Contributions of Romance Languages to Current Linguistic Theory [1st ed.] 978-3-030-11005-5, 978-3-030-11006-2

Table of contents :
Front Matter ....Pages i-x
Introduction (Julia Herschensohn)....Pages 1-10
Front Matter ....Pages 11-11
Classifier Phrase as Host for Stage-Level Spanish Adnominal Participial Adjectives (Karen Zagona, Heles Contreras)....Pages 13-37
The Syntax of Mirative Focus Fronting: Evidence from French (J.-Marc Authier, Liliane Haegeman)....Pages 39-63
Further Implications of French Devoir and Falloir for Theories of Control and Modality (Lisa A. Reed)....Pages 65-89
Front Matter ....Pages 91-91
Conceptual and Empirical Arguments for a Language Feature: Evidence from Language Mixing (Barbara E. Bullock, Almeida Jacqueline Toribio)....Pages 93-113
On the Reduction of /ʒ/ in a Minority North American Variety of French (Randall Gess)....Pages 115-136
Why Lenition Interactions Are Typically Counter-Feeding (Haike Jacobs)....Pages 137-154
The French Jespersen’s Cycle and Negative Concord (Marie Labelle)....Pages 155-172
Front Matter ....Pages 173-173
The Variable Position of Initial Subordinate Clauses in Old French: Arguments Against a Semantic Account (Bryan Donaldson)....Pages 175-201
Obviation and Old French Subjunctive Clauses (Deborah L. Arteaga)....Pages 203-229
On the Varying Fortunes of darla and darlo in Italian (Cinzia Russi)....Pages 231-251
Conclusion (Julia Herschensohn)....Pages 253-254

Citation preview

Studies in Natural Language and Linguistic Theory 95

Deborah L. Arteaga Editor

Contributions of Romance Languages to Current Linguistic Theory

Contributions of Romance Languages to Current Linguistic Theory

Studies in Natural Language and Linguistic Theory VOLUME 95

Managing Editors Marcel den Dikken, Research Institute for Linguistics, Hungarian Academy of Sciences and Department of English Linguistics, Eötvös Loránd University, Budapest, Hungary Liliane Haegeman, University of Gent, Gent, Belgium Joan Maling, Brandeis University, Waltham, MA, USA Maria Polinsky, University of Maryland, College Park, MD, USA Editorial Board Guglielmo Cinque, University of Venice, Italy Jane Grimshaw, Rutgers University, New Brunswick, USA Michael Kenstowicz, Massachusetts Institute of Technology, Cambridge, MA, USA Hilda Koopman, University of California, Los Angeles, USA Howard Lasnik, University of Maryland, College Park, MD, USA Alec Marantz, New York University, NY, USA John J. McCarthy, University of Massachusetts, Amherst, MA, USA Ian Roberts, University of Cambridge, UK

More information about this series at http://www.springer.com/series/6559

Deborah L. Arteaga Editor

Contributions of Romance Languages to Current Linguistic Theory

123

Editor Deborah L. Arteaga Department of World Languages and Cultures University of Nevada, Las Vegas Las Vegas, NV, USA

ISSN 0924-4670 ISSN 2215-0358 (electronic) Studies in Natural Language and Linguistic Theory ISBN 978-3-030-11005-5 ISBN 978-3-030-11006-2 (eBook) https://doi.org/10.1007/978-3-030-11006-2 Library of Congress Control Number: 2018965915 © Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

This volume is dedicated to Julia Rogers Herschensohn for her significant contributions to the field of Romance Linguistics and for her generous support of students and fellow linguists throughout her career.

Contents

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Julia Herschensohn Part I

Romance Syntax

Classifier Phrase as Host for Stage-Level Spanish Adnominal Participial Adjectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Karen Zagona and Heles Contreras The Syntax of Mirative Focus Fronting: Evidence from French . . . . . . J.-Marc Authier and Liliane Haegeman Further Implications of French Devoir and Falloir for Theories of Control and Modality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lisa A. Reed Part II

1

13 39

65

Language Variation in Romance

Conceptual and Empirical Arguments for a Language Feature: Evidence from Language Mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Barbara E. Bullock and Almeida Jacqueline Toribio

93

On the Reduction of /ʒ/ in a Minority North American Variety of French . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 Randall Gess Why Lenition Interactions Are Typically Counter-Feeding . . . . . . . . . . 137 Haike Jacobs The French Jespersen’s Cycle and Negative Concord . . . . . . . . . . . . . . 155 Marie Labelle

vii

viii

Part III

Contents

Diachronic Romance Studies

The Variable Position of Initial Subordinate Clauses in Old French: Arguments Against a Semantic Account . . . . . . . . . . . . . . . . . . . . . . . . . 175 Bryan Donaldson Obviation and Old French Subjunctive Clauses . . . . . . . . . . . . . . . . . . . 203 Deborah L. Arteaga On the Varying Fortunes of darla and darlo in Italian . . . . . . . . . . . . . . 231 Cinzia Russi Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253 Julia Herschensohn

Contributors

Deborah L. Arteaga Department of World Languages and Cultures, University of Nevada, Las Vegas, Las Vegas, NV, USA J.-Marc Authier Department of French and Francophone Studies, Pennsylvania State University, University Park, PA, USA Barbara E. Bullock Department of French and Italian, University of Texas at Austin, Austin, TX, USA Heles Contreras Department of Linguistics, University of Washington, Seattle, WA, USA Bryan Donaldson Department of Languages and Applied Linguistics, University of California, Santa Cruz, CA, USA Randall Gess Department of French, Carleton University, Ottawa, ON, Canada Liliane Haegeman Department of Linguistics, Ghent University, Ghent, Belgium Julia Herschensohn Department of Linguistics, University of Washington, Seattle, WA, USA Haike Jacobs Centre for Language Studies, Radboud University, Nijmegen, The Netherlands Marie Labelle Department of Linguistics, Université du Québec à Montréal, Montreal, QC, Canada Lisa A. Reed Department of French and Francophone Studies, Pennsylvania State University, University Park, PA, USA

ix

x

Contributors

Cinzia Russi Department of French and Italian, The University of Texas at Austin, Austin, TX, USA Almeida Jacqueline Toribio Department of Spanish and Portuguese, University of Texas at Austin, Austin, TX, USA Karen Zagona Department of Linguistics, University of Washington, Seattle, WA, USA

Introduction Julia Herschensohn

Abstract This chapter first gives a historical overview of the importance of Romance linguistics to the development of diachronic and synchronic theory over the past three centuries. It then presents the articles constituting the volume, highlighting the areas of morphosyntax, phonology and language variation, noting that many of them draw on new empirical databases. Several articles cover diachronic and synchronic morphosyntax, particularly focusing on the functional projections of the left periphery of CP and DP (Arteaga, Authier and Haegeman, Donaldson, Labelle, Reed, Zagona and Contreras). Others examine variation in terms of cross-linguistic, diachronic, dialectal and sociolinguistic factors (Bullock and Toribio, Gess, Jacobs, Russi).













Keywords Corpora Diachronic change Language variation Left periphery Linguistic theory Minimalist program Morphosyntax Phonology Romance linguistics













The field of Romance Linguistics has been pivotal for linguistic theory since its inception (Gess and Arteaga 2006). Studies on the Romance languages have been highly influential in expanding and refining theoretical frameworks (Aboh and Schaeffer 2015; Carrilho et al. 2016; Smith and Ihanse 2015; Tortora et al. 2016). For example, in the Principles and Parameters model, scholars using data from synchronic and diachronic Romance were at the forefront of morphosyntactic theory, furthering our understanding of topics such as the null-subject parameter (Jaeggli and Safir 1989; Vance 1997; Kato and Negrão 2000), verb movement (Emonds 1978; Pollock 1989; Rizzi and Roberts 1989; Belletti 1990), clitics (Kayne 1981; Jaeggli 1986; Rivero 1986), and case theory (Rouveret and Vergnaud 1980; Bouchard 1983; Raposo 1987; Raposo and Uriagereka 1990). With the advent of Minimalism, studies based on Romance have continued to contribute to our understanding of a wide-variety of morphosyntactic structures, J. Herschensohn (&) Department of Linguistics, University of Washington, Seattle, WA, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_1

1

2

J. Herschensohn

such as the structure of the left periphery (Rizzi 1997; Benincà 2006), feature-checking (Carstens 2000; Dobrovie-Sorin and Giurgea 2011), and phases (D’Alessandro and Roberts 2008; Rouveret 2012; Roberts 2012) Both diachronic and synchronic studies have incorporated extensive data from corpora, allowing scholars to base their analyses on more complete data (Vance 1997; Martins and Carrilho 2016; Poplack 1989). In phonology, Klausenburger (1978) and Tranel (1981) contributed to the debate regarding the underlying mental representation of French liaison, which continues to this day (e.g., Côté 2013). Harris (1983) derived Spanish stress from its syllabic structure, while Saltarelli (1983, 1984) used diachronic data to support his analysis of the mora unit in Italian, arguing for two levels of representation, segmental and syllabic. More recently, Romance phonologists have contributed to the development and refinement of classical Optimality Theory (Montreuil 2002; Colina 2009), as well as phonetically-based OT (Pons-Mall 2011; Bradley 2014), testing the relevance of the theory against both synchronic and diachronic data. The chapters in this collection continue this trend, by presenting a wealth of new data on a range of Romance languages, including Corsican, French, Italian, Sardinian, Sicilian, and Spanish, whose characteristics are examined from both synchronic and diachronic perspectives. The subdisciplines examined—syntax, language variation and diachronic studies—encompass additional linguistic threads such as morphosyntax, semantics, phonology, dialectology and sociocultural, considerations. More significantly, they extend the boundaries of linguistic theory, by adding complementary data and key interpretations to our understanding of Romance languages and language in general. Chapter by Chapter Discussion

1 Romance Syntax Karen Zagona and Heles Contreras In an exploration and elaboration of the left periphery of the DP, Zagona and Contreras extend the functions of the Classifier Phrase (CLP) to Spanish prenominal predicative adjectives. Romance predicative adjectives are canonically post-nominal and may carry stage-level (SL, temporal reading) or individual-level (IL) interpretations, whereas prenominal adjectives only allow an IL reading: este paisaje seco ‘this [intrinsically/temporarily] dry landscape’/este seco paisaje ‘this [intrinsically] dry landscape’. In contrast, a class of participial adjectives in Spanish allows SL readings in both post- and pre-nominal position: el cliente contrariado/el contrariado cliente ‘the annoyed customer’. The authors argue that the unique behavior of these participial adjectives is due to their original aspectual properties as verbs, to their syntactic structure, and to the level at which they merge. They first establish that post-nominal adjectives are reduced relatives that may license an SL reading, while pre-nominal ones are nominal specifiers that are syntactically

Introduction

3

constrained and have only an IL interpretation. The participial adjectives in question derive from intransitive pronominal psych verbs incorporating inchoative aspect. Given these characteristics, Zagona and Contreras then argue that the CLP —projected between DP and nP and hosting dimensional adjectives—can accommodate the participial adjectives in Spanish, whose SL interpretation is licensed by the [BOUNDARY] feature indicating its temporal edge. They propose that it is this participial feature that values the [uDimension] feature of the adjective’s c-commanding Degree head through Agree. They provide well-motivated analyses of the differences between pre- and post-nominal adjectives, the distinctions between IL and SL interpretations, and the constraints associated with those categories. Their proposal also furnishes a better understanding of the syntactic structures, features involved and derivational properties related to merge, involving nominal syntax and in particular its left periphery. J.-Marc Authier and Liliane Haegeman Authier and Haegeman flesh out the Left Periphery (LP) of Modern French in an exploration of Mirative Focus Fronting that makes cross-linguistic comparisons with other Romance languages. Using Rizzi’s (1997) architecture of the LP— ForceP > TopP > FocP > TopP > FinP—as a point of departure, they note characteristics of the LP projections. While TopP may be iterative, representing old information and requiring DP clitic doubling in French, generally FocP is not iterative and represents new information. Within FocP, there are two types, Corrective or CFoc, and Mirative or MFoc, the former providing clarification and the latter providing unexpected new information. Sicilian and Sardinian evince both CFoc—which can appear in both embedded and matrix clauses—and MFoc, which is limited to matrix clauses. French only allows MFoc fronting that Authier and Haegeman show allows unbounded dependencies, obeys island constraints and licenses parasitic gaps. They also demonstrate that MFoc and Top have distinct prosodic patterns and the former, but not the latter, may include complementizer que. In terms of interpretation, MFoc DPs share characteristics of expressive content and noteworthiness with wh-exclamatives, but differ in that the latter target intrinsic properties of the DP (i-level), while MFoc DPs target the character of the entire phrase, the event (e-level). Their analysis provides insight into the Left Periphery in a range of Romance languages, distinguishes the syntactic architecture of the French LP, and integrates the semantic content of the subfields of the LP with its syntax. Lisa A. Reed Reed explores the complex syntactic and semantic properties of the French modals falloir and devoir, solving residual puzzles while strengthening her 2014 account of PRO. Both verbs allow non-thematic subjects—IL faut partir ‘it is necessary to leave’; IL doit y avoir du savon ‘there must be mandatorily some soap’—but falloir requires an embedded thematic verb with PRO subject, while devoir prohibits PRO in the non-thematic construction (*IL doit partir ‘it must leave’). She reviews the extensive literature on control and raising over the past several decades, focusing on more recent treatments to show that they cannot account for all the data manifested

4

J. Herschensohn

by these French modals in terms of both syntax and semantics. She uses cross-linguistic data from Quantifier Float, clitic climbing in Romance and gerundive PRO to illustrate the problems faced by the earlier studies. She then demonstrates that her analysis of the modals, coupled with her proposal of PRO as the minimal NP ([+N, −Expletive], with no phi features or case) can fill the gaps she had earlier revealed. PRO can only first merge in a theta position and only in clauses not requiring phi feature bundle valuation (hence non-finite). Falloir can only carry deontic meaning and c-selects a FinP requiring PRO that carries deontic obligation in its proposition. In contrast, devoir enters into raising (deontic, epistemic) or control constructions. In deontic raising, it c-selects a non-thematic vP, giving for example Il doit y avoir du savon. In its epistemic raising construction, devoir c-selects TP to give, for example Paul a dû partir ‘Paul must have left’. Finally, devoir may act as a deontic control verb, c-selecting FinP and requiring a thematic matrix subject. This will usually exclude inanimates (*IL doit impérativement s’arrêter de neiger ‘It’s imperative that it stop snowing’), but may include them if they are subject to the law (Un litre de lait doit contenir 5% de crème. ‘A liter of milk must contain 5% cream’). After a final section on pragmatic approaches to modality, Reed concludes that devoir (unlike must) may be syntactically and semantically ambiguous. Her chapter provides insight into the subtleties of control and raising constructions, argument structure and the syntax-semantics interface.

2 Language Variation in Romance Barbara E. Bullock and Almeida Jacqueline Toribio Bullock and Toribio bring to bear current methodologies to adjudicate two issues related to code switching (CS), the use of a language feature (e.g. [+Spanish]) and the distinction between CS and vocabulary borrowing. The Functional Head Constraint (Belazi et al. 1994), which precludes a language switch between a head and its complement by requiring matching language features on the head and complement, has been criticized for its language feature “as a grammatical primitive”. After reviewing the CS literature of the past 25 years, the authors note that while CS involves alternation of two grammatical systems, borrowing involves only a single system with lexical items inserted from another language. A prime area of confusion is DP, for which the complement of Det is NP; they ask if a mixed language example such as el book is an example of CS or borrowing. Bullock and Toribio provide theoretical arguments countering the anti-language feature ones and adduce stronger empirical evidence to demonstrate the necessity of language marking for CS data. They point out that Natural Language Processing scholarship is dependent on indication of language, as well as part of speech (POS) marking, and then they use NLP procedures to annotate three corpora of CS data (Spanish-English). They set out clear metrics for analyzing the CS tokens in the

Introduction

5

data, using a language identity (LID) model that automatically identifies the language and CS to examine the DP examples in particular. From their quantitative analysis they are able to draw several generalizations based on the three sets of authentic empirical data (drawn from three distinct sociolinguistic settings and with three distinct proportions of Spanish-English use). They observe two types of CS, insertional (in more asymmetric corpora) and alternational (in a corpus with more balanced use of the two languages). The LID also indicates whether the CS is consistently alternational or if it is clustered in sequential sentences. As for DPs, switches are not frequent, but not prohibited either: English NPs are 90% preceded by English Det, but there are examples of Spanish Det + Engl NP in all three corpora. The demonstration of these corpora is a strong argument for the use of automatic tools and computational techniques that the authors advocate here. Randall Gess As part of the Phonology of Contemporary French (PCF) project, Gess collected and analyzed data from Maillardville, British Columbia, the Francophone community the farthest west in Canada. The PCF project has explored the phonologies of the French language in Europe, Africa and the New World, documenting the variant forms and accounting for the sources in both the new and old world. The PFC protocol, that Gess used to interview his subject “AJ”, includes the reading of two word lists (one Canadian), a reading passage, a guided interview and free conversation. The phonological area of interest is “debuccalization” of the fricative /ʒ/ which can be realized with varying degrees of buccal aperture, including [ɦ] and [h]. This phenomenon—also known as saintongeais—can be traced to today’s Charente-Maritime (formerly Saintonge) in France, and more directly to Québec, Ontario, Manitoba, the Maritimes, Louisiana and the Caribbean. AJ, whose parents were exposed to québécois, was born in Saskatchewan and moved to Maillardville at age 14; he was 83 at the time of the interview, during which he produced a good number of examples of /ʒ/ in full and reduced form. Gess gives a thorough phonetic analysis using Praat, pointing out the inherent duality of a voiced fricative, whose acoustic salience is due to making frication noise, a noise that is (somewhat counter-productively) reduced by voicing. He discusses Kingston’s idea that lenition (increased acoustic intensity) is a means of marking continuity, but he concludes that “debuccalization of /ʒ/ can occur at prosodic boundaries, a fact incompatible with Kingston’s (2008) proposal that lenition is intended to signal continuity within a prosodic constituent. We suggest instead that its articulatory complexity is what renders /ʒ/ a salient target for reduction.” Haike Jacobs Jacobs uses the phenomenon of lenition of plosives in Romance languages to compare several theoretical perspectives—rule-based approaches, Optimality Theory (OT), and Articulatory Phonology—on phonological processes. He points out problems with these earlier treatments and argues for a blend of harmonically serial Optimality that is theoretically preferable; furthermore, his proposal gives an explanation related to optimal perception on the part of the hearer (as contrasted

6

J. Herschensohn

with the frequent focus on optimal production by the speaker). Both diachronically (e.g. Latin to French) and synchronically (e.g. Corsican, Gran Canarian), lenition processes include voicing of intervocalic voiceless stops and spirantization of intervocalic voiced stops. Only by ordering spirantization before voicing will the original phonemic contrasts be preserved (as European Structuralists pointed out decades ago). In rule-based phonology, this is a counter-feeding order that is considered by some to be marked in terms of the speaker’s articulation. OT accounts can describe transparent feeding interactions, but may need to appeal to special constraints. Articulatory Phonology is able to deal with non-neutralizing lenition order, but cannot account for all Romance lenition. Finally, Jacobs introduces Harmonic Serialism, for which the underlying representation is modified in a step-like pattern with each step reevaluating the constraint ranking to achieve optimal harmony. It is more constrained that traditional OT. He revisits the lenition data, showing that the step-wise application of the phonological constraints is able to take into account markedness, applies to a limited local domain, and motivates the counterfeeding relation by ease of perception. Marie Labelle Labelle’s chapter reconsiders the well-documented Jespersen Cycle of negation morphosyntax in diachronic and synchronic French, arguing that the changes constitute a spiral rather than a circle, and that the evolution results in two branching dialects dating from the 16th century. She also provides a motivated analysis of the diachronic changes that accounts for possibilities of double negation in some dialects. The standard Jespersenian description includes five stages of negation from Medieval to modern French: ne (Old French, OF) ! ne … (pas) (Middle French) ! ne … pas (Standard French) ! (ne) … pas (Colloquial French) ! pas (Quebec French). One line of argument against the notion of cycle is evident, namely that the endpoint pas is not equivalent to the beginning ne although both are single negators. Furthermore, she demonstrates that the Quebec pas is not the outcome of the continental colloquial dialect, but rather can be traced to the 16th century, as exemplified by Molière in Les femmes savantes (1672). This play contrasts the standard French of Bélise, who corrects her servant’s stigmatized negation (“ne servent pas de rien”) in pronouncing “de pas mis avec rien, tu fais la récidive, et c’est, comme on t’a dit, trop d’une négative” (‘Of pas added to rien you are using again, and this is, as we told you, one negative too many’). This possibility of negator doubling still exists in Quebec, as for example il voit pas personne ‘he does not see anyone’. Assuming a view of diachronic development as grammar competition, Labelle provides an analysis that proposes for OF a covert negative operator in the left periphery valuing [uNEG] of both pre and postverbal negators. In the next stage, the ne loses its [uNEG] and becomes omissible, while pas becomes a negative quantifier by the 17th century. At this point, there are two dialects, the stigmatized one (that will migrate to Quebec) for which most negators retain the [uNEG] feature while the negative quantifier pas now carries [iNEG]. For the second prestige dialect, most negators are negative quantifiers and cannot

Introduction

7

co-occur (e.g. *pas personne). This analysis provides an explanation of both diachronic development and contemporary dialectal variation in the distribution of negation marking in French.

3 Diachronic Romance Studies Bryan Donaldson Through an examination of a broad selection of Old French texts (10th–14th century), Donaldson counters the traditional account attributing preposed subordinate clause word order to semantic properties by clearly demonstrating no systematic correlation between syntax and semantics for these constructions. As a verb second (V2) language, Old French (OF) displays two positions for preposed subordinate clauses, first position (Specifier of Focus Phrase, SpecFoc) immediately left of the verb, or further up in the Left Periphery (when first position is otherwise occupied) to the left of the XP occupying first position (far left). Several OF scholars of the past few decades have proposed that a high degree of semantic integration of the subordinate to the main clause is the determinant of the former’s placement in the SpecFoc position. After outlining the syntactic diagnostics for determining the subordinate’s position, Donaldson uses these criteria to examine four categories of subordinates (temporal quant ‘when’; conditional se ‘if’; anterior ainz que/ ainçois que/ avant que/ devant ce que ‘before’; causal por ce que ‘because’) in 30 OF texts. He draws 6206 tokens that he codes according to type and subordinate placement. The majority of the examples are quant (3923) and se (2012), both of which are overwhelmingly far left peripheral, while the anterior (116) prefers SpecFoc and causal (155) is even handed between the two positions. More significant, however, is the fact that there is no apparent semantic criterion to account for the differences within a given category. Altogether there is no evidence in any of the subordinator categories to indicate a link between semantic function and syntactic placement. Donaldson suggests that the varied placement of these preposed subordinates may be attributable to the syntactic variability of OF, a language that was undergoing various syntactic changes during this period. Deborah Arteaga Arteaga’s chapter explores several aspects of complex sentences: obviation between matrix and subordinate subjects, sequence of tense in the two clauses, determination of mood, nature of subordinator particles, and interpretation of the mood. The point of departure is obviation—the obligatory disjoint reference of matrix and subordinate subjects—and its absence in Old French (OF) subjunctive clauses. Obviation is observed in modern Romance and many other languages, but lack of obviation— which often clusters with independent tenses in matrix and subordinate clauses—is noted in OF and Romanian. After an overview of OF grammar and important points of the minimalist framework adopted, the chapter reviews earlier work on obviation

8

J. Herschensohn

by Picallo, Kempchinsky, San Martín and Laskova, pointing out shortcomings in each. In this paper Arteaga extends her 2015 chapter in substantial ways that solve earlier unresolved problems and account quite elegantly for synchronic Old French and its diachronic changes to Modern French. She documents several syntactic phenomena that relate to OF subjunctive structures: morphological forms are distinct from indicative ones; subjunctive mood may appear in a matrix clause with no subordinator; there is no sequence of tense between matrix and subordinate clause; and unexpressed complementizers are possible in OF. She argues that OF manifested two subordinating particles, indicative que and subjunctive que, the latter acting as the “isolator” that permitted the relative independence (in terms of coreference of subjects and freedom of tense of verbs) of the subjunctive clause. She adopts San Martín’s irrealis feature in mood that accounts for binding given a special subjunctive particle (here subjunctive que), and Laskova’s ExclF feature, which is also linked to mood and to que. In diachronic terms, IRR changes from OF, where it is interpretable on Mood and values uIRR on CF, to Modern French where it is uninterpretable on Mood, valued by matrix v after the loss of subjunctive que. As for ExclF, it is a feature on T in OF, associated with subjunctive que, and isolating it from sequence of tense restrictions. It also permits subjunctive in matrix clauses in OF. These characteristics are lost with the development into MF. Cinzia Russi In her chapter on darla and darlo ‘to give it’, Russi addresses a grammatical gender distinction that reveals socio-cultural gender differences that seemingly characterize current issues of sexual harassment in both the Old and New Worlds. She traces the emergence of both forms over several centuries, noting the distribution and usages at different stages of Italian. Her ample documentation from original sources and from appropriate dictionaries provides the basis for her discussion of the factors responsible for the distinctions between the two gendered uses. The feminine la refers to ‘vagina’, while the masculine lo designates either ‘anus’ or ‘penis’, darlo1 and darlo2 respectively. The term darla is amply documented from the 16th Century on, and defined by numerous dictionaries as (of a woman) concedersi sessualmente ‘to concede oneself sexually’, whereas darlo2—attested only in the late 20th Century and mentioned only in specialized dictionaries—is defined as possedere sessualmente ‘to possess sexually’. The term darlo1, which refers to sodomistic sex, is infrequent and defined in only specialized dictionaries. The feminine darla is associated with both sexual licentiousness and lack of power, whereas the masculine darlo2 version epitomizes sexual prowess and power. Russi supplements the dictionary definitions with an informal survey that corroborates the familiarity of darla as compared to darlo in contemporary Italian. She points out how the very definitions of what is essentially the same lexical chunk ‘to give it’ reflect the sexual biases of the linguistic and socio-cultural environment, and she concludes “the moral discriminatory connotations these verbs inherently carry devalue women and embody a prejudiced stance toward gender and certain types of sexual practice.”

Introduction

9

This collection of chapters reinforces the importance of the Romance languages as an area of investigation that can provide new empirical data while advancing linguistic theory. The chapters cover a range of languages, theoretical issues, and methodological advances. Several chapters that flesh out the syntactic Left Periphery using the relatively recent cartographic approach are complemented by corpus based studies that use substantial data to investigate theoretical issues. Altogether these chapters both shed light on particular syntax and particular languages, and build our general understanding of the Left Periphery, its functional projections, features and syntax-semantics interface. On the whole, the articles cover a range of languages, theoretical issues, and methodological advances.

References Aboh, Enoch, and Jeannette C. Schaeffer (eds.). 2015. Romance languages and linguistic theory 2013: Selected papers from going Romance Amsterdam 2013. Amsterdam: Benjamins. Benincà, Paola. 2006. A detailed map of the left periphery of medieval Romance. In Crosslinguistic research in syntax and semantics: Negation, tense, and clausal architecture, ed. Zanuttini Raffaella, Héctor Campos Héctor, Elena Herburger, and Paul Portner, 53–86. Georgetown: Georgetown University Press. Belazi Hedi M, Rubin Edward J, and Toribio A. Jacqueline. 1994. Code switching and X-Bar theory: The Functional Head Constraint. Linguistic Inquiry 25(2):221–237. Bouchard, Denis. 1983. ECM is exceptional case marking. In Proceedings of the West Coast conference on Formal linguistics, ed. M. Barlow, D. Flickinger, and M. Wescoat, 11–18. Stanford: Stanford University. Bradley, Travis. 2014. Optimality theory and Spanish phonology. Language and Linguistics Compass 8: 65–88. Carrilho, Ernestina, Alexandra Fiéis, María Lobo, and Sandra Pereira (eds.). 2016. Romance languages and linguistic theory 10: Selected papers from going Romance 28. Amsterdam: Benjamins. Carstens, Vicki. 2000. Concord in minimalist theory. Linguistic Inquiry 31: 319–355. Colina, Sonia. 2009. Spanish truncation processes: The emergence of the unmarked. Linguistics 34: 1199–1218. Côté, Marie Hélène. 2013. Understanding cohesion in French liaison. Language Sciences 39: 156–166. D’Alessandro, Roberta, and Ian Roberts. 2008. Movement and agreement in Italian past participles and defective phases. Linguistic Inquiry 39: 477–491. Dobrovie-Sorin, Carmen, and Ion Giurgea. 2011. Pronominal possessors and feature uniqueness. Language 87: 126–157. Emonds, J. 1978. The verbal complex V′-V in French. Linguistic Inquiry 9: 151–175. Gess, Randall, and Deborah Arteaga (eds.). 2006. Historical Romance linguistics: Retrospectives and perspectives. Amsterdam: Benjamins. Harris, James W. 1983. Syllable structure and stress in Spanish: A nonlinear analysis. Cambridge, Mass: MIT Press. Jaeggli, Oswaldo. 1986. Three issues in the theory of clitics: Case, doubled NPs, and extraction. In Syntax and semantics 19: The syntax of pronominal clitics, ed. Hagit Borer, 15–42. Orlando: Academic Press. Jaeggli, Oswaldo, and Ken Safir. 1989. The Null subject parameter. Amsterdam: Springer. Kato, Mary Aizawa, and Esmeralda Vailati Negrão. 2000. Brazilian Portuguese and the null subject parameter. Iberoamericana 4: 55–73.

10

J. Herschensohn

Kayne, Richard. 1981. Binding, quantifiers, clitics, and control. In Binding and filtering, ed. Frank Heny, 191–211. London: Croom Helm. Kingston, John. 2008. Lenition. In Selected proceedings of the third conference on laboratory approaches to Spanish phonology, eds. Laura Colantoni and Jeffrey Steele, 1–31. Somerville, MA: Cascadilla Proceedings Project. Klausenburger, Jürgen. 1978. French linking phenomena: A natural generative analysis. Language 54: 21–40. Martins, Ana Maria, and Ernestina Carrilho (eds.) 2016. Manual de linguistica portuguesa (Manuals of Romance Linguistics) (Portuguese Edition). Berlin: De Gruyter. Meyer-Lübke, W. 1890–1902. Grammatik der Romanischen Sprachen. Leipzig: Fues’s Verlag R. Reisland. Montreuil, Jean-Pierre. 2002. Vestigal feet in French. In Proceedings of the 2002 Texas Linguistic Society conference on stress in optimality theory, ed. Augustine Agwuele, Willis Warren, and Sang-Hoon Park, 1–20. Pollock, Jean Yves. 1989. Verb movement, universal grammar, and the structure of IP. Linguistic Inquiry 20: 365–424. Poplack, Shana. 1989. The care and handling of a mega-corpus. In Language change and variation, ed. R. Fasold and D. Schiffrin, 411–451. Amsterdam: Benjamins. Pons-Moll, C. 2011. It is all downhill from here: A typological study of the role of syllable contact in Romance languages. Probus 23: 105–173. Raposo, Eduardo. 1987. Case theory and Infl-to-Comp: The inflected infinitive in European Portuguese. Linguistic Inquiry 18: 85–109. Raposo, Eduardo, and Juan Uriagereka. 1990. Long-distance case assignment. Linguistic Inquiry 21: 505–537. Rivero, Margarita. 1986. Parameters in the typology of clitics in Romance and Old Spanish. Language 62: 774–807. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed. Lilian Haegeman, 281–337. Dordrecht: Springer. Rizzi, Luigi, and Ian Roberts. 1989. Complex inversion in French. Probus 1: 1–30. Roberts, Ian. 2012. Phases, head movement, and second-position effects. In Phases, ed. Angel Gallego, 385–440. Berlin, Boston: De Gruyter. Rouveret, Alain. 2012. VP ellipsis, phases, and the syntax of morphology. Natural Language and Linguistic Theory 30: 897–963. Rouveret, Alain, and Jean-Roger Vergnaud. 1980. Specifying reference to the subject: French causatives and conditions on representations. Linguistic Inquiry 11: 97–202. Saltarelli, Mario. 1983. The mora unit in Italian phonology. Folia Linguistica 17: 7–24. Saltarelli, Mario. 1984. Italian syllable structure. Estudis Gramaticals 1: 279–295. Smith, Jason, and Tabea Ihanse. 2015. Romance linguistics 2012: Selected papers from the 42nd linguistic symposium on Romance languages (LSRL). Amsterdam: Benjamins. Tortora, Christina, Marcel den Dikken, Ignacio Montoya, and Teresa O’Neill (eds.). 2016. Romance linguistics 2013: Selected papers from the 43rd linguistic symposium on Romance languages (LSRL). Amsterdam: Benjamins. Tranel, Bernard. 1981. The treatment of French liaison: Descriptive, methodological, and theoretical implications. In Proceedings of the tenth anniversary symposium on Romance linguistics (Papers in Romance, Volume 3, Supplement II), ed. Heles Contreras and Jürgen Klausenburger. University of Washington: Seattle. Vance, Barbara. 1997. Syntactic change in Medieval French. Dordrecht: Kluwer.

Part I

Romance Syntax

Classifier Phrase as Host for Stage-Level Spanish Adnominal Participial Adjectives Karen Zagona and Heles Contreras

Abstract This paper argues for a Classifier Phrase (CLP) as a functional layer in the Spanish DP. CLP has been proposed independently for other languages as the locus of individuation in nominals, making countable entities from masses. It has also been suggested as a host for dimensional modifiers (Svenonius 2008). We provide evidence that certain participial adjectives in Spanish are licensed as dimensional modifiers. These adjectives are exceptional in that they encode verbal (inchoative) aspect. This feature is shown to play a crucial role in the distribution of these adjectives. Our account of this generalization adopts the aspectual feature [BOUNDARY] of Marín and McNally (2011). Our analysis is that this feature provides a (temporal) value for dimensionality, thereby exceptionally licensing a stage-level reading. We further show that an account along the lines of Larson (1998), Larson and Takahashi (2007), based on a distinction between NP-layer and DP-layer modification, is not viable for the data under consideration.










Keywords Classifier phrase Participial adjectives Adnominal adjectives Aspect Phrase Inchoative adjectives Perfective adjectives [BOUNDARY] feature Stage-level adjectives













1 Introduction This paper is concerned with the analysis of adnominal adjectives, and in particular with a class of Spanish predicative adjectives that exceptionally allows stage-level readings in both pre-nominal and post-nominal positions, as noted in Bosque (1999):

K. Zagona (&)
H. Contreras Department of Linguistics, University of Washington, Seattle, WA, USA e-mail: [email protected] H. Contreras e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_2

13

14

(1)

K. Zagona and H. Contreras

a. b.

el cliente contrariado the client annoyed el contrariado cliente the annoyed client

This freedom of order with accompanying stage-level interpretation runs counter to the typical pattern for Romance. It has been widely discussed in the literature that the canonical position for predicative adjectives is to the right of the noun, where they allow both individual-level (IL) and stage-level (SL) readings; pre-nominally, they are usually restricted to IL readings.1 This is illustrated by the contrasts between the (a) and (b) examples in (2)–(3):

(2)

a. b.

(3)

a. b.

este this este this

paisaje seco landscape dry seco paisaje dry landscape

(individual- or stage-level)

esa that esa that

casa húmeda house damp húmeda casa humid house

(individual- or stage-level)

(individual-level only)

(individual-level only)

The fact that there is a small class of exceptions to the general pattern raises questions as to why stage-level readings are generally not available pre-nominally, and how this specific class of adjuncts overrides the general restriction. With respect to the first issue, two factors have been cited as potentially bearing on differences in the interpretations of pre- and post-nominal adjectives. One is the structure of the modifier itself—whether it is a bare adjective phrase or projects additional structure, such as reduced relative or small clause structure. This is particularly relevant for SL readings, since clausal functional projections are plausible candidates to license temporal readings of adjectives (Demonte 1999, 2008; Cinque 2010). The second potential factor is the layer of nominal structure at which the adjective merges. This has the potential to affect the reading of the adjective if there are functional heads in the projections of the nominal that license particular readings, or, alternatively, if attachment at a higher level allows the adjective to escape interpretive constraints that hold lower in the structure. We have argued previously (Zagona and Contreras 2017) that the structure of the adjective has a central role in determining the availability or unavailability of SL

1

For general discussion of constituent order in Romance nominals see Demonte (2008), Cinque (2010), and Picallo (2012), among others.

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

15

readings for adjectives, and that participial adjectives like contrariado ‘annoyed’ in (1b) are exceptional in that the temporal (stage-level) reading of the adjective originates in the underlying verbal participle, as we show in Sect. 2 below. In the present study, we focus on the question of where these adjectives merge, and argue that the level of attachment is also relevant for licensing the SL reading. We argue that participial adjectives merge as specifiers of a Classifier head between the nP and the DP layers of nominal structure. Svenonius (2008) discusses classifier-type functional heads in nominals, a simplified representation of which is shown in (4):

D > CL > n ]

DP

One property of Classifier is that it makes countable entities from masses. Svenonius proposes that dimensional modifiers, which cannot modify masses, merge above this head. We argue that participial adjectives like (1b) also merge as specifiers of this functional head, by virtue of their featural compatibility with other dimensional adjectives, which measure such boundaries as height or width. At the same time, the deverbal structure of these adjectives differentiates them from two other classes of predicative adjectives that cannot have SL readings pre-nominally. The discussion is organized as follows. Section 2 summarizes three background assumptions that are relevant to the issues at hand. The first concerns the structure of nP and the availability of both adjunction and specifier positions for adnominal adjectives. Second, we show that the structure of predicative adjectives is relevant for the IL/SL distinction. Third, we outline our assumptions as to the structure of participial adjectives. We then examine two possible accounts of how the level of attachment constrains the availability of the stage-level interpretation. Section 3 takes up the approach of Larson (1998), Larson and Takahashi (2007), which accounts for exceptional stage-level readings of pre-nominal adjectives in English, as in (5a): (5)

a. b.

the nonvisible visible stars the temporarily invisible stars which are visible stars (such as Capella and Sirius) ??the visible nonvisible stars

In (5a), the leftmost (higher) adjective allows a stage-level reading, while the rightmost (lower) one has an individual-level reading.2 The assumption that the

2

The generalization is reinforced by the contrast shown below (due to B. Citko, cited in Larson 1998):

(i)

a. b.

The nonvisible visible stars include Capella. (Coherent) The visible nonvisible stars include Capella. (Odd)

16

K. Zagona and H. Contreras

relative height of the adjectives is relevant for interpretation accounts for the contrast between (5a) and (5b), since a star that is inherently visible can be temporarily non-visible, but the reverse is not possible. These authors account for the relevance of level of attachment on the basis of a distinction between D-related and N-related positions, where only the former allow the modifier to be licensed relative to the discourse context associated with the determiner. Section 3 discusses this approach and shows that it does not extend naturally to account for the distribution of SL readings for Spanish prenominal modifiers. Section 4 discusses the CL head, and argues for an analysis of participial adjectives as specifiers of CL.

2 Background: Adjective Phrases, Nominal Structure and Participial Adjectives 2.1

Non-uniformity of Pre-nominal Versus Post-nominal Modifiers

We assume a “non-uniform” approach to the structure of pre-nominal versus post-nominal adjectives. This implies that pre- and post-nominal adjectives differ in structure and in their manner of attachment to the nominal.3 Following Cinque (2010), we assume that post-nominal modifiers are small clauses or reduced relatives. Ticio (2010) argues for Spanish that post-nominal adjectives attach via Pair Merge (i.e., they are syntactically adjunct-like), while pre-nominal modifiers are projections of adjective phrases that attach via Set Merge (i.e., they are specifier-like). Ticio argues convincingly that the distinction between pre-nominal and post-nominal adjectives cannot be simply a matter of linearization; if it were, order should not have any influence on whether the modifier affects syntactic processes involving the nominal. Yet only pre-nominal adjectives block processes that are sensitive to the presence of specifiers, as Ticio shows. For example, prenominal adjectives block extraction from DP:

(6)

a. b. c.

3

Compramos varios famosos libros de lingüística. we bought several famous books on linguistics */??¿De qué compramos varios famosos libros? of what we bought several famous books ¿De qué compramos varios libros famosos? ‘What did we buy several famous books about?’

The uniform approach to adjective structure is developed in Demonte (2008) and Bouchard (2002), Cinque (2010) argues that adjectives may be either adjective phrases or reduced relative or small clause structures. Cinque also argues that certain adjectives are specifiers while others are indirect modifiers.

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

17

In (6b), extraction of the nominal complement is blocked in the presence of a pre-nominal adjective. The post-nominal adjective in (6c) has no corresponding effect. If the adjectives were distinguished only by their order (after spellout), this type of contrast would be unexpected. Ticio discusses similar effects for NP ellipsis and partial cliticization. Additional evidence for a non-uniform account comes from differences in ordering and potential stacking of adjectives pre-nominally versus post-nominally (Demonte 2008). Two post-nominal adjectives can generally appear in either order, as in (7):

(7)

a.

b.

Dame el sombrero rojo redondo. give-me the hat red round ‘Give me the round red hat.’ Dame el sombrero redondo rojo. give-me the hat round red ‘Give me the red round hat.’

Pre-nominal adjectives generally have limited co-occurrence possibilities. For example, counterparts to (7) with pre-nominal adjectives are impossible, as shown in (8), (9):

(8)

a. b.

(9)

a. b.

*la delgada alta señora the thin tall woman *la alta delgada señora the tall thin woman

*la the *la the

roja olorosa rosa red fragrant rose olorosa roja rosa fragrant red rose

Demonte (1999) showed that a second pre-nominal adjective is possible only if the left-most is a modal or speaker-oriented adjective; again, the order is rigid:

(10)

a. b.

la maravillosa larga jornada the wonderful long day *la larga maravillosa jornada the long wonderful day

18

K. Zagona and H. Contreras

These differences between pre- and post-nominal distribution are expected on a non-uniform analysis. Analyzing post-nominal adjectives as adjoined, reduced relative/small clauses modifiers accounts for their canonical properties. Their restrictive interpretation, with either individual-level or stage-level readings, follows from their internal structure; their multiple co-occurrences with non-rigid ordering follow from their attachment as adjuncts. Given these considerations, we assume (11) for structure of the nominal: (11)

nP n’ n

AGENT NP

NP Prenominal AdjP

ADJUNCTS (Small clause/reduced Relatives) N’ OBJ/PPARG

The clausal analysis of post-nominal adjectives is consistent with their restrictive interpretation, and with the availability of both individual-level and stage-level interpretations. Both interpretations are expected in clausal adjuncts, since clausal functional categories (particularly TenseP and AspectP) license stage-level readings of adjectives in copular sentences. The assumption that reduced relative clauses are adjoined to NP also accounts for their non-rigid order and their potential for stacking. The analysis of pre-nominal adjectives as specifiers accounts for their limited co-occurrences and rigid order, assuming that the nominal level of structure (NP + nP) has only one specifier, and that additional adjectives attach to projections above it. The typical absence of stage-level interpretations for pre-nominal adjectives can be accounted for by assuming that adjectives have individual-level interpretations unless a stage-level interpretation is specifically licensed by functional categories external to the adjective that introduce stages and license their temporal reference. We provide support for this assumption in Sect. 2.2 below.

2.2

Structure of Adjective Phrases with Stage-Level Readings

We assume that adjectives have stage-level readings only in the context of functional categories that introduce and license them; otherwise, they have only individual-level readings. This is motivated for Spanish by several recent studies

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

19

that have argued that stage-level adjectives project a prepositional head above the adjective (Gallego and Uriagereka 2009, 2011; Brucart 2012; Zagona 2012, 2015; Camacho 2012):

(12)

a. b.

[ adj ] [ P [ adj ]]

(individual-level) (potential for stage-level reading)

Although details of the proposals differ, the general line of approach taken by these authors is that the prepositional head is a predicate of spatio-temporal location. It is the category that is selected by copular estar ‘be’. This analysis accounts for c-selection by estar and for the range of locational readings that occur with estar. Traditional approaches to the ser/estar alternation have generally suggested that the copular verbs are themselves a spell-out of stage-level or individual-level readings of adjectives. But the correlation between the copula and the availability of stage-level interpretation is inexact at best, with exceptions found in both directions. For example, ser may co-occur with transitory predicates of several types, including nominals whose subjects are eventive, as in (13), and with nominal predicates of transitory states:

(13)

La fiesta es/*está en mi casa. the party is at my house

(14)

Obama es/*está presidente desde 2009. ‘Obama is president since 2009.’

(15)

Juan es alto ahora, pero de pequeño era bajito. ‘Juan is tall now, but as a child he was short’

(16)

Juan ‘Juan

es/*está comunista. is a Communist.’

Conversely, although estar co-occurs typically with stage-level predicates, it also occurs with states that are thought of as permanent, as in (17), and with PPs of location with unmoving subjects, as in (18):

(17)

Su hermano está/*es muerto ‘dead’ his brother is dead

20

K. Zagona and H. Contreras

(18)

El edificio está en el centro de la ciudad. the building is in the center of the city

Assuming the structural distinction shown in (12) above, it is possible to account both for the broad patterns and for exceptions like (13)–(18) in terms of c-selection: estar selects prepositional complements, including PPs of spatial location, and adjective phrases headed by P, as in (12b), while ser selects other categories. The prepositional structure projected by stage-level adjectives can account for their restriction to post-nominal positions, since PP adjuncts in specifier positions are generally disallowed:

(19)

a.

*un de Madrid estudiante a from Madrid student un estudiante de Madrid a student from Madrid

b.

2.3

Structure of Participial Adjectives

The participial adjectives that allow stage-level readings pre-nominally are deverbal adjectives that are formed from intransitive se-reflexive psychological verbs (Marín and McNally 2011) such as enfadado ‘angered’, asustado ‘frightened’, asombrado ‘amazed’, ofendido ‘offended’, confundido ‘confused’ avergonzado ‘ashamed’ molestado ‘bothered’, preocupado ‘worried’, and many others.4 We assume a derivational process in which the adjective asombrado ‘amazed’ is constructed as in (20). A verbal root merges with Aspect, which in turn merges with an adjectival head that bears phi-features:

(20)

aP -a/-o

AspP Asp ad-

4

vP asombr-

Marín and McNally (2011) show that the predicates from which these adjectives are formed fall into two subclasses, punctual and non-punctual. This distinction does not seem to be relevant for the present discussion.

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

21

Zagona and Contreras (2017) (Z & C) argue that the Aspect head plays a crucial role in establishing the adjective’s interpretation and its syntactic distribution. This accounts for two properties of deverbal participial adjectives: (1) they have more restricted distribution and interpretations than typical basic adjectives, and (2), aspectually distinct subclasses of deverbal participial adjectives have different syntactic distribution. Let us examine the first generalization. Unlike typical predicative adjectives, they have only stage-level readings, and in copular sentences they co-occur only with copular estar, as is shown in (21), and summarized in (22):

(21)

a. b.

(22)

María está/*es asustada. (Stage-level only) ‘María isESTAR/isSER frightened.’ El proyecto está/*es terminado. ‘The project is ESTAR/isSER finished.’ (Stage-level only)

Properties of deverbal participial adjectives a. stage-level readings only b. co-occur only with copular estar

These properties follow from an analysis on which these adjectives contain vPs that project Aspect Phrases, since Aspect is a prepositional head of temporal location (Demirdache and Uribe-Etxebarria 2000). This accounts directly for their stage-level reading. Their obligatory co-occurrence with estar does not follow automatically, since the structure in (20) is similar to the bare adjective phrases that co-occur with copular ser. However, Z & C suggest that this co-occurrence restriction can be explained in terms of the relational nature of locational predicates. That is, the temporal location that they refer to is not absolute, but is instead established in relation to an evaluation time. Grammatical aspect, such as progressive, for example, can be described in terms of a relation between the event time and a reference time (Smith 1997; Demirdache and Uribe-Etxebarria 2000). Z & C implement this generalization as a [uLoc] feature of the Aspect head (and locational predicates more generally). The effect of this feature is that the Aspect Phrase must appear in the domain of a functional head that can check (value) its [uLoc] feature. Copular estar is such a head, but ser is not. As noted above, PPs of spatial location always co-occur with estar rather than ser. Consider now the second generalization: that aspectually distinct subclasses of participial adjectives differ in their syntactic distribution. Z & C compare two sub-classes of participial adjectives: (i) se-reflexive intransitives and (ii) perfective participial adjectives such as cumplido ‘completed’, abandonado ‘abandoned’, terminado ‘finished’. These subclasses differ in their nominal distribution. Only the se-reflexive intransitives can occur pre-nominally; the perfective participial adjectives cannot:

22

(23)

K. Zagona and H. Contreras

a. b. c. d.

una tarea cumplida a task completed *una cumplida tarea a completed task las casas vendidas the houses sold *las vendidas casas the sold houses

Z & C argue that this follows from a difference in the structure of the participles that underlie the two types of participial adjectives. The difference centers on whether the Aspect head is a ‘high’ or ‘low’ Aspect. High Aspect is above vP, while low Aspect is below v. These correspond to Travis’ (2010) distinction between Outer Aspect and Inner Aspect:

(24)

[AspP ASP(HIGH) [ v [AspP ASP(LOW) [VP V … ]]]]

We assume with Travis that low aspect (Inner Aspect) is the head that is relevant for licensing a telic reading of the VP (the verbal root and its internal argument). That is, it licenses a definite endpoint (telos) of the change of stage of the internal argument. We propose correspondingly that Outer Aspect (or high aspect) licenses an endpoint reading of the event onset, deriving an inchoative interpretation. This is supported by the fact that only the perfective participial adjectives have properties that are associated with telicity, while the se-reflexive intransitives do not (Marín and McNally 2011). Here we offer only a brief sketch of the distinction, as reflected by readings of in-adverbials with the verbs that underlie the two subclasses5:

(25)

Se {aburrió/divirtió} en un instante/en cinco minutos. SE {bored/amused} ‘It took an instant/five minutes for him to get bored/amused.’

5

As noted above (Note 4), there are two subclasses of se-reflexive inchoatives that differ in durativity. The punctual ones disallow in-adverbials:

(i)

*Se asustó/ enfadó en una hora/ en un instante. SE (frightened/angered) in an hour/ in an instant. ‘He/she got frightened/angered in an hour/in an instant.’

This is expected, since these adverbials add specification of change from one time to another, which requires at least minimal duration.

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

(26)

23

Escribió una novela en dos meses. ‘S/he wrote a novel in two months.’

In (25), the in-adverbial modifying se-reflexive intransitives specifies a time interval that leads up to the onset of the state, as shown in (27):

(27) ______________ [ONSET being bored/amused ]_______________

In (26), the in-adverb specifies a time leading up to the endpoint of the event, as in (28):

(28) ______________ [ONSET write a novel TELOS]_______________

The perfective and inchoative classes of participles then differ in the position of Aspect relative to v: for perfective participles, Aspect is below v, for se-reflexive inchoatives, it is above v:

(29)

a. b.

[ a [ v [ AspLOW V … ]]] [ a [ AspHIGH [ v V … ]]]

(Perfective Participles) (Inchoative Participles)

Z & C propose that the High Aspect head in (29b) is sufficiently high as to be accessible for s-selection or c-selection, while the Low Aspect head in (29a) is not. The inaccessibility of Low Aspect follows automatically from the assumption that v is a phase head, and features lower in the structure of the adjective cannot enter into any relations with functional heads above v.

2.4

Summary

This section has set out several assumptions that we adopt below in approaching the issue of licensing the exceptional stage-level interpretation of pre-nominal inchoative participial adjectives. We have adopted a non-uniform analysis of pre-nominal and post-nominal adjectives; pre-nominal ones merge in specifier positions while post-nominal ones have small clause or reduced relative structure, and are adjoined. This supports the assumption that post-nominal adjectives allow

24

K. Zagona and H. Contreras

stage-level readings freely because that reading is licensed internal to the modifier by clausal functional heads that license the temporal reference of the predicative adjective. By contrast, pre-nominal predicative adjectives lack these clausal functional categories, and therefore require modifier-external licensing of their stage-level reading. Evidence for this assumption was given in Sect. 2.2, based on the distribution and interpretation of P-headed versus bare adjective phrases with respect to copula selection. The contrast between the two types of participial adjectives discussed in Sect. 2.3 above further supports this assumption, since the high or low position of the Aspect Phrase within the underlying vP contributes to the reading of the event in specific ways, and therefore may be relevant for determining the availability of pre-nominal attachment. The remaining question, and the one to which we now turn, is: what precisely is the mechanism of external licensing of the stage-level reading; why is it available for inchoative adjectives but not for perfective ones?

3 NP-Internal Versus NP-External Modification We turn now to a possible analysis of the stage-level reading of participial adjectives as ‘Outer’ modifiers (Larson 1998; Larson and Takahashi 2007). Their distinction between ‘Inner’ and ‘Outer’ modifiers, proposed for certain contrasts in English, corresponds to modification within NP versus outside it.

3.1

Stage-Level Readings as NP-External Modifiers

Larson (1998) discusses Bolinger’s (1967) observation that there is a correlation between an adjective’s interpretation and its order relative to the nominal, as in (30), (31):

(30)

a. b.

the stars visible the visible stars

(temporary) (enduring)

(31)

a. b.

the person responsible the responsible person

(temporary) (enduring)

Larson shows that a pre-nominal adjective can in fact have a stage-level reading when it appears more distant from the nominal. This can be seen where two instances of the same adjective are juxtaposed pre-nominally, as in (32):

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

(32)

The

visible TEMPORARY

visible

25

stars include Capella.

ENDURING

Larson (1998) and Larson and Takahashi (2007) also show that there are additional effects of relative proximity to the noun. One is that the contrast between intersective and subsective interpretation disappears when the adjective is more distant from the noun. The prenominal adjective beautiful is ambiguous between intersective and subsective readings in (33), but the second, more distant adjective in (34) is not:

(33)

Olga is a beautiful dancer a. Olga is beautiful as a dancer (subsective) b. Olga is a beautiful person and a dancer (intersective)

(34)

Olga is a

beautiful

beautiful

INTERSECTIVE

SUBSECTIVE

dancer

On the subsective reading, it is Olga’s dancing that is beautiful; on the intersective reading, Olga has two properties, being a dancer and being beautiful. Only the intersective reading is possible when the adjective is not proximate to the noun. A third effect of proximity is a contrast between generic and deictic readings of time modifiers. In pre-nominal position, these are ambiguous, as in (35); however, the generic reading is lost when the modifier is not proximate to the noun:

(35)

a. b.

the Thursday lecture (deictic or generic) the Wednesday Thursday lecture

In (35a) the time modifier can refer either to lectures that characteristically take place on Thursdays, or to a single (deictic) Thursday lecture. In (35b) the two readings can be juxtaposed, referring to a lecture that is characteristically held on Thursdays, but is held instead on a Wednesday. These contrasts are summarized as properties of Inner and Outer modifiers:

(36)

A. Outer modifiers: temporary, deictic, intersective B. Inner modifiers: enduring, generic, non-intersective

There are two components of Larson’s account of these contrasts. The first is that nominals as well as verbs can contain a Davidsonian event argument (Davidson 1967). This implies that, dancer in (33) for example, is a relation between an individual who dances and an event of dancing. The contrast between the subsective and intersective readings in (33), (34), follows from analyzing beautiful as

26

K. Zagona and H. Contreras

predicated either of Olga or of the event of dancing. Second, Larson draws on Chierchia’s (1995) observation of a parallel between certain properties of predicate nominals and of generic sentences. Larson suggests that these properties are due to a generic quantifier C, and that the nominal expression of C occurs above NP and below DP. The contrasts between Inner and Outer modifiers follow from their position relative to this quantifier (Larson and Takahashi 2007):

(37)

[ DP D β [NP [Γe α N ]] β ]

(α = NP modifier; β = DP modifier)

Adjectives that attach below the generic quantifier will have generic, enduring, subsective readings; those that attach above C escape its effects, and have potential for temporary, deictic, intersective readings. To account for the source of intersective, temporary readings of Outer modifiers, Larson relates their interpretations to those of relative clauses, which are assumed to originate in a low position, as the closest complements of the determiner. Both of these constituents then raise:

(38)

[DP the [ visible [ [NP visible [NP stars

[

t

]

Summarizing, this account provides a potential explanation for the stage-level reading of participial adjectives, by analyzing the nominal that they modify as containing an event argument, and by raising the adjective to a position outside NP, where the event can be interpreted as a temporary state, because it escapes the effects of the generic quantifier.

3.2

Extending the Approach to Spanish

We consider now a derivation parallel to (38) for participial adjectives in Spanish. The derivation would be as in (39), where the prenominal modifier originates as a reduced relative:

(39)

[DP el [ contrariado the

annoyed

[NP cliente [ t client

t ]

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

27

There are several problems for this account, however, which show that participial adjectives neither originate as restrictive modifiers, nor surface as leftward Outer modifiers. First, participial adjectives do not have restrictive readings, as do the English Outer modifiers. They can modify proper nouns, as shown in (40):

(40)

a. b.

la aburrida María the bored María el preocupado José the worried José

Prenominal adjectives allow restrictive readings only if the adjective has contrastive intonation, as noted by Demonte (2008):

(41)

Las FÉRTILES verdes praderas de Irlanda lo deslumbraron. the FERTILE green meadows of Ireland him astonished (fértiles = restrictive)

Participial adjectives are awkward with contrastive intonation:

(42)

??Encontramos a los ASUSTADOS pacientes (We) met the FRIGHTENED patients.

More generally, the English partition between prenominal Inner and Outer modifiers (illustrated in (30) and (32) above) is a post-nominal phenomenon in Spanish:

(43)

a. b.

las estrellas visible invisibles the stars visible invisible ‘The invisible visible stars’ ??las estrellas invisibles visibles

This is shown further by the fact that pre-nominal participial adjectives are incompatible with generic sentences:

(44)

a. b.

Un agente enfadado no es productivo. an agent angry not is productivo ‘An agent who is an angry person is not productive’ *Un enfadado agente no es productivo. ‘An agent who is in an angry state is not productive.’

28

K. Zagona and H. Contreras

(45)

a. b.

Un soldado asustado no es eficaz. ‘A soldier who is a frightened person is not effective.’ *Un asustado soldado no es eficaz. ‘A soldier who is in an angry state is not effective.’

These contrasts indicate that Spanish does not prepose restrictive modifiers to a position above NP. It is notable that one other difference between Inner and Outer modifiers is nevertheless observed. Pre-nominal adjectives that modify a derived nominal such as bailarina ‘dancer’ are ambiguous between subsective and intersective readings—that is, between event modification and individual modification. However, they are not ambiguous between individual-level and stage-level readings6:

(46)

a. b.

Olga es una bailarina linda. Olga is a dancer beautiful i. Olga is beautiful and Olga is a dancer Olga es una linda bailarina. Olga is a beautiful dancer i. Olga is beautiful as a dancer. ii. Olga is beautiful and is a dancer.

(intersective) (subsective) (intersective)

Summarizing, a preposing analysis of pre-nominal participial adjectives is problematic, because there is no prenominal cluster of Outer modifier properties that might motivate such a movement, and it would not adequately characterize the properties of participial adjectives, which are non-restrictive and uniformly stage-level in interpretation. Non-participial adjectives like visible ‘visible’ cannot move to prenominal position, unless they are contrastively focused, an analysis that likewise does not apply to participial adjectives.

4 Adjective Order and Classifiers in the Projections of Nominals We present now an alternative account, on which participial adjectives merge as specifiers of a functional layer between DP and NP (or nP) that we will refer to as a Classifier (CL). The semantic contribution of this element is to derive countable

6

This ambiguity does not extend to non-derived nominals:

(i)

a.

Pepe es un buen abogado. P. is a good lawyer i. Pepe is good as a lawyer. ii. *Pepe is a good person and a lawyer

(subsective) (intersective)

This suggests that only some nominals select event arguments.

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

29

reference from the mass-like properties of NP (Svenonius 2008, among others). Svenonius argues that there are two separate Classifier positions above nP, Sortal and Unit heads7:

(47)

D [ UNIT [ SORT nP ]

The lower head sorts nominal referents by characteristics such as shape; the UNIT classifier converts them countable or quantifiable entities. Svenonius notes that, although these two combine as a single lexical head in many languages, they can sometimes appear separately. In the discussion below, we will abbreviate UNIT and SORT as one position: CL, although the features and perhaps the positions are underlyingly distinct. One of the principal issues addressed in Svenonius (2008) is the nature of ordering restrictions on adnominal adjectives.8 Is that certain generalizations about adjective order, as well as some position-dependent interpretive properties can be explained in terms of the type of modification that is possible above and below the Classifier heads. Below CL, adjectives modify masses; these include adjectives of color, origin and material; they are intersective and restrictive in interpretation. Above CL, adjectives modify countable objects; they specify their physical properties, including size, shape, and other dimensional characteristics. Since participial adjectives modify count nouns and not masses, they may be analyzed as specifiers of CL:

(48)

DP D

CL

Dimensional AdjP CL

CL nP

We argue below in Sect. 4.1 that participial adjectives merge as specifiers of CL, as in (48). In Sect. 4.2 we offer some additional evidence for the CL position for Spanish.

7

A third type of classifier is the noun class marker, which differentiates qualities or essences of the root. As these are more closely related to the root, they are nP-internal. 8 Restrictions on adjective order are illustrated for English in (i):

(i)

a. b.

a big red car *a red big car

30

4.1

K. Zagona and H. Contreras

Specifier of CLP as Host for Participial Adjectives

Distributional evidence shows that participial adjectives occupy the same position as other dimensional adjectives:

(49)

a. b.

*el the *el the

alto enfadado tall angered enfadado alto angered tall

cliente customer cliente customer

In (49), the sequence of a participial adjective alongside a dimensional adjective is ill-formed; this suggests that the two compete for the same position. This contrasts with (50), where the participial adjective co-occurs with an evaluative adjective:

(50)

a. b.

el the *el the the

pobre enfadado poor angered enfadado pobre angered poor angered client

cliente customer cliente customer

Given this initial motivation for analyzing inchoative participial adjectives as specifiers of the classifier phrase, we turn to the question of which subclasses of adjectives can merge as specifiers of the CL phrase, and what determines the availability of stage-level or individual-level readings. Recall that this contrast correlates with the morphology of participial versus basic adjectives. The former have stage-level readings in pre-nominal positions, but the latter do not:

(51)

a. b.

este seco paisaje this dry landscape el enfadado cliente the angered client

(individual-level only) (stage-level only)

The intuition that we will pursue here is that there are two separate factors that govern the relationship between dimensional adjectives and the nominal; these two factors determine (a) whether a given adjective can function as a dimensional specifier of CL; and (b) whether the adjective can be interpreted as temporal (stage-level) or not (individual-level). With respect to the first issue, we propose that the property that allows a predicative adjective to function as a dimensional modifier is a [BOUNDARY] feature. This feature, which has been proposed as a primitive feature of events in Marín and McNally (2011), refers to the moment of onset or endpoint of the event.

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

31

We analyze this feature as encoded on Aspect. Extending this to other contexts, we suggest that a boundary feature can be encoded on other functional heads. In particular, we suggest that CL also bears a boundary feature, and in this context it corresponds to the boundaries that individuate countable nouns (ones with “edges”) from masses. Dimensional adjectives also encode boundaries as part of their representation, in that they specify properties of the boundaries of objects. An adjective like tall specifies a property of the boundaries or outlines of an object or person along the dimension of height. An adjective is then expected to occur as a specifier of CL if it can be interpreted as specifying or measuring the boundaries of the nominal along some dimension. Since these adjectives are gradable, we suggest that the boundary feature is represented on the Degree Phrase, rather than being an inherent feature of adjectival roots. On these assumptions, both the individual-level adjective seco ‘dry’ in (51a) and the stage-level adjective enfadado ‘angered’ in (51b) occupy the same prenominal position: specifier of CL Phrase. The two differ only in the particular dimension that is specified by the adjective. The basic adjective seco specifies only the dimension of dryness, since there is no adjective-internal source for a temporal-dimension interpretation. Participial adjectives are exceptional in that their boundary feature originates with the Aspect Phrase of the verbal constituent from which the adjective derives. The contrast between the two is illustrated in (52):

(52)

a. Inchoatives participial adj:

b. Basic adjectives

adjP adj

adjP AspP

Asp -ad[BOUNDARY]

adj

Root

vP asombr-

The locus of the boundary feature in (52a) implies that it is specifically event-related. When a boundary feature appears in Aspect Phrase, we assume that it is inherently interpreted as an event boundary. So far we have proposed that dimensional adjectives and participial adjectives are compatible with CL Phrase modification if they have a [BOUNDARY] feature, such that they are interpreted as specifying the boundaries of the nominal along some dimension. This accounts for the fact that predicative adjectives are gradable, since the Degree Phrase is the position in which the boundary feature is normally expressed. It accounts for the contrast between basic adjectives and participial adjectives with respect to individual-level versus stage-level interpretations, as discussed above: only the participial adjectives have an event-related, or temporal dimension. There is a second factor that is relevant for licensing of dimensional adjectives. We have proposed above that the presence of a [BOUNDARY] feature on Degree allows the adjective to function as a dimensional modifier. This feature links it to the modified noun, whose boundaries are specified. We suggest further that the

32

K. Zagona and H. Contreras

Degree head is specified for the category of dimension that is contributed by the modifier. In other words, the Degree head inherits from its complement, perhaps via Agree, a feature that determines whether the dimension is a quality or property (adjectival) or a temporal dimension (aspect). This inheritance relationship is shown in (53): (53)

a.

DegP Deg [uDimension]

b. adjP

adj

DegP Deg [uDimension] adj

AspP Asp -ad-

adjP Root

vP asombr-

[BOUNDARY]

In (53a), the Degree Phrase modifier is interpreted as modifying the nominal along a temporal dimension. The Degree Phrase inherits its dimensional specification from the adjective, which in turn inherits a temporal value from Aspect Phrase. In (53b) the Degree Phrase inherits its dimensional value from the adjective, which has no temporal value. If an analysis along these lines is correct, it could account for the contrast between Inchoative and Perfective Participial adjectives. Recall from Sect. 2.3 above that the two types of participial adjective differ with respect to the position in which Aspect appears: Inchoatives have High Aspect, which is situated above vP, while Perfectives have low Aspect, which is situated below vP, and Voice, as shown in (29), repeated below:

(29)

a. b.

[ a [ v [ AspLOW V … ]]] [ a [ AspHIGH [ v V … ]]]

(Perfective Participles) (Inchoative Participles)

Given this difference in structure, it is expected that only High Aspect is accessible to an Agree relation with a higher Degree head. Low Aspect is not, because it is lower than Voice, which marks a phase boundary. Summarizing, we have argued for an analysis of inchoative participial adjectives as dimensional modifiers. This accounts for their complementary distribution with dimensional modifiers, as well as for their ability to modify only count nouns, and for their non-restrictive interpretation, which follows from merge outside nP. We have proposed two features that are relevant for the licensing of adjectives in this position. One is [BOUNDARY], which links adjective to the count noun via s-selection. The other is [uDimension], valued by Agree, which determines the type of dimension that the Degree Phrase specifies for the nominal.

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

4.2

33

CL Phrase in Spanish

We briefly consider here some Spanish-internal evidence for assuming that the Classifier Phrase is syntactically present in Spanish. Recall that Larson (1998) proposes that nominals can represent a Davidsonian event argument, accounting for the ambiguity of nominals like (46b) repeated as (54):

(54)

Olga Olga i. ii.

es una linda bailarina is a beautiful dancer Olga is beautiful as a dancer Olga is beautiful and is a dancer

(subsective) (intersective)

The subsective reading of (54) derives from predicating the adjective of the event argument. We suggest that the Classifier Phrase is the level of structure at which the event argument of the nominal is encoded. Recall that the Cl Phrase consists of two types of features: Sortal features, which differentiate subclasses of nominals, and Unit features, which make the nominal countable. These two features seem to be implicated in the interpretation of nouns like bailarina, which is a count noun, and is a type of noun that inherently refers to a relation between individual and event, in Larson’s sense. We suggest that nominals of this subtype are differentiated sortally at the Classifier Phrase level. The structure that gives rise to the two interpretations of (54) is then (55):

(55)

UnitP AdjP

Unit’

(subsective) UNIT

SortP Sort’

ArgEvent SORT

nP AdjP

n’

(intersective) … n …

34

K. Zagona and H. Contreras

In (55), a feature of the SORT head determines whether the nominal is the type that have an event argument, in a manner similar to the flavors of v that determine characteristics of its specifier. A modifier that merges below the SORT Phrase is then expected to be intersective and to modify the individual; a modifier that merges above the SORT Phase is then expected to allow a subsective interpretation. By linking this interpretation to the Classifier layer, it is possible to account for the correlation between specific nominal affixes and the presence of a subsective reading. Nominals with agentive affixes either optionally or obligatorily impose a subsective interpretation on their prenominal predicative modifiers. Derived agentive nominals with affixes like –ero, -dor, impose the subsective reading obligatorily, while other, non-derived nominals are ambiguous:

(56)

un buen colaborador a good collaborator

(subsective only)

(57)

un buen amigo a good friend

(subsective/intersective)

The syntactic analysis shown in (55) accounts for the two interpretations of nominals like amigo ‘friend’ in (57); the non-ambiguity of agentive nominals suggests that the affix itself may be a SORT morpheme, or that an nP with these affixes is always s-selected by a SORT Phrase of the type in (55), with an Event argument as its covert specifier. Notice that these subsective modifiers do not have a temporary reading. This is explained on present assumptions by the fact that the event argument is below the UNIT head. The Classifier phrase identifies countable reference for the individual, not its event argument. Another type of nominal is potentially analyzed with the structure (55). These differ in that the entire nominal refers to a temporary activity rather than to a generic kind of activity:

(58)

a.

b.

Juan está de camarero. J. isESTAR of waiter ‘Juan is working as a waiter.’ María está de niñera. M. isESTAR of babysitter ‘María is working as a babysitter.’

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

35

These nominals occur obligatorily with estar, and refer to a temporary activity of the subject, of the type specified by the de-phrase. In these nominals, the SORT Phrase can perhaps be characterized as Location. The de-phrase identifies the SORT head, or potentially is itself a manifestation of it.

5 Conclusion We have argued above that Spanish DP contains a layer between the DP and nP layers, referred to here as Classifier Phrase. This phrase allows dimensional adjectives as its specifier, and it was argued that stage-level participial adjectives also merge in the same specifier position. This analysis accounts for differences between the properties of participial adjectives and the stage-level readings of pre-nominal adjectives in English discussed in Sect. 3. We provided some very preliminary support for the Classifier Phrase above, although the question remains open as to the possible morphological realizations of this head and its syntactic properties. In recent work on Bangla, Simpson and Syed (2016) argue that Classifier Phrase (or more precisely, the Quantifier Phrase above it) functions as an escape hatch for certain movements of constituents to higher DP-internal positions, which suggests that this layer of structure is a phase boundary. The potential role of Classifier Phrase in accounting for DP-internal constituent order, and extraction phenomena of the type discussed by Ticio, are topics that call for further examination in future research. Acknowledgements We dedicate this to our friend and colleague Julia Herschensohn, who has for decades been an integral part of our life in linguistics, and to whom much is owed for her intellectual, cultural and culinary nourishment of UW students and colleagues alike. Gracias and merci, Julia. We are grateful to Barbara Citko and to members of the UW Syntax Roundtable for comments on a presentation of a preliminary version of this work. We also thank an anonymous reviewer for helpful comments. All remaining errors of commission and omission are the authors’ sole responsibility.

References Bolinger, Dwight. 1967. Adjectives in English: Attribution and predication. Lingua 18: 1–34. Bosque, Ignacio. 1999. El sintagma adjetival. In Gramática descriptiva de la lengua española, vol. 1, ed. Ignacio Bosque and Violeta Demonte, 217–310. Madrid: Espasa. Bouchard, Denis. 2002. Adjectives, numbers and interfaces. Amsterdam: North Holland. Brucart, José María. 2012. Copular alternation in Spanish and Catalan attributive sentences. Revista de Estudos Lingusticos da Universidade do Porto 7: 9–43. Camacho, José. 2012. Ser and Estar: The individual/stage-level distinction and aspectual predication. In The handbook of Hispanic linguistics, ed. José Ignacio Hualde, Antxon Olarrea, and Erin O’Rourke, 453–475. Malden, MA: Wiley-Blackwell.

36

K. Zagona and H. Contreras

Chierchia, Gennaro. 1995. Individual-level predicates as inherent generics. In The generic book, ed. Greg N. Carlson and Francis Jeffry Pelletier, 176–223. Chicago: University of Chicago Press. Cinque, Guglielmo. 2010. The syntax of adjectives: A comparative study. Cambridge, MA: MIT Press. Davidson, Donald. 1967. The logical form of action sentences. In The logic of decision and action, ed. Nicholas Rescher, 81–95. Pittsburgh: University of Pittsburgh Press. Demirdache, Hamida, and Myriam Uribe-Etxebarria. 2000. The primitives of temporal relations. In Step by step: Essays on minimalist syntax in honor of Howard Lasnik, ed. Roger Martin, David Michaels, and Juan Uriagereka, 157–186. Cambridge, Mass: MIT Press. Demonte, Violeta. 1999. A minimal account of Spanish adjective position and interpretation. In Grammatical analyses in Basque and Romance linguistics, ed. Jon Franco, Alazne Landa, and Juan Martín, 45–75. Amsterdam: John Benjamins. Demonte, Violeta. 2008. Meaning-form correlations and adjective position in Spanish. In Adjectives and adverbs: Syntax, semantics and discourse, vol. 19, ed. Louise McNally and Christopher Kennedy, 71–100. Oxford Studies in Theoretical Linguistics. Oxford: Oxford University Press, UK. Gallego, Ángel, and Juan Uriagereka. 2009. Estar = Ser + P. Paper presented at the XIX colloquium on generative grammar. Universidad del País Vasco, Vitoria-Gasteiz, Spain, 1–3 April 2009. Gallego, Ángel, and Juan Uriagereka. 2011. The lexical syntax of ser and estar. Unpublished manuscript, Universitat Autónoma de Barcelona and University of Maryland. Larson, Richard K. 1998. Events and modification in nominals. In Proceedings of semantics and linguistic theory (SALT), vol. 12, ed. Devon Strolovitch and Aaron Lawson, 145–168. Ithaca NY: CLC Publications. Larson, Richard K., and Naoko Takahashi. 2007. Order and interpretation in prenominal relative clauses. In Proceedings of the workshop on Altaic formal linguistics II. MIT working papers in linguistics, vol. 54, ed. Meltem Kelepir and Balkiz Öztürk, 101–120. Cambridge, MA: MITWPL. Marín, Rafael, and Louise McNally. 2011. Inchoativity, change of state, and telicity: Evidence from Spanish reflexive psychological verbs. Natural Language & Linguistic Theory 29 (2): 467–502. Picallo, M. Carmen. 2012. Structure of the noun phrase. In The handbook of Hispanic linguistics, ed. José Ignacio Hualde, Antxon Olarrea, and Erin O’Rourke, 263–283. Malden, MA: Wiley-Blackwell. Simpson, Andrew, and Saurov Syed. 2016. Blocking effects of higher numerals in Bangla: A phase-based analysis. Linguistic Inquiry 47 (4): 754–763. Smith, Carlota. 1992/1997. The parameter of aspect. Dordrecht: Kluwer. Svenonius, Peter. 2008. The position of adjectives and other phrasal modifiers in the decomposition of DP. In Adjectives and adverbs: Syntax, semantics and discourse, ed. Louise McNally and Christopher Kennedy, 16–64. Oxford Studies in Theoretical Linguistics. Oxford: Oxford University Press, UK. Ticio, M. Emma. 2010. Locality domains in the Spanish determiner phrase. Dordrecht: Springer. Travis, Lisa. 2010. Inner aspect: The articulation of VP. Dordrecht: Springer. Zagona, Karen. 2012. Ser and Estar: Phrase structure and aspect. In Building a bridge between linguistic communities of the Old and the New World: Cahiers Chronos, vol. 8, ed. Chiyo Nishida and Cinzia Russi, 303–327. Amsterdam: Rodopi. Zagona, Karen. 2015. Location and the ser/estar alternation. In New perspectives on the study of Ser and Estar, ed. Isabel Pérez-Jiménez, Manuel Leonetti, and Silvia Gumiel-Molina, 147–172. Amsterdam: John Benjamins. https://doi.org/10.1075/ihll.5.

Classifier Phrase as Host for Stage-Level Spanish Adnominal …

37

Zagona, Karen, and Heles Contreras. 2017. Spanish participial adjectives and individual-level/ stage-level interpretations in nominals’. In Boundaries, phases, and interfaces: Case studies in honor of Violeta Demonte. Linguistik Aktuell/Linguistics Today, ed. Olga Fernández Soriano, Elena Castroviejo Miró, and Isabel Pérez Jiménez, 280–304. Amsterdam: John Benjamins. https://doi.org/10.1075/la.239.13zag.

The Syntax of Mirative Focus Fronting: Evidence from French J.-Marc Authier and Liliane Haegeman

Abstract Cruschina (2009, 2011) argues against the general claim (cf. Rizzi 1997) that focus fronting is possible only when associated with contrast and proposes, based on data from Sicilian and Sardinian, that corrective focus (CFoc) fronting and mirative focus (MFoc) fronting target distinct left periphery projections. One of the goals of this paper will be to provide independent evidence for this claim drawn from French, a language which we will show only allows MFoc fronting. A second goal will be to provide a much-needed semantic characterization of MFoc. We will show that while MFoc fronting constructions and wh-exclamatives pattern alike in terms of information structure, the former differ semantically from the latter in having what Nouwen and Chernilovskaya (2015) call an e-level rather than an i-level expressive content.




Keywords e-level interpretation Expressive content interpretation Left periphery Mirative focus








Focus fronting
i-level

1 Introduction: Focus Movement and FocP On the basis of empirical evidence drawn mainly from Italian, French and English, Rizzi (1997) developed a first articulated structure of the clausal left periphery along the lines of (1).

Liliane Haegeman’s contribution to this work was funded through FWO project 2009-Odysseus-Haegeman-G091409. J.-M. Authier (&) Department of French and Francophone Studies, Pennsylvania State University, University Park, PA, USA e-mail: [email protected] L. Haegeman Department of Linguistics, Ghent University, Ghent, Belgium e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_3

39

40

J.-M. Authier and L. Haegeman

(1) [ForceP [TopP [FocP [TopP [FinP TP]]]]] In addition to the Force projection, ForceP, which encodes illocutionary force and the Finiteness projection, FinP, which encodes finiteness, there are two projections, TopP and FocP, which are associated with specific discourse functions and are only present when these functions are activated. Topic projections can be iterated, accounting for multiple instances of Clitic Left Dislocation in Romance. As for FocP, the hypothesis is that it can only be activated once in a single C field. Rizzi (1997) accounts for the uniqueness of FocP in terms of interpretation while Haegeman (2012) provides an explanation based on intervention. In later work, Rizzi has argued for additional projections, including ModP (Rizzi 2014) for left peripheral adjuncts, as well as IntP (Rizzi 2001) activated in polar questions, for example, but these will not be relevant to the present discussion. Rizzi’s FocP, on which we will focus, can be activated in both matrix and embedded clauses and is assumed to provide a landing site for two types of elements: wh-phrases and constituents the denotation of which is understood to “correct” previous information in the discourse.1 We illustrate the latter option (often referred to as Corrective Focus (CFoc) by e.g., Gussenhoven 2008 and Ortega-Santos 2013) in (2), using Spanish data.

(2) a. A: Jens publicó su primer artículo en Alemania. ‘Jens published his first article in Germany.’ B: No, el SEGUNDO ARTÍCULO publicó no the second article (he)published en Alemania (no el primero). in Germany (not the first) b. A: Renata dice que Jens publicó su primer artículo en Alemania. ‘Renata says that Jens published his first article in Germany.’ B: No, dice que el SEGUNDO ARTÍCULO publicó no (she)says that the second article (he)published en Alemania (no el primo). in Germany (not the first) Although this is not discussed in Rizzi (1997), semantically, Foc can be seen as a head whose function is to evoke alternatives. The compositional analysis of a whquestion as the set of alternative propositions that serve as its possible answers goes back to Hamblin (1973). In a similar vein, fronted phrases interpreted as CFoc,

1

See also Benincà (2001). FocP was subsequently argued by Lopez (2009) to exist in the Romance left periphery in general.

The Syntax of Mirative Focus Fronting: Evidence from French

41

could be argued to straightforwardly fall under the alternative-based semantics for focus advocated by Rooth (1985, 1992); that is, the fronted expression in (2) has both a regular semantic value (that picked out by el segundo artículo ‘the second article’) and a focus value, which is the set comprising its alternatives, i.e. a set of meanings of the same semantic type that includes the meaning introduced by the corresponding expression in the original sentence (see also Krifka 2007). On the basis of cross-linguistic evidence, however, many questions have emerged regarding the characterization of Rizzi’s left peripheral FocP in (1). In this paper, we will discuss some of the issues that arise from a phenomenon that has received little attention in the literature: NP/DP fronting in French. French displays overt wh-movement to the left periphery, suggesting that it instantiates FocP. But if such is the case, the question arises as to why CFoc cannot be expressed by fronting the focused constituent to the left periphery in French (cf. (3a)), as it is in Italian, but is instead expressed via clefting (cf. (3b)), which is arguably a bi-clausal structure (see Haegeman et al. 2014).

(3) A: Albert a appelé son fils. ‘Albert called his son.’ B:

a. *Non, SA MÈRE il a appelé (pas son fils) no his mother he has called (not his son) b. Non, c’est SA MÈRE qu’il a appelée (pas son fils). no it-is his mother that-he has called (not his son)

For discussion of clause internal focus in the cartographic tradition and further discussion of the alternations between focus fronting and clefting, we refer the reader to Belletti (2009). The fact that French encodes CFoc via clefting is noted by Rizzi (1997) but there is, at present, no satisfactory explanation as to why French does not allow Italian-style fronting of focused nominals. It is worth pointing out that a related problem arises with respect to English. While the left peripheral encoding of CFoc has been reported (cf. Rizzi’s own discussion of the contrast between focus fronting and topicalization, for instance) the fact remains that the most natural way to express CFoc in English is not by moving the relevant constituent to the left periphery but by identifying/marking this constituent in situ via pitch accent, as (4) illustrates.

(4) A: Cleo drives a Prius. B: a. #NO, a CAMry she drives (not a Prius). b. NO, she drives a CAMry (not a Prius). A second type of problem tied to the status of the left peripheral FocP concerns the nature of focus itself and the discourse effects that are created by the fronting of focused material. Rizzi discusses the properties of CFoc in some detail, but it has become clear over time that CFoc is not the only type of focus that gives rise to left

42

J.-M. Authier and L. Haegeman

peripheral fronting. Indeed, Cruschina (2009, 2011) shows that in Sardinian and Sicilian, left peripheral focus fronting is not restricted to CFoc, but also “obtains with constituents expressing new and unexpected information, i.e., Mirative-Fronting (Cruschina 2009: 22)”. The Sicilian example in (5) illustrate this phenomenon.

(5) Na machina s’ accattà. REFL buy.PAST.3SG a car ‘A car he bought.’ This pattern, to which we will refer as “mirative focus fronting” (MFoc fronting) is actually classified as Information Focus fronting (IFoc fronting) by Cruschina (2009, 2011). However, as pointed out by Paoli (2009), and Jiménez-Hernández (2015), among others, mirative focus cannot simply be equated with information focus because it goes beyond providing new information in that it signals that this information is unexpected or surprising in the context of an implicit comparison with alternatives. In Sardinian and Sicilian, CFoc fronting and MFoc fronting both target a left peripheral position, which one might label FocP. However, the two processes exhibit different syntactic characteristics. Specifically, while CFoc can appear in the left periphery of a complement clause, MFoc is a so-called main clause phenomenon (see Haegeman 2012 for an overview); that is, it “cannot undergo partial movement to an intermediate periphery but must move on to the left periphery of the matrix clause (Cruschina 2009: 25).” The paradigm in (6a–b) (Cruschina’s (38a–b)) illustrates this contrast in Sicilian.

no un muturinu (6) a. Dissi ca NA MACHINA s’ accattà, say.PAST.3SG that a car REFL buy.PAST.3SG not a moped ‘He said he bought a CAR, not a moped.’ b. *Dissi ca na machina s’ accattà. say.PAST.3SG that a car REFL buy.PAST.3SG ‘#He said that a car he bought.’ This suggests that MFoc fronting in Sardinian and Sicilian involves a derivational process distinct from CFoc fronting.2 In fact, Cruschina (2011) proposes that the two fronting operations target two distinct projections in the left periphery, with 2

As pointed out by Frascarelli (2010: 2139), there are at least two other languages in which non-corrective focus fronting is licensed in matrix clauses and excluded from embedded C-domains, namely, Hungarian and Basque (see Ortiz de Urbina 1999 in relation to the latter). In a similar vein, based on their study of two Cushitic languages, Somali and Afar, Frascarelli and Puglielli (2007) point out that what they call “genuine Focus information” is only available in matrix C-domains and, in line with proposals made by Haegeman (2002) for adverbial clauses and Bayer (2001) for emphatic topicalization in Bavarian, conclude that “new information is strictly connected with the Root Illocutionary Force.”

The Syntax of Mirative Focus Fronting: Evidence from French

43

the CFoc head asymmetrically c-commanding the MFoc head (IFoc in his terminology), as in (7). He further proposes (chapter 4) that the two focus projections are specialized in that non-D-linked wh-phrases move to MFocP (his IFocP) while D-linked wh-phrases target CFocP.

(7) [CFocP [TopP [IFocP [FinP TP]]]] In this article, we will take a close look at the syntactic and semantic properties of focus fronting in French. While, as mentioned earlier, CFoc in French is not encoded via fronting to the left periphery (a fact that will remain unaccounted for), Sabio (2006) and Abeillé et al. (2008) have pointed out the existence of a type of non-wh fronting that will be argued to be another instance of MFoc fronting. We will thus provide evidence that MFoc fronting is a grammatical operation that cuts across Romance languages. Our primary goal will be to document French MFoc fronting in more detail, laying out the semantic properties and discourse properties associated with this grammatical operation, but we will also discuss the issues that arise with respect to integrating MFoc fronting as a component of the cartography in French. Ultimately, we hope that the empirical evidence we provide will contribute to a better understanding of the way discourse interpretive aspects are syntacticized (in the sense of Cinque and Rizzi 2008).

2 The Syntax of MFoc Fronting in French The focus of our paper is a fronting operation in French in which a DP occurs in the left periphery in the absence of a resumptive clitic. The interpretive effect of such an operation is that of mirative focus. In this section, we examine the characteristics of MFoc fronting against the background of other fronting operations without resumption in French. We also discuss the fact that French MFoc fronting can be accompanied by an overt “complementizer” appearing to the right of the fronted constituent.

2.1

Topic Fronting in French

It has often been assumed, in the literature on French syntax, that argument fronting without clitic resumption is severely limited. That is, while English topic fronting can move NPs/DPs to the left periphery, the same operation has been argued to not be available in Modern French (cf. Zaring 1992; and Troberg 2004 among many others).3 When a DP/NP topic is fronted in French, a resumptive strategy is required instead, as shown in (8). 3

For the purposes of the present paper, we adopt the simplest and most commonly accepted definitions of the terms topic and focus. These definitions are based on information structure. That

44

J.-M. Authier and L. Haegeman

(8) a. Pierre je *(l’) ai vu. Pierre I (him) have seen ‘Pierre, I saw.’ b. Cet article je *(l’)ai lu hier. this article I (it)have read yesterday ‘This article, I read yesterday.’

(Zaring 1992: 361) (Troberg 2004: 143)

In French, topic fronting without clitic resumption is available for PPs (Sabio 1995; Kerleroux and Marandin 2001; Delais-Roussarie et al. 2004; Authier and Haegeman 2015, among others), and for TPs that are complement to modal verbs like want (Authier 2011; Authier and Haegeman 2015). The examples in (9) illustrate this point.

(9) a. Marie a réuni ses élèves. Aux filles, elle a donné des exercices Marie has gathered her pupils to-the girls she has given some exercises d'algèbre. Aux garçons, elle a dicté un problème de géométrie. of-algebra to-the boys she has read-out a problem of geometry (Delais-Roussarie, Doetjes & Sleeman 2004: 512) b. Je veux pas laver tes chaussettes, mais [PRO nettoyer l'évier] I want not to-wash your socks but to-clean the-sink je veux bien. I want well ‘I don’t want to wash your socks, but clean the sink I do want to.’ (Authier 2011: 198) Kerleroux and Marandin (2001) and Delais-Roussarie et al. (2004) point out that in French, topic fronting applied to PPs is typically used to shift from a simple discourse referent to a layered one. They explain that the context in (9a) reshapes the group denoting expression les élèves ‘the students’ found in the first sentence. The fronted PPs aux filles ‘to the girls’ and aux garçons ‘to the boys’ are topics that embody a shift from this simple referent to a layered one and this function can be encoded by the topic fronting of those PPs (though interestingly, as they point out, not by PP Clitic Left Dislocation). Note that this characterization of topic fronting appears to be very close, if not identical, to Prince’s (1981) characterization of English fronted topic phrases as representing an entity that is in a salient set-relation to something already in the discourse. Importantly, though, topic fronting may affect DPs/NPs in English, as Prince and others have noted, but in French, topic fronting of NPs/DPs remains unavailable in the absence of resumption.

is, we assume that topics are linked to old information, while focused constituents introduce new information.

The Syntax of Mirative Focus Fronting: Evidence from French

2.2

45

Topic Fronting and MFoc Fronting Instantiate Different Derivations

As just mentioned, the prevalent view in the literature is that French NPs/DPs cannot be fronted in the absence of clitic resumption and that when clitic resumption is used, the corresponding constituents in the left periphery have the status of topics. Blanche-Benveniste (1996), Sabio (2006), and Abeillé et al. (2008) have, however, pointed out the existence of data that appear to directly challenge this view. The examples they provide are of two types. The first type, illustrated in (10), is discussed in some detail in Sabio (1995: 134–135). This type of NP/DP fronting is restricted to a small class of predicates which includes verbs of appreciation (e.g., like, love, hate, can’t stand) and verbs of knowledge (e.g., know). Given that this type of fronting does not obey island constraints (cf. (11)), it is far from clear that it is indeed an instance of fronting (of the internal merge variety).4 We do not discuss this pattern here. (10) a. Les coquillages, il adore. the shellfish he loves ‘Shellfish he loves.’ b. La Troisième République, j'ai pas connu. the third republic I-have not known ‘The Third Republic I didn’t live through/experience.’ (11) a. Le chocolat, j'en connais pas beaucoup [qui aiment pas]. the chocolate I-of-them know not many that like not ‘Chocolate, I don’t know many people who don’t like it.’ b. Ce Boulevard Pasteur, il doit bien y avoir quelqu'un [qui connaisse], this boulevard Pasteur there must well be someone who knows non? no ‘This Pasteur Boulevard, there must be someone who knows it, don’t you think?’

4

Abeillé et al. (2008) and Authier and Haegeman (2015 ft.4) suggest that such examples are better analyzed as constructions containing a pro in argument position (i.e., a silent resumptive pronoun). That this hypothesis is on the right track is suggested by the fact that the relevant predicates can take a silent complement in the absence of a “left dislocated” constituent. The discourse in (i) illustrates this phenomenon.

(i)

A:Tu veux du chocolat? ‘Would you like some chocolate?’ B: Non, j’aime pas. no I-like not ‘No, I don’t like (chocolate).’

46

J.-M. Authier and L. Haegeman

The second type, illustrated in (12), is not subject to similar predicate restrictions. A nominal constituent is found in a left peripheral position and, relevant to our present purposes, is associated with mirative focus interpretation.5 This type of fronting will, from here on, be our primary focus.

(12) a. Des sauterelles grillées ils mangent dans ce pays. some grasshoppers grilled they eat in this country ‘Grilled grasshoppers they eat in this country.’ b. Même ses caleçons on lui a volé. even his underwear they him have stolen ‘Even his underwear they stole from him.’ c. Une bulle il m'a mis, ce salaud. a bubble he me-has put that jerk ‘A fat F he gave me, that jerk.’ d. Dix points de suture ils lui ont fait. ten stitches they him have made ‘Ten stitches they gave him.’ Although, as far as we know, this has not been reported in the literature, fronting with mirative focus interpretation in French exhibits a cluster of properties commonly associated with movement. It can create unbounded dependencies (13); it is sensitive to island constraints (14); and it licenses parasitic gaps (15).

(13) a. Même ses caleçons il affirme que ces voyous lui ont volé. even his underwear he claims that these hoodlums him have stolen ‘Even his underwear he claims those hoodlums stole from him.’ b. Dix points de suture l’infirmier a annoncé qu’il lui avait fait. ten stitches the-nurse has announced that-he him had done ‘Ten stitches the nurse reported to have given him.’ (14) a. *Même ses caleçons on a arrêté les voyous qui even his underwear they have arrested the hoodlums that lui avaient volé. him had stolen b. *Dix points de suture je connais l’infirmier qui lui a fait. ten stitches I know the-nurse that him has done

5

We hasten to add that it is not just nominal constituents that can be fronted in this type construction, as e.g. (29) below makes clear. We will, however, almost exclusively use examples involving NPs/DPs because this type of constituent cannot undergo topic fronting without resumption in French and is therefore unambiguously associated with mirative focus fronting.

The Syntax of Mirative Focus Fronting: Evidence from French

47

(15) a. 18 timbres sur 20 elle est arrivée à décoller [e] 18 stamps on 20 she is managed to to-unglue sans endommager [PG]. without to-damage ‘18 stamps out of 20 she managed to peel off without damaging.’ b. Même un iPhone 6 Uri Geller peut tordre [e] sans toucher [PG]. even an iPhone 6 Uri Geller can to-bend without to-touch ‘Even an iPhone 6 Uri Geller can bend without touching.’ Of further interest is the fact that fronted phrases in (12), (13) and (15) cannot undergo movement to an intermediate periphery (16) but must move on to the left periphery of the matrix clause (13), just as was the case for MFoc fronting in Sicilian (cf. (6b)).

(16) a. *J’ai appris que des sauterelles grillées ils mangent ici. I-have learned that some grasshoppers roasted they eat here b. *Les journaux ont rapporté que même son fils il avait deshérité. the newspapers have reported that even his son he had disinherited Because all of the fronted phrases in (12), (13) and (15) express new and unexpected information (i.e., are associated with mirative focus), we will, from this point on, refer to this type of fronting as MFoc fronting.6 French MFoc fronting differs from topic fronting (without clitic resumption), discussed in Sect. 2.1, in a number of ways. First, as already mentioned, unlike MFoc fronting, topic fronting of NPs/DPs requires resumption. Consequently, in our discussion of the diverging properties of MFoc fronting and topic fronting in French, we will compare MFoc fronting with the topic fronting of PPs without resumption. Second, the constituent affected by MFoc fronting in sentences like (12), (13) and (15) functions as the sole point of prosodic prominence, and the rest of the sentence receives the sort of flat intonation displayed by parentheticals. Topic fronting, on the other hand, is associated with a different type of intonation. In the sentences in (9), for example, topic fronting creates an intonational phrase boundary at the end of the topic phrase and the open sentence that follows forms its own intonational phrase. Third, unlike MFoc fronting (cf. (16)), topic fronting can target the left periphery of embedded clauses, as shown in (17).

(17) Elle a dit qu'aux filles, elle donnerait des exercices d'algèbre. she has said that-to-the girls she would-give some exercises of-algebra ‘She said that to the girls she’d give algebra exercises.’

6

A more detailed semantic characterization of MFoc fronting will be given in Sect. 3.

48

J.-M. Authier and L. Haegeman

Finally, in colloquial speech, constituents affected by MFoc fronting can optionally be followed by the “complementizer” que ‘that’, as illustrated in (18). This type of “low complementizer” is not available for constituents that have undergone topic fronting, as shown in (19).7

(18) a. Un verre de vin par personne (qu')ils nous ont servi. one glass of wine per person (that)-they to-us have served ‘One glass of wine per person they gave us.’ b. Huit ans (qu')il avait quand il a appris le russe. eight years (that)-he had when he has learned the Russian ‘Eight he was when he first learned Russian.’ c. Trois heures (qu')elle a passé chez son avocat. three hours (that)-she had spent at her lawyer ‘Three hours she spent at her lawyer’s office.’ (19) Aux filles (*qu')elle a donné des exercices d'algèbre et to-the girls (*that)-she has given some exercises of-algebra and aux garçons (*qu')elle a donné des problèmes de géométrie. to-the boys (*that)she has given some problems of geometry In relation to this contrast, we turn next to a more detailed examination of the distribution of que “complementizers” in French.

2.3

The Status of que in French MFoc Fronting Constructions

As discussed in Rizzi (2014), the initial proposal made by Rizzi (1997) that Italian che ‘that’ and French/Spanish que ‘that’ spell out the Force head does not preclude the possibility that che/que ‘that’ may serve the more general purpose of signaling that some higher portion of the C-field has been activated. In other words, it is conceivable that che/que serves to lexicalize (some of) the abstract features that make up C-domain heads. One of the better-known instances of a “low complementizer,” i.e., one that follows CLLDed topics, can be found in colloquial Iberian Spanish (see e.g., Gupton 2010) in so-called recomplementation contexts; that is, in

7

The existence of MFoc fronting sentences with que has gone largely unnoticed in the literature; the phenomenon is only mentioned in passing in Abeillé et al. (2008: 312).

The Syntax of Mirative Focus Fronting: Evidence from French

49

clauses embedded under reportative and quotative matrix verbs. The example in (20a) shows that recomplementation complementizers (in bold in (20)) follow CLLDed topics (in italics in (20)) and are optional. The iterative character of recomplementation in Spanish is illustrated in (20b) where a complementizer appears following each one of a series of CLLDed topics. (20) a. Repitió que a mi prima (que) la echaron. s/he repeated that my cousin that her threw ‘S/he repeated that they fired my cousin.’ b. Dijo que el dinero que a Juan que se lo mandaban por correo. [Escribano (1991:139)] ‘Lit. S/he said that the money that to Juan that they were sending it to him by mail.’

As Paoli (2006), Villa-García (2012) and Garrett (2014) have shown, a large body of evidence supports viewing recomplementation as the optional overt lexicalization of the head of one or several TopPs, which themselves are embedded under a ForceP whose head must be overtly realized. We thus end up with a (simplified) cartography of the extended C-domain in Romance along the lines of (21).8

(21) [ForceP que [TopP CLLD (que) [FocP wh (que) [FinP … ]]]] It is also important to note that recomplementation patterns may well extend beyond Romance. Based on data from McCloskey (2006) such as (22), Rizzi (2014) explores the plausibility of recomplementation in English. The relevant examples

8 An alternative view would be to assume, as Poletto (2000: 83–84) does for Italian che, that all instances of che/que are instances of the same element, an element which originates in head FinP and can undergo head-to-head movements of varying length (i.e., to Foc or to Foc then Top, or to Foc then Top, then Force). Such a hypothesis raises non-trivial questions, however. First, the issue of what triggers these movements is difficult to address in a non-ad hoc fashion. (Poletto does suggest that movement of che is triggered by the need to check clausal typing features but her proposal is not sufficiently spelled out to be convincing.) Second, at least in the case of recomplementation, up to three instances of Spanish que can be spelled out (see (20b)) and this raises the question of which lower copies of que can be phonologically realized and when. Finally, it is far from clear that head movement can proceed over a Top projection in the left periphery. In fact, Rizzi (1997: 303–304) suggests that, at least in English, TopP blocks head movement to the left periphery, thus accounting for contrasts such as (i).

(i)

a. If tomorrow you should talk to him… b. *Should tomorrow you talk to him… We also refer the reader to Haegeman (2012: 23–24) concerning this issue.

50

J.-M. Authier and L. Haegeman

involve the conjunction that appearing to the right of a fronted adjunct, which itself follows the Force complementizer (see also McCloskey 2006; and Haegeman 2012: 83–87 for discussion). (22) It is useful to know that once you have mastered the chosen dialect that you will be able to pick up a newspaper and read it. (McCloskey 2006: (69d))

We now return to mirative fronting in French. As the reader will recall (cf. (18)), constituents affected by MFoc fronting can be followed by the “complementizer” que ‘that’, a pattern reminiscent of recomplementation, although, since MFoc cannot be embedded, the topmost instantiation of the complementizer is not available. Let us examine to what extent other patterns of recomplementation in Romance may shed light on French MFoc fronting. We note first that it is unlikely that the occurrence of que in French MFoc fronting constructions can be assimilated to the type of Spanish-style recomplementation discussed in the literature because in Spanish, recomplementation is only possible in embedded clauses, while in French, MFoc fronting can only target the left periphery of main clauses. This suggests that Spanish recomplementation and French MFoc fronting are distinct phenomena, hence the optional que complementizers they license either lexicalize distinct heads in the C field or, alternatively, lexicalize distinct features associated with the same head. This conclusion is further supported by the fact that, unlike Spanish-style recomplementation, French MFoc fronting can never iterate, as (23) shows.

(23) *Même cet idiot (que) vice-président (qu')ils ont nommé. even this idiot (that) vice-president (that)-they have named (cf. OK Même cet idiot (qu')ils ont nommé vice-président. even this idiot (that)-they have named vice-president OK Vice-président (qu')ils ont nommé cet idiot.) vice-president (that)-they have named this idiot Finally, as illustrated in (19), French topic fronting does not display the type of Spanish-style recomplementation illustrated in (20) either: fronted topics without resumption can never be followed by que. The literature on Romance reveals, however, that other types of constituents that can target a position lower that ForceP in the left periphery may, in fact, be followed by what appears to be a “lower” complementizer, and this observation extends to French. In particular, “interrogative complementizers”, a term that we borrow from Poletto (2000: 159–160), may follow fronted wh-phrases in some

The Syntax of Mirative Focus Fronting: Evidence from French

51

varieties of French such as the colloquial French spoken in the northern half of France. Since in such cases, the relevant wh-phrases followed by que can themselves be preceded by a CLLDed topic, as shown in (24), and since fronted whphrases in Romance are standardly assumed to take Rizzi’s (1997) FocP in (1) as their landing site, it follows, by transitivity, that this lower occurrence of que cannot be taken to spell out the head of ForceP, but more likely represents the phonological realization of Rizzi’s Foc head, the specifier of which is occupied by the fronted constituent.

(24) %De ta liaison avec Claire, à qui que tu en as parlé? of your affair with Claire to whom that you of-it have spoken While at first sight, one might be tempted to equate this occurrence of que in Foc with the lower occurrence of que in French MFoc constructions, there is some empirical evidence that suggests that the two instances of que are, in fact, distinct. First, there are some speakers (e.g., speakers in Southern France) that allow lower que to follow constituents affected by MFoc, yet these same speakers disallow the occurrence of interrogative que after fronted wh-phrases. Second, as shown in (25), the interrogative que associated with fronted wh-phrases is available in embedded clauses that are complements to verbs that select embedded questions, while MFoc fronting is never available in embedded contexts (cf. (16)).

(25) %Je sais pas trop à qui qu’on doit envoyer ça. I know not much to whom that-we must to-send this ‘I don’t really know who we’re supposed to send this to.’ Thus, interrogative que and the que that follows constituents that have undergone MFoc fronting are to be kept distinct: they spell out different features and, if Cruschina (2009, 2011) is correct (see (7)), they likely instantiate distinct heads in the left periphery. Finally, as shown in (26), in Spanish, CFoc fronting is incompatible with recomplementation: constituents having undergone CFoc fronting (by hypothesis, to the specifier position of Rizzi’s FocP) can never be followed by a complementizer. This is in sharp contrast with French MFoc fronting, which is compatible with que. This observation is in line with our previous finding that French does not use NP/DP fronting to encode CFoc. More generally then, French MFoc fronting cannot be subsumed under CFoc fronting.

52

J.-M. Authier and L. Haegeman

(26) No, Juan dijo que EL COCHE ROJO (*que) quiere comprar, no Juan said that the car red that he-wants to-buy no el azul. not the blue-one ‘No, Juan said he wants to buy the RED car, not the blue one.’ (Garrett 2014)

2.4

MFoc Fronting and Clefting

Yet another possibility that must be considered is that the instantiation of que that may follow constituents affected by MFoc fronting is part of a reduced (or elliptical) cleft in the sense illustrated in (27).

(27) C’est rien que des légumes qu’ils nous ont servis. it-is nothing but some vegetables that-they us have served ‘Nothing but vegetables they served us.’ There are, however, four sets of facts that militate against such an analysis. First, in French and other languages, clefting, unlike MFoc fronting, can occur in embedded clauses, as (28a) shows. Second, in all varieties of French, clefted phrases, unlike phrases that have undergone MFoc fronting, require, rather than just allow, the presence of an overt complementizer, as illustrated in (28b). Third, clefting and MFoc fronting differ in terms of information structure and are not always felicitous in the same contexts, as shown in (29a–b). Finally, as illustrated in (30), a clefted constituent can itself undergo MFoc fronting, which again suggests that the two operations do not serve the same function nor do they compete for the same derivational space.

(28) a. Ils affirment que c'est des légumes qu'ils leur ont servis. they claim that it-is some vegetables that-they them have served ‘They claim that it was vegetables that they gave them.’ b. C'est des légumes *(qu’)ils leur ont servis. it-is some vegetables *(that)-they them have served (29) Je les ai vus tituber en sortant de la discothèque. ‘I saw them stagger out of the disco.’ a. Bourrés qu'ils étaient. (MFoc fronting) loaded that-they were b. # C'est bourrés qu'ils étaient. (Clefting) it-is loaded that-they were

The Syntax of Mirative Focus Fronting: Evidence from French

53

(30) Du bœuf c’était qu’on leur avait servi some beef it-was that-they them had served dans ce restaurant soi-disant végétarien. in that restaurant so-called vegetarian ‘Beef it was that they served them in that so-called vegetarian restaurant.’

3 The Interpretation of French MFoc Fronting In this section, we examine in some detail the interpretive properties of French MFoc fronting. Our conclusions are tentative and, given the rapid expansion of the literature on MFoc fronting, it may well be the case that future research will reveal additional properties. At first blush, French MFoc fronting sentences appear to have much in common with wh-exclamatives such as (31).

(31) Quel courage elle a montré ! which courage she has shown ‘What courage she showed!’ On the syntactic side, just like MFoc fronting sentences, French wh-exclamatives may, in colloquial speech, display a “lower” complementizer following the whphrase (cf. Authier 2013: 383; see also Zanuttini and Portner 2003 for similar facts in Paduan). This is illustrated in (32a). Thus, both MFoc fronting constructions and whexclamatives can be said to have an optional complementizer following the moved constituent. Given that French MFoc fronting has been shown to be strictly a root phenomenon, French wh-exclamatives do appear to have a slightly wider distribution since they are licit in a restricted set of embedded contexts illustrated in (32b–c). However, in embedded wh-exclamatives, the wh-phrase can never be followed by que in colloquial speech. We take this to suggest that root wh-exclamatives, which allow the lower complementizer, are to be set apart from their embedded counterparts.

(32) a. Quel courage qu’elle a montré ! which courage that-she has shown ‘What courage she showed!’ b. Il a alors réalisé quel grand danger (*qu’)ils couraient. he has then realized which great danger (*that)-they ran ‘He then realized what grave danger they were in.’ c. Tu as vu quelle jolie voiture (*que) son père lui a achetée ? you have seen which nice car (*that) his father him has bought ‘Did you see what a nice car his father bought him?’

54

J.-M. Authier and L. Haegeman

There might be additional arguments for classifying MFoc fronting constructions as exclamatives. Indeed, many of the examples of MFoc fronting that can be found on the internet are punctuated with an exclamation mark. MFoc fronting and whexclamatives also pattern alike with respect to information structure, as we will argue in Sect. 3.1, though, as it turns out, not in terms of their semantics, as we will demonstrate in Sect. 3.2.

3.1

Information Structure

As a point of departure, let us consider the interpretation of the wh-exclamative in (31). An exclamative utterance like (31) can be said to convey two implications to which we will refer, following Castroviejo Miró (2008) and Chernilovskaya et al. (2012), as its descriptive content and its expressive content. The descriptive content of (31), determined compositionally, specifies that the degree of courage exhibited by the person denoted by the subject pronoun is high. The expressive content of (31), on the other hand, specifies the emotive attitude the speaker has with respect to the descriptive content. This attitude can be surprise, awe, etc. In other words, the speaker expresses the fact that s/he finds the descriptive content noteworthy in some way. The expressive content of wh-exclamatives just described is, of course, reminiscent of the mirative character displayed by MFoc fronting constructions. Let us pursue this further. Chernilovskaya et al. (2012) (CC&L) examine the discourse properties of wh-exclamatives and make a number of observations that immediately extend to MFoc fronting constructions. As CC&L point out, while the expressive content of a wh-exclamative cannot be coherently denied by the hearer, a whexclamative can be followed by a statement that “serves to better circumscribe the attitude expressed, by excluding surprise (CC&L 2012: 112)” and by substituting to it an alternative attitude such as awe. This is illustrated in (33) (their (12)–(13)), where B’s reply excludes surprise and zeroes in on awe (in the face of the student’s remarkable achievement) as the attitude associated with the expressive content.

(33) A: How many papers she wrote in her first year! B: I am not surprised, though. She is our smartest student. This observation also applies to MFoc fronting, as (34) reveals. In a way parallel to what (33) shows for wh-exclamatives, the continuation in (34B) does not take back the mirative character of the expressive content of (34A) but serves to narrow down the way in which (34A) is to be understood to be noteworthy (e.g., by considering five articles a year to be a remarkable achievement).

The Syntax of Mirative Focus Fronting: Evidence from French

55

(34) A: Cinq articles (qu’)elle a publiés en un an ! five articles (that)-she has published in one year ‘Five articles she published in one year!’ B: Ça m’étonne pas. Elle est toujours en train de faire de la recherche. that me-surprises not she is always PROG to-do of the research ‘This doesn’t surprise me. She’s always doing research.’

Another observation made by CC&L with respect to wh-exclamatives is that while their expressive content in relation to the original speaker cannot be directly challenged by the hearer, it can give rise to debate. In (35) (their (39)), speaker B is not denying the expressive content of A’s utterance, namely that A is struck by the amount of mail, but B questions the attitude expressed by speaker A by suggesting that A should not have been surprised by the amount of mail, given the fact that A and B were away for some length of time.

(35) [Returning after a long trip abroad, finding a huge pile of mail.] A: What a huge pile of mail! B: That’s not surprising. / What did you expect? / Why are you surprised? We were gone six months. Consider now the discourse in (36), in which A’s statement displays MFoc fronting. The type of challenge expressed by B in (36) is exactly of the same kind as the one in (35). That is, in (36), B is not denying that A finds the amount of mail surprising/noteworthy but B questions A’s attitude by stating that, given the length of their absence, A should not have been surprised by it.

(36) A: Plus de cent lettres (que) j’ai trouvées dans notre boîte ! more than one-hundred letters (that) I-have found in our mailbox B: Tu m’étonnes. / Tu t’attendais à quoi? ‘No kidding. / What did you expect?’ Ça fait six mois qu’on est parti. ‘We were gone six months.’ Consider finally how wh-exclamatives and MFoc fronting constructions fare as responses to questions. As Grimshaw (1979) points out, in situations like (37), the descriptive content of a wh-exclamative cannot be used to provide a direct answer to a question. However, as Castroviejo Miró (2008) notes, wh-exclamatives can indirectly convey the answer to a question by expressing an attitude prompted by consideration of the question, as in (38), for example.

(37) A: How tall is John? B: #How tall John is!

56

J.-M. Authier and L. Haegeman

(38) A: Did you enjoy your vacation? B: What great fun we had! As illustrated in (39) and (40), MFoc fronting constructions function in a similar way. Utterances with MFoc fronting are infelicitous when their descriptive content is intended to provide a direct answer to a question, yet they can be used felicitously in reply to a question just in case they answer the question indirectly.

(39) A: Combien de chiens ont-ils ? ‘How many dogs do they have?’ B: #Trois chiens (qu’)ils ont. three dogs (that)-they have

(40) A: Votre fabrique de meubles tourne bien ? ‘Is your furniture factory a success?’ B: Deux cent personnes (qu’)on emploie. two hundred persons (that)-we employ On the basis of the examples above, we conclude that wh-exclamatives and MFoc fronting constructions pattern alike in terms of information structure.

3.2

Semantics

Given their parallelism in terms of syntax and information structure, the question arises as to whether wh-exclamatives and MFoc fronting constructions also pattern alike in terms of their semantics. Again we start from observations made in relation to wh-exclamatives and we determine to what extent these carry over to MFoc fronting constructions. Rett (2011) closely associates wh-exclamatives with concepts such as degree and scalarity. She argues that a wh-exclamative expresses that it is noteworthy that the degree property corresponding to the wh-phrase is instantiated by some value that exceeds the relevant contextual standard. In a case like (31), the degree property in question is provided by the gradable noun courage contained in the wh-phrase. As discussed in e.g., Morzycki (2009), gradable nouns like courage, smoker, etc. have degree arguments and are lexically associated with a scale. This is revealed by the fact that they admit degree readings when modified by size adjectives; for example, a heavy smoker is not a smoker who is overweight but rather, a person whose use of tobacco ranks high compared to the standard (average) consumption of tobacco attributed to smokers. Similarly, in (31), we understand that the degree property introduced by the gradable noun courage is instantiated by a value that exceeds the

The Syntax of Mirative Focus Fronting: Evidence from French

57

standard of courage with respect to humans. However, claiming that wh-exclamatives are always intrinsically associated with degree/scalarity is potentially problematic in view of examples like (41) where, given the non-gradable nature of the denotation of maison ‘house’, it is not immediately clear what kind of standard might be involved.9

(41) Quelle maison (qu')ils ont ! ‘What a house they have!’ In the face of such data, Chernilovskaya and Nouwen (2012) propose that whexclamatives express a noteworthiness evaluation only; i.e., they do not intrinsically involve a degree or scalarity restriction.10 On their approach, each wh-word introducing a wh-exclamative is specified for a kind of noteworthiness it can mark, the exceptional degree reading of e.g. (31) being but one possible instantiation of noteworthiness. Their analysis has the advantage of being immediately able to handle cases like (41) where, in the absence of a degree property introduced by maison ‘house’, one understands that the house in question is noteworthy in some way (by being extravagant or beautiful or creepy etc.). Chernilovskaya and Nouwen define “noteworthy” as in (42).

Indeed, non-gradable nouns like maison ‘house’ do not admit degree readings, but rather, are associated with size readings when modified by size adjectives; hence (i) entails (ii) but (iii), which contains a gradable noun, does not entail (iv).

9

(i) (ii) (iii)

(iv)

cette grande maison ‘this big house’ Cette maison est grande. ‘This house is big.’ ce gros fumeur this big smoker ‘this heavy smoker’ Ce fumeur est gros. this smoker is big ‘This smoker is heavy.’

10

A reviewer points out that English examples of the type in (i) suggest that house can sometimes introduce a scalar property. That is, (i) seems to be saying that it is still the case that what size house one can get in Dallas for $1 million ranks high compared to the standard (average) size of a house in the United States.

(i)

In Dallas, $1 million still buys a lot of house – a minimum of 3,300 square feet – plus the extra amenities we all crave.

In such cases, however, house appears to be synonymous with real estate. Note that since in French, maison ‘house’ cannot take on this meaning, examples like (i) do not occur in that language (cf. *beaucoup de maison ‘a lot of house’).

58

J.-M. Authier and L. Haegeman

(42) An entity is noteworthy iff its intrinsic characteristics stand out considerably with respect to a class of entities.

As (42) makes clear, the relevant noteworthiness pertains to the intrinsic characteristics of the house denoted by quelle maison ‘what a house – lit. which house’ in (41) and crucially not to the buying event that it is part of. In other words, the expressive content of a sentence like (41) targets the house denoted by the fronted wh-phrase without taking into account the sentence of which it is an argument. Thus, (41) and (43a) share the same expressive content, and so do (31) and (43b).

(43) a. Quelle maison ! ‘What a house!’ b. Quel courage ! ‘What courage!’ With this in mind, let us take a closer look at the expressive content or mirativity component of an MFoc construction like (44).

(44) Une maison, (que) Paul et Sophie ont acheté. a house (that) Paul and Sophie have bought ‘A house, Paul and Sophie bought.’ One important difference with wh-exclamatives emerges immediately: unlike the expressive content of a wh-exclamative, that of an MFoc sentence like (44) does not target the characteristics of the entity denoted by the fronted phrase. That is, while in the wh-exclamative in (41), it is the house itself and not the buying event it is part of that is understood to be noteworthy in some way, the expressive content of the mirative focus sentence in (44) targets the event denoted by the sentence. This event is implicitly compared to alternative events (such as buying an apartment, buying a car, etc.) and is judged to be noteworthy relative to those alternatives; for instance, assuming that Paul and Sophie are not well-off, their buying a house as opposed to their buying an apartment is unexpected. The semantic distinction we have just established between French wh-exclamatives and MFoc fronting constructions is, in fact, a straightforward illustration of what Nouwen and Chernilovskaya (2015) call i-level versus e-level interpretations. This distinction makes use of Rett’s (2008, 2011) so-called “card trick test,” which can be illustrated on the basis of the following scenario. Imagine Julien is doing a card trick. Having blindfolded himself, he picks two cards out of a deck that Manon just shuffled: the 4 of diamonds and the 6 of hearts. Julien gives the cards back to Manon who reshuffles the deck, then he picks two cards again: the 4 of diamonds and the 6 of hearts. Though Manon is amazed by this turn of events, she cannot then felicitously utter the wh-exclamative in (45).

The Syntax of Mirative Focus Fronting: Evidence from French

59

(45) Quelles cartes (que) Julien a tirées ! which cards (that) Julien has drawn ‘What cards Julien picked!’ This is because the expressive content of a wh-exclamative targets a property of the wh-referent’s denotatum. Specifically, (45) does not describe a state of affairs that is remarkable but rather, expresses the fact that some intrinsic property of the objects denoted by the wh-phrase quelles cartes ‘what cards’ is noteworthy; for example, the cards are high-valued, have an odd shape, etc. Thus (45) fails Rett’s card trick test because in order for this sentence to be successfully uttered in the relevant scenario, it would have to describe a state of affairs in which the cards themselves are unremarkable and in which what is remarkable is the event(s) involving the cards. The expressive content of (45), however, can only be described as attributing intrinsically surprising properties to the object(s) denoted by the whphrase. This expressive content is what Nouwen and Chernilovskaya (2015) call ilevel interpretation. As they point out, any exclamative with this type of interpretation will fail the card trick test. For a wh-exclamative to be felicitously uttered in the context of Rett’s card trick, it would need to have an e-level interpretation; that is, an interpretation on which the expressive content is directed at the high degree of noteworthiness attributable to the event described by the sentence in relation to (less remarkable) alternative events. In the case of the card trick, such alternative events are events involving Julien picking on the second try one or more cards that are not identical to the ones he picked on the first try. E-level interpretations are, however, unavailable with French wh-exclamatives. While in French (and in English), wh-exclamatives can never have an e-level interpretation (cf. (45)), it is easy to show, based on Rett’s card trick test, that French MFoc fronting constructions can. That is, a sentence like (46) can felicitously uttered in the card trick scenario.

(46) (Ça alors !) Les deux mêmes cartes (qu’)il a (encore) tirées. that then the two same cards (that)-he has (again) drawn ‘(Good grief!) The same two cards he picked.’ We have thus established a clear semantic contrast between wh-exclamatives and MFoc fronting sentences: while wh-exclamatives in French (and in English) convey an i-level interpretation (and therefore fail Rett’s card trick test), French MFoc fronting constructions are associated with an e-level interpretation (and hence pass Rett’s test). What is noteworthy in the wh-exclamative in (45) is some intrinsic property of the denotatum of the fronted constituent quelles cartes ‘which cards’, but what is noteworthy in the MFoc fronting sentence in (46) is the event the sentence describes when compared to alternative events (e.g., he drew a different set of cards).

60

J.-M. Authier and L. Haegeman

We therefore conclude that regardless of whether or not one considers MFoc constructions to be exclamatives, based on their discourse properties, they differ semantically from wh-exclamatives in having an e-level rather than an i-level expressive content.

4 Conclusion and Questions for Future Research From our discussion of the distributional and interpretive properties of the pattern that we identified as Mirative Focus fronting in French, we can draw a number of conclusions and also point to some remaining unanswered questions. First, Cruschina’s (2009, 2011) hypothesis that CFoc fronting and MFoc fronting are distinct grammatical operations is confirmed by our French data: unlike CFoc fronting, MFoc fronting appears to be restricted to the matrix left periphery cross-linguistically and French licenses MFoc fronting but not CFoc fronting. Second, as the reader will recall, we started off our discussion with a brief overview of the cartography of the left periphery developed in Rizzi (1997), raising some questions as to how French fits into this picture, particularly in relation to the role left peripheral positions play in that language. If we adopt Cruschina’s (2009, 2011) arguments adduced from Sicilian and Sardinian that place the CFoc projection above the MFoc projection, we are still left with the task of determining what place the projection that harbors non-CLLDed topics occupies in relation to MFoc. Based on examples like (47), it would appear that such a projection, to which we will refer as TopP* (TopP being reserved for CLLDed topics), must be higher than MFoc. (47) (On a un petit problème avec les commandes passées par les équipes de hockey hier. Aux Forestiers, on a bien envoyé les soixantes bâtons qu’ils avaient commandés mais…) (We have a bit of a problem with the orders placed by the hockey teams yesterday. To the Forestiers, we did send the sixty sticks they ordered but…) a. aux Gaulois, cinquante bâtons (qu’)on a envoyé to-the Gaulois fifty sticks (that)we have sent (au lieu des quinze qu’ils voulaient). (instead of-the fifteen that-they wanted) b. *cinquantes bâtons (qu’)aux Gaulois on a envoyé fifty sticks (that)to-the Gaulois we have sent (au lieu des quinze qu’ils voulaient). (instead of-the fifteen that-they wanted)

However, while the MFoc projection can thus be shown to be syntactically lower in the left periphery that Top*, it remains unclear what hierarchical position the focus projection that hosts CFoc fronted constituents occupies with respect to Top*. In line with Rizzi (1997), Cruschina (2011: 151) does show that in Sicilian, the focus projection hosting CFoc is structurally higher than the projection that hosts clitic doubled topics, as (48) demonstrates.

The Syntax of Mirative Focus Fronting: Evidence from French

(48)

A SALVO i chiavi i detti, no a Pinu. to Salvo the keys them gave not to Pinu ‘I gave the keys to SALvo, not to Pinu.’

61

(Sicilian)

However, given (a) the relative freedom of distribution of CLLDed topics in the left periphery, (b) the fact that such topics are recursive (see (1) as well as the discussion in Haegeman 2012), and (c) the fact that French disallows CFoc fronting (cf. Sect. 1), the precise hierarchy that exists between CFoc, MFoc and (French) Top* (if indeed there is one) is hard to pin down. This issue might be resolved if one could identify a language in which CFoc, MFoc and Top* co-exist. However, a complicating factor is that it might also be the case that the three types of fronting partially overlap in terms of their targeted landing sites and that the structural projections involved are set apart by their specific featural content. Competition between the various types of fronting might then be accounted for in terms of locality restrictions on movement, much in the way proposed by Haegeman (2012). Deciding between these alternatives is, however, beyond the scope of the present paper and will be left aside as an open question for future research. Finally, given the presence of wh-fronting to the left periphery in French and given the fact that NPs/DPs can undergo MFoc fronting in that language, the question arises as to why French resists the kind of CFoc fronting found in Italian and Spanish. Based on empirical evidence from Sardinian and Sicilian, Cruschina’s (2011, chapter 4) proposes that non-D-linked wh-phrases move to MFocP (his IFocP) while D-linked wh-phrases move to CFocP. This proposal is partially challenged by the French data. Indeed, given Cruschina’s hypothesis, one could assume that fronted non-D-linked wh-phrases target MFoc in French, but the fact that French allows movement of D-linked wh-phrases to the left periphery yet disallows CFoc fronting is potentially problematic if Cruschina’s proposal is assumed to hold cross-linguistically.

References Abeillé, Anne, Danièle Godard, and Frédéric Sabio. 2008. Two types of NP preposing in French. In Proceedings of the 15th HPSG conference, ed. Stefan Müller, 306–324. Stanford: CSLI Publications. Authier, J.-Marc. 2011. A movement analysis of French modal ellipsis. Probus 23: 175–216. Authier, J.-Marc. 2013. Phase-edge features and the syntax of polarity particles. Linguistic Inquiry 44: 354–389. Authier J.-Marc, and Liliane Haegeman. 2015. French adverbial clauses, rescue by ellipsis and the truncation versus intervention debate. Probus 27: 33–71. Bayer, Josef. 2001. Asymmetry in emphatic topicalization. In Studia Grammatica 52: Adiatur Vox Sapientiae, ed. Caroline Féry and Wolfgang Sternefeld, 15–47. Berlin: Akademie Verlag. Belletti, Adriana. 2009. Structures and strategies. Routledge: New York. Benincà, Paola. 2001. The position of topic and focus in the left periphery. In Current studies in Italian Syntax, ed. Guglielmo Cinque and Giampaolo Salvi, 39–64. Oxford: North Holland.

62

J.-M. Authier and L. Haegeman

Blanche-Benveniste, Claire. 1996. Trois remarques sur l’ordre des mots de la langue parlée. Langue française 111: 109–117. Castroviejo Miró, Elena. 2008. Deconstructing exclamations. Catalan Journal of Linguistics 7: 41–90. Chernilovskaya, Anna, and Rick Nouwen. 2012. On wh-exclamatives and noteworthiness. In Logic, language and meaning: Lectures notes in computer science, vol. 7218, ed. Maria Aloni, Vadim Kimmelman, Floris Roelofsen, Galit Sassoon, Kathrin Schulz, and Matthijs Westera, 271–280. Berlin: Springer. Chernilovskaya, Anna, Cleo Condoravdi, and Sven Lauer. 2012. On the discourse effects of whexclamatives. In Proceedings of the 30th West Coast conference on Formal linguistics, ed. Nathan Arnett and Ryan Bennett, 109–119. Somerville, MA: Cascadilla Proceedings Project. Cinque, Guglielmo, and Luigi Rizzi. 2008. The cartography of syntactic structure. In CISCL working papers, vol. 2, ed. Vincenzo Moscati. Siena: University of Siena. Cruschina, Silvio. 2009. The syntactic role of discourse-related features. Cambridge Occasional Papers in Linguistics 5: 15–30. Cruschina, Silvio. 2011. Discourse-related features and functional projections. Oxford: Oxford University Press. Delais-Roussarie, Elisabeth, Jenny Doetjes, and Petra Sleeman. 2004. Dislocation. In Handbook of French semantics, ed. Francis Corblin and Henriëtte de Swart, 505–530. Stanford: CSLI Publications. Escribano, José Luis Rafael. 1991. Una teoría de la oracíón. Oviedo: Servicio de Publicaciones de la Universidad de Oviedo. Frascarelli, Mara. 2010. Narrow focus, clefting and predicate inversion. Lingua 120: 2121–2147. Frascarelli, Mara, and Annarita Puglielli. 2007. Focus in the Force-Fin system: Information structure in embedded clauses. In Indiana University Linguistics Club working papers, vol. 13, ed. Kristopher Ebarb and Ken de Jong, 1–9. Bloomington: Indiana University. Garrett, Jordan. 2014. Which que is which?: A squib on reduplicative que complementizers in Iberian Spanish embedded clauses. In Indiana University Linguistics Club working papers, vol. 13, ed. Kristopher Ebarb and Ken de Jong, 1–9. Bloomington: Indiana University. Grimshaw, Jane. 1979. Complement selection and the lexicon. Linguistic Inquiry 10: 279–326. Gupton, Timothy. 2010. The syntax information structure interface: Subjects and clausal word order in Galician. Doctoral dissertation, The University of Iowa. Gussenhoven, Carlos. 2008. Types of focus in English. In Topic and focus: Cross-linguistic perspectives on meaning and intonation, ed. Chungmin Lee, Matthew Gordon, and Daniel Büring, 83–100. Dordrecht: Springer. Haegeman, Liliane. 2002. Anchoring to speaker, adverbial clauses and the structure of CP. Georgetown University Papers in Theoretical Linguistics 2: 117–180. Haegeman, Liliane. 2012. Adverbial clauses, main clause phenomena and the composition of the left periphery. Oxford: Oxford University Press. Haegeman, Liliane, André Meinuner, and Aleksandra Vercauteren. 2014. The architecture of itclefts. Journal of Linguistics 50: 269–296. Hamblin, Charles Leonard. 1973. Questions in Montague English. Foundations of Language 10: 41–53. Reprinted in Montague Grammar, ed. Barbara Partee, 247–259. New York: Academic Press. Jiménez-Hernández, Àngel. 2015. Towards a typology of focus: Subject position and microvariation at the discourse-syntax interface. Ampersand 2:49–60. Kerleroux, Françoise, and Jean-Marie Marandin. 2001. L’ordre des mots. In Cahier Jean-Claude Milner, ed. Jean-Marie Marandin, 277–303. Paris: Verdier. Krifka, Manfred. 2007. Basic notions of information structure. In Interdisciplinary studies on information structure, vol. 6, ed. Caroline Féry, Gisbert Fanselow, and Manfred Krifka, 13–55. Potsdam: SFB 632. Lopez, Luis. 2009. A derivational syntax for information structure. Oxford: Oxford University Press.

The Syntax of Mirative Focus Fronting: Evidence from French

63

McCloskey, James. 2006. Questions and questioning in a local English. In Crosslinguistic research in syntax and semantics, ed. Rafaela Zanuttini, 87–123. Georgetown: Georgetown University Press. Morzycki, Marcin. 2009. Degree modification of gradable nouns: size adjectives and adnominal degree morphemes. Natural Language Semantics 17: 175–203. Nouwen, Rick, and Anna Chernilovskaya. 2015. Two types of wh-exclamatives. Linguistic Variation 15: 201–224. Ortega-Santos, Iván. 2013. Corrective focus at the right edge in Spanish. Lingua 131: 112–135. Ortiz de Urbina, Jon. 1999. Focus in Basque. In The Grammar of focus, ed. Georges Rebuschi and Laurice Tuller, 311–334. Amsterdam: John Benjamins. Paoli, Sandra. 2006. The fine structure of the left periphery: COMPs and subjects. Evidence from Romance. Lingua 117:1057–1079. Paoli, Sandra. 2009. Contrastiveness and new information: A new view on focus. Rivista di Grammatica Generativa 34: 137–161. Poletto, Cecilia. 2000. The Higher Functional Field. Evidence from Northern Italian Dialects. Oxford: Oxford University Press. Prince, Ellen. 1981. Topicalization, Focus-movement, and Yiddish-movement: A pragmatic differentiation. In Proceedings of the seventh annual meeting of the Berkeley Linguistics Society, 249–264. Berkeley: University of California. Rett, Jessica. 2008. A degree account of exclamatives. In Proceedings of SALT XVIII, ed. Tova Friedman and Satoshi Ito, 601–618. Ithaca, NY: Cornell University. Rett, Jessica. 2011. Exclamatives, degrees and speech acts. Linguistics and Philosophy 34: 411– 442. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer. Rizzi, Luigi. 2001. On the position ‘Int(errogative)’ in the left periphery of the clause. In Current studies in Italian Syntax: Essays offered to Lorenzo Renzi, ed. Gugilelmo Cinque and Giampaolo Salvi, 287–296. Amsterdam: North Holland. Rizzi, Luigi. 2014. Some consequences of criterial freezing. In Functional structure from top to toe, the cartography of syntactic structures, vol. 9, ed. Peter Svenonius, 19–45. Oxford: Oxford University Press. Rooth, Mats. 1985. Association with focus. Doctoral dissertation, University of Massachusetts at Amherst. Rooth, Mats. 1992. A theory of focus interpretation. Natural Language Semantics 1: 75–116. Sabio, Frédéric. 1995. Micro-syntaxe et macro-syntaxe: L’exemple des “compléments antéposés” en français. Recherches sur le français parlé 13: 111–155. Publications de l’Université de Provence Aix-Marseille. Sabio, Frédéric. 2006. L’antéposition des compléments dans le français contemporain. Lingvisticae Investigationes 29: 173–182. Troberg, Michelle. 2004. Topic-comment resumptive pronouns in Modern French and Old and Middle French. Toronto Working Papers in Linguistics 23: 133–156. Villa-García, Julio. 2012. Recomplementation and locality of movement in Spanish. Probus 24: 257–314. Zanuttini, Raffaella, and Paul Portner. 2003. Exclamative clauses: At the syntax-semantics interface. Language 79: 39–81. Zaring, Laurie. 1992. Extensions of R-binding. In Theoretical analyses in Romance linguistics, ed. Christiane Laeufer and Terrell Morgan, 347–366. Amsterdam: John Benjamins.

Further Implications of French Devoir and Falloir for Theories of Control and Modality Lisa A. Reed

Abstract Authier and Reed (2009: 44–45) observe that thematic constraints involving French devoir ‘must’ and falloir ‘to be necessary’ fall out from Chomsky and Lasnik’s (1995) original null Case approach to control. However, Reed (2014: Ch. 6) shows that separate data involving Romance floating quantifiers (Baltin 1995; Sportiche 1988), pronominal clitic placement (Cinque 2001; Cardinaletti and Shlonsky 2004), and the distribution of overt and covert nominals in gerunds pose serious problems for that analysis. This article opens with the novel observation that Authier and Reed’s contrasts are problematic for theories of control ranging from those couched in terms of NP-Movement (e.g. Hornstein 1999) to those postulating a syntactically implicit external argument (e.g. Jackendoff and Culicover 2003) to those formulated in terms of Agree and/or predication (Landau 2004, 2013, 2015). This article then draws on additional data and discussion in Reed (2014, 2016) to show how a Caseless approach to PRO can accommodate these facts, but only if one also recognizes with, e.g. Kamp (1975) and Cinque (1999), and contra, e.g. Kratzer (1981, 1991) and Hacquard (2010, 2011), that the closest French equivalent of must, namely, devoir, is syntactically and semantically ambiguous. This article, therefore, sheds new light on our current understanding of control and modality. Keywords Control


devoir
falloir
Implicit arguments
Modality
PRO

1 Introduction This article explores some of the implications that French devoir ‘must’ and falloir ‘to be necessary’ hold for our understanding of control and modality, picking up, in a sense, where Reed (2016) leaves off. More specifically, in that article I survey some familiar and novel facts involving falloir and devoir and show how the theory L. A. Reed (&) Department of French and Francophone Studies, Pennsylvania State University, University Park, PA, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_4

65

66

L. A. Reed

of control developed in Reed (2014: Ch. 6) can accommodate them. However, I also acknowledge (Reed 2016: 342, 349) that the discussion is incomplete since I neither consider whether competing approaches to control might also be able to offer accounts of the data, nor do I explore whether the more lexically parsimonious approach to modals advocated in, e.g. Kratzer (1977, 1981, 1991) and Hacquard (2010, 2011), could do so. The purpose of this article is to fill those two gaps. The discussion opens in Sect. 2 with two intriguing, thematic-based contrasts associated with devoir and falloir that were first observed in Authier and Reed (2009: 44–45) and that we noted receive a straightforward account under what was then the standard (null Case) approach to control developed in e.g., Martin (1992, 2001), Chomsky and Lasnik (1995), and Bošković (1996). After demonstrating that such facts initially appear to offer support for this early Minimalist approach to control, since more recent theories cannot yet accommodate them, separate data are provided in Sect. 3 that clearly mitigate against this analysis and in favor of one that treats PRO as a Caseless, phi-featureless NP, as in Bowers (2002) and Reed (2014: Ch. 6). With this approach to the syntax of control in place, Sect. 4 turns to a closer examination of the syntax and semantics of falloir and devoir in order to explicitly argue against an influential approach to these modal verbs that assumes that they are syntactically and semantically unambiguous. Specifically, I demonstrate that such an analysis cannot accommodate certain restrictions on devoir + infinitive constructions involving weather expletives, inanimates, and idioms, necessitating a theory in which the variable argument structure of a particular modal verb interacts with control theory to produce the attested results.

2 Thematic Constraints on Expletive Il doit/ faut + Infinitive Constructions In examining the syntax of restructuring across the Romance languages, Authier and Reed (2009: 44–45) observe in passing that the then standard null Case approach to control developed e.g., in Martin (1992), Chomsky and Lasnik (1995), and Bošković (1996) offers an immediate account of the following contrasts1: [ForceP PRO faire attention où on marche ici]. (1) a. Il faut it is-necessary to-pay attention where one walks here ‘One must watch one’s step when walking through here.’ il faut [ForceP y (il) avoir du savon dans toutes b. *De par la loi, as per the law it is-necessary there to-have of-the soap in all les toilettes publiques]. the toilets public ‘By law, there must be soap in all public restrooms.’

Data parallel to (2b) were first observed in Zubizarreta (1983: 213), although the crucial contrast between that example and ones like (2a) remained unnoticed. 1

Further Implications of French Devoir and Falloir …

67

[TP PRO faire attention où on marche ici]. (2) a. *Il doit it must to-pay attention where one walks here ‘One must watch one’s step when walking through here.’ il doit [TP (il) y avoir du savon b. De par la loi, as per the law it must it there to-have of-the soap dans toutes les toilettes publiques]. in all the toilets public ‘By law, there must be soap in all public restrooms.’

Specifically, Authier and Reed (2009: 45) note that if one assumes that the French modal verbs falloir and devoir share the semantic property of not selecting for a thematic subject, but differ in their c-selection frames (today, ForceP versus a “defective” TP), then one expects that only falloir constructions will accept an embedded complement clause containing an external theta-role associated predicate, as evidenced by the (a) variants above, since only falloir’s complement clause contains the type of T that is associated with null Case, the licensor of PRO. Similarly, one would expect only devoir constructions to tolerate a complement clause containing a verbal head unassociated with an external theta-role, as made clear by the (b) variants above, since only devoir selects for a “defective” TP complement that allows movement of the expletive out of the embedded clause. (A control structure for the (b) variants is precluded by the inherently non-expletive nature of PRO, cf. Safir (1985: 33–38).2 Let us next observe that these facts appear to support a null Case approach to control over competing theories since the latter can be shown to currently offer no account of these contrasts. Consider first the NP-movement approach to control as it is developed in Hornstein (1999). Hornstein (1999: 92) advocates a non-obligatory control (NOC) approach to examples like (1a), associating them with (updated) structures of the type in (3): (3) It is impossible [ForceP pro to win at roulette].

The presence of ForceP (formerly, CP) in (3) precludes the type of NP-movement that he hypothesizes is at work in configurations of obligatory control (OC), such as the one in (4):

Safir (1985: 33–38) was the first to observe in relation to data parallel to (i)–(ii) below that PRO is inherently non-expletive.

2

(i) [Before PRO making a big decision], every option should be considered. (ii) *[PRO being obvious that John won’t be returning], we can leave.

68

L. A. Reed

(4) Mary expects [TP (Mary) to win].

As a consequence, in an example like (3), a nominal must merge and remain in the Spec of the infinitival TP in order to value the embedded T’s D-feature. As infinitival T, under his approach, is also unassociated with Case, the only candidate is (Caseless) pro, inserted as a Last Resort to ensure convergence. Although the analysis of (3) can be immediately extended, and hence account for the grammaticality of (1a), it, unfortunately, also generates ungrammatical examples like (1b), as the latter would be associated with the entirely parallel (translated) structure below in (5): (5) *By law, it is necessary [ForceP pro to be soap in all public restrooms].

An example like (1b)/(5) is generated since pro will again be inserted as a Last Resort to check T’s D feature and, importantly, the grammaticality of well-known Spanish examples of the type in (6) establishes that pro can be expletive, precluding any non-stipulative account of the contrast between (1a)/(3) and (1b)/(5) that would make reference to the non-thematic nature of the embedded verb. (6) pro parece que a Juan le gusta la música. ‘It seems that Juan likes music.’

The contrasts in (1) and (2) can also be shown to pose problems for a strict interpretation of purely semantic-based approaches to control, i.e. theories that assume that the lexical semantics of the matrix control head correlates with selection for a “bare VP” (in generative terms, a VP not dominated by a thematic vP), as has been suggested, e.g., in work by Montague (1974: Ch. 8), Bach (1979), Chierchia (1985), Dowty, (1985), Sag and Pollard (1991), and others. Jackendoff and Culicover (2003: ft. 1) clearly favor such an approach. They suggest (Jackendoff and Culicover 2003: 524, 527) that selection for complement clauses of a given semantic type correlates with the ability to syntactically select for a bare VP (i.e. control) complement clause. Predicates like be difficult/important/ necessary, for example, are analyzed as involving bare VP complementation with NOC, as in (7) below, because such heads semantically select for complement clauses of both the Action and Non-Action types, as established by the equal grammaticality of the Action-denoting complement clause in (7) and the Non-Action-denoting ones in (8a, b): (7) It is important/difficult/necessary [VP to save money].

Further Implications of French Devoir and Falloir …

69

(8) a. It is important/difficult/necessary [VP to realize that everyone makes mistakes]. b. It is important/difficult/necessary [VP to attain wisdom with age].

The devoir/falloir contrasts in (1)–(2) receive no immediate account under this theory. For example, the facts below in (9) indicate that the verb falloir is semantically parallel to the Control heads in (7) and (8) above in accepting both Action and Non-Action complement clauses, yet (1b) reveals that falloir, for some reason, is incompatible with bare VP complement clauses containing a predicate unassociated with a subject theta-role. Thus, unless one adds the stipulation that c-selection for a bare VP is contingent upon the presence of an implicit embedded thematic subject, we have no account of this restriction. (9) a. Il faut [VP économiser de l’argent]. it is-necessary to-save of the money ‘It is necessary to save money.’ se rendre compte que tout le monde fait des erreurs]. b. Il faut [VP it is-necessary oneself to-render account that all the world makes of-the errors ‘One must realize that everyone makes mistakes.’

Finally, the contrasts in (1)–(2) also appear mysterious under Landau’s (2004, 2013) original Agree approach to control, as well as his more recent Two-Tiered Theory of Control (Landau 2015). Under the former analysis, PRO is licensed not by a null Case feature associated with the type of infinitival T found in control clauses, but rather by an uninterpretable [−R] feature associated with that head. In a control structure like (1a), for example, infinitival T is associated with an uninterpretable [−R] feature because its [T] and [Agr] feature settings are not both positive. They clearly are not because the [Agr] feature setting is determined by the presence ([+Agr]) or absence ([−Agr]) of overt phi-feature (person, number, gender) morphology. Since infinitival T in (1a) is clearly [−Agr], it is impossible for it to have a combination of two [+] values for [Agr] and [T]. Because the embedded T is associated with an uninterpretable [−R] feature, PRO is the only nominal licensed in its Spec position since it is the only one associated with the required interpretable [−R] feature. (PRO’s interpretable [−R] feature is assumed to be a reflection of its lack of inherent phi-features, the latter being acquired from the controller via Agree or predication.) While this version of Landau’s theory has just been shown to account for the grammaticality of (1a), the same line of reasoning extends immediately to ungrammatical examples like (2a). That is, the modal verbs devoir and falloir both select for complement clauses in which there is no overt agreement morphology on the embedded T, so in both cases, the embedded T will be associated with a [−Agr] feature. The parallel meaning of both verbs would also lexically entail that the same sort of temporal linking obtain between the matrix and embedded clauses. Thus, both verbs would have the same [T] feature settings. Both, in short, should license PRO, but only falloir does.

70

L. A. Reed

Turning now to the more recent Two-Tiered Theory of Control (TTC), we first note, with Landau (2015: 5), that this analysis is not intended to apply to configurations of non-obligatory control (NOC). So if the control variants of (1)–(2) fall into this category, the TTC sheds no light on them. However, this theory is formulated in such a way as to, at least theoretically, cover such data since the express purpose of the TTC is to handle control into complement clauses and these examples fit that description. Namely, Landau (2015: 5) appears to support the hypothesis that NOC is restricted to sentential subject and certain adjunct positions, stating that “…it is far from clear that complement clauses ever display NOC…” This view will now be shown to follow from the form this theory takes. To begin, the TTC semantically classifies control verbs into either the attitudinal or the nonattitudinal class. It then relies on this classification to maintain the empirical coverage formerly gained by the Agree model’s [+/−T] feature. More specifically, control verbs formerly assumed to semantically select [−T] complement clauses (Exhaustive Control verbs) are now observed to be nonattitudinal, while those formerly said to select [+T] complements (Partial Control verbs) are shown to fall into the attitudinal class. The notion (non)attitudinal is semantic. A verb like plan, for example, is attitudinal because the denotation of definite descriptions in its complement clause is determined relative to the epistemic or bouletic dispositions of a participant in the matrix situation, as attested by the fact that the sentences Bill planned to talk to Ralphx. and Bill planned to talk to the new bossx. are semantically equivalent only if Bill happens to be aware of the fact that Ralph and the new boss co-refer. On the other hand, a verb like start is nonattitudinal because the denotation of these NPs is always established relative to the actual world, as evidenced by the fact that Bill started to talk to Ralphx. and Bill started to talk to the new bossx. are semantically equivalent regardless of Bill’s knowledge of the actual reference of Ralph and the new boss. As Landau (2015: 6, 18) observes, modal verbs are nonattitudinal, so falloir and devoir would also be treated as members of this class. Landau then uses this standard distinction to derive his theory of the syntax and semantics of obligatory control (OC) into complement clauses. As the name implies, OC is assumed to take one of two routes—predicative control (for nonattitude verbs) or logophoric control (for attitudes). Since devoir and falloir are both nonattitudes, only the former need be considered further here. Looking more closely, then, at predicative control as it would apply to the control variants in (1)–(2), Landau suggests that the semantic nature of nonattitudes is such that they must uniformly c-select for so-called “transitive” FinP complement clauses. I.e., the lexical semantics of nonattitudes precludes the projection of a “larger” or enhanced left periphery whose head, C, represents the context of speech or thought, as is the case for attitudes. Thus, in (1)–(2), devoir and falloir c-select transitive FinP complements. What then is a transitive FinP? It is a type of functional projection assumed to be potentially rendered predicative by the feature makeup of its head. Namely, transitive Fin is associated with an unvalued [uD] feature, that, by definition, must be valued by movement of a nominal operator to its Spec if the derivation is to converge. PRO is just such an operator since it is assumed to be inherently associated with only a valued [D] feature and a set of unvalued /-features. The [uD] feature of

Further Implications of French Devoir and Falloir …

71

transitive Fin in (1a) and (2a), therefore, acts as a probe, triggering movement of PRO to its Spec. This movement then renders the embedded clause predicative since the external theta-role of the embedded verb remains unassociated with a referential NP. In short, a potential configuration of OC is set up. Such a FinP will be directly predicated of any syntactically available controller, as is the case in an example like I must see John today, in which the property denoted by see John today is predicated of I. However, in (1a) and (2a) above, direct predication fails because there are no syntactically available controllers in the matrix clause. That is, Landau adopts a hypothesis, dating back to Chomsky (1986), Chierchia (1989), and others, to the effect that predication can only obtain of argument NPs that are syntactically projected, in other words, of explicit, theta-associated nouns.3 The lack of any such potential controller, therefore, accounts for the unacceptability of (2a). Namely, the embedded external theta-role fails to ever be discharged, causing the derivation to crash (remain uninterpretable) at LF. However, this leaves unexplained the grammaticality of (1a), which should crash for the same reason. To summarize the discussion thus far, Authier and Reed note that in expletive devoir/falloir + infinitive contexts like (1) and (2), devoir is only compatible with a complement clause containing a predicate unassociated with a subject theta-role, while falloir only accepts those associated with one, contrasts that we have now seen are expected under the original null Case approach to control, but not under popular, more recent, analyses of the phenomenon. Namely, PRO is analyzed as a non-expletive NP that can only be valued for null Case; therefore, it will only be licensed in contexts like (1a), in which its thematic and Case requirements are both met. Since these contrasts are left unexplained under more recent theories of Control, they appear to offer interesting support for Chomsky and Lasnik’s analysis. However, as we will see in the next section, there exists a distinct body of data that challenges this conclusion and points to the need for an approach to control according to which PRO is both Caseless and phi-featureless in the syntax, as in Reed (2014: Ch. 6).

3 In Favor of a Caseless, Phi-Featureless Approach to PRO What, then, are the empirical facts that indicate that PRO is not licensed by a null Case feature associated with infinitival T (or, for that matter, by some other feature, such as the EPP/D feature of embedded T in Hornstein (1999), T’s uninterpretable [−R] feature in Landau (2004, 2013), or transitive Fin’s unvalued [D] feature)?

3

Examples like the ones in (i)–(ii) were advanced in support of this view:

(i) (ii)

John ate *(the meat) raw. Chierchia (1989: 152) The room was left (*angry at the guests). Chomsky (1986: 120-121)

72

L. A. Reed

One compelling type of evidence supporting this conclusion involves Floating Quantifier placement (Q-Float) in control clauses. Baltin (1995: 211, 217) notes that under any current analysis of Q-Float, be it his own predication-style approach to the phenomenon, according to which Q-Float is limited to [+Predicative] X′s (X′s whose Spec position is occupied by an NP or its trace) or the stranding analysis first developed in Sportiche (1988), the ungrammaticality of an example like (10b) below indicates that, in these examples at least, PRO does not move to the Spec of infinitival to since such movement would license Q-Float in the pre-infinitival to position either because movement of PRO to Spec of TP makes T′ [+Predicative] (Baltin) or because PRO, like any NP, can be modified by a floating quantifier (Sportiche). If PRO does not undergo movement to Spec of TP in these Control structures, then clearly one cannot maintain the view that PRO is licensed in Control clauses because of a need to value (in earlier Minimalist terms, check off) uninterpretable features inherently and uniquely associated with infinitival T, as suggested in much of the Minimalist literature. (10)

a. b.

[To both be betrayed by their friends] would be disastrous. *[Both to be betrayed by their friends] would be disastrous.

Further Q-Float evidence uncovered even earlier in Sportiche (1988: 437–438) points in the same direction. Sportiche notes, in relation to French examples of the type in (11b), that PRO, in these types of structures, does not even undergo Move to a position higher than Spec of NegP, as again, Q-Float would be licensed in the pre-Neg position. (11)

a. b.

Ils ont décidé [de ne pas [VP tous partir à la même heure]]. all to-leave at the same hour they have decided of Neg not ‘They decided to not all leave at the same time.’ à la même heure]]. *Ils ont décidé [de tous ne pas [VP partir they have decided of all Neg not to-leave at the same hour ‘They decided to all not leave at the same time.’

A third type of Q-Float data supporting this view involves causative get. Authors like Lasnik and Fiengo (1974: 553–554) and Huang (1999: 429) have noted that the acceptability of an agent-oriented manner adverb like deliberately in (12a) below establishes that causative get not only has its well-known ECM use, but also a Control one, a claim argued for at length in Reed (2011). Interestingly, (12b) shows that the Control use of get disallows Q-Float with respect to the embedded subject position, a fact that follows if we assume, with Bošković (2004: 685), that Q-Float is restricted to non-theta positions and PRO in (12a, b) remains in its initial (theta-associated) position, namely the subject of an adjectival passive.

Further Implications of French Devoir and Falloir …

(12)

a. b.

73

John deliberately got [vP PRO/himself hurt last week]. I think those workers deliberately got [vP (*all) PRO injured] in order to collect unemployment.

In short, if these analyses of Q-Float data are on the right track, we are led to the conclusion that PRO does not consistently undergo movement to Spec of TP. In certain structures, in fact, it does not even undergo Move from its initial merge position.4 Thus, the infinitival T of control clauses cannot be assumed to be inherently associated with an uninterpretable feature or features that can only be checked off or valued by PRO, as crucially assumed in Sect. 2 in order to accommodate separate facts involving devoir and falloir. Certain data involving pronominal clitic placement in the Romance languages offer a second, distinct type of evidence for this position. Specifically, authors like Cinque (2001) and Cardinaletti and Shlonsky (2004) have argued that the clitic climbing contrasts exemplified by the Italian and French examples below in (13a, b) can receive an account in terms of arbitrary c-selection differences. Namely, clitic climbing out of the control complement clause in (13a) is allowed in Italian because the complement clause takes the form of a vP small clause. On the other hand, it is disallowed in French (13b) because the complement clause happens to take the form of ForceP, blocking movement:

4

It is important to note that nothing that has been said in the text outright precludes movement of PRO: What is being claimed is that there is no consistent movement of PRO motivated by feature valuation/checking requirements of some functional head in a control clause, as has been suggested in much of the Minimalist literature on control, i.e. Chomsky and Lasnik (1995), Martin (1992), Bošković (1996), and Landau (2004, 2013, 2015), among others. In fact, Radford (2004: 310) and Reed (2014: 260–262) discuss contexts like (ia) and (iia) below in which movement of PRO from its initial Merge position clearly is triggered, although equally clearly not by the feature valuation requirements of T, as the ungrammaticality of the (b) variants establishes:

(i) a. b. (ii) a.

[To both PRO be betrayed (PRO) by their friends] would be disastrous. *[Both PRO to (PRO) be betrayed (PRO) by their friends] would be disastrous. [To PRO seem to yourselves [to both (PRO) have been betrayed (PRO) by your friends]] is understandable in a situation like this one. b. *[Both PRO to (PRO) seem to yourselves [to have been betrayed (PRO) by your friends]] is understandable in a situation like this one.

Reed (2014) suggests that PRO undergoes Move only when this is necessary in order to satisfy some principle of the grammar, such as Principle A of the Binding Theory.

74

(13)

L. A. Reed

a.

b.

Lo volevo [vP PRO vedere subito]. it want-PAST-1SG to-see immediately ‘I wanted to see it immediately.’ *Je le veux [ForceP PRO voir immédiatement]. I it-ACC want-PRES-1SG to-see immediately ‘I want to see it immediately.’

If this approach to clitic climbing is valid, then, clearly, control complement clauses sometimes take the form of a vP small clause. And if this so, then, again, PRO cannot be assumed to be licensed by the need to check off or value a feature or features associated with a functional head outside its initial merge position. Before turning to a third type of empirical fact supporting this view, it bears mentioning that French raising verbs like sembler ‘to seem’ also disallow clitic climbing, a fact that may seem surprising given the preceding discussion. That is, sembler ‘to seem’ is also standardly assumed to select for a category smaller than ForceP (namely, TP) since NP-Movement out of the embedded clause is clearly permitted, cf. (14a), yet clitic climbing is not, cf. (14b): (14)

a. b.

Elle semble [TP (elle) l’aimer]. she seems him-ACC to-love *Elle le semble [TP aimer]. to-love she him-ACC seems ‘She seems to love him.’

The facts in (14) indicate that the appealing claim that the lack of clitic climbing in French simply reduces to arbitrary c-selection for ForceP as opposed to a “smaller” clausal complement (TP or vP) cannot be maintained. In other words, selection for a vP or TP complement is a necessary condition for clitic climbing, but it is not a sufficient one. Some further undetermined factor or set of factors must come into play. Nonetheless, for present purposes, the relevant point remains: In languages in which clitic climbing is licit, certain control verbs c-select for vP small clauses, structures lacking the type of embedded T standardly assumed to be associated with a feature or set of features that license PRO. One final empirical fact pointing in the same direction can be found in Reed (2014: 263–264) in which it is noted that current feature-based accounts of the licensing of PRO leave unanswered the important question of just why PRO and non-pronominal NPs are licensed in gerundive contexts like (15), but pronominal NPs are not5:

5

The fact that PRO and phonetically overt non-pronominal NPs exhibit an overlapping distribution in gerundive contexts has been previously noted, with respect to French and Italian, in Rizzi (1982: 163–164). However, to the best of my knowledge, the crucial fact that this overlap does not extend to pronouns remained unnoticed until Reed (2014).

Further Implications of French Devoir and Falloir …

(15)

75

Les villageois/PRO/*Ils/*Eux étant pauvres, ils n’avaient pas the villagers they them being poor they Neg had not les moyens d’engager un expert. the means of to-hire an expert ‘The villagers/PRO/*They/*Them being poor, they didn’t have the financial resources needed to hire an expert.’

Under null Case accounts of control, the acceptability of PRO in (15) must be interpreted as indicating that gerundive clauses can be introduced with a null Case feature. On the other hand, the licitness of the non-pronominal NP les villageois ‘the villagers’ indicates that introduction of this feature must be optional in these contexts. In other words, it appears from the example in (15) that gerunds can also either be entirely Caseless (in which a lexical NP is acceptable because it is associated with default Case) or they may involve structural (Nominative) Case. But if this is true, then why aren’t pronominal NPs also licit? That is, if the non-pronominal variant happens to instantiate an entirely Caseless structure, then a default Case pronoun like eux ‘them’ should be acceptable; if, alternatively, a structural Case position is available, then the structural Case pronoun ils ‘they’ should be. But all pronominal forms are rejected. Why this should be so is a question left unanswered by a null Case approach to the licensing of PRO. The same, incidentally, is true of other feature-based approaches to the licensing of PRO. Under Landau’s (2004, 2013) analysis, for example, the acceptability of PRO in (15) indicates that gerundive T either has an uninterpretable [−R] feature (which Landau assumes only PRO can check off) or that T entirely lacks an [R] feature, again licensing PRO since, under his analysis, PRO is associated with an interpretable [−R] feature and interpretable features do not need to be checked. If the former is the case, then only PRO should be licensed in gerunds, as all lexical NPs in Landau’s system have an interpretable [+R] feature; if the latter is, then all three types of NP should be licensed. Since neither situation matches the attested facts, we are again led to the conclusion that PRO is not licensed by any of the types of feature checking or valuation presently assumed in much of the Minimalist literature.6 Faced with the preceding Q-Float, clitic placement, and gerundive data, it appears, then, that an alternative approach to the licensing of PRO is needed. In Reed (2014: Ch. 6), I formulate such a theory. Following Bowers (2002: 206–207), who was himself pre-dated in certain respects by work in the Government-Binding (GB) tradition (e.g. Chomsky 1980: 18–32, 1981: 49; and Manzini (1983: 441), I assume that PRO is the “minimal NP.” More specifically, I adopt Bowers’ view that PRO is inherently associated only with a categorial [N] and an interpretable [−expletive] semantic feature. In other words, PRO has neither inherent phi-features (these are assumed to be acquired at the syntax-semantics interface, via the

6

As noted earlier, Landau’s (2015) Two-Tiered Theory of Control is not intended to apply to configurations of non-obligatory control. It, therefore, sheds no light on such data.

76

L. A. Reed

application of a Bare Output Condition that will be provided in the next section), nor is it associated with an unvalued Case (or other) feature. In contrast to Bowers, however, I make the further assumption that PRO does not, in fact, value any features of T. (Bowers, following earlier null Case analyses, assumes that PRO must check off T’s EPP feature and, as result, his theory faces the problems enumerated above with respect to Q-Float, clitic placement, and the distribution of NP types in gerunds.) Under my analysis, PRO, being inherently [−expletive], continues to only be able to undergo initial Merge in a thematic position. Furthermore, lacking inherent phi-features, it is only licensed in clauses in which there is no functional head that is itself associated with a phi-feature bundle that requires valuation by agreement with a phi-complete nominal (a nominal that the functional head, in turn, values for Case). Thus, in infinitival contexts like Colin tried to finish his work on time., the embedded functional head T (to) is assumed to have no uninterpretable EPP, [−R], or phi-features requiring valuation, nor is it assumed to value NPs for Case. Since the embedded verb does have a thematic role to discharge, PRO is licensed, PRO being the only “minimal” (i.e. relatively featureless) NP whose merger allows theta-role assignment without concomitant feature valuation. Following Bowers (2002: 206–209), the apparent overlapping distribution of PRO and lexical NPs exemplified by small clause contexts like (16a) below is attributed to the presence in (16b) versus the absence in (16c) of a matrix AgrOP (in Bowers’ terms, TrP) associated with unvalued phi-features requiring valuation by an inherently phi-feature valued NP, which AgrO/Tr, in turns, values for Case. I.e., PRO is precluded in syntactically transitive structures like (16b) because it cannot value AgrO/Tr’s uninterpretable phi-features and himself is precluded in syntactically intransitive structures like (16c) because its Case feature remains unvalued. (16)

a. b. c.

John deliberately got [PRO/himself hurt last week]. John deliberately got [AgrOP/TrP [vP himself hurt last week]]. John deliberately got [vP PRO hurt last week].

The distribution of NP types in gerundive contexts like (15) above is accommodated by modifying certain widely held assumptions regarding the mechanisms of feature valuation, as well as by rethinking the feature bundles inherently associated with certain types of heads. Specifically, in relation to (15), it is first assumed that inherent Case and phi-features must enter an agreement relationship in order for a derivation to converge. Crucially, this is assumed to be true even of inherent interpretable phi-features. Second, the various types of Ns found in the example in (15) are assumed to be inherently associated with the features in (17a–d), while those of gerundive T (étant ‘being’) are hypothesized to be those in (18):

Further Implications of French Devoir and Falloir …

(17)

(18)

a. PRO b. villageois ‘villagers’ c. ils ‘they’ d. eux ‘them’ étant ‘being’

77

[N, -expletive] [N, -expletive, φ(valued)] [N, -expletive, φ(valued), Case(unvalued)] [N, -expletive, φ(valued), Case(unvalued)] [T, +tense(valued)]

Specifically, PRO is associated only with an [N] categorial feature and an inherent non-expletive semantic feature, neither of which require valuation. Non-pronominal nouns like villageois ‘villagers’ share all of PRO’s features, plus one: They are inherently associated with an interpretable bundle of phi-features (third person, masculine, plural) that must undergo Agree. The fact that non-pronominal nouns in French and English-type languages lack overt case marking is interpreted as indicating that these nouns lack an inherent, unvalued Case feature and thus are, in this respect, like PRO—a radical departure from all previous generative theories incorporating the notion of Case, including Bowers (2002). Pronominal referential NPs like ils ‘they’ and eux ‘them’ in contrast, have all of the features of PRO and non-pronominal nouns, plus an unvalued Case feature (recoverable by the overt morphology), one that will require valuation via agreement with the closest c-commanding head able to value a noun for Case. (I continue to follow more recent generative work in assuming that Case is a by-product of Agree. I.e., phi-complete functional heads are not themselves inherently valued for a Case feature that is “matched” to that a given noun, as in earlier versions of minimalism. Instead, functional heads, like tensed T, do not possess an inherent Case feature, but rather, are able to value a N possessing an unvalued Case feature by virtue of agreement with respect to some other feature(s), such as phi-features.) As indicated in (18), gerundive T is assumed to be similar to infinitival T in lacking any features requiring valuation. I.e., it is phi-incomplete and thus unable to value Ns for Case. This accounts for the ungrammaticality of the pronominal variants in (15), as these types of nouns are inherently associated with an unvalued Case feature that must become valued via prior agreement in phi-features with a functional head. This hypothesis also explains why structures like (15) license the relatively featureless NP PRO: PRO, like gerundive and infinitival T, has no valuation requirements. Nonetheless, gerundive T is assumed to be unlike infinitival T in being partially inflected—it has a valued [+tensed] semantic feature. I suggest that this feature allows gerundive T to undergo agreement with a phi-feature associated non-pronominal NP like les villageois in examples like (15). I further argue that this difference between gerundive and infinitival T accounts for “classic” distributional contrasts of the type in (19b–c) below. Specifically, unlike gerundive and finite T, infinitival T is unable to undergo agreement with any phi-feature associated NP, licensing only Case and phi-featureless PRO.

78

(19)

L. A. Reed

a. b. c.

Mary forgot [that she/the students had sent in the abstract on Friday]. Mary forgot [to [PRO send in an abstract by Friday]]. *Mary forgot [ the students to [send in an abstract by Friday]].

In sum, this alternative analysis of PRO in terms of just the features [+N], [−expletive] accommodates all of the data introduced in this section—facts that are problematic for standard feature-based theories. Unfortunately, however, it does not offer an immediate account of the thematic constraints introduced in Sect. 2. To explain, the grammaticality of falloir examples of the type in (1a) still follows: PRO is licensed by infinitival T because it is the only referential NP that lacks both Case and phi-features. However, ungrammatical examples of the type in (1b) are now also ruled in. I.e., the projection of PRO is no longer “forced” in this theory by the need to value (or check) some feature of the embedded T, a projection that would result in PRO’s being deprived of a theta-role. Instead, such examples should converge under a derivation in which no NP merges in the embedded subject position, as in (20). (20)

*De par la loi, il faut [ y avoir du savon dans toutes as per the law it is-necessary there to-have of-the soap in all les toilettes publiques]. the toilets public ‘By law, there must be soap in all public restrooms.’

Similarly, the analysis advocated in this section cannot yet preclude the projection of PRO in devoir examples of the type in (2a). Such structures should converge since the embedded T is assumed under this theory to have no features requiring valuation and likewise with respect to PRO. How, then, can the facts overviewed in this section be reconciled with those introduced in the previous one? As we will see in Sect. 4, it is argued in Reed (2016) that this apparent paradox is resolved once one takes into account the variable argument structure of these particular French modal verbs. However, as will also be demonstrated in Sect. 4, this approach runs counter to current influential pragmatic accounts of the epistemic versus deontic interpretations associated with modals like devoir ‘must.’

4 On the Implications of Falloir and Devoir for Pragmatic Approaches to Modal Flavor In the earliest cross-linguistic work on the syntax and semantics of modals equivalent to falloir and devoir ‘must,’ it was commonly held that the contrast between epistemic and deontic modality was due to ambiguity at the semantic, and, possibly, syntactic levels. In other words, the fact that must has an epistemic flavor in (21a), but a deontic one in (21b) is assumed in Carnap (1957), Hintikka (1962),

Further Implications of French Devoir and Falloir …

79

Kripke (1963), Lewis (1968), Kamp (1975), and others to be due to the polysemous nature of this lexeme. More specifically, epistemic and deontic must are associated with distinct “accessibility relations” (R). While the R of epistemic must picks out the set of worlds in which all of the facts known in the real world (w) are true, that of deontic must picks out those in which the real world’s legal codes or moral principles are respected. Under this approach, a truth value of 1 (truth) is assigned to a sentence of the form must p if and only if the base proposition p (e.g. the one expressed by A nail punctured the front tire. in (21a)) is true in all of the relevant accessible worlds picked out by R. (21)

a. b.

Oh no! The front tire’s flat. A nail must have punctured it. i.e., Given the evidence, a nail must have punctured the tire. The person who slashed my front tire must be found and punished! i.e., Given certain moral or legal codes, the tire slasher must be punished.

In the syntactic literature, this semantic ambiguity has been argued to correlate in various ways with differences in structure. In early transformational work by Dubois (1969), Ross (1969), Perlmutter (1970), Huot (1974), and others, for example, epistemic modals are uniformly treated as raising verbs (predicative heads unassociated with a subject theta-role), while deontics are analyzed as control heads (i.e. verbs associated with an external theta-role). Although this clear-cut correlation between structure and meaning was subsequently disproven, e.g., by the licitness of expletive NPs in examples involving deontic devoir, such as the one in (2b), distinct facts led subsequent authors such as Picallo (1990), Brennan (1993), Cinque (1999), Butler (2003), Zagona (2007), and others to develop “mixed” versions of the structural ambiguity approach. Cinque (1999), for example, argues that epistemic and deontic must undergo initial merge in the distinct functional head positions in bold below: (22)

Modepis>Tense>Aspect> Modvol> Moddeontic necessity> Moddeontic possibility/ability>

This hypothesis allows him to account, among other things, for the fact that epistemic modals consistently take scope over deontic ones, as made clear in (23) below, an example in which must lacks its normally attested deontic reading, presumably because the deontic head position is filled by have (to). In other words, Cinque’s hierarchy, partially given above in (22), explains why this sentence cannot be paraphrased as It is required that John is believed to watch TV.

(23)

John must have to watch TV every night.

Although a number of the facts in this literature remain without an alternative account, it is nonetheless probably accurate to say that the current dominant

80

L. A. Reed

approach to the syntax and semantics of modals is that originally put forth in Kratzer (1977, 1981, 1991). According to this analysis, modals are structurally unambiguous, with differences in “flavor” being attributed to pragmatics.7 More specifically, she proposes that words like must are unassociated with a subject thematic role; the deontic versus epistemic interpretations attested in sentences like Mary must be in her office right now. are due to vagueness. Must may either be interpreted with respect to a conversational background (f) in which all of the facts known in the real world hold (fepistemic) or it may be interpreted with respect to one in which the set of legal or moral codes in the real world are respected (fdeontic). In this respect, then must is similar to quantifiers like every since in the sentence Every student left., every is not interpreted with respect to the set of students in the whole world, but rather with respect to some contextually salient group of students. Kratzer notes three appealing aspects of her approach. First, alternative ambiguity analyses treat the existence of deontic and epistemic readings for lexemes like must/devoir as a matter of pure chance. Yet in language after language, the equivalents of these words are “ambiguous” in exactly the same way, indicating that an important generalization may be being missed. Second, she points out that an explicit given that phrase can be added to modal statements without redundancy, cf. the felicity of Given current legal statutes, the culprit must be severely punished. Finally, she observes that hard-wiring each and every modal flavor into the lexicon entails an endless list of lexemes. For example, must p can be interpreted with respect to worlds in which laws are respected, ones in which what is good for you obtains, ones in which everyone behaves in a moral fashion, those in which we achieve our goals, those in which everyone behaves rationally, and so on. Clearly, a pragmatic approach to these multiplicities of interpretation is the null hypothesis to the degree that it is empirically tenable. The question then is whether or not the strongest form of this analysis is capable of handling the types of contrasts Reed (2016) observes in relation to deontic and epistemic devoir and falloir (in addition, of course, to the other phenomena discussed in the work cited earlier by Brennan, Cinque, Zagona, and others, which will not be discussed further here). To answer this question, we must first consider how my analysis accommodates the by now familiar contrasts in (1) and (2). In Reed (2016: 350–352), I associate falloir and devoir with the (partial) lexical entries in (24) and (25), a close examination of which reveals that falloir is assumed to differ from devoir in obligatory semantic selection for an internal argument bearing a deontic obligation. This argument may be realized as a clitic pronoun, cf. Il leur faut travailler. ‘They must work.’ or it may remain syntactically implicit, as in Il faut travailler. ‘One must work.’

7

See, e.g., Zubizarreta (1983), Brennan (1993), Bhatt (1998), Wurmbrand (1999), and Hacquard (2010).

Further Implications of French Devoir and Falloir …

81

(24)

falloir ‘to be necessary’ Meaning: It is deontically required of NP that p. (or NP is obliged to p.) c-selection: _____(Indirect Object Clitic), ______FinP s-selection:

(25)

devoir = ‘must’ Meaning: It deontically must be the case that p. c-selection: ______Non-Thematic vP (a.k.a. PrP, cf. Bowers 2002) s-selection:

Given my analysis of control overviewed in Sect. 3, these entries produce the syntactic structures below in (26) and (27). In (26a), PRO is projected because it is the only relatively featureless NP that can satisfy the thematic requirements of the embedded predicate. No nominal is projected in the embedded subject position in the remaining examples first because the embedded vPs do not semantically select for an external argument and, secondly, because the embedded functional heads are unassociated with any feature(s) that would require valuation via agree with an NP. (26)

a. b.

(27)

faire attention où on marche ici]. Il faut [FinP PRO it is-necessary to-pay attention where one walks here ‘One must watch one’s step when walking through here.’ avoir du savon dans toutes [FinP y *De par la loi, il faut as per the law it is-necessary there to-have of-the soap in all les toilettes publiques]. the toilets public *‘By law, there’s someone who must have soap in every public restroom.’

a.

*Il doit [Non-Thematic vP faire attention où on marche ici]. it must to-pay attention where one walks here ‘One must watch one’s step when walking through here.’

b.

De par la loi, il doit [Non-Thematic vP y avoir as per the law it must there to-have dans toutes les toilettes publiques]. in all the toilets public ‘By law, there must be soap in all public restrooms.’

du savon of-the soap

At LF, the BOC below in (28) applies, rounding out the account of the contrasts. Namely, (26a) is grammatical because the implicit object of falloir is available to serve as PRO’s antecedent; (26b) is not because the absurd assertion is made that this argument (understood to refer to some unspecified individual) is legally or morally obligated to ensure that there be soap in every public restroom; (27a) is

82

L. A. Reed

semantically deviant because the embedded verb fails to ever assign its external theta-role, due to the non-thematic nature of the vP/PrP that selects it. Finally, sentence (27b) is well-formed since there is no matrix implicit object that is understood to bear an obligation with respect to the predicate expressed by the embedded clause. (28)

By default, PRO must take as its antecedent a c-commanding implicit or explicit argument within the superordinate clausal domain that immediately dominates the clause in which it appears, with lexical semantic specifications ruling out potential antecedents and ForceP constituting a phase that “closes off” the search space. If there are no c-commanding potential antecedents or the search space is closed off by ForceP, PRO is assigned the index arb and its phi-features are logophorically determined. Reed (2014: 302)

The preceding discussion makes it clear that this account of the thematic-based contrasts attested with devoir and falloir is based largely on differences in argument structure: In falloir/devoir + infinitive constructions, falloir is associated with an internal argument that devoir is not. It is important to note that this aspect of the proposal is not at odds with Kratzer’s pragmatic approach to modal flavor since her core claim relates only to lexemes like must, i.e. a modal with a single phonetic form that is associated with both epistemic and deontic readings that can be shown to correlate with distinct argument structures. Nothing in her analysis precludes two phonetically distinct lexemes like falloir and devoir from being associated with deontic and epistemic readings and differing argument structures. In fact, Kratzer (1991: 650) herself briefly discusses certain data observed in e.g., Ross (1969) and Williams (1985) that appear to support the view that deontic can and may contrast with epistemic might in this respect. In Reed (2016), however, a true “test case” for the cross-linguistically uniform pragmatic approach to modal flavor can be found in devoir. To see why, we first note that devoir encodes not just deontic modality, as in (27b) above, but epistemic as well, as in (29a, b). (29)

a. b.

Ton pneu est dégonflé. Un clou doit l’avoir pénétré. your tire is flat a nail must it-to-have punctured ‘Your tire’s flat. A nail must have punctured it.’ Regarde ces gros nuages noirs sur le pic là-bas. look-at those big clouds black on the peak over-there pleuvoir. Il doit y it must there to-rain ‘Look at the big black clouds over that mountaintop. It must be raining up there.’

We next observe with Reed (2016: 353–357) that deontic and epistemic devoir contrast markedly with respect to the types of NPs that are attested in their surface subject positions. The sentences in (29a, b) above show that epistemic devoir freely tolerates inanimate and weather expletives as surface subjects. The examples in (30a, b) show that it also accepts thematic objects of split VP idioms in surface subject position, as well thematic subjects of split sentential idioms:

Further Implications of French Devoir and Falloir …

(30)

a.

b.

83

Un tel apaisement après des mois d’émeute! a such peace after some months of rioting Justice doit avoir été rendue, d’une façon ou d’une autre. justice must to-have been served of one way or of an other ‘Such peace after months of rioting! Justice must have been served in one way or another.’ Quand ils ne sont pas d’accord sur quelque chose, les plumes when they Neg are not of agreement on something the feathers doivent voler. must to-fly ‘When they don’t agree on something, the fur must really fly.’

In these respects, epistemic devoir contrasts markedly with its deontic counterpart, as the latter exhibits significant “gaps” in just these areas. Namely, deontic devoir is largely intolerant of inanimate NP subjects (31a); it entirely rejects pure weather expletive subjects (31b); and, paradoxically, it accepts as a surface subject an NP that is associated with a split VP idiom (32a), but not one associated with a split sentential idiom (32b): (31)

a.

b.

(32)

a. b.

*Pour qu’il soit légal de pêcher dans cette rivière, le niveau de for that it be legal of to-fish in this river the level of l’eau doit impérativement atteindre au moins de 2 mètres. the water must imperatively to-reach at-the least of 2 meters *‘Before one can fish in this river, the law strictly requires of the water level that it reach a minimum of 2 meters.’ *Dans cette station, il doit impérativement s’arrêter de in this ski-station it must imperatively oneself to-stop of neiger pour qu’on puisse légalement utiliser des dameuses. to-snow for that one can legally to-use of-the snow-groomers *‘At this ski station, the weather is strictly required to stop snowing before snow groomers are permitted.’

Justice doit être rendue. justice must to-be rendered ‘Justice must be served.’ Ruwet (1972: 72) *Quand je leur dis qu’ils ne valent rien en tant when I them-DAT tell that they Neg are-worth nothing in terms qu’équipe, les plumes doivent voler. that team the feathers must to-fly *‘When I tell them that they’re a nowhere team, the fur is required to fly.’

In Reed (2016), these facts are accounted for by assuming that epistemic devoir is unambiguously a raising verb, as suggested in traditional literature on English modals dating back to, e.g., Perlmutter (1970), while deontic devoir (contra the traditional view) is associated with both a raising and a control entry. More

84

L. A. Reed

specifically, epistemic devoir is associated uniquely with the lexical entry in (33), while deontic devoir has the raising entry above in (25), as well as the control one below in (34). (33)

epistemic devoir = ‘must’ Meaning: It epistemically must be the case that p. c-selection: ______TP s-selection:

(34)

control deontic devoir = ‘must’ Meaning: NP deontically must p. c-selection: ______FinP s-selection:

Given epistemic devoir’s entry in (33), the examples in (29)–(30) are all associated with a raising structure that converges because the inanimate or quasi expletive NP initially merges in a thematically compatible position in the embedded clause and then undergoes Move to its matrix surface subject position for purposes of phi- and Case feature valuation. In other words, (29a, b) and (30b) are associated with the raising structure in (35a), while the example involving a split VP idiom in (30a) is associated with the one in (35b). (35)

a. b.

NP devoirepis [TP (NP) embedded predicate] NP devoirepis [TP embedded verb (NP)]

In contrast, the lexical entries for deontic devoir in (25) and (34) interact with the theory of control outlined in Sect. 3 to associate the examples in (31)–(32) with the raising structures in (36a, b) and the control configuration in (36c). (36)

a. b. c.

NP devoirdeontic 1 [Non-Thematic vP (NP) embedded predicate] NP devoirdeontic 1 [Non-Thematic vP embedded verb (NP)] NP devoirdeontic 2 [FinP PRO embedded predicate]

In the case of (31a, b) and (32b), the raising entry in (36a) crashes at LF because the NPs in matrix surface subject position remain unassociated with a thematic role, the matrix and embedded vPs being non-thematic. Raising results in a convergent output only with respect to the VP idiom in (32a) because that NP is associated with the object theta-role of the embedded predicate, as indicated by the structure in (36b). Any attempt to “rescue” (31a, b) and (32b) via the control derivation in (36c) also fails because the matrix v assigns an external theta role of legal obligation to its thematic subject and one cannot normally impose legal obligations on inanimate

Further Implications of French Devoir and Falloir …

85

entities like the water level, the weather, or feathers. Interestingly, however, in those very few situations in which the reach of the law does extend to inanimates, as in (37a–b) below, grammaticality is expected and attested: (37)

a. b.

De par la loi, un litre de lait doit contenir au moins 5% de crème. as per the law a liter of milk must to-contain at-the least 5% of cream ‘By law, a liter of milk must contain at least 5% cream.’ Pour être exportable, un vin doit faire au moins 11 degrés. for to-be exportable a wine must to-make at-the least 11 percent ‘Wines produced for export must contain at least 11% alcohol.’

Having demonstrated how this version of the ambiguity analysis is able to accommodate the data, we are now in a position to explicitly consider whether or not some version of Kratzer’s vagueness approach to modals might be able to do so as well. Under any strict Kratzerian approach to modal flavor, lexemes like devoir are syntactically treated as raising verbs. In other words, all of the preceding sentences involving devoir would be associated with one of the two types of raising structures in (35), with the choice being determined by the thematic subject or object status of the NP in matrix surface subject position. Given the lack of any significant structural ambiguity, to what, then, would this analysis attribute the following two facts: (a) Epistemic devoir exhibits no constraints on split idioms, while deontic devoir only tolerates those of the VP-type and (b) Epistemic devoir freely accepts inanimate NPs as subjects, whereas deontic devoir only tolerates them under very limited conditions? Clearly, the nature of the approach is such that these characteristics must be tied, in some way or other, to the pragmatics of deontic, as opposed to epistemic, modality. To the best of my knowledge, only one author to date has attempted to refine Kratzer’s analysis in such a way, namely, Hacquard (2010, 2011). In the course of developing a pragmatic approach to a subset of Cinque’s (1999) data, she suggests that root readings are pragmatically licit only if the event denoted by the embedded VP is “centered” on a participant in that event, whereas epistemic modality requires anchoring to a local attitude holder, usually the speaker (Hacquard 2010: 92, 109–111; 2011: 25–26). In other words, root modality, which encompasses deontic, abilitive, teleological, bouletic, and other circumstantial flavors, inherently highlights certain properties of an argument participating in the event and, for this reason, such an entity must be made syntactically available, e.g. as a subject or an object. Hacquard suggests that this accounts for well-known contrasts involving weather expletives and split sentential idioms in English of the type in (38a, b) below. Namely, here and in this part of the world are necessary in order to provide access to an entity involved in the VP event in order for the root (in these cases, abilitive) conversational backgrounds to felicitously highlight some property of that entity.

86

(38)

L. A. Reed

a. b.

It can rain hard #(here). Hackl (1998) The shit can really hit the fan #(in this part of the world).

Could such a proposal accommodate the preceding restrictions on deontic devoir as well? That it currently cannot is most easily seen by comparing the English data in (38) with accurate English translations of the relevant devoir data provided earlier. Let us first compare the weather expletive example in (38a) above with a parallel sentence involving deontic devoir, namely, (31b), which can be translated into English as (39):

(39)

*At this ski station, the weather is legally required to stop snowing before snow groomers are permitted to operate.

In what sense could one say that here in (38a) is “available” for anchoring, but the ski station in (39) is not? Both NPs are clearly syntactically present for “highlighting,” but, for some reason, this does nothing to render licit the French deontic devoir examples. Consider next the case of “split idioms.” Why is the NP associated with the split sentential idiom “anchored” in Hacquard’s (38b), but not in the one in French example in (32b), translatable into English as in (40)? (40) *When I tell them that they’re a nowhere team, the law requires the fur to fly.

In both cases, the NP associated with the idiom is clearly syntactically projected, but, for some unknown reason, root modality fails to “highlight” its properties in the case of devoir. Finally, we observe that this proposal does not shed light on the contrasts involving inanimate subject NPs in (31a) and (37a). These are respectively equivalent to (41a, b):

(41)

a. *The law strictly requires of the water level that it reach a minimum of 2 meters before one can fish here. b. By law, a liter of milk must contain at least 5% cream.

The inanimate NP the water level is as syntactically “available” as a liter of milk is in the deontic devoir equivalents of these English examples, so why is only the French equivalent of (41b) acceptable? At this point, one may be tempted to refine Hacquard’s proposal in light of the type of root modality involved, as her data in (38) all involve abilitive conversational backgrounds (expressing what is possible, given the physical properties of some entity, such as the weather), whereas my devoir examples involve deontic modality (conveying what is necessary, given certain legal statutes).

Further Implications of French Devoir and Falloir …

87

That this tack would also fail becomes evident if one compares English must with French devoir. Must can be used to expresses deontic modality just as devoir can. Therefore, if a refinement of the notion of deontic modality is all that is needed to account for the devoir facts noted above, then one would expect deontic must to exhibit the same characteristics. The two modals are, after all, being treated as exact syntactic, semantic, and pragmatic equivalents. That this prediction is not borne out is made clear by first comparing (42) below with its devoir counterpart in (31a), and then comparing (43) with (32b). What such comparisons reveal is that must is not subject to the restrictions associated with devoir. (42) (43)

The water level must reach a minimum of 2 meters before one can fish in this river. When I tell them that they’re a nowhere team, the fur must fly.

We thus conclude that a Kratzerian approach to modality nicely accounts for must, but it cannot be extended to devoir. In other words, we have seen that the equivalents of must in other languages may be vague or they may be ambiguous. Only analysis can make the choice. Devoir, in particular, has been argued here to be syntactically and semantically ambiguous: One of its entries is similar to might in that it semantically hardwires for epistemic modality and is not associated with an external theta-role; the other is like have to in encoding only deontic modality and selecting for an external argument.

5 Conclusion This paper has explored an array of facts involving the French modal verbs falloir ‘to be necessary’ and devoir ‘must’ to argue that a more complete account of the data emerges only if one moves away from current, intuitively appealing, “unambiguous” approaches to modals and towards ones that recognize that the equivalents of must across languages may be syntactically and/or semantically ambiguous. In this way, it has shed light on current theories of the syntax and semantics of modality. This article has also contributed to our understanding of control since certain thematic contrasts involving falloir and devoir have been argued to fall out only under a theory of the phenomenon that (a) recognizes the existence of PRO (b) associates it with a minimal bundle of features, namely, [−expletive] and [+N] and (c) assumes that PRO becomes associated with phi-features via the application of a BOC at the syntax/semantics interface.

References Authier, J.-Marc, and Lisa Reed. 2009. French tough-movement revisited. Probus 21 (1): 1–21. Bach, Emmon. 1979. Control in Montague Grammar. Linguistic Inquiry 10 (4): 515–531.

88

L. A. Reed

Baltin, Mark. 1995. Floating quantifiers, PRO, and predication. Linguistic Inquiry 26 (2): 199–248. Bhatt, Rajesh. 1998. Obligation and possession. MIT Working Papers in Linguistics 32: 21–40. Bošković, Željko. 1996. Selection and the categorial status of infinitival complements. Natural Language and Linguistic Theory 14 (2): 269–304. Bošković, Željko. 2004. Be careful where you float your quantifiers. Natural Language and Linguistic Theory 22 (4): 681–742. Bowers, John. 2002. Transitivity. Linguistic Inquiry 33 (2):183–224. Brennan, Virginia. 1993. Root and epistemic modal auxiliary verbs. PhD dissertation, University of Massachusetts, Amherst. Butler, Jonny. 2003. A minimalist treatment of modality. Lingua 113 (10): 967–996. Cardinaletti, Anna, and Ur Shlonsky. 2004. Clitic positions and restructuring in Italian. Linguistic Inquiry 35 (4): 519–557. Carnap, Rudolf. 1957. Meaning and necessity. Chicago: The University of Chicago Press. Chierchia, Gennaro. 1985. Formal semantics and the grammar of predication. Linguistic Inquiry 16 (3): 417–443. Chierchia, Gennaro. 1989. Structured meanings, thematic roles and control. In Properties, types and meanings II, ed. Gennaro Chierchia, Barbara Partee, and Raymond Turner, 131–166. Dordrecht: Kluwer. Chomsky, Noam. 1980. On binding. Linguistic Inquiry 11 (1): 1–46. Chomsky, Noam. 1981. Lectures on government and binding. Dordrecht: Foris. Chomsky, Noam. 1986. Knowledge of language: Its nature, origin, and use. New York: Praeger. Chomsky, Noam, and Howard Lasnik. 1995. The theory of principles and parameters. The minimalist program, ed. Noam Chomsky, 13–127. Cambridge, Mass.: MIT Press. Cinque, Guglielmo. 1999. Adverbs and functional heads: A crosslinguistic perspective. Oxford: Oxford University Press. Cinque, Guglielmo. 2001. “Restructuring” and the order of aspectual and root modal heads. In Current studies in Italian syntax: Essays offered to Lorenzo Renzi, ed. Guglielmo Cinque and Giampaolo Salvi, 137–155. Amsterdam: Elsevier. Dowty, David. 1985. On recent analyses of the semantics of control. Linguistics & Philosophy 8 (3): 291–331. Dubois, Jean. 1969. Grammaire structurale du français: La phrase et les transformations. Paris: Larousse. Hackl, Martin. 1998. On the semantics of “ability attributions.” Ms.: MIT. Hacquard, Valentine. 2010. On the event relativity of modal auxiliaries. Natural Language Semantics 18 (1): 79–114. Hacquard, Valentine. 2011. Modality. Semantics: An international handbook of natural language meaning, vol. 2, ed. Klaus von Heusinger, Claudia Maienborn, and Paul Portner, 1484–1515. Berlin: de Gruyter. Hintikka, Jaakko. 1962. Knowledge and belief. Ithaca: Cornell University Press. Hornstein, Norbert. 1999. Movement and control. Linguistic Inquiry 30 (1): 69–96. Huang, C.-T. James. 1999. Chinese passives in comparative perspective. Tsing Hua Journal of Chinese Studies 29: 423–509. Huot, Hélène. 1974. Le verbe devoir: étude synchronique et diachronique. Paris: Klincksieck. Jackendoff, Ray, and Peter Culicover. 2003. The semantic basis of control in English. Language 79 (3): 517–556. Kamp, Hans. 1975. Reference and quantification in tense and model logic. In Pragmatics II, ed. Siegfried Schmidt, 150–197. München: Finck. Kratzer, Angelika. 1977. What must and can must and can mean. Linguistics & Philosophy 1 (3): 337–355. Kratzer, Angelika. 1981. The notional category of modality. In Words, worlds, and contexts. New approaches in word semantics, ed. Hans-Jürgen Eikmeyer and Hannes Rieser, 38–74. Berlin: de Gruyter. Kratzer, Angelika. 1991. Modality. In Semantics: An international handbook of contemporary research, ed. Arnim von Stechow and Dieter Wunderlich, 639–650. Berlin: de Gruyter.

Further Implications of French Devoir and Falloir …

89

Kripke, Saul. 1963. Semantical analysis of modal logic I. Normal propositional calculi. Zeitschrift für mathematische Logik und Grundlagen der Mathematik 9 (5–6): 67–96. Landau, Idan. 2004. The scale of finiteness and the calculus of control. Natural Language and Linguistic Theory 22 (4): 811–877. Landau, Idan. 2013. Control in generative grammar: A research companion. Cambridge: Cambridge University Press. Landau, Idan. 2015. A two-tiered theory of control. Cambridge, Mass.: The MIT Press. Lasnik, Howard, and Robert Fiengo. 1974. Complement object deletion. Linguistic Inquiry 5 (4): 535–571. Lewis, David. 1968. On the plurality of worlds. Oxford: Basil Blackwell. Manzini, Maria Rita. 1983. On control and control theory. Linguistic Inquiry 14 (3): 421–446. Martin, Roger. 1992. On the feature content and distribution of PRO. PhD dissertation, University of Connecticut, Storrs. Martin, Roger. 2001. Null Case and the distribution of PRO. Linguistic Inquiry 32 (1): 141–166. Montague, Richard. 1974. The proper treatment of quantification in ordinary English. In Formal philosophy, ed. Richard Thomason, 247–270. New Haven, Conn.: Yale University Press. Perlmutter, David. 1970. The two verbs begin. In Readings in English transformational grammar, ed. Roderick Jacobs and Peter Rosenbaum, 107–277. Waltham, Mass: Ginn and Company. Picallo, M. Carme. 1990. Modal verbs in Catalan. Natural Language and Linguistic Theory 8 (4): 285–312. Radford, Andrew. 2004. Minimalist syntax: Exploring the structure of English. Cambridge: Cambridge University Press. Reed, Lisa. 2011. Get-passives. The Linguistic Review 28 (1): 41–78. Reed, Lisa. 2014. Strengthening the PRO hypothesis. Berlin: de Gruyter. Reed, Lisa. 2016. Some notes on devoir, falloir, and the theory of control. In Romance linguistics 2013. Selected papers from the 43rd linguistic symposium on Romance languages (LSRL), ed. Christina Tortora, Marcel den Dikken, Ignacio Montoya, and Teresa O’Neill, 341–360. Amsterdam/Philadelphia: John Benjamins Publishing. Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Foris Publications. Ross, John. 1969. Auxiliaries as main verbs. In Studies in philosophical linguistics series I, ed. William Todd, 77–102. Evanston: Great Expectations Press. Ruwet, Nicolas. 1972. Théorie syntaxique et syntaxe du français. Paris: Editions du Seuil. Safir, Kenneth. 1985. Syntactic chains. Cambridge: Cambridge University Press. Sag, Ivan, and Carl Pollard. 1991. An integrated theory of complement control. Language 67 (1): 63–113. Sportiche, Dominique. 1988. A theory of floating quantifiers and its corollaries for constituent structure. Linguistic Inquiry 19 (3): 425–449. Williams, Edwin. 1985. PRO and subject of NP. Natural Language and Linguistic Theory 3 (3): 297–315. Wurmbrand, Susi. 1999. Modal verbs must be raising verbs. In Proceedings of WCCFL 18, ed. Sonya Bird, Andrew Carnie, Jason Haugen, and Peter Norquest, 599–612. Somerville: Cascadilla Press. Zagona, Karen. 2007. On the syntactic features of epistemic and root modals. In Coreference, modality, and focus: Studies on the syntax-semantics interface, ed. Luis Eguren and Olga Fernández-Soriano, 221–236. Amsterdam/Philadelphia: John Benjamins Publishing. Zubizarreta, Maria Luisa. 1983. On the notion ‘adjunct subject’ and a class of raising predicates. MIT Working Papers in Linguistics 5: 195–232.

Part II

Language Variation in Romance

Conceptual and Empirical Arguments for a Language Feature: Evidence from Language Mixing Barbara E. Bullock and Almeida Jacqueline Toribio

Abstract This chapter points to the relevance of Herschensohn’s Constructionist view of second language acquisition for the study of bilingual language mixing. In elaborating Constructionism, Herschensohn (2000) argues that the assembling of the lexicon and its attendant features constitutes the major task of the learner. The articulation of the bilingual lexicon is also invoked in formulating the Functional Head Constraint (Belazi et al. 1994), which characterizes patterns of switching in proficient bilinguals and in second language learners by appeal to the matching and checking of features, including language. While the validity of the language feature has been disputed, we underscore the positive consequences of tagging lexical items with a language label, as we move towards recruiting computational tools for effectively exploiting bilingual corpora. We provide evidence of the benefits of language tagging in quantifying language mixing profiles and in classifying bilingual phenomena such as code-switching versus borrowing.







Keywords Borrowing Code-switching Matrix language Functional Head Constraint Determiner Phrase (DP)





Constructionism


1 Introduction In the seminal monograph, The Second Time Around: Minimalism and L2 Acquisition, Herschensohn (2000) offers Constructionism, a model for describing and accounting for second language (L2) acquisition. Adopting the distinction between lexical and functional categories, Herschensohn proposes that L2 learning consists in the acqui-

B. E. Bullock (&) Department of French and Italian, University of Texas at Austin, Austin, TX, USA e-mail: [email protected] A. J. Toribio Department of Spanish and Portuguese, University of Texas at Austin, Austin, TX, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_5

93

94

B. E. Bullock and A. J. Toribio

sition of lexical items and their attendant abstract features and the linking of these to functional categories. As one example, she points to Anglophone learners of French, who must acquire L2 feature-specifications to achieve native-like French derivations. In French, strong verbal features on Tense motivate overt verb raising for feature-checking, leaving behind the adverb and negation, as in (1), whereas the weak features of English preclude raising of lexical verbs in the syntax, as in (2). In terms of Constructionism, the learner must assemble lexical items and gradually ‘realign’ the grammar through mastery of the formal features of the functional categories of the L2. 1. a) Je (*souvent) mange (souvent) des pommes. b) Je ne (*pas) mange (pas) des pommes.

2. a) I (often) eat (*often) apples. b) I do (not) eat (*not) apples.

The conceptual basis of Constructionism—the lexical-functional distinction and the specification of collections of abstract features—is also relevant for investigations of bilingual code-switching, the alternating use of languages. In a study of the syntactic regularities observed in Spanish-English and Arabic-French code-switching, Belazi et al. (1994) propose the Functional Head Constraint (FHC), which prohibits switching between a functional category and its complement, by appeal to a feature-matching process that includes language as a feature: 3. The Functional Head Constraint (Belazi, Rubin & Toribio, 1994) The language feature of the complement f-selected by a functional category, like all other relevant features, must match the corresponding features of that functional head.

In accordance with the FHC, the functional projection of Number, intermediate between DP and NP (Ritter 1991), must share a language feature with its complement NP, as illustrated in the contrasting grammaticality of the Arabic-French code-switched utterances in (4). (Consult Parafita Couto and Stadthagen-González (2017) and the chapters collected in López (2018) for methods of accessing acceptability judgments for code-switching.) 4. a) Ktib dix livres. b) *Ktib ašra livres. wrote.he ten books ‘He wrote ten books.’ (Belazi et al., 1994:229)

In addition to predicting the syntactic constraints on language alternations in proficient bilinguals, the specification of abstract features is implicated in the emergence of L2 learners’ ability to approximate target judgments regarding code-switching well-formedness. For instance, Toribio (2001) reports on the

Conceptual and Empirical Arguments for a Language Feature …

95

disparate intuitions of three groups of learners on the felicitousness of mixed language sentences that differed only in the site of the switch. In the sample set in (5), learners were asked to judge the acceptability of switching at the V-NP juncture (a) and at the AUX-V juncture (b), the latter predicted to be rejected on the FHC because the language tag of the functional element AUXSP does not match that of its lexical complement VEN. 5. a) Los estudiantes han elegido a new representative. b) Los estudiantes han elected a new representative. __(a) is acceptable __ (b) is acceptable __ neither is acceptable

Conforming to Herschensohn’s proposal, adherence to the FHC was reflected only in the introspections of the advanced group, i.e., those who had acquired the relevant features of the target language, whereas violations of the FHC were allowed by beginning and intermediate learners, whose L2 grammars arguably instantiate only a subset of the features required. Rule-governed bilingual productions and intuitions require two (or more) sufficiently articulated and distinct grammars, which L2 learners in the early stages do not yet possess (see Köppe and Meisel 1995 for similar arguments from language mixing in children acquiring two languages from birth). Speaking directly to the code-switching preferences of incipient L2 learners, Rubin and Toribio (1995: 178) surmise that “these speakers might be reinterpreting the segments of one language as having the language feature of the other, and accordingly, they interpret the code-switched sentences as wholly English or Spanish strings, and therefore, such switches do not constitute violations of the FHC since there is no switch in the language feature.” In other words, learners in the early stage of L2 acquisition adopt new words and phrases as additional entries to an extant lexicon, essentially treating them as nonce borrowings (in the sense of Poplack et al. 1988). Nevertheless, scholars have called into question the legitimacy of the language feature at the core of the FHC, pointing out that such a feature has no independent status as a grammatical primitive and that designations such as [+Spanish] or [+English] should be understood only as conventional labels for the set of formal features that define the descriptive facts of the languages (Mahootian 1993; MacSwan 2000). In addition, many have questioned the empirical adequacy of the constraint, as examples of constructions that present violations of the FHC are common. The prime example is the prevalence of language switching after a Determiner within the DP, even though switching at this juncture should be disallowed because the DET is a functional head with an NP complement (Mahootian and Santorini 1996; Jake et al. 2002; Radford et al. 2007; Cantone and Müller 2008; Liceras et al. 2008; Post 2010). However, it is precisely DET-NP switches that are often excluded from consideration by researchers investigating the efficacy of the FHC, a decision that is well-founded given that nouns are the most commonly borrowed lexical items (Appel and Muysken 1987). The issue of isolating borrowing from code-switching is one that we undertake here, and we propose to

96

B. E. Bullock and A. J. Toribio

capture the distinction quantitatively with reference to a language feature assigned at the level of the word and at the level of a span or sequence of words. In this chapter, then, we address the conceptual and empirical considerations of a language feature, operationalized as a language label, and defend its use. We begin our presentation in Sect. 2 with an overview of the notions relevant to research in language mixing, which we define as the overt interaction of language, to focus on borrowing and code-switching, which we will argue are distinguished by an abstract language feature. We make the case for positing a formal language feature in Sect. 3, by establishing, first, that it is not inconsistent with formal theories and, second, that language acquisition involves classifying the input into discrete categories, including language (Byers-Henlein 2014). In Sect. 4, we draw a parallel from the human ability to categorize languages based on perceptual cues to computational systems of classification that are used for language identification in Natural Language Processing (NLP) and Automatic Speech Recognition (ASR), where language is held to be a necessary tag for a variety of downstream processing tasks like part-of-speech (POS) annotation, syntactic parsing, and phone or word alignment. In Sect. 5, we demonstrate how automatic methods of language identification, applied at the word level, can help to empirically investigate broad patterns of bilingual behavior in replicable, accountable ways (Solorio and Liu 2008a, b; Guzmán et al. 2016, 2017a, b; Bullock et al. 2018a). The models that we advance can be exploited with any multilingual corpora for which monolingual training data exists, thus inviting additional analyses from any language pairing. In Sect. 6, we put our ideas and models to the test in experiments of language switching within the DP, the case that is often cited as counterevidence for the viability of the FHC. While the data demonstrate that switches are not frequent in DPs, when switches do arise, they are found primarily when the determiner is Spanish. Moreover, a regression analysis shows that other-language spans following a DET are significantly shorter than other-language spans following any other grammatical category. This quantitative model suggests that these short spans may constitute DP-internal borrowings, for which no switch in language, as a formal feature, need be assumed. We conclude in Sect. 7 with implications of our approach for future studies of diverse datasets in which multilingualism is at issue.

2 Reflections on Language Mixing Various typologies and hierarchies of language mixing have been proposed in the literature, but there is no universal accord for how to classify multilingual processes and outcomes (Weinreich 1953; Thomason and Kaufman 1988; Muysken 2000; Myers Scotton 2002). With the exception of code-switching, the linguistic outcomes of bilingualism can be identical to variations that arise in non-contact environments. As a consequence, we need to give weight to factors such as an individual’s language experiences and the social context of an utterance before classifying any observation (Bullock and Toribio 2018). For instance, innovations

Conceptual and Empirical Arguments for a Language Feature …

97

that appear to be attributable to language shift and bilingualism can instead be traced to the practices of fluent, monolingual speakers (Dubois and Noetzel 2005). Linguists are thus cautioned to evaluate putative contact forms against a series of diagnostics designed to assess the degree to which variation is induced language-externally rather than motivated internally (Mougeon et al. 2005; Jarvis and Pavlenko 2007; Poplack et al. 2012). At first glance, if an utterance contains lexical material from more than one language, the case for the variation as contact-induced seems to be straightforward. But classifying such data as code-switching or borrowing is not that simple. Code-switching (CS) is broadly defined as the alternation of languages within an utterance by a single speaker (Pfaff 1979; Joshi 1982; Sankoff and Poplack 1980; among others). While few would argue that the inter-sentential alternation in example in (6) comprises a full shift in language, the intra-sentential insertion of dominicana ‘Dominican woman’ and moreno ‘Black man’ in (7) and the within-word switch between the English root disappoint and the Spanish inflection -eada in (8) are more equivocal since they might ensue from lexical borrowing. 6. Inter-sentential CS (French-English Bon cop bad cop transcript) Je me suis fait installer un gadget au cerveau,| and I see subtitles under people when they speak. ‘I had a gadget installed in my brain and I see subtitles under people when they speak.’ 7. Intra-sentential (English-Spanish, Brief Wondrous Life of Oscar Wao) She was the only old-school | dominicana | who had dated a | moreno | until Oscar’s father put an end to that particular chapter of the All-African World Party. 8. Intra-word (Spanish-English, Killer Crónicas) Cierro los ojos, | disapoint|eada, horrorizada. ‘She closed her eyes, disappointed, in horror.’

The phenomena of code-switching and borrowing, implicating the overt lexical exponence of contact, appear to be distinct from language interactions at the abstract semantic or structural levels, variously described as processes of interference (Weinreich 1953), transfer (Treffers-Daller 2005), structural borrowing (Thomason and Kauffman 1988), imposition (Van Coetsem 1988), pattern replication (Matras and Sakel 2007), or convergence (Bullock and Toribio 2004). However, lexical borrowing is commonplace, particularly as the world’s major languages spread in social media, and one need not be bilingual to use a borrowed lexical item. In addition, recent research suggests that speakers do not uniformly agree on the perception of ‘foreignness’ of an other-language token. Diab and Kamboj (2011) tasked 93 native speakers of Hindi with labeling the language of word tokens in Hindi novels that featured frequent code-switches to English. The annotators agreed almost uniformly when classifying Hindi words, but 17% of the

98

B. E. Bullock and A. J. Toribio

tokens labeled as English in the expert (gold standard) annotation were labeled by participants as Hindi, inspiring the authors to observe that “English words have blended so well with Hindi that even the native Hindi speakers are not able to recognize them as English words” (Diab and Kamboj 2011: 39). For our purposes, this implies that the mere transfer of overt morpho-lexical material from one language to another is not sufficient evidence that a speaker has switched between language systems; she could instead be adding a novel token or multiword expression to her lexicon. Poplack et al. (1988) observe that bilinguals can borrow spontaneously, and that such nonce borrowing can be common in corpora. They contend that nonce borrowings, which they argue to be single token items, are distinguished from established loanwords by their frequency and distribution: Relative to loanwords that recur across speakers and texts, nonce borrowings are hapax forms. At the same time, nonce words are similar to loanwords via their integration into the grammar of the recipient language and, in this, they are distinct from single-token code-switches. The authors assert, “Whereas in code switching, the speaker alternates between one coherent grammar (and lexicon) and another, according to certain predictable syntactic constraints on switch points, in borrowing only one grammatical system is brought into play” (Poplack et al. 1988: 94). Muysken (2002) formalizes the three-way classification along two dimensions: computed versus stored and supra-lexical versus sub-lexical. On this view, code-switching and nonce-borrowing are spontaneously computed in discourse, but only code-switching involves the supra-lexical, productive syntax. In turn, nonce borrowings and established loans are sub-lexical, but only the latter are stored and familiar. Thus, of all bilingual behaviors, only code-switching involves an alternation in language at an abstract and a lexical level, implicating formal grammatical features alongside morpho-lexical exponents.

3 Making the Case for a Formal Language Feature As noted, scholars have called into question the conceptual validity of The Functional Head Constraint for its appeal to language as an abstract feature. In the Minimalist program in which the FHC was elaborated, lexical items are assumed to be endowed with features, to include u features, Case features, and scopal features. These formal features must be checked against the specifications on functional heads, motivating movement. Thus, verbs bearing strong agreement and tense features raise in order to check verbal features on INFL, DP subjects marked with [Nom] raise to check the nominal Case features of INFL, and wh-phrases raise to enter into a checking relation with a [+wh] COMP. In brief, a core principle of Minimalism holds that an element with some abstract feature [a] must be appropriately paired with another element bearing the same feature [a]. In formulating the FHC, Belazi et al. (1994) augmented the collection of features to include [language], a move that has been stridently rebuked by generativists as “conceptually quite awkward” (Bandi-Rao et al. 2014: 166) and “theoretically spurious” (MacSwan 2014: 14).

Conceptual and Empirical Arguments for a Language Feature …

99

However, a language feature has been assumed by many researchers in advancing constraint-based proposals. The Government Constraint (Di Sciullo et al. 1986) and the Complaisance Constraint (Bhatt 2014) incorporate a ‘language index’ and a ‘language feature’, respectively, for identifying the system from which items are contributed in computing a mixed utterance. A language tag is also consistent with those psycholinguistic models of bilingualism that hold that lexical entries are stored in a common stock but are discernable by their language labels (Green 1998; Dikstra and Van Heuven 2002). Whatever the status of language within generative theory or within psycholinguistic models of bilingual representation, researchers must assume that code-switching bilinguals differentiate between two systems in practice, irrespective of the level at which these systems are formally categorized. In fact, MacSwan concedes that analyses such as the FHC are improved if language labels “act as variables for bundles of features which formally characterize them” (2004: 291). The importance of differentiation is underlined by Byers-Heinlein (2014: 185) who argues that bilinguals are able to functionally discriminate between their languages, regardless of how they represent them, because they are able to “treat elements of their languages as belonging to different categories”. She submits that bilinguals do so, initially based on perceived differences between the categories; later, abstract, conceptual categories like LanguageA, LanguageB, are built from these perceptual categories. The author cites evidence that supports the construction of categories on the basis of features that are perceptible, like differences in rhythm and prosody, but learners may also compute distributional regularities, like the co-occurrence of particular words and patterns. Along similar lines, sociolinguistic literature informs us that bilingual children are able to segregate languages according to interlocutors based on external perceptual cues such as ethnicity (Beebe 1977) and to make subtle shifts in their phonetic repertoires to accommodate to the foreign accentedness of their interlocutors (Khattab 2013), demonstrating their capacity to match lexical and sociophonetic distributions. In effect, at the core of all language acquisition is a classification problem: Language acquirers sort data into separate categories based on similarity of features. For bilingual acquirers, this ultimately yields clusters labelled by language. Monolinguals must also label language as a conceptual category since they, too, can differentiate elements of their language from others and process “islands of bilingualism” within their own grammars, as with the Germanic versus Romance classification that drives dative alternation in English (Roeper 1999).

4 Language Identification in Natural Language Processing The ontological status of a formal language feature may be debated in linguistics, but language identification (LID) is considered to be a crucial preprocessing task for Natural Language Processing (NLP) and Automatic Speech Recognition (ASR) systems, where a language label is a prerequisite for a host of downstream

100

B. E. Bullock and A. J. Toribio

tasks such as part-of-speech (POS) tagging, syntactic parsing, phonetic and word alignment (Baldwin and Liu 2010; Li and Fung 2012, 2013; Vu et al. 2013; King and Abney 2013). LID is frequently accomplished at a level of ‘broad coverage’ (Jurgens et al. 2017) with techniques that assume that a document is written in a single identifiable language. This assumption is challenged when the languages are mixed, as they frequently are in social media such as Twitter (Baldwin and Lui 2010; Lui et al. 2014; Lui and Baldwin 2012; Çetinoğlu et al. 2016). In NLP, language identification is a classification task. It is similar to perceptual categorization in language acquisition in that it relies on comparing ‘bundles of features’ of the input with those of monolingual training data in order to predict which language model the input features are most similar to. Some LID models learn character sequences, or n-grams, from training data and then match the sequences in the input data to similar language sequences in the languages it has learned (Cavnar and Trenkle 1994). Others make use of lists of frequently occurring stop words, such as the in English or que ‘that’ in French and Spanish (Grefenstette 1995). Still others exploit mutual information statistics about the relative complexity of the input compared to the training model (Sibun and Reynar 1996). However, as noted by Baldwin and Liu (2010: 235–6), “all [LID] models are brittle when the assumption of strict monolingualism is broken” as with multilingual input. Recognizing a need to identify the languages of a token or even within a token (Barman et al. 2014; Das and Gambäck 2014; Maharjan et al. 2015; Çetinoğlu et al. 2016), researchers have created word-level LID models for multilingual texts to label each lexical item with its appropriate language tag (Yamaguchi and Tanaka-Ishii 2012; Lignos and Marcus 2013; King and Abney 2013; Nguyen and Dogruoz 2013; Solorio et al. 2014; Guzmán et al. 2016). This parallels the word-level language feature posited in some linguistic models of code-switching and, in the next section, we demonstrate the utility of language annotation at the word level for comparing across mixed language corpora.

5 Patterns of Language Mixing as Identified by Language Tags We share the view of other contact linguists that linguistic proposals about constraints on code-switching need to be investigated using replicable and accountable quantitative methods (Adamou 2016). Accordingly, we seek to employ automated methods of coding and analysis that can be applied to various corpora to permit reliable statistical comparisons between them. To do this, we created an automatic LID to model Spanish-English corpora which produces two levels of annotation: Language and Named Entity (NE), for identifying names for persons, organizations, and institutions (Guzmán et al. 2016, 2017a, b). It is similar to other proposals for LID applied to mixed language texts (Rosner and Farrugia 2007; Franco and Solorio 2007; Solorio and Liu 2008a, b). The Language tier includes tags for Spanish, English, punctuation and number, and the Named Entity (NE) tier labels a

Conceptual and Empirical Arguments for a Language Feature …

101

token as NE if it is recognized as such by either the Spanish or English version of the Stanford Named Entity recognizer (Finkel et al. 2005). Tokens identified as NE retain their language annotation in our system. The model works as follows: Numbers and punctuation are identified by algorithms. For the remaining tokens, a character n-gram (5-gram) model is used to identify the language of a token based on how likely that character sequence is to be encountered in a corpus of Spanish or English training data. Because code-switching is most likely to occur in informal rather than professional styles, we used corpora of film and movie subtitles to train our language models, as they more accurately represent word frequency of casual registers than do corpora composed of newspaper articles, service calls, or parliamentary minutes. The training sets for the n-gram models included a 3 million word portion of SUBTLEXUS for English, and the 3 million word ACTIV-ES corpus of Spanish (Brysbaert and New 2009a, b; Francom 2013; Francom et al. 2014). Because there is considerable orthographic overlap in the lexicons of Spanish and English (e.g., ‘me’, ‘a’, ‘has’, ‘he’) the LID uses a Hidden Markov Model to assess the probability that the language of an ambiguous token is the same or different from the token that precedes it. Our model achieves an accuracy rate of over 95% for Spanish-English corpora when evaluated against an expertly hand-annotated gold standard (Guzmán et al. 2016, 2017a, b).

5.1

The Data

There is a paucity of grammatically annotated bilingual data available to researchers, which compromises efforts to train systems capable of POS tagging code-switched input and stymies linguists’ ability to investigate broad patterns of language mixing in a systematic fashion (Çetinoğlu et al. 2016). For the present analysis, we surmount that challenge by drawing on three Spanish-U.S. English corpora, of similar size (*8,000 words), which have been manually annotated for grammatical category. The first of these, which we call S7, is the text of a recorded 40-min conversation among three individuals who regularly code-switch between southern U.S. English and the variety of Spanish spoken in the southwest of the U.S. (Solorio and Liu 2008b). For the second dataset, we searched the Spanish-English Miami Corpus in the BilingBank repository (Deuchar 2010) for an approximately 8,000-word file that contained language switching, selecting the portion that met our requirements: maria40 (M40). Our third corpus is an extract from Killer Crónicas (KC), an epistolary novel recounting the author’s experiences during a sojourn in Buenos Aires. It is written entirely in ‘Spanglish’ by Chicana writer Susana Chávez-Silverman and is used here with her permission. We processed each corpus using our LID model, ensuring that they would be tagged for language using exactly the same parameters. Punctuation is tagged for language, with the same label as the previous token, so as not to overestimate the degree of switching. Numerals are excluded from analysis, since we cannot

102

B. E. Bullock and A. J. Toribio

determine a language for numeric symbols. The LID procedure returns a time-ordered sequence of language tags that enables us to quantify the proportion of languages represented in each corpus and the probability of switching languages within each corpus by reference to the language of each token. As an illustration, consider the examples in (9) and (10), each from KC. The sentences are of roughly equal length (10–11 words total) and each contains three tokens in Spanish with the balance in English. But the dispersion of the Spanish tokens relative to the English ones differ in the two examples; in (10), there are three code-switch junctures and in (9) only one. Our metrics, described below, are meant to quantify the fact that the second example, with more transitions as marked by |, is more complex than the first. 9.

So different from when I lived in Spain, | en la secundaria.

10. Pero, | an odd opiate centeredness, | hasta orgullo, | washed over me.

The ordering of the language tags for the tokens in such examples allow us to demarcate the beginning and end of a span, or sequence of words, which bear the same language annotation. There are two language spans in (9) but four in (10). The lengths of these spans and the frequency of spans of similar length yield information on whether language switching events occur regularly through time or occur aperiodically, in bursts (Goh and Barabási 2008). Based on work by Barnett et al. (2000), we calculate a Multilingual-Index (M-Index) that returns the relative proportion of languages in a corpus. The M-Index is bounded between 0 (all word tokens in a corpus are from the same language) and 1 (every language in a corpus is represented by an equal number of word tokens). An Integration-Index (I-Index) calculates the probability of switching language between any two tokens (Guzmán et al. 2016, 2017a, b). Like the M-Index, the I-Index is bounded by 0 (no switching) and 1 (each current word token is in a different language from the previous token). Finally, we employ a formula for Burstiness that quantifies how random the spans lengths are (Goh and Barabási 2008). Burstiness is bounded between −1 (periodic signal) and 1 (aperiodic or bursty signal). Our methods and formulas for quantification, as detailed in Guzmán et al. (2016, 2017a, b) and in Bullock et al. (2018a), are motivated by our desire to characterize different patterns of corpora in a quantifiable way.

5.2

Quantitative Comparisons Between the Corpora

We use these metrics to characterize the nature of CS in the three Spanish-English corpora. Table 1 provides the results of this application. With an M-Index of 0.99, KC contains nearly an equal number of tokens in English (53%) and in Spanish (47%), unlike S7 and M40 for which the number of tokens in one language exceeds that of another. Frequency counts of these corpora based on the language of the

Conceptual and Empirical Arguments for a Language Feature … Table 1 Corpus metrics

103

Corpus

M-Index

I-Index

Burstiness

S7 M40 KC

0.60 0.63 0.99

0.06 0.10 0.17

0.32 0.26 −0.06

token show that English represents the language of the majority (75%) of tokens in S7, while Spanish is the numerically dominant language (74%) in M40. The I-Index quantifies the higher probability of switching languages in KC than in M40 and in S7. Finally, the findings for Burstiness indicate that language alternation is more regular, or periodic, in KC than it is in the corpora S7 and M40, in which language switching occurs in bursts. These metrics allow us to cluster the three corpora into two distinct types of code-switching: insertional and alternational (Muysken 2000; Guzmán et al. 2017a, b). KC is highly bilingual with short, regularly occurring spans of one language after another. This is characteristic of alternational code-switching. But S7 and M40 are asymmetric, characterized as mostly monolingual with intrusion of sporadic other-language tokens. We label this pattern as insertional code-switching. The latter corpora differ in terms of the distribution of the language within them: English serves as the predominant language for S7 while Spanish is the main language of M40. In sum, the result of our metrics combined with a frequency distribution of the tokens of each language within the corpora inform us that we have distinct code-switching signatures in these datasets. We now consider the extent and the nature of code-switching within the DP in these corpora in order to test the validity of the Functional Head Constraint.

6 Quantitative Analysis of Code-Switching in the DP Recall that the FHC is argued to be untenable on empirical grounds because code-switching after the DET is commonly observed across various language pairings. However, Belazi et al. (1994) exempt switches between the DET + NP from the FHC, arguing that NPs are the most frequently borrowed category and that borrowing does not entail a switch in the language feature. By adding POS annotation in the three corpora under analysis here, we can test the frequency of switching in the DP, and by using information pertaining to the length (in words) of monolingual spans, we can examine whether the switching that occurs after the DET is quantitatively different from switching elsewhere. We hypothesize that if switched NPs reflect other-language borrowings, rather than code-switches, they should be realized as significantly shorter spans after determiners than elsewhere. While Poplack et al. (1988) also use length as a criterion for nonce-borrowings, restricting them to insertions of a single token, we rely on relative length, reasoning that NPs can comprise multiword expressions (e.g., “graduate school”) rather than merely bare noun insertions.

104

6.1

B. E. Bullock and A. J. Toribio

Data Preprocessing

We manually tagged KC using the minimal Universal POS Tagset (UT) employed by Google, which incorporates a level of granularity that is adequate for testing our grammatical hypothesis (Petrov et al. 2011). Additionally, as a universal set, it permits structural comparisons between typologically diverse languages. For instance, it was used by Sharma et al. (2016) for Hindi-English social media texts and by Schulz and Keller (2016) to examine Latin-Middle English texts. The tagset includes 12 labels: Verb, Noun, Pronoun, Adjective, Adverb, Adposition, Conjunction, Determiner, Number, Particles, X (Unknown), and Punctuation. We remapped the S7 and M40 to the UT system. The POS tags for S7 were remapped from the modified Penn Treebank tags that had been used by the authors (Solorio and Liu 2008b) and those of M40 were remapped from the CHAT conventions of BilingBank (Deuchar 2010). Following these processing stages, each token in the three corpora bears a language tag (English or Spanish) and one of the 12 POS tags, annotations that conform to the same parameters. For each corpus, we demarcated the beginning and end of every sequence of word tokens bearing the same language label and calculated the size of the resulting span by the number of tokens with it. We identified the POS tag at the start and end of every span, as shown by the example in Table 2. We fit linear regression models with Length (of span) as the dependent variable and with Previous EndPOS and Language of the token immediately preceding the length of span as predictor variables. For instance, in Table 2, the length of span of the switched token “graduate school” is immediately preceded by a DET, labelled as Spn. Thus, length of span is assessed relative to every EndPOS of the span that precedes it. We releveled the reference of the POS variable to “Verb” from the default of punctuation (coded as “.”) since the latter is a natural point to switch the language of an utterance and could have biased the results by accounting for most of the variance in the sample.

6.2

Results

Because each corpus follows a different distribution, as discussed in Sect. 5, we modeled each corpus in a separate linear regression. The models were fit with the lm function in R (R. Core Team 2014). Table 3 provides the counts for every DET-to-NP transition, whether monolingual or switched, where the category NP is Table 2 Coding language spans Span

Length

Language

StartPOS

EndPOS

PreviousEndPOS

Corpus

la graduate school

1 2

Spn Eng

DET ADJ

DET NOUN

. DET

KC KC

Conceptual and Empirical Arguments for a Language Feature …

105

Table 3 DET- to NP-transition counts

Eng DET Spn DET

KC Eng NP

KC Spn NP

S7 Eng NP

S7 Spn NP

M40 Eng NP

M40 Spn NP

498 138

60 353

399 22

3 122

133 56

11 460

Fig. 1 KC transitions

Fig. 2 S7 transitions

defined as a span beginning with a token tagged as NOUN or ADJ. The values from Table 3 provide insight into the relative frequencies of switching between a DET and an NP within each of the corpora. In Table 3, it is clear that switching from an English determiner to a Spanish NP is relatively rare across corpora. The mosaic plots in Figs. 1 through 3 represent these results visually; the size of a tile is dependent on the observations, with larger tiles representing observations with greater counts. The asymmetric balance between Spanish and English in S7 and M40 is reflected in the relative size of the top row (English) relative to the bottom row (Spanish). Comparing between corpora, the frequency of switching can be seen to range from a minimum of .75% between English DET and Spanish NPs in S7 in Fig. 2 to a maximum of 28% between a Spanish DET and English NP in KC in Fig. 1. In KC and in S7, there is proportionally more switching permitted after a Spanish DET than after an English DET, while in M40 (Fig. 3), there appears to be no difference in the relative frequency of switching based on the previous language.

106

B. E. Bullock and A. J. Toribio

Fig. 3 M40 transitions

The results of the best fitting linear regression to predict the length of span based on the immediately preceding POS tag and on previous Language for KC finds a significant regression in which length of span is predicted by the main effects of previous POS and of previous Language (R2 = 0.02133, F(13, 1255) = 3.126, p-value: 0.0001301). Specifically, the average length of span decreased significantly in length (−1.54) only after a DET. This decrease was significant at the 95% confidence level. There is also a significant main effect for Language (p < 0.01) whereby the average length of span increases slightly (+0.73) when the language preceding a switch is Spanish. Recall that, in general, the spans in KC are shorter than in the other two corpora. The best fitting model for S7, predicting the length of span based on the immediately preceding POS tag and on previous Language, also shows significant main effects of previous POS and for previous language (R2 = 0.07241, F-statistic (12, 471) = 4.142, p-value: 3.454e−06). The average length of span in S7 decreases significantly, and by a large average number of tokens (−20.69) at the 95% confidence level, but only if the previous POS tag is a DET. In contrast, a highly significant main effect for Language (p < 0.001) shows that the average length of span increases by 19.83 if the previous language of a span is Spanish. A model fit for M40 with both POS and Language as main effects returns significance only for Language, where a switch to English results in a decrease in the length of span (−8.5) as would be expected in a corpus that is predominantly Spanish with English insertions. A model fit with only POS as a main effect shows a significant decrease (p < 0.05) in average span length (−7.48) again exclusively after the category of DET (R2: 0.0291; F(11, 728) = 3.014, p-value: 0.0006057). In sum, all three corpora demonstrate that the span lengths following DET, and only after DET, show a statistically significant decrease in length relative to the reference level. Additionally, in S7 and KC, English span lengths are increased, in general, when the previous language is Spanish. In other words, the observed insertions of shorter English NPs exclusively after Spanish DETs goes against the general trend in S7 and in KC of having longer English spans following Spanish.

Conceptual and Empirical Arguments for a Language Feature …

6.3

107

Discussion

These results indicate that switching languages in the DP, although not frequent, is not categorically prohibited. We have also confirmed asymmetries in the patterns of alternation in this context. Specifically, English determiners are overwhelmingly followed by English NPs (>90%) across these three corpora, implying that switching after a DET labelled as English is vanishingly rare, irrespective of the proportion of English-Spanish represented in the corpus or of the probability of switching between the languages overall. Yet switching to English NPs after Spanish determiners is attested in all three corpora. Similar asymmetries, in which English NPs are used after other language determiners and classifiers have been reported for Marathi-English (Joshi 1982) and Mandarin-English (Adel et al. 2014). Deuchar and her collaborators propose to account for the language asymmetry in DP-internal switching by reference to the matrix language: The matrix language provides the morphosyntactic frame for the clause (Myers-Scotton 1993) and determines the language of the determiner in mixed DPs (Herring et al. 2010; Blokzijl et al. 2017). But this explanation appears to be untenable for our data. M40 does not present a directional preference with regard to switching in the DP even though verb counts, a proxy for matrix language, show that 75% of the verbs are in Spanish. In contrast to M40, S7 and KC present English as the predominant language, with 75 and 60% of the verbs in English, but demonstrate switching only after Spanish determiners, contrary to predictions. A different account, put forward by Liceras and her colleagues, holds that the directional asymmetry in DP switches follows from the formal properties of the contributing languages: The determiner will be drawn from the language with more robust features, e.g., in the Spanish-English pairing, Spanish contributes the DET, which bears gender (Spradlin et al. 2003; Liceras et al. 2008). Other researchers have raised the possibility that the directionality reflects a social asymmetry, with switching towards the NP of the more socially prestigious language (Blokzijl et al. 2017). While the social proposal may account for the patterns of Spanish-English code-switching in our data (where English is the more prestigious language), we are not in a position to resolve the asymmetrical issue here without key corpora in which Spanish is the prestige language (Bullock et al. 2018b; Toribio 2018). We have also tested the hypothesis that switched NPs should be realized as significantly shorter spans after determiners than elsewhere. The motivation for this hypothesis was that nonce borrowings, rather than code-switches, would be reflected as shorter spans. In alternational corpora, where code-switching occurs more frequently and regularly, we would expect to find shorter spans of each language than we do in insertional corpora. In insertional texts, short insertions from one language are interspersed in longer, monolingual spans of the other. So, insertional corpora are characterizable as monolingual texts with other-language intrusions, while alternational corpora arguably have no well-defined base or matrix language. The results of the regression models demonstrate that spans following a DET are shorter than they are elsewhere, whether the corpus is characterized as

108

B. E. Bullock and A. J. Toribio

alternational (as in KC) or insertional (as in S7 and M40) and whether it is English-dominant (S7), Spanish-dominant (M40), or more balanced (KC). This finding implies that the DET presents a unique grammatical juncture, easily accommodating short other-language NPs from English. The outcome finds support in hierarchies of language contact, which identify NPs as the most frequently borrowed category, and it bolsters the argument for the exclusion of the DP from the FHC on the reckoning that inserted NPs may well be spontaneous borrowings rather than code-switches (Belazi et al. 1994).

7 Conclusions This paper has been concerned with the conceptual considerations of positing language as a grammatical feature and the empirical consequences of tagging lexical items with a language label. We advocate for the use of automatic tools and computational techniques to process mixed language corpora in ways that make statistical comparisons between them possible. The methods that we have used here are language independent and it is our hope that future researchers employ them to test our proposals in a broad range of languages. Using such methods, we have demonstrated the utility of a language label, or tag, at the lexical level to quantify the differences between bilingual corpora of the same language pairing. With this information, we have computed the distribution of these language labels in alternating spans and showed that the length of span after a language switch is dependent on grammatical category such that switched spans are significantly shorter after determiners than elsewhere. We suggest that this reflects nonce borrowing rather than code-switching, the alternation of language understood as an abstract grammatical feature. Our analysis is consistent with the spirit of Herchensohn’s Constructionism proposal: Languages are not fully differentiated by learners (or code-switched by bilinguals) unless or until they contain features that serve to formally classify their grammars as separate systems.

References Adamou, E. 2016. A corpus-driven approach to language contact: Endangered languages in a comparative perspective, vol. 12. Walter de Gruyter GmbH & Co KG. Adel, H., K. Kirchhoff, D. Telaar, N.T. Vu, T. Schlippe, and T. Schultz. 2014. Features for factored language models for code-Switching speech. In SLTU, 32–38. Appel, R., and P. Muysken. 1987. Language contact and bilingualism. London; Baltimore, Md., USA: Edward Arnold. Baldwin, T., and M. Lui. 2010. Language identification: The long and the short of the matter. In Human language technologies: The 2010 annual conference of the North American chapter of the Association for Computational Linguistics, 229–237. Association for Computational Linguistics.

Conceptual and Empirical Arguments for a Language Feature …

109

Bandi-Rao, S., M. den Dikken, and J. MacSwan. 2014. Light switches: On v as a pivot in codeswitching, and the nature of the ban on word-internal switches. In Grammatical theory and bilingual codeswitching, 161–183. Cambridge, MA: MIT Press. Barman, U., A. Das, J. Wagner, and J. Foster. 2014. Code mixing: A challenge for language identification in the language of social media. In Proceedings of the first workshop on computational approaches to code switching, 13–23. Beebe, L.M. 1977. The influence of the listener on code-switching. Language Learning 27 (2): 331–339. Belazi, H.M., E.J. Rubin, and A.J. Toribio. 1994. Code switching and X-Bar theory: The functional head constraint. Linguistic Inquiry 25 (2): 221–237. Barnett, R., Codó, E., Eppler, E., Forcadell, M., Gardner-Chloros, P., van Hout, and R., Sebba, M. 2000. The LIDES Coding Manual: A Document for Preparing and Analyzing Language Interaction Data Version 1.1–July 1999. International Journal of Bilingualism, 4 (2): 131–271. Bhatt, R.M. 2014. Argument licensing in optimal switches. In Grammatical theory and bilingual codeswitching, ed. J. MacSwan, 135–158. Cambridge, MA: MIT Press. Blokzijl, J., M. Deuchar, and M. Couto. 2017. Determiner asymmetry in mixed nominal constructions: The role of grammatical factors in data from Miami and Nicaragua. Languages 2 (4): 20. Brysbaert, M., and B. New. 2009a. Moving beyond Kučera and Francis: A critical evaluation of current word frequency norms and the introduction of a new and improved word frequency measure for American English. Behavior Research Methods 41 (4): 977–990. Brysbaert, M., and B. New. 2009b. Subtlexus: American word frequencies. Http:/Subtlexus. Lexique. Org. Bullock, B.E., and A.J. Toribio. 2004. Introduction: Convergence as an emergent property in bilingual speech. Bilingualism: Language and Cognition. Special Issue: Bilingualism and Linguistic Convergence, 7 (2), 91–93. Bullock, B.E., and A.J. Toribio. 2018. Sociolinguistics of bilingualism. In An introduction to bilingualism: Principles and processes, 2nd ed., ed. J. Altarriba and R.R. Heredia, 300–316. New York, N.Y.: Lawrence Erlbaum Associates. Bullock, B.E., G. Guzmán, V. Sharath, J. Serigos, and A.J. Toribio. 2018a. Predicting the presence of a Matrix Language in code-switching. In Proceedings of the third workshop on computational approaches to linguistic code-switching, 68–75. Association for Computational Linguistics. Bullock, B.E., G.A. Guzmán, J. Serigos, and A.J. Toribio. 2018b. Should code-switching models be asymmetric? In Interspeech 2018, 2534–2538. Byers-Heinlein, K. 2014. Languages as categories: Reframing the “One Language or Two” question in early bilingual development. Language Learning 64 (s2): 184–201. Cantone, K.F., and N. Müller. 2008. Un nase or una nase? What gender marking within switched DPs reveals about the architecture of the bilingual language faculty. Lingua 118 (6): 810–826. Cavnar, W.B., and J.M. Trenkle. 1994. N-gram-based text categorization. In Proceedings of third annual symposium on document analysis and information retrieval, vol. 48113, no. 2, 161–175. Çetinoğlu, O., S. Schulz, and N.T. Vu. 2016. Challenges of computational processing of code-switching. Presented at the EMNLP, Austin, TX. Van Coetsem, F. 1988. Loan phonology and the two transfer types in language contact. Dordrecht: Foris Publications. Das, A., and B. Gambäck. 2014. Identifying languages at the word level in code-mixed Indian social media text. Deuchar, M. (2010). BilingBank Spanish-English Miami Corpus. Di Sciullo, A.-M., P. Muysken, and R. Singh. 1986. Government and code-mixing. Journal of Linguistics 22 (1): 1–24. Diab, M., and A. Kamboj. 2011. Feasibility of leveraging crowd sourcing for the creation of a large scale annotated resource for Hindi English code switched data: A pilot annotation. DTIC Document.

110

B. E. Bullock and A. J. Toribio

Dijkstra, T., and W.J.B. Van Heuven. 2002. The architecture of the bilingual word recognition system: From identification to decision. Bilingualism: Language and Cognition 5 (3): 175–197. Dubois, S., and S. Noetzel. 2005. Intergenerational pattern of interference and internally-motivated changes in Cajun French. Bilingualism 8 (2): 131–143. Finkel, J.R., T. Grenager, and C. Manning. 2005. Incorporating non-local information into information extraction systems by Gibbs Sampling. In Proceedings of the 43rd annual meeting on association for computational linguistics, 363–370. Association for Computational Linguistics. Franco, J.C., and T. Solorio. 2007. Baby-steps towards building a Spanglish language model. In International conference on intelligent text processing and computational linguistics, 75–84. Springer. Francom, J.C. 2013. ACTIV-ES: A novel Spanish-language corpus for linguistic and cultural comparisons between communities of the Hispanic world. Francom, J., M. Hulden, and A. Ussishkin. 2014. ACTIV-ES: A comparable, cross-dialect corpus of ’everyday’ Spanish from Argentina, Mexico, and Spain. In LREC, 1733–1737. Goh, K.-I., and A.-L. Barabási. 2008. Burstiness and memory in complex systems. EPL (Europhysics Letters), 81 (4): 48002. Green, D.W. 1998. Mental control of the bilingual lexico-semantic system. Bilingualism: Language and Cognition 1 (2): 67–81. Grefenstette, G. 1995. Comparing two language identification schemes. In 3rd international conference on statistical analysis of textual data, 11–13. Guzmán, G.A., J. Serigos, B.E. Bullock, and A.J. Toribio. 2016. Simple tools for exploring variation in code-switching for linguists. In EMNLP 2016, 2–20. Guzmán, G.A., J. Ricard, J. Serigos, B. Bullock, and A.J. Toribio. 2017a. Moving code-switching research toward more empirically grounded methods. In CDH2017: Corpora in the Digital Humanities, 1–9. Guzmán, G., J. Ricard, J. Serigos, B.E. Bullock, and A.J. Toribio. 2017b. Metrics for modeling code-switching across corpora. In Proceedings of the Interspeech 2017, 67–71. Herring, J.R., M. Deuchar, M.C.P. Couto, and M.M. Quintanilla. 2010. ‘I saw the madre’: Evaluating predictions about codeswitched determiner-noun sequences using Spanish–English and Welsh–English data. International Journal of Bilingual Education and Bilingualism 13 (5): 553–573. Herschensohn, J. 2000. The second time around minimalism and L2 acquisition, vol. 21. John Benjamins Publishing. Jake, J.L., C. Myers-Scotton, and S. Gross. 2002. Making a minimalist approach to codeswitching work: Adding the matrix language. Bilingualism: Language and Cognition 5 (1): 69–91. Jarvis, S., and A. Pavlenko. 2007. Crosslinguistic influence in language and cognition. New York and London: Routledge. Joshi, A. K. 1982. Processing of sentences with intra-sentential code-switching. In Proceedings of the 9th conference on Computational linguistics, vol. 1, pp. 145–150. Academia Praha. Jurgens, D., Y. Tsvetkov, and D. Jurafsky. 2017. Incorporating dialectal variability for socially equitable language identification. In Proceedings of the 55th annual meeting of the Association for Computational Linguistics, Volume 2: Short papers, vol. 2, 51–57). Khattab, G. 2013. Phonetic convergence and divergence strategies in English-Arabic bilingual children. Linguistics 51 (2): 439–472. King, B., and S. Abney. 2013. Labeling the languages of words in mixed-language documents using weakly supervised methods. In Proceedings of NAACL-HLT, 1110–1119. Köppe, R., and J.M. Meisel. 1995. Code-switching in bilingual first language acquisition. In One speaker, two languages: Cross-disciplinary perspectives on code-switching, ed. L. Milroy and P. Muysken, 276–301. New York: Cambridge University Press. Li, Y., and P. Fung. 2012. Code-switch language model with inversion constraints for mixed language speech recognition. In COLING, 1671–1680.

Conceptual and Empirical Arguments for a Language Feature …

111

Li, Y., and P. Fung. 2013. Improved mixed language speech recognition using asymmetric acoustic model and language model with code-switch inversion constraints. In 2013 IEEE international conference on acoustics, speech and signal processing (ICASSP), 7368–7372. IEEE. Liceras, J.M., R.F. Fuertes, S. Perales, R. Pérez-Tattam, and K.T. Spradlin. 2008. Gender and gender agreement in bilingual native and non-native grammars: A view from child and adult functional–lexical mixings. Lingua 118 (6): 827–851. Lignos, C., and M. Marcus. 2013. Toward web-scale analysis of codeswitching. In Proceedings of annual meeting of the Linguistic Society of America. López, L. (ed.). 2018. Code-switching: Theoretical questions, experimental answers. A festschrift presented to Kay González-Vilbazo presented by his colleagues and students. Amsterdam: John Benjamins. Lui, M., and T. Baldwin. 2012. langid. py: An off-the-shelf language identification tool. In Proceedings of the ACL 2012 system demonstrations, 25–30. Association for Computational Linguistics. Lui, M., J.H. Lau, and T. Baldwin. 2014. Automatic detection and language identification of multilingual documents. Transactions of the Association for Computational Linguistics 2: 27–40. MacSwan, J. 2000. The architecture of the bilingual language faculty: Evidence from intrasentential code switching. Bilingualism: Language and Cognition 3 (1): 37–54. MacSwan, J. 2004. Code switching and grammatical theory. In The handbook of bilingualism, ed. T.K. Bhatia and W.C. Ritchie, 283–311. Oxford: Blackwell. MacSwan, J. 2014. Grammatical theory and bilingual codeswitching. MIT Press. Maharjan, S., E. Blair, S. Bethard, and T. Solorio. 2015. Developing Language-tagged Corpora for Code-switching Tweets. In LAW@ NAACL-HLT, 72–84. Mahootian, S. 1993. A null theory of code switching. Dissertation, Northwestern University, Evanston, IL. Mahootian, S., and B. Santorini. 1996. Code switching and the complement/adjunct distinction. Linguistic Inquiry 27 (3): 464–479. Matras, Y., and J. Sakel. 2007. Grammatical borrowing in cross-linguistic perspective, vol. 38. Walter de Gruyter. Mougeon, R., T. Nadasdi, and K. Rehner. 2005. Contact-induced linguistic innovations on the continuum of language use: The case of French in Ontario. Bilingualism: Language and Cognition 8 (2): 99–115. Muysken, P. 2000. Bilingual speech: A typology of code-mixing. Cambridge: Cambridge University Press. Muysken, P. 2002. Computation and storage in language contact. In Storage and computation in the language faculty, ed. E. Nooteboom, F. Weerman, and F. Wijnen, 157–179). Dordrecht: Kluwer. Myers-Scotton, C. 1993. Dueling languages: Grammatical structure in codeswitching. Oxford: Oxford University Press (Clarendon Press). Myers-Scotton, C. 2002. Contact linguistics: Bilingual encounters and grammatical outcomes. Oxford: Oxford University Press. Nguyen, D.-P., and A.S. Dogruoz. 2013. Word level language identification in online multilingual communication. Association for Computational Linguistics. Parafita Couto, M.C., and H. Stadthagen-González. 2017. El book or the libro? Insights from acceptability judgments into determiner/noun code-switches. International Journal of Bilingualism. Petrov, S., D. Das, and R. McDonald. 2011. A universal part-of-speech tagset. ArXiv:1104.2086. Pfaff, C.W. 1979. Constraints on language mixing: Intransentential code-switching and borrowing in Spanish/English. Language 55 (2): 291–318. Poplack, S., D. Sankoff, and C. Miller. 1988. The social correlates and linguistic processes of lexical borrowing and assimilation. Linguistics 26 (1): 47–104.

112

B. E. Bullock and A. J. Toribio

Poplack, S., L. Zentz, and N. Dion. 2012. What counts as (contact-induced) change. Bilingualism: Language and Cognition 15 (02): 247–254. Post, R.E. 2010. Code-switching in the determiner phrase: a comparison of Tunisian Arabic-French and Moroccan Arabic-French switching. Master’s Thesis, University of Texas at Austin. R. Core Team. 2014. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Radford, A., T. Kupisch, R. Köppe, and G. Azzaro. 2007. Concord, convergence and accommodation in bilingual children. Bilingualism: Language and Cognition 10 (3): 239–256. Ritter, E. (1991). Two functional categories in noun phrases: Evidence from Modern Hebrew. In Perspectives on phrase structure: Heads and licensing, ed. S. Rothstein, 37–62. San Diego: Academic Press. Roeper, T. 1999. Universal bilingualism. Bilingualism: Language and Cognition, 2 (3): 169–186. Rosner, M., and P.-J. Farrugia. 2007. A tagging algorithm for mixed language identification in a noisy domain. In Interspeech, 190–193. Rubin, E.J., and A.J. Toribio. 1995. Feature-checking and the syntax of language contact. In Amsterdam studies in the theory and history of linguistic science series, vol. 4, 177–177. Sankoff, D., and S. Poplack. 1980. A formal grammar for code-switching. In CUNY Working Papers, vol. 8. Centro de Estudios Puertorriqueños. Schulz, S., and M. Keller. 2016. Code-switching Ubique Est-language identification and part-of-speech tagging for historical mixed text. In Proceedings of the 10th SIGHUM workshop on language technology for cultural heritage, social sciences, and humanities, Proceedings of LaTeCH, 43–51. Sharma, A., S. Gupta, R. Motlani, P. Bansal, M. Srivastava, R. Mamidi, and D.M. Sharma. 2016. Shallow parsing pipeline for Hindi-English code-mixed social media text. ArXiv:1604.03136. Sibun, P., and J.C. Reynar. 1996. Language identification: Examining the issues. Solorio, T., E. Blair, S. Maharjan, S. Bethard, M. Diab, M. Gohneim, et al. 2014. Overview for the first shared task on language identification in code-switched data. In Proceedings of the first workshop on computational approaches to code switching, 62–72. Solorio, T., and Y. Liu. 2008a. Learning to predict code-switching points. In Proceedings of the conference on empirical methods in natural language processing, 973–981. Association for Computational Linguistics. Solorio, T., and Y. Liu. 2008b. Part-of-speech tagging for English-Spanish code-switched text. In Proceedings of the conference on empirical methods in natural language processing, 1051– 1060. Association for Computational Linguistics. Spradlin, K., Liceras, J., and Fernández-Fuertes, R. 2003. The grammatical features spell-out hypothesis as a diagnostic for bilingual competence. Paper presented at Presented at the The 4th International Symposium on Bilingualism (ISB4), Arizona State University, April 30-May 3, 2003. Thomason, S.G., and T. Kaufman. 1988. Language contact, creolization, and genetic linguistics. Berkeley, CA: University of California Press. Toribio, A.J. 2001. On the emergence of bilingual code-switching competence. Bilingualism: Language and Cognition 4 (3): 203–231. Toribio, Almeida Jacqueline. 2018. The future of code-switching research. Code-switching: Theoretical questions, experimental answers. In A festschrift presented to Kay González-Vilbazo presented by his colleagues and students, ed. L. López, 257–267. Amsterdam: John Benjamins. Treffers-Daller, J. 2005. Brussels French une fois: Transfer-induced innovation or system-internal development? Bilingualism: Language and Cognition 8 (2): 145–157.

Conceptual and Empirical Arguments for a Language Feature …

113

Vu, N.T., H. Adel, and T. Schultz. 2013. An investigation of code-switching attitude dependent language modeling. Statistical Language and Speech Processing 297–308. Weinreich, U. 1953. Languages in contact: findings and problems. New York; The Hague: Linguistic Circle of New York; Mouton. Yamaguchi, H., and K. Tanaka-Ishii. 2012. Text segmentation by language using minimum description length. In Proceedings of the 50th annual meeting of the association for computational linguistics: Long papers, vol. 1, 969–978. Association for Computational Linguistics.

On the Reduction of /ʒ/ in a Minority North American Variety of French Randall Gess

Abstract This paper provides a detailed examination of the acoustic properties of full and reduced forms of /ʒ/ produced by a francophone Canadian from British Columbia, including the phonological contexts in which these forms occur. Full and reduced variants form a continuum of realizations defined by aperture, ranging from full constriction (and a specified degree of voicelessness) to complete absence of constriction in the oral cavity, including also a null realization. The presence of [ɦ] and [h] on the continuum motivates our view of the process as one of gradient debuccalization—the progressive reduction of oral articulatory gestures, with realizations defined by the articulatory residue remaining after reduction (some of which may show degrees of place assimilation). Our data are considered in light of Kingston’s (2008) proposal that lenited articulations serve to mark continuity within prosodic domains, with special attention to the forms that challenge this claim. We also explore the possible factors that make /ʒ/ a target for reduction.




Keywords Debuccalization /ezh/ (voiced postalveolar fricative) Acoustic properties Articulatory gestures





French


1 Introduction Just over 200 km north of the University of Washington, in British Columbia’s Lower Mainland, lies a community called Maillardville, with a rich francophone history of over 100 years, and a significant and still-active francophone population (although now representing a small percentage of the area’s total population). It is the westernmost community in North America with such an important legacy, and yet it preserves many features of eastern varieties of Canadian French, including at least one that is relatively uncommon (i.e., is sociolinguistically limited) in the originating areas. This feature is the so-called saintongeais—a debuccalized R. Gess (&) Department of French, Carleton University, Ottawa, ON, Canada e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_6

115

116

R. Gess

variant of the fricative /ʒ/ (and of /ʃ/ in some varieties, although not in this one, according to the data available) which is assumed by some to have its ultimate source in the French provinces of Saintonge and Poitou (today’s Charente-Maritime and parts of Charentes and Deux-Sèvres) (Pignon 1960; Hull 1966, 1968; Léon 1967; Gendron 1970; Juneau 1972; Poirier 1988, as cited in Bittner (1995)), although one cannot eliminate the possibility of it occurring independently in the new world varieties. I begin this chapter with a brief review of what we know about the geographical and social distribution of debuccalized /ʒ/ (Sect. 2). Following a brief profile of the speaker who manifests this variant in his speech and an overview of the methodology used (Sect. 3), I will provide a detailed description of the acoustic properties of his full and reduced forms of /ʒ/ (debuccalized and other reduced variants) in a variety of phonological contexts (Sect. 4). I will then discuss how best to characterize the process, considering also the range of variants of /ʒ/ reported elsewhere, and how this characterization relates to treatments of lenition in the literature (Sect. 5), and finally I will discuss what may make /ʒ/ susceptible to reduction (Sect. 6). In Sect. 7, I will provide some concluding remarks, including some areas for future exploration. For now, I use the term debuccalized to refer to a variant perceived as [ɦ] or [h]. The use of the term to refer to the process encompassing all reduced realizations will be justified in Sect. 5. Until then, the term reduction will refer to the full range of realizations, including [ɦ] and [h].

2 Geographic and Social Distribution of Debuccalized /ʒ/ Besides still occurring in the region of Charentes, at least until fairly recently (Chidaine 1967; Horiot 1990; Gallard 1990, as cited in Bittner (1995)), debuccalized /ʒ/ is virtually unlimited in terms of geography in the province of Quebec (Tassé 1981), and is attested as well in Ontario, in Manitoba, in the Maritime provinces, as well as in Louisiana and the Caribbean (Hull 1966, 1968; Holder 1972; Thogmartin 1974; Massignon 1962; Lucci 1972; Klingler and Lyche 2012; Dajko and Blainey 2016). My study is the first to my knowledge to place the phenomenon west of Manitoba, although one might infer from the background of my speaker (see Sect. 3) that it also had at least a fleeting existence in Saskatchewan. Despite enjoying a wide geographic distribution, the debuccalization of /ʒ/ is nevertheless highly restricted sociolinguistically. In the most thorough sociolinguistic study of the phenomenon, Hansen (1988) found that debuccalization (what she and others refer to as “aspiration”, due to its perceptual similarity to [ɦ] in some contexts) could be most closely associated with the speech of older men of a lower socio-economic status.

On the Reduction of /ʒ/ in a Minority North American …

117

Hansen’s results must be considered in light of several issues. First, the rate of debuccalization ranges across her eight subjects from 2 to 51% (and the highest rate was in fact in the speech of the youngest of the four men (age 32)). Second, Hansen’s study was limited to one urban area in Quebec (Hull), although her general findings are largely consistent with prior observations from other areas (Vincent and Sankoff 1975; Tassé 1981). Finally, an earlier study (Charbonneau 1957), found regular production of debuccalized forms amongst 12–20-year-old male students in the Joliette region, with an important variation in reduced realizations. Differences between the two studies’ results may represent a change over the time period in question, they may be due to the limited number of subjects in Hansen’s (1988) study (only four between ages 15 and 24, two of whom were female), or they may simply be due to the different locations of the studies. (Marie-Hélène Côté (p.c.) has noted a high incidence of debuccalization in the Joliette region in the context of supervised fieldwork for the Phonology of Contemporary French project (see below), results of which are not yet available.)

3 Speaker Profile and Methodology 3.1

Speaker Profile

Hansen’s (1988) finding that debuccalization appears most closely associated with older working class males is consistent with the profile of the speaker from Maillardville (ccajl1) who produces debuccalized /ʒ/. Born in 1923, ccajl1 was in his mid sixties at the time of Hansen’s study, like two of her male subjects. They had rates of debuccalization of 10 and 21%, while ccajl1’s rate falls towards the lower end of that range.1 Our speaker is one of 12 participants in the survey point of the Phonology of Contemporary French (PFC) project (http://public.projet-pfc.net) that I conducted in Maillardville in 2006, making him 83 years old at the time of recording. ccajl1 was born in the francophone community of Willow Bunch, Saskatchewan to a mother from Quebec and a father born in Rhode Island of parents from Quebec. His schooling was in a francophone Catholic school and lasted until the age of 14, the age at which he moved to Maillardville. ccajl1 held two occupations before retirement, as a distiller (until the major distillery moved out of province) and then as a maintenance supervisor.

1

I have calculated the rate at 12% (13 out of 1070 tokens), but there are ambiguous cases as discussed in the next section. Determination of a more precise rate of debuccalization would depend on the perceptual judgments of a number of native speakers—a study for a later date.

118

3.2

R. Gess

Methodology

The data presented here were collected following the PFC protocol, which includes the reading of two word lists,2 a reading passage, a guided interview and free conversation. The four conditions are intended to elicit speech on a continuum from highly monitored to unmonitored. The recordings were done in a relatively quiet room in a retirement home in the centre of Maillardville that caters to francophone seniors. The speaker wore a head-mounted microphone so that the microphone was always at the same distance from his mouth, despite any head or body movements, and the recording was done using a PMD-660 digital recorder. The high-quality recordings were then analyzed using Praat (Boersma and Weenink 2016).

4 Acoustic Properties of Full and Reduced Forms What I present in this section constitutes the first detailed acoustic study of full versus reduced variants of /ʒ/ in the production of a single speaker. A previous study (Bittner 1995) compared reduced forms from speakers who had high rates of reduction with full forms from speakers with low rates of reduction. An earlier instrumental study (Charbonneau 1957) used kymographic data to infer articulatory aspects of the production of /ʒ/, and yet another, on consonant production in general (and principally on consonant groups), touched on articulatory aspects of both /ʒ/ and /ʃ/ as well (Rochette 1973). The tokens of /ʒ/ we will examine are from the following phonological contexts: • intervocalic, word-internal full and reduced tokens in ‘aujourd’hui’ (Sect. 4.2) • intervocalic, word-initial full token in ‘presque jour’ versus reduced token in ‘trois jours’ (Sect. 4.3) • phrase-initial, pre-vocalic full token in ‘j’ai été’ versus reduced token in ‘justement’ (Sect. 4.4) • phrase-initial, pre-sonorant ‘j’me suis’—reduced token only (Sect. 4.5) • phrase-initial, pre-obstruent full token in ‘j’parlerai’ versus reduced token in ‘j’pensais’ (Sect. 4.6) • word-final, pre-C[+voice] full token in ‘neige dans’ versus reduced token in ‘rouges, b-bondés’ (Sect. 4.7) Before turning to these examples of full and reduced forms, I will first discuss an important feature in the production of /ʒ/, which is voicing.

2

Canadian survey points use two word lists, the second of which is specifically designed to test for features associated with Canadian varieties. The first word list is common to all survey points.

On the Reduction of /ʒ/ in a Minority North American …

4.1

119

An Important Note on Voicing

Despite being positive for the phonological feature [±voice], /ʒ/ is regularly produced with an important degree of voicelessness. Our speaker produces tokens with significant voicelessness across conditions, from the most careful speech style (word lists, 79%) to the most casual (54 and 52% for the guided interview and free conversation (based on 6 min of each)). In addition to occurring across speech conditions, voicelessness also occurs across phonological environments in which /ʒ/ appears. Table 1 shows the average percentage of unvoiced frames for each phonological context in which relevant tokens occur. This measure was taken using Praat’s voice report (tokens from the total 19 min of free conversation). Although voicelessness occurs across the environments in which /ʒ/ appears, this does not mean that that each token in a given environment includes a period of voicelessness. For example, in the second context ([…V_#V…]), one of the two tokens is fully voiced, while the other is 15.15% unvoiced. The range in the first environment ([…V_V…]) is from fully voiced (ten tokens) to 28% unvoiced (in the proper noun, ‘Roger’). The reason why voicelessness occurs in the realization of /ʒ/ is fairly well understood. The production of a voiced fricative involves two competing needs: the need to maintain voicing, which is facilitated by low pressure in the oral cavity; and the need to produce frication noise, which requires high pressure in the oral cavity. As Ohala (1983) explains, “[t]o the extent that the segment retains voicing it may be less of a fricative, and if it is a good fricative it runs the risk of being devoiced”. As Hualde and Prieto (2014) point out, “[d]iachronically, this may lead to the recategorization of voiced fricatives as voiceless”. While I do not wish to claim that the phoneme /ʒ/ has been recategorized as voiceless for this speaker, I do assume that Table 1 Average percentage of locally unvoiced frames by context

Environment

n

% unvoiced

Example

[…V_V…] […V_#V…] […V#_V…] [… C[+vc]#_V…] […V_#C[+vc]…] […r_V…] [pause #_V…] […V_G…] […V_# pause] [pause #_C[+vc]…] [C[+vc]#_C[−vc]…] [pause #_C[−vc]…] […V#_C[−vc]…] […V_#C[−vc]…]

14 2 43 5 4 6 10 1 1 1 3 4 9 2

5.47 7.58 22.16 24.72 31.30 37.7 48.04 48.49 56.25 56.52 80.17 81.75 83.33 85

aujourd’hui college à les jeunes encore j’avais l’âge de argent J’en avais régional recyclage J’me suis adressé comme j’te dis J’parlerai pas pis j’ toussais engage quelqu’un

120

R. Gess

Table 2 Voicing in reduced forms

Example

% unvoiced

‘trois jours’ ‘trois jours’ ‘aujourd’hui’ ‘des soeurs, je les aime’ ‘quand j’ai arrivé ‘quand j’attrapais’ ‘euh, j’avais’ ‘aujourd’hui’ ‘rouge, b-bondé’ ‘demain, je sais pas’ ‘avec euh … justement’ ‘Je me suis’ ‘J’pensais’

0 0 0 0 0 15.15 15.39 19.23 22.22 29.63 50 56.52 100

the conflicting aerodynamic constraints on its realization are relevant to the establishment of production targets. For now, let us simply note the variation in rates of voicelessness in the production of /ʒ/ across phonological contexts, with lower rates of voicelessness (i.e., fuller voicing) corresponding to contexts with voiced elements both preceding and following /ʒ/. We have thus far established that tokens of /ʒ/ are often produced with some degree of voicelessness, with likelihood of devoicing dependent on phonological context. This holds true also if we are considering only reduced forms, as shown in Table 2. It is noteworthy that there is no contextual [−voice] trigger for two of the forms with the highest amount of voicelessness, ‘justement’ and ‘Je me suis’. Rather, the high degree of voicelessness seems related to the initial (post-pausal) position of the affected fricative. This suggests that the production target may have a glottal configuration that naturally impedes full phonological voicing (i.e., that induces devoicing).

4.2

Full and Reduced Forms in Intervocalic, Word-Internal Position

The first context we examine is intervocalic, word-internal position, with both tokens coming from the same word, ‘aujourd’hui’. The full form is shown in Fig. 1, and the reduced form in Fig. 2. A summary of the differences in their acoustic profiles follows. For both forms, the segment in question falls mostly within the large oval. In both the full and reduced forms, we see evidence of source mixing during the production of the segment, in the upper rectangle of the two comprising the

On the Reduction of /ʒ/ in a Minority North American …

Fig. 1 Full form in intervocalic, word-internal: ‘aujourd’hui’ (415)

Fig. 2 Reduced form in intervocalic, word-internal: ‘aujourd’hui’ (473)

121

122

R. Gess

spectrogram. There is both a noticeable waveform and the presence of aperiodic noise superimposed on it. In the full form, the aperiodic noise is much stronger than in the reduced form. This is reflected in the noise-to-harmonics ratio of 0.696 for the full form, versus 0.206 for the reduced one. The full form is also partially devoiced, with a noticeable absence of voice bars on the spectrogram (bottom of lower rectangle), and a percentage of voicelessness according to Praat’s Voice Report of 22.73% unvoiced, compared to the reduced form’s 0%. Another difference between the full form and the reduced form is that the former has an intensity that is sharply lower than that of surrounding vowels, whereas the latter manifests only a low drop in intensity compared to vowels before and after. This is indicated by the light line near the middle of the spectrogram, which approaches a U shape in Fig. 1, but which is rather flat in Fig. 2. The reduced form also shows some continuity in formant structure during its realization compared to the full form, where formant structure appears disrupted. This is indicated by the dotted lines in Fig. 2 versus the more scattered dots in Fig. 1. Finally, there is a difference in duration between the full and reduced forms, with the former representing 34.5% of the VCV interval, while the latter represents only 25.9% of the VCV interval. For the reduced form in Fig. 2, I would argue for a transcription as [ʝ], indicating a constriction in the palatal area, but not one involving the tongue apex and the alveolar ridge (the full form involves constrictions in both areas as we discuss in Sect. 6). I would expect that the percept of the reduced form comes from contact between the sides of the tongue and the teeth, with a concave top of the tongue and retracted apex, but articulatory study would be needed to verify this (see Sect. 7).

4.3

Full and Reduced Forms in Intervocalic, Word-Initial Position

In intervocalic, word-initial position, we have a full form in ‘jour’ (‘presque jour et nuit’), shown in Fig. 3, and a reduced form in ‘jours’ (‘trois jours pour revenir’), shown in Fig. 4. In the full form (Fig. 3), the source mixing is again characterized by strongly evident aperiodic noise superimposed on the waveform. Aperiodicity is difficult to perceive visually in the waveform of the reduced form (Fig. 4). Again this difference is reflected in the noise-to-harmonics ratios of the two tokens, which is at 0.411 for the full form, and 0.055 for the reduced one. Similarly, the full form manifests some devoicing, even if only slight (3.23%), while the reduced form is fully voiced. As was the case when the token was in word-internal position, we see a very different profile with respect to intensity drop between the full and reduced forms. In the full form, there is a very noticeable transition into and out of the fricative, quite a bit more noticeable coming out of it in this case (the syllable beginning with

On the Reduction of /ʒ/ in a Minority North American …

123

Fig. 3 Full form in intervocalic, word-initial: ‘jour’ (334)

Fig. 4 Reduced form in intervocalic, word-initial: ‘jours’ (371)

/ʒ/ is accentually prominent in the string ‘presque jour et nuit’). In the reduced form, the intensity line is again much flatter. Similarly, the full form also shows a very noticeable interruption in formant structure in the F1–F3 regions, whereas in the reduced one the formants are much more continuous. Finally, in terms of duration,

124

R. Gess

the full form this time is longer at 0.114 versus 0.084 ms, but constitutes only 39.4% of the VCV interval versus 49.7% for the reduced form. Again, the accentual profile of the collocation would account for this difference with respect to the word-internal cases, where the /ʒ/ begins an unstressed syllable in both cases. For the reduced form in Fig. 4, which represents a percept appropriate to the label “aspirated”, a transcription as [ɦ] is called for. This is entirely compatible with the strong continuity of the formant structure from the preceding vowel and into the following one.

4.4

Full and Reduced Forms in Phrase-Initial, Pre-vocalic Position

In the following figures, we see tokens in phrase-initial and pre-vocalic position, in the sequences ‘j’ai été’ (Fig. 5) and ‘justement’ (Fig. 6). In phrase-initial, pre-vocalic position, both forms have a relatively high noise-to-harmonics ratio, but it is the reduced form with the higher one, unlike what we saw in intervocalic position (word-internal and word-initial). In this case, the full form’s ratio is at 0.404 whereas the reduced form’s is at 0.453. In the spectrograms we see a clear dominance of aperiodic noise, with just a hint of periodicity and the beginning and end of the form in Fig. 5, and not even such a hint in Fig. 6. Note that in Fig. 5, the fricative is preceded by a period of pre-voicing, with clear voice bars apparent on the spectrogram and waveform (the long vertical lines).

Fig. 5 Full form in phrase-initial, pre-vocalic position: ‘j’ai été’

On the Reduction of /ʒ/ in a Minority North American …

125

Fig. 6 Reduced form in phrase-initial, pre-vocalic position: ‘justement’

There is no visible voicing in the reduced form, although it measures as only 50% voiceless according to Praat’s Voice Report, compared to 39.45% for the full form, which also has the period of pre-voicing. In terms of intensity, the full form sees a rapid increase from 0 during the period of pre-voicing and increases ever so slightly during the period of frication, rising again, significantly, at the beginning of the following vowel. The reduced form sees a more gradual rise from 0 during the frication noise, approximating a straightforward extrapolated line between 0 and the peak of the following vowel. The reduced form is not surprisingly less intense, at 66.02 dB, versus 73.84 dB for the full form. The reduced form is also shorter, at 64% of the CV interval, versus 70.8% for the full form. In the case of phrase-initial /ʒ/ in ‘justement’, the percept suggests a transcription as [h], not inconsistent with the degree of voicelessness measured (50%) in Voice Report.

4.5

Reduced Form in Phrase-Initial, Pre-sonorant Position

A comparison cannot be made between full and reduced forms in phrase-initial, pre-sonorant position, as there is only one token of /ʒ/ which occurs before a sonorant consonant, and which is reduced. The example appears in the sequence ‘Je me suis adressé’, and is shown in Fig. 7.

126

R. Gess

Fig. 7 Reduced form in phrase-initial, pre-sonorant position: ‘je me suis adressé’

In this case, the target segment, not very salient on the spectrogram, is produced with 56.52% voicelessness, according to Praat’s Voice Report. Associated with the segment is very low-intensity noise that is fairly diffuse over the lower frequency range. As in the preceding case, the percept here is most appropriately transcribed as [h].

4.6

Full and Reduced Forms in Phrase-Initial, Pre-obstruent Position

In Figs. 8 and 9, we compare full and reduced forms of /ʒ/ in phrase-initial, pre-obstruent position. In each case, the following obstruent is /p/. In this case, the difference between the full and reduced forms is very simply the presence versus absence of noise. The non-realization of /ʒ/ is transcribed as Ø. The realization of the full form in Fig. 8 is voiceless (81.75% according to Voice Report), so transcribed as [ʃ].

On the Reduction of /ʒ/ in a Minority North American …

Fig. 8 Full form in phrase-initial, pre-obstruent position: ‘je parlerai pas’

Fig. 9 Reduced form in phrase-initial, pre-obstruent position: ‘ pensais moi que je’

127

128

4.7

R. Gess

Full and Reduced Forms in Word-Final Position (Before a Voiced Consonant)

Our speaker also produced full and reduced tokens of /ʒ/ in word-final position. We compare, in Figs. 10 and 11, instances of the token before a voiced consonant, in ‘neige dans’ and ‘rouges b-bondés’ (the speaker stutters slightly over the initial stop of the second word). This case is similar to the preceding one, in that the most salient difference between the full and reduced forms is the presence versus absence of audible noise. However, in this case, the reduced form is noticeable on the spectrogram, with very low-intensity noise appearing at the higher end of the spectrogram right before the period of silence corresponding to the stop closure associated with the following /b/. Measuring the period of visible high-frequency noise associated with the reduced form yields a noise-to-harmonics rate of 0.223, versus 0.448 for the full form. The rates of voicelessness are similar, at 27.27% for the full form and 23.53% for the reduced form. As one might expect, the reduced form is also shorter, at 15.9% of the VC interval, versus 19.2% for the full form. The fact that these measures are possible, despite the reduced segment’s lack of audibility suggests that some articulation associated with /ʒ/ is made by the speaker. The absence of audible sound may be due to a labial closure overlapping with the constriction associated with the fricative. The apparent ‘velar pinch’ (coming together of the second and third formants) suggests a posterior articulation of /ʒ/, and a transcription as [ɣ].

Fig. 10 Full form in word-final position: ‘neige dans’

On the Reduction of /ʒ/ in a Minority North American …

129

Fig. 11 Reduced form in word-final position: ‘tout rouges b-bondés’

5 Characterizing the Debuccalization of /ʒ/ We have seen thus far that our speaker produces a variety of reduced forms of /ʒ/. In this section we will examine the variants we have seen, considering them as systematically related, principally in terms of aperture, but also in terms of place (Sect. 5.1). We will then attempt to characterize the process that produces them, bearing in mind the way it has been described previously (Sect. 5.2). Finally (Sect. 5.3), we will discuss the process in the context of theoretical discussions of lenition in general.

5.1

Reduction on an Aperture Continuum

Figure 12 shows the variants of /ʒ/ on a continuum with respect to the openness of the stricture in the oral cavity associated with frication noise. The variants in (a) are voiced and the ones in (b) are voiceless. Note the implicit assumption in the presentation of forms that the “basic” one on the left includes voicelessness. Let us make this assumption explicit by saying that the production target for /ʒ/ includes some spreading of the glottis. As Vaux (1998: 509) points out, even voiced fricatives “require some spreading of the glottis in order to produce sufficient

130

R. Gess

a. [ʒ̥]

[ʒ]

[ɣ]

[j]

[ɦ]

[ʝ]

Ø

b. [ʃ]

[h]

Fig. 12 Reduction on an aperture continuum

airflow to yield frication noise”. Although Vaux concludes that voiced fricatives are still specified phonologically as [−spread glottis], my position here is that the phonetic implementation component includes gradient information regarding spreading of the glottis in the production target (cf. Gess 2009: 233). Further comment on Fig. 12 is warranted. First, the presentation implies that the devoiced variant [ʒ̥] has a higher degree of stricture than fully voiced [ʒ]. This is compatible with Ohala’s (1983) observation discussed in Sect. 3.1 (following Table 1), that a “good” fricative (presumably with a closer approximation between the active and passive articulators) lends itself to devoicing. Second, one might expect [ʒ], [ʝ] and [ɣ] to occupy the same position on the continuum. However, Keating (1988) and others have argued that palatals are complex segments. Following Grammont (1933), I assume that the segment /ʒ/ in French also has a complex articulation, involving constrictions between the tongue tip and alveolar ridge (coronal) and between the tongue body and the velar region (dorsal) (see his Fig. 80). I know of no more recent work that treats the articulatory details of /ʒ/ (and /ʃ/) in French, although clearly such work is needed (see Sect. 7). Since [ʒ] and [ʝ] are complex (albeit with different configurational details), and [ɣ] is not, placement of the former to the left of the latter on the aperture continuum is justified. We might expect voiceless realizations corresponding to [ʝ] and [ɣ], but none are found in the data from the speaker reported on here, even though there is one apparent potential consonantal trigger for [x], in ‘engage quelqu’un’ (presented in Table 1). /ʒ/ does not assimilate in this sequence, but as we will see in the following section, Charbonneau reports a realization [x], in ‘agi’. It is interesting that the trigger in this case is a vowel, as our place-assimilated forms (in ‘aujourd’hui’ and ‘rouges’) also involve triggering vowels, and we have no cases of consonantally triggered place assimilation. To the extent that vowels serve as triggers, and not consonants, place-assimilated voiceless realizations are not expected, since vowels also entail greater voicing, as Tables 1 and 2 demonstrate. [ɦ] and [h] represent an absence of oral place and result from reduction, not assimilation, as we discuss below.

On the Reduction of /ʒ/ in a Minority North American …

5.2

131

Characterizing the Process

Before turning to how we might formally characterize the process producing the variants reported on in the previous section, it is instructive to see how prior studies have characterized it. Alternative labels have included ‘aspiration’ (Tassé 1981; Hansen 1988), ‘velarization’ (Holder 1972), ‘posteriorization’ (Tassé 1981), ‘spirantization’ (intended as general weakening, Charbonneau 1957), ‘mellowing’ (Brent 1971), and ‘debuccalization’ (‘désarticulation’ in French, Gendron 1970; Tassé 1981). Debuccalization seems the most apt term, as for our speaker, the variants can be seen as resulting from a reduction of the supralaryngeal gestures (and only the supralaryngeal ones) associated with /ʒ/. Normally ‘debuccalization’ (in English) refers to the complete absence of supralaryngeal features, but the French term ‘désarticulation’ lends itself to a more gradient interpretation, which I think is suitable to the case at hand. I consider it crucial in using the term debuccalization as I am here, that the range of variants produced by the reduction process include [ɦ] and [h] as members, on a continuum of decreasing magnitude of the oral articulatory gestures. However, the presence of variants other than the laryngeals should not prevent us from using this label. For each variant, the percept comes from the residue of articulation once reduction of oral gestures has taken place. This residue includes the reduced oral gestures themselves, to the extent that any remain (and whether or not they are place assimilated). It also includes any labial gestures (rounding and/or protrusion), and any laryngeal gestures (state of the glottis), such as spreading of the glottis. Another residual feature is power of airflow, which we have not mentioned previously, but which seems important, and is used to describe some of the variants listed in Charbonneau (1957). Power of airflow may produce different percepts ([h] versus [ɦ], for example) depending on, for example, the prosodic strength of the position occupied by /ʒ/. Importantly, the percept for each variant appears to come from residue of articulation only, with no compensatory modification of gestures (cf. the discussion of this issue in Gess (2009) (231–241), referring to arguments made by Kirchner (2001, 2004), suggesting that modification of gestures accompanying reduction are expected). This does not preclude the possibility that modification of gestures might occur in a process of phonologization—i.e., when speakers reproduce a salient percept in a different way. The process is also characterized by the fact that its occurrence is largely determined by lexical frequency. The words classified as reduced were ‘je’ (seven occurrences) ‘jour(s)’ (two occurrences), ‘aujourd’hui’ (two occurrences), ‘rouge’ and ‘justement’. Also relevant to its implementation is speech context—it occurs only in spontaneous speech, not in reading. This raises the possibility that rate of articulation affects the complex articulation of the fricative.

132

5.3

R. Gess

This Case of Debuccalization as Lenition

In this section I address a very interesting claim made by Kingston (2008: 27), in a piece entitled “Lenition”, that when lenition occurs, “a more open articulation is chosen to increase acoustic intensity, reduce interruption, and thereby convey that the current prosodic constituent is continuing rather than a new one beginning”. This claim is made in opposition to an effort-based explanation. The case of debuccalization discussed here is very interesting in the context of Kingston’s proposal, as it occurs in environments counter to his predictions. We have reduced variants occurring in phrase-initial position in ‘je me suis’ and ‘justement’, as well as in phrase-final position in ‘rouge’. It seems quite clear that our case of debuccalization represents a case of lenition (in the synchronic sense intended by Kingston (2008)) (“what motivates a speaker to lenite”). So of particular interest here are the reduced variants that occur in non-leniting position. This list includes most importantly ‘je’, which as a subject clitic is naturally occurring in domain-initial position. Perhaps this is just a case of hyper frequency (is there a more frequently occurring element?), that trumps Kingston’s generalization on position with respect to domain. This leaves us with ‘justement’ (domain-initial) and ‘rouge’ (domain-final). As we saw previously, with ‘rouge’, there appears to be some articulation here on the part of the speaker (the ‘velar pinch’) that is masked, perhaps by an early labial closure for the following segment. This example shows us that some instances of the laryngeal variants may be purely perceptual, with the presence of some oral articulatory movement representing significant undershoot. This may indeed be the case with ‘justement’, which has an interesting articulatory sequence, in which /ʒ/ precedes a high vowel itself preceding /s/. High vowels are prone to devoicing in French (as heard often in the affirmative ‘oui’ with aspiration at the end), which explains the realization as [h] rather than [ɦ]. The reduction in the supralaryngeal cavity results in a narrow passage created by the roundedness of the vowel, enhancing the rounding associated with /ʒ/ as a secondary feature. The prosodic strength of the position, and resultant force of airflow, together with the shape of the cavity create a quite salient percept despite an apparent gestural reduction involving the front of the tongue. Before ending this discussion, we need to address a question that was asked at the PFC conference in 2016—could what we are calling debuccalization be simple assimilation to place? If this were the case, the laryngeal percepts would presumably result from triggering posterior vowels and/ or consonants. Evidence against this view is the existence of a form with debbucalized /ʒ/ before a labial (e.g., ‘je me suis adressé’), which does not involve tongue displacement. Although reduction of hyper frequent ‘je’ should perhaps not count against Kingston’s generalization regarding lenition generally, it is instructive nevertheless to examine its articulatory implementation, which we have done in this study. Another interesting question is why /ʒ/ is susceptible to reduction in the first place, which we will address in the following section.

On the Reduction of /ʒ/ in a Minority North American …

133

6 Why Is /ʒ/ Susceptible to Reduction? A common view is that segments that are susceptible to reduction are already perceptually weak (cf. work on phonetically based phonology, e.g., Hayes et al. (2004))—but /ʒ/ is not a perceptually weak segment. It is characterized by high-frequency noise across a large part of the spectrum (2,000 Hz and above). On the other hand, we have suggested that /ʒ/ is an articulatorily complex segment, as it is both coronal and dorsal. It follows that /ʒ/ is also more effortful than any sound containing only one such articulation (only coronal or only dorsal). I assume that the complexity of /ʒ/ is the principal reason it is targeted for reduction. A reduced articulation is less effortful, and if only one of the two articulations is targeted for reduction, there remains an articulatory configuration capable of creating an identifiable percept (ensuring message continuity). We have already mentioned an additional complexity involving /ʒ/ as a voiced fricative, which is the inherent conflict between the maintenance of voicing and the creation of a turbulent airstream (Ohala 1983). This complexity also makes /ʒ/ susceptible to reduction (one of two choices to resolve the conflict, the other being devoicing (Solé 2002)). I do not assume this as the principal motivation for reduction, as we do not see a similar targeting of the voiced fricative /z/. Phonologically, it must be observed that /ʒ/ has no competition in the posterior articulatory space (Hull 1966), except for varieties with the posterior rhotic (our speaker has an anterior rhotic). In order to see if this phonological factor also renders /ʒ/ susceptible to debuccalization, it would be interesting to know if we find the phenomenon in varieties with the uvular /ʁ/. To my knowledge, this question has not previously been raised. Finally, we must note that, for the form ‘je’ in particular, perhaps the most frequently used item in the language, once schwa deletion has occurred there is only the one target left for reduction, so that also contributes to the susceptibility of /ʒ/ to debuccalization.

7 Conclusions and Opportunities for Future Explorations In our acoustic study of full and reduced forms of /ʒ/ in the production of one speaker, we have seen that a range of realizations are produced on a continuum of aperture from [ʒ̥] and [ʒ] to [ɦ] and [h]. Place-assimilated variants also occur along the continuum. The inclusion of [ɦ] and [h] on the continuum, with their total lack of supralaryngeal features, suggests that debuccalization is an appropriate term to characterize the process. As a subcategory of lenition, debuccalization of /ʒ/ can occur at prosodic boundaries, a fact incompatible with Kingston’s (2008) proposal that lenition is intended to signal continuity within a prosodic constituent. We suggest instead that its articulatory complexity is what renders /ʒ/ a salient target for reduction.

134

R. Gess

A number of questions remain unanswered by this study. Of particular importance is to know the precise nature of the articulatory aspects of full and reduced forms. Without knowing exactly how they are pronounced articulatorily, it is difficult to reach sound conclusions with respect to the implementation of reduction. An ultrasound study would be invaluable in this regard. Given the invasiveness of ultrasound equipment, it may be impossible to get reduced forms that are not imitated, but even knowing the details of articulation with respect to the full form would be beneficial. Other questions may be fruitfully answered by ongoing work being conducted in Quebec under the guise of the PFC project, supervised by Marie-Hélène Côté. In particular, there is a survey point in progress in the Joliette area, which is known for a relatively high rate of production of debuccalized variants (including in the name of the town).3 Study in this area may tell us if there is as much variation in realization in communities in which the phenomenon is phonologized, as there is in those in which it occurs in the speech of just a handful of speakers and in limited contexts. We might hypothesize that variation would be more limited in this case, perhaps to the extent of involving substitution rather than reduction. PFC survey points may also be good testing grounds for seeing if posterior articulations of /ʒ/ are inhibited by the presence of a dorsal rhotic. Acknowledgements I would like to thank for feedback on various earlier versions, audiences at Français Langue Orale et Linguistique – Phonologie du français contemporain: schwa, phonologie de corpus, Paris, November 17–19, 2016; Montreal-Ottawa-Toronto Phonology Workshop, Université du Québec à Montréal, Montréal, March 24–26, 2017; and the 47th Linguistic Symposium on Romance Languages, University of Delaware, April 20–23, 2017.

References Bittner, Michèle. 1995. Réalisation des constrictives ʃ et ʒ en parler saguenéen. Étude acoustique. Master’s report, Université du Québec à Chicoutimi. Boersma, Paul, and David Weenink. 2016. Praat: Doing phonetics by computer [Computer program]. Version 6.0.22. http://www.praat.org/. Accessed 15 Nov 2016. Brent, Edmund. 1971. Canadian French: A synthesis. Ph.D. dissertation, Cornell University. Charbonneau, René. 1957. La spirantisation du (ž). Journal of the Canadian Linguistic Association 3 (14–17): 71–77. Chidaine, Jean G. 1967. CH et J en saintongeais et en français canadien. In Etudes de linguistique franco-canadienne, ed. Jean-Denis Gendron and Georges Straka, 143–151. Paris: Klincksieck and Quebec: Les Presses de l’Université Laval. Dajko, Nathalie, and Darcie Blainey. 2016. Variation in Louisiana French: Prairie Cajuns and Bayou Cajuns. In Varieties of spoken French, ed. Sylvain Detey, Jacques Durand, Bernard Laks, and Chantal Lyche, 478–490. Oxford: Oxford University Press.

3

So much so that the Huffington Post, Québec version, offered a (2015) quiz on how much the participant knows the town, subtitled “14 questions pour démystifier la ville de Joliette (ou Holiette)” (http://quebec.huffingtonpost.ca/2015/10/23/quiz-connaissez-vous-bien-laval_n_8373816.html).

On the Reduction of /ʒ/ in a Minority North American …

135

Gallard, Philippe. 1990. Evolutions phonétiques et phonologiques d’un parler poitevin. La linguistique 26: 79–87. Gendron, Jean-Denis. 1970. Origine de quelques traits de prononciation du parler du franco-québécois. In Phonétique et linguistique romanes. Mélanges offerts à M. Georges Straka, 339–352. Lyon & Strasbourg: Société de linguistique romane. Gess, Randall. 2009. Reductive sound change and the perception/production interface. Canadian Journal of Linguistics 54: 229–253. Grammont, Maurice. 1933. Traité de phonétique. Paris: Delagrave. Hansen, Anita Berit. 1988. L’aspiration du /ʒ/ à Hull (Québec). Approche sociolinguistique. Revue Romane 23: 178–197. Hayes, Bruce, Donca Steriade, and Robert Kirchner. 2004. Phonetically based phonology. Cambridge: Cambridge University Press. Holder, Maurice A. 1972. Le parler populaire franco-canadien. Phonetica 26: 3349. Horiot, Brigitte. 1990. Les aires linguistiques II. Dialectes de l’Ouest. In Lexikon der Romanistischen Linguistik, vol. V, I, ed. Günter Holtus, Michael Metzeltin, and Christian Schmitt, 615–637. Tübingen: Max Niemeyer Verlag. Hualde, José Ignacio, and Pilar Prieto. 2014. Lenition of intervocalic alveolar fricatives in Catalan and Spanish. Phonetica 71: 109–127. Hull, Alexander. 1966. The structure of the Canadian French consonant system. La Linguistique 2: 103–110. Hull, Alexander. 1968. The origins of the New World French phonology. Word 24: 225–269. Juneau. 1972. Contribution à l’histoire de la prononciation française au Québec. Québec: Les Presses de l’Université Laval. Keating, Patricia A. 1988. Palatals as complex segments: X-ray evidence. UCLA Working Papers in Phonetics 69: 77–91. Kingston, John. 2008. Lenition. In Selected proceedings of the 3rd conference on laboratory approaches to Spanish phonology, ed. Laura Colantoni and Jeffrey Steele, 1–31. Somerville, MA: Cascadilla Proceedings Project. Kirchner, Robert. 2001. An effort-based approach to lenition. New York: Routledge. Kirchner, Robert. 2004. Consonant lenition. In Perceptually based phonology, ed. Bruce Hayes, Robert Kirchner, and Donca Steriade, 113–145. Cambridge: Cambridge University Press. Klingler, Thomas A., and Chantal Lyche. 2012. “Cajun” French in a non-Acadian community. A phonological study of the French of Ville Platte, Louisiana. In Phonological variation in French: Illustrations from three continents, ed. Randall Gess, Chantal Lyche, and Trudel Meisenburg, 275–312. Amsterdam & Philadelphia: John Benjamins. Léon, Pierre R. 1967. H et R en patois normand et en français canadien. In Etudes de linguistique franco-canadienne, ed. Jean-Denis Gendron and Georges Straka, 125–142. Paris: Klincksieck and Quebec: Les Presses de l’Université Laval. Lucci, Vincent. 1972. Phonologie de l’Acadien (Parler de la région de Moncton, N.-B., Canada). Montréal: Didier. Massignon, Geneviève. 1962. Les parlers français d’Acadie, enquête linguistique, vol. I. Paris: Klincksieck. Ohala, John. J. 1983. The origin of sound patterns in vocal tract constraints. In The production of speech, ed. P.F. MacNeilage, 189–216. New York: Springer. Pignon, Jacques. 1960. L’évolution phonétique des parlers du Poitou (Vienne et Deux-Sèvres). Paris: D’Artrey. Poirier, Claude. 1988. L’expansion du français hors de France. In Dictionnaire du français plus, ed. A.E. Shiaty, Pierre Auger, Normand Beauchemin, and Claude Poirier, 11854–11856. Beauceville: Centre Éducatif et Culturel. Rochette, Claude-E. 1973. Les groupes de consonnes en français., vol 2, Paris: Klincksieck et Québec: Les Presses de l'Université Laval. Solé, Maria-Josep. 2002. Assimilatory processes and aerodynamic factors. In Papers in Laboratory phonology 7, ed. Carlos Gussenhoven and Natasha Warner, 351–386. Berlin: Mouton de Gruyter.

136

R. Gess

Tassé, Georges. 1981. Le parler rural de St-Prosper de Beauce (Étude des phénomènes phonologiques caractéristiques). Master’s report, Université du Québec à Montréal. Thogmartin, Clyde O. 1974. The phonology of three varieties of French in Manitoba. Orbis 23: 335–349. Vaux, Bert. 1998. The laryngeal specifications of fricatives. Linguistic Inquiry 29: 497–511. Vincent, Diane, and David Sankoff. 1975. The geographic dimension of phonological variation within an urban speech community. Montréal: U.

Why Lenition Interactions Are Typically Counter-Feeding Haike Jacobs

Abstract Typologically, the interaction between voicing and spirantization processes applies predominantly in a counter-feeding fashion, and, more rarely in a feeding one. After providing some relevant data from contemporary Romance varieties that illustrate this state of affairs, this paper first discusses why this is problematic for previous theoretical analyses, both from a rule-based and from a constraint-based perspective of phonology. A novel way of evaluating constraints will be proposed which locally evaluates only output candidates that have undergone one single change to satisfy the relevant markedness constraint at hand. On the one hand, this allows to describe both types of interaction (feeding and counter-feeding) which thus far was quite problematic for OT. On the other hand, we will illustrate that, in perception, this makes a feeding interaction computationally more complicated than a counter-feeding one, which is, we claim, the reason for the typological unmarkedness of the counter-feeding interaction between voicing and spirantization.








Keywords Counter-feeding opacity Contrast maintenance Serial optimality theory Local constraint evaluation Perception of voicing and spirantization Harmonic Serialism Optimality theory with candidate chains










1 Introduction In this paper, we will address the observation that lenition processes involving voicing of intervocalic voiceless plosives and spirantization of intervocalic voiced plosives typically apply in a counter-feeding interaction (Gurevich 2004), that is the voiceless plosives surface as voiced ones, but do not partake in the modification the

H. Jacobs (&) Centre for Language Studies, Radboud University, Nijmegen, The Netherlands e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_7

137

138

H. Jacobs

underlying voiced plosives are subject to: they are not further modified so as to be realized as voiced fricatives. This paper is organized as follows. In Sect. 2, we will provide some relevant diachronic and synchronic Romance data sets. Section 3 critically reviews previous (both rule-based and optimality theory (OT)-based) theoretical analyses of counter-feeding lenition interactions. Rule-based models are unable to differentiate between preferred or common and less preferred or uncommon lenition interactions. OT fits uneasily with counter-feeding relationships in general, and, constraints that directly militate against a neutralization of contrast are, on the one hand, too complicated to be convincing, and, on the other hand, make the wrong empirical predictions. Articulatory Phonology has come up with phonetically grounded reasons for the general process of consonant weakening or lenition, but has no principled way of accounting for, or even of distinguishing between, a preferred counter-feeding and a less preferred feeding interaction. After reviewing these previous accounts, we will in Sect. 4, propose a bi-directional, harmonically serial version of OT, to shed new light on this issue. Basically, we will argue that in perception a counter-feeding interaction is computationally less complicated than a feeding one, which, as we propose, is the reason why the counter-feeding interaction is the unmarked one.

2 Diachronic and Synchronic Romance The way consonant lenition (a set of changes comprising the spirantization of voiced intervocalic plosives, the voicing of intervocalic voiceless obstruents and the degemination of intervocalic geminates) took place in the historical phonology of the Western Romance languages is illustrated for the history of French in (1).

Why Lenition Interactions Are Typically Counter-Feeding

139

The way voicing and spirantization occurred diachronically can also be synchronically observed in contemporary Romance varieties, as illustrated for Sisco Corsican (Cravens 2000; Gurevich 2004) in (2) and for Gran Canarian Spanish (Oftedal 1985; Broś 2016 (providing new experimental data confirming Oftedal 1985)) in (3).

140

H. Jacobs

In (2) and (3), the voiced plosives are realized as voiced fricatives and the voiceless plosives as voiced plosives when occurring in intervocalic position. The interactions in (2) and (3) are counter-feeding and opaque, given that it is not clear why, if from a surface perspective voiced plosives are changed into voiced fricatives (as in (2b) and (3b)), the resulting voiced plosives from voicing in (2a) and (3a) are not turned into voiced fricatives as well. In Campidanian Sardinian (Bolognesi 1998; Cravens 2000), the voiceless obstruents are realized as voiced fricatives, as illustrated in (4), whereas the voiced plosives are unaffected or sporadically deleted (with lexical restrictions) in fluent speech (Bolognesi 1998: 37).

Why Lenition Interactions Are Typically Counter-Feeding

141

Cravens (2000: 56) and Bolognesi (1998: 222), referring to Molinu (1992) and Virdis (1978), mention Logudorese as a variety in which the voiced stops too are turned into voiced fricatives. Campidanian surface contrasts such as [sːɑbɔrtːɑ] (from underlying /sːɑpɔrtːɑ/ ‘the door’) versus [sːɑbɔrtːɑ] (from underlying /sːɑbɔrtːɑ/ ‘the time’) are thus neutralized and both of them are realized identically as [sːɑbɔrtːɑ] (Bolognesi 1998: 36). The same counter-feeding tendency illustrated for Sisco Corsican and Gran Canarian Spanish above is observed from a broader typological perspective. Gurevich (2004) observed for 230 lenition processes in 153 languages that in the majority (92%) of cases, lenition tended to be contrast preserving. Structuralists’ claims (Martinet 1955) that the maintenance of contrast is the explanatory factor behind it, have thus far escaped a theoretically satisfactory formalization. In the next section, we will briefly review some previous descriptions.

3 Previous Rule-Based and Constraint-Based Analyses Within rule-based generative phonology the feeding and counter-feeding lenition interactions can easily be described by the relative order of the two informal rules in (5) and (6).

In rule-based phonology, however, there is no clear way in which it could be expressed that a counter-feeding lenition interaction (first rule (5) then rule (6)) is typologically the unmarked one and that a feeding interaction (first rule (6) then rule

142

H. Jacobs

(5)) is more marked. On the contrary, the principle of maximal rule application (Kiparsky 1982) predicts that a feeding interaction should be the unmarked state of affairs. Similarly, Hale and Reiss (2008) remark that “Opaque rules are not surface true, rules that are not surface true are harder to learn, failure to learn aspects of the ambient language constitutes a diachronic change, therefore, languages are more likely to lose a given instance of opacity than to gain one.” But if this is so, why is it then that all the interactions above, except for Logudorese Sardinian, are counter-feeding? Kaye (1975) observes that opacity has a functional motivation: avoiding neutralization and keeping intact lexical contrasts. This idea harks back to structural linguistics (Martinet 1955) in which consonant lenition, weakening, was seen as articulatory effort reduction and in which the historical changes in (1) have been analyzed in terms of push and pull chains. Latin geminates exerted pressure on the system, which caused the spirantization of the voiced plosives, thus making room in the intervocalic context for the voiceless obstruents to be realized as voiced without losing lexical contrasts. Finally, after the original voiceless single plosives had changed into fricatives, degemination could take place without loss of lexical contrasts. Jacobs and Wetzels (1988) tried to formalize the neutralization avoidance and lexical contrast maintaining aspect of diachronic lenition by relying on Steriade’s (1987) theory of underspecification. Given the absence of voiced fricatives [z] and [v] in Latin, the feature [−voice] can be considered to be a predictable or redundant value for the class of fricative consonants and also the feature [−continuant] can be considered to be a redundant one for the class of voiced plosives. The basic idea was that each step in the lenition process affected a predictable, redundant feature value and changed that value into the unmarked feature value for the intervocalic position. The first step in the lenition process, the spirantization of voiced intervocalic stops was seen as the change from a predictable feature value, [−continuant], to a feature value that within the context in which lenition occurred is a less marked one, or a preferred one from the point of view of the speaker’s ease of articulation, that is, [+continuant]. The second step of lenition was analyzed in a similar way. Once spirantization was completed, the [−voice] feature value could no longer be considered a predictable feature value for the class of fricative consonants and a new system of underlying specifications was set up for the intervocalic context exclusively, given that the older system continued to exist in other (post-consonantal and word-initial) positions. It was supposed that this new intervocalic underlying feature representation was one in which the feature [+/−voice] was entirely predictable for the entire class of obstruents (plosives [p, t, k] and strident fricatives [s, f] being redundantly [−voice] and non-strident fricatives [b, ð, ɣ] being redundantly [+voice]). Again, without loss of lexical contrast, the second diachronic step, voicing of intervocalic voiceless obstruents, was formalized as a change from a predictable feature value, [−voice], to a feature value that intervocalically is a preferred one from the ease of articulation perspective, that is, [+voice].

Why Lenition Interactions Are Typically Counter-Feeding

143

However, although the under-specification-based analysis formally expresses the neutralization avoidance characteristic of diachronic lenition, it cannot be extended to synchronic lenition interactions, such as Sisco Corsican or Gran Canarian Spanish, as it would imply different underlying intervocalic representations for each of the two synchronic processes involved: one representation for spirantization, in which [−continuant] is predictable for voiced stops, and a different one for voicing, in which [−voice] is predictable for obstruents. Classical OT can only describe transparent, feeding lenition interactions, but not opaque, counter-feeding ones, given that the markedness constraints are output-based constraints. For clarity’s sake, this is illustrated in tableau (7), where four constraints are used: (SPIRANTIZATION: No voiced intervocalic plosive; VOICING: No voiceless intervocalic plosive; IDENT (VOICE): A consonant in the output has the same ‘voice’ specification as in the input and, IDENT (CONT): A consonant in the output has the same ‘continuant’ specification as in the input). Tableau (7) shows that the SPIRANTIZATION needs to be ranked above IDENT (CONT).

As illustrated in tableau (8), the ranking assumed in (7) produces a feeding output for forms like (3c).

The constraint ranking in (7) incorrectly predicts that /la kama/ is optimally realized as [laɣama] rather than as actual [laɡama]. If the relative ranking of the constraints responsible for voicing and spirantization were reversed, /la kama/ would incorrectly be optimally realized as [lakama] instead of [laɡama], as illustrated in (9).

144

H. Jacobs

Within OT, a number of different ways have been proposed to describe counter-feeding lenition interactions. A Stratal or Level OT approach (Rubach 2000) and an OT with candidate chains (OT-CC, McCarthy 2007) approach are discussed for Gran Canarian Spanish in Gussenhoven and Jacobs (2017). The Stratal OT approach suffers from much the same drawback as the under-specification analysis. Different constraint rankings and different levels have to be assumed for spirantization and voicing, whereas they both apply in the same environment: intervocalically and across word boundaries, pointing clearly to post-lexical modifications belonging to the same level. The OT-CC analysis requires a special type of PRECEDENCE constraint, that imposes an extrinsic order on the violations of faithfulness constraints in a candidate chain, which, again, cannot be clearly motivated independently (we refer the reader to Gussenhoven and Jacobs (2017: 226–230) for a more detailed discussion). For Gran Canarian Spanish, Broś (2016) proposes an analysis applying the model of constraint conjunction (cf. Kirchner 2001), that is, by combining two constraints under a joint name as a conjunction of two constraints and by inserting that conjoined constraint at the relevant place in the hierarchy. Similar to PRECEDENCE constraints, the conjunction of constraints cannot really be motivated independently. Also, what all the proposals share is that the description of a counter-feeding lenition interaction is formally more complicated (given that additional mechanisms are required) than the description of feeding lenition interactions, which would predict the opposite of what is typologically observed: feeding should be the unmarked and counter-feeding the marked state of affairs. A more direct way to formalize the neutralization avoidance and lexical contrast maintaining aspect of counter-feeding lenition interactions has been to invoke a constraint like the one in (10), which directly militates against neutralization (Brandão de Carvalho 2008: 215; Bolognesi 1998: 195).

The constraint in (10), however, is problematic for two reasons. First, it has to be considered as an Output-Output constraint in the sense that an output form [naða]

Why Lenition Interactions Are Typically Counter-Feeding

145

for, say, underlying /nata/ from (3c) above, needs to be compared to every output of every lexical item of the language in order to establish whether there is a possible output in the language, [naða] for underlying /nada/, and, that therefore /nata/ may not surface as [naða] by virtue of (10), but must instead surface as [nada]. Not only does comparing a possible output form of a given lexical item to the optimal outputs of all other lexical items lead to an extremely complicated analysis, it also makes the wrong empirical predictions. For instance, given the absence of the word tibigo in Gran Canarian Spanish, the constraint in (10) would for the possible output form [tibiɣo] for underlying /tipiko/ ((3) above) not come across the same output form for a possible, but not actually occurring, underlying /tibigo/ and hence would not be able to block spirantization from applying. It should be observed that even if (10) were to be reformulated more locally so as to avoid neutralization of a VCV sequence, the same two problematic aspects would still be effective. Let us next briefly discuss an Articulatory Phonology-based account of the typical non-neutralizing lenition interactions. Blevins (2004) has argued that there is no need at all to express the non-neutralizing characteristic of the synchronic lenition interaction between voicing and spirantization given that it follows, as she claims from the general mechanism of gestural reduction. For the contrast maintenance between degemination and voicing, which can be observed synchronically in Malayalam (Mohanan 1993), she remarks (ibid. 289). This gestural reduction takes the long closure of the geminate to a point where it approaches the duration of short stop in other contexts. For the short t, gestural reduction results in durational shortening as well; voicing in the t > d shift may be due to voice leakage, spontaneous voicing, or a percept of voicing […] No reference to “contrast-maintenance” is needed: the paired changes have parallel phonetic explanations based on parameters of hypo- versus hyper-articulated speech.

However, if we look at the synchronic Romance interaction from Sect. 2 above, it is clear that phonetics alone is not enough to account for them. In Gran Canarian Spanish, [ʧ] is subject to voicing (flecha [fleʤa] and una chica [unaʤiɡa]), but fricative [s] is unaffected (música [musiɡa] ‘music’ and not *[muziɡa]). In Campidanian Sardinian, on the other hand, both fricatives and affricates are affected by voicing and spirantization ([sɛr:a]*[s:azɛr:a] ‘the saw’ and [ʧiða]*[s:aʒiða] ‘the week’). Why would gestural reduction operate differently in the two Romance varieties? Also, for Campidanian Sardinian gestural reduction would only affect the intermediate series (the voiceless obstruents), but leave unaffected the most sonorous (voiced plosives) sonorous and the least sonorous segments (geminates). Finally, if in gestural reduction, reduction in gestural magnitude is the key explanatory factor and no reference to ‘contrast-maintenance” is required, it is not clear why the interactions sometimes indeed are feeding and neutralizing as in Logudurese Sardinian. After having reviewed a number of rule-based and constraint-based descriptions of the typical counter-feeding lenition interaction between voicing and spirantization in this section, the next section addresses and proposes an account based on Harmonic Serialism.

146

H. Jacobs

4 Voicing and Spirantization in Harmonic Serialism In Harmonic Serialism (McCarthy 2010), much like in OT-CC, an underlying form is step-wise modified in its way to becoming the optimal surface form. The first step, or the first harmonic improvement, is dependent on the constraint ranking and is assumed to be one single modification that incurs a faithfulness violation, but improves in harmony. The output of this first single modification is then input for the next round of evaluation and so forth until no further harmonic improvement is possible. McCarthy (2010) illustrates Harmonic Serialism with the opaque stress and epenthesis interaction in Levantine Arabic in (11).

The two forms in (11) both end up in having the same prosodic surface shape CV.CV.CVC.CV, but yet they are stressed differently. Besides the constraint HEADEDNESS (a word must have a prosodic head, it is must have a stress), McCarthy uses two more constraints, NOCOMPLEX CODA (complex codas are not allowed) and DEP-IO-V (no vowel insertion). These constraints are ranked in that order and the first best possible modification for underlying /katab-t-lu/ is illustrated in tableau (12).

As illustrated in (12), with Harmonic Serialism, the generation of output candidates by GEN is drastically limited in comparison with parallel OT (Prince and Smolensky 1993). Only output forms that have undergone one single change are taken into account. Of the three possible first, single, modifications, the candidate

Why Lenition Interactions Are Typically Counter-Feeding

147

(12-1), katábtlu, is the most harmonic candidate given that candidates (12-2) and (12-3) fail to abide by the top ranked constraint. The optimal output form [katábtlu] from (12) becomes input to the next step in the evaluation process, as illustrated in (13).

The optimal output form from this second round of evaluation in (13), [katábitlu], becomes input to the next step in the evaluation process, as shown in (14). At this point no further harmonic improvement is possible and the derivation, as McCarthy puts it, “converges” at step 3.

As illustrated in (12), and this is crucial for the further argumentation below, it is the ranking of the markedness constraints, HEADEDNESS above NOCOMPLEX CODA, that determines the order in which the processes take place: first stressing (at step 1) and then inserting a vowel (at step 2). In Jacobs (2016), it was illustrated that Harmonic Serialism not only has problems in accounting for counter-bleeding cases of opacity, but cannot describe counter-feeding opacity, which we illustrate in (15) for the selection of the first single modification for Gran Canarian Spanish underlying /la kama/.

148

H. Jacobs

Given that candidate 3 involves two, instead of one single, modifications at the same time, a change from [–voice] to [+voice] and a second change from [−continuant] to [+continuant], it is ruled out as a best possible first step. This leaves [lagama] as the optimal first step, and, input into the next round of evaluation, where it will have no chance of escaping the effects of SPIRANTIZATION. In order to describe formally counter-feeding and counter-bleeding opacity, it was proposed in Jacobs (2016) to further constrain the evaluation by making it more local. Here is how it works. In (12), HEADEDNESS as the top-ranked constraint leads to an output candidate as input for the second round of evaluation that obeys this top-ranked constraint. That is, no other output candidates are taken into account, but still, in principle, any number of otherwise modified forms could be evaluated, but to no purpose. That is, if the first constraint is HEADEDNESS and its antagonistic constraint DEP-FT (no insertion/assignment of feet), the best first possible foot is going to be assigned. For the constraint HEADEDNESS and its antagonistic DEP-FT faithfulness constraint, a larger part of the constraint grammar needs to be inspected in order to determine whether, say, an initial or a final foot is the best first step, or whether an iamb or a trochee is needed and also whether the foot is quantity-sensitive or quantity-insensitive. McCarthy’s observation that the ranking of the markedness constraints determines the order of the processes can be translated directly into the generation and evaluation of output candidates. Evaluation then can work as follows: a chain or an input-output derivation is step-wise built by taking into account only that portion of the grammar that is relevant. For the evaluation of a markedness constraint only a smaller part of the entire constraint grammar is required. That is, only two candidates, a candidate that abides by it and one that does not, are evaluated. By doing so, the global power of OT is further restrained and retained only when required, as for instance in the evaluation of the best first foot, that is, when a larger part of the grammar needs to be inspected. This is visualized in (16), for the selection of the first single modification for Gran Canarian Spanish underlying /la kama/, where, deviating from standard practice, the relevant part for the constraint evaluation at hand is not shaded, but where the irrelevant part is shaded.

Why Lenition Interactions Are Typically Counter-Feeding

149

The first constraint /la kama/ encounters, SPIRANTIZATION, does not offer a way for harmonic improvement, and the input for the next step will therefore be unmodified [lakama]. The last output candidate (16-3) has two modifications at the same time and is therefore ruled out. The evaluation then proceeds, with [lakama] being the input form for the second round of evaluation. This is illustrated in (17), where the two relevant constraints in (16), as they already have been inspected, are shaded, and, where the next two constraints, VOICING and IDENT (VOICE) are evaluated.

At this second step, harmonic improvement for input /la kama/ is possible, and, the optimal output [lagama] will be input for the next step, which is illustrated in (18).

At step 3, no further harmonic improvement is possible; turning /la gama/ back into /la kama/ will lead to two violations, one for VOICING and one for IDENT (VOICE). There will be convergence on the output form [laɡama]. A feeding voicing and spirantization interaction will require the first two constraints to be VOICING and

150

H. Jacobs

IDENT (VOICE). Two observations are crucial. First, it is crucial that input /laɡama/ at step 3 is not subject to the complete ranking again, because otherwise [laɣama] would still be preferred. Secondly. it is important to observe that, in this way, we have brought back the segmental constraints and their antagonistic faithfulness constraints to the descriptive power of derivational phonological rules whenever that is required, which allows for combining desired global power with its required limitations (cf. Jacobs 2016 for a more detailed account). In the next section, we will argue that this way of constraint evaluation, when logically applied to perception, leads to the observation that a feeding interaction between voicing and spirantization is, computationally, more complicated than a counter-feeding one.

5 The Perception of Voicing and Spirantization in Harmonic Serialism In Calabrese (2010), working within a rule-based approach, it has been argued that, in perception, a listener constructs an underlying representation (UR) by the reverse application of the phonological derivation. That is, the surface shape of a word AED (from underlying ABC and two rules C ! D/_# and B ! E /_D) requires the reverse application in the reversed order E ! B /_D and D ! C /_# and checking cohorts of UR’s from the Mental lexicon against the hypothetical surface representation provided by the phonological module. Translated into an OT grammar, this would mean that a surface shape of a word is run backwards through the same OT-grammar, where it becomes gradually less harmonical, that is, it deteriorates, but, it becomes, gradually, more faithful to the UR until a match is found. We will demonstrate next that this implies that a counter-feeding relation between voicing and spirantization is, from a perception point of view, less complicated than a feeding interaction. That is, for a feeding interaction, perception is forced to go back into the constraint hierarchy, but not for a counter-feeding interaction, which is, we claim, the reason why the latter type is unmarked. The first step in the perception of counter-feeding [laɡama] is illustrated in (19).

During the first step of the evaluation, perceived [laɡama] cannot become less harmonic and more faithful because it already violates the first markedness

Why Lenition Interactions Are Typically Counter-Feeding

151

constraint. The perceived form then moves on to the next two constraints in the serial evaluation of constraint pairs, as illustrated in (20).

Perception of counter-feeding [laɡama] can by incurring a violation of VOICING and undoing a violation of IDENT (VOICE), become closer to an underlying form which will lead to a matching underlying representation at this second step: /la kama/. Let us next turn to the perception of feeding [laɣama] from underlying /la kama/. The feeding constraint ranking, as remarked above, is one in which the first two constraints are VOICING and IDENT (VOICE). The first step in the perception of feeding [laɣama] from underlying /la kama/ is given in (21).

The second output candidate (21-2) has two modifications at the same time and is therefore ruled out. The first constraint VOICING is not violated by changing perceived [laɣama] into /laɡama/, so no deterioration is possible and the perceived form must move down in the hierarchy and is going to be inspected by the next pair of constraints, as illustrated in (22).

152

H. Jacobs

By changing perceived [laɣama] into /lagama/ the form is deteriorating, that is, becoming less harmonic and becoming more faithful to a potential UR. After this second step there is still no matching UR for perceived [laɣama]. The underlying representation /la gama/ does exist, but will only be a matching lexical representation for perceived [laɣama] from underlying /la gama/, but not for perceived [laɣama] from underlying /la kama/. In order to find the corresponding UR, perception is forced to go back up into the hierarchy and start again at the top. This is illustrated in (23).

At step 3, the perception of [laɣama] as /la ɡama/ is deteriorating and is becoming more faithful to a potential UR by incurring a violation of VOICING and by undoing a faithfulness violation of IDENT (VOICE). It is then perceived as [laɣama] > /la ɡama/ > /la kama/ and has found a matching UR. To summarize, in perception a feeding lenition interaction is more complicated than a counter-feeding one, as perception is forced to go back and forth through the constraint-grammar. We surmise that this is the reason why a counter-feeding relationship is what languages typically show: it is less complicated for the hearer.

6 Conclusion In this paper, we have addressed the question why, between voicing and spirantization lenition processes, one typically observes a counter-feeding relationship. After reviewing some relevant Romance data, we have briefly discussed the relative merits of previous rule-based and constraint-based analyses. After reviewing these previous accounts, we have proposed a novel way of evaluating constraint, which, locally, evaluates only output candidates that have undergone one single change, a single change that is limited to satisfy the markedness constraint at hand. The ranking of the markedness constraints thus effectively determines the order of the processes that are triggered by those markedness constraints and the single change output candidates. The proposal keeps the possibility of global optimization, but only if required (stress, epenthesis etc.), while at the same time bringing OT’s descriptive power back to serial rule order whenever that is required. In Sect. 5, we have demonstrated that in perception

Why Lenition Interactions Are Typically Counter-Feeding

153

a counter-feeding interaction is computationally less complicated than a feeding one, which, as we have argued, is the reason why the counter-feeding interaction is the unmarked one.

References Blevins, Juliette. 2004. Evolutionary phonology. Cambridge: Cambridge University Press. Bolognesi, Roberto. 1998. The phonology of Campidanian Sardinian. The Hague: HAG. Brandão de Carvalho, Joaquim. 2008. Western Romance. In Lenition and fortition, ed. Joaquim Brandão de Carvalho, Tobias Scheer, and Philippe Ségéral, 207–234. Berlin: Mouton de Gruyter. Broś, Karolina. 2016. Between phonology and morphosyntax: Voicing and spirantization in the Spanish of Gran Canaria. In Phonology, its faces and interfaces, ed. Jolanta Szpyra-Kozłowska and Eugeniusz Cyran, 173–200. Frankfurt am Main: Peter Lang. Calabrese, A. 2010. Perception, production and acoustic inputs in loanword phonology. In Loan phonology, ed. Andrea Calabrese and Leo Wetzels, 59–113. Amsterdam: John Benjamins. Cravens, Thomas. 2000. Romance lenition. In New approaches to old problems: Issues in Romance historical linguistics, ed. Steven Dworkin and Dieter Wanner, 51–64. Amsterdam: John Benjamins. Gurevich, Naomi. 2004. Lenition and contrast. The functional consequences of certain phonetically motivated sound changes. New York/London: Routledge. Gussenhoven, Carlos, and Haike Jacobs. 2017. Understanding phonology. New York/London: Routledge. Hale, Mark, and Charles Reiss. 2008. The phonological enterprise. Oxford: Oxford University Press. Jacobs, Haike. 2016. Serial OT and segmental opacity. In Spotlight on melody and structure in syntax and phonology, ed. Anna Bloch-Rozmej, Anna Bondaruk, and Anna Pramożwska, 233– 256. Lublin: Wydawnictwo KUL. Jacobs, Haike, and Leo Wetzels. 1988. Early French lenition: A formal account of an integrated sound change. In Features, segmental structure and harmony processes I, ed. Harry van der Hulst and Norval Smith, 105–129. Dordrecht: Foris. Kaye, Jonathan. 1975. A functional explanation of rule ordering in phonology. Parasession on Functionalism CLS: 244–252. Kiparsky, Paul. 1982. Explanation in phonology. Dordrecht: Foris. Kirchner, Robert. 2001. An effort based approach to consonant lenition. New York/London: Garland. Martinet, André. 1955. Économie des changements phonétiques. Bern: Francke. McCarthy, John. 2007. Hidden generalizations. Phonological opacity in optimality theory. Equinox. McCarthy, John. 2010. Harmonic Serialism supplement to doing optimality theory. http://works. bepress.com/john_j_mccarthy/108. Mohanan, K.P. 1993. Fields of attraction in phonology. In The last phonological rule, ed. John Goldsmith, 61–116. Chicago: The University of Chicago Press. Molinu, Lucia. 1992. Gli esiti fonosintattici del dialetto di Buddusó. L’Italia Dialettale 15: 123– 153. Oftedal, Magne. 1985. Lenition in Celtic and in Insular Spanish: The secondary voicing of stops in Gran Canaria. Oslo: Universitetsforlaget. Prince, Alan, and Paul Smolensky. 1993 [2004]. Optimality theory. Constraint interaction in generative grammar. Oxford: Blackwell.

154

H. Jacobs

Rubach, Jerzy. 2000. Glide and glottal stop insertion in Slavic languages. A DOT analysis. Linguistic Inquiry 31: 271–317. Steriade, Donca. 1987. Redundant values. Papers from the twenty-third regional meeting of the Chicago Linguistics Society, vol. 2, 339–362. Virdis, Maurizio. 1978. Fonetica del dialetto sardo campidanese. Cagliari: Edizione della Torre.

The French Jespersen’s Cycle and Negative Concord Marie Labelle

Abstract French is a classical example of Jespersen’s cycle. The term cycle suggests that the endpoint of the cycle is identical to the initial point. By discussing the various historical changes underlying the French cycle, we will show that the evolution is more spiral-like than cyclic: the language that seems to have completed the cycle has properties different from that of the initial language. We will also argue that, while a number of authors view the evolution as linear, going from Old French to Standard French, then to Colloquial French, Quebec French illustrating the end of the cycle, the facts rather suggest that a split between two dialects had occurred during the 16th century One dialect corresponds to Standard French, the other one led to Quebec French. We will propose an analysis of the facts.

1 Introduction The facts underlying the French Jespersen’s cycle (Jespersen 1917) illustrated in Fig. 1 are well known and widely discussed, but an examination of them taking into account changes relative to negative concord will allow us to propose a theoretical account that we think sheds some light on the evolution. Zeijlstra (2004) and Martineau and Déprez (2004), among others, present the evolution in Fig. 1 as going from Old and Middle French, where ‘ne’ is the sole marker of clausal negation (jeo ne sais—stages 1 and 2), to standard French, where ‘ne’ co-occurs with the negative word ‘pas’ (je ne sais pas—stage 3), to colloquial French, where ‘ne’ is omissible (je (ne) sais pas—stage 4), to Quebec French and similar varieties, where ‘ne’ has disappeared (je sais pas—stage 5). Stage 5, featuring a single marker of clausal negation, ‘pas’, completes the cycle. We will argue that the facts are more complex. The reanalysis of pas and point as negative quantifiers, which triggered the reanalysis of ne and its omissibility, is also linked to the contemporaneous emergence, during the 16th century, of two M. Labelle (&) Department of Linguistics, Université du Québec à Montréal, Montreal, QC, Canada e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_8

155

156

M. Labelle

Fig. 1 The French Jespersen’s cycle

distinct dialects, a normative one, of which the most common variety of colloquial European French (henceforth “European French”) is a continuation, and a stigmatized one continued in Quebec French and in similar varieties (henceforth “Quebec French”). As will be seen in Sect. 2, these dialects differ with respect to the possibility of negative concord: contrary to European French, Quebec French is a non-strict negative concord language. Thus, the evolution is branching. We will also see that the evolution is not cyclic but spiral-like since the single marker of clausal negation in stage 5 is quite different from the one in stage 1. The reader should not expect to discover new facts in the present paper, our aim being rather to frame known facts into a theory of negative concord that will bring to light aspects of the historical change. Our account will remain largely speculative, and we hope that it will trigger more research on the question.

2 Negative Concord It is well known that Medieval French, contemporary European French and Quebec French differ with respect to negative doubling, the negative concord observed between negative indefinites and the negation in a clause having a single negation reading (den Besten 1986). Medieval French classifies as a strict negative concord language (Giannakidou 2006): in Old and Middle French, the marker of clausal negation ne cooccurred with both preverbal and postverbal negative indefinites. This is illustrated in (1) (for details, cf. Labelle and Espinal 2014 and references therein). By contrast, in the common variety of colloquial European French (2), there is no negative doubling: pas is the marker of clausal negation (ne is

The French Jespersen’s Cycle and Negative Concord

157

omissible), but it is not compatible with pronominal negative indefinites like rien or personne without leading to a double negation reading.1 Finally, Quebec French is a non-strict negative concord language: there is negative doubling with the negative marker pas, but not from the subject position (3) (for details, cf. Daoust-Blais 1975: 57; di Sciullo and Tremblay 1996: 4; Labelle 2017); the apparent optional realization of negative doubling in this non-standard variant of French could be due to diglossia with the standard variety (Labelle 2017). (1)

a.

b.

(2)

a. b.

(3)

a. b.

Je ne trouvai nully […] qui…. ‘ I found nobody […] who…’ (1309, Joinville, 361.4156) Nul n’est si sainz cume li Sires ‘Nobody is as saint as the Lord.’ (c. 1170, Quatre Livres des Rois, 6.110)

Il (ne) voit (*pas) personne. ‘He doesn’t see anybody.’ Personne (n’) est (*pas) venu. ‘Nobody came.’

I voit pas personne. ‘He doesn’t see anyone.’ Personne est (*pas) venu. ‘Nobody came.’

The objective of the present paper is not to compare colloquial French and Quebecois French, but to explore the changes leading to the differences illustrated in (1)–(3) with respect to negative concord and their impact on our understanding of Jespersen’s cycle. We assume that, in languages without negative doubling, like contemporary European French, both the negative marker (pas in French) and negative indefinites are negative quantifiers, incorporating a negation in their semantics (Baunaz 2011, chapter 4). In (4), pas is a negative quantifier over events, and the negative quantifier personne quantifies over humans. This explains why they cannot cooccur without a double negation reading.2

In this paper, we leave aside the temporal adverbs plus ‘no more’ and jamais ‘never’, whose special behaviors set them apart and require independent studies. For a discussion, see Muller (1991). 2 We assume that the negative spread observed between two or more negative indefinites, e.g. in Personne (ne) fait rien ‘Nobody does anything’, is due to the creation of a branching quantifier (for French: Moritz and Valois 1994; Déprez 1999, 2000; de Swart and Sag 2002; Labelle 2017; for a different analysis, see Zeijlstra 2004, 2010). 1

158

(4)

M. Labelle

Il (ne) voit pasNege personneNegPERSON. (There is no event of him seeing no person) ‘He does not see nobody.’

The analysis of negative concord is more complex. Negative concord items are peculiar words: on the one hand they cooccur with the negation in a clause having a single negation reading, in which case they do not contribute a negation; on the other hand, they have a negative reading in isolation (Giannakidou 2006). This behavior led Ladusaw (1992) to qualify them as self-licensing negative polarity items. Capitalizing on that intuition, Zeijlstra (2004; see also Chierchia 2013) analyzes negative concord items as non-negative indefinites bearing a syntactic uninterpretable formal negative feature [uNEG]. The [uNEG] feature must be deleted before Spell-Out through Agreement with a c-commanding negative operator bearing a corresponding interpretable feature [iNEG]. Here, we adopt this analysis, in line with Labelle and Espinal’s (2014) study of Medieval French negation and Labelle’s (2017) study of negative concord in Quebec French. In languages exhibiting a preverbal/postverbal asymmetry in negative concord, like Quebec French, the marker of clausal negation—pas in Quebec French—is analyzed as a negative operator bearing [iNEG]. It c-commands and Agrees with postverbal negative expressions, licensing them by deleting their [uNEG] feature (5). But it does not c-command preverbal negative indefinites (6). This explains why there is no negative doubling with negative indefinites in subject position.

(5) (6)

I voit pasNeg[iNEG] personne[uNEG]. (=3a)

Following Labelle’s (2017) analysis of Quebec French, we analyze pas in (5) as a negative quantifier over the entities denoted by the negative concord items with which it Agrees. In (5) pas quantifies over persons: NOx(PERSONx)(…). In negative sentences without negative concord, pas is a negative quantifier over events, entering a tripartite structure of type NOx(eventx)(…) (e.g. Krifka 1989; Beghelli 1997). Multiple Agreement allows the same operator to license all the [uNEG] bearing elements on which it has scope.

(7)

a.

I a pas rien dit à personne. ‘He didn’t say anything to anybody’

b. c.

Pas[iNEG] … rien[uNEG] … personne[uNEG] NOx,y (THINGx, PERSONy)(…)

In non-strict negative concord languages, there is no negative doubling with negative indefinites in subject position, as we saw in (6). However, as shown in (8), a [uNEG] bearing indefinite in subject position is unlicensed, and we would expect the sentence to be ungrammatical.

The French Jespersen’s Cycle and Negative Concord

159

Zeijlstra (2004) accounts for the grammaticality of sentences like (8) by proposing that the [uNEG] feature of unlicensed negative indefinites triggers the merge of a covert negative operator bearing [iNEG] to their immediate left. This is what happens with negative indefinites in subject position (9).

(9)

ØNeg[iNEG] personne[uNEG] est venu. (=3b)

This self-licensing operation prevents the derivation from crashing. The analysis also accounts for the negative reading of negative indefinites in fragment answers:

Thus, the [uNEG] feature restricts negative concord items to the scope of a negative operator and allows them to have a negative reading in isolation. This feature distinguishes negative concord items from negative polarity items, which are licensed not only in negative contexts, but also in non-negative polarity contexts (e.g. questions, conditionals) and are not allowed in fragment answers for instance. To account for strict negative concord languages, like medieval French, Zeijlstra (2004) argues that the marker of clausal negation carries [uNEG] and must be licensed by a covert negative operator bearing [iNEG]. Since, in medieval French, the verb to which the negative marker ne was cliticized raised to T and from there to the left periphery in main clauses (e.g. Vance 1997), we assume, for concreteness, that the covert negative operator occupied a position within the left periphery of the clause (perhaps Laka’s [1990] RP), from where it c-commanded and licensed every [uNEG] bearing item, both preverbal and postverbal (cf. Labelle and Espinal 2014).

(11) [ØNege,x,y[iNEG] [ nulx[uNEG]…ne[uNEG]…nullyy[uNEG]]] In this approach, medieval French ne is not a negative operator: it does not have negative force by itself. Rather, this marker is an overt reflex of Agreement with a covert negative operator. The term cycle in Jespersen’s cycle suggests that the cycle ends where it started (see Fig. 1). However, in the present framework, the word functioning as the single marker of clausal negation at the end of the cycle (namely pas) does not have the same properties as the initial word (ne): medieval French ne bears a [uNEG] feature, but pas is a negative operator in both European French and Quebec French. Thus, the evolution is spiral-like rather than cyclic, since the endpoint is distinct from the initial point. In Sect. 3, we propose an account of the changes that affected ne and pas. In addition, Quebec French, pas, contrary to European French pas, carries [iNEG], a feature allowing it to license negative concord items. Assuming that Quebec French (and similar dialects) is a development from European colloquial

160

M. Labelle

French would imply that negative indefinites would have gone from carrying a medieval [uNEG] feature (stage 2), to being negative quantifiers (stage 4/standard French and common colloquial French), and then back to a [uNEG] stage (stage 5/ Quebec French). We believe that this back-and-forth type of change did not happen and that there was a branching evolution. In Sects. 4 and 5, we will argue that the historical changes leading to the emergence of two different grammars occurred during the 16th century and are the consequence of the reanalyses of ne and pas.

3 The Reanalyses of ne and pas We view change as grammar competition (Kroch 2000). In the old grammar (stages 1 and 2 of Fig. 1), ne bears [uNEG], overtly marking the presence of a covert negative operator, and pas is an optional negation reinforcer (12a) (Price 1962). In the new grammar (stage 4), pas is the operator of clausal negation and ne is omissible because, the negative interpretation of the clause being overtly expressed by pas, ne has become a morphological marker without substantive features (12b). Stage 3 of Fig. 1 is ambiguous between these two grammars.

(12) a. b.

Old grammar: ØNege[iNEG] Il ne[uNEG] dort (pasreinf). New grammar: Il (ne) dort pasNege.

We treat ne-drop as a spell-out option permitted by the new grammar. Ne may be dropped because it carries no substantive feature, but nothing in the grammar forces it to be dropped. Speakers may continue to spell out ne without consequences. This is what happens in standard French.3 While ne-drop is a sign of the new grammar, the presence of ne in Stage 3 does not allow us to distinguish between the old and the new grammar. The first indication of a change in the status of ne is the fact that it becomes omissible, indicating that the negative force is now on pas. There is a debate as to when the change happened. Hirschbühler and Labelle (2003, 2006), in their discussion of negative imperatives, arrive at the conclusion that the grammar with nedrop observed in Quebec French was imported to North America by settlers during the 17th century. Although Martineau (2009) found only 0.2% (2/877) of ne omission in 17th century texts—the two examples representing low class speech—, Dufter and Stark (2007), discussing Jean Heroard’s (Heroard 1601–1628; Ernst 1985) transcripts of the speech of the child who would become Louis XIII, born in 1601, show that the omission of ne increases between ages 1–4 and ages 5–9, going from 57 to 73% in clauses containing pas and a pronominal subject. This increase in ne-drop suggests that ne-drop, in this corpus, is not a childhood feature reflecting

3

It would lead us too far to discuss the various syntactic approaches to the relation between ne and pas in French. Standardly, ne is merged under Neg0, and pas fills SpecNegP (Pollock 1989). For a review, see Meisner et al. (2015).

The French Jespersen’s Cycle and Negative Concord

161

an incomplete acquisition of ne; rather, it reflects the acquisition of a variable rule of ne omission. This conclusion is supported by the fact that adult members of the high society interacting with the child are also quoted by Héroard as optionally omitting ne (Ayres-Bennett 1994: 75; Dufter and Stark 2007: 125): (13) [Me de Montglat] ha jhesu monsieu j fau pa faire cela, on vou recognetré pa pou le fi du Roy seulemen [. . .] ha monsieu j fau pa dire cela, j fau pa palé ainsi au gouvernante . . . cela n'e pa beau Monsieu (23.09.1607) ‘Ah Jesus, Sir, one must not do that, you will not be recognized as the son of the king [….] ah, Sir, one must not say that, one must not speak that way to governesses… it is not nice, Sir.’ (From Ayres-Bennett 1994: 75)

If adults omitted ne in 1607, the change must have started as early as the 16th century in some sociolects.4 We hypothesize that the reanalysis of pas/point from negation reinforcers to negative quantifiers triggered the reanalysis of ne as not carrying [uNEG]. In medieval French, pas was a negation reinforcer used with events/situations (e.g. Il ne dort pas ‘He doesn’t sleep at all’), while the miminizer point was found in partitive constructions (e.g. Je n’ai point de vin ‘I do not have a bit of wine’) (Price 1962). It appears that pas evolved from an event minimizer to a negative quantifier over events and eventually to a negative quantifier unspecified for a specific domain and able to quantify over entities (e.g. pas de pain ‘no bread’); conversely, point evolved from a nominal denoting a minimal quantity to a negative quantifier over entities in partitive constructions to, finally, a negative quantifier unspecified for a specific domain and able to quantify over events. In the last stage, the two items competed for the same contexts (Price 1962; Martineau 2005).

(14) pas: minimizer reinforcing the negation Nege point: nominal denoting a minimal quantity Negx (where e ranges over events and x ranges over entities)

Neg(e or x) Neg(e or x)

These reanalyses appear to have occurred between the second half of the 15th century and the end of the 16th century. Ayres-Bennett (1994) mentions various constructions in which pas/point had acquired a negative reading before the 17th century. For instance, already during Middle French, non was replaced by pas in oppositions (15), and, during the 16th century pas and point negated preposed infinitives (16). The negative quantifiers pas/point could have spread from such constructions to finite negative clauses.

4

See Boerm (2008) for the observation that ne-drop was frequent in Judeo-French long before this feature appeared in French writings. For instance, in La Reine Esther, written in 1300, pas is used in the majority of negative clauses (63%, 76/121), and in 58% of these cases (44/76) ne is omitted (p. 249). Boerm argues that the Jewish merchants speaking Judeo-French had the sociolinguistic features of spreaders of linguistic change. If he is correct, the reanalysis of pas as a negative quantifier discussed in the text could have been influenced by Judeo-French.

162

M. Labelle

(15) Me convient obtemperer aux jugemens et consentir d'aucuns et pas aux miens. ‘I must follow and agree with the judgments of others, and not with mine.’ (Ménagier de Paris, 14th c.; in Ayres-Bennett 1994: 70) (16) Poinct soupper (respondit Pantagruel) seroit le meilleur. ‘Not to have supper (answered Pantagruel) would be the best.' (Rabelais; in Ayres-Bennet 1994: 70)

Interrogatives like (17) are frequently discussed in this context (Ayres-Bennett 1994 and references therein; Muller 1991: 225). In medieval French, point functioned as a minimizer in that construction.

‘

(17) Quenois le tu point? ‘Do you know him a bit?

(Rose, CFMA, 4223; SATF, 4353; in Price 1978: 601; 1993: 192-193)

Price (1978, 1993) argues that the construction was reanalyzed as a negative interrogative towards the second half of the 15th century, when pas started to be used more often in the construction (18).

(18) Sont ilz pas douze heures au jour? 'Are there not twelve hours in the day?' (1486, Michel, Passion, 13454; in Price 1978: 603; 1993: 193)

Price is perhaps too categoric in concluding that such interrogatives have “nothing to do with the dropping of ne in modern French” (1993: 194). Once the construction started to be interpreted as a negative interrogative, it provided evidence for a grammar in which pas/point are negative operators and in which ne may be omitted. If the reanalysis took place during the 15th century, this new grammar could have spread to declaratives during the 16th century, that is, before 17th century grammarians like Ménage (1675) imposed ne in negative clauses. In support of a reanalysis of the construction during the second part of the 15th century, we note that the percentage of yes-no interrogatives containing pas/point without ne climbed from 15% during Old and Middle French to 26% during the 16th and 17th century (Martineau and Vinet 2005; Martineau 2011). This steep increase could reflect the new coexistence of two constructions: the old one, with pas, point functioning as minimizers, and a negative interrogative one that, before, was expressed exclusively with ne, and in which pas, point now functioned as negative quantifiers over events.5

5

Negative interrogatives without ne seem to have found support in various levels of the society. On the one side, Vaugelas (1647) stated that questions without ne meant the same thing as those with ne, but were more elegant; on the other side, Martineau (2011) found that they were more frequent in low-class speech (19% in the 17th century) than in middle and high class speech (6%).

The French Jespersen’s Cycle and Negative Concord

163

We thus hypothesize that, probably during the 16th century, the negative quantifiers pas and point started to be used in negative declaratives, and that this triggered the reanalysis of ne. As shown in (19b)–(20b), the cooccurrence of pas/ point reanalyzed as negative quantifiers with ne[uNEG] requiring licensing from a clause-initial negative operator can only yield a double negation reading. This holds whether the negative quantifier quantifies over events or over entities:

(19) a. b.

(20) a. b.

Il ne dort pas. ‘He doesn’t sleep.’ ØNege[iNEG] … ne[uNEG] … pasNege ‘There is no event of him not sleeping.’ Je n’ai point de vin. ‘I have no wine’ ØNege [iNEG]… ne [uNEG] … point Negx ‘There is no event of I having no wine’

The fact that these sentences did not have a double negation reading must have triggered the reanalysis of ne as not bearing [uNEG], which led to the disappearance of the covert negative operator licensing it, marking the end of the strict negative concord grammar. (21) Il ne dort pas. ‘He doesn’t sleep.’ ØNege[iNEG] ne[uNEG] pasreinf. ne

pasNege

‘There is no event of him sleeping at all’ ‘There is no event of him sleeping’

(22) Je n’ai point de vin. ‘I have no wine’ ØNege[iNEG] ne[uNEG] pointreinf. ‘There is no event of I having a bit of wine’ ne

pointNegx

‘I have no wine.’

The loss of [uNEG] left ne without a formal role to play in determining negative force and paved the way for its omission. This scenario explains why, when pas replaced ne as marker of clausal negation, it did not adopt the [uNEG] feature of ne, but functioned as a negative operator in its own right. Linguistic change typically takes the form of an s-curve: at the beginning, very few cases are observed; as the change gains momentum, there is a steep increase in the number of cases; finally the curve flattens when the feature has spread to almost every relevant example (Kroch 2000). In the case of ne omission, Martineau and Mougeon (2003) and Martineau (2009) showed that the steep increase in the curve happened in written texts reproducing the vernacular between the 18th and the 20th century with differences according to sociolects: it is observed sooner in Quebec French and in low class French than in mainstream French, a fact that, following Larrivée (2014), we attribute to differences in normative pressures.

164

M. Labelle

4 Pronominal Negative Indefinites We now turn to the changes that affected negative indefinites. Labelle and Espinal (2014) argued that medieval French negative indefinites evolved from negative polarity items to negative concord items, with, in many cases, competition between the two variants. We assume that once a negative concord variant existed, it was used by default in negative clauses, the polarity variant being restricted to non-negative polarity contexts. Medieval French pronominal negative indefinites were compatible with pas and point. This is to be expected since negative indefinites were licensed by the null operator licensing ne, pas and point being respectively a negation reinforcer and a word denoting a minimal quantity. Martineau and Déprez (2004) quote various examples of this, including the following one from the 15th century.

(23) Nul n'est pas sage qui s'y fie ‘Nobody is wise in the least who trusts him/it.’ (Arrêts d'amour, 220, 20, in Martineau and Déprez 2004: 36)

Martineau and Déprez treat the negative indefinites in such examples as negative polarity items, but Labelle and Espinal (2014) showed that a negative reading in isolation indicative of a negative concord item status started to be attested during the 14th century for rien and nul (during the 16th century for personne). Assuming that nul is a negative concord item bearing a [uNEG] feature, (23) is analyzed as in (24). The negative indefinite nul in subject position is licensed by the clause-initial covert operator also licensing ne, and pas co-occurs with these elements without a double negation reading.

(24) ØNeg[iNEG] Nuls[uNEG] n'[uNEG] est pasreinf. sage qui s’y fie During the 17th century, French grammarians started to reject the construction in (23). Vaugelas (1647), for instance, stated that: “…, rien, nul, aucun nient assez d’eux-mêmes sans y ajouter ny pas ny point” [rien ‘nothing’, nul ‘no, none’, aucun ‘no, none’ negate sufficiently by themselves without adding neither pas nor point]

In his commentary of this text in the same document, T. Corneille added “il faut y ajouster personne” [one must add personne ‘nobody’ (to that list)]. The stigmatization of the construction at the time is clearly observed in the oft-cited dialogue between Bélise (the mistress) and Martine (the servant), from Molière’s Les femmes savantes, written in 1672:

The French Jespersen’s Cycle and Negative Concord

165

(25) Martine: […] et tous vos biaux dictons ne servent pas de rien. ‘And all your beautiful sayings do not serve for anything at all’

Bélise:

[…] De pas mis avec rien tu fais la récidive, et c'est, comme on t'a dit, trop d'une négative. ‘Of pas added to rien you are using again, and this is, as we told you, one negative too many’

Martine:

Mon dieu! Je n'avons pas étugué comme vous, et je parlons tout droit comme on parle cheux nous. My God! I have not studied like you, and I speak exactly like we speak at home.

The dialogue shows that two variants of French were in competition during the 17th century: a normative variant illustrated by Bélise, and a stigmatized one illustrated by Martine who speaks a non-standard dialect with verbs in the first person singular ending in –ons. Notice that both Vaugelas and Bélise in Molière’s dialogue see a problem in the cooccurrence of rien (nul, aucun, personne) with pas (point), both items being considered negative. Ne is not mentioned, suggesting that ne is not perceived as the source of the problem. From this, we conclude that, for 17th century speakers of the normative variant, pas had replaced ne as marker of clausal negation. If pas (point) were a negation reinforcer, it would be compatible with rien, as we saw in (23). A second conclusion that we reach is that negative indefinites were interpreted neither as negative concord items nor as polarity items in negative clauses. Whether as negative concord items or as polarity items, they would be licensed by the marker of clausal negation pas, and would therefore be compatible with it. The fact that normative speakers rejected the cooccurrence of pas (point) with negative indefinites shows that negative indefinites were interpreted as negative quantifiers, yielding a double negation reading with the negative operator pas, as illustrated in (26).6

(26) tous vos biaux dictons ne servent pasNege de rienNegTHING. ‘all your beautiful sayings do not serve for nothing’ This explains why normative speakers rejected sentences containing both pas and a negative indefinite. What is the source of the reanalysis of negative indefinites as negative quantifiers? According to Ingham (2011), in 14th century Anglo-Norman, both preverbal and postverbal indefinites frequently allowed ne-drop (35%):

6

Notice that this is the case even though polarity uses of the same indefinites continued to be attested in questions and in conditional sentences (e.g. Martineau and Déprez 2004). Following Labelle and Espinal (2014), we assume that two variants of the words coexisted for a while: a negative quantifier variant and a residual polarity variant limited to non-negative contexts.

166

M. Labelle

(27) Nous voloms rien faire ‘We do not want to do anything’ (1340, Edw III, p. 313; in Ingham 2011:151, ex. 6) The author argues that the negative indefinites had been reanalyzed as negative quantifiers, probably under the influence of English. Since negative sentence particles like pas/point resisted this change and very rarely allowed ne-drop (2.6%), it appears that the omission of ne, in that dialect, was led by the reanalysis of negative indefinites as negative quantifiers. We cannot presume that the change that happened in the mainland about two centuries later has the same cause. On the contrary, according to Dufter and Stark’s (2007: 122–123) study of the future king Louis XIII, in the mainland, sentences without ne were more frequent with pas than with other negative words (73% vs. 64% for the period covering ages 5 to 9—their counts of ‘other’ including pronominal negative indefinites, and also plus ‘no more’ and jamais ‘never’). We are therefore tempted to believe that the reanalysis of ne as not carrying [uNEG]— that is, as being no longer dependent on the presence of a covert negative operator —caused the reanalysis of negative indefinites into negative quantifiers. Once ne lost its [uNEG] feature, the negative indefinites in clauses like those in (28) were reinterpreted as the only elements with negative force:

(28) a. b.

Personne Neg PERSON n’est parti. ‘Nobody left.’ Je ne vois rien Neg THING ‘I don’t see anything.’

To summarize, we hypothesize that the reanalysis of pas/point as quantifiers triggered the reanalysis of ne as not carrying [uNEG], which in turn triggered the reanalysis of negative indefinites from negative concord items to negative quantifiers.

5 The Stigmatized Variant Turning to the stigmatized variant illustrated by Martine’s sentence in (23), it has two possible analyses in the present framework. The possibility that Martine’s dialect is conservative and continues the medieval strict negative concord grammar cannot be ruled out. In that case, ne bears [uNEG], pas is a negation reinforcer, and rien is a negative concord item:

(29) ØNeg[iNEG] tous vos biaux dictons ne[uNEG] servent pasreinf de rien[uNEG]. However, since we saw that pas was reanalyzed as a negative quantifier before the 17th century, we favor a second analysis, according to which negative

The French Jespersen’s Cycle and Negative Concord

167

indefinites kept the [uNEG] feature they had inherited from medieval French, but they now Agree with the negative quantifier pas, reinterpreted as carrying [iNEG]:

(30) tous vos biaux dictons ne servent pasNeg[iNEG] de rien[uNEG]. The grammar in (30) is a non-strict negative concord grammar. If we are correct in supposing that the grammar of the stigmatized variant is as in (30), we predict that negative indefinites in subject position should not be compatible with pas (there should be no negative doubling from the subject position). Supporting this prediction is the fact that, in their discussion of 17th century sentences containing pas/ point and a negative indefinite, Martineau and Déprez (2004: 40–41) list many examples, all of them with postverbal indefinites. (As shown in [23], preverbal ones were attested until the 15th century.) Larrivée (2014: 250) quotes an example of negative doubling without ne in a 16th century text reproducing the vernacular: (31) – Voire mais, vostre femme est toute faschée; que luy avez-vous faict? – J’aury pas rien faict, ma dam; ‘– True, but your wife is very upset, what did you do to her? – I didn’t do nothing, my lady’ (Bonaventure des Périers, 1558, Nouvelles récréations; in Larrivée 2014:250)

The absence of ne shows that the grammar is not the medieval one and supports the idea that the non-strict negative concord grammar typical of Quebec French existed in France during the 16th (in addition to confirming that ne-omission in tensed clauses existed during the 16th century in some sociolects). Indeed, Bouvier (2002: 68) states without more precision that the asymmetric negative doubling typical of Quebec French exists in “a territorially widespread popular ‘register’ of European French”. This strongly suggests that the non-strict negative concord grammar was brought to the new continent by settlers. We conclude that, in the 16th century, the medieval grammar had evolved into two distinct variants: a normative one without negative concord, and a stigmatized one, with non-strict negative concord.

6 A Branching Evolution We argued that the reanalysis of pas/point as negative quantifiers started during the 15th century in some constructions and spread to negative declaratives in subsequent years. This change triggered the reanalysis of the other words that participated in the system of negation and marked the end of the medieval strict negative concord grammar. On the one hand, ne lost the [uNEG] feature that made it syntactically dependent on the presence of a covert negative operator, and it became omissible. On the other hand, negative indefinites like personne, rien were reanalyzed because they were no longer licensed by the clause-initial covert operator that

168

M. Labelle

used to license ne, this operator having disappeared with the reanalysis of ne. We believe that two distinct grammars emerged concurrently, as shown in (32). (32) ØNeg[iNEG] ne[uNEG] (pas) personne[uNEG]

(ne) pasNeg (ne) pasNeg[iNEG]

personneNegperson personne[uNEG]

In the normative variant, negative indefinites were reanalyzed as negative quantifiers. In the stigmatized variant, they retained the [uNEG] feature they had inherited from medieval French and started to Agree with the new negative operator, pas, reanalyzed as carrying [iNEG]. The normative variant is continued in standard French and in the common variety of colloquial European French; the stigmatized one is continued in Quebec French and in dialects similar to it. Thus, when framed into a theory of negative concord, an apparent simple cycle turns out to be a spiral-like change with two ending points, both different from the starting point. Did the medieval grammar survive in some French dialects? Unfortunately, there is a dearth of detailed studies on negative concord in non-standard varieties of French that would help answering this question. Various authors mention that negative doubling from the subject position is attested in Europe (Larrivée 2014: 146; Martineau and Déprez 2004; Muller 1991: 261; Rowlett 1998: 167), something that is not grammatical in Quebec, but they only quote anecdotal evidence:

One would like to know whether these examples are representative of specific dialects, or whether they correspond to exceptional uses by authors normally using the standard grammar (the example from Proust in [31a] could represent the second case). Dagnac (2015; Dagnac and Burnett 2016) answers this question for Picard by showing that this dialect allows the cooccurrence of negative indefinites (6% with rien, 16% with personne) with the negative words mie/point, not only from the object position, but also from the subject position, as shown in (34).7 This pattern of negative doubling could be a remnant of the medieval grammar, with mie functioning as negation reinforcer. If so, (34) has the grammar schematized in (35).

7

Dagnac observes cross-speakers differences in the possibility of negative doubling according to whether the negative word is mie or point.

The French Jespersen’s Cycle and Negative Concord

169

(34) Pèrson-ne i n' voreu mie d'élle pour és mèrieu. (A. Depoilly) Nobody he NE would_want MIE of her to REFL marry ‘Nobody would want to marry her.’ (Dagnac 2015, ex. 18a) (35) ØNeg[iNEG] Personne[uNEG]… ne[uNEG] … miereinf. … This raises the question of whether there exists in France a strict negative concord dialect without ne, exhibiting negative doubling between pas/point and both preverbal and postverbal indefinites. If such a dialect exists, the present framework would lead us to conjecture that pas/point acquired the [uNEG] feature of medieval ne instead of being reanalyzed as negative quantifiers.

7 Conclusion Our diachronic study of French negative concord leads us to two conclusions. Firstly, an apparent simple cycle, like the French Jespersen’s cycle, may hide more complexity than it appears. The single negative marker pas at the end of the cycle has properties that distinguish it from the original single negative marker ne: pas is a negative operator, but ne carried [uNEG]. Thus, the change is more spiral-like than cyclic. Secondly, the change from medieval French to the common variety of colloquial European French and to Quebec French (and similar dialects) is branching rather than linear. During the 16th century, two different dialects emerged: a normative one that is continued into the current variety of contemporary European French, where pas is a negative operator, and a stigmatized one of which Quebec French and similar dialects are a continuation, where pas is a negative operator carrying [iNEG]. This is of course not the end of the story. The present account is largely speculative. Careful studies of patterns of negative doubling in French and French-related dialects are required to validate the present hypotheses, and to better understand the grammar and the evolution of negative doubling. In particular, it would be interesting to know which patterns of negative concord exist or existed in the various French dialects, and what is or was their geographical distribution. This might allow us to localize the sources of the various grammars and their spread, and possibly, to shed light on the moment of their emergence. We hope that the present paper will spur work on that topic. Acknowledgements We thank the reviewers of this paper for their comments on previous versions.

170

M. Labelle

References Ayres-Bennett, Wendy. 1994. Negative evidence: Or another look at the non-use of ne in 17th century French. French Studies 48 (1): 63–85. Baunaz, Lena. 2011. The grammar of French quantification. Dordrecht: Springer. Beghelli, Filippo. 1997. The syntax of distributivity and pair-list readings. In Ways of scope taking, ed. Ana Szabolcsi, 379–408. Dordrecht: Kluwer. den Besten, Hans. 1986. Double negation and the genesis of Afrikaans. In Substrata versus universals in Creole languages. Papers from the Amsterdam Creole workshop, April 1985, ed. Peter Muysken and Norval Smith, 185–230. Amsterdam: Benjamins. Boerm, Michael Lloyd. 2008. Pourquoi ‘Pas’: The socio-historical linguistics behind the grammaticalization of the French negative marker. Ph.D. dissertation, University of Texas at Austin. Bouvier, Yves-Ferdinand. 2002. A featural account of polarity phenomena. Doctoral dissertation, Université de Genève, no. L 498. http://archive-ouverte.unige.ch/unige:553. Chierchia, Gennaro. 2013. Logic in grammar: Polarity, free choice, and intervention. Oxford: Oxford University Press. Dagnac, Anne. 2015. ‘Pas’, ‘mie’, ‘point’ et autres riens: de la négation verbale en picard. In La négation: études linguistiques pragmatiques et didactiques, actes du colloque franco-roumain du 15–17 mai 2013, ed. Jan Goes and Mariana Pitar. Timişoara, Arras: Artois Presses Université. https://hal-univ-tlse2.archives-ouvertes.fr/hal-00988781. Dagnac, Anne, and Heather Burnett. 2016. Concordance négative optionnelle: contrastes forts et faibles entre picard et québécois. In Proceedings of the 5th Congrès Mondial de Linguistique Française. Tours, France: Université François Rabelais. http://www.shs-conferences.org/ articles/shsconf/pdf/2016/05/shsconf_cmlf2016_13003.pdf. Daoust-Blais, Denise. 1975. L’influence de la négation sur certains indéfinis en français québécois. Ph.D. Thesis, Université de Montréal. Déprez, Viviane. 1999. The roots of negative concord in French and French based Creoles. In Language creation and language change: Creole, diachrony and development, ed. Michel DeGraff, 375–428. Cambridge, Mass.: MIT Press. Déprez, Viviane. 2000. Parallel (a)symmetries and the internal structure of negative expressions. Natural Language and Linguistic Theory 18 (2): 253–342. Di Sciullo, Anna-Maria, and Mireille Tremblay. 1996. Configurations et interprétation: les morphèmes de négation. Recherches linguistiques de Vincennes 25: 27–52. Dufter, Andreas, and Elisabeth Stark. 2007. La linguistique variationnelle et les changements linguistiques ‘mal compris’: le cas du ne de négation. In Études sur le changement linguistique en français, ed. Bernard Combettes and Christiane Marchello-Nizia, 115–128. Nancy: Presses Universitaires de Nancy. Ernst, Gerhard. 1985. Gesprochenes Franzosisch zu Beginn des 17. Jahrhunderts. Direkte Rede in Jean Héroards “Histoire particulière de Louis XIII” (1605–1610). Tübingen: Niemeyer. Giannakidou, Anastasia. 2006. N-words and negative concord. In The Blackwell companion to syntax, vol. III, chapter 45, ed. Martin Everaert, Henk Van Riemsdijk, Rob Goedemans, and Bart Hollebrandse, 327–391. Oxford: Blackwell. Héroard, Jean. 1989. Journal de Jean Héroard, 1601–1628, 2 vol., ed. Madeleine Foisil. Paris: Fayard. Hirschbühler, Paul, and Marie Labelle. 2003. Residual Tobler-Mussafia effects in French dialects. In Romance linguistics: Theory and acquisition, ed. Ana Maria Pérez-Leroux and Yves Roberge, 149–164. Amsterdam: John Benjamins. Hirschbühler, Paul, and Marie Labelle. 2006. Enclisis and proclisis of object pronouns at the turn of the 17th century in the speech of the future Louis XIIIth. In Historical Romance linguistics: Retrospective and perspectives, ed. Deborah Arteaga and Randall Gess, 187–207. Amsterdam: John Benjamins.

The French Jespersen’s Cycle and Negative Concord

171

Ingham, Richard. 2011. Ne-drop and indefinites in Anglo-Norman and Middle English. In The evolution of negation: Beyond the Jespersen cycle, ed. Pierre Larrivée and Richard P. Ingham, 145–163. Berlin/Boston: Walter de Gruyter. Jespersen, Otto. 1917. Negation in English and other languages. Copenhagen: A.F. Høst, 1966. Reprinted in Selected Writings of Otto Jespersen. London: Allen del Orwin, 1961. Krifka, Manfred. 1989. Nominal reference, temporal constitution and quantification in event semantics. In Semantics and contextual expressions, ed. Renate Bartsch, 75–115. Dordrecht: Foris. Kroch, Anthony. 2000. Syntactic change. In The handbook of contemporary syntactic theory, ed. Mark Baltin and Chris Collins, 699–729. Oxford: Blackwell. Labelle, Marie. 2017. Negative concord in Quebec French. Probus 29 (2): 323–354. Labelle, Marie, and Maria Teresa Espinal. 2014. Diachronic changes in negative expressions: The case of French. Lingua 145: 194–225. Ladusaw, William A. 1992. Expressing negation. In Proceedings of the 2nd semantics and linguistic theory conference (SALT II), held May 1–3, 1992 at the Ohio State University. Ohio State University Working Papers in Linguistics, vol. 40, ed. Chris Barker and David Dowty, 237–259. Columbus, Ohio. journals.linguisticsociety.org/proceedings/index.php/SALT/issue/ view/81. Laka, Itziar. 1990. Negation in syntax: On the nature of functional categories and projections. Ph. D. dissertation, MIT. Published in 1994 by Garland, New York. Larrivée, Pierre. 2014. The continuity of the vernacular: The evolution of negative doubling in French. In The diachrony of negation, ed. Maj-Britt Mosegaard Hansen and Jacqueline Visconti, 235–256. Amsterdam: Benjamins. Martineau, France. 2005. Perspectives sur le changement linguistique: aux sources du français canadien. Canadian Journal of Linguistics 50: 173–213. Martineau, France. 2009. Modeling change: A historical sociolinguistics perspective on French negation. In Corpus analysis and variation in linguistics, ed. Yuji Kawaguchi, Makoto Minegishi, and Jacques Durand. Amsterdam: Benjamins. Martineau, France. 2011. Ne-absence in declarative and yes/no interrogative contexts: Some patterns of change. In The evolution of negation: Beyond the Jespersen cycle, ed. Pierre Larrivée and Richard P. Ingham, 179–207. Berlin/Boston: Walter de Gruyter. Martineau, France, and Viviane Déprez. 2004. ‘Pas rien’/‘Pas aucun’ en français classique: variation dialectale et historique. Langue française 143: 33–47. Martineau, France, and Raymond Mougeon. 2003. Sociolinguistic research on the origins of ne deletion in European and Quebec French. Language 79 (1): 118–152. Martineau, France, and Marie-Thérèse Vinet. 2005. Microvariation in French negation markers: A historical perspective. In Grammaticalization and parametric variation, ed. Montserrat Battlori, Maria-Lluïsa Hernanz, Carme Picallo, and Francesc Roca, 194–205. Oxford: Oxford University Press. Meisner, Charlotte, Aurélia Robert-Tissot, and Elisabeth Stark. 2015. L’absence et la présence du NE de négation. In Encyclopédie grammaticale du français online. http://encyclogram.fr. Ménage, Gilles. 1675. Observations de Monsieur Ménage sur la langue française, Paris. Moritz, Luc, and Daniel Valois. 1994. Pied-piping and specifier-head agreement. Linguistic Inquiry 25 (4): 667–707. Muller, Claude. 1991. La négation en français. Genève: Droz. Pollock, Jean-Yves. 1989. Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20 (3): 365–424. Price, Glanville. 1962. The negative particles pas, mie and point in French. Archivum Linguisticum 14: 14–34. Price, Glanville. 1978. L’interrogative sans ne. Studii şi cercetări lingvistice 29: 599–606. Price, Glanville. 1993. Pas (point) without ne in interrogative clauses. French Language Studies 3 (2): 191–195. Rowlett, Paul. 1998. Sentential negation in French. Oxford: Oxford University Press.

172

M. Labelle

de Swart, Henriëtte, and Ivan A. Sag. 2002. Negation and negative concord in Romance. Linguistics and Philosophy 25 (4): 373–417. Vance, Barbara S. 1997. Syntactic change in Medieval French: Verb-second and null subjects. Dordrecht: Kluwer. de Vaugelas, Claude Favre. 1647 Remarques sur la langue françoise. Fac-similé de l’édition originale, 1934. Genève: Droz. Zeijlstra, Hedde H. 2004. Sentential negation and negative concord. Ph.D. Dissertation, University of Amsterdam. Zeijlstra, Hedde H. 2010. On French negation. In Proceedings of the 35th annual meeting of the Berkeley Linguistics Society, ed. Iksoo Kwon, Hannah Pritchett, and Justin Spence, 447–458. Berkeley, CA: BLS. http://ling.auf.net/lingbuzz/000885.

Part III

Diachronic Romance Studies

The Variable Position of Initial Subordinate Clauses in Old French: Arguments Against a Semantic Account Bryan Donaldson

Abstract This paper examines the influence of semantics in a case of syntactic variation in Old French, a verb-second language in which subordinate clauses that immediately precede a main declarative occur in either the left periphery of the declarative, or in the core of the declarative itself, in the first (preverbal) position (Combettes 2010; Donaldson 2012; Skårup 1975; Vance 1997; Vance et al. 2010). Explanations of this variation have sometimes implicated the nature of the link between the initial subordinate and the declarative as well as the semantic properties of the subordinate (e.g., Combettes 2010; Imbs 1956; Ménard 1988). Previous analyses have claimed that semantic relationships like concession, anteriority, and causality represent a particularly close relationship between the subordinate and the declarative, a claim that has been used to explain why such initial subordinates occur in the core of the declarative rather than in its left periphery. The present paper challenges this hypothesis, using data from 30 major texts spanning the 10th to early 14th centuries. The results reveal that semantic relations do not uniformly predict the syntactic position of the initial subordinate.













Keywords Syntax Semantics Variation Left periphery Subordinate clause Verb-second Old French







1 Introduction This paper examines the influence of semantics on the clausal position of initial subordinate clauses in Old French (OF). Initial subordinates sometimes appear inside the core of the main declarative, in the first (preverbal) position, and sometimes to the left of the declarative per se (Combettes 2010; Donaldson 2012; Grad 1956, 1961; Nissen 1943; Skårup 1975; Vance et al. 2010). Some analyses B. Donaldson (&) Department of Languages and Applied Linguistics, University of California, 1156 High Street, Santa Cruz, CA 95064, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_9

175

176

B. Donaldson

(e.g., Combettes 2010; Imbs 1956; Ménard 1988) link the position of the initial subordinate to the type of semantic relationship between the subordinate and declarative. This study examines some of these semantic hypotheses empirically, using a corpus of 30 major texts that span the OF period (10th–early 14th centuries). The results fail to identify a clear connection between semantic relations and the position of the subordinate. In particular, the results reveal that the tightness of the semantic relation—as measured by Van Valin and LaPolla’s (1997) Interclausal Semantic Relations Hierarchy—does not predict the position of the subordinate.

1.1

Initial Subordinate Clauses

An initial subordinate clause is a tensed dependent clause that immediately precedes a main declarative and whose propositional content is somehow related to that of the declarative. All the initial subordinates in this study are introduced by an overt subordinating conjunction (Foulet 1928: 290), such as ainz que “before” in (1) and (2) and quant “when” in (3) and (4). (1)

Mais, [ainz qu’il l’ait fors del fuerre gitee,] [li FU] Bauduc devant a l’encontree. “But, before he could pull the sword from the sheath, Bauduc was in front of him.” (Aliscans, 7347–7348)

(2)

[Ainz qu’il soit outre,] [grant estor] [i AVRA.] “Before he gets away, there will be a big skirmish.” (Aliscans, 1131)

(3)

E [quant li moinie vindrent lur complie chanter,] [QUIDIERENT il] pur veir que se dormist li ber. “And when the monks came to sing their Compline, they truly believed that the baron was sleeping.” (Becket, 1991–1992)

(4)

[Quant il est vostre huem liges,] [il] [vus DEIT] fei porter. “When he is your vassal, he must show you fidelity.”(Becket, 1856–1857)

Initial subordinates interact in one of two ways with the OF verb-second (V2) grammar. In one configuration, the initial subordinate occurs as the first constituent of the V2 declarative, occupying the position immediately in front of the finite verb (traditionally, the “first position”). Thus, in (1) and (3), the initial subordinate is followed immediately by the matrix verb and any associated preverbal object clitic(s). In the second configuration, the initial subordinate occurs not in the first position of

The Variable Position of Initial Subordinate Clauses …

177

the V2 declarative but rather in a left-peripheral position, outside the declarative per se. In this case, exemplified by (2) and (4), the initial subordinate is typically separated from the matrix verb by (minimally) one constituent, which occupies the matrix first position: the NP grant estor “big skirmish” in (2) and the subject pronoun il “he” in (4). As examples (1)–(4) show, certain types of initial subordinate can occur in either position within a single text. However, as further data will reveal, some types (e.g., ainz que “before”) show a much stronger tendency to occur in SpecFocus than others (e.g., quant “when”). This paper tests hypotheses about the relationship between semantics and syntactic position for four classes of initial subordinates: (a) temporal subordinates formed with quant “when;” (b) conditional subordinates formed with se “if;” (c) four subordinates that express anteriority: ainz que, ainçois que, avant que, and devant ce que “before;” and (d) the polysemous subordinate por ce que “because, so that, since.” The paper is organized as follows: Sect. 2 discusses the verb-second (V2) grammar of OF and the clausal architecture of main declaratives. Section 3 discusses the nature of the semantic relation between an initial subordinate and the following declarative and introduces Van Valin and LaPolla’s (1997) Interclausal Semantic Relations Hierarchy (ISRH). Section 4 presents the diagnostics used to identify the syntactic position of initial subordinates. Section 5 details the texts used and analysis of the data. Section 6 contains the results for initial quant (Sect. 6.1), se (Sect. 6.2), ainz que, ainçois que, avant que, and devant que (Sect. 6.3), and por ce que (Sect. 6.4). Section 7 presents a general discussion and conclusions.

2 Old French V2 and Clausal Architecture Old French is usually analyzed as a V2 language (Adams 1987; Buridant 2000; Combettes 2010; Marchello-Nizia 1985; Roberts 1993; Vance 1997). In a V2 language, especially as instantiated in modern Germanic, the finite verb is usually the second constituent in a main declarative (Holmberg 2015). The first position of main declaratives is flexible, able to host nearly any type of maximal projection (XP), as Buridant (2000: 741–752) details for OF. Each of the following examples respects this generalization; brackets delimit the first and second positions, and the verb (in small capitals) consistently appears in second position. In (5), the first position hosts the adverb lors “then.” (5)

[Lors] [DIT] la dame de rechief. “Then the woman spoke again.” (Yvain, 4615; in Buridant 2000: 741)

In (6), the first position contains the prepositional phrase en le cambre “in the bedroom.”

178

(6)

B. Donaldson

[En le cambre] [ENTRE] Aucassins. “Aucassin enters the bedroom.” (Aucassin et Nicolette, 29, 1; in Buridant 2000: 741)

In (7), the nominal subject Aucassins appears in first position. (7)

[Aucassins] [TENT] le main. “Aucassin holds out his hand.” (Aucassin et Nicolette, 10, 35; in Buridant 2000: 741)

Furthermore, an initial subordinate can occupy the first position, as examples (1) and (3) demonstrate. Old French also evinces main declaratives in which multiple XPs precede the finite verb; in such examples, the verb appears as the third (or higher) constituent. Example (8) contains a left-dislocated subject, and (9) a left-dislocated direct object. (8)

[Domine Dieu] [il] [les LUCRAT]. “God, He won them.” (Vie de Saint Léger, 214)

(9)

[Ceste bataille] [veirement] [la FERUM.] “Truly, we will wage this battle.” (Roland, 882; in Marchello-Nizia 1998: 176)

Benincà’s (2006) model of the left periphery in medieval Romance follows Rizzi (1997) in positing a series of syntactic positions to the left of the main clause proper.1 These leftward projections are an extension of the CP domain and host syntactically optional XPs like left dislocations and circumstantial clauses, which contextualize the main declarative and link it to previous discourse, without belonging to its core syntactic structure. Slight variations notwithstanding, this approach underlies recent analyses of verb-third declaratives in OF and other medieval Romance varieties (e.g., Donaldson 2015; Mathieu 2012; Vance et al. 2010). Henceforth, I adopt Benincà’s (2006) terminology by referring to the

1

Skårup (1975) anticipates this approach; his extraposition zone is roughly equivalent to— although considerably less complex than—recent models of the left periphery.

The Variable Position of Initial Subordinate Clauses …

179

traditional “first position” of main declaratives as SpecFocus; the second position, which hosts the verb, is Focus’. In Donaldson (2012), I claim that OF initial subordinates that precede an independently fully formed main declarative occupy a SpecFrame position in the left periphery of the declarative (see also Salvesen 2013; Skårup 1975). For the present study, it is sufficient to differentiate left-peripheral initial subordinates from those in SpecFocus, adopting a simplified view of the left periphery. Accordingly, examples (2) and (4) possess the clausal architecture given in (10): the initial subordinate is left-peripheral and an independent XP occupies SpecFocus. (10)

[Left periphery initial subordinate] [ SpecFocus XP ] [Focus’ V ]

In contrast, in (1) and (3), the initial subordinate appears in SpecFocus, inside the V2 core of the main declarative, as in (11): (11)

[Left periphery ...] [SpecFocus initial subordinate] [ Focus’ V ]

3 Semantics Some previous analyses of the position of initial subordinates discuss the nature of the relation between the subordinate and the declarative. A consensus has emerged that, when the initial subordinate is “integrated” into SpecFocus (or its terminological equivalent) of the main declarative, the subordinate possesses a tight link with the declarative. Thus, Imbs (1956: 504) speaks of such subordinates as “tightly integrated” with the declarative, and Ménard (1988: 53) considers them to be “tightly linked.” However, left-peripheral initial subordinates, by definition not integrated into the core structure of the main declarative, can also present a tight link with the main clause. Initial quant “when” clauses, very frequent in OF, are routinely followed by a main declarative introduced by the sentence adverb si (Grad 1961). In this context, according to Ménard (1988: 213), si serves to “reinforce the link between the subordinate and the main clause,” which corroborates Grad’s (1961: 12) claim that (left-peripheral) quant subordinates enjoy just as tight a relation with the main clause as integrated (SpecFocus) ainz que subordinates. As Skårup (1975: 416) observes, left-peripheral XPs, including initial subordinates, are still connected to the following main declarative. To assess the strength of the semantic link between initial subordinates and main declaratives, Combettes (2010) references Van Valin and LaPolla’s (1997: 480) Interclausal Semantic Relations Hierarchy (ISRH), which provides a graded hierarchy of semantic relations between two propositions, shown in (12).

180

(12)

B. Donaldson

Interclausal Semantic Relations Hierarchy (Van Valin & LaPolla 1997: 480) Tightest relation Causative Aspectual Psych-action Purposive Jussive Direct perception Propositional attitude Cognition Indirect discourse Conditional Simultaneous states of affairs Sequential states of affairs a. overlapping b. non-overlapping c. non-overlapping, with interval Unspecified temporal order Loosest relation

The ISRH serves as a scale of “semantic cohesion” (Van Valin and LaPolla 1997: 480), with the degree of semantic cohesion decreasing progressively from the highest position (causatives) to the lowest (events with unspecified temporal relations). Thus, whereas propositions linked by a causative relationship evince a high degree of semantic cohesion, the semantic link between two temporally unordered propositions is much weaker. Van Valin and LaPolla (1997) contend that propositions with tighter semantic relations display a higher degree of syntactic integration than propositions with less semantic cohesion. Because the ISRH permits fine-grained distinctions in qualifying the link between two clauses, and because Combettes (2010) has already used it to analyze initial subordinates in OF, it is adopted in the present analysis.

The Variable Position of Initial Subordinate Clauses …

181

4 Diagnostics The position of initial subordinates (left-peripheral vs. SpecFocus) is determined on the basis of accepted theoretical assumptions about the position of subjects and object or adverbial clitics in OF. In the following conditions, an initial subordinate is necessarily left-peripheral: First, if another XP occupies SpecFocus, and the subordinate occurs to the left of this XP, the subordinate is necessarily to the left of SpecFocus. Accordingly, the initial quant subordinate in (13) is left-peripheral, as it occurs to the left of the sentence adverb si “thus,” which occupies SpecFocus (Benincà 2006: 64; Skårup 1975: 302). (13)

[Left periphery Quant il vinrent la,] [SpecFocus si] [Focus’ PARLERENT] au marchis… “When they came there, they spoke to the marquis…” (Clari, §3)

Second, when object or adverbial clitics appear postverbally in the main clause, SpecFocus is understood to be empty (Skårup 1975: 383–392; Benincà 2006: 69). As such, in (14), the initial subordinate occupies the left periphery rather than SpecFocus, despite the apparent (surface) adjacency of the subordinate and the declarative. (14)

E [Left periphery së il poet,] [SpecFocus ...] [Focus’ MURRAT i] veirement. “And if he can, he will certainly kill him there.” (Roland, 615)

Table 1 summarizes the orders in which an initial subordinate occupies a position in the left periphery rather than SpecFocus. Evidence that an initial subordinate occupies SpecFocus comes from the presence of certain types of postverbal subject and from the position of object and adverbial clitics in the main declarative. Postverbal subjects, specifically those in Germanic inversion, attest to the saturation of SpecFocus in OF (Vance et al. 2010: 303). Germanic inversion (as opposed to Romance inversion, discussed subsequently), involves a postverbal subject that occurs before other postverbal VP material such as past participles, short adverbs, or nominal objects. In OF, postposed pronominal subjects always involve Germanic inversion (Roberts 1993: 122; Vance et al. 2010: 303). When no XP intervenes between an initial subordinate and a matrix verb with Germanic inversion, as in (15), the initial subordinate occupies SpecFocus.

Table 1 Initial subordinate is left-peripheral

Initial subordinate subordinate +

Declarative word order • XP + verb (+ subject) • verb + clitic(s)

182

(15)

B. Donaldson

E [Left periphery...][SpecFocus quant li moinie vindrent lur complie chanter,] [Focus’ QUIDIERENT il] pur veir que... “And when the monks came to sing their Compline, they truly believed that...” (Becket, 1991–1992)

Preverbal object and adverbial clitics attest to the saturation of SpecFocus (Benincà 2006). If the matrix verb evinces preverbal clitic(s), and no XP intervenes between the initial subordinate and the matrix verb, the subordinate occupies SpecFocus. Example (1) is repeated as (16); the indirect object clitic li “to him” is preverbal and is immediately preceded by the initial ainz que subordinate. Note the contrast in clitic position between (14), in which the subordinate is left-peripheral (and SpecFocus remains empty), and (16), where the subordinate occupies SpecFocus. (16)

Mais, [Left periphery...] [SpecFocus ainz qu’il l’ait fors del fuerre gitee,] [ Focus’ li FU] Bauduc devant a l’encontree. “But, before he could pull the sword from the sheath, Bauduc ” was in front of him. (Aliscans, 7347–7348)

Table 2 shows the orders in which the initial subordinate unambiguously occupies SpecFocus. In three ambiguous contexts, the position of the initial subordinate cannot be determined with certainty. In the first context, the main declarative contains only the initial subordinate and negative ne before the verb. Ne is problematic, acting sometimes like a clitic but sometimes like a XP (Foulet 1928; Grad 1956; Ingham 2014). As a result, one cannot know if ne saturates SpecFocus, in which case a preceding subordinate would be left-peripheral, or if ne is a clitic, in which case a preceding subordinate would occupy SpecFocus. The second context involves a main declarative without clitics or any preverbal XP other than the initial subordinate (see Grad 1956: 71); in this context, there are no clues about the saturation of SpecFocus. Finally, postverbal nominal subjects that do not unambiguously represent Germanic inversion do not provide information about the saturation of SpecFocus (Nissen 1943: 56; Skårup 1975: 383; Vance et al. 2010: 303). Table 3 summarizes the contexts in which the position of the initial subordinate cannot be determined with certainty.

Table 2 Initial subordinate occupies SpecFocus Initial subordinate subordinate +

Declarative word order • verb + subject pronoun • verb + NP subject + past participle/short adverb/NP object • clitic(s) + verb

The Variable Position of Initial Subordinate Clauses … Table 3 Orders in which the syntactic position of the initial subordinate is ambiguous

183

• ne + (clitic/s) + verb • verb • verb + (XP) + NP subject Note XP = maximal projection other than past participle, short adverb, NP object subordinate +

5 Data Data came from 30 OF texts from the 10th to the early 14th century, with a concentration of texts in the mid- to late-12th century, when there appears to be the greatest variation in the position of initial subordinates. Each text was analyzed in its entirely (Table 4). All sequences of initial quant, se, ainz que, ainçois que, avant que, devant ce que, and por ce que + main declarative were collected. Orthographic variants (e.g., einçois que, pour ce que) were included. To be considered initial, the subordinate could be preceded only by a coordinating conjunction. Rare instances in which the subordinate was separated from the main declarative by et “and” were excluded. A total of 6206 tokens were analyzed: all were coded for the type of subordinate, the word order of the main declarative, and, using the diagnostics in Sect. 4, the clausal position of the subordinate. Table 4 Corpus Date

Text

980 1040 1100 1121

Vie de Saint Léger (Koschwitz 1964) Vie de Saint Alexis (Perugi 2000) Chanson de Roland (Segre 1971) Voyage de Saint Brendan (Short and Merrilees 1979) Gormont et Isembart (Bayot 1914) Li coronemenz Looïs (Langlois 1968) Chanson de Guillaume (Suard 2008)

1130 1130 Before 1150 1150 1150 1160 1165 1170 1174 1180 1180

Roman de Thèbes (Petit 2008) Charroi de Nîmes (De Poerck et al. 1970) Roman d’Enéas (Salverda de Grave 1964) Erec et Enide (Fritz and Zink 1994) Renart (earliest branches; Lodge and Varty 2001) Vie de Saint Thomas Becket (Walberg 1964) Roman de Tristan (Muret and Defourques 1974) Lais (Harf-Lancner 1990)

Author (if known)

Benedeit

Verse/ prose Verse Verse Verse Verse Verse Verse Verse

Chrétien de Troyes

G. de Pont-Sainte-Maxence Béroul Marie de France

Verse Verse Verse Verse Verse Verse Verse Verse

(continued)

184

B. Donaldson

Table 4 (continued) Date

Text

1180 1185 1188

Aliscans(Régnier 2007) Perceval (Conte du Graal; Méla and Zink 1994) Roman d’Alexandre (Foster and Short 1976– 1977) Prise d’Orange (Katz 1947) Jeu de Saint Nicolas (Dufournet 2005) Aucassin et Nicolette (Dufournet 1984)

1190 1200 1200 1200 1205 1208 1214 1225 1250 1267 1276 1308

Roman du graal (Joseph only; Cerquiglini 1981) Conquête de Constantinople (Lauer 1956) Conquête de Constantinople (de Wailly 1874) Le bel inconnu (Perrie Williams 1967) La queste del saint graal (Marchello-Nizia 2011) L’âtre périlleux (Woledge 1936) Roman de Cassidorus (Palermo 1963) Jeu de la Feuillée (Langlois 1966) Vie de Saint Louis (Corbett 1977)

Author (if known)

Chrétien de Troyes Thomas Kent

Jean Bodel

based on Robert de Boron Robert de Clari G. de Villehardouin Renaut de Beaujeu

Adam de la Halle Jehan de Joinville

Verse/ prose Verse Verse Verse Verse Verse Verse/ prose Prose Prose Prose Verse Prose Verse Prose Verse Prose

6 Results This section presents results for, respectively, each type or class of initial subordinate in question: quant, se, anteriority (ainz que, ainçois que, avant que, devant ce que), and por ce que.

6.1

Initial quant

Temporal subordinates formed with quant “when” can encode either (a) an action that occurs simultaneously with the action of the main clause, or (b) a sequence of actions, with the action in the quant subordinate preceding that of the main clause (Combettes 2010: 121; Imbs 1956: 271–272, 337, passim); both possibilities are illustrated subsequently in (18) through (21). Given that sequences of simultaneous and sequential events are differentiated (albeit very slightly) on the ISRH, it is possible that OF distinguishes them via the syntactic position of initial quant subordinates. Skårup (1975: 27) pointed to another possible semantic nuance, arguing that, when an initial quant clause occurs in SpecFocus of the declarative, the quant clause sometimes carries a concessive reading (“even when”), as in (17). (Note that such examples appear to be rare.)

The Variable Position of Initial Subordinate Clauses …

(17)

185

Et [quant il avra compliment de ce que il avra desiret,] [ n’AVRA il] pas tant gaaignié que une fausse maille ne vaille plus. “And even when he will have accomplished what he desires, the value of his accomplishment will not be more than a false half-penny coin.” (Roman de Troie en prose, 198:32, in Skårup 1975: 270)

Initial quant “when” subordinates are attested in every text except the Jeu de la Feuillée (1276) and are, with a total of 3923 tokens, the most frequent type. However, there are only 12 unambiguous cases of an initial quant clause in SpecFocus; initial quant subordinates are almost categorically left-peripheral. Table 5 reports data only from those texts that present variation in position and contain at least one SpecFocus example. A first observation is that initial quant occurs so infrequently in SpecFocus and so predominantly in the left periphery as to render implausible the possibility of a productive strategy of encoding semantic nuances via syntactic position. Furthermore, of the texts that employ SpecFocus variants, only Becket contains multiple tokens. The remaining texts present no clear evidence at all for SpecFocus variants. Finally, SpecFocus examples occur only in verse texts in these data, raising the possibility that they arise in response to versification requirements. Earlier observations about concession are not supported: Whereas Skårup (1975: 270) noted that some SpecFocus examples possess a concessive reading, none of the SpecFocus tokens in the present data do so. These data suggest that OF did not consistently and productively place quant subordinates in SpecFocus to impart a concessive reading. Likewise, no consistent distinctions emerge with respect to temporal readings. Whereas Van Valin and LaPolla (1997: 480) consider simultaneity to involve a (marginally) closer semantic relation than sequentiality, OF does not appear to encode this distinction via syntactic position, as both senses are frequent in left-peripheral quant. Consider (18) and (19), where the action expressed by the quant clause occurs simultaneously with that of the main declarative, and (20) and (21), where the actions are sequential, with that of the main declarative occurring after that of the initial quant clause. Table 5 Texts with initial quant clause attested in SpecFocus

Text, date Brendan 1121 Renart 1170 Becket 1174 Aliscans 1180 Perceval 1185 Alexandre 1188 L’âtre pér. 1250 a Unambiguous cases

Totala

Clausal position Left periphery

SpecFocus

23 28 138 99 93 120 76

1 1 6 1 1 1 1

24 29 144 100 94 121 77

186

(18)

B. Donaldson

simultaneous actions [Quant le dut prendre,] [si] [li CAÏT] a tere. “When he was about to take it, it fell to the ground.” (Roland, 333)

(19)

simultaneous actions E [quant il s’en parti de la cambre le rei,] [justises e baruns tels que numer ne dei] [l’ ESCRÏERENT] en haut, a hu e a desrei: ‘Li traïtres s’en vait: vëez lei, vëez lei!” “And when he went out of the king’s chamber, judges and barons such that I should not name cried out at him, tumultuously and insolently: ‘The traitor is leaving, see him there!’” (Becket, 1916)

(20)

sequential actions [Quant li empereres ad faite sa venjance,] [sin] [APELAT] ses evesques de France. “Once the emperor had accomplished his revenge, he summoned his bishops from France.” (Roland, 3975–3976)

(21)

sequential actions [Quant ecrites serunt,] [puis] [les nus MUSTERREZ]. “Once they are written, then you will read them to us.” (Becket, 1003)

The handful of SpecFocus examples do not reveal an obvious semantico-temporal distinction from the left-peripheral occurrences. Both simultaneous and sequential actions appear in the SpecFocus examples, in approximately equal numbers. In each text, regardless of whether the SpecFocus example(s) encode simultaneous or sequential actions, both types of action are found in the left-peripheral quant data in that text. Becket stands out as the only text in the corpus with multiple tokens of initial quant in SpecFocus; the SpecFocus tokens present both simultaneous (n = 4) and sequential (n = 2) actions.

6.2

Initial se

Subordinates with se “if” express a condition or hypothesis in OF (Buridant 2000: 627; Ménard 1988: 235). The ISRH situates conditionals (slightly) higher than temporal relations. It follows that, if the position of initial subordinates is sensitive to semantics in OF, then initial se clauses should occupy SpecFocus more frequently than any type of initial temporal subordinate.

The Variable Position of Initial Subordinate Clauses …

187

As with initial quant, concession has also been evoked: Skårup (1975: 273, 459) and Grad (1956: 77) both note that se subordinates in SpecFocus sometimes carry a concessive reading, as in (22). (22)

Dist Amböyns: ‘[Se Dex l’avoit plevi,] [l ’ OCIRAI je], ainz que il part de ci.’ “Amboyn said: ‘Even if God protected his life, I will kill him before he leaves from here.’” (Gaydon, 6177–6178; in Skårup 1975: 273)

Initial se “if” clauses are attested in every text of the corpus. Like quant, they are very frequent, with a total of 2012 tokens. Again like quant, initial se occurs overwhelmingly in the left periphery, with only five unambiguous cases in SpecFocus. Table 6 presents data from only those texts that evince variation. As was the case with quant, the SpecFocus tokens in this corpus are limited to verse texts. The acute imbalance between left-peripheral and SpecFocus examples casts serious doubt on whether OF productively encoded a semantic distinction via the syntactic position of initial se clauses. Although Skårup (1975: 273) and Grad (1956: 77) remarked that se clauses in SpecFocus could express a concessive value, only two of the five SpecFocus examples in the present corpus are concessive (in Enéas and Tristan); the remaining three are not. Furthermore, Buridant (2000: 655), Ménard (1988: 235, 242, 344), and Skårup (1975: 460–461) present left-peripheral examples of concessive se. Taken together, these observations suggest that OF did not systematically mark concession in initial subordinates via syntactic position. Another argument against a semantic influence on the syntactic position of initial se subordinates draws on the ISRH. Imbs (1956) and Combettes (2010) contend that initial ainz que subordinates often occupy SpecFocus because they possess a particularly tight semantic link with the declarative, as discussed subsequently. Combettes (2010) cites the ISRH, on which anteriority is considered a (marginally) tighter relation than temporal succession. However, by this logic, initial se, which expresses condition, should be more likely than ainz que to occupy SpecFocus (condition is higher than anteriority on the ISRH, although both are rather low). However, in OF, the reverse is true: initial se subordinates almost never occupy SpecFocus, whereas ainz que and synonymous expressions frequently do.

Table 6 Texts with initial se clause attested in SpecFocus

Text, date

Clausal position Left periphery

Thèbes 1150 114 Enéas 1160 126 Becket 1174 97 Tristan 1180 82 Marie 1180 54 a Unambiguous cases

Totala SpecFocus 1 1 1 1 1

115 127 98 83 55

188

6.3

B. Donaldson

Initial Subordinates of Anteriority (ainz que, ainçois que, avant que, devant ce que)

This section examines four initial subordinates that express anteriority (“before…”): ainz que, ainçois que, avant que, and devant ce que. These expressions are synonymous in expressing anteriority (Buridant 2000: 610; Imbs 1956: 479–498). However, a diachronic evolution is attested throughout the OF period, with ainz que and ainçois que gradually being replaced by devant ce que and then avant que (Imbs 1956: 490). If the syntactic position of initial subordinates is semantically conditioned, these expressions of anteriority should pattern similarly. Researchers have observed that initial ainz que, ainçois que, and avant que frequently occupy SpecFocus. Grad (1956: 82; 1961: 11–12), and Combettes (2010: 121) found avant que to systematically occupy SpecFocus in the Roman de Tristan en prose (ca. 1235). Nissen (1943: 57–62), examining the 14th-century Chronique de Jean d’Outremeuse, reported that ainçois que appears both in SpecFocus and in the left periphery, and although Skårup (1975: 459) reported that ainz que frequently occupies SpecFocus, he also found left-peripheral examples. Globally, however, of all the OF initial subordinate types, ainz que and ainçois que are the most likely to occupy SpecFocus. Previous accounts of this behavior cite semantic properties, identifying the expression of anteriority as a particularly tight semantic link between the subordinate and the declarative. Imbs (1956: 504) speaks of “semantic continuity” between initial ainz que and the declarative, arguing that ainz que clauses that occur in SpecFocus are a semantically integral part of the predicate. Combettes (2010) concurs that anteriority is responsible for the syntactic position (in SpecFocus) of initial subordinates that express anteriority. Citing the ISRH, Combettes (2010: 117–122) observes that, among temporal relations of sequentiality, anteriority outranks posteriority, thereby explaining the occurrence of ainz que and its synonyms in SpecFocus. Furthermore, Combettes (2010: 122) argues that initial subordinates of anteriority introduce new information to the discourse and are thus distinct from other types of initial subordinates and other left-peripheral material, whose content is already familiar from previous discourse. Others have questioned the status of anteriority, however. Grad (1961: 12) rejected Imbs’ (1956) analysis, claiming that initial quant “when” subordinates enjoy just as tight a semantic relation with the main declarative, despite their typical position outside SpecFocus. A further difficulty arises with devant ce que, which expresses anteriority but typically occurs outside SpecFocus. Imbs (1956: 505) acknowledges this fact but—with little principled reasoning to support it—attributes the (left-peripheral) syntactic position of devant ce que to the presence of ce “this” as part of the subordinating conjunction. Each example of ainz que, ainçois que, avant que, and devant ce que in the present data clearly conveys anteriority. Of the four types, ainz que is the most frequent, occurring in 21 texts, for a total of 77 occurrences whose syntactic position can be determined with certainty (Table 7).

The Variable Position of Initial Subordinate Clauses … Table 7 Ainz que

Text, date Alexis 1040 Roland 1100 Brendan 1121 C. Looïs 1130 Guillaume < 1150 Thèbes 1150 Charroi 1150 Enéas 1160 Erec 1165 Renart 1170 Becket 1174 Tristan 1180 Marie 1180 Aliscans 1180 Perceval 1185 Alexandre 1188 Orange 1190 Nicolas 1200 Villehardouin Bel inconnu 1214 L’âtre pér. 1250 Total

189 Clausal position Left periphery

SpecFocus

Total

1 3 2 0 0 1 0 0 1 1 1 2 0 4 1 4 1 0 0 0 1 23

1 1 0 3 4 5 2 7 0 2 0 0 3 13 3 2 3 1 1 3 0 54

2 4 2 3 4 6 2 7 1 3 1 2 3 17 4 6 4 1 1 3 1 77

Globally, the data confirm previous observations, revealing a moderately strong preference toward SpecFocus. Nonetheless, the numerous left-peripheral tokens establish that the distribution is not categorical. Furthermore, there is variation across texts: whereas some place ainz que only in SpecFocus (e.g., Enéas, Guillaume), others evince both positions (e.g., Thèbes, Aliscans), and others contain only left-peripheral examples (e.g., Tristan, Erec). Initial ainçois que is attested less frequently, in only seven texts, but shows similar patterns, with a preference for SpecFocus variants (Table 8). Avant que, which eventually emerges over time as the dominant subordinator to mark anteriority (Imbs 1956), is attested—unsurprisingly—only in two late texts in the corpus (Table 9). Cassidorus evinces a strong preference for avant que in SpecFocus, but in Joinville, the picture is less clear, with an even split between left-peripheral and SpecFocus variants. Low token counts aside, the data for initial devant ce que (Table 10) reveal a very different behavior, in the form of a strong preference for left-peripheral variants (see Imbs 1956: 504). These data suggest that anteriority does not unilaterally influence the syntactic position of initial subordinates; none of these four initial subordinates are found

190 Table 8 Ainçois que

Table 9 Avant que

B. Donaldson Text, date

Clausal position Left periphery SpecFocus

Total

Thèbes 1150 Enéas 1160 Becket 1174 Villehardouin 1208 Bel inconnu 1214 L’âtre pér. 1250 Feuillée 1276 Total

1 1 0 0 2 0 0 4

1 3 1 2 2 1 1 11

Text, date Cassidorus 1267 Joinville 1308 Total

Table 10 Devant ce que

0 2 1 2 0 1 1 7

Clausal position Left periphery

SpecFocus

Total

1 8 9

5 8 13

6 16 22

Text, date

Clausal position Left periphery SpecFocus

Total

Alexandre 1188 Clari 1205 Villehardouin 1208 Queste 1225 Total

1 2 2 0 5

1 2 2 1 6

0 0 0 1 1

uniformly and categorically—or even near-categorically—in SpecFocus. Note that OF possesses other elements, such as adverbial ainz or the sentence adverb si, that do show a strong, even categorical, affinity for SpecFocus (e.g., Foulet 1928: 309; Marchello-Nizia 1985). Although ainz que, ainçois que, and, to a lesser degree, avant que, frequently occupy SpecFocus, the left-peripheral tokens are too numerous to be exceptions, especially as certain texts contain only left-peripheral cases. Next, there is the issue of devant ce que, which expresses anteriority but most often occupies the left periphery.2 Although devant ce que and ainz que share the semantic value of anteriority, as subordinating conjunctions, they are formed from

2

Imbs (1956) speculated that the behavior of devant ce que was due to the presence of ce, and Combettes (p.c.) suggested, in this vein, that contrasting the behavior of synonymous devant que and devant ce que may be informative in determining the eventual role of ce. Unfortunately, the present corpus does not allow such an investigation, as the rare attestations of devant que occur only in non-initial position. Other initial subordinates formed with ce, however, regularly occur in both the left periphery and SpecFocus; a clear example of the latter is por ce que (Sect. 6.4).

The Variable Position of Initial Subordinate Clauses …

191

words (ainz and devant) with markedly different behaviors. On its own, the adverb ainz “rather, on the contrary” typically occupies SpecFocus of a main declarative (Foulet 1928: 309), whereas devant “in front of, before” does not. Villehardouin provides an informative test case, as it contains both initial ainz que and devant ce que. In this text, ainz occurs 33 times, excluding tokens of ainz que and the fixed expression qui ainz ainz “in emulation, more than the other.” In all 33 instances, ainz occupies SpecFocus. By contrast, devant occurs 105 times (outside of devant ce que): 10 tokens are temporal (“before”), and 95 are spatial (“in front of”). Crucially, in this text, devant never occupies SpecFocus. The behavior that ainz and devant display as independent lexemes is partially mirrored in the behavior of ainz que, which displays a tendency to occupy SpecFocus, and devant ce que, which tends not to occur in SpecFocus. Nissen (1943: 63–64) and Grad (1961: 13) suggest that ainz exerts analogical force on ainz que, attracting it to SpecFocus (also Foulet, 1928: 312). If this reasoning is correct, semantics would have no effect on the position of these initial subordinates. Although the ISRH, which places anteriority above other sequential temporal relations, provides a plausible explanation for the behavior of most initial subordinates of anteriority, the behavior of initial quant subordinates introduces a complication to the hypothesis of semantic influence. As shown in Sect. 6.1, initial quant clauses exhibit no variation in syntactic position according to gradations of the ISRH. First, expressions of simultaneity and sequentiality are both overwhelmingly left-peripheral. Second, following the ISRH, quant clauses that express simultaneity represent a tighter semantic link with the declarative than anteriority. This relationship does not map straightforwardly onto syntactic position, however, as initial quant clauses (regardless of their semantic value) are overwhelmingly left-peripheral, with only extremely rare examples in SpecFocus, in contrast to ainz que, ainçois que, and avant que. Finally, attestations of après ce que “after…” in SpecFocus in Cassidorus and Joinville cast further doubt on the status of anteriority. Combettes (2010: 122) points to the role of discourse structure, noting that subordinates of anteriority typically present discourse-new information, which he considers compatible with the information structural properties of SpecFocus (see Steiner 2014 for a detailed analysis of the information structure of preverbal XPs). A full investigation of the information structural properties of initial temporal subordinates exceeds the scope of this paper, but a preliminary investigation was conducted on two texts: Aliscans, because it attests initial ainz que in both the left periphery and SpecFocus (Table 7), and Joinville, because examples of initial avant que are evenly split between the left periphery and SpecFocus (Table 9). The results do not reveal a clear effect of information structure. In Aliscans, the four left-peripheral tokens of ainz que all present discourse-new information; of the SpecFocus examples, nine contain discourse-new information, but the remaining four contain discourse-old information. In Joinville, only five of the 16 initial avant que subordinates contain discourse-new information, differing from Combettes’ (2010) findings (for the Roman de Tristan en prose). These five examples are divided between the left-periphery (n = 3) and SpecFocus (n = 2). It remains for

192

B. Donaldson

future work to determine whether other texts show a more systematic relationship between the position of initial subordinates and information structure.

6.4

Initial por ce que

Initial por ce que is polysemous in OF, capable of conveying at least three different semantic values. First, por ce que can express purpose (Buridant 2000: 624; Einhorn 1974: 100), where it is equivalent to English so that or modern French pour que or afin que, as in (23). (23)

Et [por ce que l’en poïst conoistre les uns des autres] [FIST] Elyezer les suens covrir de covertures blanches. “And so that one could tell them apart, Elyezer had his [knights] covered with white blankets.” (Queste, §182)

Second, por ce que can express causativity (Buridant 2000: 614; Einhorn 1974: 100), equivalent to English because or modern French parce que, as in (24). (24)

[Por ce que cil pardons fut issi granz,] [si] [s’en ESMURENT] mult li cuer des genz. “Because the indulgence was so generous, the hearts of the people were (Villehardouin, §2)greatly moved.”

Finally, por ce que can provide an explanation or justification of an utterance, without expressing true cause (Combettes 2010: 119), akin to English since or modern French puisque, as in (25). (25)

[‘Pour ce,’ dist li enfes, ‘que vous avez aucunes choses oïes de ma mere, comment elle se parti de ceste terre et l’en couvint aler en estrange paÿs,] [il AVINT,] aprés ce que partie se fu de cest empire, que nouveles li furent dites d’un saint hermite... “The child said, ‘Since you have heard various things said about my mother, how she left this land and how it was necessary for her to goto a foreign country, it happened, after leaving this kingdom, that she was given news about a holy hermit...’” (Cassidorus, §391)

Previous research has observed that initial por ce que subordinates often occupy SpecFocus. Nissen (1943: 56) found this situation to be categorical in the

The Variable Position of Initial Subordinate Clauses …

193

Chronique de Jean d’Outremeuse. Vance et al. (2010: 309) reported a similar, albeit less categorical, behavior for por ce que in their 13th- and early 14th-century prose data. Neither of these studies examined the semantics of por ce que. On the other hand, in the 13th-century Roman de Tristan en prose, Combettes (2010: 119– 120) found that causative por ce que, exemplified in (24), occupied SpecFocus, whereas instances of por ce que that provided a justification or explanation of an utterance (without expressing cause), as in (25), occurred left-peripherally. Por ce que clauses offer a further opportunity to test if, and how, OF distinguishes degrees of semantic cohesion via distinctions in the syntactic position of initial subordinates. Causality occupies the highest position on the ISRH, whereas purpose is several positions lower.3 Both these relations, however, are considerably higher in the hierarchy than all the other types of semantic relations examined in previous sections (conditionality, all varieties of temporality). In particular, the ISRH predicts that initial causative or purposive por ce que should occur in SpecFocus more frequently than any other type of initial subordinate investigated, including ainz que. Initial por ce que occurs in 20 of the texts, with a total of 155 tokens whose clausal position can be determined with certainty. Without considering semantics, occurrences are evenly divided between the left periphery and SpecFocus, as Table 11 shows.

6.4.1

Purposive Por ce que

When por ce que expresses purpose, its verb appears in the subjunctive (Einhorn 1974: 100; Ménard 1988: 152), as with modern French pour que. This diagnostic revealed 14 purposive examples (Table 12); the purposive interpretation was always verified by the context. The data reveal a slight preference toward SpecFocus variants, but left-peripheral variants are nearly as well attested. Most of the texts evince only one or the other of the variants, but in Joinville, purposive initial por ce que subordinates appear in both positions, witness (26), where por ce que is left-peripheral, and (27), where it occupies SpecFocus. (26)

Et [pource que l’en ne le veist,] [il] [se MIST] parmi les mirales de la cité de Rames. “And so that he would not be seen, he slipped inside the walls of the city of Ramla.” (Joinville, §542)

3

I do not attempt to position instances of por ce que that provide justification/explanation with respect to the ISRH, but will nonetheless examine whether the semantics of por ce que corresponds to differences in clausal position.

194 Table 11 Overall distribution of initial por ce que

Table 12 Purposive por ce que

B. Donaldson Text, date

Clausal position LP SpecFocus

Total

Brendan 1121 Thèbes 1150 Enéas 1160 Erec 1165 Renart 1170 Becket 1174 Tristan 1180 Marie 1180 Aliscans 1180 Perceval 1185 Alexandre 1188 Graal Boron 1200 Clari 1205 Villehard. 1208 Bel inconnu 1214 Queste 1225 L’âtre pér. 1250 Cassidorus 1267 Feuillée 1276 Joinville Total

1 1 1 2 1 5 0 4 1 3 3 3 0 5 2 4 1 25 1 16 77

1 1 1 2 2 10 1 5 2 7 5 4 2 6 5 30 5 41 1 24 155

Text, date Thèbes 1150 Becket 1174 Perceval 1185 Queste 1225 L’âtre pér. 1250 Cassidorus 1267 Joinville 1308 Total

0 0 0 0 1 5 1 1 1 4 2 1 2 1 3 26 4 16 0 8 78

Clausal position Left periphery

SpecFocus

Total

1 1 0 0 0 1 3 6

0 0 1 4 1 0 2 8

1 1 1 4 1 1 5 14

The Variable Position of Initial Subordinate Clauses …

(27)

195

[Pource que vous puissiés veoir que c’estoit le plus loiaus homme qui onques feust en son temps,] [vous VUEIL-je] dire que il fu si loiaus...” “So that you might see that he was the most honorable man who ever lived in his time, I want to tell you that he was so honorable...” (Joinville, §764)

6.4.2

Causative por ce que

Instances of causative por ce que were identified following established criteria and observations about causation. Thus, whereas the contents of a causal clause can be focused (Dancygier and Sweetser 2000: 136; Grevisse and Goosse 2008: 1489), a justification or explanation cannot. Likewise, a causal can respond to the question “why did X transpire?” (Grevisse and Goosse 2008: 1489), whereas a justification or explanation cannot. Finally, a causal communicates the “crucial contributing factor” (Neeleman and van de Koot 2012: 23). These criteria differentiate the causal example in (24) from that in (25), in which the por ce que clause provides an explanation about why the main declarative was uttered. Examples in which a causal relation could not be clearly established, and which were not purposive, were classified as justification/ explanation. Table 13 presents the data for causal por ce que.

Table 13 Causal initial por ce que

Text, date Enéas 1160 Erec 1165 Renart 1170 Becket 1174 Tristan 1180 Marie 1180 Aliscans 1180 Perceval 1185 Alexandre 1188 Graal Boron 1200 Clari 1205 Villehard. 1208 Bel inconnu 1214 Queste 1225 L’âtre pér. 1250 Cassidorus 1267

Clausal position Left periphery 1 2 1 3 0 4 0 2 3 2 0 5 2 2 1 8

Total SpecFocus 0 0 0 5 1 1 1 3 2 1 2 1 3 17 2 13

1 2 1 8 1 5 1 5 5 3 2 6 5 19 3 21 (continued)

196

B. Donaldson

Table 13 (continued)

Text, date Feuillée 1276 Joinville Total

Clausal position Left periphery

SpecFocus

Total

1 11 48

0 4 56

1 15 104

The data reveal only a slight propensity to place initial causative por ce que in SpecFocus. Predictions from the ISRH about the behavior of por ce que are not borne out. First, purposive por ce que clauses occupy SpecFocus more frequently (64.3%) than causal por ce que clauses (53.8%), whereas the ISRH would predict the opposite, given that causation represents a tighter semantic link than purpose. Next, consider the behavior of causal subordinates versus subordinates of anteriority, recalling from the ISRH that causality is a significantly tighter semantic link than anteriority. By this logic, causative por ce que should evince the highest rates of occurrence in SpecFocus of any type of initial subordinate. What the data reveal, however, is that initial ainz que (anteriority) occurs in SpecFocus at a rate of 70.1%, whereas the figure for causative por ce que is considerably lower (53.8%). A final observation underscores the inapplicability of the ISRH in determining the clausal position of initial subordinates: The majority of the texts evince variation in their treatment of causative por ce que, evincing both left-peripheral and SpecFocus variants. Consider the following causative examples from Becket: in (28), the initial subordinate is left-peripheral; in (29), it occupies SpecFocus. (28)

[Pur ço qu’iert desarmez] [tut premiers] [le SIWI]. “Because he was unarmed, he was the first to follow him.” (Becket, 5601)

(29)

Car [pur ceo qu’il ert clers] [l’AVEIËNT] deporté. “For, because he was a cleric, they treated him gently.” (Becket, 818)

The lais of Marie de France present similar variation with initial causative por ce que; in (30), the subordinate is left-peripheral, and in (31), it occupies SpecFocus. (30)

Vielz huem e anciëns esteit.

“He was an elderly man

Charnel heir madle nen aveit;

who had no male heir .

une fille ot a mariër.

He had a daughter to marry off.

[Pur ceo qu’il ne la volt doner

Because he did not want to give her

a sun per,] [cil] [le GUERREIOT ].

to his neighbor, the latter warred against him.”

(Marie de France, Lanval, 93–97)

The Variable Position of Initial Subordinate Clauses …

(31)

[Pur ceo que pruz fu e curteis

Because he was valiant and noble

e que mult le preisot li reis,]

and because the king esteemed him greatly,

[li OTRIA] sa druërie.

she granted him her love.”

197

(Marie de France, Les deux amants , 57–69)

In Joinville, most instances of causal por ce que are left-peripheral, as in (32). (32)

Et [pource que il parloient haut en ma chapelle et que il faisoient noise au prestre,] [je] [leur ALAI] dire que il se teussent et leur dis que vileinne chose estoit de chevaliers et de gentils homes qui parloient tandis que l’en chantoit la messe. “And because they were speaking loudly in my chapel and because they were quarreling with the priest, I went to tell them to be quiet and I told them that it was a shameful thing to see knights and noblemen speaking while the mass was being sung.” (Joinville, §297)

In sum, data from initial por ce que clauses do not reveal an influence of semantics on the syntactic position of the subordinate.

7 General Discussion and Conclusion Results from the initial subordinate types examined here across a large selection of texts do not support the view that the semantic value of an initial subordinate —at least as measured by the ISRH—influences its clausal position (left periphery vs. SpecFocus) in OF. In the case of initial quant, the distribution between left-peripheral and SpecFocus variants does not pattern according to semantic values, as neither concession nor the distinction between simultaneous and sequential temporal events corresponds to differences in the clausal position of initial quant. For initial se, there is no uniform correspondence between concession and syntactic position. For initial subordinates of anteriority, the propensity to occupy SpecFocus is not as strong as has been supposed previously (c.f., Combettes 2010; Imbs 1956). First, the distribution of ainz que, ainçois que, and avant que between SpecFocus and the left periphery is far from systematic, and second, initial subordinates of anteriority, as a class, do not present a unified syntactic behavior (e.g., ainz que vs. devant ce que). Finally, the position of initial por ce que does not pattern according to semantics, either. Although it is true that por ce que and ainz que, ainçois que, and avant que occur in SpecFocus with some frequency, this is not the case for quant and se. For these two types, the proportions between left-peripheral and SpecFocus variants are so skewed that it is unlikely, on purely statistical grounds, that the OF grammar productively

198

B. Donaldson

signaled any semantic distinction via the clausal position of the initial subordinate. Further arguments against the role of semantics come from the behavior of initial subordinates vis-à-vis the ISRH. Individually, none of the initial subordinate types evinces the type of behavior predicted by the ISRH with respect to clausal position. These data suggest that OF does not differentiate, via syntactic position, between initial quant subordinates that mark simultaneity versus those that signal succession. However, this is precisely the type of distinction that Combettes (2010) cites in distinguishing between subordinates that mark anteriority versus those that encode posteriority and other temporal relations. It seems implausible that OF would target a narrow subset of the ISRH and distinguish, via clausal position, different types of sequentiality (anteriority vs. posteriority) but then fail to make an analogous distinction between simultaneity and sequentiality. When viewed collectively with respect to the ISRH, the types of initial subordinates examined do not pattern as anticipated. If the distinction between SpecFocus and left-peripheral variants followed straightforwardly from the tightness of the semantic link between the subordinate and the declarative, the ISRH would predict causative por ce que to manifest the highest proportions of SpecFocus variants, followed by purposive por ce que, conditional se, quant indicating simultaneity, and finally, anteriority and then succession. As the results in Sect. 6 revealed, this prediction is not upheld. A further argument against signaling semantic nuances via the syntactic position of initial subordinates comes from other V2 languages. In modern German, for example, all initial subordinates appear in the first position of the declarative, regardless of their semantic value (Salvesen 2013). That is, the position of initial subordinates is determined by purely syntactic constraints, and no variation is possible. Among the medieval Romance varieties, all of which passed through a V2 stage (Benincà 2006), Old Occitan only rarely places initial subordinates in SpecFocus, and the left periphery appears to be the default position (Vance et al. 2010: 316). Nevertheless, further research is needed on other medieval Romance varieties and on other V2 languages. If the semantic value of the initial subordinate does not explain differences in its clausal position, what does? Combettes (2010: 122) mentions information structure, in particular distinctions between discourse-new and discourse-old information. A preliminary examination of the information structure of initial ainz que subordinates did not reveal a correspondence between discourse-new or –old information and syntactic position, although a more extensive study is necessary before drawing strong conclusions. Nonetheless, any explanation based on information structure would need to account for the fact that variation in clausal position is largely confined to certain types of subordinates, and that se and quant, for example, are largely invariable, occurring in the left periphery almost without exception. Textual genre may also be important. In the case of initial se and quant, all the unambiguous SpecFocus examples occur in verse texts, whose syntactic behavior sometimes

The Variable Position of Initial Subordinate Clauses …

199

differs from that of prose (e.g., Hirschbühler and Junker 1988). Similarly, van Reenen and Schøsler (1992) noted differences in topic marking with the sentence adverb si in prose versus verse texts. Metrical concerns may partially explain variation in clausal position. Grammatical instability and change also characterize the OF period, and although initial subordinates placed in SpecFocus are a minority variant in OF, their presence demonstrates a certain flexibility in the OF grammar with respect to the position of initial subordinates. The timing of SpecFocus variants is noteworthy. The data show that most instances of initial subordinates in SpecFocus occur between 1160 and 1200, a period during which the V2 grammar of OF underwent changes. After this period, OF no longer allowed SpecFocus to remain empty in declaratives, and verb-initial declaratives disappeared (Simonenko and Hirschbühler 2012), as did matrix postposed object and adverbial clitics. One hypothesis is that some types of initial subordinates are temporarily drawn in SpecFocus as a means of fulfilling the emerging syntactic requirement that this position be filled. Similarly, Vance et al. (2010) tie another aspect of the behavior of OF initial subordinates to syntactic change, in particular the gradual loss of V2 in favor of SVO over the course of the 13th century. Future research might uncover a level of semantic or discourse-pragmatic meaning that consistently explains the position of initial subordinates. Alternately, if the variability stems from grammatical instability or change, it may be largely devoid of meaning—semantic, discursive, or otherwise—at a given point in time.

References Primary Sources Bayot, Alphonse (ed.). 1914. Gormont et Isembart: Fragment de chanson de geste du XIIe siècle. Paris: Honoré Champion. Cerquiglini, Bernard (ed.). 1981. Le roman du graal. Paris: Union Générale. Corbett, Noel L. (ed.). 1977. La vie de Saint Louis by Jehan de Joinville. Sherbrooke: Naaman. De Poerck, Guy, R. van Deyck, and R. Zwaenepoel. ed. 1970. Le charroi de Nîmes: Chanson de geste. Saint-Aquilin-de-Pacy: Mallier. Dufournet, Jean (ed.). 1984. Aucassin et Nicolette. Paris: Flammarion. Dufournet, Jean (ed.). 2005. Le jeu de saint Nicolas. Paris: Flammarion. Foster, Brian, and Ian Short. ed. 1976–1977. The Anglo-Norman Alexander: Le roman de toute chevalerie. London: Anglo-Norman Text Society. Fritz, Jean-Marie, and Michel Zink (eds.). 1994. Erec et Enide. Paris: Librairie Générale Française. Harf-Lancner, Laurence (ed.). 1990. Lais de Marie de France. Paris: Librairie Générale Française. Katz, Blanche (ed.). 1947. La prise d’Orange according to MS. A1. Morningside Heights, NY: King’s Crown. Koschwitz, Eduard (ed.). 1964. La vie de Saint Léger (Les plus anciens monuments de la langue française, 6th ed. Munich: Max Hueber Verlag. Langlois, Ernest (ed.). 1966. Le jeu de la feuillée. Paris: Honoré Champion.

200

B. Donaldson

Langlois, Ernest (ed.). 1968. Le couronnement de Louis, chanson de geste du XIIe siècle. Paris: Honoré Champion. Lauer, Philippe (ed.). 1956. La conquête de Constantinople. Paris: Honoré Champion. Lodge, R.Anthony, and Kenneth Varty (eds.). 2001. The earliest branches of the Roman de Renart. Leuven: Peters. Marchello-Nizia, Christiane. ed. 2011. La queste del saint graal. http://portal.textometrie.org/txm/. Accessed 15 June 2016. Méla, Charles, and Michel Zink (eds.). 1994. Le conte du graal ou le roman de Perceval. Paris: Librairie Générale Française. Muret, Ernest, and L.M. Defourques (eds.). 1974. Le roman de Tristan: Poème du XIIe siècle. Paris: Honoré Champion. Palermo, Joseph (ed.). 1963. Le roman de Cassidorus. Paris: Société des anciens textes français. Perrie Williams, Gwladys. ed. 1967. Le bel inconnu: Roman d’aventures. Paris: Honoré Champion. Perugi, Maurizio (ed.). 2000. La vie de Saint Alexis. Geneva: Droz. Petit, Aimé (ed.). 2008. Le roman de Thèbes. Paris: Honoré Champion. Régnier, Claude (ed.). 2007. Aliscans. Paris: Honoré Champion. Salverda de Grave, Jean-Jacques. (ed.). 1964. Enéas: Roman du XIIe siècle. Paris: Honoré Champion. Segre, Cesare (ed.). 1971. La chanson de Roland. Milan: Riccardo Ricciardi. Short, Ian, and Brian S. Merrilees (eds.). 1979. The Anglo-Norman voyage of St Brendan. Manchester: Manchester University Press. Suard, François (ed.). 2008. La chanson de Guillaume. Paris: Librairie Générale Française. de Wailly, Natalis (ed.). 1874. La conquête de Constantinople. Paris: Firmin Didot. Walberg, Emmanuel (ed.). 1964. La vie de Saint Thomas Becket. Paris: Honoré Champion. Woledge, Brian (ed.). 1936. L’âtre périlleux. Paris: Honoré Champion.

Secondary Sources Adams, Marianne. 1987. From Old French to the theory of pro-drop. Natural Language and Linguistic Theory 5: 1–32. Benincà, Paola. 2006. A detailed map of the left periphery of medieval Romance. In Crosslinguistic research in syntax and semantics: Negation, tense, and clausal architecture, ed. Raffaella Zanuttini, Héctor Campos, Elena Herburger, and Paul Portner, 53–86. Washington, DC: Georgetown University Press. Buridant, Claude. 2000. Grammaire nouvelle de l’ancien français. Paris: Sedes. Combettes, Bernard. 2010. Aspects diachroniques de la parataxe: Les propositions temporelles en position initiale en ancien français. In La parataxe: entre dépendance et intégration, vol. 1, ed. Marie-José Béguelin, Mathieu Avanzi, and Gilles Corminbœuf, 115–137. Bern: Peter Lang. Dancygier, Barbara, and Eve Sweetser. 2000. Constructions with if, since and because: Causality, epistemic stance, and clause order. In Cause, condition, concession, contrast, ed. Elizabeth Couper-Kuhlen and Bernd Kortmann, 111–142. Berlin: Mouton de Gruyter. Donaldson, Bryan. 2012. Initial subordinate clauses in Old French: Syntactic variation and the clausal left periphery. Lingua 122: 1021–1046. Donaldson, Bryan. 2015. Discourse functions of subject left dislocation in Old Occitan. Journal of Historical Pragmatics 16: 159–186. Einhord, E. 1974. Old French: A concise handbook. Cambridge: Cambridge University Press. Foulet, Lucien. 1928. Petite syntaxe de l’ancien français. Paris: Honoré Champion. Grad, Anton. 1956. L’inversion du sujet dans la principale précédée d’une subordonnée en ancien français. Razprave SAZU 2: 65–90. Grad, Anton. 1961. Remarques sur l’emploi de l’adverbe de reprise si en ancien français. Linguistica 4: 5–20.

The Variable Position of Initial Subordinate Clauses …

201

Grevisse, Maurice, and André Goosse. 2008. Le bon usage. Brussels: De Boeck. Hirschbühler, Paul, and Marie-Odile Junker. 1988. Remarques sur les sujets nuls en subordonnées en ancien et en moyen français. Revue québécoise de linguistique théorique et appliquée 7: 63–84. Holmberg, Anders. 2015. Verb second. In Syntax—theory and analysis: An international handbook, vol. 1, ed. Tibor Kiss and Artemis, 342–383. Berlin: De Gruyter. Imbs, Paul. 1956. Les propositions temporelles en ancien français: La détermination du moment. Paris: Les Belles Lettres. Ingham, Richard. 2014. Old French negation, the Tobler/Mussafia law, and V2. Lingua 147: 25–39. Marchello-Nizia, Christiane. 1985. Dire le vrai: l’adverbe si en français médiéval – essai de linguistique historique. Geneva: Droz. Marchello-Nizia, Christiane. 1998. Dislocations en ancien français: Thématisation ou rhématisation? Cahiers de Praxématique 30: 161–178. Mathieu, Eric. 2012. The left periphery in Old French. In Research on Old French: The state of the art, ed. Deborah Arteaga, 327–350. Dordrecht: Springer. Ménard, Philippe. 1988. Syntaxe de l’ancien français. Pompignac: Editions Bière. Neeleman, Ad, and Hans van de Koot. 2012. The linguistic expression of causation. In The theta system: Argument structure at the interface, ed. Martin Everaert, Marijana Marelj, and Tal Siloni, 20–51. Oxford: Oxford University Press. Nissen, Harald. 1943. L’ordre des mots dans la chronique de Jean d’Outremeuse. Uppsala: Almqvist & Wiksells. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer. Roberts, Ian. 1993. Verbs and diachronic syntax: A comparative history of English and French. Dordrecht: Kluwer. Salvesen, Christine M. 2013. Topics and the left periphery: A comparison of Old French and Modern Germanic. In In search of Universal Grammar: From Old Norse to Zoque, ed. Terje Lohndal, 131–172. Amsterdam: John Benjamins. Simonenko, Alexandra, and Paul Hirschbühler. 2012. Placement de clitiques dans les propositions V1 et évolution de la structure de la proposition en ancien français. In Typologie, ordre des mots et groupe verbal en français médiéval, ed. Monique Dufresne, 11–53. Laval: Presses de L’Université Laval. Skårup, Povl. 1975. Les premières zones de la proposition en ancien français. Revue Romane [special issue #6]. Steiner, B. Devan. 2014. The evolution of information structure and verb second in the history of French. Dissertation, Indiana University. van Reenen, Pieter, and Lene Schøsler. 1992. Ancien et moyen français: Si “thématique”—analyse exhaustive d’une série de textes. Vox Romanica 51: 101–127. Valin, Van, D. Robert, and Randy J. LaPolla. 1997. Syntax: Structure, meaning, and function. Cambridge: Cambridge University Press. Vance, Barbara. 1997. Syntactic change in medieval French: Verb-second and null-subjects. Dordrecht: Kluwer. Vance, Barbara, Bryan Donaldson, and B. Devan Steiner. 2010. V2 loss in Old French and Old Occitan: The role of fronted clauses. In Romance linguistics 2009: Selected papers from the 39th linguistic symposium on Romance languages, ed. Sonia Colina, Antxon Olarrea, and Ana Maria Carvalho, 301–320. Amsterdam: John Benjamins.

Obviation and Old French Subjunctive Clauses Deborah L. Arteaga

Abstract Obviation, or the fact that the subject of the subjunctival subordinate clause cannot be coreferential with the subject of the main clause, as seen by the French example, *je veux que je parte ‘I want that I leave’ (cf. Ruwet 1984), is a topic that has been discussed widely in the literature, especially for the Romance languages, including Catalan (Picallo 1985), French (Jackubowicz 1985), Old French (Martineau 1994; Arteaga 2015), Italian (Costantini 2005, 2016), Portuguese (Raposo 1987), Romanian (Farkas 1984), and Spanish (Kempchinsky 2009). This chapter considers the lack of obviation in Old French. After reviewing analyses proposing that obviation effects are related to tense construal in the subordinate clause, it is proposed that the analyses in San Martín (2007) and Laskova (2017) can be adapted to account for the lack of obviation effects in Old French and Romanian. Our analysis is an extension of Arteaga (2015), in that it is valid cross-linguistically and does not rely on the morphological development from Old French to Modern French. Keywords Old French Bulgarian


Obviation
Sequence-of-tense
Binding Theory


1 Introduction Obviation—or obligatory disjoint reference—holds where the subjects of the main and subordinate clauses cannot be coreferential. One of the first modern scholars to discuss this in Modern French (MF) was Ruwet (1984). Consider (1) and (2) below:

D. L. Arteaga (&) Department of World Languages and Cultures, University of Nevada, Las Vegas, Las Vegas, NV, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_10

203

204

D. L. Arteaga

(1) * Je veux que je sorte I want-1sg that I go-out-1sg-SUBJ ‘I want to go out.’

(2) Je veux sortir I want-1sg to.go.out ‘I want to go out.’ The example in (1) is ungrammatical, because after a verb of volition, which takes the subjunctive, co-reference between the subject of the main clause and that of the subordinate clause is impossible in MF. Instead, an infinitive is used, as illustrated by the grammatical (2). The same obviation restrictions also exist in the other modern Romance languages, with the exception of Romanian. Farkas (1984) notes that the sentence of the subordinate clause can be coreferential with that of the main clause in that language:

să (3) Vreau plec want-1sg SUBJ leave-1sg ‘I want to leave.’ (literally, ‘I want that I leave.’ (cf. (1) above) As can be seen from the above example, the Romanian subjunctive, unlike the subjunctive in the other Modern Romance languages, contains a separate particle introducing the subjunctive complement, namely să.1 In this chapter, we will first present relevant morphosyntactic features in Old French (OF). After reviewing several previous analyses of obviation, we will follow Arteaga (2015), who derives lack of obviation from the feature IRR in MP, as proposed by San Martín (2007). However, we will argue that the indicative and subjunctive complementizers were distinct in OF. We also adopt Laskova’s (2017) analysis of the subjunctive, which captures a broad cross-linguistic snapshot of the subjunctive/indicative moods, in that we will also posit an Exclusion Feature (ExclF) as the temporal anchor of the subjunctive. We will give a preliminary account by formalizing her analysis based on feature valuation.

1

Indicative clauses are introduced by the complementizer ca, which does not indicate mood.

Obviation and Old French Subjunctive Clauses

205

2 Old French Morpho-Syntax 2.1

Null Subject Language

In this section, we will consider several morphosyntactic characteristics of which differ from MF. One, it was a null-subject language2:

(4) Bel avret corps Beautiful-m-sg-OBL had-3sg pro body-m-sg-OBL ‘(She) had a beautiful body.’ (La Cantilène de Ste Eulalie, L2)

(5) Or revendrai now will.return-1sg pro al padre ed a la medre to.the-m-sg-OBL father-m-sg-OBL and to the-f-sg-OBL mother-f-sg-OBL ‘Now (I) will return to the father and to the mother.’ (La vie de St. Alexis, L 101) The subjects, elle ‘she’ in (4) and je ‘I’ in (5) are both unexpressed.

2.2

Case System

The second characteristic relevant to our discussion is that OF had a case system, differentiating the nominative and oblique cases, as demonstrated by the following examples3:

2

The Old French examples are from Arteaga (2015). Feminine nouns did not decline. There was a difference between singular and plural only, such as la rose, ‘the rose,’ les roses, ‘the roses.’

3

206

D. L. Arteaga

(6) Li chevalier de la The-m-pl-NOM knights-m-pl-NOM of the-f-sg-OBL Table Reonde distrent table-f-sg-OBL round-f-SG-OBL said-3pl le jor de Pentecoste m-sg-OBL day-m-sg-OBL of Pentecost-f-sg-OBL ‘The knights of the Round Table said on the day of Pentecost’ (La Queste 25 L 5)

(7) Si ot il assez en la place so had-3sg it enough in the-f-sg-OBL place-f-sg-OBL barons et chevaliers barons-m-pl-OBL and knights-m-pl-OBL ‘There were enough barons and knights in the place.’ (La Queste, 13 L15) In (6), li chevalier is in the nominative plural, while in (7) chevaliers is in the oblique plural.

2.3

Word Order

Third, with respect to word order, it is generally accepted by both philologists (cf. Anglade 1965; Einhorn 1975; Ménard 1988; Moignet 1976; Jensen 1990) and syntacticians (cf. Adams 1987; Arteaga 1994; Arteaga and Herschensohn 2004, 2006; Vance 1997; Roberts 1993; Mathieu 2013; Ingham 2013) that in main clauses, OF had V2 word order, as seen by the following examples:

(8) Toutes ces choses all-f-pl-OBL these-f-pl-OBL things-f-pl-OBL te presta Nostre Sires to you loaned-3sg our-m-sg-NOM lord-m-sg-NOM ‘Our lord loaned all those things to you.’ (La Queste 68.28) (Togeby 1974:58)

Obviation and Old French Subjunctive Clauses

207

(9) Lors vient li rois a Galaad Then comes-3-sg the-m-sg-NOM king-m-sg-NOM to Gilead. ‘Then the King comes to Gilead.’ (La Queste 11, 12) (Togeby 1974:58) In (8), the verb presta is in the second position, the subject being the DP toutes ces choses. In (9), the adverb lors is in the first position, followed by the verb vient. Vance (1997) considers OF to be an “asymmetrical V2 language,” as in subordinate clauses with que, SVO word order is typically found,4

(10) Or voi

je bien que ma

chetivetez

now see-1SG I well that my.FM.SG.NOM m’ a

confondu

capture.OBL.SG.NOM

plus que

me has-3SG confounded.MG.SG.OBL more than nule

autre

chose.

any.FM.SG.NOM other.FM.SG.NOM chose.FM.SG.NOM ‘Now I tell you that my capture has confounded me more than anything else.’ (La Queste 61.30) The example in (10) illustrates the word order typical of OF, both in main and subordinate clauses. The main clause Or voi je bien ‘I see now’ (literally ‘now see I well’) shows V2 word order, while the subordinate clause introduced by que, que ma chetivetez m’a confondu ‘that my capture has confounded me,’ has SVO word order. In this section, we have seen the following morphosyntactic characteristics of, namely that it was a null subject language with a case-system that distinguished between two cases, in addition to having V2 word order in main clauses.

4

This is example is from Arteaga (2009).

208

2.4

D. L. Arteaga

Subjunctive Morphology

As Einhorn (1975) notes, in OF, the subjunctive morphology contrasts in some forms with that of the indicative, as verbal endings were pronounced. In er verbs, the 2nd and 3rd persons singular are differentiated, as in the verb soner ‘to ring’: indicative 2sg/3sg sones/sont vs. subjunctive 2sg/3sg sons/sont. In the case of ir and re verbs, all singular forms were distinct, as illustrated by finir ‘to finish,’ and ‘vendre’ to sell: 1sg/2sg/3sg finis/finis/finist vs finesse/finisses/finisset; 1sg/2sg/3sg vent/venz/vent vs. vendre/vendes/vendet.5 In addition, suppletive verbs, such as avoir ‘to have,’ or être ‘to be,’ also showed different conjugation patterns in the indicative and subjunctive, in all persons, such as 1sg/2sg/3sg (avoir) indicative ai/at/at vs. subjunctive aie/aies/ait and (être) indicative vs. sui/es/est subjunctive seie/seies/seit.6

2.5

Non-expression of que

An additional difference between indicative and subjunctive clauses is that subjunctive clauses in OF can occur without que, often referred to as parataxis (Arteaga 2015: 156). Unexpressed complementizers in OF are quite widespread in subjunctive environments.7 A variety of verb types allowed parataxis, including epistemics, verbs of volition, main clause subjunctives, and contrary-to-fact subjunctives (see inter alia Anglade 1965; Brunot 1966; Einhorn 1975; Foulet 1982; Jensen 1990; Moignet 1976; Nyrop 1930; Togeby 1974)8: Indeed, Glikman and Verjans (2013) note that the expression of que in subjunctive complements freely alternated with its non-expression. Consider (11)–(14): (11) Dex confonde tote lor gente. God-m-sg-NOM destroy-3-sg-SUBJ all-fsg-OBL their-f-sg-OBL people-f-sg-OBL ‘May God destroy all of their people!’ (Brut 13756) (Kowalski 1882:3)

5

See Appendix 1 for the complete paradigms of regular verbs. See Appendix 2 for the conjugation of avoir and être. 7 Costantini (2007) shows that in Italian, complementizer deletion is only found after verbs of believing, like credere ‘to believe.’ These verbs also take the subjunctive in Italian. We will not discuss his analysis due to space limitations. 8 As previously discussed in Arteaga (2009), an unexpressed que is the norm in main clause subjunctives. 6

Obviation and Old French Subjunctive Clauses

209

(12) Car fusse je en Babiloine so were-1sg-SUBJ I in Babylon! ‘If only I were in Babylon-f-sg-OBL!’ (Fabliaux 9.352) (Jensen 1990:357) (13) Cuida [chose fust par li sole] thought-3sg thing-F-SG-OBL done-3sg-SUBJ by her alone ‘He thought (that) she alone did that.’ (although Renart actually did) (Renart 2603) (Jensen 1974:55) (14) Nos volons we want-1pl [son nevo en chast] his-m-sg-OBL nephew-m-sg-OBL from.there banish-3sg- SUBJ ‘We want him to banish his nephew.’ (literally, ‘We want (that) he banish his nephew.’) (Béroul, 602) (Graeme-Ritchie 137)

In the main clause subjunctive in (11), the complementizer que is not present, nor is it in the contrary-to-fact clause in (12). The subordinate clause is not introduced by que in (13), which contains the verb cuidier ‘to believe,’ which expressed a false belief and took the subjunctive. The example in (14) has the verb of volition nous volons ‘we want,’ no complementizer introduces the subordinate clause. Note that in (14), V2 word order is found; son nevo is in the oblique case and therefore is the object.9 In all of these instances, the expression of que in MF is obligatory. These examples indicate that the lack of complementizer can independently mark the subjunctive (cf. Jensen 1990; Foulet 1982).10 In subjunctive clauses where que is not expressed, V2 order is found, such as in (13) and (14) (Arteaga 2015: 156).

9

The nominative form would be ses nies. Giorgi and Pianesi (1997, 2001), and Costantini (2005, 2016), among others, consider the complementizer che, noting that only verbs of believing allow Complementizer Deletion. They note that many other verbs that take the subjunctive in Italian do not allow for Complementizer Deletion, nor do verbs that take indicative complements. Italian CD is beyond the scope of this paper. 10

210

2.6

D. L. Arteaga

Sequence of Tenses in Subjunctive Clauses

It is important to note that the tense in OF subjunctive clauses was not anaphoric to that of the main clause, as seen by the following11:

(15) Ço voelt li reis that wants-3-sg the-m-sg-NOM king-m-sg-NOM par amur cunvertisset out.of love-m-sg-OBL convert-3-sg-SUBJ ‘The king wants him to convert out of love.’ (Roland 3674) (16) Mais lui ert tart qued il s’ en fust alez but to.him was-3-sg-SUBJ late that he himself from.there was gone-m-sg ‘He looked forward to having left’ (literally that he had gone) (La vie de St. Alexis 13)

(17) Lors desira tantost then desired-3sg proi immediately k’ eüst Mahom guerpi1 that hei had-3-sg-SUBJ Mohammed rejected ‘Then hei immediately wished that hei had rejected Mohammed.’ (Buev 2678) (Busse 1905:24) In (15) above, the tense of the matrix clause is the present, but that of the subordinate clause is the imperfect subjunctive. In the example (16), the subjunctive verb is in the pluperfect subjunctive, which is not construed as subsequent to the tense of the main clause, which is in the imperfect. The subordinate clause in (17) is in the pluperfect subjunctive, although the main clause verb is in the passé simple. We next turn to obviation in OF.

Note the lack of obviation between the unexpressed il ‘he’ of the main and subordinate subjects.

11

Obviation and Old French Subjunctive Clauses

211

3 Obviation in Old French In OF, obviation did not hold in volitional complements. Consider the following: (18) Encore ainme je mix que je still like-1-sg I better that I

muire die-1-sg-SUBJ

ci here

que tos li poules me regardast that all-m-sg the-m-sg-NOM public-m-sg-NOM me watched-3sg-SUBJ demain a merveilles tomorrow to marvel-f-pl-OBL ‘I prefer to die here.’ (literally, ‘that I die here’) than to be watched by all of the public tomorrow.’ (Aucassin 16, 14)

(19) N’ a talent que il rie not has-3sg proi desire-m-sg-OBL that hei laughs-3sg-SUBJ ‘He doesn’t want to laugh.’ (La chanson de Antioche. 156,8) (Busse 1905:24) In (18), the pronoun je is used in the subjunctive subordinate clause despite the fact that it is coreferential with je of the main clause. In (19), the pronoun il in the main clause (which is unexpressed) is coreferential with il in the lower clause. However, infinitival constructions after verbs of volition are also found in OF:

(20) Mielz voill murir better will-1sg pro to.die proi que jo ne l’ alge ocire that Ii not him go-1sg-SUBJ to.kill ‘I prefer to die than to not go kill him.’ (La Chanson de Roland 1646) (Quiel 1881:20) The example in (20) shows both options open to OF after a verb of volition with coreferential pronouns in the main and subordinate clauses, namely an infinitive, such as murir after voill, and a clause que jo no alge ocire.

212

D. L. Arteaga

4 Theoretical Framework A relevant theoretical construct is the Binding Theory, which addresses the reference of anaphors, pronouns, and referring expressions. According to Principle A, anaphors, such as reflexive pronouns, must be bound within their governing category. Pronouns, on the other hand, must be free, as per Principle B. Referring expressions, such as proper nouns, must be free (Principle C). This becomes relevant for analyses of obviation, because subordinate clauses typically constitute a separate Binding Domain. Coreference between the matrix and subjunctive clauses should, in theory, be possible. The analysis presented here is within the Minimalist framework. Two important aspects of Minimalism are Merge, External and Internal, and Agree (Chomsky 2001). External merge combines two syntactic elements which combine to form a new syntactic unit, while Internal Merge allows for the movement of one syntactic object within the domain of a second. Agree is the operation by which features, interpretable and uninterpretable, are valued. Uninterpretable features include, for example, case, [ucase], which is valued by the noun. An uninterpretable feature is a probe, searching for its goal, which deletes it, preventing the derivation from crashing. According to Chomsky (2005), Probe-Goal relations must be local, which is accounted for by the Phase Impenetrability Condition, which states the following (Chomsky 2001: 13): (21) If X is dominated by a complement of a phase YP, X cannot move out of YP.

We next turn to theoretical accounts of obviation that have been proposed in the literature.

5 Previous Analyses of Obviation Many scholars have discussed how to account theoretically for obviation within Romance. The general idea has been to extend the Binding Domain so that the pronouns (subject to Principle B) cannot be coreferential. By extending the Binding Domain to the entire clause, co-reference becomes impossible.

5.1

Picallo (1985)

There have been various theoretical mechanisms used to explain the extension of the Binding Domain to the main clause. Some scholars have argued that obviation

Obviation and Old French Subjunctive Clauses

213

is a result of restrictions on the Sequence of Tense (SOT) in subjunctive clauses. For example, Picallo noted that obviation effects hold in Catalan, as illustrated by the following examples (Picallo 1985: 34):

(22) * Jo vull que jo vagi a veure acquesta pellícula I want-1-sg that I go-1-sg-SUBJ to to.see that-f-sg movie-f-sg ‘I want that I go see that movie.’ According to Picallo, the sentence in (22) is ungrammatical because the pronouns jo in the matrix and subordinate clauses are coreferential. Picallo accounts for this by proposing that subjunctive clauses lack tense, and that a Binding Domain must have tense. Under such an analysis, coreference between the matrix and subordinate subjects is ruled out. Jakubowiz (1985) proposes a similar analysis for French. One shortcoming of both analyses is that the lack of obviation in Romanian is not addressed. Many scholars have rejected the notion that the tense of subjunctive clauses is anaphoric, for example Kempchinsky (1986), Suñer and Padilla Rivera (1987), and Costantini (2005). As we saw above in Sect. 2.6, tense construal in OF subjunctive clauses was not determined by the tense of the main clause.

5.2

Kempchinsky (1986)

In Kempchinsky’s view, both subjunctive clauses and indicative clauses can have restricted tense, as the moment of speech and the time of hearing affect interpretation of tense. In order to account for obviation, she argues that it is due to the presence of a subjunctive/imperative operator in Comp, which is identified by INFL after it raises to COMP to identify the subjunctive operator, and is coindexed with the matrix verb (Kempchinsky 1986: 167):

(23) [NP INFL Vi[S[CINFLi][SNPti . . . ]]]] Because of the coindexing, Kempchinsky assumes that it is the syntactic relationship between the subjunctive/imperative operator, INFL, and the main verb, that extends the Binding Domain to the matrix clause. Unlike Picallo (1985) and Jakubowicz (1985), she accounts for the lack of obviation in Romanian, claiming that because sa® is a feature on a head, it, along with V and INFL, front to COMP. In such a case, the Binding Domain is the subordinate clause, because the verbs are not coindexed. Yet Kempchinsky’s analysis cannot account for the lack of obviation in OF, because, as we will argue, the earlier stage of the language had two different complementizers.

214

5.3

D. L. Arteaga

San Martín (2007)

San Martín (2007) looks elsewhere to propose an account of obviation within the Minimalist Framework, which takes into account cross-linguistic evidence for Greek. Situating her analysis within the Minimalist framework, she proposes that subjunctive clauses are headed by CForce (CF) and a Mood Phrase (MP). The central idea is that languages that have a complementizer or special particle that solely introduces the subjunctive clause, such as Romanian să or Greek (hina), do not have obviation. These subjunctive markers are in the MP. In languages without a special subjunctive complementizer or subjunctive particle, obviation holds. San Martín follows Chomsky (2005), in arguing that T’s phi and Tense features derive from C. This means that that T has no independent phi features, unless selected by C. One of these features, she argues, is the [IRR] (irrealis) feature, which we will adopt.12 San Martín assumes that in cases of volitional and desiderative complements, true hypotaxis, in her view, there is an interpretable [IRR] feature on Mood (cf. Rivero 1994; Rizzi 1997). She provides the following structure (San Martín 2007: 185):

(24) CForce . . . (Topic) . . . (Focus) . . . Fin/Mood San Martín further assumes, as in Chomsky (2005), that T only has phi and Tense features when they are derived from C. She adopts the structure proposed by Rizzi (1997):

(25) FORCE

(TOP*) FOC

(TOP*)

FIN

IP

In her view, the representation of the subordinate MP like Greek or Romanian is the following in volitional and desiderative clauses (San Martín 2007: 161):

(26) . . . v . . . [C F . . . . . Mood . . . T . . . . v]AGREE u[IRR] [IRR] (27) [v . . . . [C F . . . Mood . . . T . . .DP . . . v]]

12

Other scholars have argued against as a simplistic view of the subjunctive (cf. Kempchinsky 1986; Costantini 2005).

Obviation and Old French Subjunctive Clauses

215

(28) [v . . . [C F . . . Mood . . . T . . . DP . . . v]] [u ] [Case]

For San Martín, it the interpretable feature IRR (irrealis) that values the uninterpretable feature on CF (cf. Rizzi 1997) by means of the operation Agree (cf. Chomsky 2005). In such a case, the Binding Domain is the lower clause, based on her redefinition of the Binding Phrase using Probes and Goals, which incorporates the idea of a Phase Head (San Martín 2007: 186)13: (29) X is the binding domain D for Y iff X is the minimal maximal projection which includes Y, Z, and PH, where Z is Y’s probe and the Phase Head(s) that transfer(s) Agreeing Features to Z.2

If there is a special subjunctive particle or complementizer, the Binding Domain is not extended to the main clause. In other words, the pronoun can be co-referential with the subject of the main clause, as Principle B of the Binding theory does not apply. The properties of C, including irrealis, get transferred to the subordinate T. For languages that have obviation effects, which in her view are “neutral” to the subjunctive/indicative distinction, it is the higher PH that is relevant, in this case v, it is v that contains u[IRR] in C-Force. Obviation follows from this analysis, as T only inherits [IRR] when the next matrix v is merged with its argument structure. This is shown below: (30) [DPi . . . v . . . [CF . . . Mood . . . T . . . DP*i/j . . . v]

One advantage of San Martín’s analysis is that it can formally account for not only obviation, but also for its diachronic evolution. According to San Martín, in Classical Greek, no special particle was associated with subjunctive particles, so that coreference was only possible with indicative clauses, as is the case in MF. At this stage, volitional and desiderative verbs took an infinitive where the matrix and lower clauses were coreferential. However, in early Hellenistic Greek, hina became the subjunctive complementizer, and obviation effects were lost. This is relevant to our discussion, as we will claim below that there were two different complementizers in OF, one introducing indicative clauses and one introducing subjunctive clauses.

13

As San Martín notes, this is very similar to the formulation of the Binding Theory in Lasnik and Uriagereka (2005).

216

5.4

D. L. Arteaga

Arteaga (2015)

Using data from OF, Arteaga (2015) presents an analysis of obviation based on the work of San Martín, arguing that Mood via Agree can value the uninterpretable feature irrealis on CF in the earlier stage of the language. She bases her analysis on the fact that the subjunctive morphology was robust in OF, as there were morphological differences between the two moods in the earlier stage of that language, as discussed in Sect. 2.4 above. The morphology, combined with the fact that subjunctive que is optional, leads Arteaga to argue that in OF, the subjunctive was more “independent.” In order to provide a theoretical explanation of this, she adopts San Martín’s analysis, claiming that these differences are due to the fact that the IRR feature in CF is interpretable in OF. Given the emergence of obviation in French, Arteaga claims that in OF, the IRR on MP becomes uninterpretable due to the loss of the morphological features described in Sects. 2–2.2, namely the case system and verbal conjugations. Following Lightfoot’s (1999) cue-based analysis of language change, she argues that because of the morphological leveling, children would come to rely on obviation and the presence of que in order to identify subjunctive complements. One advantage of Arteaga’s (2015) analysis is that it allows the existence of infinitives and subjunctive complements, unlike Arteaga (1990), who claimed that infinitival complements were rare in OF, a claim that cannot be maintained if large corpora are considered. Arteaga (2015) further captures the fact that the subjunctive in OF was fundamentally different from that in MF, because of its syntactic independence, which it allowed it to occur without que. Yet there are two central problems with Arteaga’s account. One is its focus on morphology, and the second is her claim that OF had only one complementizer for both the indicative and the subjective, que. Arteaga argues that there is a marked morphological difference between the OF subjunctive and the MF subjunctive. But it is not the case that OF evinced more morphological differences than MF. For example, in the case of er verbs, while OF distinguished the 2nd and 3rd persons singular, MF shows a difference between the first and second persons plural ([ɔ̃] vs. [jɔ̃], [e] vs. [je]). In the ir and re conjugations, in the subjunctive, the final consonants are pronounced, yielding contrasts throughout the indicative/subjunctive paradigms such as je finis ([fini] vs je finisse [finis] and je vends [vɛ̃] vs. je vende [vɛ̃d]. Similar morphological contrasts can be seen with suppletive verbs (i.e., indicative être ‘to be’ suis/es/est vs. subjunctive soie/sois/soit and indicative avoir ‘to have ai/as/a vs. subjunctive aie/aies/ait).14 In our view, it is therefore untenable to claim that French lost the morphological indicative/subjunctive distinction. Another piece of evidence suggesting that Arteaga (2015) should be revised is the fact that parataxis, or the omission of the 14

See Appendices 3–4 for the complete conjugation of regular and suppletive subjunctive forms in Modern French.

Obviation and Old French Subjunctive Clauses

217

complementizer, such as the English I think I can come can always occur in OF when the subordinate clause is in the subjunctive, but not in the indicative.15 The examples (11)–(14) above, repeated below as (31)–(34), illustrate this syntactic phenomenon: (31) Dex confonde tote lor gente. God-m-sg-NOM destroy-3-sg-SUBJ all-fsg-OBL their-f-sg-OBL people-f-sg-OBL ‘May God destroy all of their people!’ (Brut 13756) (Kowalski 1882:3)

(32) Car fusse je en Babiloine so were-1sg-SUBJ I in Babylon! ‘If only I were in Babylon-f-sg-OBL!’ (Fabliaux 9.352) (Jensen 1990:357) (33) Cuida [chose fust par li sole] thought-3sg thing-F-SG-OBL done-3sg-SUBJ by her alone ‘He thought (that) she alone did that.’ (although Renart actually did) (Renart 2603) (Jensen 1974:55)

(34) Nos volons we want-1pl [son nevo en chast] his-m-sg-OBL nephew-m-sg-OBL from.there banish-3sg- SUBJ ‘We want him to banish his nephew.’ (literally, ‘We want (that) he banish his nephew.’) (Béroul, 602) (Graeme-Ritchie 137)

As discussed in Sect. 2.5 above, the complementizer que can be unexpressed in main clause subjunctives and contrary-to-fact clauses, as in (30) and (31). The subordinate clauses in (3) and (34) are not introduced by que; in both cases, V2 word order was found. In our view, these examples illustrate the independence of the subjunctive. 15

There are isolated cases of omission of que when it heads an indicative clause, but the generalization is that parataxis is always possible with subjunctive clauses. See Anglade (1965), Jensen (1990).

218

5.5

D. L. Arteaga

Lasokva (2017)

Laskova (2017) is the first to discuss the subjunctive from a broad cross-linguistic perspective, considering data from the Balkan languages, the Slavic languages, the Germanic languages, and the Romance languages. Her goal is to isolate what the subjunctive is, in so doing, tease apart the many characteristics that have been attributed to the subjunctive mood. Laskova notes that the subjunctive occurs in a wide variety of languages, but its use is not uniform. For example, she points out that in most languages of Europe, the subjunctive is closely related to whether the utterance is counterfactual, while in the case of Persian, it only occurs if there is uncertainty about whether or not something will happen in the future (Laskova 2017: 1). As she notes, in the Germanic languages, the subjunctive can be used for reported speech after verba dicendi. As is the case in OF, (cf. (15)–(17)) SOT restrictions are not found in Balkan languages, as the following examples from Bulgarian illustrate (Laskova 2017: 6)16,17:

(35) Predi edna before one

sedmica week

Ivan Ivan

ne vjarva e NEG believe-past

Maria da e v rne tolkova skoro. Maria DA.Subj. come back so soon ‘One week ago Ivan did not believe that Maria would come back so soon.’

(36) Ivan Ivan

ne Neg

vjarva believe.Pres

Maria da e v rne tolkova skoro. Maria DA.Subj. come back so soon ‘Ivan doesn’t believe that Maria will come back so soon.’ In the complement clause in (35), although the matrix verb has a past morpheme, še, the embedded verb does not; similarly, in (36), the subordinate verb does not share the present morpheme of the main clause.18 These data lead Laskova to

16

Laskova uses the abbreviation DA.Subj to refer to the subjunctive particle, da. The present morpheme in this case is va. 18 The exception to this is in cases where deictic forms such as ‘now’ sega or ‘yesterday’ vc̆era are present. 17

Obviation and Old French Subjunctive Clauses

219

conclude that the morphological realization and the temporal properties of the subjunctive must be teased apart. Therefore, in her view, in the Romance languages, temporal agreement is independent of the subjunctive. Laskova’s main proposal is that the subjunctive lacks temporal anchoring to the time of the matrix clause, due to an Exclusion Feature (ExclF) feature, in the sense of Iatridou (2000). In languages that have a particle indicating mood, there can be no temporal anchoring. One claim that is relevant to our discussion is that in the Romance languages, the agreement morphology of the subjunctive is independently present. Specifically, she proposes that the subjunctive particle ExclF is what gives subjunctive clauses the possibility of temporal anchoring. In her analysis, the role played by ExclF is to block the subjunctive from referring to the time of an utterance or the actual world. It is important to note that her proposal allows for cross-linguistic variation, in that in many languages, including OF (see Sect. 2.6 above) the subjunctive does have independent tense. In other words, where there are no SOT restrictions, the time reference comes from a particle that fills the function of Exc1F. To support her analysis, she presents extensive data from Bulgarian which illustrate the differences between three related verbal structures, the subjunctive, the indicative, and the non-past perfective (NPF), in order to disambiguate what have been traditionally considered to be properties of the subjunctive. She notes that while the NPF appears to carry the same semantic properties as the subjunctive, in Bulgarian, the NPF and the subjunctive are not found in the same contexts, and in fact constitute morphosyntactic minimal pairs. Because of the semantic similarity, she argues that the differences can reveal what constitutes the subjunctive (Laskova 2017: 10): (37) (Nosja si kr stoslovica./Bring Refl. cross-word puzzle.) Taka i da čakam, thus even SUBJ wait na opa ka on a.queue njama da mi e not DA to.me is skučno. boring (‘I bring with me a cross-word puzzle.’) ‘In this way, even if I wait in a queue, ’ I will not be bored.

220

D. L. Arteaga

(38) (Za vseki slučaj si nosja kra˘stoslovica./ In any case, I have taken with me a cross-word puzzle.) Taka, dokatodojde mojat red, njama da mi Thus until come.NPF my turn, Fut Neg to me ‘In this way I will not get bored until my turn comes.’

e be

skučno. boring

In the above examples, a lack of presupposition is not relevant. The speaker is merely telling another person that she plans to go to the post office, but she doesn’t know if it is open. In other words, it is a hypothetical situation. No other verb forms are grammatical in Bulgarian. Either the subjunctive particle da is found or the NPF, but not both. Thus, the crucial piece of the puzzle is how the NPF alternates with the subjunctive. Laskova presents the following comparison of the NPF, the subjunctive, and the indicative in Bulgarian. The following table gives a summary of the syntactic similarities and differences that are relevant to our discussion (adapted from Laskova 2017: 18): (39)

Possibility to form a main clause as a single verb form Possibility to form complement clauses as a single verb form Possibility to form a declarative clause Possibility to temporally locate the event with respect to the matrix clause event

NPF NO

Subjunctive YES

Indicative YES

NO

YES

YES

YES NO Future-shifted interpretation only

NO YES

YES YES

The NPF and the indicative share certain characteristics in that they can both form a declarative clause. Yet the NPF cannot appear alone in a main clause or in a subordinate clause, unlike both the indicative and the subjunctive. Both the subjunctive and the indicative can also agree temporally with the tense of the matrix clause, which differentiates them from the NPF, which has no temporal element, and needs either the modal particle da or an event in the matrix clause that anchors it. Finally, both the subjunctive and the indicative are found in complement clauses. Because of these differences, Laskova argues that in Bulgarian, the subjunctive is not ambiguous, as it in the Modern Romance languages. It is clear from the data that we have seen, that the OF subjunctive was syntactically unambiguous, as well, in that it shares the features of the Bulgarian subjunctive. It can appear as a main clause, as the examples in (11) and (12) show, and can appear in complement clauses, as in (13)–(14). Finally, although it is not required, in OF, the clause can agree temporally with that of the main clause (cf. (13)–(14)). Laskova’s main argument for the Romance languages, then, is that SOT restrictions in the subjunctive is independent of the subjunctive morphology, but is

Obviation and Old French Subjunctive Clauses

221

rather due to the attitude predicate in the main clause. Our examples in (15)–(17) above also show OF also lacked temporal anchoring in the subjunctive, in that there are no SOT restrictions. Specifically, Laskova, following Giorgi (2010), proposes the following structure (Laskova 2017: 23):

(40) T (e1, e2)

e2 (Φ)

T (e1, e’)

T

V (e1)

ExclF V

e1

For Giorgi, U represents the matrix event, which in his view, must anchor the event of the matrix clause, e being the embedded event, e′, being the matrix eventuality. The feature e2 is a variable that is in turn defined by U. For Laskova, this feature does not serve as an anchor for the embedded event in subjunctive clauses. She proposes that the temporal relation between the two clauses is the result of ExclF in the lower T, instead of precedence/follow/overlap (Giorgi). Her analysis captures the generalization that there is an element that prevents the subjunctive from being temporally anchored, and is in accordance with several semantic-based analyses of the subjunctive (cf. Higginbotham 1995; Giorgi and Panesi 2001). Laskova, citing Tomaszewicz’s (2009), analysis of Polish, notes that her account is not incompatible with a MP (cf. San Martín). Yet for our purposes, the main disadvantage of Laskova’s analysis is that it does not formally account for obviation. Indeed, she argues that obviation cannot be a marker of the subjunctive, which is valid, as OF lacked obviation yet had a special complementizer that marked the subjunctive. Moreover, she does not provide a formal analysis regarding how subjunctive particles can “substitute” for ExclF, because her analysis is not feature-based. In the next section, we give an account of obviation that incorporates both San Martín’s and Laskova’s analyses.

222

D. L. Arteaga

6 Proposal In this section, we sketch a preliminary feature-based account of obviation. Because of the cross-linguistic validity of Laskova’s analysis, we assume that any account of the subjunctive in Romance must incorporate the notion that the subjunctive is not temporally anchored (as we have demonstrated above, Sect. 2.6), and we will consider the specific subjunctive temporal anchor to be ExclF. For OF, however, we argue that it is the subjunctive complementizer itself (instead of a subjunctive particle) that marks the subjunctive. We base our analysis on the features of OF que that MF lacks, namely that it can be deleted in any subjunctive context, including contrary-to-fact subjunctives, verbs of believing, main clause subjunctives, and verbs of volition. When que is unexpressed, which only occurs in subjunctive clauses, V2 order is found, as in (12)–(14) above. Adapting Laskova’s analysis (cf. Laskova 2017: 234), the structure we therefore propose is the following (cf. (30) and (40):

(41)

que

MoodP i[IRR]

i[ExclF] T (e1, e2)

e2 (Φ)

T (e1, e’)

V (e1)

V

e1

This structure captures Laskova’s analysis of temporal interpretation, but also allows us to account for obviation, by incorporating San Martín’s account. We adapt Laskova’s hypothesis regarding the subjunctive particle da, which she claims can play the role of the temporal anchor in the subjunctive; in OF, this is the subjunctive que. As proposed by San Martín, CF selects the feature IRR, which is valued by the IRR feature on MP via Agree. Assuming San Martín’s reformulation of the Binding Domain (29), obviation does not hold in languages that have special complementizers for the subjunctive, as it is the lower clause that is relevant, as it contains the embedded subject, its Probe (embedded T), and the Phase Head (CF).

Obviation and Old French Subjunctive Clauses

223

We posit ExclF as an uninterpretable feature on T valued by the interpretable feature on the subjunctive complementizer que. We adopt the analysis in Arteaga (2015) in that we assume that CF is associated with IRR in the lexicon. As we saw in Sect. 5.4, Arteaga does not differentiate between subjunctive and indicative que. We propose that it is subjunctive que that values ExclF, as would be the case for to Balkan da or Greek (hi)na. We have shown above that the subjunctive is syntactically unambiguous in OF, because the subjunctive can form a main (non-declarative) clause without que, (11) a complement clause as a single verb (19), and it shows no SOT restrictions (cf. (39) above). The question arises how subjunctive que can value ExclF, when it is unexpressed. Here we adopt the analysis in Arteaga (2009), who argues against true parataxis in OF, positing instead a null complementizer. She presents a wide range of evidence for this, including the distribution of expletive il, the possibility of V2 word order when que is unexpressed, and the fact that verbs of believing, in affirmative constructions, even in the absence of the complementizer, can take either the indicative or the subjunctive: (42) Je cui-t

[plus sot

de ti

n

’i

a]

I believe-1SG more stupid.M.SG.OBJ of you not there has.3SG pro ‘I think (that) there isn’t anyone more stupid than you.’ (cf. MF Je pense qu’il n’y a personne plus.sot que toi (Feuillée 341) (Togeby 1974 § 151)

(43) Cui-t

[ne

soi-ent

mort

thinks-3SG pro NEG.EXPL are-3PL.SUBJ dead.M.PL.NOM l-i

autre]

the-M.PL.NOM others.M.PL.NOM ‘I think that the others are dead (but I am mistaken).’ (Cf. MF Je crois que les autres sont (indicative) morts. (Mort Artu 98, 11) (Jensen 1990 §756)

In (43), the subjunctive in the lower clause indicates a false belief, reinforcing the fact that the subjunctive in OF had independent meaning. This suggests, in turn, that C can value the ExclF feature in T, regardless of whether it is expressed. In the present analysis, the fact that the subjunctive/indicative morphology is different in OF is incidental to the evolution, as it continues to be fairly robust. What is crucial is the fact that in MF, it is no longer possible for que to be

224

D. L. Arteaga

unexpressed in subjunctive complements. Adopting the analysis of Laskova, we propose that que no longer values ExclF. In our account, subjunctive que became unable to serve as a temporal anchor, as it merged with indicative que. The feature IRR becomes uninterpretable in Modern French for the same reasons, explaining why verbs take must either the subjunctive or the indicative, but not both (cf. 42–43 above). In order to account for obviation, we further follow San Martín in proposing that in languages without a specific marker of the subjunctive, such as Modern French, the Binding Domain becomes the main clause, because MP cannot value the [IRR] feature on CF. The Phase Head is therefore v in the main clause, meaning that co-reference is impossible (San Martín 2009: 187):

(44) [DPi . . . v . . . [CF . . . Mood . . . T . . . DP . . . . v]] [IRR] [uIRR]

[u ] [Case] [IRR] [ ]

A piece of evidence that supports our analysis is that que ceased to be optional in most subjunctive contexts in the 13th century (Jensen 1990), and V2 word order in subordinate clauses was lost. For children learning MF, que alone does not distinguish the indicative from the subjunctive, as it must be expressed in either context. V2 word order is not found in subordinate clauses, expletive il is obligatory, and mood choice is largely fixed. Instead, children must rely on verbal morphology, obviation, and the lack of independent meaning (cf. 42–43 above). In Kempchinsky’s (2009) terms, the subjunctive has become less independent in MF. Our analysis has the advantage of not relying solely on morphological cues to differentiate the indicative from the subjunctive. It also takes into account Laskova’s extensive discussion of the subjunctive cross-linguistically. Moreover, it captures the spirit of the proposal made by San Martín, which is summed up below (2007: 182): (45) The correct generalization is that free reference is displayed by languages that encode the subjunctive Mood in the Left periphery of embedded clauses.

7 Conclusion In this chapter, we have considered the lack of obviation in OF and its emergence. We have reviewed several previous analyses of obviation, noting that none accounts for the fact that OF did not evince obviation. Considering the

Obviation and Old French Subjunctive Clauses

225

cross-linguistic validity of Laskova’s (2017) analysis of the subjunctive, we gave a preliminary account based on feature valuation, proposing that subjunctive que in OF values the uninterpretable feature ExclF on T. We have followed San Martín (2007), by attributing the lack of obviation in OF to the feature IRR in MP, which was interpretable and could value the uninterpretable IRR on CF. This allows for the limitation of the binding domain to the lower clause, which in turn, can account for the OF data. The fact that subjunctive que merged with indicative que meant that it can no longer value the uninterpretable feature of ExclF on T. Our analysis extends that of Arteaga (2015), because it does not depend on the morphological evolution of the subjunctive in French, which still contrasts the indicative from the subjunctive. Further research is needed to test our account of obviation cross-linguistically. Acknowledgements I want to thank Julia Herschensohn for her extensive comments on earlier drafts of this paper, which I dedicate to her. She has been unconditionally supportive of both my professional and personal lives, and I owe her a great deal of gratitude.

Appendix 1: Old French Indicative/Subjunctive Verbal Paradigms, Regular Verbs er verbs soner ‘to ring’ present indicative 1sg son1pl sonons 2sg sones 2pl sonez 3sg sonet 3pl sonent

present subjunctive 1sg 1pl 2sg sons 2pl 3sg sont 3pl

sonons sonez sonent

ir verbs finir/fenir ‘to finish’ present indicative 1sg finis 1pl finissons 2sg finis 2pl finissez 3sg finist 3pl finissent

present subjunctive 1sg finisse 1pl 2sg finisses 2pl 3sg finisset 3pl

finissons finissez finissent

re verbs vendre ‘to sell’ present indicative 1sg vent1pl vendons 2sg venz 2pl vendeiz/ vendez 3sg vent 3pl vendent

present subjunctive 1sg vende 1pl 2sg vendes 2pl 3sg vendet 3pl

vendons vendez vendent

226

D. L. Arteaga

Appendix 2: Old French Indicative/Subjunctive Verbal Paradigms, Suppletive Verbs avoir ‘to have’ present indicative 1sg ai 1pl avons 2sg as 2pl avez 3sg at 3pl ont

être ‘to be’ present indicative 1sg suis 1pl somes 2sg es 2pl estes 3sg est 3pl sont

present subjunctive 1sg aie 1pl 2sg aies 2pl 3sg aiet/ait 3pl

present subjunctive 1sg seie 1pl 2sg seies 2pl 3sg seit 3pl

aiiens/aiions aiiez aient

seions seiiez seient

Appendix 3: Modern French Indicative/Subjunctive Verbal Paradigms, Regular Verbs er verbs sonner ‘to ring’ present indicative 1sg je sonne 1pl nous sonn ons 2sg tu sonnes 2pl vous sonn ez 3sg il sonne 3pl ils sonnent

ir verbs finir ‘to finish’ present indicative 1sg je finis 1pl nous finissons 2sg tu finis 2pl vous finissez 3sg il finit 3pl ils finissent

present subjunctive 1sg je sonne 1pl nous sonn ions 2sg tu sonnes 2pl vous sonniez 3sg il sonne 3pl ils sonnent

present subjunctive 1sg je finisse 1pl nous finiss ions 2sg tu fini sses 2pl vous finissiez 3sg il finisse 3pl ils finissent

Obviation and Old French Subjunctive Clauses

re verbs vendre ‘to sell’ present in dicative 1sg vends 1pl vendons 2sg vends 2pl vendez 3sg vend 3pl vendent

227

present subjunctive 1sg vende 1pl 2sg vendes 2pl 3sg vende 3pl

vendions vendiez vendent

Appendix 4: Modern French Indicative/Subjunctive Verbal Paradigms, Suppletive Verbs être ‘to be’ present indicative 1sg je suis 1pl nous sommes 2sg tu es 2pl vous êtes 3sg il est 3pl ilssont

avoir ‘to have’ present indicative 1sg j’ai 1pl nous avons 2sg tu as 2pl vous avez 3sg il a 3pl ils ont

present subjunctive 1sg je sois 1pl 2sg tu sois 2pl 3sg il soit 3pl

present subjunctive 1sg j’aie 2sg tu aies 3sg il ait

1pl 2pl 3pl

nous soyons vous soyez ils soit

nous ayons vous ayez ils aient

References Adams, Marianne. 1987. From Old French to the theory of pro-drop. Natural Language & Linguistic Theory 5: 1–32. Anglade, Joseph. 1965. Grammaire élémentaire de l’ancien français. Paris: Armand Colin. Arteaga, Deborah (ed.). 1990. The disjoint reference requirement in subjunctive clauses: Diachronic evidence from Romance. Unpublished Ph.D. thesis, University of Washington. Arteaga, Deborah (ed.). 1994. Impersonal constructions in Old French. In Issues and theory in Romance linguistics, ed. Michael Mazzola, 141–156. Washington, D.C.: Georgetown University Press. Arteaga, Deborah (ed.). 2009. On the existence of null complementizers in Old French. In Romance linguistics: Structure, interfaces, and microparametric variation, ed. Pascual Masullo, Erin O’Rourke, and Chia-Hui Huang, 19–36. Amsterdam: Benjamins. Arteaga, Deborah. 2015. On the Old French subjunctive. In Romance linguistics 2012, ed. Jason Smith and Tabea Ihsane, 147–170. Amsterdam: Benjamins. Arteaga, Deborah, and Julia Herschensohn. 2006. Il était une fois: Diachronic development of expletives, case, and agreement from Latin to Modern French. In Historical Romance linguistics: Retrospectives and perspectives, ed. Randall Gess and Deborah Arteaga, 267–286.

228

D. L. Arteaga

Arteaga, Deborah, and Julia Herschensohn. 2004. Case, agreement, and expletives. A parametric difference in Old French and Modern French. In Contemporary approaches to Romance linguistics, ed. Julie Auger, Clancy Clements, and Barbara Vance, 1–15. Amsterdam and Philadelphia: Benjamins. Brunot, Ferdinand. 1966. Histoire de la langue française des origines à nos jours. Paris: Librairie Armand Colin. Chomsky, N. (2001). Derivation by Phase. In M. Kenstowicz (Ed.), Ken Hale: A Life in Language, 1–52. Cambridge, MA: MIT Press. Chomsky, Noam. 2005. On phases. Ms: MIT. Costantini, Franceso. 2005. On obviation in subjunctive clauses: The state of the art. Annali di Ca’Foscari 44: 97–132. Costantini, Francesco. 2007. On Tense Mismatch and a Morphosyntactic Theory of Sequence of Tenses. Rivista di Grammatica Generativa 32:39–64. Costantini, Franceso. 2016. Subject obviation in subjunctive clauses as a semantic failure. Annali di Ca’ Foscari. Serie occidentale 50: 109–131. Einhorn, Elsabe Carmen. 1975. Old French: A concise handbook. Cambridge: Cambridge University Press. Farkas, Donka. 1984. Subjunctive complements in Rumanian. In Papers from the XIIth linguistic symposium on Romance languages, ed. Philip Baldi, 354–372. Amsterdam and Philadelphia: Benjamins. Foulet, Lucien. 1982. Petite syntaxe de l’ancien français. Paris: Librairie Honoré Champion. Giorgi, Alessandra. 2010. About the speaker. New York: Oxford University Press. Giorgi, Alessandra, and Fabio Panesi. (1997). Tense and Aspect. In From Semantics to Morphosyntax. New York/Oxford: Oxford University Press. Giorgi, Alessandra, and Fabio Panesi. 2001. Tense, attitudes, and subjects. In Proceedings of SALT, vol. 11, ed. B. Hastings, B. Jackson, and Z. Zvolensky, 226–249. Ithaca: University of Cornell Press. Glikman, Julie, and Thomas Verjans. 2013. Old French parataxis: Syntactic variant or stylistic variation? In Historical research on Old French: The state of the art, ed. Deborah Arteaga, 243–260. Amsterdam: Springer. Higginbotham, James. 1995. Tensed thoughts. Mind and Language 10: 226–249. Iatridou, Sabine. 2000. The grammatical ingredients of counterfactuality. Linguistic Inquiry 31: 231–270. Ingham, Richard. 2013. A derivational approach to negative polarity licensing in Old French. In Historical research on Old French: The state of the art, ed. Deborah Arteaga, 262–282. Amsterdam: Springer. Jakubowicz, Celia. 1985. Do binding principles apply to INFL? Paper presented at NELS, vol. 15. Jensen, Frede. 1990. Old French and comparative Gallo-Romance syntax. Tübingen: Niemeyer. Kempchinsky, Paula. 1986. Romance subjunctive clauses and logical form. Unpublished Ph.D. thesis, UCLA. Kempchinsky, Paula. 2009. What can the subjunctive disjoint reference effect tell us about the subjunctive? Lingua 119: 1788–1810. Laskova, Vasselina. 2017. On the nature of the subjunctive. Lingua 189–90: 19–45. Lasnik, Howard, and Juan Uriagereka. 2005. A course in minimalist syntax. Oxford: Blackwell. Lightfoot, David. 1999. The development of language: Acquisition, change, and evolution. Oxford: Blackwell. Martineau, France. 1994. The expression of the subjunctive in older French. Catalan Working Papers in Linguistics. 32.45–69. Mathieu, Éric. 2013. The left-periphery in Old French. In Historical research on Old French: The State of the art, ed. Deborah Arteaga, 327–350. Amsterdam: Springer. Ménard, Pierre. 1988. Syntaxe de l’ancien français. Paris: Bordeaux Éditions Bière. Moignet, Gérard. 1976. Grammaire de l’ancien français. Paris: Klincksieck. Nyrop, Kristofer. 1930–1935. Grammaire historique de la langue française, 6 vols. Copenhagen: Gyldendalske Boghandel.

Obviation and Old French Subjunctive Clauses

229

Picallo, M. Carme. 1985. Opaque domains. Unpublished Ph.D. thesis, CUNY. Raposo, Eduardo. 1987. Case theory and Infl-to-Comp: The inflected infinitive in European Portuguese. Linguistic Inquiry 18: 85–109. Rivero, Marisa. 1994. Clause Structure and V-movement in the Languages of the Balkans. Natural Language and Linguistic Theory 12:63–120. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Elements of grammar: Handbook in generative syntax, ed. Liliane Haegeman, 281–337. Dordrecht: Kluwer. Roberts, Ian. 1993. Verbs and diachronic syntax: A comparative history of English and French. Norwell: Kluwer. Ruwet, N. 1984. ‘Je veux partir/Je veux que je parte’. À propos de la distribution des complétives à temps fini et des compléments à l’infinitif en français. Cahiers de grammaire 7: 76–138. San Martín, Itziar. 2007. Beyond the infinitive vs. subjunctive rivalry: Surviving changes in mood. In Coreference modality, and focus: Studies on the syntax-semantics interface, ed. Luis Eguren and Olga Fernández-Soriano, 171–190. Amsterdam: Benjamins. Suñer, Margarita, and José Padilla Rivera. 1987. Sequence of tenses and the subjunctive, again. Hispania 70: 634–42. Togeby, Knud. 1974. Précis historique de grammaire française. Odense: Akademisk Forlag. Tomaszewicz, Barbara. 2009. Subjunctive mood in Polish. In Studies in formal Slavic phonology, morphology, syntax, semantics and information structure: Proceedings of FDSL 7, Leipzig 2007 (Linguistik International 21), ed. Gerhild Zybatow, Uwe Junghanns, Denisa Lenertová, and Petr Biskup, 221–234. Frankfurt am Main: Peter Lang. Vance, Barbara. 1997. Syntactic change in Medieval French. Dordrecht: Kluwer.

On the Varying Fortunes of darla and darlo in Italian Cinzia Russi

Abstract Darla (‘give it-fem.’) ‘of women, to have sexual relations with someone, especially in a loose manner’ (De Mauro 1999–2000: II, 469) and darlo (‘give it-masc.’), which is attributed two meanings, ‘to lend oneself to sodomy’ (De Mauro 1999–2000: II, 470) and ‘of men, to sexually possess a woman’ (Boggione and Casalegno 2000: 152; 2004: 166), have enjoyed varying fortunes in the Italian language. While darla is well-established in the average speaker’s lexicon, darlo (in both meanings) remains highly marginal. These two verbs attest to a lexical asymmetry resulting from the misogynistic view ingrained in Western culture of women as inferior, inherently immoral, and wanton, and their different fates show that sociocultural attitudes and stereotypes shape language change.










Keywords Italian Lexical asymmetry Lexical change Socio-cultural attitudes Stereotypes Linguistic sexism verbi procomplementari










1 Introduction The goal of this study is twofold: first, it aims to trace the emergence of the verbs darla “di donna, avere rapporti sessuali con qcn., specialmente con facilità” ‘of women, to have sexual relations with someone, especially in a loose manner’ (De Mauro 1999–2000: II, 469) and darlo, to which two meanings are assigned, “prestarsi a pratiche sessuali sodomitiche” ‘to lend oneself to sodomy’ (De Mauro 1999–2000: II, 470) and “di uomo, possedere sessualmente una donna” ‘of men, to sexually possess a woman’ (Boggione and Casalegno 2000: 152; 2004: 166; henceforth, darlo2).1 Second, it investigates the differences in distribution and usage of these two verbs at different stages of Italian. 1

Darlo without subscript is used to refer to darlo1 and darlo2 jointly.

C. Russi (&) Department of French and Italian, The University of Texas at Austin, Austin, TX, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_11

231

232

C. Russi

Diachronically, darla appears to be attested since (at least) the sixteenth century. In contrast, darlo2 seems to have entered the language in the second half of the twentieth century, although dating the appearance of darlo1 is rather challenging. In contemporary Italian, darla shows a considerably wider distribution and higher usage frequency than darlo. This is evidenced by, among other things, the absence of darlo in CORIS/CODIS, an extensive online corpus of written modern Italian, where, in contrast, numerous occurrences of darla are found.2 Furthermore, an informal survey among native speakers of different ages and socio-cultural background reveals that darla is well-established in their lexicon, whereas darlo1 is overall unknown and darlo2 is marginally recognized and restricted to younger speakers.3 The main questions I address in this chapter are the following: (a) What accounts for the negligible presence (if not absence) of darlo1 in Italian through time? (b) What factors are responsible for the unequal distribution of darla and darlo2? That is, why does darlo2 continue to be so peripheral in present-day Italian? Uncovering the dynamics of the developments of these two verbs sheds light on the role that socio-cultural attitudes and stereotypical prejudiced beliefs pertaining to the representation and perception of gender and sexuality can play in shaping language use as well as language change. I propose that darla and darlo instantiate a case of linguistic sexism by attesting to a lexical asymmetry in terms of gender (women vs. men), conventional gender roles (submissive women vs. dominant men), and sexual practices (heterosexual or ‘normal’ vs. homosexual or ‘deviant’). Darlo2 has not actually eliminated or leveled out the asymmetry since it still lingers in the fringes of the contemporary Italian lexicon. The chapter is organized as follows: Sect. 2 reviews the definitions of darla and darlo provided in Italian dictionaries (both general and specialized) and points out some inconsistencies in the definitions of darlo. Section 3 examines the earliest attestations of darla and darlo on the basis of evidence from the dictionaries taken into consideration and literary sources. Section 4 assesses and discusses the asymmetric status of darla and darlo in contemporary Italian. Section 5 addresses the relevance of stereotypical socio-cultural attitudes and prejudices toward women and homosexuality/sodomy for the development and persistence of the asymmetry between darla and darlo. Section 6 summarizes the main points of the study and makes suggestions for further research.

2

CORIS/CODIS (http://dslo.unibo.it/), is a corpus of modern written Italian developed at the CILTA (Centre for Theoretical and Applied Linguistics, University of Bologna). Presently, it contains 130 million words and consists of a collection of authentic and commonly occurring texts in electronic format, which are considered to be highly representative of modern Italian, and includes an assortment of diverse genres (press, fiction, academic and legal and administrative prose, miscellanea). 3 I asked about thirty family members and friends, all native speakers of Italian, to provide a definition of darla and darlo and their opinion on appropriate contexts for the use of the verbs.

On the Varying Fortunes of darla and darlo in Italian

233

2 Definitions Darla and darlo belong to the class of verbs labeled verbi procomplementari in Tullio de Mauro Grande dizionario italiano dell’uso (GRADIT, 1999–2000). These are defined as “verbi dotati di un proprio significato non riconducibile al verbo principale o molto cristallizzato” (De Mauro 1999–2000: xxxiv)4 and their distinctive feature is a lexicalized clitic pronoun or pronoun cluster. Darla and darlo are comprised of the verb dare ‘to give’ and the third person singular direct object clitic pronouns la (feminine) and lo (masculine), whose referents are, respectively, ‘vagina’ and ‘anus’ or ‘penis’. These verbs exemplify an instance of euphemization, a very common phenomenon across languages, whereby a taboo term is replaced by a very generic, ‘neutral’ one, a pronoun in this case (Galli de’ Paratesi 1969: 41; Toscan 1981). The definitions of darla found in general dictionaries are given in (1) and the ones from specialized ones are listed in (2).5,6 (1) a. De Mauro (1999–2000: II, 469): “di donna, avere rapporti sessuali con qcn., specialmente con facilità” ‘of women, to have sexual relations with someone, especially in a loose manner’; the verb is marked comune ‘common’ in terms of usage and gergale ‘slang’ in terms of register. b. Battaglia (1961–: IV, 35): “concedersi carnalmente (una donna)” ‘to have sexual intercourse (of women)’.

(2) a. Boggione and Casalegno (1996: 75; 2000: 150–151; 2004: 165–166): “[d]are, darla, darne: concedersi sessualmente, il pronome dimostrativo [sic] allude per eufemismo all’organo femminile” ‘dare, darla, darne: to give oneself sexually, the demonstrative pronoun euphemistically refers to the female sexual organ’. b. Ambrogio and Casalegno (2004: 139): “darla, darla via: concedersi sessualmente, per lo più con facilità (una donna)” ‘to give it, to give it away: to give oneself sexually, in general easily (of women)’.

‘[V]erbs characterized by a meaning of their own, which cannot be inferred from the source verb or have become highly fossilized’. 5 Note that neither darla nor darlo appear in the dictionaries of Accademia della Crusca (online versions, http://www.accademiadellacrusca.it) nor in Vocabolario Treccani (1997) and Grande dizionario della lingua italiana moderna (1998) is the only single-volume general dictionary that includes darla with the same definition as in Battaglia (1961–). 6 Lexemes labeled ‘common’ are “vocaboli che sono usati e compresi indipendentemente dalla professione e mestiere che esercitiamo o dalla collocazione regionale e che sono generalmente noti a chiunque abbia un livello mediosuperiore di istruzione” ‘words that are used and understood independently of our profession, trade or geographical domain, and that are generally known by anyone with a mid-high level of education’ (De Mauro 1999–2000: xx). 4

234

C. Russi

c. Lotti (1992: 125): “darla (via): (rif[erito] a donna) concedersi sessualmente” ‘to give it (away): (referred to women) give oneself sexually”, marked as volgare ‘vulgar’.

d. Forconi (1988: 67): “darla, darla via: concedersi sessualmente, specie con una certa facilità, detto di donna” ‘to give it, to give it away: give oneself sexually, especially with easiness, said of women’.

Darlo1 is found only in De Mauro (1999–2000): (3) “prestarsi a pratiche sessuali sodomitiche” ‘to lend oneself to sodomy’; the verb is marked comune ‘common’ in terms of usage and volgare ‘vulgar’ in terms of register, and the synonym “dare via il culo” ‘to get fucked in the ass [lit. to give away one’s ass]’ is provided (II, 470).

Lotti (1992: 126) reports the expression “darlo via” ‘to give it away’ defined as “praticare pederastia passiva” ‘to practice passive pederasty’ and marks it volgare ‘vulgar’. Finally, the definitions of darlo2, all from specialized dictionaries, are given in (4): (4) a. Lotti (1992: 125): “possedere carnalmente una donna” ‘to sexually possess women’; this is given as meaning 1, while “essere pederastam attivo” ‘be an active pederast’ is given as meaning 2; both are marked volgare ‘vulgar’. b. Boggione and Casalegno (2000: 152; 2004: 166): “di uomo, possedere sessualmente una donna” ‘of men, to sexually possess a woman’. c. Ambrogio and Casalegno’s (2004: 139): “possedere sessualmente” ‘to possess sexually’.

The items in (1)–(4) reveal a disparity between darla and darlo in terms of (a) the number and type of dictionaries that include them, and (b) the definitions they are assigned. Concerning (a), darla is included in more dictionaries (both general and specialized). In contrast, neither darlo1 nor darlo2 are found in Battaglia (1961–), which is the most comprehensive dictionary of the Italian language comprising twenty four volumes, and only one (specialized) dictionary features both darlo1 and darlo2 (Lotti 1992). Darlo1 is found only in two dictionaries (one general and one specialilzed, in the latter as the expression darlo via) but, quite surprisingly, is absent in Boggione and Casalegno (2000, 2004), which are dictionaries of erotic lexicon.7 Finally, darlo2 is found only in specialized dictionaries.

Boggione and Casalegno include the expression dare di dietro ‘to get fucked [lit. to give] from behind’ under the entry for dare with the definition of “sodomizzare” to ‘sodomize’ (2000: 152; 2004: 167), while the expression dare il culo ‘to get fucked in the ass’ is included in the index of voices by semantic field under Omosessualità e sodomia ‘Homosexuality and sodomy’ (2000: 681; 2004: 742). 7

On the Varying Fortunes of darla and darlo in Italian

235

As for (b), the definitions of darla are consistent and unambiguous as to the referents of the subject (‘women’) and the clitic pronoun (‘vagina’), whereas the definitions of darlo display discrepancies regarding the referents of both the subject and lo, as well as with respect to the absence/presence of a homosexual connotation. De Mauro’s definition—where the synonym given indicates that sodomy denotes both homo- and heterosexual anal intercourse (see (3) above)—does not restrict darlo1 in terms of subject (i.e. it can be either ‘men’ or ‘women’), but the referent of lo can only be ‘anus’. On the other hand, Lotti’s definition restricts darlo1 (which he has as darlo via) also in terms of subject (only ‘men’). In the case of darlo2, all the definitions conform in restricting the subject to ‘men’ and the referent of lo to ‘penis’, but they diverge in terms of the presence/absence of a homosexual connotation: Lotti includes it (as meaning 2), Boggione and Casalegno do not, and Ambrogio and Casalegno do not take an explicit position (though all the examples of darlo2 they provide clearly denote heterosexual intercourse; see Sect. 3.3).8

3 First Attestations 3.1

Darla

The earliest attestation of darla provided by Battaglia (1961– IV, 35), given in (5), is from Pietro Aretino’s Dialogo nel quale la Nanna insegna alla Pippa (1536).9 (5) e se pure gliene [glie ‘to.them’-ne ‘of it’] dai talvolta, fatti pagare innanzi (Dialogo: 193, 18–20) ‘and if you ever give some to them sometimes, make them pay you first’

The sentence in (5) features darne (i.e., ‘give’ plus the partitive clitic pronoun ne ‘of it’), not darla; however, Sei giornate (see fn. 8) includes numerous instances of ne standing for the third person singular feminine direct object pronoun la, independently of its referent (i.e. even for referents other than ‘female sexual organ’, as shown in (6):

Galli de’ Paratesi (1969: 42) reports the expressions quello che lo dà (lit. ‘the one who gives it’) and quello che lo prende (lit. ‘the one who takes it), glossed respectively as ‘active pederast’ and ‘passive pederast’, and states that lo “sta per ano” ‘lo stands for ‘anus’’; in a footnote, however, she states that the two expressions are ambiguous since the pronoun “secondo alcuni sta per ano, secondo altri per pene” ‘according to some [lo] stands for ‘anus’, according to others for ‘penis’’(original italics) (fn. 5, p. 42). 9 The edition of Aretino’s work I use as a reference is Sei giornate. Ragionamento della Nanna e della Antonia (1534). Dialogo nel quale la Nanna insegna a la Pippa (1536), edited by Giovanni Aquilecchia. Bari: Laterza, 1969. 8

236

C. Russi

(6) ella ricolse la lettera …, la lesse e rilesse mille volte; … la mostrò a colui che mi mandò a portargnele (Dialogo: 320, 34 – 321, 1) ‘she picked up the letter …, read and reread it a thousand times; … showed it to the one who sent me to give it to her’

In any case, as shown in (7), darla itself is indeed attested in these works by Aretino, which makes us wonder why Battaglia quotes darne instead. (7) Il fatto sta nel fracasso che ne fate voi altri uomini da sarti, e non in quello che ne facciamo noi donne da scacchi, che per nonnulla la diamo e ridiamo, e per nonnulla la togliamo e ritogliamo (Dialogo: 261, 6–9) ‘The point is the mess you dummies make, not in what we fickle women make of it, who for nothing give it again and again, and for nothing take and retake it away’

The second example of darla provided by Battaglia is from the mid-twentieth century, precisely from Cesare Pavese’s novel Il mestiere di vivere: (8) Tutto si riduce alla sacramentale astuzia della fidanzata che non deve ‘dargliela’ al morso, altrimenti lui la pianta (Pavese 1952: 8, 109) ‘It all comes down to the fiancée’s sacramental cunning, who must not give it to him to bite, or he’ll dump her’

The earliest attestation of darla reported by Boggione and Casalegno (2000: 150–151; 2004 165–166) are also in the form of darne and are both from the sixteenth century: one from Aretino and one from Benedetto Varchi’s La suocera (1569) (9)10: (9) cotesta è una fantaccia sudicia, sporca, spilorcia, che è come la pila dell’acqua: pensa quando ella ne dà a te! (Varchi 1569: III, 4, 22) ‘that one is an awful maid filthy, unkempt and miserly; she’s like the holy water font: just imagine when she gives it to you!’

De Mauro (1999–2000: II, 469) does not provide examples of darla; he only gives the date of its first attestation, 1536, which he takes from Battaglia (1961–),11

Essere come la pila dell’acquasanta ‘to be like the holy water font’ is a well-known Italian saying meaning ‘something that is touched by everyone’ and, more specifically ‘a woman of loose morals’. 11 Regarding his sources, De Mauro states that when the date of the first attestation of an entry is not followed by examples or comments, it comes from previous works, primarily Battaglia (1961–) and Nicolò Tommaseo Dizionario della lingua italiana (1861–1879); given that Tommaseo’s dictionary does not list darla, the date 1536 must come from Battaglia. 10

On the Varying Fortunes of darla and darlo in Italian

237

while Lotti (1992) and Forconi (1988) do not give dates and their examples seem to be constructed because no source is cited.12 An earlier attestation of darla/darne, however, is found Giovanni Sercambi’s Novelle (beginning of fifteenth century): (10) Ella n’ha che a tutti ne potrà dare (XI, 17, 165) ‘She has enough of it [lit. she has of it] to give some to all of us’

The context of the sentence is the following: a monk is lured by a woman into a storeroom with the promise of sex. Once inside, he finds two fellow monks who had been similarly tricked by the woman (with the intention of exposing them). Upon discovering that his expectations are the same as the other two’s, the monk speaks the line quoted in (10). Such a context indicates that darne here is a variant of darla, just as in (6) and (9), suggesting that the verb was already attested by the early fifteenth century. Actually, an even earlier precursor of darla could be found in the verb darsi ‘give oneself’, which is already well attested by the late thirteenth–early fourteenth century with a meaning quite equivalent to that of darla.13,14 (11) a. Una notte li venne in visione che li parea vedere le dee della scienzia a guisa di belle donne: e stavano al bordello e davansi a chi le volea (Novellino: 78, 307; from OVI) ‘One night he had a vision in which he thought he saw the goddesses of science in the form of beautiful women: and they were in the brothel and gave themselves to whoever wanted them’ b. ove femmina si dà così di lieve, che ivi non può essere amore, ma desidera molto d’usare con uno e con un altro per sodisfare alla sua luxuria (De Amore di Andrea Cappellano volgarizzato: I, 22, 211; from OVI) ‘when a woman gives herself so easily, that there cannot be love, but she much desires to have sex with one man and the other to satisfy her lust’

12 In Lotti (1992: 125) Aretino in parentheses precedes the definition of darla, and the following examples:

(i) a. Non la darebbe neanche al padre eterno ‘She wouldn’t give it to the holy father’ b. La dà via come niente ‘She gives it away like nothing’

Cf. Battaglia’s (1961–: IV, 32) definition of darsi: “acconsentire alle profferte d’amore, alle richieste di matrimonio, abbandonarsi all’essere amato con assoluta dedizione. In particolare ‘concedersi’ (una donna)” ‘to consent to amorous advances, to marriage requests, surrender to the loved one with absolute devotion. In particular, ‘to give oneself [sexually]’ (of women)’. 14 The Opera del vocabolario italiano (OVI) database “contains 1960 vernacular texts (22.3 million words, 456,000 unique forms) the majority of which are dated prior to 1375, the year of Boccaccio's death. The verse and prose works include early masters of Italian literature like Dante, Petrarch, and Boccaccio, as well as lesser-known and obscure texts by poets, merchants, and medieval chroniclers” (http://artfl-project.uchicago.edu/content/ovi). 13

238

C. Russi

Darsi with women as subjects is indeed robustly attested in OVI in contexts referring unmistakably to sexual intercourse and characterized by an immoral connotation and censorious overtones (whereas darsi with men as subjects is practically absent), and it has endured in the language with the same meaning until today.15

3.2

Darlo1

As date of the first attestation of darlo1 ‘lend oneself to sodomy’, De Mauro (1999– 2000: II, 470) gives 1536, which is the date of Aretino’s Dialogo. However, again he gives no examples, leading us to assume that, as in the case of darla, he is following Battaglia (1961–) (see fn. 11). Yet, as noted above, darlo is not found in Battaglia (1961–), and no occurrences are found in Aquilecchia’s edition of Aretino’s work nor in the Bàrberi Squarotti’s edition (1988). De Mauro’s date, thus, appears disputable. As for Lotti (1992), he gives no date of first attestation for darlo via and provides one example (Sono travestiti, gente che lo dà via ‘They are transvestites, people who give it away’, p. 126) for which no source is mentioned, which suggests that it is constructed by himself. Moreover, the screening of a sample of erotic/obscene literary classics from the fifteenth to the seventeenth century revealed no occurrences of darlo1 (nor of darlo2); this would suggest that darlo was not widespread in older stages of Italian and, since it appears overall unattested in modern/contemporary Italian as well (as we will see in Sect. 4), it makes one wonder about the viability of its characterization as comune ‘common’ (see (3a) above).16

3.3

Darlo2

The earliest attestation of darlo2 ‘of men, sexually possess a woman’ reported in dictionaries dates from 1978 (12) and is found in Ambrogio and Casalegno (2004: 139), who give four more examples from the 1990s:

15

Boggione and Casalegno (2000: 683) list darsi in their Indice delle voci per ambito semantico ‘Index of voices by semantic field’ under Prostituzione, disponibilità sessuale ‘Prostitution, sexual availability’ and provide several examples from renowned modern Italian writers, such as Gabriele D’annunzio, Alberto Bevilacqua, Pier Paolo Pasolini. 16 It may be worth noting that darlo1 is unattested even in classic texts where sex in general or more specifically homosexuality and sodomy are at center stage and talked about in quite explicit terms, such as Aretino’s Sonetti sopra i XVI modi’ (1526), Antonio Vignali’s La Cazzaria (1525–1527), Francesco Berni’s Rime (early sixteenth century), or Antonio Rocco’s L’Alcibiabe fanciullo a scola (1630), to name a few.

On the Varying Fortunes of darla and darlo in Italian

239

(12) Oggi ho chiavato come un dio, te lo darò tutto e ancora un pezzo se mi avanza (Garelli 1978: 145) ‘Today I fucked like a god, I will give it to you, all of it and another piece, if I have some left’

Boggione and Casalegno (2000: 152; 2004: 166) provide two examples of darlo2, both from the contemporary post-feminist writer Rossana Campo’s literary debut In principrio erano le mutande, one of which is the following (found also in Ambrogio and Casalegno 2004: 139): (13) mi fa l’elenco dettagliato delle posizioni e dei bellissimi romantici modi di come lui gliel’ha dato (Campo 1992: 15) ‘she makes me a detailed list of the positions and the most romantic ways he gave it to her’

Finally, once again, Lotti (1992) gives no date of first attestation and gives only one constructed example (Secondo me aspetta che tu glielo dia ‘I think she’s waiting for you to give it to her’, p. 125). In summary, while tracing the emergence of darla and darlo2 is overall straightforward, tracking the first attestations of darlo1 seems to be quite challenging. Also, darla has been attested in Italian for over five hundred years (if not longer), whereas darlo2 appears to be a fairly recent addition.17

4 Darla and Darlo in Contemporary Italian 4.1

Darla

Darla is undoubtedly well-established among Italian speakers (independently of age and socio-cultural background) and is widely attested in the language, as evidenced by abundant examples from both recognized literary works (14) (see also (8)) and non-literary texts (15) from CORIS/CODIS (see fn. 2):

The ‘young age’ of darlo2 is also evidenced by the fact that it is not included in Boggione and Casalegno (1996) which is an earlier version of their two later dictionaries. 17

240

C. Russi

(14) a. sempre in mezzo a fìmmine che appena lo vedevano gliela davano subito (Camilleri 1987: 101) ‘always among women who, as soon as they saw him, would give it to him’ b. Vanno con tutti; la danno a chiunque (Bevilacqua 1999: 68) ‘They go with everyone; they give it to anyone’ c. “E gliela dai?” Erica si mette in piedi e si stiracchia. “Che domande fai? No. Di solito no. Però lui insiste e così ogni tanto, alla fine … Gliela do” (Ammaniti 2000: 30) ‘“And do you give it to him?” Erica stands up and stretches “What kind of question is that? No. Usually I don’t. But he insists and so once in a while … I end up giving it to him”’ d. Tutto avrebbe voluto tranne vedere l’amplesso fra Gigi er cesso e l’amica nigeriana. “Te la danno, Gigi?”. “Pagando …” (Manzini 2016: 117) ‘The last thing he wanted to see was that dog Gigi his Nigerian girlfriend in a sexual embrace. “Are you getting laid, Gigi?”. “If I pay …”’

(15) a. Drim ti mette a disposizione ragazze che te la danno subito (STAMPAPeriodici) ‘Drim will get you girls who’ll give it to you right away’ b. Le donne si dividono in tre categorie: le puttane, le stronze e le rompiballe. La puttana la dà a tutti. La stronza la dà a tutti meno che a te. La rompiballe la dà a te, solo a te, sempre a te (MISCVolumi) ‘Women can be divided into three categories: sluts, bitches and ballbusting nags. Sluts give it to everybody. Bitches give it to everybody but you. Ballbusting nags give it to you, only you, always you’ c. era stato esiliato da Caligola a causa della relazione che aveva avuto con la sorella dell’imperatore, Giulia Livilla (una che la dava un po’ a tutti) (PRACCVolumi) ‘he had been exiled by Caligula because of the relationship he had had with the emperor’s sister, Julia Lavilla (who practically gave it to everybody)’

A remarkably enlightening manifestation of the prominent status (and impact) of darla in contemporary Italian is the saying in (16), a pun based on the name of the celebrated songwriter Lucio Dalla.18 (16) Dalla. Non è un cantante ma un consiglio ‘Dalla. It’s not a singer but a piece of advice’

The earliest reference to this phrase that I was able to uncover, comes from a short article on the opening concert for the sixth anti-AIDS campaign in Rome, published in the newspaper Repubblica on December 2, 1999.19 The article mentions a sixteen-year-old boy, Raffaello, who wears a t-shirt featuring the aphorism quoted above and explains it saying, “Ma è anche importante lasciarsi andare con

18

Many thanks to Giuseppe Lagani for pointing out this to me. http://ricerca.repubblica.it/repubblica/archivio/repubblica/1999/12/02/musica-anti-aids.html?ref= search.

19

On the Varying Fortunes of darla and darlo in Italian

241

Fig. 1 ‘Dalla’ t-shirt for sale on Amazon.com (https:// www.amazon.it/T-shirtuomo-Dalla-cantanteconsiglio/dp/B00W56RX5M? ie=UTF8&*Version*= 1&*entries*=0)

la giusta leggerezza e ironia” ‘But it’s also important to let oneself go with the right amount of lightness and irony’. Four years later, the controversial journalist and anchor-woman Alda D’Eusanio caused quite an uproar when she wore a t-shirt with this phrase on her TV show broadcasted on RAI, Italy’s national public broadcasting company. Currently, various other merchandise displaying the phrase is available for sale by street vendors and online, and the phrase is included in the book Non se ne può più (2012) by journalist and writer Stefano Bartezzaghi under the discussion of “tormentoni linguistici” ‘linguistic fads’ (i.e. words/expressions used thoughtlessly because they are perceived as fashionable, to the point of humorous, ‘cool’, etc.) (Fig. 1).20

4.2

Darlo

In contemporary Italian, darlo is far less visible than darla. Darlo1 is practically unattested and darlo2 holds an overall marginal status.21 The only example I was 20

Worthy of note is also the Facebook page https://www.facebook.com/-Dalla-non-%C3%A8-uncantantema-un-consiglio-120653937990249/, featuring the photo of the worldwide famed porn star Rocco Siffredi. 21 Note that neither darlo1 nor darlo2 are found in CORIS/CODIS.

242

C. Russi

able to find is from Andrea Camilleri’s novel Un filo di fumo (1997 [1980]), where it occurs next to darla: (17) Totuzzo, […], aveva quasi distintamente detto: “gioia mia”. A distanza di due anni, adesso articolava con sicurezza: “dammelo”, “dammela” (Camilleri 1997: 51) ‘Totuzzo, […], had almost distinctly said: “precious”. Now, after two years, he could articulate: “give it to me”, “give it to me’”

In (17) darlo undoubtedly refers to sodomitic, though heterosexual, intercourse since the referent of its subject is Helke, the wife of one of the main characters, who has been entertaining a lively ‘romantic’ relationship with Totuzzo, a deaf-dumb young man. Further evidence for the reduced prominence of darlo is given by the fact that native speakers (of any age, level of education and social class) do not recognize darlo1, and adult speakers recognize darlo2 only if they know writer/novelist Rossana Campo’s works where both darla and darlo2 occur regularly, the latter also in the form darne, as shown in (18): (18) lo vedo che è cambiato e già scopavamo poco prima e adesso non me ne dà quasi più (Campo 2010: 110) ‘I can see that he has changed and we already didn’t fuck much before and now he almost never gives it to me’

A Google search conducted in May 2016 yielded two attestations of darlo2 from two songs by two Italian hip-hop/rap singers and songwriters, J-Ax (19a) and Sfera Ebbasta (19b). (19) a. La ragazza mi guarda in tasca. Io non te lo do/ La tua malia su di me si è esaurita. Io non te lo do/Sono diverso dagli uomini là fuori. Io non te lo do (Io non te lo do, 2009) ‘The girl looks at my pocket. I won’t give it to you/Your spell on me is broken. I won’t give it to you/ I’m different from the men out there. I won’t give it to you’ b. Basta che chiedi, baby, vieni e te lo do/So cosa cerchi, baby, vieni e te lo do/Ho tutto quello che vuoi, vieni e te lo do/Non preoccuparti, baby, vieni e te lo do/So che lo vuoi, io te lo do (Te lo do, 2015) ‘You just have to ask, baby, come here and I’ll give it to you/I know what you’re looking for, baby, come here and I’ll give it to you/I have I’ve got all you want, come and I’ll give it to you/Don’t worry, baby, come and I’ll give it to you/I know you want it, I’ll give it to you’

The meaning of darlo in (19b) may be ambiguous because, technically, two overt referents of lo are identifiable in cosa cerchi ‘what you’re looking for’ and tutto quello che vuoi ‘all you want’, yet the hint at the sexual meaning is certainly plausible. In (19a), on the other hand, the meaning of darlo is rather unequivocal, given the absence of an overt referent of lo and the subject of the song: a young man is addressing a former girlfriend who has become ‘easy’, saying that he will not ‘give it’ to her because he has understood what true love is and has found a ‘good’ (i.e. respectable) and beautiful girl (besides, the girl is looking at his pocket).

On the Varying Fortunes of darla and darlo in Italian

243

It can thus be concluded that in present-day Italian darlo2 is robustly represented in Campo’s work and fairly established in the lexicon of younger speakers.22 Nonetheless, it seems unquestionable that, as for degree of visibility/prominence and distribution as well as in terms of register, a noticeable asymmetry distinguishes the two verbs.

4.3

A Case of ‘Sexist’ Asymmetry

I would like to argue that the ‘concrete’ asymmetry that pertains to darla and darlo stems from a semantic connotative asymmetry which manifests itself on three different levels. The first two emerge clearly from the dictionary definitions in terms of (a) the verbs used in the definitions and (b) the linguistic register darla and darlo2 are associated with. Concerning (a), we observe the use of verb concedersi ‘to give oneself’ (< Latin CONCEDĔRE ‘to retreat, withdraw, give up, concede’) for darla vis-à-vis the use of possedere ‘to possess, own’ (< Latin POSSIDĒRE, a compound of the root of potis ‘one who possesses, an owner; one who has power’ and 23 SEDERE ‘to sit, stand’) for darlo2. Given that the verb concedersi ‘to give oneself’ entails the notions of giving or allowing something as a favor, agreeing to someone’s desire’, darla and darlo2 are asymmetrical with respect to the connotation of dominance within the domain of sexual relationships: when men lo danno ‘give it (to women)’, women are possessed by men, but when women la danno ‘give it (to men)’, men are not possessed by women; on the contrary, women are still possessed by men. Regarding (b), the asymmetry is evinced by De Mauro’s differentiation in terms of register since darla is marked as slang and darlo1 as vulgar, which suggests that overall, darla is proper, whereas darlo1 is relegated to a lower, more offensive and inappropriate or indecent domain.24 Thus, from a socio-cultural perspective, that is, in terms of shared views of conventional behaviors and customs, what darla refers

22

I found the following occurrence of darlo in Aldo Busi’s Sodomie in corpo 11 (1988: 255): arrivederci, Busi, e buon soggiorno in Kenya e mi raccomando: sia casto, e soprattutto eviti di andare con le bianche! Non lo dia in bocca al lupo! ‘goodbye Busi, and have a nice stay in Kenya and remember: be chaste and above all avoid going with white women! Don’t give it in the wolf’s jaws!’. The sentence eviti di andare con le bianche ‘avoid going with white women’ signals a HETEROSEXUAL context, indicating that this is an instance of darlo2. 23 Note that Vocabolario Treccani defines concedersi as “darsi, consentire a un rapporto sessuale” ‘to give oneself, to consent to sex’, and illustrates with the sentence “si è concessa al suo amante” ‘she gave herself to her lover’ (http://www.treccani.it/vocabolario/concedere/; my emphasis, CR). 24 The wide and varied distribution of darla would indicate that the label gergale ‘slang’ is accurate only in the extended sense of ‘allusive, indirect’, not in the main meaning of ‘characteristic of a specific social group’.

244

C. Russi

to (women engaging easily in ‘orthodox’ heterosexual activity) is less contemptible than what darlo1 refers to (‘deviant’ sodomitic sexual practices). The third level of the asymmetry involves once again darla and darlo2 and is manifested by the fact that darlo2 does not entail a negative connotation of its subject while darla does. This indicates that darlo2 has not become fully equivalent to darla, not even in Campo’s works, as shown by the comparison between the sentences with darlo in (20) and those with darla (21), which are all uttered by women. (20) a. Lei dice, Guarda quello ha un modo di darmelo che mai nessuno in vita mia me l’ha dato così (Campo 1992: 25) ‘She says, Look he has a way of giving it to me that nobody else in my life has ever given it to me that way’ b. lei lì che si sbava e lui che non gliel’ha mai dato! (Campo 1995: 34) ‘she’s there drooling and he’s never given it to her!’ c. A questo punto mi sembra di essere venuta al mondo per questo: per farfugliare parole insensate, per inarcare la schiena, muovere i fianchi e farmelo dare da questo russo di Pigalle (Campo 2007: 54) ‘At this point I feel like I came into this world for this: to mumble meaningless words, arch my back, move my hips and let this Russian from Pigalle give it to me’

(21) a. un uomo si scoccia se la vai a dare in giro, se non sei solo per lui (Campo 1995: 90) ‘a man gets annoyed if you give it around, if you are you’re not for him only’ b. Lui mi ha detto che fra di voi era già finita da un po’, che non scopavate mai, che gliela davi una volta ogni due tre mesi (Campo 1995: 109) ‘He told me that it’d been over for a while between you, that you never screwed, that you gave it to him once every two three months’

It is clear from the examples in (20) that darlo2 does not connote men as sexually ‘loose’ or ‘easy’; on the contrary, it portrays men positively by connoting their sexual prowess. Moreover, I would argue that darlo2 carries the same negative connotation as darla (and perhaps it even reinforces it) in that it still hints at women’s loose sexual habits; this is evinced by the fact that the contexts of (20a)– (20c) clearly highlight the importance of sexual relationships for the female characters involved (e.g., in (20b) through the verb sbavarsi ‘to drool’). This is also illustrated (and perhaps even more clearly) by (22), where darne replaces darlo2 due to the quantified contexts and co-occurs with darla. (22) Eh, però te ne ho dato proprio tanto, vero? Te ne ho dato un bel po’, così quando sei a Parigi non vai a darla subito a qualche negrone o qualche francesino (Campo 2014: 42) ‘Well, I really gave you a lot of it, didn’t I? I gave it to you quite a bit, so that when you are in Paris you don’t go right away to give it to some black dude or some Frenchie’

On the Varying Fortunes of darla and darlo in Italian

245

By overtly referring to a large quantity, darne tanto ‘give a lot of it’, darne un bel po’ ‘give quite a bit of it’ underscore women’s sexual ‘insatiability’, just like darla subito ‘give it right away’. In conclusion, through their never-ending chats and cogitations about love and intimate relationships, Campo’s women may be “reasserting their fundamental independence from [men]” (Lombardi 1997: 39), it appears that concerning their sexual lives, they remain anchored to conventional views of women and the traditional construction of femininity whereby women are encouraged to cede agency and give into men.

5 Socio-cultural Foundations of the Asymmetry Linguistic asymmetries pertaining to the representation of women and men reflect women’s and men’s perceived values and status in society; they are sexist because they are centered on the notion “that woman is a sexual being dependent on men” (Pauwels 2003: 553). The asymmetry between darla and darlo represents an instance of linguistic sexism because the moral discriminatory connotations these verbs inherently carry devalue women and embody a prejudiced stance toward gender and certain types of sexual practice: when it comes to their sex life, women relent because they are innately weak, submissive, powerless, easily conquered by men’s charm and physical strength (as well as intrinsically immoral and loose, and sexually greedy), whereas men possess since they are strong, assertive and dominant. Furthermore, ‘normal’ heterosexual intercourse is not vulgar, while sodomy is. The development and persistence of the asymmetry between darla and darlo can be linked to enduring dominant socio-cultural beliefs and deep-rooted prejudicial views of women and ‘normative’ (heterosexual) versus ‘deviant’ sexual behaviors (homosexuality and sodomy). Misogyny and homophobia have permeated male dominated Western societies and cultures since antiquity but they differ crucially in that the former carries no stigma, since it is socially and culturally accepted even among women, whereas the latter does. And it is precisely this difference, I suggest, that accounts for the asymmetry between these verbs. The impact and incidence of misogyny on (Western) societies is strongly evidenced by proverbs, which express truth based on common sense or experience and describe basic rules of conduct. Proverbs about women abound across languages and cultures that consistently reflect a marginalized, trivialized and demeaning view of women: “[l]a donna è certamente il tema più trattato dai proverbi. […] Questo ci fa capire il peso che riveste nella visione popolare del mondo, dove però la sua straordinaria importanza e il suo potere sono arginati da diffidenze, avvertimenti, analisi spietate, malignità, spesso vere cattiverie” ‘women certainly represent the most widespread subject of proverbs. […] This underscores the weight women carry in the popular vision of the world, where, however, their extraordinary importance and power stem from distrust, warnings, ruthless scrutiny, spite, and

246

C. Russi

often genuine malice’ (Lapucci 2006: 343; cf. also Nuessel 2000; Boggione 2005, among others). Women are greedy (“Il gioco, il letto, la donna e il fuoco, non si contentan mai di poco” ‘Gambling, bed, women and fire, are never easily satisfied’; Bellonzi 1968: 9), dangerous and wicked (“La donna è la rovina dell’uomo” ‘Women are men’s ruin’, “Una buona donna è sempre una cattiva bestia” ‘A good woman is a bad beast nonetheless’; Lapucci 2006: 350) and must please men (“La donna deve essere brava in casa, seria in strada, santa in chiesa e matta a letto” ‘Women must be good at home, serious in the street, saintly in church and wild in bed’; Lapucci 2006: 346). Though it may seem that Italian women have attained equality, they actually continue to hold an inferior position within the family, in the work place and in public life in general (Krause 2003: 341–342; Pasquino 2011: 183–185), and are sexualized and objectified considerably, both privately and publicly. The sexualization and objectification of women that pervade Italian society is clearly attested in the mass media and is cleverly illustrated by Beppe Servegnini in his La bella figura. A field guide to the Italian mind (2006).25 In the chapter Television, where the Semi-Undressed Signorina acquires a cloak of significance, Severgnini writes “The new Italian icon is the Semi-Undressed Signorina. We ought to put her on coins and stamps. Her face is interchangeable, but from the neck down everything stays the same. She turns up in every TV show program, wiggles her hips, and every so often gets to speak, especially when she has nothing to say” (77). He then continues by remarking that Italian television advertising has made sex its “mechanism of choice”, since either explicitly or allusively sexual references are employed in advertisements for basically any kind of product (78). Quite revealing for our purposes is how Severgnini explains Italian women’s acceptance of the sexualization of which they are victims, namely that they tolerate it “out of habit, resignation, and a lack of awareness. Thirty years ago, feminists complained if anyone reminded them they were women. Today women watch programs with scantily clad Barbie Doll lookalikes, bemoaning the fact that they cannot look like them.” (79) The pervasiveness of sexism in contemporary Italy is, perhaps, even more tangibly attested by episodes involving high-profile public figures, particularly politicians, such as Silvio Berlusconi’s bunga parties or Beppe Grillo’s Facebook post about Laura Boldrini (for the latter and other episodes of sexism on the web see Frau and Guzzetta 2015). Women’s subordinate role in the sphere of sexual life and sexuality persists widely in the Western world (hence Italy), where “it is the socially sanctioned role of boys to encourage girls to let them sample the fruit, and the role of girls to resist the ‘harvest’” (Daniluk 1998: 108) and, although girls are expected to show some resistance not to appear overly ‘easy’ or ‘loose’, excessive resistance is also a problem since “[g]irls who do not go all the way may be viewed as ‘frigid’ or ‘cock teases’” (Daniluk 1998: 109). Moreover, “[n]ormative heterosexuality and the

25

Beppe Severgnini is an acclaimed journalist, essayist and columnist who has written extensively on social and cultural aspects of contemporary Italy.

On the Varying Fortunes of darla and darlo in Italian

247

normative heterosexual act define young people in relation to each other: him as actor, her as acted upon; his agency, her subordination; her body for his pleasure. His body for his pleasure too” (Holland et al. 2003: 87). In sum, in the domain of sexual ‘conduct’, women are still seen as easily lured; men’s ‘romantic’, enticing manners are aimed at making women fall. Women are well aware of this, but accept it so as to conform to sanctioned socio-cultural norms. Therefore, I would propose that misogyny has played a decisive role in the diffusion and entrenchment of darla in Italian, on the one hand, and promoted the emergence of darlo2 on the other. Let’s now address the issue of the (highly) negligible distribution of darlo1 in Italian. The ghost of homophobia has forever haunted Western societies. In medieval Europe, sodomy (which comprised both homo- and heterosexual anal intercourse as well as any type of non-vaginal intercourse) was considered the vilest of sexual sins (Brundage 2013: 43) and was prosecuted until the nineteenth century (think of Oscar Wilde, to give but one example). The stigma associated with homosexuality and sodomy continues to exist in contemporary Western societies, in particular among Christians; as Johansson and Percy (2013: 159) aptly summarize, “[f]rom St. Paul onward, Christians deemed sodomy sinful and, once they gained hegemony over governments, criminal as well” and “[t]he sodomy delusion remained in full vigor until the 1960s and only now is being challenged” (176).26 In (popular) Western cultures, then, sodomy has remained the most ‘deviant’ of sexual acts, and the strong stigma attached to it may have contributed to hinder the diffusion of darlo1 in Italian. Regarding the asymmetry between darla and darlo2, for darlo2 to become truly equivalent to darla (hence, for the asymmetry between the two to be leveled out), the subject of darlo2 would have to acquire the same negative connotation of sexual easiness and promiscuity as the subject of darla. But associating men with sexual easiness or promiscuity is a socially and culturally accepted, ingrained stereotyped view, which does not carry a negative, prejudiced connotation; rather, it is the ‘normative standard’: men are to seduce and conquer, while women are to be seduced and conquered. Of course, it could well be the case that darlo2 will catch up with darla in terms of distribution and frequency of use, but the sexist asymmetry will most likely live on. I would like to conclude this section by drawing attention to some expressions found in Aretino’s works—and also in earlier (fourteenth-century) narrative texts— which are comprised of dare and a feminine nominal direct object and metaphorically/euphemistically denote sexual intercourse.

26

See also Herzog (2011) and Dall’Orto (1988, 2015).

248

C. Russi

(23) a. dare la acqua al molino ‘give water to the mill’, with water and mill respectively denoting ‘semen’ and ‘vagina’ (Sei giornate, 530): Intanto il compagnone […] diede la acqua al molino ‘Meanwhile the old pal […] gave water to the mill’ (Sei giornate, 82, 2–3). b. dare la biada ‘give the fodder’, with fodder denoting ‘semen’: le diè la biada ‘he gave her the fodder’ (Sei giornate, 81, 27–28). c. dare la stretta/strettina ‘give the squeeze/the little squeeze’, glossed as ‘have sex, of men’ (Sei giornate, 577): a qualche fante darà la stretta costui ‘this guy will give a squeeze to some maidservant’ (Sei giornate, 330, 30); mi dava alle volte qualche strettina ‘sometime he gave me a little squeeze’ (Sei giornate, 73, 25–26). d. dare la imbeccata ‘give the imbeccata’ (imbeccata ‘the quantity of food that is fed in a bird’s beak by another bird’), glossed by Aquilecchia as ‘undergo intercourse, of women’ (Sei giornate, 553); mi gitto seco nel letto con intenzione di non dargli la imbeccata ‘I throw myself in bed with him with the intention of not having sex’ (Sei giornate, 108, 21–22). e. dare merenda ‘give a snack’, glossed as ‘concede oneself to the sexual act, of women’ (Sei giornate, 559): e datogli una o due volte merenda ‘and having given him a snack once or twice’ (Sei giornate, 109, 15). f. dare una/la chiavatura ‘nail down (lit. give a/the nailing down’): La Doralice a un medico promesse dargli una chiavatura ‘Doralice promised a doctor to give him a screwing’ (Sonetti lussuriosi e pasquinate: 137).

The expressions in (23) differ with respect to the gender of the subject they allow. On the one hand, (23a)–(23c) restrict their subject to men, whereas in (23d) and (23e) the subject must be a woman. On the other hand, the phrase in (23f), even though in the example given (and in the other two found in the text) the subject is a woman, can also have a man as subject since chiavatura and its modern counterpart chiavata derive from the verb chiavare ‘possess someone sexually, have sexual intercourse’, whose subject is unrestricted in terms of gender. All the expressions in (23)—none of which survives in contemporary Italian—represent conceivable/ potential sources for darla, that is, plausible collocations from which darla could have originated. What is relevant to our purposes is that (23a)–(23c) could have given rise to a different meaning of darla, namely ‘of men, to have sexual relations with someone, especially in a loose manner’; the fact that they did not provides empirical linguistic evidence attesting to the import of socio-cultural factors in the shaping of the lexicon.

On the Varying Fortunes of darla and darlo in Italian

249

6 Conclusion This chapter has addressed the varying fortunes the verbs darla ‘of women, to have sexual relations with someone, especially in a loose manner’, and darlo1 ‘to lend oneself to sodomy’, and darlo2 ‘of men, to sexually possess a woman’ have endured in the Italian language. I have shown that while darla has been attested since (at least) the sixteenth century, darlo2 made its appearance in the language only in the second half of the twentieth century and tracing the emergence of darlo1 is quite problematic. Moreover, darla appears to be well established in the lexicon of Italian speakers independently of age, education and social class and exhibits a robust distribution and high usage frequency. In contrast, darlo1 is overall unattested and unknown, and darlo2, although attested, remains marginally known and restricted to younger speakers. More importantly, I have addressed the questions of how to account for (a) the insignificant presence (if not absence) of darlo1 in Italian through time and (b) the limited distribution of darlo2 in contemporary Italian. After reviewing the definitions of darla and darlo provided in Italian dictionaries (both general and specialized) and calling attention to some inconsistencies in the definitions of darlo, I have discussed the earliest attestations of these two verbs on the basis of data from both the dictionaries considered and from additional literary sources. The unequal status of darla and darlo in modern Italian has then been assessed and it has been attributed to stereotypical socio-cultural attitudes and prejudice toward women and homosexuality/sodomy. I have proposed that the varying fortunes of darla and darlo embody an instance of linguistic sexism in that they reflect a lexical asymmetry in terms of gender (women vs. men), traditional views of gender roles (inferior, submissive women vs. superior, dominant men) and sexual practices (‘normal’ heterosexual vs. ‘abnormal’ homosexual); that is, the moral discriminatory connotations inherently carried by these verbs reflect a demeaning view of women and a biased intolerant attitude toward gender and ‘unorthodox’ sexual practices. The emergence and persistence of the asymmetry between darla and darlo has been related to still prevailing socio-cultural beliefs and entrenched prejudicial views of women and ‘normative’ (heterosexual) vs. ‘deviant’ sexual behaviors (homosexuality and sodomy). By examining the different fortunes of darla and darlo, the present study has provided evidence in support of the claim that socio-cultural attitudes and deep-rooted stereotypical representations and views about gender and sexuality can impact on language use and language change. However, the present study has only scratched the surface. The most pressing questions to address are un-clouding the mystery surrounding the ‘fate’ of darlo1 and properly assess the status of darlo2 in contemporary Italian. To answer these questions, the careful analysis of a comprehensive corpus of erotic literature from (at least) the fifteenth century to the present is needed, as well as an in-depth scrutiny of queer and ‘urban’ literature and culture to assess if the sexist asymmetry between darla and darlo can be better (or differently) defined if analyzed against these scenarios.

250

C. Russi

Acknowledgements I would like to express my gratitude to Mikołaj Nkollo who, albeit indirectly, suggested to me the topic of this study. Parts of this chapter were presented at the Gender and Language roundtable, the annual meeting of the American Association for Italian Studies, Baton Rouge, April 21, 2016 and the text of the presentation appears in http://www. gendersexualityitaly.com. I thank the organizers for giving me the opportunity to share my ideas and for their helpful feedback. I am also deeply grateful to Adria Frizzi, Chiara McCarty and Case McCarty, and Laura Seccia for assisting me in the English translation of contemporary Italian texts and/or providing insightful comments, and to David Birdsong and Chiyo Nishida for reading the final draft. Any errors or misunderstandings remain of course my own.

References Ammaniti, Niccolò. 2000. Ti prendo e ti porto via. Milano: Mondadori. Anonymous. 1970. II Novellino, ed. Guido Favati. Genova: Bozzi. Anonymous. 1980. De Amore di Andrea Cappellano volgarizzato, ed. Graziano Ruffini. Milano: Guanda. Ambrogio, Renzo, and Giovanni Casalegno. 2004. Scrostati gaggio! Dizionario storico dei linguaggi giovanili. Torino: UTET. Arbasino, Alberto. 1980. Un paese senza. Milano: Garzanti. Aretino, Pietro. 1969. Sei giornate. Ragionamento della Nanna e della Antonia (1534). Dialogo nel quale la Nanna insegna a la Pippa (1536), ed. Giovanni Aquilecchia. Bari: Laterza. Aretino, Pietro. 1980. Sonetti lussuriosi e pasquinate. Roma: New Compton. Aretino, Pietro. 1988. Ragionamento e Dialogo, ed. Giorgio Bàrberi Squarotti. Milano: Rizzoli. Battaglia, Salvatore. 1961–. Grande dizionario della lingua italiana. Torino: UTET. Bellonzi, Fortunato. 1968. Proverbi toscani. Milano: Aldo martello Editore. Berni, Francesco. 1985. Rime, ed. Danilo Romei. Milano: Mursia. http://www.liberliber.it/ mediateca/libri/b/berni/rime/pdf/rime_p.pdf. Bevilacqua, Alberto. 1994. L’eros. Milano: Mondadori. Boggione, Valter. 2005. Chi dice donna … 3587 proverbi sull’amore, il matrimonio, il tradimento, la gelosia. Torino: UTET. Boggione, Valter, and Giovanni Casalegno. 1996. Dizionario storico del lessico erotico italiano: metafore, eufemismi, oscenità, doppi sensi, parole dotte e parole basse in otto secoli di letteratura italiana. Milano: TEA. Boggione, Valter, and Giovanni Casalegno. 2000. Dizionario letterario del lessico amoroso: metafore, eufemismi, trivialismi. Torino: UTET. Boggione, Valter, and Giovanni Casalegno. 2004. Dizionario del lessico erotico. Torino: UTET. Brundage, James A. 2013. Sex and canon law. In Handbook Medieval sexuality, ed. Vern L. Bullough and James A. Brundage, 33–50. Hoboken: Taylor and Francis. Busi, Also. 1988. Sodomie in corpo 11. Milano: Mondadori. Camilleri, Andrea. 1987. Il corso delle cose. Palermo: Sellerio. Camilleri, Andrea. 1997. Un filo di fumo. Palermo: Sellerio. Campo, Rossana. 1992. In principio erano le mutande. Milano: Feltrinelli. Campo, Rossana. 1993. Il pieno di super. Milano: Feltrinelli. Campo, Rossana. 1995. Mai sentita così bene. Milano: Feltrinelli. Campo, Rossana. 2007. Più forte di me. Milano: Feltrinelli. Campo, Rossana. 2010. Lezioni di arabo. Milano: Feltrinelli. Campo, Rossana. 2014. Fare l’amore. Milano: Ponte alle Grazie. Dall’Orto, Giovanni. 1988. La ‘tolleranza repressiva’ dell’omosessualità. Quando un atteggiamento legale diviene tradizione. http://www.giovannidallorto.com/saggistoria/tollera/tollera. html.

On the Varying Fortunes of darla and darlo in Italian

251

Dall’Orto, Giovanni. 2015. Tutta un’altra storia. L’omosessualità dall’antichità al secondo dopoguerra. Milano: Il Saggiatore. Daniluk, Judith C. 1998. Women sexuality across life span. Challenging myths, creating meanings. New York/London: The Guildford Press. De Mauro, Tullio. 1999–2000. Grande dizionario italiano dell’uso. Torino: UTET. Forconi, Augusta. 1988. La mala lingua. Dizionario dello ‘slang’ italiano: i termini e le espressioni gergali, popolari, colloquiali. Milano: SugarCo. Frau, Ombretta, and Juliet Guzzetta. 2015. “Il blog è mio e lo gestisco io:” Dominio di genere nel web italiano. gender/sexuality/Italy 2. http://www.gendersexualityitaly.com. Galli de’ Paratesi, Nora. 1969. Le brutte parole. Semantica dell’eufemismo. Milano: Mondadori. Garelli, Cesare. 1978. Il linguaggio murale. Milano: Garzanti. Herzog, Dagmar. 2011. Sexuality in Europe. A twentieth century history. Cambridge: Cambridge University Press. Holland, Janet, Caroline Romazanoglu, Rachel Thomson, and Sue Sharpe. 2003. When bodies come together: Power, control and desire. In Sexuality and society. A reader, ed. Jeffrey Weeks, Janet Holland, and Matthew Waites, 84–94. Cambridge, UK/Malden, MA: Blackwell, 2003. Il vocabolario Treccani. 1997. Roma: Istituto dell’Enciclopedia Italiana. Johansson, Warren, and William A. Percy. 2013. Homosexuality. In Handbook of Medieval sexuality, ed. Vern L. Bullough and James A. Brundage, 155–189. Hoboken: Taylor and Francis. Krause, Elizabeth L. 2003. Italy. The Greenwood encyclopedia of women’s issues worldwide. Vol. 3, Europe, ed. Lynn Walter, 341–372. Westport, CT: Greenwood Press. Lapucci, Carlo. 2006. Dizionario dei proverbi italiani. Firenze: Le Monnier. Lombardi, Giancarlo. 1997. Campo, Rossana (1963–). In The feminist encyclopedia of Italian literature, ed. Rinaldina Russel, 39–40. Westport, CT: Greenwood Press. Lotti, Gianfranco. 1992. Le parole della gente. Dizionario dell’italiano gergale dalle voci burlesche medievali ai linguaggi contemporanei dei giovani. Milano: Mondadori. Manzini, Antonio. 2016. 7–7–2007. Palermo: Sellerio. Nuessel, Frank. 2000. The depiction of women in Italian proverbs. Romance Languages Annuals XI: 288–294. Pasquino, Monica. 2011. Femminismo e femminismi dagli anni Ottanta al XXI secolo. Identità e differenze. Introduzione agli studi delle donne e di genere, ed. Maria Serena Sapegno, 179– 209. Milano: Mondadori. Pauwels, Anne. 2003. Linguistic sexism and feminine linguistic activism. In The handbook of language and gender, ed. Janet Holmes and Miriam Meyerhoff, 550–570. Malden, MA/ Oxford, UK: Blackwell. Rocco, Antonio. 2003. L’Alcibiade fanciullo a scola, ed. Laura Coci. Roma: Salerno Editrice. Sercambi, Giovanni. 1995. Novelle, ed. Giovanni Sinicropi. Firenze: Le Lettere. Severgnini, Beppe. 2006. La bella figura. A field guide to the Italian mind. Translated by Giles Watson. New York: Broadway Books. Tommaseo, Nicolò. 1861–1879. Dizionario delle lingua italiana. Torino: UTET. http://www. dizionario.org/. Toscan, Jean. 1981. Le carnaval du langage: le lexique érotique des poètes de l’équivoque de Burchiello à Marino (XVe-XVIIe siècles). Lille: Atelier. Varchi, Benedetto. 1858. La suocera. Triste: Lloyd Austriaco. Veniero, Lorenzo. ca. 1531. La Zaffetta. http://www.classicitaliani.it/cinquecento/Veniero_La% 20Zaffetta.htm. Vignali, Antonio. 1984. La Cazzaria, ed. Pasquale Stoppelli. Roma: Edizioni dell’Elefante.

Conclusion Julia Herschensohn

This collection of chapters reinforces the importance of the Romance languages as an area of investigation that can provide new empirical data while advancing linguistic theory. The chapters cover a range of languages, theoretical issues, and methodological advances. Several chapters that flesh out the syntactic Left Periphery using the relatively recent cartographic approach are complemented by corpus based studies that use substantial data to investigate theoretical issues. Despite the particular focus of each chapter, they also intersect in important ways. What makes this volume unique is that by approaching current theoretical topics from vastly different perspectives, the chapters inform our understanding of language in general. Altogether, these chapters both shed light on particular syntax and particular languages, and build our general understanding of the Left Periphery, its functional projections, features and syntax-semantics interface. A cartographic approach to the Left Periphery (LP) of the clause and the noun phrase is explored in several chapters. Zagona and Contreras solve an apparent syntactic dilemma, the temporal interpretation of preverbal gerundive adjectives, by elaborating the syntactic and featural make-up of the LP of the Spanish noun phrase. The DP is often compared to the CP, but its functional projections have been less elaborated, so this clarification is welcome both for Romance languages, but more generally for the cartographic enterprise. Authier and Haegeman show similarities and differences among several Romance languages with respect to Focus and Topic Phrases located in the clausal LP; they also demonstrate similarities with wh-exclamatives and link focus to distinct prosody. Labelle’s treatment of negation begins in Old French with a negative operator in the LP, and depends on shifting features on the negative elements, ne and pas over the centuries. Donaldson contrasts two preverbal positions—Spec of Focus and a position far left of Focus—of preposed subordinate J. Herschensohn (&) Department of Linguistics, University of Washington, Seattle, WA, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 D. L. Arteaga (ed.), Contributions of Romance Languages to Current Linguistic Theory, Studies in Natural Language and Linguistic Theory 95, https://doi.org/10.1007/978-3-030-11006-2_12

253

254

J. Herschensohn

clauses that are fronted in verb second Old French. Arteaga’s discussion of obviation considers the features and links between the LP of the matrix and subordinate subjunctive clause in OF. Cross-linguistic diachronic and synchronic variation are the realm of several of the chapters, some of which use corpora to explore theoretical questions that they address. Jacobs traces lenition of intervocalic stops (including voicing of voiceless stops and spirantization) in both diachronic (French) and cross-linguistic (Corsican, Gran Canarian) contexts and from Latin through Old French, using them as a test for different theoretical frameworks. On the one hand, his essay hinges on theoretical questions of feeding order of rules, but on the other, he is able to bring the theory back to phonetic basics in tying his “best solution” to perceptual optimality. Donaldson uses his corpora of 6000 tokens from 30 Old French texts to evaluate the hypothesis that it is semantic force (concession, anteriority, or cause) that determines the preverbal position of preposed subordinate clauses in Old French. His carefully examined evidence shows that semantics (in the left periphery or the core of the declarative) is not at all in play in any of the adverbial clauses. Labelle, in a study of original texts from Old French, Middle French, Modern Continental French and Québécois spanning a millennium, demonstrates the variants of negation. She argues that Jespersen’s Cycle is spiral-like, given the fact that French negation does not return to its original form, but rather results in two variants on either side of the Atlantic. She also gives an account of the featural changes motivating the diachronic evolution from simple negation (ne) to double (ne…pas) and back to simple (pas). Arteaga discusses obviation in Old French, contrasting it with the same phenomenon in Greek and the Balkan languages and adducing theoretical arguments to support her views. She argues that morphological change by itself cannot account for the emergence of obviation in French, instead tying it to feature valuation, proposing that subjunctive que in OF (but not in MF) values the uninterpretable feature ExclF on T. Bullock and Toribio explore code switching in French and Spanish, advocating the use of corpora and computational approaches to document the range of variation. Gess discusses debuccalization of fricatives in Canadian French, using data from his corpus, as well as other corpora, to argue against a view of lenition as perceptually facilitative on empirical and theoretical grounds. Finally, Russi documents the lack of sociocultural change in contrasting the persistence of bias in the usage of feminine darla against masculine darlo over 500 years by examining dictionary entries and original sources through several centuries.