable, but that do not get a lot of attention in style manuals. In this study, the English passive has been framed as a specific version of a construction that exists in many languages and that is shaped by principles of Universal Grammar. However, it is not possible (nor desirable) to separate the general aspects of the construction from those that are particularly “English.” For example, if English had not developed into a preposition stranding language, it would not allow prepositional passives. If the properties of expletive it were different, English, like Dutch or German, would probably allow passives based on unergative verbs. And if the verb get in English had remained a verb that allowed only nominal or prepositional complements, there would be no get-passive, and whatever alternative to the be-passive might have developed, it would not be influenced by the semantic and syntactic facets of get. In addition, there are functional aspects of the English passive that could not be duplicated in another language. One of them concerns the status of English as an international language. For example, English is the dominant language of science. Most high-impact journals, especially in the natural sciences, are published in English, and many international scientists have been trained in the United States and have thus absorbed, or at least been exposed to, the rhetoric of scientific discourse in the United States, which has little regard for the passive. Although it is not clear which impact style sheets in journals like Science or Nature really have, every effort should be made to build bridges between those who study the structure of language and those who teach the use of language and shape its public perception. As long as the passive construction is described as agentless and wordy, confusion about the appropriateness of the construction will arise. If style manuals, textbooks on writing, instructors’ comments and such really do have an effect on whether or not a whole construction is self-edited out, it matters even more that the advice that is given is based on a solid understanding of the structural basis of the construction.

Appendix: Exemplary abstracts Abstract (A), from the journal Cell, is a typical abstract from the natural sciences. A large portion of the abstract is devoted to the description of how the data were collected. These events are reported in past tense. There is just one reporting event, and it is a fact construction in present tense (These results strongly suggest . . . ). Typically, the reporting event occurs towards the end of the abstract. (A) Signal Recognition Particle to the Translocation Channel Targeting of ribosome–nascent chain complexes to the translocon in the endoplasmic reticulum is mediated by the concerted action of the signal recognition particle (SRP) and the SRP receptor (SR). Ribosome-stripped microsomes were digested with proteases to sever cytoplasmic domains of SR, SR, TRAM, and the Sec61 complex. We characterized protein translocation intermediates that accumulate when Sec61 or SR is inactivated by proteolysis. In the absence of a functional Sec61 complex, dissociation of SRP54 from the signal sequence is blocked. Experiments using SR proteoliposomes confirmed the assembly of a membrane-bound posttargeting intermediate. These results strongly suggest that the Sec61 complex regulates the GTP hydrolysis cycle of the SRP-SR complex at the stage of signal sequence dissociation from SRP54. (Cell12)

Abstract (B) is a typical abstract from the Cambridge Journal of Economics. It has three reporting events (present, stress, suggest), one of them realized in the passive. (B) ‘Work first’: workfare and the regulation of contingent labour markets The paper presents a critical review of UK and US welfare-to-work strategies, stressing their implications for changing forms of labour regulation. The favoured policy orientation – ‘work first’ - forcefully redistributes the risks and burdens of job-market instability from the state to unemployed individuals, the solution to whose ‘welfare dependency’ is presented in terms of a one-way transition into (low) waged work. At a systemic level, the analysis suggests that a regressive regulatory accommodation may be emerging between mandatory welfare-to-work programming on the one hand and the lowest reaches of deregulated, ‘flexible’ labour markets on the other, as the destabilisation of welfare via work-activation measures creates a forced labour supply for contingent jobs. (CJE6)

Abstract (C), also from the Cambridge Journal of Economics, is an example of an abstract in which the research questions are spelled out as direct whquestions, but there is no hint to the answer that will be given.

204

Appendix: Exemplary abstracts

(C) Experimental economics under the microscope Although experimental economics has been one of the most rapidly expanding fields in economics in recent years, it has so far attracted little sustained methodological discussion. This paper is intended as a step towards filling the gap, and provides preliminary answers to the following questions. (i) What are distinctive characteristics of the experimental method in economics? (ii) To what extent are the results obtained in the laboratory transferable to non-laboratory situations? (iii) What are the limits of the experimental method in economics? (iv) Why is it that experimentation, which has been so successful in the natural sciences, remains so controversial in the social sciences. (CJE15)

Abstract (D) is a rare example of a lengthy abstract (140 words) without a single reporting event. There is no reference to the author(s), nor does it become clear that this is the construction of an argument, not merely a report of well-established facts. This kind of abstract – long, but no reporting events – was only found in the Journal of Women’s History. (D) Feminist Maternalists and the French State: Two Inspectresses General in the Pre-World War I Third Republic Although recent historians of women reformers’ contributions to the development of welfare states have underscored the importance of maternalist arguments for opening new roles to women in the public sphere, French examples of women filling such roles have been neglected. The careers of inspectresses general Pauline Kergomard and Olympe Gevin-Cassal provide case studies that illustrate the link between maternalism and women’s access to positions of responsibility in public administration in pre-World War I France. Kergomard, an inspectress general of nursery schools, and Gevin-Cassal, an inspectress general of children’s services for the Ministry of the Interior, utilized maternalist discourse to defend their positions and advocate new professional opportunities for other women. Their secular outlooks suited the anticlerical Third Republic but differentiated them from Catholic women. Gender-specific assignments gave women a place in some inspectorates before 1914 but their numbers were restricted. (JWH8)

Finally, abstract (E) from English Language and Linguistics shows a mix of reporting strategies, among them two uses of the paper construction (The present study focuses on . . . , The study also shows . . . ) and the use of one agent construction (I propose that . . . ). It also shows that reporting by somebody else becomes a part of the reported material (have been generally regarded) and that summary references occur without a by-phrase. (E) Actually and in fact in American English: a data-based analysis Actually and in fact have been generally regarded as interchangeable without leading to any significant differences in the meaning of the containing utterances:

Appendix: Exemplary abstracts

205

as a result, no serious attempt has been made to discover potential differences between the two. The present study focuses on the differences as well as the similarities between actually and in fact in their distribution and use in spoken and written American English. Based upon an analysis of tokens from the Switchboard Corpus and the Brown Corpus, I propose that ‘unexpectedness’ is the core meaning shared by actually and in fact, and that the difference between the two lies in the typical association of each with one or the other way of signalling ‘unexpectedness’. The study also shows that in real discourse contexts, actually and in fact develop a number of different uses that are more or less remote from this core meaning. (ELL9)

References Aarts, Bas 1992

Small Clauses in English: The Nonverbal Types. (Topics in English Linguistics 8). Berlin/New York: Mouton de Gruyter. Abraham, Werner 2003 Ergative diagnostics: temptatio redux. Linguistik Online 13. http://www.linguistik-online.de/13 01/abraham.html. Last accessed April 15, 2009. Ackema, Peter and Maaike Schoorlemmer 1995 Middles and nonmovement. Linguistic Inquiry 26: 173–197. Åfarli, Tor 1989 Passive in Norwegian and English. Linguistic Inquiry 20: 101–108. Alley, Michael 1996 The Craft of Scientific Writing. New York, Berlin, Heidelberg: Springer. Anderson, Mona 1978 NP-preposing in noun phrases. Proceedings of the 8th Annual Meeting of the North-Eastern Linguistic Society: 12–21. APA (ed.) 2001 Publication Manual. Washington: American Psychological Association. Arce-Arenales, Manuel, Melissa Axelrod and Barbara Fox 1994 Active voice and middle diathesis. In: Barbara Fox and Paul Hopper (eds.), Voice. Form and Function, 1–21. Amsterdam: John Benjamins. Aronoff, Mark 1976 Word Formation in Generative Grammar. Cambridge, Mass: MIT Press. Atkinson, Dwight 1992 The evolution of medical research writing from 1735 to 1985. Applied Linguistics 13: 337–374. Atkinson, Dwight 1996 The Philosophical Transactions of the Royal Society of London, 1675–1975: A sociohistorical discourse analysis. Language in Society 25: 333–371. Aue-Apaikul, Panida 2006 Transitivized intransitives: Syntactic, semantic and constructional issues. Ph.D. Dissertation, University of Wisconsin, Madison. Azvedo, Milton Mariano 1980 Passive Sentences in English and Portuguese. Washington, D.C.: Georgetown University Press.

208

References

Baker, Mark 1988a Incorporation. A Theory of Grammatical Function Changing. Chicago: University of Chicago Press. Baker, Mark 1988b Morphology and syntax: An interlocking independence. In: Martin Everaert, A. Evers, M. Huybregts and M. Trommelen (eds.), Morphology and Modularity, 9–32. Baker, Mark, Kyle Johnson and Ian Roberts 1989 Passive arguments raised. Linguistic Inquiry 20: 219–251. Balcom, Patricia 1997 Why is this happened? Passive morphology and unaccusativity. Second Language Research 13: 1–9. Baron, Dennis 1989 Declining Grammar. Urbana, IL: National Council of Teachers of English. Baxter, Charles 1998 Dysfunctional narratives, or ’Mistakes were made’. In: Charles Baxter (ed.), Burning Down the House. Essays on Fiction, 1–26. St. Paul: Graywolf. Bazerman, Charles 1988 Shaping Written Knowledge: The Genre and Activity of the Experimental Article in Science. Madison: The University of Wisconsin Press. Belletti, Adriana 1988 The case of unaccusatives. Linguistic Inquiry 19: 1–34. Bhatt, Rajesh and Roumyana Pancheva 2001 Implicit arguments. Ms. University of Texas, Austin/Santa Cruz. Biber, Douglas 1988 Variation across Speech and Writing. Cambridge: Cambridge University Press. Biber, Douglas, Susan Conrad and Randi Reppen 1998 Corpus Linguistics. Cambridge: Cambridge University Press. Biber, Douglas, Edward Finegan and Dwight Atkinson 1994 ARCHER and its challenges: Compiling and exploring a representative corpus of historical English registers. In: Udo Fries, Gunnel Tottie and Peter Schneider (eds.), Creating and Using English Language Corpora. Papers from the Fourteenth International Conference on English Language Research on Computerized Corpora, Zürich 1993, 1–13. Amsterdam/Atlanta: Rodopi. Biber, Douglas, Stig Johansson, Geoffrey Leech, Susan Conrad and Edward Finegan 1999 Longman Grammar of Spoken and Written English. Harlow: Longman.

References

209

Booij, Geert 1992 Morphology, semantics and argument structure. In: I. M. Roca (ed.), Thematic Structure: Its Role in Grammar, 47–64. Berlin/New York: Foris. Borer, Hagit 1994 The projection of arguments. In: E. Benedicto and J. Runner (eds.), Functional Projections: University of Massachusetts Occasional Papers 17, Amherst, MA: GLSA. Borer, Hagit 1998 Deriving passives without theta roles. In: Steven Lapointe, Diane Brentari and Patrick Farrell (eds.), Morphology and Its Relation to Phonology and Syntax, 60–99. Stanford, CA: CSLI Publications. Borer, Hagit and Kenneth Wexler 1987 The maturation of syntax. In: Thomas Roeper and Edward Williams (eds.), Parameter Setting, 123–172. Dordrecht: D. Reidel Publishing Company. Borer, Hagit and Kenneth Wexler 1992 Bi-unique relations and the maturation of grammatical principles. Natural Language and Linguistic Theory 10: 147–189. Bresnan, Joan 1982 The passive in lexical theory. In: Joan Bresnan (ed.), The Mental Representation of Grammatical Theories, 3–86. Cambridge, MA: MIT Press. Broder, John M. 2007 Familiar fallback for officials: ’Mistakes were made’. The New York Times, March 14, 2007. Brown, Penelope and Stephen Levinson 1978 Universals in language usage: Politeness phenomena. In: Esther Goody (ed.), Questions and Politeness: Strategies in Social Interaction, 56–289. New York: Cambridge University Press. Brown, Penelope and Stephen C. Levinson 1987 Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Budwig, Nancy 1990 The linguistic marking of nonprototypical agency: An exploration into children’s use of passives. Linguistics 28: 1221–1252. Burzio, Luigi 1986 Italian Syntax. A Government-Binding Approach. Dordrecht: Kluwer. Burzio, Luigi 2000 Anatomy of a generalization. In: Eric Reuland (ed.), Arguments and Case. Explaining Burzio’s Generalization., Amsterdam: John Benjamins.

210

References

Bußmann, Hadumod 2002 Lexikon der Sprachwissenschaft. Stuttgart: Kröner. Bybee, Joan and Paul Hopper 2001 Introduction to frequency and the emergence of linguistic structures. In: Joan Bybee and Paul Hopper (eds.), Frequency and the Emergence of Linguistic Structures, 1–23. Amsterdam: Benjamins. Carnie, Andrew 2007 Syntax. A Generative Introduction. Malden, MA: Blackwell. Carter, Ronald and Michael McCarthy 1999 The English get-passive in spoken discourse: Description and implications for an interpersonal grammar. English Language and Linguistics 3: 41–48. Cassidy, Frederic (ed.) 1985 Dictionary of American Regional English. Vol. 1. Cambridge, MA: The Belknap Press of Harvard University Press. Chappell, Hilary 1980 Is the get-passive adversative? International Journal of Human Communication 13: 411–452. Chomsky, Noam 1957 Syntactic Structures. The Hague: Mouton. Chomsky, Noam 1965 Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam 1970 Remarks on nominalization. In: Noam Chomsky (ed.), Studies on Semantics in Generative Grammar, 11–61. The Hague: Mouton. Chomsky, Noam 1981 Lectures on Government and Binding. Dordrecht: Foris. Chomsky, Noam 1986 Knowledge of Language. Westport: Praeger. Chomsky, Noam 1995 The Minimalist Program. Cambridge, MA: MIT Press. Chomsky, Noam and Howard Lasnik 1993 The theory of principles and parameters. In: Joachim Jacobs, Arnim von Stechow, Wolfgang Sternefeld and Theo Vennemann (eds.), Syntax: An International Handbook of Contemporary Research, 506–569. Berlin: Walter de Gruyter. Collins, Peter 1996 Get-passives in English. World Englishes 15: 43–56. Comrie, Bernard 1988 Passive and voice. In: Masayoshi Shibatani (ed.), Passive and Voice, 9–23. Amsterdam: John Benjamins.

References

211

Cornelis, Louise 1996 English and Dutch: The passive difference. Language Sciences 18: 247–264. Couper-Kuhlen, Elizabeth 1979 The Prepositional Passive in English: A Semantic-Syntactic Analysis, with a Lexicon of Prepositional Verbs. Tübingen: Niemeyer. Croft, William 2001 Radical Construction Grammar. Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. Crystal, David 1998 English as a Global Language. Cambridge: Cambridge University Press. Czepluch, Hartmut 1982 Case theory and the dative construction. The Linguistic Review 2: 1–38. Davies, William and Stanley Dubinksy 2004 The Grammar of Raising and Control: A Course in Syntactic Argumentation. Oxford: Blackwell. Denison, David 1993 English Historical Syntax: Verbal Constructions. London/New York: Longman. DiSciullo, Anna-Maria and Edwin Williams 1987 On the Definition of Word. Cambridge, MA: MIT Press. Dixon, R. and Alexandra Aikhenvald 1997 A typology of argument-determined constructions. In: Joan Bybee, John Haiman and Sandra Thompson (eds.), Essays on Language Function and Language Type, 73–113. Amsterdam: John Benjamins. Donadio, Rachel. 2006 Dumbing up. The New York Times, Sep. 24, 2006. Dorgeloh, Heidrun and Anja Wanner 2003 Too abstract for agents? The syntax and semantics of agentivity in English research articles. In: Holden Härtl and Heike Tappe (eds.), Mediating between Concepts and Grammar, 433–453. Berlin: Mouton de Gruyter. Dorgeloh, Heidrun and Anja Wanner 2009 Formulaic argumentation in scientific discourse. In: Roberta Corrigan, Edith Moravcsik, Hamid Ouali and Kathleen Wheatley (eds.), Formulaic Language. (Vol. 2: Acquisition, Loss, Psychological Reality, and Functional Explanations), 523–544. Amsterdam: John Benjamins. Dowty, David 1979 Word Meaning and Montague Grammar. Dordrecht: Reidel. Dowty, David 1991 Thematic proto-roles and argument selection. Language 67: 547–619.

212

References

Dryer, Matthew 1982 In defense of a universal passive. Linguistic Analysis 10: 53–60. Fabb, Nigel 1984 Syntactic affixation. Ph.D. Dissertation, Massachusetts Institute of Technology. Fagan, Sarah 1988 The English middle. Linguistic Inquiry 19: 181–203. Fagan, Sarah 1992 The Syntax and Semantics of Middle Constructions. A Study with Special Reference to German. Cambridge: Cambridge University Press. Felser, Claudia 1999 Verbal Complement Clauses. Amsterdam/Philadelphia: John Benjamins. Felser, Claudia and Anja Wanner 2001 The syntax of cognate and other unselected objects. In: Nicole Deh´e and Anja Wanner (eds.), Structural Aspects of Semantically Complex Verbs, 105–130. Frankfurt/Berlin: Peter Lang. Fischer, Olga, Ans van Kemenade, Willem Koopman and Wim van der Wurff 2000 The Syntax of Early English. Cambridge: Cambridge University Press. Garces-Conejos, Pilar and Antonia Sanchez-Maccaro 1998 Scientific discourse as interaction: Scientific articles vs. popularisations. In: Antonia Sanchez-Maccaro and Ronald Carter (eds.), Linguistic Choice across Genres: Variation in Spoken and Written English, 173–190. Amsterdam/Philadelphia: John Benjamins. G´ıvon, Talmy 1979 On Understanding Grammar. New York: Academic Press. G´ıvon, Talmy 1993 English Grammar: A Function-Based Introduction. Amsterdam/ Philadelphia: John Benjamins. G´ıvon, Talmy and Lynne Yang 1994 The rise of the English get-passive. In: Barbara Fox and Paul Hopper (eds.), Voice. Form and Function, 119–149. Amsterdam: John Benjamins. Goldberg, Adele 1995 Constructions: A Construction Grammar Approach to Argument Structure. Chicago: University of Chicago Press. Goldberg, Adele 2002 Surface generalization. An alternative to alternations. Cognitive Linguistics 13: 327–356. Goldberg, Adele and Farrell Ackerman 2001 The pragmatics of obligatory adjuncts. Language 77: 798–814.

References

213

Goodall, Grant 1993 On case and the passive morpheme. Natural Language and Linguistic Theory 11: 31–44. Grimshaw, Jane 1986 Unaccusatives: An overview. Proceedings of the 17th Annual Meeting of the North-Eastern Linguistic Society: 244–258. Grimshaw, Jane 1990 Argument Structure. Cambridge, MA: MIT Press. Grimshaw, Jane and Sten Vikner 1991 Obligatory adjuncts and the structure of events. Ms. Brandeis University, Waltham, MA. Grimshaw, Jane and Sten Vikner 1993 Obligatory adjuncts and the structure of events. In: Eric Reuland and Werner Abraham (eds.), Knowledge and Language, 143–156. Dordrecht: Kluwer Academic Publishers. Gronemeyer, Claire 1999 On deriving complex polysemy: The grammaticalization of get. English Language and Linguistics 3: 1–39. Gross, Alan, Joseph Harmon and Michael Reidy 2002 Communicating Science: The Scientific Article from the 17th Century to the Present. Oxford: Oxford University Press. Guasti, Maria Teresa 2002 Language Acquisition. The Growth of Grammar. Cambridge, Mass: MIT Press. Haegeman, Liliane 1985 The get-passive and Burzio’s generalization. Lingua 66: 53–77. Haegeman, Liliane 1994 Introduction to Government and Binding Theory. Oxford: Blackwell. Haegeman, Liliane 2006 Thinking Syntactically. A Guide to Argumentation and Analysis. Oxford: Blackwell. Haegeman, Liliane and Jacqueline Gu´eron 1999 English Grammar. A Generative Perspective. Oxford: Blackwell. Hale, Ken and Samuel J. Keyser 1993 On argument structure and lexical expression of syntactic relations. In: Ken Hale and Samuel J. Keyser (eds.), The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger, 53–110. Cambridge, MA: MIT Press. Halpers, Aaron 1998 Clitics. In: Andrew Spencer and Arnold Zwicky (eds.), The Handbook of Morphology, 101–122. Oxford: Blackwell.

214

References

Han, Zhao-Hong 2000 Persistence of the implicit influence of NL: The case of the pseudopassive. Applied Linguistics 21: 78–105. Hansen, Kristine 1998 A Rhetoric for the Social Sciences. A Guide to Academic and Professional Communication. Upper Saddle River: Prentice Hall. Haspelmath, Martin 1990 The grammaticization of passive morphology. Studies in Language 14: 25–72. Hawkins, John 1986 A Comparative Typology of English and German: Unifying the Contrasts. London: Croom Helm. Herold, R. 1986 A quantitative study of the alternation between the be- and the getpassives. Paper presented at the 15th NWAV (New Ways of Analyzing Variation) Conference. Stanford University. Hoekstra, Teun 2000 The nature of verbs and Burzio’s Generalization. In: Eric Reuland (ed.), Arguments and Case. Explaining Burzio’s Generalization, 57–78. Amsterdam: John Benjamins. Hopper, Paul and Sandra Thompson 1980 Transitivity in grammar and discourse. Language Sciences 56: 251– 299. Horgan, Diane 1978 The development of the full passive. Journal of Child Language 5: 65– 80. Hornstein, Norbert 2000 Control in GB and minimalism. In: Lisa Cheng and Rint Sybesma (eds.), The First GLOT State-of-the-Article Book, 27–45. Berlin/New York: Mouton de Gruyter. Hornstein, Norbert and Amy Weinberg 1981 Case theory and preposition stranding. Linguistic Inquiry 12: 55–91. Huang, James 1984 On the distribution and reference of empty pronouns. Linguistic Inquiry 15: 531–574. Huddleston, Rodney and Geoffrey Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Huddleston, Rodney and Geoffrey Pullum 2005 A Student’s Introduction to English Grammar. Cambridge: Cambridge University Press.

References Hust, Joel 1977

215

The syntax of the unpassive construction in English. Linguistic Analysis 3: 31–63. Institute for Scientific Information 2007 Journal Citation Reports, Science Edition. ISI Web of Knowledge. Iwakura, Kunihiro 1987 A government approach to double object constructions. Linguistic Analysis 17: 78–98. Iwata, Seizi 1999 On the status of implicit arguments in middles. Journal of Linguistics 35: 527–553. Jackendoff, Ray 1972 Semantic Interpretation in Generative Grammar. Cambridge, MA: MIT Press. Jackendoff, Ray 1983 Semantics and Cognition. Cambridge, MA: MIT Press. Jackendoff, Ray 1987 The status of thematic relations in linguistic theory. Linguistic Inquiry 18: 369–411. Jackendoff, Ray 1990 Semantic Structures. Cambridge, MA: MIT Press. Jackendoff, Ray 1992 Babe Ruth homered his way into the hearts of America. In: Tim Stowell and Eric Wehrli (eds.), Syntax and the Lexicon, 155–178. San Diego: Academic Press. Jaeggli, Osvaldo 1986 Passive. Linguistic Inquiry 17: 587–622. Jaeggli, Osvaldo and Ken Safir 1989 Introduction: The null subject parameter and parametric theory. In: Osvaldo Jaeggli and Ken Safir (eds.), The Null Subject Parameter, 1– 44. Dordrecht: Kluwer. Jespersen, Otto 1924 The Philosophy of Grammar. New York: Holt & Co. Jespersen, Otto 1927 A Modern English Grammar on Historical Principles. Reprinted 1954. London: Bradford & Dickens. Johansson, Stig and Signe Oksefjell 1996 Towards a unified account of the syntax and semantics of GET. In: Jenny Thomas and Mick Short (eds.), Using Corpora for Language Research, 57–75. London/New York: Longman. Jones, Charles 1988 Thematic relations in control. In: Wendy Wilkins (ed.), Thematic Relations, 75–89. San Diego: Academic Press.

216

References

Ju, Min Kyong 2000 Overpassivization errors by second language learners. Studies in Second Language Acquisition 22: 85–111. Keyser, Samuel and Thomas. Roeper 1984 On the middle and ergative constructions in English. Linguistic Inquiry 15: 381–416. Kirkman, John 1996 First, teach the teachers. Nature 384: 20. Kratzer, Angelika 1996 Severing the external argument from its verb. In: Johan Rooryck and Laurie Zaring (eds.), Phrase Structure and the Lexicon, 109–137. Dordrecht: Kluwer. Lakoff, Robin 1971 Passive resistance. Papers from the Regional Meeting of the Chicago Linguistic Society 7: 149–162. Lambrecht, Knud 1994 Information Structure and Sentence Form. Topic, Focus and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. Larson, Richard K. 1988 On the double-object construction. Linguistic Inquiry 19: 335–391. Larson, Richard K. 1991 Promise and the theory of control. Linguistic Inquiry 22: 103–139. Leather, Simon 1996 The case for the passive voice. Nature 381: 467. Levin, Beth 1993 English Verb Classes and Alternations. A Preliminary Investigation. Chicago: University of Chicago Press. Levin, Beth and Malka Rappaport 1986 The formation of adjectival passives. Linguistic Inquiry 17: 623–661. Levin, Beth and Malka Rappaport Hovav 1995 Unaccusativity. At the Syntax-Semantics Interface. Cambridge, MA: MIT Press. Levin, Beth and Malka Rappaport Hovav 1998 Morphology and lexical semantics. In: Arnold Zwicky and Andrew Spencer (eds.), Handbook of Morphology, 248–271. Oxford: Blackwell. Levin, Beth and Malka Rappaport Hovav 2005 Argument Realization. Cambridge: Cambridge University Press. Li, Charles and Thompson, Sandra 1976 Subject and topic: A new typology of language. In: Charles Li (ed.), Subject and Topic, 457–489. New York: Academic Press.

References

217

Lipson, Mimi 2000 Co-variance of form and meaning in the loss of auxiliary selection in English. In: Regine Eckardt and Klaus von Heusinger (eds.), Meaning Variation – Meaning Change, 111–122. Working Papers of the Fachbereich Sprachwissenschaft 106, University of Konstanz. Mair, Christian and Geoffrey Leech 2006 Current changes in English syntax. In: Bas Aarts and April McMahon (eds.), Handbook of English Linguistics, 318–342. Malden, MA: Blackwell. Manzini, Maria Rita 1983 On control and control theory. Linguistic Inquiry 14: 421–446. Maratsos, M., D. Fox, J. Becker and M. Chalkley 1985 Semantic restrictions on children’s passives. Cognition 19: 167–191. Marin-Arrese, Juana, Elena Martinez-Caro and Soledad Perez de Ayala Becerril 2001 A corpus study of impersonalization strategies in English and Spanish. In: Paul Rayson, Andrew Wilson, Tony McEnery, Andrew Hardie and Shereen Khoja (eds.), Proceedings of Corpus Linguistics 2001, 369–377. Lancaster: UCREL (Technical Papers Vol. 13). Mauner, Gail and Jean-Pierre Koenig 1999 Lexical encoding of event participant information. Brain and Language 68: 178–184. Mauner, Gail, Michael Tanenhaus and Greg Carlson 1995 Implicit arguments in sentence processing. Journal of Memory and Language 34: 357–382. Meints, Kerstin 2003 To get or to be? Use and acquisition of get- versus be-passives: Evidence from children and adults. In: Hubert Cuyckens, Thomas Berg, Ren´e Dirven and Klaus-Uwe Panther (eds.), Motivation in Language. Studies in Honor of Günter Radden, 123–150. Amsterdam: Benjamins. Melinger, Alissa and Gail Mauner 1999 When are implicit agents encoded? Evidence from cross-modal naming. Brain and Language 68: 185–191. Meyer, Charles F. 2002 English Corpus Linguistics. An Introduction. Cambridge: Cambridge University Press. Myers, Greg 1985 Text as knowledge claims: The social construction of two biology articles. Social Studies of Science 15: 593–630. Myers, Greg 1989 The pragmatics of politeness in scientific articles. Applied Linguistics 10: 1–35.

218

References

Myers, Greg 1992 “In this paper we report”: Speech acts and scientific facts. Journal of Pragmatics 17: 295–313. Newmeyer, Frederick 1998 Language Form and Language Function. Cambridge, MA: MIT Press. O’Grady, William 1997 Syntactic Development. Chicago: University of Chicago Press. Oehrle, Richard 1976 The grammatical status of the English dative alternation. Ph.D. Disssertation, Massachusetts Institute of Technology. Oosten, Jeanne van 1986 The nature of subjects, topics and agents: A cognitive explanation. Ms. Indiana University Linguistics Club, Bloomington. Orasan, Constantin 2001 Patterns in scientific abstracts. In: Paul Rayson, Andrew Wilson, Tony McEnery, Andrew Hardie and Shereen Khoja (eds.), Proceedings of Corpus Linguistics 2001, 433–442. Lancaster: UCREL (Technical Papers Vol. 13). Oshita, Hiroyuki 2000 What is happened may not be what appears to be happening: A corpus study of ’passive’ unaccusatives in L2 English. Second Language Research 16: 293–324. Penrose, Ann and Steven Katz 1998 Writing in the Sciences. Exploring Conventions of Scientific Discourse. New York: St. Martin’s Press. Perlman, Alan 1996 Passive voice. Nature 382: 108. Perlmutter, David 1978 Impersonal passives and the unaccusative hypothesis. Proceedings of the Annual Meeting of the Berkeley Linguistic Society 4: 157–189. Pinker, Steven 1989 Learnability and Cognition. The Acquisition of Argument Structure. Cambrige, MA: MIT Press. Pinker, Steven, David Lebeaux and Loren Frost 1987 Productivity and constraints in the acquisition of the passive. Cognition 26: 195–267. Pullum, Geoffrey 2008 Some egg preceded every chicken. Language Log. http://itre.cis.upenn.edu/∼myl/languagelog/archives/005431.html. Last accessed August 24, 2008.

References

219

Pustejovsky, James 1991 The syntax of event structure. In: Beth Levin and Steven Pinker (eds.), Lexical and Conceptual Semantics, 47–81. Cambridge, MA: Blackwell. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1985 A Comprehensive Grammar of the English Language. London: Longman. Rappaport Hovav, Malka and Beth Levin 1998 Building verb meanings. In: Miriam Butt and Wilhelm Geuder (eds.), The Projection of Arguments: Lexical and Compositional Factors, 97– 134. Stanford, CA: CSLI Publications. Rappaport Hovav, Malka and Beth Levin 2001 An event structure account of English resultatives. Language 77: 766– 797. Rappaport, Malka and Beth Levin 1988 What to do with theta-roles. In: Wendy Wilkins (ed.), Thematic Relations, 7–36. San Diego, CA: Academic Press. Reinhart, Tanja 2000 The theta system: Syntactic realization of verbal concepts. OTS Working Papers in Linguistics, Utrecht University. Reuland, Eric 2000 Explaining Burzio’s Generalization. In: Eric Reuland (ed.), Arguments and Case, 1–9. Amsterdam: John Benjamins. Ritter, Elizabeth and Sara Thomas Rosen 2000 Event structure and ergativity. In: Carol Tenny and James Pustejovsky (eds.), Events as Grammatical Objects, 187–238. Stanford, CA: CSLI. Roberts, Ian 1987 The Representation of Implicit and Dethematized Subjects. Dordrecht: Foris. Roeper, Thomas 1984 Implicit arguments. Ms. University of Massachusetts, Amherst. Roeper, Thomas 1987 Implicit arguments and the head-complement relation. Linguistic Inquiry 18: 267–310. Rosen, Sara Thomas 2003 The syntactic representation of linguistic events. In: Lisa Cheng and Rint Sybesma (eds.), The Second GLOT International State-of-theArticle Book, 323–365. Berlin/New York: Mouton de Gruyter. Rudwick, Martin 1985 The Great Devonian Controversy. The Shaping of Scientific Knowledge among Gentlemanly Specialists. Chicago: University of Chicago Press. Safire, William 2003 On language. The New York Times Magazine, November 2, 2003.

220

References

Science (ed.) 2007 Some notes on Science style. Science Magazine. http://www.sciencemag.org/about/authors/prep/res/style.dtl. Last accessed June 20, 2007. Selkirk, Elisabeth 1982 The Syntax of Words. Cambridge, MA: MIT Press. Seoane, Elena 2006 Changing styles: On the recent evolution of scientific British and American English. In: Christiane Dalton-Puffer, Dieter Kastovsky, Nikolaus Ritt and Herbert Schendel (eds.), Syntax, Style and Grammatical Norms: English from 1500–2000, Bern: Peter Lang. Shibatani, Masayoshi 1985a Passives and related constructions: A prototype analysis. Language Sciences 61: 821–848. Shibatani, Masayoshi 1985b Voice in Philippine languages. In: Masayoshi Shibatani (ed.), Passive and Voice, 85–142. Amsterdam: John Benjamins. Siewierska, Anna 1984 The Passive: A Comparative Linguistic Analysis. London: Croom Helm. Simpson, John and Edmund Weiner (eds.) 1989 The Oxford English Dictionary, 2nd Edition. Oxford: Oxford University Press. Slobin, Dan 1966 Grammatical transformations and sentence comprehension in childhood and adulthood. Journal of Verbal Learning and Verbal Behavior 5: 219–227. Speas, Margaret 1990 Phrase Structure in Natural Language. Dordrecht/Boston: Kluwer Academic Publishers. Stein, Gabriele 1979 Studies in the Function of the Passive. Tübingen: Narr. Stowell, Tim and Eric Wehrli 1992 Syntax and the Lexicon. San Diego, CA: Academic Press. Stroik, Thomas 1992 Middles and movement. Linguistic Inquiry 23: 127–137. Sussex, Roland 1982 A note on the get-passive construction. Australian Journal of Linguistics 2: 83–92. Svartvik, Jan 1966 On Voice in the English Verb. The Hague: Mouton.

References

221

Swales, John 1990 Genre Analysis: English in Academic and Research Settings. Cambridge: Cambridge University Press. Swales, John 2001 EAP-related linguistic research: An intellectual history. In: John Flowerdew and Matthew Peacock (eds.), Research Perspectives on English for Academic Purposes, 42–54. Cambridge: Cambridge University Press. Taranto, Gina 2002 An event-structure analysis of causative and passive get. Ms. University of California, San Diego. Tenny, Carol 1994 Aspectual Roles and the Syntax-Semantics Interface. Dordrecht/Boston: Kluwer. Thompson, Sandra 1978 Modern English from a typological point of view: Some implications of the function of word order. Linguistische Berichte 54: 19–35. Thompson, Sandra 1987 The passive in English: A discourse perspective. In: Robert Channon and Linda Shockey (eds.), In Honor of Ilse Lehiste, 497–511. Dordrecht: Foris. Tomasello, Michael 2003 Constructing a Language: A Usage-Based Theory of Language Acquisition. Cambridge, MA: Harvard University Press. Traugott, Elizabeth 1972 A History of English Syntax. A Transformational Approach to the History of English Sentence Structure. New York: Holt, Rinehart and Winston. Traugott, Elizabeth and Bernd Heine 1991 Introduction. In: Elizabeth Traugott and Bernd Heine (eds.), Approaches to Grammaticalization, Vol. 2, 1–14. Amsterdam: John Benjamins. Trumble, William R. and Angus Stevenson (eds.) 2002 Shorter Oxford English Dictionary, 5th Edition. Oxford: Oxford University Press. Turner, E. and R. Rommetveit 1967 The acquisition of sentence voice and reversibility. Child Development 38: 649–660. Vendler, Zeno 1967 Linguistics in Philosophy. Ithaca, New York: Cornell University Press. Verkuyl, Henk J. 1993 A Theory of Aspectuality. Cambridge: Cambridge University Press.

222

References

Verrips, Maike 2000 Passives and implicit arguments in child language. In: Catherine Howell, Sarah Fish and Sarah Keith Lucas (eds.), BUCLD 24: Proceedings of the 24th Annual Boston University Conference on Language Development, 749–760. Boston: Somerville, MA: Cascadilla Press. Wanner, Anja 1999 Verbklassifizierung und aspektuelle Alternationen im Englischen. Tübingen: Max Niemeyer. Wasow, Thomas 1977 Transformations and the lexicon. In: Peter Culicover, Thomas Wasow and Adrian Akmajian (eds.), Formal Syntax, 327–360. New York: Academic Press. Wegener, Heide 1985 “Er bekommt widersprochen” – Argumente für die Existenz eines Dativpassivs im Deutschen. Linguistische Berichte 96: 127–139. Williams, Edwin 1981a Argument structure and morphology. The Linguistic Review 1: 81– 114. Williams, Edwin 1981b On the notions ’lexically related’ and ’head of a word’. Linguistic Inquiry 12: 245–274. Williams, Edwin 1987 Implicit arguments, the binding theory, and control. Natural Language and Linguistic Theory 5: 151–180. Williams, Joseph 1999 Style: Ten Lessons in Clarity and Grace. New York: Longman. Wood, Alistair 2001 International scientific English: The language of research scientists around the world. In: John Flowerdew and Matthew Peacock (eds.), Research Perspectives on English for Academic Purposes, 71–83. Cambridge: Cambridge University Press. Wunderlich, Dieter 1997 Cause and the structure of verbs. Linguistic Inquiry 28: 27–68. Wyngaerd, Guido vanden 1988 Passive and the analysis of auxiliary verbs. Linguistics in the Netherlands 19 (= AVT Publications 4): 159–168. Xiao, Richard, Tony McEnery and Yufang Qian 2006 Passive constructions in English and Chinese: A corpus-based contrastive study. Languages in Contrast 6: 109–149. Yakhontova, Tatyana 2002 Titles of conference presentation abstracts: A cross-cultural perspective. In: Eija Ventola, Celia Shalom and Susan Thompson (ed.), The Language of Conferencing, 301–313. Frankfurt: Peter Lang.

References

223

Zaenen, Annie 1993 Unaccusativity in Dutch: Integrating syntax and lexical semantics. In: James Pustejovsky (ed.), Semantics and the Lexicon, 129–161. Dordrecht/Boston: Kluwer. Zandvoort, Reinard 1957 A Handbook of English Grammar. London: Longman. Zobl, Helmut 1989 Canonical typological structures and ergativity in English L2 acquisition. In: Susan Gass and Jacqueline Schachter (eds.), Linguistic Perspectives on Second Language Acquisition, 203–221. Cambridge: Cambridge University Press. Zubizaretta, Maria Luisa 1985 The relation between morphophonology and morphosyntax. The case of Roman causatives. Linguistic Inquiry 16: 695–727. Zubizaretta, Maria Luisa 1998 Prosody, Focus, and Word Order. Cambridge, MA: MIT Press. Zwicky, Arnold 2006 How long have we been avoiding the passive, and why? Language Log. http://itre.cis.upenn.edu/ %7Emyl/languagelog/archives/003380.html. Last accessed October 29, 2007.

Sources of examples Sources of examples Austen, Jane (1985): Emma. Harmondsworth: Penguin. (Original work published 1816) Austen, Jane (1985): Mansfield Park. Harmondsworth: Penguin. (Original work published 1814) Bront¨e, Emily (1985): Wuthering Heights. Harmondsworth: Penguin. (Original work published 1847) Bacon, Charlotte (2002): Lost Biography. New York: Picador. Davis, William (1994): A Government of Our Own: The Making of the Confederacy. New York: Free Press Fielding, Helen (1996): Bridget Jones’s Diary. London: Picador. Fielding, Helen (1999): Bridget Jones – The Edge of Reason. London: Picador. Hitchens, Christopher: “Stenographer to the Stars.” Retrieved January 1, 2008, from http://www.salon.com/weekly/woodward960701.html. Taylor, Alan (Director) and Liz Tuccillo (Writer) (2003): The post-it always sticks twice [Television series episode]. Michael Patrick King (Producer), Sex and the City. New York: Home Box Office. Klemperer, Viktor (1995): Ich will Zeugnis ablegen bis zum letzten [I will bear witness until the end]. Berlin: Aufbau-Verlag. Leduff, Charlie (2003, February 1): A wanderer for peace has run out of road. The New York Times. Retrieved January 1, 2008, from http://www.nytimes.com/2003/02/01/us/a-wanderer-for-peace-has-run-out-ofroad.html Obama, Barack (2009, January 20): Inaugural Address. Retrieved January 22, 2009, from http://www.nytimes.com/2009/01/20/us/politics/20text-obama.html. Reagan, Ronald (1986, December 6): Radio Address to the Nation on the Iran Arms and Contra Aid Controversy. Retrieved January 1, 2008, from http://www.reagan.utexas.edu/archives/speeches/1986/120686a.htm. Penchaszade, Analia: ‘El Desaperecido’ as a Terror Tactic that Lasts Past State Terrorism. Retrieved January 1, 2008, from http://www.iisd.org/YOUTH/ysbk047a.htm.

Subject and author index a-adjunct 149 absorption of Case 56f., 60, 62, 68, 91, 154 of theta role 121, 136, 141f., 148f., 154 abstract as a genre 82, 156–159, 205–207 A-chain 44, 65 (Fn. 31) adjectival passive 19f., 32, 72–75, 90 misclassified as verbal passive 97, 168 with get 95–98, 102f. acquisition of the passive 42–48, 55 affectedness 21, 32, 36, 55, 69f., 111, 144, 183 agent as conceptual argument 128f. as external argument 17f., 32f. as implicit argument see implicit argument demotion/backgrounding of 11, 38f., 58, 188 in scientific writing 158, 161, 170, 174f., 179–182, 190, 196 non-agent as subject 9f., 14, 40, 79, 82 relationship to by-phrase 21, 76f., 85 secondary agent 86, 105f., 112 aktionsart 36, 125f., 144 animacy 80, 100, 102, 104, 112, 158, 179 argument alternation 8f., 36, 49 argument structure 4, 33, 66, 121, 124, 129, 147–150, 153 argument suppression 59, 148–150 Bazerman, Charles 193f. be (auxiliary) 15, 17, 83–85, 139, 163

auxiliary selection 23 (Fn. 14), 44, 60, 72, 85 in second language acquisition 138–140 passives without be (bare passives) 17f., 83–85, 93, 110, 170–172, 198 Biber, Douglas 9f., 12, 38–41, 77f., 80, 88, 97f., 107, 158 (Fn. 83), 168 (Fn. 88), 174 Binding Theory 122–124 Borer, Hagit 43f., 144–146 Burzio, Luigi 28, 31f., 65 Burzio’s Generalization 28, 53, 61, 134 By-phrase 11, 18f., 27, 40f., 46, 76–80, 101, 120–123, 126f., 142f., 168f., 172, 185 Case absorption see absorption accusative 8, 15f., 54, 56f., 61f., 144, 154 case vs. Case 8 (Fn. 5) Case Filter 8, 10, 41, 53, 115, 145, 188 dative 16f., 57, 61, 91 exceptional Case marking (ECM) 54 inherent Case 53, 56f., 61 nominative 57, 61f., 65, 91f., 135, 139, 154 structural Case 53, 57, 60–62, 92, 154, 198 causative alternation 25, 31 (Fn. 16), 32, 127, 149, 197 chˆomeur 152 Chomsky, Noam 2f., 8, 13, 33, 53, 115, 122, 133f. clitic chain 142, 153

228

Subject and author index

cohesion 11, 38, 40f., 69, 82, 110, 186–188 conative alternation 9, 51 Construction Grammar 3, 28f., 49f., 87, 112 control 42, 105, 115–120, 132, 143f., 146, 164, 188 thematic vs. argument control 118 corpus-linguistic studies of the passive abstract corpus 79, 132, 165–170, 175–183, 186–188, 192, 200 ARCHER 189–191, 194, 201 BROWN 89, 171 FROWN 99–108, 168–172, 175 London-Lund 86 LSWE (Longman Grammar) 12, 38–41, 167, 169 Croft, William 29, 49 DARE (Dictionary of American Regional English) 98, 101f., 109 Denison, David 57, 61, 83 detached persona 160–162, 191 ditransitive verbs 50, 56f., 73, 76 Dowty, David 34, 55, 107, 125 Dutch 12, 19, 60–62, 91, 136 ergative languages 29–31, 34 ergative verbs see causative alternation and unaccusative verbs expletive chain 65 expletive it 12 (Fn. 7), 16, 36, 41, 62–66, 111, 144, 188, 202 Extended Projection Principle (EPP) 9f., 49, 53 (Fn. 27), 62, 135 extraposition 63, 81 fact construction 179, 181f., 185, 190, 194, 201 focus 11, 39, 41, 126, 130 functions of the passive 2, 37–42, 44, 82, 130f., 184, 186–188, 198f. German 12, 19, 58f., 61f., 65, 91, 136–138

gerundial passive 17f., 75f. get-passive 4, 17, 25f., 45, 51, 85–88, 99–102, 199f. adjectival 95–97, 102f. adversity effect 86, 100, 102, 107–108, 202 ergative vs. causative get 91–93 history 89f., 106 reflexive 90, 94f., 101, 106 semantic and stylistic restrictions 86, 88, 98, 101f., 109, 200 subject responsibility 86, 100, 103–107, 112, 200 Goldberg, Adele 3, 9, 28f., 50, 56, 126f. Grimshaw, Jane 34, 117, 120, 125f., 148–150, 152 Haegeman, Liliane 63, 73, 84, 87, 89, 92 Huddleston Rodney 17, 26, 63f., 69, 71, 75, 93, 97, 105 impersonal passive 12 (Fn. 7), 14, 16, 38, 58–66 implicit argument 3f., 9f., 18, 22f., 28, 45, 84, 95, 98, 105, 112f., 180, 191, 199 processing of 127–130 representation of 140–151 speakers’ awareness of 136–138 syntactic potential 115–124, 132, 152, 188 incorporated argument 151 information structure 9f., 38f., 41, 77f., 82, 113, 169 intransitive verbs 12, 34, 44, 46 (Fn. 25), 58–60, 75, 135 Jackendoff, Ray 14, 46, 147, 150f. Jaeggli, Osvaldo 8, 57, 60, 105, 116–119, 121, 141–143 Jespersen, Otto 40, 57, 75f. Kratzer, Angelika

33f., 142f. (Fn. 72)

Subject and author index Lambrecht, Knud 9, 12, 38–41 Levin, Beth 49–51, 73, 75, 90, 124–126, 149, 151 lexical-conceptual structure 124, 147, 150–152 linking rules 8, 29, 40, 46, 76f., 124–127 markedness 9, 14, 25, 30f., 34, 43, 154, 191, 194 Mauner, Gail 127–129 middle construction 10, 21, 25, 32, 116, 124 misclassifications of passives 26, 98, 163 mistakes were made 1, 24f., 201 nominalization 15, 21, 150, 184, 160 NP-Movement 8, 10, 13, 18, 22, 31f., 34f., 44, 53f., 67f., 76, 93, 119, 135, 139, 188 acquisition of 43, 45 passives without 16, 28, 41, 58 paper construction 179, 181f., 185, 191, 194, 201 participial relative clause 18, 82, 84, 110, 170 passive morpheme (-en) 13–17, 59, 70f., 114, 121, 141–143, 154 passive reading 4, 22, 24, 28, 47, 84, 96, 136, 140, 153, 180, 183 passive transformation 7 passives without actives 26f. Perlmutter, David 133–135 Pinker, Steven 14, 37f., 45f., 136f. politeness 183–185, 193 predicate argument structure (PAS) see argument structure prepositional passive 16f., 22, 66–70 principle of end-weight 10, 40, 77f., 81, 169 Principles and Parameters 3, 7, 42, 154, 204 pro 144–146

229

promotion of theme 10, 40f., 58 prototypical agent and patient 34, 37, 46, 48, 55, 59, 111, 198 Pullum, Geoffrey 17, 26, 63f., 69, 71, 75, 93, 97, 105, 164 Quirk, Randolph 95

13, 16, 29, 52, 64,

Rappaport Hovav, Malka 73, 75, 124–126, 149, 151 reanalysis 68f. recipient passive 16f., 19, 56f., 91 reported events 172f., 186–188 reporting events 173–176, 180–183, 189f., 192 resultative construction 60, 70, 74 reversibility 37, 43, 106 Roberts, Ian 61f., 65, 72, 117, 121, 123, 142 Roeper, Thomas 32, 86, 113 Safire, William 23 scientific writing 160–165, 172–177, 189, 191, 193, 200, 202 Seoane, Elena 84f., 171f., 175, 189, 200 Shibatani, Masayoshi 29, 41 Siewierska, Anna 12, 97, 143 stylistic advice (“Avoid passive”) 24, 80, 160, 163f., 172, 191, 194f., 200, 202 sub-event identification 125–127 subject 3f., 18f., 28, 30f., 35, 40–42, 44, 54, 62f., 80–83, 106, 144 Svartvik, Jan 94, 130, 173f. Swales, John 156f., 174 telicity 9, 36f., 144 Tenny, Carol 9, 37 thematic hierarchy 46, 77 theta-roles 14, 18, 33, 115f., 121, 140, 147 assignment to an affix 141f. transmission 72, 121, 143

230

Subject and author index

Thompson, Sandra 11, 36f., 78, 131, 180 topic 9, 36, 40f., 43, 69, 78, 82, 139f., 186–188 transitivity 35–37, 55, 197 unaccusativity 32, 44, 74, 133f. unaccusative verbs 34f., 58f., 72, 74, 89, 133f., 136–140

unergative verbs see intransitive verbs voice

29–34, 175, 201

Williams, Edwin

20, 32, 73

Zobl, Helmut 138–140 Zwicky, Arnold 80