Copyright by Melissa Dae Murphy 2008
The Dissertation Committee for Melissa Dae Murphy certifies that this is the approved version of the following dissertation:
THE ROLE OF TYPOLOGICAL DRIFT IN THE DEVELOPMENT OF THE ROMANCE SUBJUNCTIVE: A STUDY IN WORD-ORDER CHANGE, GRAMMATICALIZATION AND SYNTHESIS
Committee: ____________________________________ Brigitte L. M. Bauer, Supervisor ____________________________________ Frederick Hensey, Supervisor ____________________________________ Chiyo Nishida ____________________________________ Knud Lambrecht ____________________________________ Máximo R. Salaberry
THE ROLE OF TYPOLOGICAL DRIFT IN THE DEVELOPMENT OF THE ROMANCE SUBJUNCTIVE: A STUDY IN WORD-ORDER CHANGE, GRAMMATICALIZATION AND SYNTHESIS
by Melissa Dae Murphy, B.A.; M.A.
Dissertation Presented to the Faculty of the Graduate School of the University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy
The University of Texas at Austin August 2008
For my family
Acknowledgements
I consider myself extremely fortunate to have been influenced by a wide range of academic experiences and would like to thank the people who, over the years, have contributed in some way to the achievement of this goal. First and foremost, I would like to express my deepest gratitude to Brigitte L.M. Bauer. As my dissertation supervisor, she continuously provided thoughtful comments and suggestions, yet always encouraged me to work independently, developing my own hypotheses and reaching my own conclusions. I truly believe this process has been much more meaningful as a result. Brigitte has been incredibly generous with her time, both while in the Netherlands and while in the United States. I am extremely indebted to her for her guidance and support. I would also like to thank my co-supervisor, Frederick Hensey, for his energetic enthusiasm toward my somewhat unconventional topic; it is greatly appreciated. Many thanks also go to Knud Lambrecht, for his attention to detail and constant willingness to provide valuable feedback, and to Chiyo Nishida and Rafael Salaberry, who offered perspectives that were extremely useful during the revision of my final manuscript. Finally, I would like to acknowledge several professors who have made an impact on me over the years. Special thanks go to: Jean-Pierre Montreuil and Carlos Solé, for their advice and encouragement during the early stages of my
v
doctoral program and candidacy, John B. Jensen, from Florida International University, who was my link to the world of linguistics while I was pursuing my M.A. in Spanish literature, and Robert Trammell, my first linguistics professor at Florida Atlantic University, without whom I might not have embarked on this journey.
vi
THE ROLE OF TYPOLOGICAL DRIFT IN THE DEVELOPMENT OF THE ROMANCE SUBJUNCTIVE: A STUDY IN WORD-ORDER CHANGE, GRAMMATICALIZATION AND SYNTHESIS
Publication No.__________
Melissa Dae Murphy, Ph.D. The University of Texas at Austin, 2008 Supervisors: Brigitte L. M. Bauer and Frederick Hensey
In spite of the vast amount of research on mood in Romance languages, certain fundamental issues are clearly underrepresented, particularly in the field of diachronic linguistics. With this in mind, the primary goal of this dissertation is to provide a comprehensive explanation for developments in Romance mood distribution. Unlike the majority of existing research, this approach does not analyze mood in isolation, nor does it look outward for language-external explanations. Instead, changes in mood usage are related to major typological developments via several interconnected analyses which rely heavily on data from Latin and medieval Spanish and French. This investigation, which takes as its starting point the well-attested typological shift from OV to VO word order, addresses four major issues. The
vii
first of these is branching congruency, whereby post-posed subordinate clauses are more closely associated with explicit subordinating conjunctions.
This
hypothesis is tested via a quantitative analysis of Latin data, in order to establish a link between conjunctions and VO word order.
The development of these
subordinating elements is then analyzed within the grammaticalization framework, which provides insight into the nature of specific Latin and Romance forms, in addition to demonstrating the usefulness of certain theoretical notions. The outcome of this process is a highly generalized Romance subordinator, which is argued to have undergone partial synthesis with the subjunctive, evidenced by an increase in both obligatoriness and contiguity. Finally, these cumulative changes in the linguistic system are shown to have had substantial destabilizing effects on the existing subjunctive / indicative contrast. The significance of this claim is that, already in Latin, mood selection is characterized by a loss of motivation and an increase in automaticity. As a result, subsequent changes in mood distribution in Romance languages are not viewed merely as reductive phenomena, but rather as signs of the refunctionalization of a destabilized, yet viable, paradigm.
viii
Table of Contents LIST OF TABLES AND ILLUSTRATIONS...................................................... xiii ABBREVIATIONS FOR GRAMMATICAL TERMS......................................... xv ABBREVIATIONS FOR LANGUAGES............................................................ xvi ABBREVIATIONS FOR TEXTS CITED.......................................................... xvii CHAPTER 1: THE ROMANCE SUBJUNCTIVE 1.1. Introduction........................................................................................... 1 1.2. Mood and Modality...............................................................................2 1.3. The Latin Subjunctive........................................................................... 6 1.4. Romance “Mood Loss” Phenomenon................................................. 13 1.4.1. Reduction in Subjunctive Morphology................................ 14 1.4.2. Reduction in Subjunctive Usage.......................................... 16 1.5. The Present Study............................................................................... 23 1.5.1. Hypotheses........................................................................... 23 1.5.2. Methodology........................................................................ 25 CHAPTER 2: REVIEW OF THE LITERATURE 2.1. Introduction......................................................................................... 27 2.2. Synchronic Research on the Romance Subjunctive............................28 2.2.1. Syntactic Analyses............................................................... 29 2.2.2. Semantic Analyses............................................................... 38 2.2.3. Pragmatic Analyses.............................................................. 54 2.2.4. Contributions of Synchronic Research................................ 62 2.3. Diachronic Research on the Romance Subjunctive............................ 64 2.3.1. Language-Specific Approaches........................................... 65 2.3.1.1. The Role of Phonology......................................... 66 2.3.1.2. The Role of Language Contact............................. 68 2.3.1.3. The Role of Semantic Shift................................... 72 2.3.2. Pan-Romance Approaches................................................... 76 ix
2.3.2.1. The Role of Tense Prominence............................. 78 2.3.2.2. The Role of Hypotaxis.......................................... 82 2.3.2.3. The Role of Word-Order Change......................... 85 2.3.3. Contributions of Diachronic Research................................. 87 CHAPTER 3: TYPOLOGICAL DRIFT AND LATIN CONJUNCTIONS 3.1. Introduction......................................................................................... 88 3.2. Latin Word Order................................................................................ 88 3.3. The Role of Branching........................................................................ 93 3.4. Increase in the Frequency of Latin Conjunctions............................... 98 3.4.1. Word Order and Subordination Techniques........................ 99 3.4.2. Analysis of Latin Texts...................................................... 101 3.4.2.1. Senatus Consultum de Bacchanalibus................ 106 3.4.2.2. De Bello Gallico................................................. 109 3.4.2.3. Peregrinatio Egeriae.......................................... 113 3.5. Observations and Conclusion........................................................... 117 CHAPTER 4: GRAMMATICALIZATION OF SUBORDINATING ELEMENTS 4.1. Introduction....................................................................................... 121 4.2. The Latin Conjunction System......................................................... 122 4.3. Overview of the Grammaticalization Framework ........................... 127 4.4. Analysis of Latin and Romance Subordinating Elements................ 135 4.4.1. Generalization ................................................................... 136 4.4.1.1. Cum..................................................................... 138 4.4.1.2. Quando................................................................ 140 4.4.1.3. Quoniam.............................................................. 141 4.4.1.4. Quia.....................................................................143 4.4.1.5. Quod.................................................................... 145 4.4.1.6. Que...................................................................... 149 x
4.4.2. Decategorialization............................................................ 153 4.4.3. Specialization..................................................................... 161 4.4.3.1. Latin Specialization: Quod................................. 163 4.4.3.2. Romance Specialization: Que............................. 165 4.4.4. Phonetic Erosion................................................................ 168 4.4.5. Layering and Divergence................................................... 170 4.4.6. Renewal..............................................................................173 4.5. Cross-Linguistic Support.................................................................. 178 4.6. Conclusion........................................................................................ 179 CHAPTER 5: SYNTHESIS OF NON-CONTIGUOUS ELEMENTS 5.1. Introduction....................................................................................... 180 5.2. The Analytic / Synthetic Dichotomy................................................ 181 5.3. Schwegler’s (1990) Analyticity and Syntheticity.............................. 185 5.4. A New Perspective on Synthesis...................................................... 191 5.4.1. Synthesis as a Semantic Process........................................ 192 5.4.2. Synthesis as a Unidirectional Process................................ 193 5.4.3. Synthesis versus Grammaticalization................................ 195 5.4.4. Signs of Synthesis: Conditions and Criteria...................... 197 5.5. Que and Subjunctive: Synthesis of Non-Contiguous Elements........ 199 5.5.1. Semantic Synthesis............................................................ 200 5.5.2. Obligatoriness.................................................................... 204 5.5.2.1. Use of que as Attraction to Subjunctive............. 204 5.5.2.2. Use of Subjunctive as Attraction to que............. 211 5.5.3. Displaceability, Linearity and Separability........................215 5.6. Conclusion........................................................................................ 218 CHAPTER 6: DESTABILIZATION AND REFUNCTIONALIZATION OF ROMANCE MOOD 6.1. Introduction....................................................................................... 220 xi
6.2. Destabilizing Effects of Typological Drift....................................... 221 6.2.1. Shift to Right-Branching ................................................... 221 6.2.2. Grammaticalization of Conjunctions ................................ 233 6.2.3. Synthesis of que and Subjunctive...................................... 236 6.3. Refunctionalization of Romance Mood Contrast............................. 238 6.3.1. Stable (Prototypical) Subjunctive...................................... 239 6.3.2. Unstable Subjunctive......................................................... 243 6.3.2.1. With Verbs of Saying and Verbs of Opinion...... 244 6.3.2.2. In Result Clauses.................................................247 6.3.2.3. In Causal Clauses................................................ 251 6.3.2.4. In Adversative Clauses....................................... 255 6.3.2.5. With Expressions of Emotion............................. 260 6.3.3. General Observations......................................................... 267 6.4. Conclusion........................................................................................ 270 CHAPTER 7: CONCLUSION AND IMPLICATIONS 7.1. Background....................................................................................... 274 7.2. Recapitulation of Hypotheses and Findings..................................... 276 7.2.1. Hypothesis #1.....................................................................277 7.2.2. Hypothesis #2.....................................................................280 7.2.3. Hypothesis #3.....................................................................283 7.2.4. Hypothesis #4.....................................................................284 7.3. Concluding Remarks......................................................................... 287 7.3.1. General Conclusions regarding Romance Mood............... 288 7.3.2. Implications for Diachronic Romance Linguistics............ 290 BIBLIOGRAPHY................................................................................................ 293 VITA.................................................................................................................... 306
xii
LIST OF TABLES AND ILLUSTRATIONS
Table 1: Classical Latin Subjunctive Paradigms................................................... 14 Table 2: Modern Spanish Subjunctive Paradigms................................................. 15 Table 3: Modern French Subjunctive Paradigms...................................................15 Table 4: Syncretism in Modern French Indicative and Subjunctive Paradigms.... 16 Table 5: Latin Subjunctive Uses in Main Clauses................................................. 17 Table 6: Latin Subjunctive Uses in Subordinate Clauses...................................... 19 Table 7: Terrell and Hooper’s (1974) Semantic Classification System................ 39 Table 8: Summary of Synchronic Research on Romance Mood........................... 63 Table 9: Typological Drift and Branching Patterns............................................... 97 Table 10: Subordination Patterns in Senatus Consultum de Bacchanalibus....... 109 Table 11: Subordination Patterns in De Bello Gallico.........................................112 Table 12: Subordination Patterns in Peregrinatio Egeriae..................................116 Table 13: L-B and R-B Subordinate Structures................................................... 119 Table 14: Latin Conjunction System................................................................... 123 Table 15: C. Lehmann’s (1995) Grammaticalization Criteria............................. 133 Table 16: Grammaticalization Cline for Subordinating Elements....................... 154 Table 17: Latin Relative Pronouns...................................................................... 157 Table 18: Types of Grammaticalization.............................................................. 160 Table 19: Candidates for Specialization among Subordinating Elements........... 162 Table 20: Specialization of Quod in Merovingian Gaul...................................... 164 Table 21: Layering and Divergence of Que......................................................... 172 Table 22: Renewal of Romance Generalized Subordinator................................. 173 Table 23: Renewal of Que in Medieval Spanish Texts........................................175 Table 24: Modern Spanish and French Subordinating Conjunctions.................. 177 Table 25: Criteria for Synthesis........................................................................... 198 Table 26: Modern Spanish Indicative System..................................................... 225 xiii
Table 27: Modern French Indicative System....................................................... 225 Table 28: Latin Conjunctions and Mood in De Bello Gallico............................. 233 Table 29: Latin Adversative Conjunctions and Corresponding Moods...............256 Table 30: AcI versus Quod/Quia with Verbs of Emotion................................... 261 Table 31: Subordinate Structures from Latin to Modern Romance..................... 268 Table 32: Mood Distribution from Latin to Modern Romance........................... 268 Table 33: Summary of Subordination Patterns in Latin Data.............................. 279 Table 34: L-B and R-B Subordinate Structures (Table 13)................................. 279 Table 35: Grammaticalization Cline for Subordinating Elements (Table 16)..... 281
Illustration 1: Analytic / Synthetic Continuum.................................................... 188 Illustration 2: Scope of Subjunctive Mood.......................................................... 266 Illustration 3: Pendular Development of Mood Motivation................................ 269 Illustration 4: Pendular Development of Mood Motivation (Illustration 3)........ 287
xiv
ABBREVIATIONS FOR GRAMMATICAL TERMS
ABL
ablative
NMNL
nominalizer
ACC
accusative
NOM
nominative
ART
article
NON-PAST
non-past
AUX
auxiliary
OBL
oblique
COMP
complementizer
PART
participle
COND
conditional
PASS
passive
CONJ
conjunction
PAST
past
CP
clitic pronoun
PERF
perfective
DAT
dative
PTV
partitive
DEF
definite
PL
plural
DEM
demonstrative
PLUPERF
pluperfect
F
feminine
POSS
possessive
FUT
future
PP
past participle
GEN
genitive
PREP
preposition
GER
gerund
PRES
present
GERV
gerundive
PRET
preterit
IMP
imperative
PRON
pronoun
IMPERF
imperfect
PTC
particle
IND
indicative
REFL
reflexive
INDEF
indefinite
REL
relative
INF
infinitive
SG
singular
LOC
locative
SUB
subordinator
M
masculine
SUBJ
subjunctive
N
neuter
TOP
topic
NEG
negative
xv
ABBREVIATIONS FOR LANGUAGES
Eg.
Egyptian
Eng.
English
FFr.
Future French
Fr.
French
FSp.
Future Spanish
Heb.
Hebrew
It.
Italian
Jp.
Japanese
Lt.
Latin
Lu.
Luiseño (Uto-Aztecan)
OFr.
Old French
OSp.
Old Spanish
PIE
Proto-Indo-European
Po.
Portuguese
Sp.
Spanish
Ved.
Vedic Sanskrit
xvi
ABBREVIATIONS FOR TEXTS CITED Alex.
La vie de Saint Alexis
Avell.
Collectio Avellana
Ayala, Pal.
Rimado de palacio (Pero López de Ayala)
Caes., D.B.C.
De Bello Civili (Caesar)
Caes., D.B.G.
De Bello Gallico (Caesar)
Cato, R.R.
De Re Rustica (Cato)
Celest.
La Celestina
Ch.O., Poésies
Poésies (Charles d’Orléans)
Chrét., Chev.
Le chevalier de la charrete (Chrétien de Troyes)
Cid
Cantar de mío Cid
Conq.
Conqueste de Constantinople (Robert de Clari)
Egip.
Vida de Santa María Egipciaqua
Eul.
La cantilène de Sainte Eulalie
Firm., Math.
Matheseos Libri Octo (Julius Firmicus Maternus)
GdT, H.F.
Historia Francorum (Gregory of Tours)
Hita, L.B.A.
El libro de buen amor (Arcipreste de Hita)
Miracles
Miracles de Nostre Dame
Nîmes
Charroi de Nîmes
Ov., Pont.
Epistulae ex Ponto (Ovid)
Paix
Livre de la paix (Christine de Pisan)
Passion
Passion du Palatinus
Pereg.
Peregrinatio Egeriae
Petr., Sat.
Satyricon (Petronius)
Piramus
Piramus et Tisbé
Pl., Amph.
Amphitruo (Plautus)
Pl., Asin.
Asinaria (Plautus) xvii
Pl., Bacch.
Bacchides (Plautus)
Pl., Capt.
Captivi (Plautus)
Pl., Cist.
Cistellaria (Plautus)
Pl., Curc.
Curculio (Plautus)
Pl., Men.
Menaechmi (Plautus)
Pl., Pseud.
Pseudolus (Plautus)
Pl., Stich.
Stichus (Plautus)
Quat. Liv. R.
Quatre livres des rois
Quinze
Quinze joies de mariage
Rol.
La chanson de Roland
Scaev., Dig.
Digesta (Q. Cervidius Scaevola)
S.C. Bacch.
Senatus Consultum de Bacchanalibus
Sen., N.Q.
Naturales Quaestiones (Seneca)
Tab.
Lex Duodecim Tabularum
Talav., Corv.
Corvacho (Arcipreste de Talavera)
Vill., Test.
Grand testament (François Villon)
Thèbes
Roman de Thèbes
Troyana
Historia troyana en prosa y verso
Vulg. Joan.
Evangelium Joannis (Biblia Vulgatae Editionis)
xviii
CHAPTER 1 THE ROMANCE SUBJUNCTIVE 1.1. Introduction Research in the field of mood and modality, particularly with respect to Romance languages, has resulted in an unquestionably impressive body of work. Nevertheless, in spite of the abundance of existing research, linguists continue to be intrigued by this aspect of the Romance verbal system, since it offers countless opportunities for both descriptive and theoretical research in various fields, including, but not limited to, syntax, semantics, pragmatics, cognition, language acquisition, language typology and diachronic change. The attention of linguists, however, has centered primarily around a small set of salient research questions, often resulting in the reformulation of previous hypotheses and/or rewording of existing definitions and terminology. Although several linguists have examined more peripheral issues such as tense/mood interaction, sociolinguistic variation, dialectal variation, and diachronic change, work in these areas has been fairly limited and there are ample opportunities for future research. In this dissertation, I will focus specifically on a diachronic aspect of mood in Romance languages which I consider to be severely underrepresented in the field: the role of morpho-syntactic changes related to typological drift in the development of the Romance subjunctive.
1
In this introductory chapter I will: (1) introduce the notions of mood and modality, (2) present an overview of issues related to the subjunctive in Latin, (3) illustrate in detail the Romance “mood loss” phenomenon to be explored in this dissertation, and (4) present my hypotheses along with the methodology I will use to carry out my analysis.
1.2. Mood and Modality The study of verbal categories such as tense, mood and aspect must first overcome difficulties encountered due to the terminology itself. The distinction between modality and mood, in particular, tends to be troublesome, since the terms are often used interchangeably in spite of the fact that they are not synonymous. Modality is, first and foremost, a semantic category that “covers a broad range of semantic nuances—jussive, desiderative, intentive, hypothetical, potential, necessitative, inferential, hortatory, exclamative, etc.—whose common denominator is the addition of an overlay of meaning to the most neutral semantic value of the proposition of an utterance, namely factual and declarative” (Fleischman 1982: 13). Modality can also be divided into two major types: epistemic (knowledge-based) and deontic (action-based).
Epistemic modality
involves notions of speaker commitment, opinion, judgment and evidence, while deontic modality represents either internal or external need or compulsion (Bhat 1999: 63). Furthermore, unlike the category of tense, modal notions relate not to 2
the verb itself, but more loosely to the sentence as a whole (Palmer 1986: 2). Without a doubt, this wider—and often unspecifiable—scope associated with modality contributes to its complex nature. Since modality is a sentence-level feature, its surface expression is not restricted to the verb phrase and, as a result, we find a substantial amount of cross-linguistic variation.
According to Palmer (1986), there are three
fundamental ways in which modality may be expressed: (1) modal verbs, (2) mood, and (3) clitics and particles. These three techniques, in order, are illustrated below: (1-1)
Eng.
Must I come? (Palmer 1986: 33)
(1-2)
Lt.
Naviget! sail-3SG-PRES-SUBJ
‘let him sail’ (Palmer 1986: 39) (1-3)
Lu.
noo
xu
n
po
heyi
I
PTC
1-SG-CP
FUT
dig
‘I should dig’ (Palmer 1986: 44) Of these three means of modal expression, mood is the most widespread in European languages and is typically characterized by a distinct set of verbal paradigms (Palmer 1986: 39).
3
Mood, specifically, is the morphological expression of modality and, as such, is a strictly formal category. Unfortunately, mood is no easier to categorize than modality, as it can be divided into multiple subcategories. The most widely accepted classification divides mood into three types: impersonal/non-temporal (infinitives, participles and gerunds), personal/non-temporal (subjunctives and imperatives), and personal/temporal (indicatives) (Makamina 2003: 134). Not surprisingly, of the three personal—or finite—forms, the indicative is considered “maximally finite” (Palmer 1986: 82), since it typically contains all tense/aspect paradigms, likely due to its unmarked declarative function. Of all possible expressions of modality (i.e. particles, clitics, mood, modal verbs), mood is generally thought to be the most complex. According to Palmer, this is due to its grammatical status, since a natural consequence of grammaticalization is a high level of arbitrariness (1986: 5).
Additionally,
compared to the categories of tense and aspect, Bhat argues that the category of mood is less united, less coherent in its encoding, and more speaker-oriented (1999: 130). All of these characteristics relate to the somewhat elusive nature of this verbal category. The family of Romance languages presently has the following moods: indicative, subjunctive, conditional and imperative. One of these (i.e. conditional) is innovative, while the other three are generally considered to be the result of a
4
gradual conflation of the following four reconstructed Proto-Indo-European moods1 (cf. Szemerényi 1996: 231-265; Bennett 1910: 145-161): 1. Indicative (used to express certainty) 2. Imperative (used to express 2nd person singular affirmative commands) 3. Subjunctive (used to express will/requisition and futurity) 4. Optative (used to express hope/desire and potentiality/possibility) It is evident that the imperative, subjunctive and optative moods are closely related in terms of semantic and/or pragmatic values. Furthermore, this obvious functional overlap appears to have resulted in variation even in very early stages of Indo-European. In his discussion of the trajectories of the imperative, optative and subjunctive, Lehmann argues that “it is scarcely remarkable that the three moods with their comparable meanings are attested in only a few dialects” (1993: 183). Regarding the Romance family in particular, the modern imperative serves the functions of the PIE imperative (and injunctive), while the modern subjunctive corresponds roughly to the PIE subjunctive and optative moods. However, the question of whether or not any or all of these particular functions continue to operate in modern Romance languages is a difficult one to answer.
Some
linguists argue that the subjunctive is now a completely dependent marker of 1
The injunctive (an unaugmented imperfect with modal function) is not included here as a separate PIE mood, although some linguists (in the minority) would do so. The injunctive clearly forms part of the modal system, yet it is not formally independent and is found only in Old Indic (Szemerényi 1996: 263-264).
5
subordination (e.g. Sabanééva 1993), while others argue the contrary. Makamina, for example, refutes such claims of dependency, and states: “le subjonctif n’est pas le mode de la subordination” (Makamina 2003: 139), thereby agreeing with Le Bidois and Le Bidois, who assert that the subjunctive is “indépendant, pleinement autonome, tout à fait spontané” (1967: 501). It becomes clear, even before conducting a closer examination of the literature, that opposing viewpoints are the norm rather than the exception with regard to the subjunctive mood in Romance languages. Although contradictory opinions are to be expected in any study of language, this particular topic is often further complicated by issues including terminology, dialectal variation, and the complex relationship between syntax and semantics.
Additionally, debates
regarding synchronic inconsistencies represent problems which are fundamentally diachronic in nature, as even the Latin subjunctive presents challenges for linguistic research.
1.3. The Latin Subjunctive Since Latin is typically taken as a point of reference in studies on diachronic developments in Romance languages, I will briefly comment on the Latin subjunctive, in an attempt to demonstrate that the problematic nature of Romance mood is not at all recent. To begin with, when one sets out to study the grammar of Latin, one typically encounters overly simplistic explanations of undoubtedly 6
complex grammatical forms and functions. By looking exclusively at traditional grammar books, one might erroneously conclude that the subjunctive can be mastered by simply memorizing a finite number of black-and-white rules. Many monographs present the meaning, forms, and uses of the subjunctive in a misleadingly abridged fashion, relying instead on impeccable organization and presentation of the material to create a semblance of order to what is, in reality, quite disjointed. Wheelock, for example, introduces the subjunctive by stating: “In contrast to the indicative, the mood of actuality and factuality, the subjunctive is in general (though not always) the mood of potential, tentative, hypothetical, ideal, or even unreal action” (2000: 186). He then proceeds to equate the meaning behind the Latin subjunctive to English modal auxiliaries such as may, might, should, would, etc. In addition, Latin grammars typically enumerate subjunctive uses via lists which include concrete examples to illustrate the “rule” (cf. Gildersleeve and Lodge 1997). At first glance, then, the Latin subjunctive appears quite straight forward; however, there are several indications that the matter is much more troublesome than it initially appears. First of all, the sheer quantity of pages devoted to the subjunctive mood is overwhelming; in Bennett (1910), sixty-two pages are dedicated to subjunctive usage in main clauses and one hundred thirty-nine pages are devoted to subjunctive in subordinate clauses. Furthermore, there is often a striking increase 7
in the number of footnotes included in sections on the subjunctive, which are used to explain difficult concepts in more detail, mention ongoing scholarly debates, acknowledge the existence of counter-examples, etc. Finally, the categorization of subjunctive usage appears to be, at times, unmotivated and/or haphazard. Consider the following categories from Bennett’s (1910) table of contents: “subjunctive of compliance,” “subjunctive of helplessness,” “present prohibitive introduced by numquam, nemo, nullus, non, etc.,” “present prohibitive with connecting negatives neve, neu, nive, neque, nec.,” “subjunctive of duty or fitness,” etc. In short, a brief and over-simplified definition of the subjunctive mood is presented in sharp contrast to an extensive inventory of seemingly unrelated functions. This visible contradiction mirrors, in effect, the debates that have existed among scholars for centuries. The complexity of the Latin subjunctive stems not only from the fact that modal categories have subtle semantic nuances, but also from its origins, since the Latin subjunctive is the result of a conflation of both the PIE subjunctive and optative moods.2 In terms of morphology, the following subjunctive paradigms in Latin evolved from PIE subjunctives and optatives:
2
A drastically different viewpoint is held by Touratier, who argues that both the subjunctive and optative continue to exist in Latin and that there are two distinct, but homophonous, morphemes: one to express “volonté” (1977: 370) and another to express “possibilité” (1977: 370).
8
PIE subjunctive forms > 1. Latin regular presents (e.g. amem ‘love’, moneam ‘warn’) 2. Latin imperfects (e.g. amarem ‘love’, essem ‘be’) 3. Latin pluperfects (e.g. amavissem ‘love’, dixissem ‘say’) PIE optative forms > 1. Latin presents in –im (e.g. possim ‘be able’, velim ‘wish’) 2. Latin perfects (e.g. viderim ‘see’, amaverim ‘love’) (Bennett 1910: 146) In terms of function, it is generally believed that the PIE subjunctive was used to express will and futurity and the optative was used to convey wish and potential force (Bennett 1910: 148). These semantic categories, while distinct, are similar enough to cause confusion and complicate an already nuanced system.
As
summarized by Meillet and Vendryes (1953): Bien que les emplois du subjonctif et de l’optatif fussent distincts en indo-européen, ils ont fini par se confondre notamment comme marque de la subordination et le voisinage des sens a entraîné la simplification des modes. Seuls le grec ancien et l’indo-iranien offrent la coexistence du subjonctif et de l’optatif. Dans toutes les autres langues, un seul des deux modes a été conservé. (Meillet et Vendryes 1953: 196) It is no surprise, then, that the academic goal of uncovering a single, underlying meaning for the hybrid, Latin subjunctive paradigm would face many obstacles. The reality of the Latin subjunctive is that, even today, its fundamental meaning remains unclear. For Sonnenschein (1910), the meaning shared by all 9
subjunctive usage is obligation (in a broad sense, including moral obligation, natural obligation, logical obligation, etc.). He remains quite convinced that all usage can be reduced to this meaning and justifies any counter-example as a “stylistic peculiarity,” which “cannot be regarded as inherent in the Latin language itself” (1910: 49). For Hahn (1953), on the other hand, the most basic meaning of the Latin subjunctive is not obligation, but futurity.3 She claims that the “vivid future” was expressed by the PIE subjunctive and came to be expressed by the Latin future tense, while the “remote future” was expressed by the PIE optative and, later, by the Latin subjunctive (Hahn 1953: 149). Unlike those linguists who have offered a unified account of subjunctive usage, Handford (1947) makes an important distinction between subjunctive with modal value (i.e. will, futurity, wish and potentiality) and subjunctive with no modal value; that is to say, a syntactically governed subjunctive. He goes so far as to argue that the subjunctive may appear in contexts where it is “merely a stylistic substitute for an indicative, employed at the dictate of habit or fashion” (1947: 139). The search for the meaning(s) of the Latin subjunctive is further complicated by the existence of surprisingly common non-modal usage, often referred to as “modal attraction.” In this case, we are dealing with verbs that
3
This argument is supported by evidence that the Latin subjunctive and future forms are often interchangeable. Hahn (1953) provides examples in which the subjunctive is used in place of the future and vice versa. Additionally, she argues that both forms appear in similar contexts (e.g. in apodoses of hypothetical sentences following protases with a present indicative) (Hahn 1953: chapter 7).
10
would normally appear in the indicative, but appear instead in the subjunctive mood as a result of assimilation to a verb upon which they are dependent. This is, therefore, a purely formal—rather than semantic—phenomenon.4 Consider the following examples of this assimilatory subjunctive usage: (1-4)
Lt.
nemo
avarus
adhuc
no one-NOM
greedy-NOM-SG
hitherto
inventus
est,
find-PERF-PART-NOM-SG
be-3SG-PRES-IND
cui,
quod
REL-PRON-DAT-SG
REL-PRON-NOM-SG
haberet,
esset
satis
have-3SG-IMPERF-SUBJ
be-3SG-IMPERF-SUBJ
sufficient
‘no miser has yet been found who was satisfied with what he had’ (1-5)
Lt.
cum
diversas
causas
as
different-ACC
reasons-ACC
afferrent,
dum
formam
assert-3PL-IMPERF-SUBJ
while
nature-ACC
sui
quisque
et
animi
REFL-PRON-GEN-SG
each one-NOM
and
mind-GEN
et
ingenii
redderent
and
natural talent-GEN
give back-3PL-IMPERF-SUBJ
‘as they brought forward different arguments, while each mirrored his own individual type of mind and natural bent’
4
Frank (1904) claims that many purported examples of “attraction” are, instead, examples of sematically-motivated (often, “anticipatory”) subjunctive. Bennett, on the other hand, refutes this claim and insists that the phenomenon is purely mechanical (1910: 305-307).
11
(1-6)
Lt.
quod
ego
fatear,
REL-PRON-ACC-SG
I-NOM
admit-1SG-PRES-SUBJ
pudeat? cause shame-3SG-PRES-SUBJ
‘should I be ashamed of a thing which I admit?’ (1-7)
Lt.
mos
est
Athenis
custom-NOM
be-3SG-PRES-IND
Athens-ABL
quotannis
in
contione
laudari
annually
in
assembly-ABL
praise-PRES-PASS-INF
eos
qui
sint
those-ACC
who-NOM-PL
be-3PL-PRES-SUBJ
in
proeliis
interfecti
in
battles-ABL
kill-PERF-PART-NOM-PL
‘it is the custom of Athens every year for those who have been killed in battle to be publicly eulogized’ (Bennett 2000: 212) According to Latin mood usage, the verbs shown in bold above should— according to normative rules—appear in the indicative, yet they are assimilated. Furthermore, this assimilatory subjunctive may be “attracted” not only by subjunctives, but also by commands and infinitives (as seen in the AccusativeInfinitive [AcI] construction in example 1-7). This may be motivated in terms of common features, as the subjunctive and imperative share similar modal force and the subjunctive and infinitive are both commonly used in subordinate structures. Handford (1947) sees this phenomenon as evidence of “the remarkable tendency of Latin to extend the use of the subjunctive to its furthest limits” (148). What is 12
most important is that these debates about usage and examples of non-modal subjunctive underscore the already variable nature of the subjunctive mood in Latin.5
It does not seem surprising, therefore, that this verbal category has
experienced a substantial amount of change over the centuries. This apparent lack of motivation and absolute systematicity in Latin may have quite possibly made the mood vulnerable to changes that might better serve the system as a whole.
1.4. Romance “Mood Loss” Phenomenon In addition to the fundamental differences between the PIE and Latin modal systems (i.e. conflation), there are further reductionist tendencies found in the Romance systems that continue to operate even today. Although many linguists have, indeed, commented on these developments, their work has been primarily descriptive, with little emphasis placed on the goal of uncovering potential causes of the phenomenon. First and foremost, one of the major problems found in the literature is that of ambiguous terminology. The term “mood loss” tends to be ambiguous since it is used to refer to at least two distinct, albeit related, phenomena. The term is often used to refer to (1) a reduction in subjunctive morphology and (2) a reduction in the number of contexts in which the subjunctive appears. 5
Since the mid-twentieth century, there has been a limited amount of comprehensive work on the Latin subjunctive. Instead, scholars have narrowed their focus, presenting analyses of specific clause types, tenses, etc.
13
Below, I will illustrate both types of reduction and will assess the extent to which they are present in both Spanish and French.
1.4.1. Reduction in Subjunctive Morphology It is well known that the subjunctive paradigm in Classical Latin was morphologically robust. There were four subjunctive tenses (present, perfect, imperfect and pluperfect) and there were no examples of syncretism within the subjunctive itself or between subjunctive and indicative forms. The following tables illustrate the subjunctive paradigms of the verb Lt. cantare ‘to sing’ in Classical Latin, and the corresponding forms in Modern Spanish and Modern French, where morphological subjunctive “loss” is evident:
Table 1: Classical Latin Subjunctive Paradigms CLASSICAL LATIN PRESENT cantem cantes cantet cantemus cantetis cantent IMPERFECT cantarem cantares cantaret cantaremus cantaretis cantarent
PERFECT cantaverim cantaveris cantaverit cantaverimus cantaveritis cantaverint PLUPERFECT cantavissem cantavisses cantavisset cantavissemus cantavissetis cantavissent
14
Table 2: Modern Spanish Subjunctive Paradigms MODERN SPANISH PRESENT yo cante tú cantes él cante nosotros cantemos vosotros cantéis ellos canten IMPERFECT yo cantara tú cantaras él cantara nosotros cantáramos vosotros cantarais ellos cantaran
PRESENT PERFECT yo haya cantado tú hayas cantado él haya cantado nosotros hayamos cantado vosotros hayáis cantado ellos hayan cantado PLUPERFECT yo hubiera cantado tú hubieras cantado él hubiera cantado nosotros hubiéramos cantado vosotros hubierais cantado ellos hubieran cantado
Table 3: Modern French Subjunctive Paradigms MODERN FRENCH PRESENT je chante tu chantes il chante nous chantions vous chantiez ils chantent
PAST j’aie chanté tu aies chanté il ait chanté nous ayons chanté vous ayez chanté ils aient chanté
While the subjunctive mood in Spanish remains quite strong, there is syncretism between first and third-person singular forms in all four tenses. French, on the other hand, has undergone two significant changes. Firstly, the imperfect and pluperfect subjunctive tenses—while still present to some extent in literary language—have disappeared completely from contemporary spoken French, resulting in a greatly reduced system (Grevisse 1993: 1271, Soutet 2000: 7).
15
Furthermore, the remaining paradigms are characterized by a substantial amount of phonological syncretism between subjunctive and indicative moods, as shown below: Table 4: Syncretism in Modern French Indicative and Subjunctive Paradigms MODERN FRENCH INDICATIVE SUBJUNCTIVE /šãt/ * /šãt/ * /šãt/ * /šãt/ * /šãt/ * /šãt/ * /šãtõ/ /šãtjõ/ /šãte/ /šãtje/ /šãt/ * /šãt/ * /ešãte/ * /ešãte/ * Past /ašãte/ /ešãte/ /ašãte/ /ešãte/ /avõšãte/ /ejõšãte/ /avešãte/ /eješãte/ /õšãte/ /ešãte/ Note: Transcriptions with an asterisk (*) indicate syncretism between corresponding indicative and subjunctive forms.6 Present
The tables above illustrate quite effectively why linguists—particularly those researching French—often use the term “loss” in reference to these reductive morphological changes.
1.4.2. Reduction in Subjunctive Usage However significant this morphological reduction may be, the most vexing and widely discussed aspect of Romance “mood loss” is the reduction in the number 6
The tendency in Modern French to use on ‘one’ as the marker of first-person plural (at the expense of nous) further increases the level of syncretism in these paradigms.
16
of subjunctive uses or contexts. It is not uncommon for descriptions of major developments in the Romance verbal system to include overly-simplified characterizations (e.g. shift from synthetic to analytic forms, fewer aspectual distinctions, more temporal distinctions, fewer subjunctive uses, etc.).
With
regard to mood distributional patterns, in particular, the precise nature of the reduction in subjunctive usage is typically left unspecified. In an attempt to clarify the exact subjunctive contexts in which there has been a shift to indicative, I have created the following tables based on descriptions presented in several well-respected Latin grammars (i.e. Bennett 1910, Handford 1947, Gildersleeve and Lodge 1997, and Wheelock 2000). Listed are the various uses (both in main and subordinate clauses) for the subjunctive in Latin, along with the corresponding mood(s) in modern Spanish and French.7
Table 5: Latin Subjunctive Uses in Main Clauses
Latin Subjunctive Uses in Main Clauses
Sp.
Fr.
SUBJ
SUBJ
1. Optative (expressions of wishing) Lt. Sp. Fr.
di istaec prohibeant [SUBJ] que los dioses lo impidan [SUBJ] que les dieux l’empêchent [SUBJ] ‘may the gods prevent that!’
7
The examples provided for each category in the following tables are for illustrative purposes and are not intended to be exhaustive.
17
2. Volitive: Jussive (command) dicant [SUBJ] que lo digan [SUBJ] qu’ils le disent [SUBJ] ‘let them tell’ 3. Volitive: Hortatory (1st person plural) Lt. Sp. Fr.
amemus [SUBJ] patriam amemos [SUBJ] nuestro país aimons [IMP] notre pays ‘let us love our country’ 4. Volitive: Prohibitive (prohibition) Lt. Sp. Fr.
ne repugnetis [SUBJ] no se resistan [SUBJ] ne résistez [IMP] pas ‘do not resist!’ 5. Volitive: Deliberative (questions, exclamations) Lt. Sp. Fr.
SUBJ
SUBJ
SUBJ
SUBJ / IMP
SUBJ
IMP
quid faciam? [SUBJ] ¿qué debería [COND] hacer? que’est-ce que je ferai [IND]? ‘What shall I do?’ 6. Volitive: Concessive (something granted)
IND / IND / INF / INF / COND COND
sit [SUBJ] hoc verum concedo que eso es [IND] verdad j’accepte que ce soit [SUBJ] vrai ‘I grant that this is true’ 7. Potential: “may” (a mere possibility)
IND / SUBJ
Lt. Sp. Fr.
Lt. Sp. Fr.
Lt. Sp. Fr.
dicat [SUBJ] aliquis alguien podría [COND] decir quelqu’un pourrait [COND] dire ‘some one may say’
18
IND / SUBJ
IND / IND / SUBJ / SUBJ / COND COND
8. Potential: “would” (understood condition) dies deficiat [SUBJ] el tiempo fallaría [COND] le temps manquerait [COND] ‘time would fail’ 9. Potential: “could” (perceiving, thinking) Lt. Sp. Fr.
crederes [SUBJ] uno podría [COND] creer on pourrait [COND] croire ‘one could believe’ 10. Potential: “should/would/could have” (irrealis) Lt. Sp. Fr.
Lt. Sp. Fr.
fuissem [SUBJ] philosophus habría sido [COND] filósofo j’aurais été [COND] philosophe ‘I would have been a philosopher’
COND COND
COND COND
COND COND
Table 6: Latin Subjunctive Uses in Subordinate Clauses
Latin Subjunctive Uses in Subordinate Clauses
Sp.
Fr.
1. Final (purpose) clauses (‘in order that’, ‘lest’) ut omnes intellegant [SUBJ] dico lo digo para que todos entiendan [SUBJ] je le dis pour que tout le monde le comprenne [SUBJ] ‘I say it so that everyone understands’ 2. Consecutive (result) clauses (‘that [not]’, ‘so that’) Lt. Sp. Fr.
Lt. Sp. Fr.
SUBJ / SUBJ / INF INF
Siciliam ita vastavit ut restitui in antiquum statum non possit [SUBJ] destrozó Sicilia de tal manera que no puede IND / [IND] ser restaurada a su condición original SUBJ il a tant ravagé la Sicile qu’on ne peut [SUBJ] pas la restaurer à ce qu’elle était avant. ‘he so ravaged Sicily that it cannot be restored to its former condition’ 19
IND / SUBJ
3. “Anticipatory” temporal clauses dum litteras veniant [SUBJ], morabor esperaré a que llegue [SUBJ] la carta j’attend que la lettre arrive [SUBJ] ‘I shall wait until the letter comes/for the letter to come’ 4. Protasis of “ideal” conditional clauses Lt. Sp. Fr.
si hoc dicas [SUBJ], erres si dijeras [SUBJ] esto, estarías equivocado si vouz disiez [IND] cela, vous vous tromperiez ‘if you should say this, you’d be mistaken’ 5. Protasis of “unreal” conditional clauses Lt. Sp. Fr.
sapientia non expeteretur, si nihil efficeret [SUBJ] Sp. la sabiduría no sería buscada si no consiguiera [SUBJ] nada Fr. on ne chercherait pas la sagesse, si elle ne comptait [IND] pour rien ‘wisdom would not be sought, if it accomplished nothing’ 6. Comparison conditional clause (‘as if’)
IND / IND / SUBJ / SUBJ / INF INF
SUBJ
IND
SUBJ
IND
SUBJ
IND
Lt.
Lt. Sp. Fr.
serviam tibi tam quasi emeris [SUBJ] me argento te serviré como si me hubieras comprado [SUBJ] por dinero je te servirai, comme si tu m’avais acheté [IND] ‘I will serve you as though you had bought me for money’
20
7. Proviso clauses (‘provided that’) nubant, dum ne dos fiat [SUBJ] comes que se casen, siempre que no haya [SUBJ] dote Fr. qu’ils se marient, pourvu qu’aucune dot ne soit [SUBJ] donnée ‘let them marry, provided that no dowry goes with it’ 8. Adversative & Concessive clauses (‘although’) Lt. Sp.
licet omnes terrores impendeant [SUBJ], succurram Sp. aunque todos los miedos me acechen [SUBJ], prestaré ayuda Fr. quoique toute terreur me menace [SUBJ], je vous aiderai ‘although all terrors hang over me, I will lend aid’ 9. Clauses of characteristic
SUBJ
SUBJ
IND / SUBJ
IND / SUBJ
IND / SUBJ
IND / SUBJ
IND
IND
Lt.
multa sunt, quae mentem acuant [SUBJ] hay muchas cosas que agudizan [IND] el ingenio Fr. il y a beaucoup de choses qui rendent [IND] plus sage ‘there are many things which sharpen the wits’ 10. Causal clauses Lt. Sp.
Lt. Sp. Fr.
cum sis [SUBJ] mortalis, quae mortalia sunt cura ya que eres [IND] mortal, preocúpate por lo mortal puisque tu es [IND] mortel, occupe-toi de ce qui est mortel ‘since you are mortal, care for what is mortal’
21
11. Oratio Obliqua (indirect discourse) Paetus omnes libros quos pater suus relinquisset [SUBJ] mihi donavit Sp. Paetus me dio todos los libros que su padre le IND / IND / había dejado [IND] COND COND Fr. Paetus m’a donné tous les livres que son père lui aurait donné [COND] ‘Paetus gave me all the books which (as he said) his father had left’ 12. Circumstantial “cum” clauses Lt.
cum ver appetat [SUBJ], ex hibernis movendum est Sp. como se acerca [IND] la primavera, debemos trasladarnos de nuestro campamento de invierno Fr. comme le printemps approche [IND], il faut quitter notre campement d’hiver ‘as Spring is approaching, we must move out of our winter-quarters’ 13. Indirect questions Lt.
Lt. Sp. Fr.
dic mihi ubi fueris [SUBJ], quid feceris [SUBJ] dime dónde estuviste [IND], qué hiciste [IND] dis-moi où tu as été [IND] et ce que tu as fait [IND] ‘tell me where you were, what you did’
IND
IND
IND
IND
This summary illustrates the fact that the verbs in approximately half of all subjunctive contexts in Latin have been replaced by non-subjunctive forms in Spanish and French. It is precisely this phenomenon that has motivated the present study, in which I will attempt to provide a multi-faceted, typological explanation for this seemingly reductionist trend.
22
1.5. The Present Study In this dissertation, I will present and analyze three major changes in Latin and Romance morpho-syntax which, together, I believe to be the fundamental cause of the apparent reduction in subjunctive contexts. The analysis presented here will differ from the existing subjunctive literature in several ways. It will differ substantially from synchronic research, in that its primary aim is not to define or categorize the subjunctive mood. It will also differ from studies on languagecontact, by focusing exclusively on language-internal developments. Finally, it will differ from the majority of diachronic analyses by arguing for a complex, multi-layer explanation for changes in Romance mood distribution, and by substantiating these claims both on theoretical grounds and via the analysis of authentic Latin and Romance data. This dissertation will, in short, address the motivation(s) behind the phenomenon in a way that I believe best reflects its true nature as a gradual typological shift involving several different, yet interconnected, processes.
1.5.1. Hypotheses My initial research on Romance mood loss was based primarily on the role of reductive morphology, particularly with regard to French.
I soon became
convinced, however, that the changes taking place were rooted not in morphology, but in syntax, and that an in-depth study involving issues in Latin 23
subordination techniques would be necessary. What is absolutely fundamental is that the work presented here will not begin with the assumption that there was a significant increase in subordination (or language complexity) that took place in Latin. Compare the older Accusative-Infinitive clause in (1-8) with the more modern finite subordinate clause in (1-9): (1-8)
Lt.
patrem
advenisse
scio
father-ACC
come-PERF-INF
know-1SG-PRES-IND
‘I know that my father has come’ (Herman 2000: 88) (1-9)
Lt.
scio
iam
filius
quod
know-1SG-PRES-IND
already
son-NOM
that
amet
meus
istanc
meretricem
love-3SG-PRES-SUBJ
my-NOM
this-ACC
prostitute-ACC
‘I already know that my son loves this prostitute’ (Pl., Asin. 52-53 in Herman 2000: 88) The major development shown above does not represent an increase in subordination but rather a shift in subordination technique (i.e. the presence of an explicit subordinating conjunction). Following a review of the literature (in chapter 2), the following four interrelated hypotheses will be tested: 1)
Due to the tendency toward branching congruency, the typological shift from left-branching to right-branching structures led to an increase in subordinating conjunctions in Latin (chapter 3). 24
2)
This increase in frequency accelerated the development of subordinating conjunctions in Latin, a process which can be analyzed effectively within the grammaticalization framework (chapter 4).
3)
As a result of this grammaticalization process, the subordinator que and the subjunctive were able to undergo partial synthesis (chapter 5).
4)
These morpho-syntactic changes resulted in a significant amount of instability in mood distribution, thus triggering the refunctionalization of the subjunctive/indicative contrast in the Romance languages (chapter 6).
1.5.2. Methodology In an attempt to offer a new approach to subjunctive developments in Romance languages, this dissertation will analyze the causes of “mood loss” via three relevant perspectives: word order / branching (cf. Adams 1977; Bauer 1995), grammaticalization (cf. Heine, Claudi and Hünnemeyer 1991; Hopper and Traugott 1993), and synthesis of non-contiguous elements (cf. Schwegler 1990). Additionally, for both accuracy and stronger explanatory value, these phenomena will be illustrated using written data from several stages of Latin, Spanish and French, as opposed to recycling the few—and often artificial— examples that are repeatedly cited in the literature. In order to establish the effect of word-order change on conjunction frequency (chapter 3), I will present a quantitative analysis of conjunctions in left25
branching and right-branching subordinate clauses in Latin. With regard to the grammaticalization of conjunctions (chapter 4), I will illustrate each of the processes (i.e. generalization, decategorialization, specialization, erosion, layering, divergence and renewal), relying on written data from Early Latin (2nd century B.C.), Classical Latin (75 B.C. to 1st century A.D.), and Late Latin (2nd – 6th centuries A.D.), as well as data from medieval Spanish and French. I will then present arguments for the synthesis of que and the subjunctive (chapter 5), relying on Romance data from various periods. Subsequently (in chapter 6), I will investigate the way in which these morpho-syntactic developments have weakened the motivation behind mood selection, thus prompting the refunctionalization of the subjunctive / indicative contrast in Romance languages.
26
CHAPTER 2 REVIEW OF THE LITERATURE 2.1. Introduction Although the subjunctive has been an area of interest for centuries, advances in theoretical frameworks in recent decades have made possible much more rigorous analyses. Beginning in the 1970’s, subjunctive research (which had previously been primarily based on syntax) began to focus on the role of semantics, and by the 1980’s, pragmatic considerations had come to the forefront. Since most of the investigations to date have been synchronic, research questions have primarily addressed the distributional patterns of subjunctive and indicative in modern Romance languages.
This common focus has resulted in a body of work
comprised of solid theories and thoroughly-tested hypotheses. In sharp contrast, there are many aspects of Romance mood that remain almost entirely unexplored or whose analysis has been both quantitatively and qualitatively insufficient. In this section, therefore, I will attempt to make evident the need for new research questions concerning Romance mood, by illustrating the overlap—and perhaps redundancy—among synchronic studies, as well as the relatively superficial nature of diachronic studies, which point to several interesting trends, yet are often limited in terms of scope and/or supporting evidence.
27
2.2. Synchronic Research on the Romance Subjunctive The overwhelming majority of subjunctive research has been synchronic in nature, as most linguists have focused on the mood’s status in modern Romance languages with the primary aim of accurately isolating its function and/or meaning. The enormous quantity of research in this area can be attributed to both a high level of interest in mood and modality in general, and the fact that it is possible to approach the topic from various perspectives and theoretical frameworks.
There is substantial evidence that mood selection is related to
syntax, semantics and pragmatics and, therefore, linguists from a wide range of fields have contributed to our understanding of mood in Romance languages. In the following subsections, I will present an overview of synchronic research related to the subjunctive in Spanish and French, concentrating on the three predominant synchronic approaches: syntactic, semantic, and pragmatic. Most syntactic work on Romance mood begins with the assumption that the subjunctive is entirely context-driven, and thus, devoid of its own meaning. This traditional approach “claims that the subjunctive or indicative forms do not function meaningfully, because the choice of mood is determined automatically by the type of phrase found in the matrix” (Terrell and Hooper 1974: 485). Therefore, the goal in this case is to accurately identify the environments in which the subjunctive appears. Semantic and pragmatic analyses, on the other hand, reject the “automatic” nature of syntactic explanations and, instead, seek to 28
uncover the beliefs, judgements, etc. expressed by the speaker via the subjunctive mood. The following summary of these approaches will focus on those studies which are most representative of the approach as a whole.
2.2.1. Syntactic Analyses The interpretation of mood as a syntactic phenomenon was quite common among traditional grammarians (e.g. Bello 1988 [1847]), who saw mood selection and contrast as context dependent and, as a result, highly predictable. Instead of finding meaning in the subjunctive mood itself, certain linguists continue to treat the subjunctive as an “automatic” reflex of a lexical feature.
This view is
supported by the fact that the subjunctive appears predominantly in subordinate clauses and, in most cases, is not controlled by the speaker. In many ways, subjunctive and indicative appear to be in complementary distribution and it is the task of the syntactician to determine the environments in which each appears. An excellent example of this type of analysis is found in Foster (1973), who not only states that the subjunctive in Spanish is merely a matter of surface syntax, but also makes the strong claim that, in all languages, non-indicative dependent moods are devoid of underlying meaning (1973: 191). He provides the following examples as evidence that the subjunctive mood is non-contrastive and determined by its syntactic environment:
29
(2-1)
Sp.
es
cierto
que
Ud.
lo
be-3SG-PRES-IND
true
that
you-NOM
it-ACC
hace do-3SG-PRES-IND
‘it is true that you do it’ (2-2)
Sp.
es
importante
que
Ud.
lo
be-3SG-PRES-IND
important
that
you-NOM
it-ACC
haga do-3SG-PRES-SUBJ
‘it is important that you do it’ (Foster 1973: 192) The indicative and subjunctive moods are automatically generated in sentences (2-1) and (2-2), respectively. To further support his argument, he explains that selecting an incorrect surface verb cannot “result in a ‘wrong’ message” (Foster 1973: 192). Consider the following sentences: (2-3)
Sp.
*quiero
que
Ud.
lo
want-1SG-PRES-IND
that
you-NOM
it-ACC
hace do-3SG-PRES-IND
‘I want you to do it’ (2-4)
Sp.
*percibo
que
Ud.
lo
perceive-1SG-PRES-IND
that
you-NOM
it-ACC
haga do-3SG-PRES-SUBJ
‘I perceive that you do it’ (Foster 1973: 192) 30
While the surface structure of the above sentences would be deemed incorrect by a native speaker, the alternate mood does not change the message of the utterance. There are, however, environments in Spanish where both moods are possible, as shown in the following nominal, adjectival and adverbial clauses: (2-5)
a. Sp. le you-DAT
digo
que
lo
say-1SG-PRES-IND
that
it-ACC
hace
Ud.
do-3SG-PRES-IND
you
‘I’m telling you that you’re doing it’ b. Sp. le you-DAT
digo
que
lo
say-1SG-PRES-IND
that
it-ACC
haga
Ud.
do-3SG-PRES-SUBJ
you
‘I’m telling you to do it’ (2-6)
a. Sp. busco look for-1SG-PRES-IND
al
hombre
que
PTC-the
man
that
me
ayuda
me-DAT
help-3SG- PRES-IND
‘I’m looking for the man that helps me’ b. Sp. busco look for-1SG-PRES-IND
el
hombre
que
me
the
man
that
me-DAT
ayude help-3SG- PRES-SUBJ
‘I’m looking for the man who will/can help me’ (2-7)
a. Sp. cuando when
llegue
Pablo,
lo
arrive-3SG-PRES-SUBJ
Pablo
him-ACC
31
saludamos greet-1PL-PRES-IND
‘when Pablo arrives, we’ll greet him’ b. Sp. cuando when
llega
Pablo,
lo
arrive-3SG-PRES-IND
Pablo,
him-ACC
saludamos greet-1PL-PRES-IND
‘when Pablo arrives, we greet him’ (Foster 1973: 193) In (2-5a&b), the contrast is based on the meaning of the main verb (informative or imperative).
In (2-6a&b), the contrast relates to the specificity of the noun
antecedent. In (2-7a&b), the contrast is based on the temporal property of the adverbial clause. In short, while Foster does recognize the semantic nature of these contrasts, he asserts that the semantic category is not actually part of the dependent verb but resides elsewhere in the utterance (e.g. in the main verb). The type of examples provided by Foster are relatively common, as attempts to explain subjunctive usage based on syntactic criteria generally rely heavily on distinctions between adjectival, adverbial, and nominal clauses. Shawl (1975), for example, claims that certain syntactic structures, along with specific binary features, are sufficient mood selection criteria. In adjectival clauses, nouns with the feature [+experience] will trigger indicative in the relative clause, while the feature [-experience] will trigger subjunctive, as shown below:
32
(2-8) Det
NP N
S NP Det
Predicate Phrase N tenga farolito
Indef Indef +noun -animate -experience una
casa
una ‘a
casa house
+noun -animate -experience una
casa que that
tenga farolito has a lantern’ (Shawl 1975: 324)
Similarly, the binary feature [α subsequence] is related to temporal clauses, with [+subsequence] and [-subsequence] triggering the subjunctive and indicative, respectively. In a sentence such as Sp. Rosa cantará cuando llegue él ‘Rosa will sing when he arrives’, the subjunctive is chosen due to the feature [+subsequence] of the verb Sp. llegar ‘arrive’. In nominal clauses, the feature involved in mood selection is [α bias].8 Since the verb Sp. querer ‘want’, for example, has the feature [+bias], the subjunctive is selected in Sp. Yo quiero que Rosa hable
8
This binary feature is quite broad. According to Shawl, the feature [- bias] refers to the getting or giving of information, observation or belief, while [+ bias] includes notions of suasion, nonbelief and emotion (1975: 326).
33
despacio ‘I want Rosa to speak slowly’ (Shawl 1975: 325-27). Although this is arguably a semantic feature, it is not a feature of the subordinate verb itself, but of the main verb, which then automatically generates the appropriate mood in the subordinate clause. In an attempt to explain all uses of the subjunctive with just one “rule,” Bergen (1978) claims that “the subjunctive is the surface structure manifestation of the speaker’s (or actor’s) subjective reservation concerning the factual character of the proposition in the subordinate clause” (224).
The feature
[+reservation] is an external condition present within a lexical entry, which adds [+subjunctive] transformationally to the verb of the lower clause. The process is shown below, where (2-9a) represents deep structure, and (2-9b) represents an intermediate structure, which is the product of the transformational rule: (2-9) a.
1 2 ________ ________ Yo [-perf] quer[-past] [+res]
3 ________
4 ________ él [+subs] [-past]
5 ________ ven-
6 ________ mañana
b.
1 2 ________ ________ Yo [-perf] quer[-past] [+res]
3 ________
4 ________ él [+subs] [-past]
5 ________ ven[+subj]
6 ________ mañana
(Bergen 1978: 225) The lexical feature—in this case, [α reservation]—lacks specificity and does very little to predict mood selection, particularly with regard to those instances of 34
subjunctive in factive predicates (e.g. after expressions of emotion, Sp. el hecho de que / Fr. le fait que ‘the fact that’, etc.). In spite of its limitations, this type of analysis is quite common in the literature. As expected, we find a similar line of research related to the subjunctive in French.
Abouda (2002), for example, treats the subjunctive mood as an
agreement phenomenon with very few opportunities for mood alternation on the part of the speaker.
The mood of the subordinate verb is determined by an
abstract semantic trait [T] of the main verb [V1]. This trait may be manifested in verbal morphology (i.e. mood) as well as main verbs (e.g. Fr. vouloir ‘want’) and complementizers (e.g. Fr. sans que ‘without’). He refers to the syntactic relationship between the two verbs as both selectional and projectional, and argues that a complement is not acceptable unless there is agreement between these two systems (Abouda 2002: 5-6). The weakness of Abouda’s claim is that the precise nature of the semantic trait [T] remains unspecified.
What is
interesting, however, is the notion of mood as an agreement phenomenon, and the fact that he includes not only main and subordinate verbs, but also recognizes the role of other elements, such as conjunctions. To highlight the syntactic nature of the subjunctive, linguists often choose to focus on those cases in which there is no apparent semantic motivation behind mood selection.
Arteaga (1993), for example, focuses on subjunctive in
35
ungoverned que- clauses in French in order to support this position. She presents the following examples: (2-10) Fr.
qu’il
fût
arrivé
depuis
that—he-NOM
be-3SG-IMPERF-SUBJ
arrive-PP
since
le
matin,
c’est
vrai
the-SG
morning
it—be-3SG-PRES-IND
true
‘that he’d arrived since morning is true’ (2-11) Fr.
que
le
problème
soit
politique
that
the-SG
problem
be-3SG-PRES-SUBJ
political
est
hors
de
doute
be-3SG-PRES-IND
outside
of
doubt
‘that the problem is political is beyond doubt’ (2-12) Fr.
mais
s’ils
se
but
if—they-NOM
3PL-REFL-PRON
marient
et
qu’ils
marry-3PL-PRES-IND
and
that—they-NOM
aient
(*ont)
des
have-3PL-PRES-SUBJ
(*have-3PL-PRES-IND)
some
enfants
tout de suite…
children
all at once…
‘but if they get married and have children right away’ (2-13) Fr.
si
je
suis
malade
et
if
I-NOM
be-1SG-PRES-IND
sick
and
qu’il
n’y
ait
that-it
NEG—LOC-PRON
have-3SG-PRES-SUBJ
(*a)
personne
pour
(*have-3SG-PRES-IND)
person
in order to
36
me
remplacer
me-ACC
replace-INF
‘if I’m sick and there’s no one to replace me’ (Arteaga 1993: 200-204) According to this analysis, subjunctive is considered to be the default mood, and as a result, is assigned to ungoverned que- clauses such as preverbal que- clauses, shown in (2-10) and (2-11), and conjoined si clauses, shown in (2-12) and (2-13). In these cases, the subjunctive mood cannot be attributed to any semantic properties of the main verb and is therefore merely a “function of the syntax” (Arteaga 1993: 202). It is my view that many syntactic analyses—such as those presented above—are successful in emphasizing the largely automatic nature of the subjunctive in modern Romance languages, yet fail to do so without relying on semantic categories which have not been clearly identified.
In recent years,
syntactic analyses have become increasingly rigorous, relying more heavily on theory than on description, but to what end? The vast majority of the research does not provide speakers (or learners) of Spanish with a better understanding of the cases in which the subjunctive is “generated”. On the other hand, analyses without references to semantic categories (e.g. Arteaga 1993) make significant contributions within the field of syntax. By conducting an analysis in relation to
37
formal features such as word order and conjunctions, one is able to effectively isolate the role of syntax in mood selection.
2.2.2. Semantic Analyses Semantic analyses of the subjunctive mood in Romance languages are, without question, the most prevalent. This is undoubtedly due to the mood’s obvious semantic ties, but may also reflect a desire to provide an adequate explanation for L2 learners. However, although the field of semantics would seem to be the key component to any study of mood and/or modality, it has not provided a simple solution to the problem, as there are several distinct perspectives, each with its own set of strengths and weaknesses. Semantic analyses that attach multiple meanings to the subjunctive with no common thread may potentially result in claims that are too disjointed to be of any real explanatory value, while analyses that seek a single, unified definition of the subjunctive may be prone to oversimplification. Terrell and Hooper’s (1974) seminal work in the field provides an explanation of mood selection based on the following semantic categories:
38
Table 7: Terrell and Hooper’s (1974) Semantic Classification System SEMANTIC NOTION SEMANTIC CLASS MOOD Assertion
Presupposition
Neither
1) Assertion
IND
2) Report
IND
3) Mental Act
IND
4) Comment
SUBJ
5) Doubt
SUBJ
6) Imperative
SUBJ (Terrell and Hooper 1974: 488)
The notion of presupposition is crucial to this study, as it focuses specifically on subjunctive / indicative contexts often considered to be problematic. There is also sufficient syntactic evidence to motivate the above classification system. For example, Terrell and Hooper (1974) distinguish presupposed complements from asserted and doubted complements in two ways. The first relates to negation, as mood selection is not affected by negation in pressuposed complements. Assertive complements, on the other hand, become dubative when negated, and vice versa. Secondly, presupposed complements—unlike assertive, dubative and imperative complements—are compatible with the expression the fact that (Sp. el hecho de que). Compare the following four examples: (2-14) Sp.
estoy
contento
del
be-1SG-PRES-IND
happy-M-SG
of—the-SG
hecho
de
que
María
haya
fact
of
that
Maria
have-3SG-PRES-SUBJ
39
venido
a
visitarnos
come-PP
to
visit-INF—us-ACC
‘I am happy about the fact that Mary came to visit us’ (2-15) Sp.
*dudo
el
hecho
de
doubt-1SG-PRES-IND
the-SG
fact
of
que
María
haya
venido
that
Maria
have-3SG-PRES-SUBJ
come-PP
a
visitarnos
to
visit-INF—us-ACC
‘I doubt the fact that Mary came to visit us’ (2-16) Sp.
*quiero
el
hecho
de
want-1SG-PRES-IND
the-SG
fact
of
que
María
haya
venido
that
Maria
have-3SG-PRES-SUBJ
come-PP
a
visitarnos
to
visit-INF—us-ACC
‘I want the fact that Mary came to visit us’ (2-17) Sp.
*sé
el
hecho
de
know-1SG-PRES-IND
the-SG
fact
of
que
María
haya
venido
that
Maria
have-3SG-PRES-SUBJ
come-PP
a
visitarnos
to
visit-INF—us-ACC
‘I know the fact that Mary came to visit us’ (Terrell and Hooper 1974: 489) In addition to emphasizing the role of presupposition as a semantic category, this classification system also makes a clear distinction between mental acts and 40
comments, which combine with indicative and subjunctive moods, respectively. Furthermore, Hooper and Terrell (1974) argue that this distinction between indicative and subjunctive in presupposed complements may point to an inherent instability in this category, which is also reflected in a higher degree of dialectal variation in these cases. Goldin (1974) offers a similar semantic analysis, but presents it as a process whereby the speaker must apply two ordered principles in order to select the appropriate mood. First, the speaker will assess whether or not he/she is expressing an evaluative reaction; the presence of an evaluative reaction will trigger the subjunctive. If the main clause does not involve a reaction, then the speaker will evaluate the presupposition of the subordinate clause. If there is negative or indefinite presupposition (e.g. I don’t think that… / It is possible that…), the subjunctive will be chosen. If, on the other hand, the presupposition is positive (e.g. I think that…), he/she will select the indicative (Goldin 1974: 296-297).
The following examples illustrate mood selection involving the
reaction principle and the presupposition principle: (2-18) a. Sp. me 1SG-REFL-PRON
alegro
de
make happy-1SG-PRES-IND
about
que
Miguel
venga
a
that
Michael
come-3SG-PRES-SUBJ
to
la
fiesta
the-SG
party
‘I’m happy that Michael is coming to the party.’ 41
b. Sp. no NEG
creo
que
Miguel
believe-1SG-PRES-IND
that
Michael
venga
a
la
fiesta
come-3SG-PRES-SUBJ
to
the-SG
party
‘I don’t think that Michael is coming to the party.’ c. Sp. creo believe-1SG-PRES-IND
que
Miguel
that
Michael
viene
a
la
fiesta
come-3SG-PRES-IND
to
the-SG
party
‘I think that Michael is coming to the party’ (Goldin 1974: 296-297) In (2-18a&b), the reaction and presupposition principles have elicited the subjunctive in the subordinate clause. On the other hand, in (2-18c), there is neither evaluative reaction nor negative/indefinite presupposition, and therefore, the indicative is chosen. While this analysis relies heavily on instances of complementary distribution, other linguists focus on cases where both moods are possible. Looking exclusively at examples where the subjunctive and indicative alternate, Connors (1978) claims that its fundamental meaning is potentiality, which she then divides into two categories: volition and hypothesis. When a matrix verb has two possible interpretations (volitional and assertive), volition will be marked by the subjunctive. Consider the examples below, where (2-19a) is assertive and (2-19b) is volitive:
42
(2-19) a. Fr. je I-NOM
dis
que
Pierre
say-1SG-PRES-IND
that
Pierre
fait
ça
do-3SG-PRES-IND
that
‘I say that Pierre does that’ b. Fr. je I-NOM
dis
que
Pierre
say-1SG-PRES-IND
that
Pierre
fasse
ça
do-3SG-PRES-SUBJ
that
‘I say that Pierre is to do that’ (Connors 1978: 46) On the other hand, when the matrix verb only allows for an assertive interpretation, the indicative/subjunctive alternation becomes possible when negated. In this case, the subjunctive is used to treat the embedded clause as mere hypothesis. Consider the following examples: (2-20) a. Fr. je I-NOM
ne
pense
pas
qu’il
NEG
think-1SG-PRES-IND
NEG
that—he-NOM
vient;
j’en
suis
sûr
come-3SG-PRES-IND
I-NOM—PTV
be-1SG-PRES-IND
sure
‘I don’t think he’s coming; I’m sure of it’ b. Fr. je I-NOM
ne
pense
pas
qu’il
NEG
think-1SG-PRES-IND
NEG
that—he-NOM
vienne come-3SG-PRES-SUBJ
‘I don’t think he’s coming’ (Connors 1978: 46) 43
Connors presents additional examples as further support for the hypothetical nature of the subjunctive: (2-21) a. Fr. que that
Jean
parte,
et
toute
John
leave-3SG-PRES-SUBJ
and
all
l’affaire
est
foutue
the—affaire-F
be-3SG-PRES-IND
damned-F-SG
‘John leaves and the game is up’ (‘if John leaves the game is up’) b. Fr. supposons suppose-1PL-IMP
que
Paul
soit
là
that
Paul
be-3SG-PRES-SUBJ
there
‘let’s suppose that Paul is there’ (Connors 1978: 47-48) Based on the data shown above, Connors concludes that the subjunctive is capable of expressing both volition and hypothesis. She combines these two categories into “potentiality,” in keeping with the well-established trend to provide a unified account of subjunctive usage. Bolinger (1974, 1975) is another linguist who defines the subjunctive as the morphological expression of a semantic feature of the subordinate clause. In contrast to syntacticians, he claims that “certain features are compatible with subjunctive, others with indicative, but none of these features determine either indicative or subjunctive.” (1975: 48, emphasis original). He also argues that the typical division of subjunctive into two semantic categories is inaccurate and even
44
“artificial” (1974: 463). He supports his claim in two ways. First, he presents examples of subjunctive which are neither dubitative nor optative: (2-22) Sp.
es
típicamente
profesorial
que
be-3SG-PRES-IND
typically
professorial
that
D. Andrés
se
haya
D. Andrés
3SG-REFL-PRON
have-3SG-PRES-SUBJ
olvidado
de
sus
gafas
forget-PP
about
his-PL
glasses
‘it’s typically professorial that D. Andrés has forgotten his glasses’ (2-23) Sp.
es
desdeñable
que
be-3SG-PRES-IND
despicable
that
hayan
hablado
mal
de
mí
have-3PL-PRES-SUBJ
speak-PP
badly
about
me
‘it’s despicable that they’ve spoken badly about me’ (Bolinger 1974: 463) Secondly, he argues that if [dubitative] and [optative] were truly distinct features, then conjoining them as shown below should produce a zeugma (e.g. She left in high spirits and a Cadillac): (2-24) Sp.
es
posible,
tal vez
necesario,
pero
be-3SG-PRES-IND
possible,
perhaps
necessary,
but
sin embargo
deplorable,
que
él
however
deplorable,
that
he-NOM
sea
nuestro
representante
be-3SG-PRES-SUBJ
our-SG
representative
45
‘it’s possible, perhaps necessary, but however deplorable, that he is our representative’ (Bolinger 1974: 462) Since no such zeugma occurs in (2-24), Bolinger concludes that the subjunctive must be the manifestation of a single, much broader meaning. His assertion is that the indicative is an intellectual mood, while the subjunctive is attitudinal (1974: 469). The unified nature of this definition does allow it to account for a wider range of examples; however, it may also be too vague to be of any true predictive value. Nowikow (1995), on the other hand, claims that there are two types of subjunctive: subjunctive of subordination and subjunctive of subjective evaluation. She considers the subjunctive of subordination to be all cases of obligatory subjunctive triggered by expressions of need, possibility, fear and emotion, while the second type is “valuative” (1995: 212) and motivated by the attitude of the speaker. According to Nowikow, it is precisely the flexibility of the speaker in these cases that allows linguists to account for variation and socalled “counter-examples.” Consider the sentences below, in which the choice of mood may be justified according to Nowikow’s criteria of valuation: (2-25) Sp.
lo increíble
era
que
Pablo
the incredible thing
be-3SG-IMPERF-IND
that
Pablo
no
lo
sabía
/
NEG
it-ACC
know-3SG-IMPERF-IND
/
46
supiera know-3SG-IMPERF-SUBJ
‘the incredible thing was that Pablo didn’t know it’ (2-26) Sp.
se
queja
de
que
3SG-REFL-PRON
complain-3SG-PRES-IND
about
that
no
le
tratan
/
NEG
him-DAT
treat-3PL-PRES-IND
/
traten treat-3PL-PRES-SUBJ
bien well
‘he complains that they don’t treat him well’ (Nowikow 1995: 211-212) Nowikow claims that the speaker who chooses indicative in these cases is intentionally removing evaluation and emphasizing the assertive value of the utterance (1995: 212). These semantic analyses have, indeed, illustrated and clarified some of the more prevalent notions associated with the subjunctive. However, more recent semantic analyses on the Romance subjunctive have been based on specific theoretical frameworks such as Discourse Representation Theory, Possible-World Semantics and Conditional Semantics and have provided yet additional perspectives. Farkas, for example, set out to explain why so-called fiction verbs (e.g. dream, imagine) pattern like positive categorical epistemic verbs (e.g. believe, think). What these verbs have in common is that “the proposition
47
expressed by their complement is true in a particular world anchored to their subjects” (Farkas 1992: 84). Consider the following representations, in which WR and Wd represent the real world and the dream world, respectively: (2-27)
a. X ---------------------------------x=John believe (x, p) p: y ---------y=Mary sick(y) WR(x) WR b. X ---------------------------------x=John dream (x, p) p: y ---------y=Mary sick(y) Wd(x) WR (Farkas 1992: 87)
In both cases, the propositions are extensionally anchored, since their world variables are a particular world. However, consider the following case, where the 48
proposition is intensionally anchored; in other words, its world variable ranges over a set of worlds (with Wf(x) representing the set of future possibilities): (2-28) X ---------------------------------x=John order (x, p) p: y ---------y=Mary leave(y) Wf(x) WR (Farkas 1992: 92) There is, therefore, a direct relationship between intensional anchoring and the subjunctive, and extensional anchoring and the indicative. This approach is able to account for the use of indicative in fiction verbs as well as the variability associated with factive-emotives, which must be true in the real world (WR) yet do not introduce any particular world. In a related study, Quer (2001) distinguishes between three types of weak intensional predicates (the epistemic model ME(x), the dream model MD(x), and the model of reported conversation MRC(x)) and strong intensional predicates (model of buletic alternatives MBul(x)). presented below:
49
Examples of each model type are
(2-29) ME(x) Sp.
Ana
cree
que
los
Ana
believe-3SG-PRES-IND
that
DEF-ART-PL
pingüinos
vuelen
penguins
fly-3PL-PRES-IND
‘Ana believes that penguins fly’ MD(x) Sp.
Ana
ha
soñado
que
Ana
have-3SG-PRES-IND
dream-PP
that
los
pingüinos
volaban
DEF-ART-PL
penguins
fly-3PL-IMPERF-IND
‘Ana has dreamt that penguins were flying’ MRC(x) Sp. Ana Ana
dice
que
los
say-3SG-PRES-IND
that
DEF-ART-PL
pingüinos
vuelen
penguins
fly-3PL-PRES-IND
‘Ana says that penguins fly’ MBul(x) Sp. Juan John
nos
pidió
que
no
us-DAT
ask-3SG-PRET-IND
that
NEG
le
diéramos
tarea
him-DAT
give-1PL-IMPERF-SUBJ
homework
‘Juan asked us not to give him homework’ (Quer 2001: 85-86) Like Farkas (1992), he associates strong intensional predicates (which introduce a set of worlds) with the subjunctive. He also uses the buletic model to account for other instances of subjunctive usage. For example, the subjunctive is consistently used in purpose clauses, as shown below: 50
(2-30) Sp.
le
voy
a
dar
mil
him-DAT
go-1SG-PRES-IND
to
give-INF
thousand
dólares
para
que
pueda
dollars
in order
that
be able-3SG-PRES-SUBJ
comprar
una
computadora
buy-INF
a
computer
‘I’m going to give him a thousand dollars so that he can buy a computer’ Purpose clauses always include a set of possible worlds and, therefore, appear consistently with verbs in the subjunctive.
On the other hand, consider the
following restrictive relative clauses: (2-31) a. Sp. necesito need-1SG-PRES-IND
un
libro
que
a
book
that
explica
bien
la
guerra
civil
explain-3SG-PRES-IND
well
the
war
civil
‘I need a book that explains the Civil War well.’ b. Sp. necesito need-1SG-PRES-IND
un
libro
que
a
book
that
explique
bien
la
guerra
civil
explain-3SG-PRES-SUBJ
well
the
war
civil
‘I need a book that explains the Civil War well’ Within this framework, the mood contrast illustrated above is considered a case of model shift; (2-31a) is evaluated in the epistemic model (ME) of the speaker, while (2-31b) is evaluated in the buletic model (MBul).
51
Several linguists have used similar criteria in attempts to provide a unified explanation for subjunctive and indicative usage. Villalta (2000) claims that the subjunctive appears when a comparison of alternative propositions is required. She refers to predicates that select the indicative as “acceptance predicates” and predicates that select the subjunctive as “comparative predicates” (236).
To
support her claim that the subjunctive represents a comparison of alternatives on a scale, she provides three tests, illustrated by the data below: (2-32) a. Sp. *Violeta Violeta
concluye/cree/sabe
mucho
conclude/believe/know-3SG-PRES-IND
a lot
que
no
es
posible
that
NEG
be-3SG-PRES-IND
possible
‘Violeta concludes/believes/knows a lot that it’s not possible’ b. Sp. Violeta Violeta
quiere/espera/duda
mucho
want/hope/doubt-3SG-PRES-IND
a lot
que
sea
posible
that
be-3SG-PRES-SUBJ
possible
‘Violeta wants/hopes/doubts a lot that it’s possible’ (2-33) a. Eng. What she said/affirmed/believed turned out to be correct b. Eng. *What she desired/advised/regretted turned out to be correct (2-34) a. Eng. I believe they’re ready > They’re ready, I believe b. Eng. I’m sorry you lost it > *You lost it, I’m sorry (Villalta 2000: 238-239) 52
These data reveal the following characteristics about mood selection: subjunctiveselecting predicates are generally compatible with degree modification, but are generally incompatible with “correct attitudes” (238) and pre-posed complements. As a result, Villalta (2000) argues that the two predicate types do indeed form natural classes, a claim that is consistent with her unified account of mood selection. Related to the notion of comparison is the notion of debate. According to Donaire (2001, 2003), the subjunctive mood is a structure of debate about which the speaker expresses his point of view.
In her own words, “lo propio del
subjuntivo es definir una estructura de discursos en forma de debate entre un punto de vista favorable y un punto de vista desfavorable” (2001: 79). For example, Fr. vienne (the present subjunctive of venir ‘to come’) simultaneously represents both a favorable and unfavorable orientation. The speaker then selects a point of view via expressions such as Fr. vouloir ‘to want’ (favorable), Fr. regretter ‘to regret’ (unfavorable), etc. (Donaire 2003: 125-126). In my view, semantic studies such as those outlined above are characterized by two major strengths: (1) they tend to be more speaker-centered than syntactic analyses, and (2) they present useful terminology, thereby offering the linguist a wide range of descriptive tools. The terminology itself, however, occasionally presents its own set of problems. After years of research, there is now a vast array of terms referring to identical or nearly identical semantic values. 53
One could argue that it is possible to reach a point where additional terminology, rather than serving as a useful linguistic tool, may simply become cumbersome.
2.2.3. Pragmatic Analyses Pragmatic analyses may be the most revealing in terms of speaker motivation behind subjunctive usage. Unfortunately, distinguishing clearly between semantic and pragmatic analyses is often a difficult task, as studies are commonly referred to as both semantic and pragmatic, by critics and by the authors themselves. For the purposes of this literature review, pragmatic analyses of subjunctive will include topics related to speaker assertion, information structure, and reformulation. Already in the 1970s, the pragmatic nature of the subjunctive mood was a fruitful area of investigation.
Klein (1975), for instance, claims that the
subjunctive is the mood of non-assertion (with assertion being expressed by the indicative). Furthermore, while all cases of subjunctive represent non-assertion, the precise pragmatic interpretation depends on real-world possibilities and the nature of the complement in question. Consider the following examples:
54
(2-35) a. Sp. dudo doubt-1SG-PRES-IND
que
aprenda
that
learn-3SG-PRES-SUBJ
‘I doubt that he’s learning’ b. Sp. deseo want-1SG-PRES-IND
que
aprenda
that
learn-3SG-PRES-SUBJ
‘I want him to learn’ c. Sp. lamento regret-1SG-PRES-IND
que
aprenda
that
learn-3SG-PRES-SUBJ
‘I regret that he’s learning’ (Klein 1975: 353-355) In (2-35a), the lack of assertion is related to the uncertainty of the proposition, while the lack of assertion in (2-35b) is due to the fact that the event is not yet realized or assured. In (2-35c), the subjunctive is chosen even though we are dealing with a fact, since the clause does not have informative value, as the speaker is merely commenting on or reacting to its content (Klein 1975: 355). Crucial to this type of analysis is a clear understanding of the notion of assertion. Lunn (1989) attempts to clarify the concept by providing the following scale: less assertable……………...assertable………………….less assertable untrue…………..…...both new and true…………………..old SUBJUNCTIVE……..…..INDICATIVE……………..SUBJUNCTIVE (Lunn 1989: 691) According to Lunn (1989), assertable information is generally both new and true and will be encoded in the indicative mood. Old and/or untrue information, on 55
the other hand, is less assertable and will be encoded in the subjunctive mood (691). In a later study, Lunn (1995) continues this line of reasoning and provides several examples of journalistic language in which the subjunctive is chosen to express old, less assertable information: (2-36) Sp.
la
pareja
que
se
the
couple
that
3SG-REFL-PRON
hiciera
famosa
por
interpretar
make-3SG-IMPERF-SUBJ
famous
for
interpret-INF
el
papel
de
marido
y
mujer
en
the
role
of
husband
and
wife
on
“El pájaro espino”,
es
en
la
“El pájaro espino”,
be-3SG-PRES-IND
en
the
vida
real
un
matrimonio
feliz
life
real
a
couple
happy
‘the couple that became famous for playing the role of husband and wife on “El pájaro espino” is in real life a happy couple’ (2-37) Sp.
casi
once
años
después de
que
el
almost
eleven
years
after
that
the
Bayern
de
Munich
alcanzara
su
Bayern
of
Munich
seize-3SG-IMPERF-SUBJ
its
tercera
y
última
Copa
de
Europa,
los
third
and
last
Cup
of
Europe,
the
vigentes
campeones
de
la
RFA
y
established
champions
of
the
RFA
and
España
van
a
volver
a
Spain
go-3PL-PRES-IND
to
return-INF
to
56
enfrentarse
en
un
encuentro
oficial
face-INF—3PL-REFL-PRON
en
an
encounter
official
‘almost eleven years after the Bayern of Munich seized its third and last European Cup, the established champions of the RFA and Spain are going to face each other in an official encounter’ (Lunn 1995: 432-433) According to Lunn, examples such as these reveal the “semantically minimizing” (1995: 436) function of the subjunctive.
Furthermore, this is one of the
subjunctive contexts that exhibits a high degree of variability. Mejías-Bikandi (1994) also attempts to define the notion of assertion, and explains that the most common misconception is that non-presupposition and assertion are equal, which they are not. He argues that “a speaker asserts a proposition P when the speaker intends to indicate that P is contained in some space R, that is, when the speaker intends to indicate that P provides information about some individual’s view of reality (1994: 895). Using this definition of assertion we can understand why the indicative is used in (2-38) and (2-39), while the subjunctive is used in (2-40): (2-38) Sp.
creo
que
María
está
believe-1SG-PRES-IND
that
Maria
be-3SG-PRES-IND
enferma sick
‘I believe that Maria is sick’ (2-39) Sp.
María
se
ha
dado
Maria
3SG-REFL-PRON
have-3SG-PRES-IND
give-PP
57
cuenta
de
que
Pedro
está
account
of
that
Pedro
be-3SG-PRES-IND
enfermo sick
‘Maria has realized that Pedro is sick’ (2-40) Sp.
me
alegro
de
1SG-REFL-PRON
make happy-1SG-PRES-IND
about
que
Pedro
esté
enfermo
that
Pedro
be-3SG-PRES-SUBJ
sick
‘I’m happy that Pedro is sick’ (Mejías-Bikandi 1994: 897-899) In (2-38), the speaker may not be entirely certain about María’s well-being, but is asserting his belief nevertheless.
The sentence in (2-39) provides a classic
example of presupposed information, yet the indicative mood indicates that the speaker is committing himself to the truth of the utterance. In (2-40), however, the emphasis is on the emotional reaction to Pedro’s condition, and the fact that Pedro is sick, although true, is not being asserted. Another pragmatic approach that has received a significant amount of attention in recent years is the study of the relationship between mood and thematic structure. The claim that the subjunctive is associated with theme and the indicative with rheme is supported by several facts.
First of all, the
subjunctive is often found with information considered to be common/public knowledge, as shown below:
58
(2-41) Sp.
que
haya
pobres,
por
cierto,
that
be-3SG-PRES-SUBJ
poor-PL
for
certain
es
una
imperfección
y
be-3SG-PRES-IND
an
imperfection
and
el
capitalismo,
como
todo
DEF-ART-SG
capitalism
like
all
lo
humano,
es
imperfecto
it
human
be-3SG-PRES-IND
imperfect
‘that there are poor people, of course, is an imperfection, and capitalism, like everything human, is imperfect’ (Krakusin and Cedeño 1992: 1290) Furthermore, there is a parallel between thematic organization (i.e. rheme in sentence-final position) and mood selection. Krakusin and Cedeño’s (1992) analysis of Sp. el hecho de que ‘the fact that’ reveals that thematic information is generally found in initial position and in the subjunctive mood, while rhematic information is more commonly found in final position and in the indicative mood. Compare, for instance, the two examples below: (2-42) Sp.
el
hecho
de
que
en
los
cuatro
el
the
fact
of
that
in
the
four
the
régimen
democrático
se
regime
democratic
3SG-REFL-PRON
impusiera
facilitaba
impose-3SG-IMPERF-SUBJ
facilitate-3SG-IMPERF-IND
59
la
tarea
the
task
‘the fact that in the four (countries) the democratic regime was imposed facilitated the task’ (2-43) Sp.
esta
tesis
ignora
el
hecho
this
thesis
ignore-3SG-PRES-IND
the
fact
decisivo
de
que
la
gran
mayoría
de
decisive
of
that
the
great
majority
of
norteamericanos
vive
arriba
y
North Americans
live-3SG-PRES-IND
above
and
no
abajo
de
la
frontera
de
NEG
below
of
the
border
of
la
pobreza
DEF-ART-SG
poverty
‘this thesis ignores the decisive fact that the vast majority of North Americans lives above and not below the poverty level’ (Krakusin and Cedeño 1992: 1290-1291) In example (2-42), el hecho de que introduces a subjunctive verb in thematic position, while in (2-43), el hecho de que appears in rhematic position with a verb in the indicative. This contrast serves to underscore the pragmatic nature of mood distinctions. In yet another pragmatic account of the subjunctive mood, Ligatto argues that the subjunctive mood can be used as an indicator of reformulation (2002: 140). She provides the following dialog to illustrate the phenomenon:
60
(2-44) A: Sp. Juan John
se
marcha
hoy
3SG-REFL-PRON
walk-3SG-PRES-IND
today
‘John is leaving today’ B: Sp. no NEG
puede
ser:
be able-3SG-PRES-IND
be-INF
hoy
llegan
sus
amigos
today
arrive-3PL-PRES-IND
his-PL
friends
‘that can’t be: his friends are arriving today’ A: Sp. está be-3SG-PRES-IND
decidido
a
determined
to
marcharse
aunque
hoy
walk-INF—3SG-REFL-PRON
even though
today
vengan
sus
amigos
come-3PL-PRES-SUBJ
his-PL
friends
‘he’s determined to leave even though his friends are coming today’ (Ligatto 2002: 142) It is clear that the purpose of the subjunctive form (Sp. vengan) is not related to the factivity of the proposition, since both speakers A and B are aware of the fact that the friends are arriving today. Instead, the subjunctive mood in this case reflects the pragmatic “weight” of the utterance, which is “devoid of informative value” (Ligatto 2002: 142), as speaker A is merely reformulating information already presented earlier in the discourse. Pragmatic accounts such as these are extremely intriguing, primarily due to their radical departure from the more traditional syntactic and semantic 61
analyses. One of the primary strengths of a pragmatic approach is that it is highly speaker centered and, therefore, allows the linguist to account for several subjunctive uses that would normally be termed “counter-examples.” In addition, a focus on information structure succesfully merges the fields of syntax and semantics, by identifying patterns of meaning in particular syntactic structures. On the other hand, pragmatic explanations occasionally overlook other possible factors in mood selection. For example, the subjunctive in “que haya pobres, por cierto, es una imperfección” (Krakusin and Cedeño 1992: 1290), could easily be attributed to the comment es una imperfección rather than the thematic position of the subordinate clause. Furthermore, pragmatic explanations tend to have substantial descriptive, but little predictive value. Many of the examples above could have appeared in both moods (indicative and subjunctive), and a reasonable justification would have likely been provided for both.
2.2.4. Contributions of Synchronic Research In spite of the staggering amount of research dealing with the function(s) of the subjunctive, there is quite a bit of overlap, both in goals and findings. In fact, the principle claims (from linguists attempting to provide a mostly unified explanation) can be presented in summary form as follows:
62
Table 8: Summary of Synchronic Research on Romance Mood Linguists IND SUBJ Shawl +experience -experience (1975) -subsequence +subsequence -bias +bias Bergen -reservation +reservation (1978) Connors reification potentiality (1978) assertion volition non-hypothesis hypothesis Bolinger intellectual attitudinal (1974 & 1975) Nowikow -valuative +valuative (1995) Farkas extensional anchoring intensional anchoring (1992) Quer weak intensional predicates strong intensional predicates (2001) Villalta acceptance comparison (2000) Donaire -debate +debate (2003) Klein +assertion -assertion (1975) Krakusin & rhematic thematic Cedeño (1992) Ligatto -reformulation +reformulation (2002) From a comparison such as this, it becomes clear that the synchronic studies cited above have much in common. First of all, they all share a common goal: to accurately describe the contexts in which the subjunctive is selected. Although they do indeed appear to reveal the basic underlying role of the subjunctive, they fail to account for certain uses and variation. There are several reasons for this. First of all, semantic distinctions such as those found in modal and aspectual
63
categories are typically difficult to define. Secondly, the pragmatic nature of the subjunctive and the subsequent importance of the speaker make idiolectal variation the norm rather than the exception.
Finally, from a diachronic
standpoint, Romance mood appears to be particularly unstable, thus resulting in an unusually high degree of synchronic variation. It is, therefore, my (somewhat pessimistic) view that the goal of defining the subjunctive in this way may be impossible to reach.
In fact, I believe that our collective knowledge of the
Romance subjunctive does not suffer from a lack of answers, but a lack of questions. A few select areas of research continue to draw attention, while others are rarely, if ever, addressed. It is for this reason that new perspectives regarding Romance mood are sorely needed. Comparative and diachronic studies, for example, are particulary useful in terms of uncovering larger trends and/or accounting for directions of change. Looking at cross-linguistic developments diachronically enables us to understand how a given feature functions within the larger system and, as a result, allows us to explain phenomena that would otherwise be considered irregular or exceptional within a synchronic framework.
2.3. Diachronic Research on the Romance Subjunctive While synchronic research typically focuses on the distributional patterns of indicative and subjunctive, diachronic research explores changes in such 64
patterning. These changes are often depicted as a gradual “loss” of subjunctive in the history of Romance languages. Studies in this area, however, are often less rigorous, in addition to being primarily language-specific. Work on subjunctive loss in Spanish, for example, focuses primarily on language-contact situations, while the same phenomenon in French is typically—and not surprisingly— analyzed from a phonological perspective. In the sections that follow, I will present an overview of these language-specific studies, followed by a summary of the existing pan-Romance explanations for “mood loss.”
2.3.1. Language-Specific Approaches Among the attempts to explain changes in mood distribution from Latin to modern Romance, the majority of studies are language specific. This is, in my opinion, the main limitation of the diachronic research to date. Although there have been some interesting data presented illustrating the loss of subjunctive morphology and/or motivation, these analyses are not comparative and therefore lack additional data which would potentially serve to support or refute the authors’ claims. Below, the three prevailing approaches to “mood loss” in Romance languages will be presented: (1) phonological reduction, which appears repeatedly in research in French linguistics, (2) language contact, which is often cited as an
65
explanation for mood loss in Spanish, and (3) semantic change, which is more descriptive than explanatory, yet may have the most cross-linguistic applicability.
2.3.1.1. The Role of Phonology With regard to the very noticeable loss of mood distinctions in French, the general tendency has been to claim that phonetic erosion led to the downfall of the subjunctive paradigms.
Due to its radical nature in the history of French,
phonological reduction is often cited as an explanation for morphological and even syntactic change. Focusing specifically on the imperfect subjunctive, Van Vliet (1983) discusses several factors that are often used to explain its disappearance in French.
First of all, he points out that the passé antérieur, the imperfect
subjunctive and the pluperfect subjunctive are all structurally tied to the passé simple.
The head tense—the passé simple—weakened for several reasons.
Suffixal morphology had become less common, and an alternative analytic construction existed that would be capable of replacing it. In addition, there was syncretism between the first-person singular forms Fr. chantai (simple past) and chantais (imperfect). Finally, CV was the preferred syllable structure, which was not consistent with forms such as Fr. chantâmes, chantâtes, and chantèrent. Once the passé simple came to be replaced by the passé composé, the entire family of paradigms collapsed (Van Vliet 1983: 91). Furthermore, Van Vliet claims that 66
there were two problems inherent in the imperfect subjunctive ending -asse: homonymy and aesthetics. Neither of these arguments, however, is convincing from a linguistic standpoint.
First of all, homonyms are quite abundant in
language and result in little if any miscommunication. Any possible confusion between the verb Fr. limasse ‘to file’ in the clause Fr. que je limasse and the noun Fr. limace ‘slug’ (Van Vliet 1983: 97) is completely inconceivable. Furthermore, the claim that the phoneme sequence /-as-/ (as in Fr. il veut que vous l'assassinassiez ‘he wants you to assassinate him’) is "unpleasant to the French ear" (Van Vliet 1983: 96) is without any scientific basis whatsoever. The problem often associated with phonological explanations such as this is that correlation and causation are often confused. In other words, the fact that a given phonological change coincides with a change in morpho-syntax does not prove that one was the cause of the other. This claim is also supported by Harris (1978a), who refutes Vennemann’s (1975) argument that phonological reduction causes syntactic drift. Focusing on issues of case loss and obligatory subject pronouns in Romance Languages, he demonstrates that phonetic erosion should not be taken as the cause for such changes. His work emphasizes the importance of comparative studies, as such research reveals that case loss, for example, occurred faster in Spanish in spite of the fact that phonetic erosion was more widespread in French. When dealing with phonological change, therefore, it is
67
useful to acknowledge that it is often merely a contributing factor to, or symptom of, morpho-syntactic change.
2.3.1.2. The Role of Language Contact Among studies on the loss of subjunctive mood in Spanish, the prevailing approach has not been based on phonology, but rather on language contact. Diachronic change among Spanish-English bilinguals has been a particularly prolific area of research over the last few decades. Guitart, for example, proposes the following hypothesis: "The more a Spanish-English bilingual is influenced by English in his use of Spanish, the less he will use the subjunctive in sentences in which the matrix comments on the clause" (1982: 61). Of the Hispanics living in the United States who participated in his study, the Venezuelan informants—who consider themselves Spanishdominant—used the subjunctive more than the Miami Cubans, who have more exposure to English. Similarly, the Miami Cubans used the subjunctive more than the Mexican-American informants, who consider themselves English-dominant (Guitart 1982: 64). However, in order for this argument to be convincing, we would need data indicating no corresponding reduction in subjunctive usage among monolinguals in Venezuela, Cuba and Mexico. It may, therefore, be the case that the bilingual data presented by Guitart (1982) merely reflect dialectal variation within the Spanish language itself, albeit at a more accelerated rate. 68
The tendency to attribute these types of change to language contact is quite common in the literature. In his study of Puerto Rican-American Spanish, Lantolf (1978) reveals an ongoing reduction in subjunctive usage among SpanishEnglish bilinguals. In volitional contexts, the indicative was accepted by only 1% of informants aged 40-67, but by 9% of informants between the ages of 12 and 19. In dubitative contexts, the indicative was accepted by 7% of informants aged 40-67, but by 20% of informants aged 12-19. Finally, informants between the ages of 40 and 67 accepted the indicative in 22% of presuppositional contexts, but informants aged 12-19 accepted indicative in 36% of these cases. Clearly, there is a shift from subjunctive to indicative occurring among this group of bilingual speakers, arguably due to a loss of “contact with the norm of their language” (Lantolf 1978: 209).
Once again, the phenomenon (i.e. mood usage among
bilingual speakers) is studied without consideration given to internal factors and, therefore, changes have been attributed exclusively to external causes. Similar studies have been done with respect to Spanish in the Southwest United States (cf. Sánchez 1972, Hensey 1973 & 1976, Solé 1977, Floyd 1978, Studerus 1995). While the general trend seems to be a reduction in subjunctive mood, very few clear patterns have emerged and no strong claims have been made (cf. Floyd 1978). For example, in a study on border varieties of Spanish, Studerus (1995) compared subjunctive usage among Spanish-speakers from Laredo (Texas) and Spanish-speakers from Nuevo Laredo (Mexico) by providing subjects with a 69
list of sentences including two options (indicative and subjunctive) and asking them to choose the form that seemed most natural.
Studerus came to the
somewhat broad conclusion that “substantial rule variation does exist” (1995: 94) and Mexican speakers select subjunctive mood more frequently than MexicanAmerican speakers. In a related study, García and Terrell (1977) administered judgement tasks to both Mexican and Mexican-American students aged 13-19. Indicative was accepted by the Mexican informants for 8% of volitional contexts, 25% of dubitative contexts, and 48% of commentative (presuppositional) contexts, while the Mexican-American informants accepted indicative in these same contexts 26%, 47% and 57% of the time.
The authors then propose two possible
explanations for these data: either the mood system is static for monolinguals, but bilingual speakers are losing the mood contrast, or the mood system is variable for all speakers and bilinguals merely represent a more advanced state due to lack of normative pressure. The first of these proposals is quite surprising, since the data provided by Mexican speakers is far from static. The second proposal, however, is completely in line with my own perspective regarding the tendency for external influences to be secondary to internal changes. In a more recent study, Silva-Corvalán (1994) claims that contact with English has indeed resulted in a loss of subjunctive morphology. However, her argument is not based on the notion of interference, but rather on what she 70
considers a general principle of bilingualism: phenomena such as regularization and loss are "motivated by bilingual speakers' need to lessen their cognitive load when having to communicate rather frequently in two or more languages" (SilvaCorvalán 1994: 268). In other words, simplification is merely a strategy whose function is to maintain the non-dominant language. While there is no doubt that the notion of "cognitive load" plays a role in the shift from subjunctive to indicative, one important question remains: Which features of a language does this simplification strategy target, and why? In simple terms, if the subjunctive mood were not formally unstable and/or did not have a weak semantic load, it would not undergo processes of simplification such as we find among SpanishEnglish bilinguals. Silva-Corvalán's data do, indeed, support this possibility. Her monolingual informants used the indicative where the subjunctive was expected 6.5% of the time, revealing that subjunctive loss was not triggered by contact with English, but merely accelerated by it. In fact, Silva-Corvalán states that modern standard varieties of Spanish undergoing this shift have been documented in Argentina, Mexico, Paraguay, Uruguay, and Venezuela and she does acknowledge that “the phenomenon of gradual loss of mood distinctions examined here represents part of an evolutionary trend in Spanish and other Romance Languages” (Silva-Corvalán 1994: 268). In conclusion, this line of research has been quite fruitful in terms of identifying and characterizing a linguistic trend found in bilingual communities. 71
The data presented by these researchers indicate that bilingualism does indeed have a significant impact on linguistic features. However, while recent studies have shown that contact-induced grammatical change is indeed possible (cf. Heine and Kuteva 2005), morpho-syntactic interference is much less common than lexical or phonological interference. For this reason, it is particulary important, when dealing with morpho-syntactic change, to consider the possibility of underlying internal factors, a step which is often not included in studies such as those discussed above. In fact, one could argue that any change due to contact must also have a language-internal component. In other words, there must be a predisposition for change, the purpose of which may often be a move toward system congruency.9
2.3.1.3. The Role of Semantic Shift Several linguists, with regard to Spanish in particular, have presented data to illustrate that reduction in subjunctive usage is not random or categorical, but rather that there is an ongoing change in distributional patterns linked to semantic and/or pragmatic factors. This argument is based on the notion that the links between different semantic categories and the subjunctive mood are not equally strong.
Consider the following hierarchy (corroborated by both Terrell and
Hooper 1974 and Lantolf 1978): assertion > reporting > presupposition 9
See Danchev (1988) for an overview of various perspectives on contact-induced change. He argues that “languages can borrow what is ‘in harmony’ with their ‘wants’ and ‘needs’” (39).
72
(mental acts and comments) > doubt > volition. The category of assertion is associated most strongly with indicative, while the category of volition is tied most strongly to commands, and hence, the subjunctive mood.
As a result,
subjunctive loss will logically occur in a top-down fashion, which is precisely what has been attested in the literature. Instances of assertion expressed by subjunctive or volition expressed by indicative are extremely scarce, yet pressuppositional clauses—in sharp contrast—demonstrate a high degree of variability. Goldin (1974), for example, claims that, for most speakers of Spanish, the subjunctive is governed by two ordered principles: the reaction principle and the presupposition principle (see section 2.2.2), illustrated respectively in examples (2-45) and (2-46) below: (2-45) a. Sp. me 1SG-REFL-PRON
alegro
de
que
make happy-1SG-PRES-IND
about
that
sea
fácil
be-3SG-PRES-SUBJ
easy
'I'm happy that it's easy' b. Sp. es be-3SG-PRES-IND
bueno
que
te
good
that
you-DAT
gusten
tus
clases
please-3PL-PRES-SUBJ
your-PL
classes
'it's good that you like your classes' (2-46) a. Sp. es be-3SG-PRES-IND
73
posible
que
Alicia
possible
that
Alicia
haya
llegado
a
casa
ya
have-3SG-PRES-SUBJ
arrive-PP
to
house
already
'it's possible that Alicia has already gotten home' b. Sp. Pedro Pedro
espera
que
María
le
hope-3SG-PRES-IND
that
Maria
him-DAT
escriba write-3SG-PRES-SUBJ
'Pedro hopes that Maria will write to him' (Goldin 1974: 296-97) According to Goldin, cases of subjunctive "loss" are actually cases in which the native speakers' grammar appears not to include the reaction principle. He cites the following examples: (2-47) a. Sp. me 1SG-REFL-PRON
alegro
de
que
make happy-1SG-PRES-IND
about
that
está
aquí
be-3SG-PRES-IND
here
'I'm glad you're here' b. Sp. es be-3SG-PRES-IND
una
lástima
que
a
shame
that
ha
ido
have-3SG-PRES-IND
go-PP
‘it's a shame that he has gone' (Goldin 1974: 300) While these speakers continue to use examples of subjunctive “triggered” by the presupposition principle, the reaction principle is simply not part of their
74
grammar, and therefore, they do not use subjunctive after expressions such as Sp. alegrarse de que ‘be happy that’, Sp. es bueno que ‘it’s good that’, Sp. es una lástima que ‘it’s a shame that’, etc. Guitart (1982) presents data of a similar nature, as shown in the dialog below: (2-48) Luis: Sp.
Oye,
Blas, algo
te
listen-2SG-IMP
Blas
you-DAT
something
molesta.
¿Qué
cosa
bother-3SG-PRES-IND
what
thing
es? be-3SG-PRES-IND
'Listen, Blas, something is bothering you. What is it?' Blas: Sp.
que
Juan
no
está
aquí
that
John
NEG
be-3SG-PRES-IND
here
'that Juan's not here' (Guitart 1982: 59) He supports the claim previously made by García and Terrell (1977) and Lantolf (1978) regarding Spanish-English bilinguals, that commentative sentences in particular undergo a shift from subjunctive to indicative (Guitart 1982: 61). Along the same lines, Studerus claims that of the five "subjunctive triggering matrices" (1981: 97)—doubt, uncertainty, comment, optative, imperative—the comment matrix in particular exhibits a shift to the indicative, as illustrated in the following examples: 75
(2-49) Sp.
era
curioso
que
be-3SG-IMPERF-IND
curious
that
bailaba dance-3SG-IMPERF-IND
'it was odd that she used to dance' (2-50) Sp.
me
alegré
que
1SG-REFL-PRON
make happy-1SG-PRET-IND
that
bailó dance-3SG-PRET-IND
'I was happy that she danced.' (Studerus 1981: 98-99) In conclusion, while these approaches may accurately describe the patterns involved in mood loss, they generally do not possess explanatory value. To state that the subjunctive is being lost following reaction clauses, for example, merely addresses what is happening, not why. More specifically, many studies have presented data showing a specific pattern of loss, yet they have not addressed the initial or fundamental motivation behind the change.
2.3.2. Pan-Romance Approaches Aside from truly language-specific factors, such as contact with English and radical phonetic erosion, there have been very few attempts to explain changes in Romance mood distribution. To do so would require an analysis of linguistic features dating back to Classical Latin in addition to an analysis of features
76
outside the scope of mood itself. As a result, pan-Romance claims often lack a solid framework and/or sufficient supporting evidence. Journoud (1971), for example, attributes subjunctive loss to a shift in mentality. He argues that modern mentality has become concrete and fact-based and states: l’organisation de nos sociétés modernes et de leurs formes de vie repose de plus en plus sur le fait, sur l’omniprésence du fait, dont les gens sont invités (par les “slogans”, par les “mass media”), à prendre conscience collectivement, c’est-à-dire sans y rien ajouter d’eux mêmes, sans que l’appréciation, l’interprétation personnelles viennent “repenser” la réalité. C’est, dès lors, dans le langage, le triomphe de l’indicatif, mode du réel et du concret. Plus besoin, pour le grand publique, de suppositions, de doutes ni de nuances, plus besoin d’abstraction. Partant, plus besoin de subjonctif. (Journoud 1971: 551) The argument that modern-day society has eliminated the need for evaluation, interpretation, or abstraction—and therefore, subjunctive—is not convincing. Although there has indeed been a drastic reduction in subjunctive (as a morphological paradigm), this in no way implies a concomitant reduction in evaluative comments or speaker interpretations.
In short, abstract thoughts
continue to be expressed, in spite of a lack of overt morphological marking in the subordinate clause. Although the flaws in diachronic studies are not typically this apparent, studies that offer one or two unsupported hypotheses or make no reference to early stages in the development do not offer satisfactory contributions to this
77
complex topic. Therefore, I will consider in the following sections only those linguists who have set out with the intention of systematically addressing subjunctive developments in the Romance languages as a group.
2.3.2.1. The Role of Tense Prominence In this section, I will refer to studies in which mood loss is presented as a logical result of a tense-prominent verbal system. In general terms, this line of research asserts that the subjunctive mood can be seen as compatible with the Latin verbal system, which was predominantly aspect-based, but not with the modern Romance systems, which are predominantly tense-based. According to Hewson and Bubenik (1997), aspect in Latin was perceived to be so fundamental that grammarians as early as the 1st century B.C. had divided the verbal system into two main categories: infectum and perfectum. Each of these two aspectual categories had three corresponding indicative paradigms (present, past, and future), and two corresponding subjunctive paradigms (present and past) (Hewson and Bubenik 1997: 189-202). In Vulgar Latin, however, there was an increase in tense prominence marked by the creation of new future and conditional paradigms with Lt. habere (cf. Fleischman 1982). In concise terms, Hewson and Bubenik explain that “the Latin verbal system was completely revamped.
The distinction between Perfectum (Retrospective) and Infectum
78
(Non-Retrospective), an aspectual contrast in Classical Latin, was transformed into differences of tense” (1997: 314). As a result of this development, the newer Romance system based on values of past, present and future created limitations for the subjunctive, which was not able to express all of these in explicit ways.
Furthermore, the
development of the analytic forms, Sp. voy a cantar / Fr. je vais chanter ‘I’m going to sing’ and Sp. iba a cantar / Fr. j’allais chanter ‘I was going to sing’, allowed the older, synthetic future and conditional to acquire modal value. Thus, future and conditional paradigms in modern Romance languages may be used for conjecture and hypothesis, respectively. In fact, Boysen asserts that subjunctive loss directly corresponds to the presence of a conditional paradigm (1966: 22). Along similar lines, Harris argues that the new conditional forms of Vulgar Latin came to “rival and at times replace the subjunctive as markers of attenuation” (1974: 176). This argument can easily be extended to Modern Romance, where tense prominence appears to have a definite effect on mood choice. Although this is an area of research that has not been treated thoroughly, several linguists have noticed a trend whereby indicative replaces subjunctive precisely when the speaker wishes to express a tense with no corresponding subjunctive paradigm. García and Terrell, for example, observed that Spanish-speakers who used indicative forms in subjunctive contexts were typically using preterit and future 79
indicative forms (1977: 222).
In their study, examples of preterit used in
subjunctive contexts were accepted sixty-seven percent of the time, while examples of imperfect indicative in subjunctive contexts were only accepted six percent of the time (223). Similar work has been done by Studerus (1981), who illustrates the temporal “blurring” that occurs due to the subjunctive in the following sentences: (2-51) a. Sp. era be-3SG-IMPERF-IND
curioso
que
curious
that
viniera come-3SG-IMPERF-SUBJ
‘it was odd [at time b] that he used to come [at time a]’ b. Sp. era be-3SG-IMPERF-IND
curioso
que
curious
that
viniera come-3SG-IMPERF-SUBJ
‘it was odd [at time b] that he would come [at time c]’ (Studerus 1981: 100) The inability of the subjunctive mood to distinguish between anteriority and subsequence in cases such as this may be a possible motivation for some speakers to select indicative for “contrast”: (2-52) a. Sp. era be-3SG-IMPERF-IND
80
curioso
que
curious
that
venía come-3SG-IMPERF-IND
‘it was odd [at time b] that he used to come [at time a]’ b. Sp. era be-3SG-IMPERF-IND
curioso
que
curious
that
viniera come-3SG-IMPERF-SUBJ
‘it was odd [at time b] that he would come [at time c]’ (Studerus 1981: 100) In the examples above, the speaker’s desire to express temporal notions of anteriority and subsequence apparently takes precedence over the expression of mood. This type of analysis is extremely significant since it takes into account two very important concepts in diachronic linguistics: system congruency and the coexistence of old and new forms. These linguists do not examine the subjunctive in isolation, but instead consider its function in relation to the verbal system as a whole; in this case, one in which tense has become increasingly prominent. Furthermore, the existence of two rival forms (e.g. synthetic and analytic future) makes possible a shift in meaning and function. If older forms have acquired modal value (a well-documented phenomenon), then it is possible for them to usurp the role of the subjunctive. These are definitely important considerations. However, while there may be a tendency to replace subjunctive with indicative in order to express explicit temporal values, this alone cannot account for a 81
reduction in subjunctive usage, since indicative often replaces subjunctive in cases where no temporal advantage is apparent.
2.3.2.2. The Role of Hypotaxis The most widespread explanation provided for the reduction in subjunctive in Romance languages is the shift from parataxis to hypotaxis attributed to Classical Latin. As explained by Harris (1974, 1978b), the development of hypotactic structures is typically justified in functionalist terms. In other words, the more literate the society, the more complex (i.e. hypotactic) the form of speech must be (Harris 1974: 177). An expression such as that shown in (2-53a) came to be replaced by (2-53b) below: (2-53) a. Lt. volo;
veniat
wish-1SG-PRES-IND
come-3SG-PRES-SUBJ
‘I wish it; may he come’ b. Lt. volo wish-1SG-PRES-IND
ut
veniat
that
come-3SG-PRES-SUBJ
‘I wish that he comes’ (‘I want him to come’) (Harris 1978b: 168) As a result of this change, Harris states that the subjunctive “gradually ceased to be linked to a semantic value inherent in the matrix verb but came rather to be conceived of as a much more general marker of the grammatical process of subordination itself” (Harris 1978b: 169). For this reason, it is widely argued that
82
Latin was already dealing with two distinct subjunctives: one which was semantically motivated and one which was automatic and devoid of meaning. Handford (1947), for example, divides the Latin subjunctive into two main types, which he refers to as “subjunctive with modal force” (29-138) and “dependent subjunctive with weakened meaning” (139-179). Although Harris has perhaps articulated this point of view the most explicitly, he is clearly not without support. The subjunctive is often referred to as the mood of subordination (e.g. McCray 1988: 180), and references similar to Harris’ have been made regarding a “subjunctive of hypotaxis” (cf. Nowikow 1995).
Sabanééva, for example, claims that “le subjonctif en subordonnée
demeure essentiellement amodal, détaché de l’attitude du locuteur” (1993: 65), while Handford argues that the subjunctive in Classical Latin “tended to have its modal force more and more weakened, until in many cases it functioned merely as a mood of subordination” (1947: 23). The strength of the argument regarding hypotaxis lies in the recognition of the “subjunctive loss” phenomenon as being tied to morpho-syntactic developments in Latin, as opposed to a recent innovation or result of external factors such as language contact. There is no doubt that patterns of subordination played a prominent role in the development of mood distinctions in modern Romance languages. Furthermore, the strong claim that subjunctive was already “meaningless” (Harris 1974: 171) in Latin gives us a much needed perspective 83
regarding new developments. Classic Latin did not have the invariable, highly motivated subjunctive paradigm that traditional grammars would lead us to believe. Nor are we dealing with abrupt changes marking a rapid decline in the modern Romance verbal systems. For this reason, a closer examination of subordination will serve as a valuable tool in the study of the Romance subjunctive. In my opinion, however, this argument—based on an increasing level of complexity in literate society—is not completely objective or quantifiable, and it leaves us with following questions: How do we define complexity? How do we define hypotaxis?
Is, for example, a finite subordinate clause truly more
“subordinate” than a non-finite Accusative-Infinitive construction? Furthermore, we need to look beyond the obvious presence of subordination and ask ourselves if there are other differences between volo; veniat and volo ut veniat which may be of importance. In my view, it is not the mere presence of subordination, but instead the specific type of subordination technique that ultimately has an effect on mood. As with the well-known shifts from case to prepositions (in Spanish and French), from synthetic to analytic tenses (in Spanish and French), from person/number morphology to subject pronouns (in French), and from gender/number marking to definite articles (in French), we are dealing not only with the addition of a new element, but also with the much more significant, typological shift in word order. 84
2.3.2.3. The Role of Word-Order Change The main characteristic lacking in Harris’ account of subjunctive (in 2.3.2.2 above) is provided by Raible (1992), who presents a thorough analysis of the major changes in subordination techniques that occurred due to typological drift from Proto-Indo-European to Romance. He supports the belief—widely held by Latin grammarians—that one of the original functions of the subjunctive in Classical Latin was to mark subordination.
Raible explains that there are three
basic techniques used to mark subordination: (1) non-finite verbs, (2) “special forms” (Raible 1992: 308) and (3) subordinating conjunctions. Since linking features should ideally be placed between the elements the speaker wishes to link together, Raible argues that a marker of subordination at the end of a subordinate clause (Cl*)10 is most consistent with SOV word order (as are post-determinative case marking, genitive-noun sequencing, etc.) (Raible 1992: 306). Therefore, both subjunctive morphology and non-finite constructions such as Accusative-Infinitive and Ablative Absolute were ideal subordination techniques in Latin SOV sentences.
Raible points out that without such
techniques, SOV word order would be especially troublesome because the recursive structure, (SOV)V is a potential “trap for the hearer” (1992: 307). There was, of course, a gradual shift taking place in Latin from SOV to SVO ordering, and this shift led to a variety of morpho-syntactic changes, one of 10
In Raible’s (1992) analysis, Cl is used as an abbreviation for clause, with an asterix representing the marker of subordination.
85
them involving subordination techniques. In this new sentence type, non-finite verb forms and subjunctive morphology would not be ideal candidates for linking elements, as the marker of subordination would not appear in medial position. Subordinating conjunctions, however, are capable of fulfilling this role quite effectively, resulting in the construction V *Cl. A marker of subordination at the beginning of the clause (along with pre-determinative case marking, noungenitive sequencing, etc.) is, therefore, highly compatible with SVO word order (Raible 1992: 306-314). In conclusion, subjunctive mood can be seen as a subordination technique specifically compatible with OV structures. Along with a shift from OV to VO structures in Latin came a preference for marking subordination pre-verbally (via conjunctions) rather than post-verbally (via mood or non-finite forms). According to Raible, this increase in subordinators resulted in a “syntactical overcharacterization” (1992: 320), and the subjunctive has then become “a second or third order subordinating device—the most important signal being the conjunction at the head of the clause. Consequently, the subjunctive can be reduced in the verbal system” (Raible 1992: 327).
Of course, what has occurred is not a
complete loss of subjunctive, but rather an increase in fluctuation, inconsistency and cross-linguistic variation.
While the syntactic function (i.e. marker of
subordination) of the subjunctive may be redundant in languages with predominantly SVO typology, its modal function cannot be ignored. As Raible 86
explains, “it is precisely the genuine Janus-facedness of the subjunctive which will turn out to be a congenital defect” (311). Raible’s (1992) argument is perhaps the strongest to date with regard to changes in the modal system in Romance languages, and ironically so, since the focus of his work is not mood, but rather subordination. This focus is crucial, as it is specifically the type of subordination technique, and not merely the presence of subordination, which contributes to the problematic nature of the subjunctive mood in Romance languages.
2.3.3. Contributions of Diachronic Research Language-specific diachronic research has explored several interesting variables in the development of the subjunctive mood. It is my belief, however, that these variables (e.g. phonology, language contact) are not primary, but secondary factors contributing to a change already underway.
To determine why the
Romance languages have been susceptible to these external influences requires a broader, comparative approach.
Although several linguists have attempted
typological explanations of this type, many of their ideas have not yet been fully developed in terms of contextualization and/or supporting evidence. It is with this in mind that I will begin my own analysis—based primarily on changes in Latin morpho-syntax—with the goal of identifying an underlying, typological impetus for changes in mood distribution in Romance languages. 87
CHAPTER 3 TYPOLOGICAL DRIFT AND LATIN CONJUNCTIONS 3.1. Introduction As outlined in the introductory chapter, I believe there are three developments of fundamental importance which, together, have contributed to many of the changes in Romance mood: (1) an increase in subordinating elements, (2) the grammaticalization of these subordinating elements, and (3) the synthesis of que and the subjunctive. Furthermore, I interpret these changes as a natural outcome of fluctuations and incongruities in Latin, which should not be viewed as a starting point, but rather an intermediary stage between Proto-Indo-European and modern Romance languages. In this chapter, I will discuss changes in Latin word order, drawing particular attention to the notion of branching as it relates to typological drift. Subsequently, I will present my own analysis of data from three Latin texts (Senatus Consultum de Bacchanalibus, Caesar’s De Bello Gallico and Peregrinatio Egeriae), with the aim of establishing a clear relationship between branching patterns and frequency of conjunctions.
3.2. Latin Word Order Latin word order has traditionally been termed “free,” and Latin grammars generally include an overview of syntax in which word order is treated as 88
insignificant due to Latin’s highly efficient nominal morphology. Consider the following examples from Wheelock’s Latin: (3-1) (a) Lt. puer boy-NOM
(b) Lt. bellae pretty-DAT
(c) Lt. bellae pretty-DAT
(d) Lt. rosam rose-ACC
puellae
bellae
rosam
dat
girl-DAT
pretty-DAT
rose-ACC
give-3SG-PRES-IND
puellae
puer
rosam
dat
girl-DAT
boy-NOM
rose-ACC
give-3SG-PRES-IND
puellae
rosam
dat
puer
girl-DAT
rose-ACC
give-3SG-PRES-IND
boy-NOM
puer
puellae
bellae
dat
boy-NOM
girl-DAT
pretty-DAT
give-3SG-PRES-IND
‘the boy is giving the pretty girl a rose’ (Wheelock 2000: 20) While this characterization is true to a certain extent—as word order was not fixed—it is also the case that the preferred order was clearly verb final (XYV), and that deviations from this order were typically motivated either by stylistic factors, pragmatic considerations (e.g. information structure) or by the presence of specific lexical items (e.g. Lt. enim ‘for’) which had a fixed order within the sentence (see Pinkster 1990: chapter 9; Bauer 1995: 93-97). Taking issues such as these into account, Adams makes the following assessment regarding Latin word order:
89
It is worth stressing finally that the word order of Latin is not genuinely ‘free’ in any useful sense of the term. Though in an inflecting language the elements S, O and V theoretically can—and indeed in practice do—occur in a variety of orders, predominating patterns can be identified: SOV in classical Latin and SVO in late texts. Since variants are usually stylistically marked, there is not free variation of the possible alternatives. (1977: 99) Above all, it is important to keep in mind that the presence of both VO and OV patterning in a language is a sign of ongoing syntactic change and, as a consequence of the dynamic nature of language, “ideal” language types—such as Greenberg’s (1963) I, II, and III—are extremely rare. The variability found in Latin, in particular, is due to a gradual shift away from the OV word order of PIE and toward the VO order characteristic of modern Romance languages. With regard to the OV status of Proto-Indo-European word order, several claims have been made, though with varying degrees of assertiveness. Comrie (1989), for example, rejects the notion that PIE itself was typologically consistent or “ideal” (Comrie 1989: 212-213). Nevertheless, there is substantial evidence that OV was indeed the unmarked word order, since similar—if not identical— verb-phrase patterns are seen in several ancient Indo-European languages (cf. Bauer 1995: section 4.1.1). Consider, for example, the following Vedic Sanskrit (1st millennium BC) sentence from the Śatapathabrāhmaṇa, which serves to illustrate this canonical PIE word order:
90
(3-2)
Ved. víšaḥ villagers-NOM
kṣatríyāya
balíṁ
haranti
prince-DAT
tax-ACC
pay-3PL-PRES-IND
‘the villagers pay taxes to the prince’ (Lehmann 1993: 203) Based on data such as this from Vedic Sanskrit, Hittite, Old Irish, Latin and Greek, Watkins (1964) concludes that the basic PIE word order was #. (E) . . . V#,11 with a stylistically marked variant of #V(E) . . . # (Watkins 1964: 1041). Lehmann (1973, 1974) has also found comparative diachronic evidence to support OV order for PIE. In order to isolate the unmarked order, Lehmann has focused not only on the placement of the verb, but also on features of OV languages that are not easily influenced by literary effect (e.g. comparative constructions, prepositions / postpositions). His analysis of Sanskrit, Hittite, early Greek and early Latin revealed similar word-order patterns: standard-pivotcomparative, genitive-noun, postpositions, along with pre-posed adjectives and relative clauses (see Lehmann 1974: chapter 2). Along similar lines, Luraghi’s (1990) research on Hittite has led her to conclude that while PIE word order was not fixed, SOV may have been the “statistically most frequent order” (Luraghi 1990: 123).
In short, whether word order in PIE was syntactically or
pragmatically determined is an issue that cannot be resolved here. Nevertheless,
11
In this representation, V is a finite verb, while E represents an enclitic pronominal element and # marks sentence boundaries.
91
the essential observation is that OV structures were, in fact, predominant in the proto-language. The shift from OV to VO is a well-attested Indo-European phenomenon. With particular reference to Latin, Adams points out that early texts are consistently OV, while late texts are consistently VO (Adams 1977: 92). Bauer’s (1995) research on the development of word order from PIE to Modern French confirms this trajectory. An unmarked verb-final order is found in Early Latin texts such as Senatus Consultum de Bacchanalibus, Lex Duodecim Tabularum and Columna Rostrata, while VO patterns are seen emerging during the Classical Latin period, ultimately becoming the norm in Late Latin (Bauer 1995: 88-103). Bauer asserts that “evidence from the Peregrinatio strongly suggests the increasing importance of postposed verbal complements: the verbal complement increasingly tends to follow the verb” (Bauer 1995: 101). The examples below illustrate this shift from the older (OV) to the newer (VO) order: (3-3)
Lt.
torcularia
bona
habere
wine presses-ACC
good-ACC-PL
have-PRES-INF
‘to have good wine presses’ (Cato, R.R. 3.2) (3-4)
Lt.
ascendere
montes
ascend-PRES-INF
mountains-ACC
‘to ascend the mountains’ (Pereg. 3.1)
92
3.3. The Role of Branching Although the syntactic change in question is most commonly regarded as a shift from OV to VO word order, a closer examination reveals that the shift taking place is a much broader one and can be characterized in terms of branching. The term branching, according to Bauer, refers to “the linear ordering of the hierarchical elements in a grammatical unit” (1995: 24).
Additionally, the
relationship between these elements may be either one of complementation (i.e. required) or one of modification (i.e. optional). Left-branching and rightbranching structures can be illustrated using brackets, as shown below: (3-5)
a.
[[
b.
[
]
] [
]]
The head assigns a grammatical function to its complement, which either branches to the left (as in 3-5a) or to the right (as in 3-5b). Therefore, we are not dealing exclusively with the placement of the object in relation to the verb, but with the ordering of many constituents including nouns, adjectives, verbal morphology, prepositions, etc. along the lines of Greenberg’s (1963) implicational Universals.
It is useful to keep in mind,
however, that this typological description, like that of Greenberg (1963), is not categorical. Just as languages typically do not represent an ideal “type” (I, II, III), languages also do not tend to be exclusively left or right-branching. Nevertheless,
93
branching is a typologically useful term, in that many languages do indeed exhibit strong patterns in either direction. Turning now to the development of Romance branching patterns, in specific, the archaic left-branching PIE structures have gradually been—and continue to be—replaced by equivalent right-branching structures. The argument that a shift in branching surpasses OV>VO as a fundamental characterization of diachronic Romance typology is supported on chronological grounds.
The
position of the verb in relation to its object was clearly not the “leader,” as it were, in this typological shift. Instead, the change from OV to VO word order ocurred relatively late in the overall movement toward toward right-branching structures (see Bauer 1995: section 4.1.3). Furthermore, this change in branching patterns is attested not only in Romance languages, but throughout Indo-European. In order to shed some light on the motivation behind such a fundamental shift from L-B to R-B, researchers have searched for independent evidence pointing to the advantages of right over left-branching. Generally speaking, L-B structures are believed to be more difficult in terms of processing.12 Levin and Garrett (1990)—taking advantage of the association between the level of difficulty of a linguistic structure and the level of formality—set out to assess the relative difficulty of various sentences by asking native speakers of English to judge the formality of left-branching and right-branching options. The results 12
Justus, for example, refers to “perceptually difficult OV object complementations” (1980: 188).
94
confirmed their hypothesis; left-branching complex sentences (e.g. Because they had had such a long journey, the passengers were tired) were consistently judged to be more formal than their right-branching counterparts (e.g. The passengers were tired because they had had such a long journey).13 Research has also been done from the perspective of first-language acquisition. Bauer (1995) has explored the acquisition of left and right-branching structures in various languages and has found that right-branching structures appear to have a clear advantage. Although the acquisition of both branching types is similar when dealing with basic structures (i.e. simple sentences), Bauer asserts that “left branching causes quasi-insurmountable problems when structures are complex” (1995: 207). For example, instead of producing the left-branching relative construction of their language,14 Turkish children often rely on a rightbranching structure including a marker of subordination and a finite verb. Bauer summarizes her findings as follows: The development of language is therefore oriented and nonarbitrary: complex LB structures have been replaced by their RB equivalents, which are less difficult to acquire and which are mastered more quickly. Child language, therefore, provides independent evidence to demonstrate in what respect right branching shows advantages over left branching. (1995: 211)
13
This reasoning, unfortunately, is somewhat circular. After all, the L-B sentences may be judged as more formal (i.e. difficult) precisely because they do not represent the predominant order in English. A similar study conducted for native speakers of predominantly L-B languages would be very revealing in this regard. 14 The Turkish L-B relative construction consists of a nominal verb form plus a particle (see Bauer 1995: 205).
95
Based on findings such as these, it appears as though language change moves in the direction of earlier-acquired features. It is important to point out, however, that not all languages undergo the same changes nor is the rate of change constant.
For example, Modern English is VO, yet pre-poses its
adjectives. Early Latin was predominantly OV yet already had prepositions. What we are seeing in these cases is diachronic change in action, or in the words of Raible, “die Gleichzeitigkeit des Ungleichzeitigen” ‘the simultaneousness of the non-simultaneous’ (1992: 301). In order to illustrate this more generalized typological drift, I have developed the table below—as several others have done before me—contrasting ten archaic left-branching structures from early Latin texts (Lex Duodecim Tabularum, Senatus Consultum de Bacchanalibus, and Cato’s De Re Rustica) with their modern right-branching counterparts in both Spanish and French:
96
Table 9: Typological Drift and Branching Patterns LEFT-BRANCHING RIGHT-BRANCHING object senatum
PP ad villam
adverb
consoluerunt15
verb
verb
object
consultaron ont consulté
el senado le sénat
verb
verb
PP
viene vient
a la finca a la ferme
venit16 substernatur17
verb
adverb
diligenter
verb
ser cubierto être couvert
cuidadosamente attentivement
adjective
copula
copula
adjective
está est
escrito écrit
noun
adjective
corona couronne
larga longue
noun
genitive
hijo fils
de Lucius de Lucius
degree
adjective
más plus
frío froid
scriptum
est18
adjective
noun
longae
genitive Luci
adjective
19
coronae
noun
20
filius
degree
frigid
-ior21
noun
ending
operari
-orum22
verb
ending
cogno
-vit23
noun
conjunction
aliarum
-que24
15
S. C. Bacch. 1 Cato, R.R. 2.1 17 Cato, R.R. 5.7 18 S. C. Bacch. 22 19 Tab. 10.6 20 S. C. Bacch. 1 21 Cato, R.R. 6.2 22 Cato, R.R. 1.3 23 Cato, R.R. 2.1 24 Cato, R.R. 2.1 16
97
preposition
noun
de d’
obreros ouvriers
pronoun/aux
verb
(él) ha il a
aprendido appris
conjunction
noun
y et
de otros d’autres
The phenomenon presented above illustrates the Mirror Image Principle, which asserts that any two constituents ordered XY in an OV language will be ordered YX in a VO language (Parker 1980: 269-270). Vennemann (1975) has used this apparent preference for symmetry as a justification for language change, which he refers to as the Principle of Natural Serialization,25 while Justus—in a similar fashion—refers to “OV or VO word order harmonics” (1980: 185).
In this
dissertation, it is precisely this typological shift toward right-branching structures which will be taken as the starting point for a chain of events that progressively destabilized the Romance subjunctive.
3.4. Increase in the Frequency of Latin Conjunctions I am now able to address the first of the three developments that led to substantial changes in subjunctive usage in Romance languages. This section will therefore test the first hypothesis of this dissertation: that the typological shift from leftbranching to right-branching structures led to an increase in subordinating conjunctions in Latin.
While Raible’s (1992) analysis (see section 2.3.2.3)
focuses primarily on the shift from OV to VO word order, and thus addresses complement clauses (i.e. subject and object) in particular, this analysis will relate instead to branching as a broader phenomenon. This approach enables us to: 25
The majority of criticism against Vennemann’s (1975) theory is due to his claim that the initial “asymmetries” arise as a result of phonological change, a belief which is no longer widely accepted due to overwhelming chronological counter-evidence (cf. Harris 1978a and Comrie 1989: 214).
98
(1) include a wider range of subordinate clauses instead of limiting ourselves to complement clauses, and (2) examine much earlier texts, since the presence of right-branching (i.e. post-posed) subordinate clauses was an early development in Indo-European, while VO word order was not. I will explain why, theoretically, changes in branching would lead to new preferences in subordination techniques, and will subsequently analyze Latin data to determine if there is, in fact, evidence to corroborate this claim.
3.4.1. Word Order and Subordination Techniques Cross-linguistically, there are a variety of ways in which a speaker can mark subordination,26 and those available to speakers of Latin are as follows: (1) special (modal) forms, (2) non-finite forms, and (3) conjunctions. With these options in mind, we must now consider their specific structures. Both special and non-finite forms rely on morphology alone and result in a clause in which subordination is marked at the end of the subordinate verb (SC*),27 while the use of overt subordinating conjunctions in Latin results in just the opposite (*SC). What I propose is a principle of branching congruency, whereby a leftbranching (i.e. pre-posed) subordinate clause is more consistent with a leftbranching subordination technique, and a right-branching (i.e. post-posed) 26
Lehmann has identified twelve characteristics of subordinate clauses including, but not limited to, connecting words, mood shift and clause arrangement (1980: 141). 27 The abbreviations MC and SC are used here to represent the main clause and the subordinate clause, while the asterix represents the marker of subordination.
99
subordinate clause is more consistent with a right-branching subordination technique. The following illustrations in (3-6) represent left-branching structures, while the illustrations in (3-7) represent right-branching structures: (3-6)
(3-7)
a.
[ [ SC ]
MC ]
b.
[ [ SC ]
* ]
a.
[ MC
[ SC ] ]
b.
[ *
[ SC ] ]
Consider, for example, the placement of the marker of subordination in the following complex sentences. The first illustrates an OV structure in Japanese and the second, a VO structure in Ancient Egyptian: (3-8)
Jp.
Boku
wa
Taroo
ga
kekkonsita
koto
I
TOP
Taroo
NOM
marry-PAST
NMNL
o
sitte
iru
ACC
know-GER
be-NON-PAST
‘I know that Taroo got married’ (Justus 1980: 188) (3-9)
Eg.
rḫ · kwi
ntt
ḥtp
· f
ḥr
· s
know-PERF-I
PTC
at-rest
he
face
(of) it
‘I knew that he would be pleased about it’ (Justus 1980: 190) In the Japanese example, a left-branching subordinate clause appears with a leftbranching subordination technique, following the congruent branching patterns
100
shown in (3-6), while in the Egyptian example, a right-branching subordinate clause combines with a right-branching subordination technique, as in (3-7). Returning to Latin, then, one can see how this strong preference for typological consistency may have played a role in the development of subordination techniques in Latin. As long as the subordinate clause appeared before the main clause, the marker of subordination was likely to appear at the end of the subordinate clause, solely via verbal morphology (subjunctive mood or non-finite forms). However, when the subordinate clause came to follow the main clause, markers of subordination that preceeded the subordinate clause (i.e. conjunctions) would be preferred, thus ensuring branching congruency.
3.4.2. Analysis of Latin Texts The goal of this section is to test the validity of the hypothesis developed above; in other words, to determine whether or not Latin subordinating conjunctions are more closely associated with right-branching subordinate clauses (i.e. main clause + subordinate clause). The aim of this section is not chronological, since a shift to right-branching is an underlying assumption of this dissertation. The focus is, instead, on subordination techniques as they relate to word order. In the analysis to follow, left-branching subordination techniques will include: (1) infinitival constructions, (2) participial constructions, and (3) subjunctive mood with no additional marker of subordination. 101
Right-
branching subordination techniques will include all overt pre-posed subordinating conjunctions. Illustrative examples of each of these possiblilities are provided below: L-B Technique: Infinitival Constructions (3-10) Lt.
fores
pultare
nescis
doors-ACC
knock-PRES-INF
not know-2SG-PRES-IND
‘you don’t know how to knock at a door’ (Pl., Bacch. 581 in Bennett 1910: 403) (3-11) Lt.
venerat
aurum
petere
come-3SG-PLUPERF-IND
gold-ACC
seek-PRES-INF
‘he had come to seek gold’ (Pl., Bacch. 631 in Bennett 1910: 419) (3-12) Lt.
adsimulabo
me
esse
pretend-1SG-FUT-IND
1SG-PRON-ACC
be-PRES-INF
ebrium drunk-ACC-M-SG
‘I will pretend to be drunk’ (Pl., Amph. 999 in Bennett 1910: 367) Example (3-10) contains a complementary infinitive, while (3-11) contains an infinitive of purpose, a usage which developed early due to the notion of direction associated, from the outset, with all infinitives (cf. Bennett 1910: 366). Finally, example (3-12) contains an Accusative-Infinitive, which was widespread in Latin.
102
In this construction, the subordinate verb is an infinitive, and the subject of this infinitive appears in the accusative. The second type of construction including a left-branching subordination technique is a participial construction, as shown below: L-B Technique: Participial Constructions (3-13) Lt.
Romam
videntes,
viri
Rome-ACC
see-PRES-PART-NOM-PL
men-NOM
gaudebant rejoice-3PL-IMPERF-IND
‘seeing Rome, the men rejoiced’ (Wheelock 2000: 156) (3-14) Lt.
Roma
visa,
viri
Rome-ABL
see-PERF-PART-ABL-SG
men-NOM
gaudebant rejoice-3PL-IMPERF-IND
‘with Rome having been seen / when Rome was seen, the men rejoiced’ (Wheelock 2000: 156) While (3-13) represents a standard participial phrase, (3-14) is a specific sub-type referred to as an Ablative Absolute. This particular construction consists of a noun and a participle in the ablative case. It can be distinguished from the standard participal clause in that “the ablative absolute is always self-contained” (Wheelock 2000: 155), as there is no additional reference to the ablative noun in the main clause. 103
Finally, subjunctive morphology alone (i.e. without any additional marker of subordination) offers a third instance of left-branching markers of subordination: L-B Technique: Subjunctive Mood (3-15) Lt.
taceat
oportet
be silent-3SG-PRES-SUBJ
be proper-3SG-PRES-IND
‘it is necessary to be quiet’ (Bennett 1910: 210) Subordination in this case is marked at the end of the verb (as in infinitives and participles) and this, alone, serves to mark the subordinate status of the clause. In terms of right-branching, there is only one R-B subordination technique in Latin: the conjunction, which appears before the subordinate verb. Consider the following examples: R-B Technique: Subordinating Conjunction28 (3-16) Lt.
mane
dum
scribit
wait-2SG-IMP
until
write-3SG-PRES-IND
‘wait until he writes’ (Pl., Bacch. 737 in Bennett 1910: 98) (3-17) Lt.
hoc
dicit
ut
eos
this-ACC-N
say-3SG-PRES-IND
that
them-ACC
28
I have chosen to include circumstantial/adverbial clauses in my analysis—in spite of their tendency to rely more heavily on pragmatic factors (cf. Luraghi 1990: 78-79 and Diessel 2001: 448)—since the majority of finite subordinate clauses in Latin were, in fact, circumstantial. I have, however, excluded relative clauses from this analysis since their placement within the sentence relies on quite different factors, and they depend on pronouns rather than conjunctions.
104
iuvet help-3SG-PRES-SUBJ
‘he says this to help them’ (Wheelock 2000: 189) Therefore, in contrast to examples (3-10)-(3-15), subordination in examples (3-16)-(3-17) is marked pre-verbally. This distinction will be the focus of the analysis to follow, where the placement of the marker of subordination (L-B or R-B) will be related to the placement of the subordinate clause (L-B or R-B). For this analysis, I have selected three Latin texts: Senatus Consultum de Bacchanalibus, Caesar’s De Bello Gallico, and Egeria’s Peregrinatio, which represent archaic, classical, and late Latin, respectively. For each of the texts, I will present examples of all four possible combinations shown below: Type A: L-B subordinate clause + L-B subordination technique Type B: L-B subordinate clause + R-B subordination technique Type C: R-B subordinate clause + L-B subordination technique Type D: R-B subordinate clause + R-B subordination technique After the presentation of illustrative examples, I will report my quantitative findings in summary form. My expectation is that there will be a high degree of branching congruency; that is to say, a higher frequency of conjunctions in R-B subordinate clauses, and a greater amount of infitives, participles and subjunctives
105
in L-B subordinate clauses. In short, I expect that, of the four combinations presented above, types A and D will be more frequent than types B and C.
3.4.2.1. Senatus Consultum de Bacchanalibus This senate decree inscribed on a bronze tablet served to prohibit the bacchanalia—mystic festivals for the god Bacchus—throughout Italy. Since the text was written in 186 B.C., the language is quite archaic. Nevertheless, a shift toward right-branching structures was already underway. There are examples of both left-branching and right-branching subordinate clauses, as well as both types of subordination techniques: verbal morphology and conjunctions. Consider the illustrative examples below: TYPE A: L-B subordinate clause + L-B subordination technique (3-18)
[[figier]
ioubeatis]
fix-PRES-PASS-INF
command-2PL-PRES-SUBJ
‘you are to give orders that it be set up’ (S.C. Bacch. 27) (3-19)
[[neiquis
eorum
Bacanal
none-NOM
them-GEN
shrine of Bacchus-ACC
habuise]
velet]
have-PERF-INF
wish-3SG-PRES-SUBJ
‘let none of them be minded to have a shrine of Bacchus’ (S.C. Bacch. 3)
106
TYPE B: L-B subordinate clause + R-B subordination technique (3-20) Lt.
[[sei
ques
esent,
if
REL-PRON-NOM-SG
be-3PL-PRES-SUBJ
quei
arvorsum
ead
REL-PRON-NOM-SG
against
this-ABL
fecisent,
quam
suprad
do-3PL-PLUPERF-SUBJ
as
above
scriptum
est],
write-PERF-PART-NOM-SG
be-3SG-PRES-IND
eeis them-DAT
rem
caputalem
thing-ACC
capital-ACC
faciendam
censuere]
make-GERV
decree-3PL-PERF-IND
‘if there are any who transgress against the decrees set out above, a capital charge is to be brought against them’ (S.C. Bacch. 24-25) (3-21) Lt.
[[exstrad
quam sei
quid
ibei
except
unless
INDEF-PRON
there
sacri
est ], . . .
faciatis
sacred-NOM
be-3SG-PRES-IND
make-2PL-PRES-SUBJ
utei
dismota
sient]
so that
abolish-PERF-PART-PL
be-3PL-PRES-SUBJ
‘unless they contain something sacred, . . . see to it that they are disbanded’ (S.C. Bacch. 28-30)
107
TYPE C: R-B subordinate clause + L-B subordination technique (3-22) Lt.
[deicerent
[necesus
ese]]
say-3PL-IMPERF-SUBJ
necessary
be-PRES-INF
‘they say that it is necessary’ (S.C. Bacch. 4) TYPE D: R-B subordinate clause + R-B subordination technique (3-23) Lt.
[faciatis
[utei
make-2PL-PRES-SUBJ
so that
dismota
sient]]
abolish-PERF-PART-PL
be-3PL-PRES-SUBJ
‘see to it that they are disbanded’ (S.C. Bacch. 30) (3-24)
[quisquam
fecise
velet, . . .
no one
do-PERF-INF
want-3SG-PRES-SUBJ
[nisei
praitorem
urbanum
unless
magistrate-ACC
urban-ACC
adieset]] go-3SG-PLUPERF-SUBJ
‘let none of them do (this) unless they have gone to the city magistrate’ (S.C. Bacch. 16-17) The following table takes into account all twenty-six subordinate clauses found in the text and summarizes the distribution of morphological marking and conjunctions among both left-branching subordinate clauses (SC+MC) and rightbranching subordinate clauses (MC+SC): 108
Table 10: Subordination Patterns in Senatus Consultum de Bacchanalibus morphology (L-B) conjunction (R-B) subord. + main (L-B)
(A) 11 / 14 (79%)
(B) 3 / 14 (21%)
main + subord. (R-B)
(C) 1 / 12 (8 %)
(D) 11 / 12 (92 %)
The results here are as expected. There are signs of both branching patterns in the text, and all four subordination types (A-D) are, indeed, present. Nevertheless, the combinations are not in any way random, as there appears to be a clear preference for branching congruency in the selection of subordination techniques. Subordinate clauses marked solely via verbal morphology appear almost exclusively in pre-posed subordinate clauses, while subordinating conjunctions are strongly preferred in post-posed subordinate clauses.
3.4.2.2. De Bello Gallico Written during the 1st century B.C., this third-person narrative by Julius Caesar is divided into eight books and recounts the events that ocurred during the Gallic Wars, from 58 B.C. to 51 B.C. This series is typically considered to be a prime example of Classical Latin, on account of its precise syntax, polished style and clarity of expression. My analysis will include the first one hundred subordinate clauses from Book I of De Bello Gallico. Below are examples of each of the possible four combinations (Types A-D):
109
TYPE A: L-B subordinate clause + L-B subordination technique (3-25) Lt.
[[sese
potiri
3PL-REFL-PRON-ACC
become master of-PRES-INF
posse]
sperant]
be able-PRES-INF
hope-3PL-PRES-IND
‘they hope to be able to gain mastery’ (Caes., D.B.G. 1.3) (3-26) Lt.
damnatum
[[poenam
condemn-PERF-PART-NOM-SG
punishment-ACC
sequi]
oportebat]
follow-PRES-INF
need-3SG-IMPERF-IND
‘it was necessary that the punishment overtake him if condemned’ (Caes., D.B.G. 1.4) TYPE B: L-B subordinate clause + R-B subordination technique (3-27) Lt.
[[ne
causam
diceret],
that not
case-ACC
say-3SG-IMPERF-SUBJ
se
eripuit]
3SG-REFL-PRON-ACC
rescue-3SG-PERF-IND
‘he saved himself from the necessity of pleading his case’ (Caes., D.B.G. 1.4) (3-28) Lt.
tamen,
[[ut
spatium
yet
so that
space of time-NOM
intercedere
posset] . . .
intervene-PRES-INF
be able-3SG-IMPERF-SUBJ
110
legatis
respondit . . .]
ambassadors-DAT
reply-3SG-PERF-IND
‘yet, so that a period of time might intervene, he replied to the ambassadors . . .’ (Caes., D.B.G. 1.7) TYPE C: R-B subordinate clause + L-B subordination technique (3-29) Lt.
pontem
qui
bridge-ACC
REL-PRON-NOM-SG
erat
ad
Genavam
be-3SG-IMPERF-IND
at
Geneva-ACC
[iubet
[rescindi]]
order-3SG-PRES-IND
break down-PRES-PASS-INF
‘the bridge which was near Geneva he orders to be broken down’ (Caes., D.B.G. 1.7) (3-30) Lt.
[maturat
ab
urbe
[profisci]]
hasten-3SG-PRES-IND
from
city-ABL
set out-PRES-INF
‘he makes haste to leave the city’ (Caes., D.B.G. 1.7) TYPE D: R-B subordinate clause + R-B subordination technique (3-31)
[persuasit,
[ut
de
persuade-3SG-PERF-IND
that
from
finibus
suis
cum
boundaries-ABL
REFL-PRON-ABL-PL
with
111
omnibus
copiis
exirent]]
all-ABL-PL
resources-ABL
exit-3PL-IMPERF-SUBJ
‘he persuaded them to migrate from their territories with all their possessions’ (Caes., D.B.G. 1.2) (3-32)
[non
esse
dubium,
[quin
NEG
be-PRES-INF
doubtful-ACC-SG
that
totius
Galliae
plurimum
all-GEN-SG
Gaul-GEN-SG
most
Helvetii
possent]]
Helvetii-NOM-PL
be able-3PL-IMPERF-SUBJ
‘there was no doubt that the Helvetii were the most powerful of all the Gauls’ (Caes., D.B.G. 1.3) The reader of this text will immediately realize why—aside from its date of publication—De Bello Gallico is considered a classical text. subordination
techniques
such
as
Accusative-Infinitives
Left-branching and
participial
constructions are abundant. Nevertheless, the typological shift underway is also apparent, as right-branching (i.e. post-posed) subordinate clauses are quite prevalent (37%) as well: Table 11: Subordination Patterns in De Bello Gallico morphology (L-B)
conjunction (R-B)
subord. + main (L-B)
(A) 49 / 63 (78 %)
(B) 14 / 63 (22%)
main + subord. (R-B)
(C) 8/37 (22 %)
(D) 29/37 (78 %)
112
The table above illustrates that this classical text also exhibits consistent branching congruency.
Morphological marking alone is typically found in
pre-posed (L-B) subordinate clauses, while conjunctions are, without a doubt, the preferred subordination technique for post-posed (R-B) subordinate clauses.
3.4.2.3. Peregrinatio Egeriae This late Latin text is Egeria’a first-person account of her pilgrimage to the Holy Land during the 4th century A.D. She describes her experiences on her four-year journey in great detail, including acquaintances, religious practices and natural wonders. As with Caesar’s De Bello Gallico, I will analyze the first one hundred subordinate clauses of this text, and will provide illustrative examples followed by a quantitative summary: TYPE A: L-B subordinate clause + L-B subordination technique (3-33) Lt.
sed
[[ non
ipsa
parte
but
NEG
that same-ABL
part-ABL
exire]
habebamus],
exit-PRES-INF
have-1PL-IMPERF-IND
qua
intraveramus
REL-PRON-ABL-SG
enter-1PL-PLUPERF-IND
‘but we did not have to exit on the same side that we had entered’ (Pereg. 4.5)
113
(3-34) Lt.
nam
[[ mihi
credat]
for
me-DAT
believe-3SG-PRES-SUBJ
volo]
affectio
vestra
want-1SG-PRES-IND
affection-NOM
your-NOM-SG
‘for I hope that you (your affection) will believe me’ (Pereg. 7.3) TYPE B: L-B subordinate clause + R-B subordination technique (3-35) Lt.
[[quod
ingrederis],
plures
as
enter-2SG-PRES-IND
many-NOM-PL
sunt] be-3PL-PRES-IND
‘as you enter, (you see that) there are more than one’ (Pereg. 2.5) (3-36) Lt.
[[quia
iam
sera
erat],
since
already
late
be-3SG-IMPERF-IND
oblationem
facere
non
potuimus]
offering-ACC
make-PRES-INF
NEG
be able-1PL-PERF-IND
‘since it was already late, we couldn’t make the offering’ (Pereg. 4.8) TYPE C: R-B subordinate clause + L-B subordination technique (3-37) Lt.
tunc
[coepi
[eos
then
begin-1SG-PERF-IND
them-ACC
rogare]] beg-PRES-INF
‘then I began to ask them’ (Pereg. 3.7) 114
(3-38) Lt.
illud
autem
vos
this-ACC
however
you-ACC-SG
[volo
[scire]]
want-1SG-PRES-IND
know-PRES-INF
‘this, however, I want you to know’ (Pereg. 3.8) TYPE D: R-B subordinate clause + R-B subordination technique (3-39) Lt.
[rogavimus
presbyteros,
[ut
beg-1PL-PERF-IND
priests-ACC
that
ibi
fieret
oblatio]]
there
be made-3SG-IMPERF-SUBJ
offering-NOM
‘we asked the priests to make the offering there’ (Pereg. 4.8) (3-40) Lt.
[montes
cum
infinito
labore
mountain-NOM-PL
with
infinite-ABL
work-ABL
ascenduntur,
[quoniam
non
eos
ascend-3PL-PRES-PASS
since
NEG
them-ACC
subis
lente
et
lente
per
climb-2SG-PRES-IND
slowly
and
slowly
by
girum] circle-ACC
‘these mountains are ascended with infinite toil, for you cannot go up slowly in a spiral’ (Pereg. 3.1) In this much later text, there are many more examples of right-branching subordinate clauses (i.e. main clause + subordinate clause), as fifty-nine out of
115
one hundred subordinate clauses appeared after the main clause. Nevertheless, we still see a strong tendency toward branching congruency, as the majority of conjunctions appear in post-posed subordinate clauses. Table 12: Subordination Patterns in Peregrinatio Egeriae morphology (L-B) conjunction (R-B) subord. + main (L-B)
(A) 23 / 41 (56%)
(B) 18 / 41 (44%)
main + subord. (R-B)
(C) 14 / 59 (24%)
(D) 45 / 59 (76%)
There is a peculiar feature of this data that, in my opinion, needs to be addressed further. Compared with the two previous texts (Senatus Consultum de Bacchanalibus and De Bello Gallico), the Peregrinatio displays a much greater quantity of Type B subordination (i.e. conjunctions in subordinate clauses which precede the main clause). Upon reading this text, the reason becomes clear: there is an extremely high number of temporal clauses serving as a sequential list of Egeria’s activities and experiences on her journey. These clauses tend to come at the beginning of the sentence due mainly to pragmatic reasons. Furthermore, the scope of temporal clauses tends to be wider, since in addition to sentence-level linking between the main and subordinate clauses, they also provide cohesion between the clauses and the preceding sentence. Several linguists have observed this phenomenon and offer similar justifications. Luraghi, for example, argues that temporal clauses provide the setting for the predication (1990: 79), while Diessel explains that they offer a framework or orientation (2001: 448). Perrochat 116
sees adverbial clauses as being a more autonomous type of subordinate clause and—referring to the rigidity of word order—concludes: “l’on voit la stabilité de la construction diminuer à mesure que le caractère de subordination s’affaiblit” (Perrochat 1926: 58).
This characteristic of temporal clauses, and their
idiosyncratic abundance in the Peregrinatio, in no way negates the claims made here. In this late Latin text, explicit conjunctions are present in only forty-four percent of left-branching subordinate clauses, but in seventy-six percent of rightbranching subordinate clauses. This difference in percentages is substantial and, as a result, carries with it significant implications.
3.5. Observations and Conclusion In this chapter, I have provided an overview of the notion of branching, which I believe to be at the center of many of the morpho-syntactic developments that have occurred in the history of Romance languages. My hypothesis regarding branching
congruency—the preference
techniques
in
right-branching
for
subordinate
right-branching subordination clauses,
and
left-branching
subordination techniques in left-branching subordinate clauses—was tested by analyzing data from three Latin texts: Senatus Consultum de Bacchanalibus (Old Latin), Caesar’s De Bello Gallico (Classical Latin) and Egeria’s Peregrinatio (Late Latin).
117
The results from all three texts are, indeed, consistent with my hypothesis. The development of the conjunction system in Romance languages is not, then, an isolated development. On the contrary, one cannot ignore the fact that it is influenced by changes in word order, as there appears to be a definite correlation between branching and frequency of explicit subordinating conjunctions. However, one must always keep in mind that these are merely tendencies related to an innate preference for congruency, and therefore, there are examples in all three texts of L-B/R-B (Type B) and R-B/L-B (Type C) combinations. The presence of conjunctions in pre-posed subordinate clauses is accounted for almost entirely by temporal conjunctions, whose placement relies more heavily on textual and pragmatic considerations. Nevertheless, a clear pattern emerges, and one can see why the number of conjunctions would increase following an increase in the number of post-posed subordinate clauses.29 In order to briefly illustrate how—in spite of a moderate amount of pre-posed adverbial clauses (Type B) and post-posed non-finite forms (Type C)— there was an overall increase in subordinating conjunctions, consider the following three common structural changes (where subordinate clauses are in bold):
29
Herman (1989) found a similar pattern in his analysis of complement clauses. While AcI constructions were found both before and after the main verb, finite clauses with quod and quia were only found following the main verb.
118
Table 13: L-B and R-B Subordinate Structures LEFT-BRANCHING
[[faciendam]
censuere]
RIGHT-BRANCHING
S.C. Bacch. 25 [[habere]
Pereg. 4.8
arbitrabantur]
[quia videbantur]]
[scire
Caes., D.B.G. 1.2 [[rebus comparatis]
[ut fieret]]
[rogavimus
Pereg. 3.8
diem dicunt]
[fuit Moyses
Caes., D.B.G. 1.6
[cum ascendisset]]
Pereg. 3.7
The complex sentences above represent archaic left-branching and modern rightbranching structures, respectively.30 In the first set of examples, we see final clauses, expressed by a gerundive in Senatus Consultum de Bacchanalibus, and later, by a conjunction in Peregrinatio Egeriae. The second set of examples includes basic object clauses, expressed via an Accusative-Infinitive construction in Caesar but by a conjunction in the Peregrinatio. Finally, the third set of examples have subordinate clauses with temporal value, expressed by an Ablative Absolute construction in Caesar but again by a conjunction in the Peregrinatio. These examples are extremely important as they illustrate not only an increase in conjunctions, but more importantly, an increase in conjunctions associated with a shift from left to right-branching. In fact, this claim regarding 30
These examples parallel the table on page 97, as they represent the same typological shift in branching.
119
word order and explicit conjunctions is borne out in the data. Thirty-seven out of one-hundred subordinate clauses in Caesar are post-posed, as opposed to fiftynine out of one-hundred in the Peregrinatio. Similarly, only forty-three out of one hundred subordinate clauses from Caesar’s De Bello Gallico have conjunctions, while conjunctions are present in sixty-three out of one hundred subordinate clauses in Peregrinatio Egeriae. The tendency uncovered in this analysis, while not categorical, provides us with the evidence necessary to conclude that a shift in branching patterns did indeed contribute to an increase in subordinating conjunctions, a development whose consequences will be explored in the following chapter dealing with the grammaticalization of Latin and Romance subordinating elements.31
31
From this point onward, it will be necessary to distinguish between different types of subordinating elements. I will use the term conjunction to refer to subordinating elements that introduce subordinate clauses. The term complementizer will refer specifically to subordinating elements that introduce complement (i.e. subject or object) clauses. A highly generalized subordinating element with a wide range of functions (i.e. relative pronoun, conjunction, complementizer) will be called a generalized subordinator.
120
CHAPTER 4 GRAMMATICALIZATION OF SUBORDINATING ELEMENTS 4.1. Introduction The previous chapter explored the gradual shift toward right-branching structures and the concomitant increase in the frequency of explicit subordinating conjunctions. This more prevalent right-branching subordination technique relied on a well-established, yet somewhat problematic, Latin conjunction system. The focus of this chapter, then, will be the way in which this conjunction system developed, ultimately resulting in a completely restructured inventory of forms. Furthermore, although the processes discussed in this chapter involve minor categories (categories other than V, N, and A), there is a discernible change in the nature of these conjunctions which I believe can be analyzed quite effectively within the grammaticalization framework. In this chapter I will first briefly characterize the Classical Latin conjunction system.
Subsequently, I will provide an overview of the
grammaticalization framework to be used, followed by a detailed analysis of the Latin subordinating elements relevant to this discussion.
The analysis will
include seven of the most widely accepted processes associated with grammaticalization and will include illustrative data from Latin, Old Spanish and Old French texts.
121
4.2. The Latin Conjunction System It is clear, from even a cursory glance at the vast inventory of conjunctions in Latin, that the forms are extremely opaque. Consider the following short list of conjunctions: cum, ut, quod, ne, dum, donec, velut si. One immediately notices that the paradigm lacks morphological regularity; that is to say, there is no visible structural similarity between the forms. Related to this notion of opacity is the following
observation
by
Herman:
“Les
nombreuses
conjonctions
de
subordination du latin classique forment une série atomisée, constituée d’éléments qui . . . sont plus ou moins isolés les uns des autres” (1963: 20). Furthermore, speakers of Latin had at their disposal an extremely wide range of conjunctions. One could argue that this variety may have made possible a higher degree of expressivity. However, it is clearly not an economical system, as it represents the opposite of a cognitively ideal one-to-one relationship between form and function. Consider the following categories of conjunctions along with their most common corresponding forms and the mood(s) of the finite verbs they introduce (compiled from Bennett 1910, Leiper 1913, Herman 1963 and Gildersleeve and Lodge 1997):
122
Table 14: Latin Conjunction System CATEGORY FORM Causal
Temporal
Final (Purpose)
Consecutive (Result)
MEANING
MOOD
quando
because, since
I
quia
because, since
I/S
quoniam
because, since
I/S
quod
because, since, in that
I/S
cum
because, since, as
S
postquam
after
I
ut
when
I
ubi
when
I
quando
when
I
cum
when
I/S
simul ac
as soon as
I
ut primum
the first moment that
I
cum primum
the first moment that
I
dum
until, as long as, while
I/S
donec
until, as long as
I/S
quoad
until, as long as
I/S
priusquam
before
I/S
antequam
before
I/S
ut
that, in order that
S
quo
that thereby
S
ut eo
that thereby
S
ne
that not, lest
S
ut
that, so that
S
ut non
that, so that, as not
S
quin
that, so that, as not
S
123
CATEGORY Conditional
Comparison
Conditional Comparison
Concessive / Adversative
FORM
MEANING
MOOD
si
if
I/S
si non
if not
I/S
nisi
unless
I/S
ut
as
I/S
sicut
as, as for example, as just
I/S
velut
as, as for example, as just
I/S
quam
as, than
I/S
atque
as, than
I/S
ut si
as if
S
velut si
as if
S
ac si
as if
S
quam si / quasi
as if
S
tamquam si
as if
S
tamquam
as if
S
quamquam
although
I
etsi
although, even now if
I/S
etiamsi
although
I/S
tametsi
although
I/S
quamvis
however much, although
S
quantumvis
to what amount you choose
S
ut
although, granted that
S
ne
although not
S
cum
although
S
licet
although
S
124
This table effectively illustrates that Latin exhibits overlap in both directions; not only are there functions with several competing forms, but there are also forms with multiple corresponding functions. For example, one of the most widespread conjunctions—ut—is used in various contexts with drastically different meanings. The following Latin data demonstrate several of these quite distinct uses: (4-1)
Lt.
inventa
sunt
specula,
PERF-PART-NOM-PL
be-3PL-PRES-IND
mirrors-NOM
ut
homo
ipse
se
so that
man-NOM
himself-NOM
3SG-REFL-PRON-ACC
nosset recognize-3SG-PLUPERF-SUBJ
‘mirrors were invented, (in order) to make man acquainted with himself’ (Sen., N.Q. 1.17,4 in Gildersleeve and Lodge 1997: 345) (4-2)
Lt.
Pompeius
ut
equitatum
suum
Pompey-NOM
when/as
cavalry-ACC
his-ACC
pulsum
vidit,
acie
beat-PERF-PART-ACC
see-3S-PERF-IND
battle line-ABL
excessit leave-3SG-PERF-IND
‘when/as Pompey saw his cavalry beaten, he left the line of battle’ (Caes., D.B.C. 3.94,5 in Gildersleeve and Lodge 1997: 360) (4-3)
Lt.
ut
desint
vires,
granted that
be lacking-3PL-PRES-SUBJ
strength-NOM-PL
tamen
est
laudanda
nevertheless
be-3SG-PRES-IND
praise-GERV
125
voluntas will-NOM
‘granted that strength be lacking, nevertheless you must praise my good will’ (Ov., Pont. 3.4,79 in Gildersleeve and Lodge 1997: 392-3) The examples above reveal the polyvalent nature of Latin conjunctions, which include—but are not limited to—temporal, causal, final, and concessive functions. It is, thus, my belief that the Latin conjunction system, in terms of opacity and iconicity, is far from ideal.
Not only do the forms themselves appear
unmotivated and isolated from each other (at least synchronically), but most of the conjunctions are able to operate in several, often unrelated, clause types. It is not surprising, therefore, that such a system would be easily influenced by common speaker-driven forces such as analogy, leveling and grammaticalization. What occurred, in fact, was a restructuring of the conjunction system from Latin to modern Romance languages, involving two essential points. Firstly, in addition to the many adverbial conjunction types that existed in Latin (i.e. temporal, causal, conditional, etc.), a new category emerged: the complementizer. Secondly, one form in particular has taken over this function and has served as the basis of an entirely new set of Romance conjunctions. I believe that the processes involved
in
this
development
can
grammaticalization framework.
126
be
better
understood
within
the
4.3. Overview of the Grammaticalization Framework The grammaticalization framework is able to account for processes whereby linguistic forms experience an increase in grammatical function along with a decrease in lexical content. While early work in the field focused, primarily, on the more salient development of open-class elements (i.e. nouns and verbs) into grammatical morphemes, the framework is now also frequently applied to closedclass elements, which—following the same principles—can become even further grammaticalized. In addition, changes occur on many levels including semantics, morphology and syntax. In short, grammaticalization is simultaneously a formal and semantic process, one which can be described as the “driving force of so much morphosyntactic change” (Haspelmath 1999: 1064). According to Heine, Claudi and Hünnemeyer (1991), an awareness of the distinction between major and minor categories can be traced to 10th century China, and the view that all function words develop from content words dates back to the 14th-century writings of Zhou Bo-qi (cf. Heine, Claudi and Hünnemeyer 1991: 5).
In the Western world, the notions which would be
essential to modern theories of grammaticalization were developed by scholars during the 18th and 19th centuries. Although a precise characterization of the framework had not yet been developed at this time, scholars undoubtedly had a conceptual grasp on the fundamental nature of the diachronic development of grammatical forms. Gabelentz, for example, pointed out the tendency for words 127
to fade or bleach over time, and Humboldt argued that grammar develops via concrete words, which pass through four stages on their way to affixes (Hopper and Traugott 1993: 19-20). These diachronic tendencies were given full attention by Meillet in his 1912 article, “L’évolution des formes grammaticales,” in which he used the term “grammaticalisation” (1958 [1912]: 133) to refer to “le passage d’un mot autonome au rôle d’élément grammatical” (Meillet 1958 [1912]: 131). Perhaps the most significant contribution of his essay is the clear—and fundamentally important—distinction made between grammaticalization and analogy. Unlike analogical change, which is purely formal, the effects of grammaticalization have far-reaching implications in terms of the underlying linguistic system, since the process is capable of creating new grammatical categories (e.g. Romance definite articles, conditional tense, etc.). In the words of Meillet: Tandis que l’analogie peut renouveler le détail des formes, mais laisse le plus souvent intact le plan d’ensemble du système existant, la “grammaticalisation” de certains mots crée des formes neuves, introduit des catégories qui n’avaient pas d’expression linguistique, transforme l’ensemble du système. (Meillet 1958 [1912]: 133) In spite of the fact that a basic theory of grammaticalization has existed for roughly a century, it is only during the last few decades that it has truly flourished. Before the 1970’s, grammaticalization was not often adopted as a framework within which to study language, due primarily to the overwhelming
128
popularity of
generative linguistics. Hopper and Traugott argue that
“grammaticalization presents a challenge to approaches to language which assume discrete categories embedded in fixed, stable systems” (1993: 25). A similar commentary is provided by Fischer and Rosenbach, who explain that grammaticalization was not generally of interest to followers of generative linguistics, since the framework focuses on processes rather than states (Fischer and Rosenbach 2000: 9). However, by the 1980’s, grammaticalization had reemerged as a topic of interest among linguists, demonstrated by a marked increase in the number of applications, research questions and sub-fields. As a result of this research, many grammaticalization processes in Romance languages have been identified, including: A) noun > adverbial suffix [Lt. mente ‘mind’ > Sp. rápidamente, Fr. rapidement] B) noun > negative particle [Lt. passum ‘step’ > Fr. je ne sais pas] C) lexical verb > auxiliary verb [Lt. librum habeo ‘have’ > Sp. he cantado, Fr. j’ai chanté] D) lexical verb > inflectional morphology [Lt. librum habeo ‘have’ > Sp. cantaré, Fr. chanterai] E) demonstrative pronoun > definite article [Lt. illum ‘this’ > Sp. el, Fr. le] 129
The recent abundance of grammaticalization studies32 has undoubtedly elucidated various issues, but has also resulted in an overwhelmingly vast array of concepts and terminology to contend with. Therefore, I feel it is necessary to clarify my perspective with regard to three of grammaticalization’s most important components: (1) the source/target dichotomy, (2) the notion of cline and (3) the various micro-processes involved in grammaticalization. The terms source and target are typically used to refer to the older— typically lexical—items and the grammaticalized elements into which they develop.
While these terms are relatively clear and do indeed indicate a
diachronic change or transition, they are also somewhat misleading.
It is
important to recognize that the term source must be used in a relative sense, as we do not have access to absolute starting points, but rather arbitrary starting points with which we compare future linguistic elements. For example, the source of Fr. le is Lt. illum, yet Lt. illum must surely have a source of its own, albeit outside the scope of study. The term target has similar shortcomings, since we are not necessarily dealing with a true endpoint, as the item in question may continue to grammaticalize. Furthermore, the term target implies an intention that is not, in fact, part of natural diachronic processes, as a language does not have intentions 32
Many of these are comprehensive monographs (e.g. Heine, Claudi and Hünnemeyer 1991, Hopper and Traugott 1993, C. Lehmann 1995, Klausenburger 2000). There have also been attempts to reconcile grammaticalization with generative syntax (e.g. Roberts and Roussou 2003, van Gelderen 2004), as well as inventories of grammaticalization clines (e.g. Heine and Kuteva 2002) and studies on specific linguistic elements (e.g. Kilroe 1994, Epstein 1994, Bauer 2003 & 2006).
130
or goals of its own. While it is true that the speakers of a language do have intentions (e.g. clarity of expression, ease of articulation, etc.), one would not say that speakers ever intend to grammaticalize a linguistic element. In spite of these unfortunate—though unintended—connotations of the terms source and target, I will continue to use this terminology for the sake of convention, as it is so frequent in the literature. With regard to the grammaticalization process itself, it transforms the source element into the target element along a path typically referred to as a cline. A prototypical, yet greatly simplified, grammaticalization cline is shown below: content item > grammatical word > clitic > inflectional affix (Hopper and Traugott 1993: 7) There are two important correlates to this grammaticalization cline. Firstly, the process is inherently unidirectional, with grammaticalized elements remaining grammatical or being lost completely, as opposed to reverting to lexical items. A reversal in the grammaticalization process would run counter not only to principles of conformity, but also to principles of clarity, since grammaticalized elements are generally less explicit. Haspelmath (1999) argues that lexical items are “freely manipulable” while functional items are “processed automatically and unconsciously” (Haspelmath 1999: 1059).33
33
Secondly, the cline represents
While this reversal of the process is indeed impossible, linguists have attempted to justify the existence of degrammaticalization citing examples such as ifs and buts or isms. In my view, however, the term lexicalization would be more accurate in these cases.
131
gradual movement along a lexical-functional continuum, which cannot easily be broken down into discrete categories.
Haspelmath explains that “linguistic
expressions cannot always be categorized clearly into categories such a N, P, V, Aux, C, D, and that instead what we have are continua (N ↔ P, V ↔ Aux, etc.)” and argues that “grammaticalization changes can be described without invoking the concept of abrupt categorical reanalysis” (Haspelmath 1999: 1045). This is particularly important in this study, because not all changes in the development of conjunction systems are marked by clear shifts from one word class to another. There are fundamentally two ways to approach a framework of grammaticalization, one involving a characterization of grammaticalized elements, and another involving a description of the processes themselves. The first approach is best found in the work of C. Lehmann (1995), who provides six parameters34 by which one may assess the degree to which an element is grammaticalized. The parameters are based on basic notions of weight, cohesion and variability, as shown in the table below:
34
These parameters have been called “diagnostics” by Fischer and Rosenbach (2000: 23). However, this term implies a certain testability which the grammaticalization framework has not yet attained.
132
Table 15: C. Lehmann’s (1995) Grammaticalization Criteria PARADIGMATIC SYNTAGMATIC WEIGHT
integrity
structural scope
COHESION
paradigmaticity
bondedness
VARIABILITY
paradigmatic variability
syntagmatic variability (C. Lehmann 1995: 123)
This framework includes three paradigmatic parameters, which attempt to characterize the degree of grammaticalization of a specific linguistic element: 1) Integrity:
the overall “weight” or size of the linguistic element
2) Paradigmaticity:
the extent to which the linguistic element is integrated into the paradigm to which it belongs
3) Paradigmatic variability: the degree to which it and other elements are interchangeable There are also three syntagmatic parameters which, unlike paradigmatic parameters, offer a description of linguistic elements not in isolation but in relation to other elements: 1) Structural scope:
the size of the construction in which the element functions
2) Bondedness:
the cohesion (i.e. degree of attachment or dependence) between an element and other elements in the syntagm
3) Syntagmatic Variability:
the level of rigidity or flexibility of the element’s position within the construction (C. Lehmann 1995: 123-160) 133
In my view, the three paradigmatic parameters proposed by C. Lehmann lack explicitness (e.g. integrity refers to both semantic and phonological properties). More importantly, the syntagmatic parameters appear not to be indicators of grammaticalization, but rather criteria of synthesis, which I view as an entirely independent process.35 Therefore, I will adopt the second—and more widespread—approach, which is found in the earlier work of Heine, Claudi and Hünnemeyer (1991) and Hopper and Traugott (1993), among others. They provide descriptions that are general enough to be applicable to an extremely wide range of diachronic developments in various language families, yet are specific enough to address multiple
independent
changes
(e.g.
semantic,
morpho-syntactic
and
phonological). In the following section of this chapter I will present my analysis of
Latin
subordinating
elements
in
relation
to
seven
well-known
grammaticalization processes described by Heine, Claudi and Hünnemeyer (1991) and Hopper and Traugott (1993):
35
1) Generalization:
the gradual increase in the semantic range of meanings or appropriate contexts of a linguistic unit
2) Decategorialization:
the loss of morphological and syntactic properties (often seen as a shift from major to minor categories)
For this reason, I will address the role of synthesis in the following chapter.
134
3) Specialization:
the reduction in the number of candidates with overlapping (competing) functions
4) Phonetic erosion:
the reduction in phonological weight, including the possible loss of morpheme boundaries and even loss of the entire element36
5) Layering:
the existence of one function with two forms (one older and one newer)
6) Divergence:
the existence of one form with two functions (one older and one newer)
7) Renewal:
the elaboration of the generalized element for the purpose of ensuring expressivity
4.4. Analysis of Latin and Romance Subordinating Elements In spite of the fact that the Latin conjunction system was relatively opaque (as illustrated in section 4.2 above), it is evident that several conjunctions developed from various case forms of the relative pronoun (< PIE *kw-). These closed-class elements have undergone a very slow process of grammaticalization which continues even in modern Romance languages. Rather than looking exclusively at the Modern Spanish and Modern French subordinator que in isolation, I will analyze several related forms within the grammaticalization framework. In my view, this wider approach is more appropriate, if not necessary, since grammaticalization typically involves competition between rival forms. 36
Consider, for example, highly grammaticalized elements such as case marking and verbal inflection, which can be completely lost.
135
Therefore, I will present an analysis of cum (< quom), quando, quoniam, quia, quod and the newer Romance form, que, in relation to the following processes: generalization, decategorialization, specialization, phonetic erosion, layering, divergence and renewal.
4.4.1. Generalization The generalization process is often the focal point of grammaticalization literature. Its prominence is likely due to two facts: (1) the phenomenon is welldocumented and has been discussed—albeit with different terminology—for centuries, and (2) the phenomenon is a fundamental component of the grammaticalization process, and in many cases, appears to serve as the impetus for further developments.
Generalization, in short, refers to the semantic
component of diachronic developments whereby linguistic units lose specificity and acquire a broader range of meanings and contextual applicability.37 For example, the auxiliary verb of the periphrastic future (i.e. Sp. ir Fr. aller ‘go’) and the French negative particle (i.e. Fr. pas) were at first only compatible with motion verbs, but later on spread to other contexts. These semantic changes are often viewed exclusively in terms of attrition or loss, and as a result, terms such as desemanticization and bleaching are quite 37
Generalization is often broken down into two sub-categories, one semantic and one syntactic (cf. Hopper and Traugott 1993: 96-103). In this analysis, however, I have chosen to make a clear distinction between semantic processes (i.e. generalization) and morpho-syntactic processes (i.e. decategorialization).
136
common in the literature. This focus is deeply rooted in scholarly tradition, since even in the late 19th century, the terms verblassen ‘fade’ and verbleichen ‘bleach’ were used by Gabelentz to refer to this process (cf. Hopper and Traugott 1993: 20). Nevertheless, one must keep in mind that although grammaticalized forms may lose a certain amount of specificity, they also experience substantial functional gains. There are two primary factors involved in the generalization process: frequency and reinterpretation. Meillet had already acknowledged the role of frequency when he stated: “A chaque fois qu’un élément linguistique est employé, sa valeur expressive diminue et la répétition en devient plus aisée . . . C’est l’effet ordinaire
de
l’habitude”
(Meillet
1958
[1912]:135).
However,
while
generalization may indeed be facilitated by frequency, it is ultimately motivated by metaphor or context-induced reinterpretation, which refers to instances where “some nonfocal sense is highlighted in a specific context and develops into a new focal sense” (Heine, Claudi and Hünnemeyer 1991: 101). In a similar way, a linguistic unit originally used to express concept A may come to express concept B, due to some underlying semantic similarity between A and B. In the sections that follow, I will provide data for each of the six Latin and Romance subordinating conjunctions in order to illustrate this generalization process.
137
4.4.1.1. Cum An archaic accusative form of the relative pronoun, cum (< quom) was used in Latin as a temporal conjunction. This usage can be seen even in very early Latin texts, such as the Senatus Consultum de Bacchanalibus: (4-4)
Lt.
dum
ne
minus
senatoribus
C
provided that
NEG
less than
senators-ABL
one hundred
adesent,
quom
ea
be present-3PL-IMPERF-SUBJ
when
this-NOM
res
consoleretur
matter-NOM
discuss-3SG-PASS-IMPERF-SUBJ
‘provided that not less than one hundred senators were present when the matter was discussed’ (S.C. Bacch. 6) As an extension of this usage, quom came to express cause, a development which is not surprising given the similarity between the two semantic notions.38 In some instances, this causal notion was subtle and, perhaps, unintended. Consider the following example from Caesar in which the precise value of cum is difficult to ascertain: (4-5)
Lt.
his
cum
sua
sponte
them-DAT
when/since
their-ABL
will-ABL
persuadere
non
possent,
persuade-PRES-INF
NEG
be able to-3PL-IMPERF-SUBJ
legatos
ad
Dumnorigem
Aeduum
deputies-ACC
to
Dumnorix-ACC
Aeduan-ACC
38
For instance, the English temporal conjunction, since, also developed into a causal conjunction due to an overlap in meaning.
138
mittunt send-3PL-PRES-IND
‘when/since by their own influence they could not persuade them, they sent ambassadors to Dumnorix, the Aeduan’ (Caes., D.B.G. 1.9) In other cases, the causal meaning of cum is quite explicit, with little or no trace of its original temporal meaning, as shown in the example below: (4-6)
Lt.
. . . perfacile
esse,
cum
virtute
very easy-ACC
be-PRES-INF
since
valor-ABL
omnibus
praestarent,
totius
all-DAT
surpass-3PL-IMPERF-SUBJ
all-GEN
Galliae
imperio
potiri
Gaul-GEN
power-ABL
seize-PRES-INF
‘. . . it was very easy, since they surpassed all the Gauls in valor, to gain sovereignty of all of Gaul’ (Caes., D.B.G. 1.2) What we have, then, is a widening of scope, since one form has a fairly broad range of applicability. Furthermore, with regard to the common shift from temporal to causal conjunction, Mellet states: “rien d’étonnant par conséquent à ce que des conjunctions temporelles se soient également chargées d’un sens causal” (Mellet 1995: 212). In fact, the motivation for this shift is so strong that it is not an isolated phenomenon, but rather occurs in several similar conjunctions, as will be illustrated in the following sections.
139
4.4.1.2. Quando The conjunction quando also derives from an accusative form of the relative pronoun, but with the addition of the Proto-Indo-European demonstrative stem *do (cf. Watkins 2000: 14). Not surprisingly, we see a similar trajectory for this conjunction, which is commonly used in both temporal and causal clauses, as shown below: (4-7)
Lt.
consuetudo
est,
ut
custom-NOM
be-3SG-PRES-IND
that
fiat
hic
oratio
ab
be done-3SG-PRES-SUBJ
this-NOM
oration-NOM
by
his,
qui
veniunt,
these-ABL
REL-PRON-NOM-PL
come-3PL-PRES-IND
quando
de
eo
loco
primitus
when
from
this-ABL
place-ABL
at first
videtur
mons
Dei
see-3SG-PRES-IND-PASS
mountain-NOM
God-GEN
‘the custom is that prayer should be made by those who arrive here, when from this place the mount of God is first seen’ (Pereg. 1.2) (4-8)
Lt.
quando
non
licet,
since
NEG
be allowed-3SG-PRES-IND
perdam
operam
lose-1SG-FUT-IND
work-ACC
‘since it’s not allowed, I will lose work’ (Pl., Cist. 532 in Bennett 1910: 132)
140
If we compare the two examples above, the polysemy of quando is apparent. In example (4-7), the conjunction is used to express time. However, in example (4-8), while a temporal reading is possible, it is overshadowed by causal value.
4.4.1.3. Quoniam Quoniam is yet another example of a conjunction based on the relative pronoun. In this case, the form in question can be traced to an accusative relative pronoun plus the adverb iam ‘now’ (cf. Meillet and Vendryes 1953: 669). Since one of the components of the conjunction is a temporal adverb, it is not surprising that the conjunction functions in temporal clauses, as shown below: (4-9)
Lt.
quoniam
nuntiatum
now that
announce-PERF-PART-NOM
est . . .
festinamus
be-3SG-PRES-IND
hurry-1PL-PRES-IND
‘when it was announced . . . we hurried’ (Pl., Stich. 676-677 in Bennett 1910: 93) Again, we are able to see a gradual shift taking place and, as a result, there are two possible interpretations for the following examples: (4-10) Lt.
quoniam
sentio
errare,
now that/since
feel-1SG-PRES-IND
err-PRES-INF
coepi
adsentari
begin-1SG-PERF-IND
agree-PRES-INF
141
‘now that/since I feel I am mistaken, I have begun to agree’ (Pl., Men. 481 in Bennett 1910: 92) (4-11) Lt.
. . . fieri be done-PRES-INF
licere,
quoniam
be allowed-PRES-INF,
now that/since
propius
accessisset,
seque
nearer
approach-3SG-PLUPERF-SUBJ
himself—and
id
sine
periculo
facere
it-ACC
without
danger-ABL
do-PRES-INF
posse
existimare
be able-PRES-INF
think-PRES-INF
‘ . . . it could be done, now that/since he (Caesar) had come nearer, and he thought he might be able to do it without danger’ (Caes., D.B.G. 1.42) Finally, in the following example from the Peregrinatio Egeriae, we can see a purely causal conjunction where a temporal reading is difficult, if not impossible: (4-12) Lt.
et
quoniam
episcopus
illius
civitatis
and
since
bishop-NOM
that-GEN
city-GEN
valde
instructus
est
de
greatly
instructed-NOM
be-3SG-PRES-IND
of
scripturis,
requisivi . . .
scriptures-ABL
ask-1SG-PERF-IND
‘and since the bishop of that city is very learned in the Scriptures, I asked . . .’ (Pereg. 20.9)
142
This generalization process eventually extended even further and quoniam came to be used as a complementizer primarily devoid of its own semantic content: (4-13) Lt.
qui
credit
quoniam
who-NOM
believe-3SG-PRES-IND
that
Iesus
est
filius
Dei
Jesus-NOM
be-3SG-PRES-IND
son-NOM
God-GEN
‘he who believes that Jesus is the son of God’ (Vulg. Joan. 1.5,5, in Meillet and Vendryes 1953: 662) Via the above examples, we have seen the gradual spread of semantic content and loss of specificity of the subordinating conjunction, quoniam.
4.4.1.4. Quia The conjunction quia is a reflex of the plural neuter relative pronoun.39 However, in this case it appears that the conjunction began not as a temporal conjunction but as a causal conjunction, as shown below: (4-14) Lt.
pinus
eo,
quia
semen
pine-NOM
for that reason
since
seed- ACC
viride
et
maturum
habet . . .
green-ACC
and
mature-ACC
have-3SG-PRES-IND
item
quidvis
anni
matura
likewise
whatever
season-GEN
ripe-NOM
39
This is the etymology provided by Meillet and Vendryes (1953: 499). Lewis and Short (1879), on the other hand, claim that quia stems from a relative pronoun plus the adverb Lt. jam ‘now’. This seems unlikely, however, since the inital function of the conjunction appears to have been causal, not temporal.
143
est
et
tempestiva
be-3SG-PRES-IND
and
seasonable-NOM
‘the pine, since it has both green and ripe seed…is also ripe and ready at any season’ (Cato, R.R. 17.1) Temporal examples of quia are much less frequent and appear to have developed as an extension of its causal usage, perhaps due to analogy with similar conjunctions. According to Herman, the majority of examples of temporal quia appear in the Collectio Auellana, which is from the 6th century A.D. (Herman 1963: 55). Below is an example of quia with temporal value from this late Latin text: (4-15) Lt.
et
quia
and
when
pervenerunt arrive-3PL-PERF-IND
in
civitatem,
in
city-ACC
nuntiata
est
Dorotheo . . .
announce-PERF-PART
be-3SG-PRES-IND
Dorothy-DAT
‘and when they arrived in the city, it was announced to Dorothy . . .’ (Avell. 225.5 in Herman 1963: 55) Not unlike quoniam, quia also lost enough specificity to be able to serve as a general linking device with no semantic content of its own: (4-16) Lt.
solum
scio,
quia
postmodum
only
know-1SG-PRES-IND
that
afterwards
puer
Abrahae,
ut
peteret
boy-NOM
Abraham-GEN
in order to
seek-3SG-IMP-SUBJ
144
Rebeccam . . .
in
Charra
venerit
Rebecca-ACC
in
Charrae
come-3SG-PERF-SUBJ
‘I know only that afterwards Abraham's servant came to Charrae to seek Rebecca’
(Pereg. 20.9) When discussing this particular development from a specific conjunction to a “bleached” complementizer, Herman draws attention to “le flottement psychologique entre subordonnée complément d’objet et subordonnée causale” (Herman 1963: 44), and Moreno Sánchez refers to “la estrecha relación que existe entre la oración completiva y la causal” (Moreno Sánchez 1995: 76). Similar to the transition from temporal conjunction to causal conjunction, it does not require a great semantic leap to shift from causal conjunction to complementizer.
4.4.1.5. Quod It is fair to say that quod—the neuter nominative singular form of the Latin relative pronoun—played an extremely important role in Latin subordination. I will therefore trace the entire trajectory of its usage beginning with its original role as relative pronoun, which is illustrated below: (4-17) Lt.
hoc
est
praedium
this-NOM
be-3SG-PRES-IND
farm-NOM-N
quod
ubi vis
REL-PRON-NOM-N-SG
anywhere you like
expedit
facere
be profitable-3SG-PRES-IND
make-PRES-INF
145
‘this is the sort of farm which is profitable to make everywhere’ (Cato, R.R. 9) It is, in fact, possible to find instances where quod appears to have an additional overlay of causal meaning while still functioning as a relative pronoun. Examples such as the following serve to shed light on the way in which such a transfer of meaning could occur: (4-18) Lt.
occupandum
Vesontionem,
occupy-GRDV-ACC-M
Vesontio-ACC-M
quod
est
REL-PRON-NOM-N-SG/because
be-3SG-PRES-IND
oppidum
maximum
Sequanorum
town-NOM-N
greatest-NOM
Sequani-GEN-PL
‘to seize Vesontio, which/because (it) is the largest town of the Sequani’ (Caes., D.B.G. 1.38) In the above example from Caesar, it is not difficult to acknowledge a possible causal reading for quod. In the following example, however, we can see that the causal value takes over completely.
Additionally, quod can no longer be
interpreted as a relative pronoun, due to lack of agreement between neuter quod and masculine amicus: (4-19) Lt.
Helvetiis
erat
amicus,
Helvetii-DAT
be-3SG-IMPERF-IND
friendly-NOM-M
quod
ex
ea
civitate
Orgetorigis
because
from
that-ABL
state-ABL
Ortegorix-GEN
146
filiam
in
matrimonium
daughter-ACC
into
marriage-ACC
duxerat lead-3SG-PLUPERF-IND
friendly to the Helvetii because from that state he had taken Orgetorix’s daughter in marriage’
‘he was
(Caes., D.B.G. 1.9) Parallel to the development of quoniam and quia, we also find examples of quod shifting from a causal conjunction to a more generalized conjunction with multifaceted meaning.
Consider the example below, in which several readings
(i.e. temporal, causal and semantically “empty”) are possible: (4-20) Lt.
haec
est
autem
vallis
this-NOM
be-3SG-PRES-IND
moreover
valley-NOM
ingens
et
planissima,
in
qua
great-NOM
and
flat-NOM
in
REL-PRON-ABL
filii
Israhel
commorati sunt
children-NOM
Israel-GEN
remain-3PL-PERF-IND
his
diebus,
quod
sanctus
these-ABL
days-ABL
when/because/that
holy-NOM
Moyses
ascendit
in
montem
Moses-NOM
ascend-3SG-PERF-IND
into
mountain-ACC
Domini Lord-GEN
‘now this is the great and flat valley wherein the children of Israel waited during those days, when/because/that holy Moses went up into the mount of the Lord’ (Pereg. 2.2)
147
This example from the Peregrinatio Egeriae shows the extent to which quod has generalized. The action in the subordinate clause could clearly be seen as the justification for the action in the main clause. It could also represent the time at which the action in the main clause occurred. Furthermore, quod in this instance is sufficiently vague to simply serve as a linking element with no additional overlay of meaning. There are clear instances, however, where quod has made a full transition to a semantically “bleached” complementizer as shown below: (4-21) Lt.
Pharao,
quando
vidit,
quod
Pharaoh
when
see-3SG-PERF-IND
that
filii
Israhel
dimiserant
children-NOM
Israel-GEN
abandon-3PL-PLUPERF-IND
eum, . . .
isset
cum
omni
him-ACC
go-3SG-PLUPERF-SUBJ
with
all-ABL
exercitu
suo
intra
Ramesse
army-ABL
his-ABL
inside
Rameses-ACC
‘when Pharaoh saw that the children of Israel had escaped him, . . . he went with all his army into Rameses’ (Pereg. 8.5) In this example, quod simply links the two clauses and no longer contains any temporal or causal value. What we have here is the end of the generalization process, which first imbued quod with additional values (i.e. temporal / causal), ultimately making it non-specific enough to serve as a general linking device.
148
4.4.1.6. Que The Romance form, que, is both the newest and the most difficult to trace. Although it has often been cited as a reflex of quod (cf. Bichakjian 1982: 17), this is a phonetic impossibility. Most scholars agree that it is a combination of several relative/interrogative elements—namely, quae, quem, and quid—which merged early on in the grammaticalization process.40 This hybrid, uninflected form was pronounced /ke/, although orthographic variations are found in earlier stages of the Romance languages. Initially, these forms (before merging), functioned as relative pronouns, as shown below in an example from the Satyricon: (4-22) Lt.
alter
numerabat
pilas,
other-NOM
count-3SG-IMPERF-IND
balls-ACC-F-PL
non
quidem
eas
quae
NEG
indeed
those-ACC
REL-PRON-NOM-F-PL
inter
manus . . .
vibrabant
between
hands-ACC
move rapidy-3PL-IMPERF-IND
‘the other counted the balls, but not those which were caught’ (Petr., Sat. 27) Early on in the history of the daughter Romance languages, many examples of this form can be found in an ever-increasing variety of contexts. The extremely
40
For example, Herman (1963) attributes que to a merger of quae and quem, while Togeby (1980) cites quem and quid and Schlieben-Lange (1992) cites quia and quem.
149
rapid generalization of this form can be seen in the following data from Old Spanish and Old French: (4-23) OSp. sabam
quantos
esta
carta
how many
this
letter
know-3PL-PRES-SUBJ
uirem
&
oyrem.
Que
see-3PL-FUT-SUBJ
and
hear-3PL-FUT-SUBJ
that
you
don
Arias
pela
gracia
de
I-NOM
don
Arias
by the
grace
of
dios . . . Damos
uos
um
poulo
God
you-DAT
a
piece of land
give-1PL-PRES-IND
‘whoever sees and hears this letter should know that I, don Arias, by the grace of God . . . give you a piece of land’ (13th century text in Gifford and Hodcroft 1966: 96) (4-24) OSp. al at—the
tiempo
que
los
godos
time
that/when
the
Goths
passoron
mar . . .
cross-3PL-PRET-IND
sea
‘at the time that/when the Goths crossed the sea . . .’ (13th century text in Gifford and Hodcroft 1966: 181) (4-25) OSp. mando order-1SG-PRES-IND
que
si
algun
uaron
o
that
if
some
man
or
mugier
christiano
uendiere
e
woman
Christian
sell-3SG-FUT-SUBJ
and
prouadol
fuere
sea
prove-PP
be-3SG-FUT-SUBJ
be-3SG-PRES-SUBJ
quemado burn-PP
150
‘I order that if any Christian man or woman should sell (it) and it is proven, he/she is to be burned’ (14th century text in Gifford and Hodcroft 1966: 192) (4-26) OFr.
n’i ad
icel
ne
demeint
NEG—there is
DEM-PRON
NEG
show-3PL-PRES-IND
irance anger-OBL
que that/because
il
ne
they-NOM
NEG
sunt
a
Rollant
le
be-3PL-PRES-IND
at
Rolland-OBL
the-OBL
cataigne captain-OBL
‘there isn’t anyone that does not display anger that/because he is not with Roland the captain’ (Rol. 1845-1846) (4-27) OFr.
melz
me
venist,
amis,
better
me-DAT
come-3SG-IMPERF-SUBJ
friend-NOM
que
morte
fusse
that
dead-NOM
be-1SG-IMPERF-SUBJ
‘it would be better for me, my beloved, that/if I were dead” (Alex. 485) These examples effectively illustrate the generalized nature of the conjunction que, which can serve as a complementizer, but can also express causal, temporal and conditional values.
What is important here is the generalization of the
conjunction que and its compatibility with many different clause types.
151
This section has shown the way in which certain Latin / Romance conjunctions (in specific, those derived from relative pronouns) have undergone a gradual process of generalization over a long period of time. The result of this process is twofold: (1) each conjunction is able to express a fairly wide range of meanings and, subsequently, (2) there is a substantial amount of overlap between competing forms. Before concluding this section, I would like to address an important characteristic of this process in relation to mood. There appears to be a distinct pattern with regard to generalization, in terms of mood compatibility. Of the five Latin forms described in this section, three of them (i.e. quoniam, quia and quod) continued their generalization from causal clauses to the point where they were used with no semantic value of their own (i.e. in 4-13, 4-16 and 4-21). These conjunctions, in particular, were able to combine freely in causal clauses with both the indicative and the subjunctive.
Cum, on the other hand, was only
compatible with the subjunctive in causal clauses (cf. Leiper 1913: 16; Gildersleeve and Lodge 1997: 338-339, 370-371), while quando was only compatible with the indicative.
It is quite likely, therefore, that this wider
applicability favoured the further generalization of the conjunctions quoniam, quia and quod.
152
4.4.2. Decategorialization While generalization refers to the semantic component of the grammaticalization process, decategorialization refers to the morpho-syntactic component. Generally speaking, decategorialization is the gradual reduction in morpho-syntactic properties. The target form will function differently and will typically have less morphological weight than its source.
For example, the development of the
ablative case of the Latin noun mens ‘mind’, into the adverbial suffix (Sp. –mente and Fr. -ment) was accompanied by a loss of certain nominal properties, namely, the ability to be inflected for number and be accompanied by a determiner. It is important to point out, however, that this shift should be seen as a change in category rather than a loss. Hopper and Traugott (1993) insist: “In ascribing ‘decategorialization’ to a form we are not tracing the decay or deterioration of that form, but its functional shift from one kind of role to another in the organization of discourse” (Hopper and Traugott 1993: 104). The essence of decategorialization, then, is the movement of a form along a grammaticalization cline (e.g. verb > auxiliary > clitic > affix). In the specific case of Latin relative pronouns, the phenomenon we are examining involves a relatively subtle shift (from minor category to minor category). Nevertheless, features of the decategorialization process can still be successfully identified. Below, I propose a unified grammaticalization cline for the six closed-class elements discussed in this chapter: 153
Table 16: Grammaticalization Cline for Subordinating Elements A B C relative > temporal < > causal > complementizer pronoun conjunction conjunction ACC form
>
quom
< >
quom
ACC form + DEM stem
>
quando
< >
quando
ACC form + jam ‘now’
>
quoniam
< >
quoniam
>
quoniam
ACC N PL form
>
quia
< >
quia
>
quia
< >
quod
>
quod
>
que
ACC N SG form hybrid form
There are two general characteristics of this cline that need to be addressed. Firstly, the overlap displayed in the three main categories (A-C) is to be expected, due—at least in part—to the coexistence of older and newer forms. Secondly, the fact that not all forms continue along the cline to stage C is also quite normal. Hopper and Traugott explain that since “the particular course of events in any cline that is presented is not predetermined, once the ‘slippery slope’ is embarked upon, continued grammaticalization is not inevitable, but may be suspended indefinitely at any point” (Hopper and Traugott 1993:106). With respect to the specific linguistic forms in this analysis, the categories I am proposing as part of the cline can be distinguished in syntactic terms.
154
Regarding the integration of subordinate clauses, C. Lehmann (1989) discusses the notion of hierarchical downgrading, with parataxis at one end of the spectrum, and embedding at the other: independent - adjoined - correlative - medial41 - participial - governed clause clause diptych clause clause clause (C. Lehmann 1989: 157) Although the focus here is on words and not clauses, we can witness the decategorialization of these subordinating elements via the types of clauses they introduce. Consider the following three examples: (4-28) Lt.
ei
gratias
egit,
him-DAT
thanks-ACC
act-3SG-PERF-IND
quod
de
se
optimum
because
of
REFL-PRON-ABL
excellent-ACC
iudicium
fecisset
opinion- ACC
make-3SG-PLUPERF-SUBJ
‘(the legion) gave thanks to him, because he had expressed a very high opinion of them’ (Caes., D.B.G. 1.41) (4-29) Lt.
quod
antea
de
colloquio
what
before
about
conference-ABL
postulasset,
id
per
request-3SG-PLUPERF-SUBJ
DEM-PRON-ACC
through
se
fieri
licere
REFL-PRON-ACC
be done-PRES-INF
be allowed-PRES-INF
41
Medial clauses—which appear specifically in systems of clause chaining—are not typically found in European languages (Matthews 1997: 55, 221).
155
‘what he had requested before regarding a conference, (it) might be allowed to be done as far as he was concerned’ (Caes., D.B.G. 1.42) (4-30) Lt.
tamen
verum
quod
frunitus est,
nevertheless
true
that
enjoy-3SG-PERF-IND
quam diu
vixit
as long as
live-3SG-PERF-IND
‘neverthless (it’s) true that he enjoyed himself as long as he lived’ (Petr., Sat. 43) In the first example (4-28), quod is introducing an adjoined clause—following C. Lehmann’s (1989) criteria—since it is not integrated into the main clause. In (4-29), however, quod is functioning within a correlative diptych, in which the clauses are clearly interdependent. Finally, in example (4-30), quod introduces a complement clause which is fully embedded. In fact, Hopper and Traugott make a clear distinction between hypotaxis, which includes adverbial clauses, and subordination, which includes complement clauses (Hopper and Traugott 1993: 170). Similarly, C. Lehmann distinguishes between syndesis and asyndesis,42 with universal subordinators closer to asyndesis than specific conjunctions (C. Lehmann 1989: 174). In this way, we are able to use syntactic criteria to
42
The terms syndesis and asyndesis (as opposed to hypotaxis and parataxis) relate to the explicitness of the linking device.
156
establish that the form quod did, in fact, undergo a process of decategorialization and not merely a semantic process of generalization.43 The decategorialization of relative pronouns can also be established on morphological grounds, since they have undergone a reduction in (or even total loss of) nominal inflection.
Consider the following Latin relative pronoun
paradigm: Table 17: Latin Relative Pronouns SINGULAR
PLURAL
M
F
N
NOM
qui
quae
quod
GEN
cuius
cuius
cuius
DAT
cui
cui
cui
ACC
quem
quam
quod
ABL
quo
qua
quo
NOM
qui
quae
quae
GEN
quorum
quarum
quorum
DAT
quibus
quibus
quibus
ACC
quos
quas
quae
ABL
quibus
quibus
quibus
In spite of a moderate amount of syncretism, grammatical categories of gender, number and case are expressed via fifteen distinct forms. Reflexes of these forms exist in Romance languages both as relative pronouns and as conjunctions and
43
Although I have chosen to illustrate the hierarchical downgrading of quod, the same development can be traced for other forms as well (e.g. quia and que).
157
complementizers. Not surprisingly, both the rate and extent of decategorialization is quite different for these two different functions. The following examples from Old French and Old Spanish show that, used as a relative pronoun, the form was still inflected for case, as shown below: (4-31) OFr.
Maximiien,
chi
rex
Maximian-NOM
REL-PRON-NOM
king-NOM
eret
a
cels
dis
be-3SG-IMPERF-IND
at
those-OBL
days-OBL
‘Maximian, who was king in those days’ (Eul. 11-12) (4-32) OFr.
d’icest
saint
cors,
que
of this-OBL
holy-OBL
body-OBL
REL-PRON-ACC
avum have-1PL-PRES-IND
am in
bailide possession-OBL
‘by this holy body which we have in our possession’ (Alex. 534) (4-33) OSp. Cosdroe, Cosdroe,
qui
era
REL-PRON-NOM
be-3SG-IMPERF-IND
rei
de
Persia
king
of
Persia
‘Cosdroe, who was king of Persia’ (13th century text in Gifford and Hodcroft 1966: 181) (4-34) OSp. mia my
aldea
ke
dizen
village
REL-PRON-ACC
say-3PL-PRES-IND
158
Uilla
Algariua
Villa
Algariua
‘my village, which they call Villa Algariua’ (12th century text in Gifford and Hodcroft 1966: 30) As a conjunction or complementizer, however, the form was completely uninflected (consistently /ke/ but with orthographic variants such as que, che, qued and ke), as illustrated in the following examples: (4-35) OFr.
s’il
veit
que
jo
if—he-NOM
see-3SG-PRES-IND
that
I-NOM
lui
serve
him-DAT
serve-1SG-PRES-IND
‘if he sees that I serve him’ (Alex. 495) (4-36) OSp. sabida know-PP
cosa
es
que
Iague
thing
be-3SG-PRES-IND
that
Iague
de
Sarraton
gano
el
moleo
of
Sarraton
win-3SG-PRET-IND
the
mill
‘it
is known that Iague of Sarraton won the mill’ (12th century text in Gifford and Hodcroft 1966: 31)
Here we see that the decategorialization process has extended even farther. Unlike the relative pronoun, which still has ties to a noun—and therefore, is somewhat resistant to decategorialization—the complementizer is no longer a nominal element and thus retains none of its nominal properties. In conclusion, although both the source and target in this instance belong to minor categories, the 159
criteria of decategorialization can still serve as a strong indicator of the grammaticalization process. Before I conclude this section, I would like to address one important issue regarding the decategorialization of the six forms.
When examining the
grammaticalization cline for these forms (see table on page 154), a major difference becomes apparent. Que, unlike the other forms, reaches its target with incredible speed. I believe this contrast can be adequately motivated within the grammaticalization framework. The older forms grammaticalized very slowly, since the target grammatical function (i.e. complementizer) did not previously exist. However, once finite complement clauses were a productive part of Latin syntax, que was able to fill this existing function almost immediately. This is extremely significant because it highlights the fact that there is, in effect, more than one type of grammaticalization. There is a type where new forms come to perform new functions and one where new forms come to perform old functions. Consider the following table adapted from Ramat (1998: 108): Table 18: Types of Grammaticalization function
form
type 1
new
new
type 2
old
new
In other words, although quod and que both underwent decategorialization and have similar sources and targets (i.e. relative pronouns and complementizers), there is a subtle, yet important, difference. 160
The development of quod is an
example of the first type of grammaticalization, since it was involved in the creation of a new category (i.e. complementizer), while the development of que is an example of a new form with an old (i.e. already existing) function, and thus, was able to proceed much more rapidly.
4.4.3. Specialization The consequence of the gradual semantic and morpho-syntactic processes described above is the expansion of a grammatical paradigm. As a result of the generalization of meaning, several distinct forms come to serve the same function(s). Specialization can therefore be seen as a means by which speakers resolve ambiguities and/or reduce the variety of somewhat cumbersome paradigms. In other words, this is a process involving the “‘thinning out’ of the field of candidates” (Hopper and Traugott 1993: 116). The specialization process is often cited with regard to negation in French. After a period of generalization and decategorialization, multiple nouns had come to serve as negative particles: pas ‘step, pace’, point ‘dot, point’, mie ‘crumb’, gote ‘drop’, amende ‘almond’, areste ‘fish-bone’, beloce ‘sloe’ and eschalope ‘pea-pod’ (Hopper and Traugott 1993: 115). Consequently, the inventory was greatly reduced, with pas surfacing as the clear “winner.” This reductive phenomenon also occurred with the Latin conjunction system. The six candidates examined in this chapter are presented in summary 161
form in the following table, which provides the three main functions along with the inventory of forms (numbered 1 through 6) used to express them: Table 19: Candidates for Specialization among Subordinating Elements temporal conjunction causal conjunction complementizer ‘when/now that’ ‘because/since’ ‘that’ 1. cum 2. quando 3. quoniam 4. quia
1. cum 2. quando 3. quoniam 4. quia 5. quod
3. quoniam 4. quia 5. quod 6. que
In general terms, the specialization process affected the entire Latin conjunction system.
Of vital importance to this discussion, however, is the development of
the complementizers in particular, which reveals something quite significant regarding the process of specialization. It is an extremely fluid process which, I would argue, is cyclical in nature. A group of candidates may be pared down at one stage, yet new candidates may enter the paradigm, triggering a second round of specialization. This, in my view, is precisely what occurred with Latin / Romance complementizers. Below, I will briefly discuss these two separate, but related, specialization processes.
162
4.4.3.1. Latin Specialization: Quod Speakers of Latin had at their disposal three forms which could serve as complementizers: quod, quia and quoniam. Of these three candidates, quoniam was the least common, possibly due to its bi-syllabic form, including a CVC syllable structure, as well as its specificity (i.e. it was used primarily to introduce causal clauses which presented information already known by the listener [cf. Herman 1963: 44]). Quia and quod, on the other hand, were both strong candidates, yet quod had several advantages. Firstly, the form was extremely pervasive and, due to its multiple functions, the overall frequency of the form quod (including all of its uses) well surpassed quia (and quoniam). In Book I of De Bello Gallico, for example, I have counted one-hundred sixteen examples of quod, thirty-one examples of quia and four examples of quoniam.44 Secondly, it was the first form to function as a complementizer. Although this usage was not widespread until late Latin, it can be seen sporadically in Classical Latin, as illustrated via the following example from Caesar: (4-37) Lt.
testimonium
esse
quod
nisi
proof-NOM
be-PRES-INF
that
unless
rogatus
non
venerit
ask-PERF-PART-NOM
NEG
come-3SG-PERF-SUBJ
et
quod
bellum
non
intulerit,
and
that
war-ACC
NEG
bring on-3SG-PERF-SUBJ
44
Although I counted one-hundred sixteen examples of quod in Book I of De Bello Gallico, only two were used as complementizers. The majority (seventy-one) were causal conjunctions and relative pronouns.
163
sed
defenderit
but
defend-3SG-PERF-SUBJ
‘the evidence was that he did not come without an invitation and that he had not waged war, but prevented it’ (Caes., D.B.G. 1.44) Furthermore, the form quod appears to have been the unmarked form, since quod was used in virtually all contexts, while quia was primarily used in causal clauses. These factors may have ultimately favored quod during the specialization process. According to Herman (1963), quod was the complementizer of choice in Merovingian Gaul, as illustrated via the following numbers from four 6th-8th century texts: Table 20: Specialization of Quod in Merovingian Gaul TEXT quod
quia
Formulae Andecavenses
7
2
Cartae Senonicae
11
1
Lex Ribvaria
26
0
Liber Historiae Francorum
15
1
(Herman 1963: 41) These data reflect the specialization process whereby one form (out of two or more candidates) becomes the most predominant. The emptiest form (in terms of both semantics and phonology) was selected, resulting in a reduction in overlap and a shift toward a more consistent one-to-one relationship between form and function. 164
4.4.3.2. Romance Specialization: Que There was, however, a second specialization process, drawing attention to the way in which several simultaneous grammaticalization processes may overlap. A new indeclinable form (i.e. que) emerged beginning in the 6th century (cf. Herman 1963: 125) and once again, speakers had at their disposal a choice of complementizers (now primarily quod and que). Herman (1963) states: “Il est donc à peu près certain que le relatif indéclinable à valeur générale que (quae, quem, quid, que dans les textes) avait pour concurrent et quasi-synonyme un autre relatif à valeur générale, quod, et cela jusque vers le VIIe siècle au moins” (Herman 1963: 128-129). A period of coexistence between quod and que resulted, ultimately, in a second specialization process out of which que emerged as the new generalized subordinator.45 In Gifford and Hodcroft’s (1966) monograph, Textos Lingüísticos del medioevo español—which includes one hundred twenty-three texts from the beginning of the 9th century through the end of the 15th century—there is evidence of strong specialization, as there are very few instances of quod as a complementizer or generalized subordinator. Below is one of the scarce examples of quod used in this way:
45
The exception to this is found in Southern Italian and Rumanian, where two general subordinators are found. In Rumanian, for example, the reflex of Lt. quod (i.e. că) persists, and is used strictly after declarative and emotive verbs. The reflex of Lt. si, (i.e. să), follows volitive verbs. This difference has been explained in terms of the Greek adstratum (see Togeby 1980: 134).
165
(4-38) OSp. si if
negarent
&
dixerint
deny-3PL-FUT-SUBJ
and
say-3PL-FUT-SUBJ
quod
non
inciderunt
ligna
that
NEG
cut-3PL-PERF-IND
firewood
‘if they deny (it) and say that they didn’t cut firewood’ (11th century text in Gifford and Hodcroft 1996: 25) Que, on the other hand, is extremely frequent as a complementizer and generalized subordinator in Old Spanish texts. Consider the following data: (4-39) OSp. fazemos make-1PL-PRES-IND
saber
a
todos
quantos
know-INF
to
all
how many
esta
present
carta
verran
et
this
present
letter
see-3PL-FUT-IND
and
orran
que
nos,
veyendo
et
hear-3PL-FUT-IND
that
we
see-GER
and
conosciendo
que
muchas
de
vegadas . . .
know-GER
that
many
of
times
‘we inform whomever reads and hears this letter that we, seeing and knowing that many times…’ (13th century text in Gifford and Hodcroft 1966: 143) (4-40) OSp. et and
si
el
filio
de
don
Garcia
if
the
son
of
don
Garcia
muriere
que
non
aia
die-3SG-FUT-SUBJ
that
NEG
have-3SG-PRES-SUBJ
filios,
toda
la
auandicta
heredat
sons
all
the
aforementioned
property
assi quemo
la
deuie
in the same manner as
it-DAT
should-3SG-IMPERF-IND
166
auer
el
filio,
con
quanto
que
have-INF
the
son
with
as much
that
ouiere,
remanezca
have-3SG-FUT-SUBJ
remain-3SG-PRES-SUBJ
a at
la the
orden
de
sancto
Iacobo. . .
et
que
order
of
saint
Iacobo
and
that
partan
todos
los
ganados
share-3PL-PRES-SUBJ
all
the
possessions
‘and if the son of don García should die and has no children, all the aforementioned property that the son would have received, however much there may be, shall remain with the order of Saint Iacobo . . . and they should share all the proceeds’ (13th century text in Gifford and Hodcroft 1966: 32) The examples above illustrate the wide range of functions and high frequency of the specialized form que. We find a similar preference for que in Old French texts, as shown in the representative example below: (4-41) OFr.
tuit
oram
que
por
nos
all-NOM
pray-1PL-IMP
that
for
us-DAT
degnet
preier
qued
deign-3SG-PRES-SUBJ
pray-INF
that
auuisset
de
nos
Christus
have-3SG-IMP-SUBJ
of
us-OBL
Christ-NOM
mercit mercy-OBL
‘let us all pray that she will deign to pray for us that Christ will have mercy on us’ (Eul. 26-27)
167
The reason(s) for which que was selected—thereby ousting quod—cannot be known for certain.
However, definite advantages can be identified.
Most
important is the fact that in relation to quod, que was possibly felt to be the unmarked member of the pair. First of all, although quod had a wide range of semantic values, they were, nevertheless, relatively specific. Que, on the other hand, was a new form and, thus, had no such additional overlay of meaning. Furthermore, in terms of phonology, que was also shorter, and contained a more neutral vowel.46
4.4.4. Phonetic Erosion It is widely accepted that phonological correlates to changes in semantics, morphology and syntax are quite common.
This component of the
grammaticalization process has been recognized by scholars for over a century. In the late nineteenth century, for example, Gabelentz recognized a gradual Abnutzung ‘abrasion’ of phonological substance (cf. Heine, Claudi and Hünnemeyer 1991: 214). Furthermore, Meillet acknowledged the relationship between this change and the reduction in semantic content, arguing that “l’affaiblissement du sens et l’affaiblissement de la forme des mots accessoires vont de pair” (Meillet 1958 [1912]: 139).
46
The vowels in que (Sp. /e/ and Fr. / ə/) are also “hesitation vowels” in these languages.
168
Although the grammaticalized subordinator was monosyllabic from the start, there is evidence of further phonetic reduction. First of all, this hybrid form (< quae, quem and quid) was reduced across Romance languages to a CV structure (> /ke/), although final consonants were occasionally preserved in medieval orthography. In French, however, we find an even greater reductive tendency: loss of autonomy. The subordinator que underwent elision, a linking phenomenon primarily involving highly grammaticalized monosyllabic forms (i.e. je, me, te, le, se, ce, de and ne). Elision of que with the following word occurred quite early, and it is therefore seen often in Old French texts: (4-42) OFr.
qu’elle
Deo
raneiet
that—she-NOM
God-OBL
deny-3SG-PRES-SUBJ
‘that she deny God’ (Eul. 6) (4-43) OFr.
…qu’a
tei
ansemble
that—at
you-DAT
together
n’oüsse conversét NEG—converse-PLUPERF-SUBJ
‘that I hadn’t spoken with you’ (Alex. 489) This reductive trend reflects an important aspect of natural morphology: iconicity. Iconicity refers to the non-arbitrary relationship between form and meaning, whereby longer forms will be used to mark “heavier” content, while
169
shorter forms will be used for emptier, or unmarked, elements.47 Within this context we can understand why the reduction in semantic and morphological content which took place during the grammaticalization of Latin / Romance subordinators would eventually lead to a reduction in phonological content as well. As Haspelmath states, “there is an iconic relationship between form and meaning in grammaticalization” (Haspelmath 1999: 1050).
4.4.5. Layering and Divergence Both layering and divergence are widespread synchronic results of the grammaticalization process. I have included them together in one section because I view them as two sides of the same coin. Both reflect the way in which a language system—at any stage—contains older and newer linguistic features working simultaneously. Layering refers to the existence of one function with two corresponding forms and divergence refers to the existence of one form with two different functions. The grammaticalized form que exhibits both of these tendencies in modern Romance languages. In terms of layering, one function—namely, the complement clause—is expressed via old and new forms, as shown below:
47
For example, with regard to number marking, singular is the unmarked member of the pair. Therefore, we typically see iconic relationships such as that between book and books, where the marked plural form carries additional weight. Less commonly, relationships are non-iconic, as in the word fish (both singular and plural), while a counter-iconic relationship is seen between French SG. [øf] ‘egg’ and PL. [ø] ‘eggs’, where the plural form is lighter than the singular form.
170
(4-44) Sp. (4-45) Sp.
pienso
tener
razón
think-1SG-PRES-IND
have-INF
reason
pienso
que
tengo
razón
think-1SG-PRES-IND
that
have-1SG-PRES-IND
reason
‘I think I’m right’ (4-46) Fr. (4-47) Fr.
je
crois
être
malade
I-NOM
believe-1SG-PRES-IND
be-INF
sick
je
crois
que
je
I-NOM
believe-1SG-PRES-IND
that
I-NOM
suis
malade
be-1SG-PRES-IND
sick
‘I think I’m sick’ Examples (4-44) and (4-46) represent an older form of complementation, which was originally a Latin Accusative-Infinitive construction. The main difference in this case is that in modern Spanish and French, the infinitive lacks an overt subject.
Examples (4-45) and (4-47), however, contain the newer Romance
complementizer: que. It is important to point out that layering does not imply equality among all layers, neither in precise usage nor in frequency. In this case, for instance, the non-finite clause is less common, being subjected to syntactic constraints; namely, that the understood subject of the subordinate clause is typically controlled by the subject of the main clause. Divergence represents the opposite phenomenon, where one form has both an older function and a newer function. In the case of que, it serves its new role
171
as conjunction / complementizer, while still serving its original function as a relative pronoun. Consider the following examples: (4-48) Sp.
espero
que
compres
el
hope-1SG-PRES-IND
that
buy-2SG-PRES-SUBJ
the
libro
que
te
recomendé
book
that
you-DAT
recommend-1SG-PRET-IND
‘I hope that you buy the book that I recommended to you’ (4-49) Fr.
je
pense
qu’il
I-NOM
think-1SG-PRES-IND
that—he-NOM
utilise
l’ordinateur
qu’il
use-3SG-PRES-IND
the—computer
that—he-NOM
a
acheté
en
septembre
have-3SG-PRES-IND
buy-PP
in
September
‘I think that he’s using the computer that he bought in September’ In both Spanish and French, we see the same form, que, used both as a relative pronoun and as a complementizer. In summary, this synchronic state of affairs, in terms of both layering and divergence, can be represented as follows: Table 21: Layering and Divergence of Que FUNCTION LAYERING
DIVERGENCE
complement clause relative pronoun conjunction / complementizer
172
FORM infinitive que + finite verb que
4.4.6. Renewal Renewal is an aspect of language change in which the tug-of-war between simplification and explicitness is most evident.
In short, it is an innovative
process which elaborates or expands upon the grammaticalized (i.e. generalized, decategorialized and specialized) form. Hopper and Traugott (1993) present it succinctly as follows: “If all grammaticalization leads to decategorialization and ultimately to minimal, compacted forms, how is it that language users can ensure that languages continue to serve their purposes of organizing cognition and achieving communication?” (Hopper and Traugott 1993: 121).
The renewal
process provides one answer to this question. For example, Fr. pas was originally used for emphasis within negative constructions, yet lost its emphatic value via grammaticalization, developing into a generalized negative particle. Renewal has since taken place, however, and new emphatic forms have developed (e.g. Fr. pas du tout ‘not at all’). Not only can we see this renewal process in the development of the Romance conjunction system, but it is an exceptionally widespread phenomenon. Consider the following Latin > Romance trajectory: Table 22: Renewal of Romance Generalized Subordinator Latin forms grammaticalized (no longer in use) subordinator quod quoniam quia cum
>
que
173
>
Romance forms (cases of renewal) Sp. porque Fr. parce que
With regard to French, we can see new forms surfacing throughout Old and Middle French periods. The following examples of new conjunctions based on que are from 10th-16th century French texts: (4-50) OFr.
por qe
Deus
cel
edre
li
because
God-NOM
this-OBL
ivy-OBL
him-DAT
donat give-3SG-PRET-IND
‘because God gave him this ivy’ (Ayres-Bennett 1996: 41) (4-51) OFr.
jusques a ce que
vous
eussiez
until
you-NOM-PL
have-2PL-IMPERF-SUBJ
plus
veu
et
apris
more-OBL
see-PP-OBL
and
learn-PP-OBL
‘until
you had seen and learned more’ (Ayres-Bennett 1996: 123)
(4-52) OFr.
à fin qu’il
ne
semble
so that—it-NOM
NEG
seem-3SG-PRES-IND
‘so that it doesn’t seem’ (Ayres-Bennett 1996: 153) In order to illustrate the same process in Spanish, I collected all examples of renewal (i.e. new conjunctions based on que) from the thirty-seven Castilian texts in Gifford and Hodcroft’s (1966) Textos lingüísticos del medioevo español. I have summarized the results in the following table, in which the extension of the renewal process is abundantly clear: 174
Table 23: Renewal of Que in Medieval Spanish Texts 9-11th C 12th C 13th C texts texts texts por que / porque ‘because’ (de) pues que / despues que ‘after’ des que / desque ‘since’ (temporal) fasta que ‘until’ ante que ‘before’ pues que / puesto que ‘since’ (causal) para que ‘so that’ (purpose) ansi que / assi que ‘so that’ (consecutive) aun que / aunque ‘although’ TOTAL
14th C texts
15th C texts
0
1
3
10
6
0
1
1
1
1
0
0
2
3
2
0
0
1
2
0
0
0
0
1
0
0
0
0
0
3
0
0
0
0
4
0
0
0
0
2
0
0
0
0
5
total 0
total 2
total 7
total 17
total 23
Not only is the overall frequency of conjunctions much higher in the 15th century texts (twenty-three instances in the 15th century versus zero before the 12th century), but the number of forms has increased substantially as well. There are only examples of two renewed forms in the twelfth-century texts, while there are seven different forms in the texts from the fifteenth century. The results provided above may lead one to question the “gap” evident before the twelfth century, in which no renewed forms are seen. An important consideration here is that this table reflects a relatively small sample of data and
175
renewed forms (particularly causal and temporal conjunctions) may indeed have existed during this period. Furthermore, if this sample does accurately represent the linguistic reality of the medieval period, this does not imply incompatibility with the grammaticalization process outlined in this chapter.
After all, the
specialization process was still ongoing during the medieval period and examples of quod as a competing complementizer can still be found in the eleventh century (see example 4-38).
In this way, one could argue that it was not until the
specialization of que had reached its completion (and the use of que was virtually categorical) that renewed forms were able to surface. Finally, once the renewal process was underway, additional forms would likely be created at an increasingly rapid rate due to analogy, thus justifying the drastic increase in forms during the fourteenth and fifteenth centuries. It is clear that the overwhelming majority of conjunctions in modern Romance languages are the result of this renewal process. In fact, Grevisse provides a list of sixty French subordinating conjunctions, fifty-eight of which are based on que (1993: 1536). The table below lists some of the most common Spanish and French conjunctions, grouped into their respective categories:
176
Table 24: Modern Spanish and French Subordinating Conjunctions Conjunction type English Spanish French Causal
Temporal
Consecutive
‘because’
porque
parçe que
‘since’
ya que
puisque
‘before’
antes de que
avant que
‘after’
después de que
après que
‘until’
hasta que
jusqu’à ce que
‘that’
que
que
Purpose
‘so that’
para que
pour que
Conditional
‘unless’
a menos que
à moins que
Adversative
‘although’
aunque
quoique
In essence, the grammaticalization process, as outlined in this chapter,48 effectively tore down the Latin system, and then rebuilt it around the most highly generalized form.
From my perspective, however, this is not a meaningless
occurrence, since the resulting conjunction system is both transparent and rightbranching, which is not at all surprising given the nature of Romance typological developments.
48
The grammaticalization process described in this chapter did not have the same effects on the Latin conjunctions quando and si, which still exist in modern Romance. They were not generalized to the extent of the other conjunctions, as quando was consistently used for ‘when’ and si was categorically conditional in nature. This relatively low level of ambiguity is quite possibly the reason they have not been ousted from the system.
177
4.5. Cross-Linguistic Support The
grammaticalization
of
relative
pronouns
into
conjunctions
and
complementizers in Romance languages is not without precedent. In fact, parallel developments are attested in several languages. In Biblical Hebrew, for example, a non-restrictive relative clause, as in (4-53), was reinterpreted as a complement clause, as in (4-54): (4-53) Heb.
?al
tir?u-ni
she-?ani
shaxoret
NEG
see-3SG-IMPERF—me
REL-PRON—I
dark
‘don’t see me that I am dark skinned’ (Givón 1991: 289 in Cristofaro 1998: 64) (4-54) Heb.
yada”ti
she-gam
zeh
hu
ra”yon
know-1SG-PERF
COMP—also
this
it
notion
ruah of wind
‘I knew that this too was total nonsense’ (Givón 1991: 297 in Cristofaro 1998: 65) Similar trajectories are found in Vedic Sanskrit, Hittite and Ancient Greek. In Sanskrit, the conjunction yád (from the PIE neuter accusative relative *yod), was quite generalized and appeared in temporal, causal and conditional clauses (cf. Lehmann 1974: 167-172). The Hittite particle ku- (derived from the PIE stem *kw), developed into a causal and temporal conjunction (kuit), and finally, into a general subordinator (Cristofaro 1998: 65-66).
In Ancient Greek, the
complementizer o{ti was first used in correlative dyptichs, then in complement 178
clauses, and finally as a general subordinator functioning in a variety of clause types (Cristofaro 1998: 66-68). These data lend additional support to the claims presented in this chapter. The developments posited for the Latin conjunctions are both semantically and pragmatically motivated, and as a result, crosslinguistic similarities are not difficult to find.
4.6. Conclusion In this chapter, my goal was to provide a thorough treatment of Latin subordinating elements within the framework of grammaticalization. Using such a framework to carry out the analysis has been valuable for several reasons. The generalization process reveals some of the significant underlying psychological connections between notions such as time and cause, while decategorialization focuses on a specific morpho-syntactic change of fundamental importance: the creation of the complementizer. The processes of specialization and renewal address the way in which the old system was broken down, along with the way in which the new Romance system was developed around a single unifying element. It is my hope that the above discussion has drawn attention to the significant role that seemingly subtle changes in minor categories can play in the structural development of a language. It is precisely this new status of grammaticalized que which will provide the foundation for the issues to be discussed in chapter five.
179
CHAPTER 5 SYNTHESIS OF NON-CONTIGUOUS ELEMENTS 5.1. Introduction In chapters three and four, I analyzed several interrelated developments involving subordination techniques in Latin and Romance languages. I argued that an increase in right-branching structures led to an increase in subordinating conjunctions, thereby facilitating the grammaticalization of these conjunctions, which ultimately—specifically via the process of specialization—led to an increase in the frequency of que as a generalized subordinator. In my opinion, the chain of events does not end here. In this chapter, I will address one of the repercussions of the grammaticalization process and corresponding specialization of a single subordinating element. I believe that the high frequency (and almost categorical use) of the element que has had significant effects on Romance mood, in part as a result of what I view as synthesis between que and the subjunctive. This synthesis has contributed to the destabilization of the subjunctive paradigm due to what may be perceived by speakers as a redundant feature. In this chapter, I will first attempt to clarify the terms analytic and synthetic, whose history has been plagued with a vast amount of ambiguity and controversy. I will then provide an assessment of the contribution of Schwegler’s (1990) Analyticity and Syntheticity, particularly in relation to his claims regarding the synthesis of non-contiguous elements. Subsequently, I will present my own 180
views on synthesis, which differ in several ways from the mainstream, and will then explore the extent to which this perspective can be applied to que and the subjunctive.
5.2. The Analytic / Synthetic Dichotomy The attempt at a typological classification of languages not according to genetic criteria but according to notions of syntheticity and analyticity was initiated by Schlegel in the early nineteenth century. According to Schlegel (1971 [1818]), languages can be categorized into three types: 1) languages with no grammatical structure 2) languages with affixes 3) languages with inflection (Schlegel 1971 [1818]: 14) He further divides inflected languages into analytic (e.g. French) and synthetic (e.g. Latin) groups. He defines analytic languages as those which use articles with nouns, personal pronouns with verbs and have auxiliary verbs, and defines synthetic languages simply as “celles qui se passent de tous ces moyens de circonlocution” (1971 [1818]: 16). Furthermore, Schlegel is quite categorical in his assessments, stating that all daughter languages of Latin have an entirely analytic grammar. His bias toward synthetic languages is evident, as he discusses the widespread admiration for the beauty of synthetic languages (1971 [1818]: 181
17), while claiming that all analytic languages known to us “sont nées de la décomposition des langues synthétiques” (1971 [1818]: 16). Although the work of Schlegel was rejected by many,49 interest in synthesis and analysis was revived by Sapir (1949 [1921]) in chapter six of his monograph, Language, entitled “Types of Linguistic Structure”. He classifies languages into three types: analytic, synthetic and polysynthetic.
This
classification, however, is still somewhat ambiguous. According to Sapir, an analytic language “does not combine concepts into single words at all (Chinese) or does so economically (English, French)” (Sapir 1949 [1921]: 128). A synthetic language clusters concepts “more thickly” (128), while a polysynthetic language is “more than ordinarily synthetic” (128). It is difficult for linguists to define “words” and “concepts,” let alone vague notions such as “economically” or “more than ordinarily.” In spite of difficulties regarding terminology, Sapir does indeed offer a useful perspective on the terms synthetic and analytic. First of all, he points out that they are relative terms, by explaining that “a language may be ‘analytic’ from one standpoint, ‘synthetic’ from another” (Sapir 1949 [1921]: 128).
Secondly, he highlights their usefulness in diachronic rather than
synchronic linguistics, stating: “I believe the terms are more useful in defining certain drifts than as absolute counters” (1949 [1921]: 128).
49
He also
Humboldt, for example, criticized Schlegel’s non-genetic classification, which he believed to be invalid (see Schwegler 1990: 8-10).
182
acknowledges the widespread bias among linguists, who commonly exalt the superiority of synthetic languages. He refers to this as an “evolutionary prejudice . . . which is only now beginning to abate its tyrannical hold on our mind” (123). Ironically, Sapir himself is unable to entirely avoid this biased terminology, stating that synthetic languages “have frequently broken down into analytic forms” (1949 [1921]: 145, emphasis mine). Subsequently, research in the field became increasingly rigorous, as linguists attempted to offer precise criteria and linguistic tools to assess syntheticity and analyticity. A short, but very significant, article was published by Tauli (1945), providing a list of essential criteria to be considered: 1) whether a form is linear or alinear 2) whether a common or alternant stem is present 3) whether a common or alternant morpheme appears 4) whether the stem occurs independently or has independent meaning 5) whether the morpheme occurs independently or has independent meaning 6) whether a form is phonetically strong or weak 7) whether the elements are separable 8) whether the order of the elements may be changed 9) whether the morpheme appears with all words in coordinative constructions 10) whether there is agreement in subordinating constructions
183
11) whether the order of elements is progressive or regressive (Tauli 1945: 83-84) Clearly, these criteria focus on specific types of constructions of particular interest to Tauli (1945), since the underlying assumption is that there is indeed a stem and a morpheme present. Nevertheless, this is a great step toward identifying specific features which may be useful in comparing various linguistic forms. The focus of research from this point forward was to provide linguists with accurate quantitative tools to measure degrees of syntheticity. Greenberg’s (1960) well-known contribution is the synthetic index, or, the number of morphemes divided by the number of words (M/W). For any given language, the ratio generally falls somewhere between 1.00 and 3.00.50 His analysis of various languages yielded the following results: Annamite Persian English Anglo-Saxon Yakut Swahili Sanskrit Eskimo
1.06 1.52 1.68 2.12 2.17 2.55 2.59 3.72 (Greenberg 1960: 193)
50
I question Greenberg’s theoretical lower limit of 1.00 (Greenberg 1960: 185). How would this account for instances of redundancy such as negation in French? One could argue that one morpheme (NEG) expressed by two words (ne and pas) would result in an index of 0.50. I believe this issue serves to highlight the potential difficulties encountered in calculating Greenberg’s mathematical index.
184
In spite of the undoubtedly rigorous nature of such an analysis, it is nevertheless characterized by several limitations.
The inability to accurately define the
morpheme and the word makes the synthetic index difficult to calculate with precision. In addition, assigning one single synthetic index to an entire language reveals little about the language itself. For example, two languages with a score of 1.50 may have surprisingly little in common. One may be consistently analytic while the other may contain a combination of forms on either end of the spectrum. That is to say, there may be several polysynthetic forms canceled out, as it were, by competing analytic tendencies also present in the language.
In short,
languages can only be understood if we examine specific structures rather than attempt to assign a numeric “score” to languages as a whole. This issue casts doubt on the usefulness of such a mathematical approach in providing meaningful typological descriptions.51
5.3. Schwegler’s (1990) Analyticity and Syntheticity Departing from this index-based approach, Schwegler’s (1990) monograph dedicated to the topic of analyticity and syntheticity has made a significant
51
There were various subsequent attempts at a successful index of syntheticity. For example, Krupa (1965) inverts Greenberg’s formula, preferring to divide the number of words by the number of morphemes. Slavičková (1968) divides the number of root morphemes by the total number of morphemes, while Kelemen’s (1970) index is based on the number of analytical words divided by the total number of words (see Schwegler 1990: 24).
185
contribution to the field. From my perspective, Schwegler’s analysis has three main strengths. Firstly, he views synthesis primarily as a process and, as a result, of more use to diachronic rather than synchronic analysis. Secondly, he claims that synthesis is not measurable, quantifiable or absolute and flatly rejects synthetic indices or formulas based on identification of the word. Thirdly, he discards the traditional restriction that synthesis can only occur between contiguous elements, offering instead a much more innovative approach to the notions of syntheticity and analyticity. Schwegler provides several concise definitions for the terms analyticity, syntheticity and synthesis.
Analyticity is defined as “the semantic, syntactic,
morphological and phonological autonomy of morphemes within a speech unit” (Schwegler 1990: 48). Syntheticity, on the other hand, is “the semantic, syntactic, morphological, and phonological interdependency (or relatedness) of morphemes within a speech unit” (Schwegler 1990: 48). The term synthesis, then, is used to refer to the corresponding process in which the “tightness” of relationships increases. Unfortunately for the researcher, semantic synthesis occurs in the mind of the speaker and is not “visible” or testable. However, there are two crucially important factors which facilitate semantic synthesis: generality and relevance (Schwegler 1990: 52-54). Generality refers to the specificity—or lack thereof— of a given lexeme. In order for synthesis to occur between a linguistic element and other elements, its semantic content must be relatively non-specific. 186
The linguistic element in question must also be relevant to another element in order for synthesis to occur. For example, person markers are relevant to verbal stems, but not to nominal stems. Schwegler explains his perspective as follows: “The question of relevance is particularly essential to the idea of syntheticity because it directly conditions the likelihood of a change in the analytic/synthetic tension of speech units” (Schwegler 1990: 53). An increase in semantic synthesis, according to Schwegler, is likely to lead to syntactic, morphological and phonological synthesis. Schwegler (1990) describes this trajectory as follows: Initially, synthesis is always of a purely semantic nature. Two or more items enter a state of mutual relevance, and, because of the cultural or cognitive salience of the new semantic association, typically begin to occur with an increased frequency. At this point, syntactic factors begin to play a decisive role. (Schwegler 1990: 72) The morpho-syntactic developments are related to the following criteria: 1) Separability:
whether the two items are separate (+ analytic) or contiguous (+ synthetic)
2) Linearity:
whether the two items are linear (+ analytic) or fused (+ synthetic)
3) Displaceability:
whether the order of elements is variable (+ analytic) or fixed (+ synthetic)
4) Isolatedness:
whether an item can appear alone (+ analytic) or not (+ synthetic)
5) Obligatoriness:
whether an item requires the presence of another (+ synthetic) or not (+ analytic) 187
6) Transparency:
whether an item exhibits allomorphic variation (+ synthetic) or not (+ analytic)
Since, in terms of diachronic change, the relevance of these notions lies in their movement along a continuum, I have created the following visual representation: Illustration 1: Analytic / Synthetic Continuum ANALYTIC + separable
SYNTHETIC - separable
+ linear
- linear
+ displaceable
- displaceable
+ isolated
- isolated
- obligatory
+ obligatory
+ transparent
- transparent
In short, an element undergoing a process of synthesis will become less separable, more fused, less displaceable, less likely to stand alone, more obligatory and less transparent. Using these six criteria above as a guide, Schwegler (1990) has analyzed several classic examples of synthesis in Romance languages affecting both verb phrases and noun phrases. For example, he addresses in detail the development of future and present perfect tenses with Latin habere, describing the semantic, syntactic, morphological and phonological developments that took place. He considers the major impetus for synthesis in both cases to be the reanalysis and subsequent semantic integration of habere with the infinitive and participle, respectively. This development can be illustrated as follows:
188
(5-1)
[ [ quid habemus ] dicere ]
>
[ quid [ habemus dicere ] ]
[ [ bona parta ] habemus ]
>
[ bona [ parta habemus ] ] (Schwegler 1990: 126)
The semantic synthesis of these elements paved the way, so to speak, for morphosyntactic synthesis, whereby word order became fixed and habere became bound.52 The two elements, having undergone semantic and morphosyntactic synthesis, also became increasingly unified phonologically. However, unlike the left-branching future construction, the elements which make up the rightbranching present perfect do not always occur within the same cursus, due to the potential insertion of adverbs. Schwegler (1990) explains that “phonological unity is often overriden by stronger, syntactic considerations” (Schwegler 1990: 136). In addition to the analysis of these more well-established phenomena, Schwegler (1990) includes an analysis of predicate negation in French, presenting an innovative argument for the possible synthesis of non-contiguous elements. He claims that “some relevant units may undergo the analytic/synthetic cycle without ever being juxtaposed” (Schwegler 1990: 151). In spite of the noncontiguous position of the elements ne and pas within the sentence, several criteria for synthesis are identified. First of all, there was a marked increase in 52
The obvious difference between these two constructions in modern Romance languages is that the future is left-branching while the present perfect is right-branching. For each periphrasis, the prevalent word order at the time at which the structure became the “the priviledged structure” (Bauer 2006: 301) has been preserved.
189
syntactic restrictions. Although they originally exhibited substantial syntactic freedom, the two elements—ne and pas—were restricted to the verb phrase by the Middle French period.
Furthermore, ne and the finite verb could only be
separated by certain object clitics, while pas was less frequently separated from the verb by lexical items. A second feature of negation in French which is indicative of synthesis relates to obligatoriness and morphological freedom. While pas was entirely optional in its early stages as a negative particle, by the seventeenth century, it was well on its way to being obligatory (Schwegler 1990: 155-156). A third indication of the synthesis of ne…pas is the phonological unity of the two elements, which are almost always part of a single cursus, as shown below: (5-2)
Fr.
/ilnapakõpri/ il
n’a
pas
compris
he
NEG—have-3SG-PRES-IND
NEG
understand-PP
‘he hasn’t understood’ (5-3)
Fr.
/ilnatužurpakõpri/ il
n’a
toujours
pas
he
NEG—have-3SG-PRES-IND
still
NEG
compris understand-PP
‘he still has not understood’ (Schwegler 1990: 156)
190
Finally, it is well known that instead of maintaining two markers of negation, ne tends to be eliminated in spoken Modern French.
Schwegler explains this
phenomenon in a quite straight-forward way: “languages tend to eliminate nonfunctional redundancy” (Schwegler 1990: 158). Analyticity and Syntheticity offers not only a thorough presentation of widely accepted instances of synthesis (e.g. tenses with habere), but also offers a relatively new approach to the subject, by acknowledging the possibility of synthesis between non-contiguous elements.
This contribution is of great
significance for two main reasons. Firstly, it underscores the role of continuum and process rather than dichotomy in the notions of synthesis and analysis. Secondly, by considering non-contiguous elements, Schwegler refocuses our attention away from inflection, which is all too often immediately associated with syntheticity. For this reason, this perspective is much more valuable, particularly for diachronic linguists.
5.4. A New Perspective on Synthesis Although Schwegler’s (1990) broader view of synthesis enables us to recognize synthesis between non-contiguous elements, the inclusive nature of his work in some ways lacks precision with regard to diachronic processes themselves. Schwegler refers to synthesis, cycles, grammaticalization and drift with little or no clarification or useful distinctions. In my view, it is essential that we make our 191
definitions as explicit as possible, thereby avoiding overlap and ambiguity. If grammaticalization and synthesis are one and the same phenomenon, then our terminology should be updated to reflect this. If, however, they are independent processes—as I believe they are—then we must be diligent in isolating and describing them independently of one another. In the sections that follow, I will offer my own views on synthesis, some of which are in line with existing theories on the subject, and some of which differ in significant ways.
5.4.1. Synthesis as a Semantic Process My perspective on synthesis is based primarily on the notion that synthesis is fundamentally a semantic process involving an increase in the speaker’s perceived tightness of the relationship between two elements. Morpho-syntactic changes may occur as a result of this synthesis, but not necessarily so. A view such as this avoids categorizing elements as synthetic and analytic based exclusively on their surface structure. For example, can we really say that Sp. del ‘of the’ (M-SG) or Sp. conmigo ‘with me’ are more synthetic than Sp. de la ‘of the’ (F-SG) or Sp. con ella simply because orthographic convention would have them written as one word instead of two? We could argue, of course, that morpho-phonological synthesis/fusion has occurred, but certainly not a deeper, more fundamental, semantic synthesis. And therefore, what significance would such a claim have?
192
There is, of course, the issue of measurability to contend with. Synthesis—as a semantic phenomenon—is not measurable in the traditional sense. Nevertheless, it deals with speaker perceptions and, in this way, may ultimately be responsible for significant changes such as reanalysis, which reflect this underlying psychological process. On the other hand, synthesis in a morphosyntactic sense is quantifiable and measurable.
Unfortunately, it is also
influenced by language-specific factors that may prevent the results of such measurements from accurately reflecting the true nature of the relationship between the linguistic elements. As Tesnière noted so perceptively three-quarters of a century ago, the problem with most studies on syntheticity is that they rely on the word in an orthographic sense. Discussing this tendency, he states: “Il semble qu’elle se fonde sur une apparence plus que sur la réalité” (Tesnière 1932: 62). He also asserts: “Notre orthographe continue à analyser, d’après le modèle latin, des mots que notre conscience linguistique a synthétisés depuis longtemps” (Tesnière 1932: 63, emphasis original). For this reason, I believe it is more meaningful to view synthesis as a purely semantic phenomenon, in spite of its lack of measurability.
5.4.2. Synthesis as a Unidirectional Process Also fundamental to my perspective is the belief that synthesis—like grammaticalization—is a unidirectional process. 193
It has been argued that the
process is cyclical in nature (cf. Schwegler 1990), due to changes such as those involved in the development of future tenses in the history of Romance language, as shown below: *ama bho > amabo > amare habeo > amaré > voy a amar > ? analytic
synthetic
analytic
synthetic
analytic
Upon closer analysis, however, such a claim is clearly untenable. It is true that future tense was expressed via analytic forms, then synthetic forms, and again by analytic forms. However, we are not dealing with the continuous development of a single form, but rather the development of several forms appearing at different time periods. What we have in the trajectory shown above, then, is not a cycle, but three separate unidirectional paths involving synthesis, which would be better represented as follows: 1. *ama bho
>
amabo
2. amare habeo
>
amaré
3. voy a amar
>
??
The fact that these are three distinct processes is supported by the fact that they involve different auxiliary verbs (i.e. PIE. *bheu ‘be’, Lt. habere ‘have’ and Sp. ir ‘go’) in addition to exhibiting different branching patterns: the first two, leftbranching and the third, right-branching. Each of these three developments is, without a doubt, unidirectional, moving always towards a higher degree of synthesis. There exists no parallel development which increases the analyticity of
194
a form or construction.53 This unidirectionality is one of the similarities between synthesis and grammaticalization which often results in an unfortunate lack of clear distinctions between the two processes.
5.4.3. Synthesis versus Grammaticalization One of the common characteristics of research on synthesis is confusion between synthesis and similar or related processes. This is not surprising, however, since in many instances of linguistic change we do find evidence of both processes. This is due, in part, to the fact that grammaticalization results in closed-class elements, which tend to be more dependent upon other linguistic elements. For example,
Romance
adverbs
formed
with
–mente
involve
both
the
grammaticalization of mente and the synthesis of mente with the adjective. It is difficult to imagine one process occurring without the other. On the other hand, the grammaticalization of the Romance definite article is much easier to analyze exclusively in terms of grammaticalization.
While there is a high level of
synthesis between the article and the noun, the grammaticalization and synthesis processes are not as closely interconnected. It is, therefore, not uncommon for scholars to refer simultaneously to both grammaticalization and synthesis in their descriptions of a single development. 53
A similar view is held by Pulgram (1963), who makes a sharp distinction between the gradual synthesis of forms, which “connotes a linguistic change which for the most part escapes the control and the attention of the speaker,” while a shift to innovative analytic forms “testifies to the occurrence of an act of choice” (Pulgram 1963: 37).
195
For example, in Schwegler’s (1990) discussion of present perfect and future tenses with habere, instead of focusing on the synthesis of habere with participle and infinitive, he often refers to the development of habere itself. He points out that habere was originally used with a more specific modal value (in the case of future) or aspectual value (in the case of present perfect), but that “semantic extension” (Schwegler 1990: 127) occured. In his discussion of negation, he mentions that the Latin nouns passum ‘step’, micam ‘crumb’, etc. lost their nominal properties (e.g. ability to inflect for number) and “their grammatical status shifted” (Schwegler 1990: 156). In spite of the fact that Schwegler’s (1990) monograph is a study on synthesis, he spends a great deal of time referring to grammaticalization processes such as generalization and decategorialization. Furthermore, two of Schwegler’s (1990) criteria for synthesis—isolatedness and transparency—are better suited as indicators of grammaticalization.
Along
similar lines, C. Lehmann (1995) refers to three syntagmatic parameters in his work on grammaticalization:
structural scope, bondedness and syntagmatic
variability (see section 4.3). These criteria, in my view, are strictly criteria of synthesis. There is no question that it is often necessary to refer to both processes in the same analysis (if one wishes to account for a full diachronic trajectory). However, I believe it is important to recognize that there are two processes at
196
work, one paradigmatic and one syntagmatic,54 if only in order to prevent the misconception that one process automatically entails the other.
5.4.4. Signs of Synthesis: Conditions and Criteria In my view, there are several pre-conditions which facilitate the synthesis process as well as certain morpho-syntactic signs which may indicate that synthesis has occurred. I believe semantic synthesis is facilitated by a higher frequency of occurrence, which increases the likelihood of the speaker feeling a closer association between two elements. Following Schwegler (1990), I also believe that generality and relevance are important conditions for two elements to undergo synthesis. Although Schwegler (1990) claims that there are six morpho-syntactic criteria indicative of synthesis (see section 5.3 above), I believe this list needs to be further categorized, and perhaps, reduced.
I propose the following
classification:
54
In this context, the term paradigmatic refers to the nature of a single element in isolation, while the term syntagmatic refers to the way in which a given element interacts with other elements (see pp. 132-134 for a discussion of C. Lehmann’s [1995] work in this area).
197
Table 25: Criteria for Synthesis I obligatoriness true indicator of synthesis separability II
may occur if synthesis has occurred, but only given
linearity displaceability
III
specific syntactic environments
isolatedness
indicators of grammaticalization (paradigmatic rather
transparency
than syntagmatic)
In other words, I would eliminate category III completely, as I believe these are paradigmatic criteria which indicate grammaticalization. The notions in category II are not inaccurate, but they are not truly “criteria” since they do not necessarily occur in synthetic forms and, thus, their absence does not preclude synthesis. Category I, on the other hand, is the strongest indicator of synthesis, since it will occur if two elements are felt to be strongly connected, and it is not restricted by issues of word order or branching tendencies.
For example, consider the
following examples of future and present perfect with habere: (5-4)
Sp.
trabajaré
(5-5)
Fr.
travaillerai
(5-6)
Sp.
he trabajado
(5-7)
Fr.
j’ai travaillé
First and foremost, semantic synthesis has occurred, and the morpho-syntactic condition of obligatoriness is met. In this way, these forms can all be considered synthetic, in spite of the apparent differences in separability, linearity and
198
displaceability. The lack of displaceability in all four constructions is not a sign of synthesis—as Schwegler (1990) has argued—but rather a sign of relatively fixed word order in modern Romance languages (i.e. L-B in earlier stages and R-B in later stages). Furthermore, the differences in terms of separability and linearity between present perfect forms in (5-6 and 5-7) and future forms (in 5-4 and 5-5) are due not to differences regarding syntheticity but to differences in branching, since right-branching forms are not likely to fuse.
5.5. Que and Subjunctive: Synthesis of Non-Contiguous Elements Following the framework presented above (which relies most heavily on semantic association and obligatoriness), I will attempt to show evidence that the subordinator que and the subjunctive have moved closer to the synthetic end of the synthetic-analytic spectrum. First, I will discuss the pre-existing conditions which facilitated this process, along with signs of a closer semantic association between the two elements.
Secondly, I will examine the extent to which
obligatoriness can be seen among the elements.
Finally, I will consider
Schwegler’s morpho-syntactic signs of displaceability, linearity and separability. Although I do not believe these to be true tests for synthesis, the trend toward contiguity can nevertheless be observed.
199
5.5.1. Semantic Synthesis Frequency, generality and relevance are all favorable conditions for semantic synthesis. In the case at hand, the first two of these conditions are clearly met, as the Romance subordinator, que, is both frequent in discourse and extremely general in meaning due to the extended grammaticalization process explored in the previous chapter. Furthermore—and more importantly—there is arguably a high degree of relevance between que and the subjunctive. While que is the predominant marker of syntactic subordination in Romance languages, the subjunctive mood is also used to subordinate propositions, both semantically and syntactically (cf. sections 2.3.2.3 and 3.4.1). Although que appears with all finite verbs, there is an indication that the relevance between que and subjunctive is stronger than that between que and the indicative. By looking at two medieval texts, one can see that combinations of que with subjunctive are more abundant than combinations of que with indicative. Consider the following Old Spanish text, which I have reproduced from Gifford and Hodcroft (1966). All examples of que are in bold, with indicative verbs in italics and subjunctive verbs underlined: (5-8)
OSp. Ego donna Sol, abbadessa de Sancta Maria la Real de Burgos, do una terra que es en Duraton, a medias, a poner maiolo, a don Feles et a Lobo & con toda sua frontada del rio; & que fagan en el rio de duos molinos fata tres, otro si
200
a medias, & los molinos que sean fectos fata Sant Michael, & que los faga don Feles & don Lobo, asi quomo molinos deuen seder con todo suo apareiamento. Et de pues que los molinos fueren fectos, si agua abinere que crebante en la pesquera alguna cosa, que lo fagan a medias; & si portello en los molinos alguno crebantaret, otrosi a medias; & si la pesquera o los molinos leuare el agua ques assolen, que los faga don Feles et don Lobo. Et esta terra que sea la media oganno posta & que lo labren lo doganno si maes non puderen duas uices, et lo al que remanecere que sea posto logo otro anno, & descend arriba que lo labren cadaano tres uices. Et quando el maiolo leuare & los molinos fueren fectos & moleren, cuando el abadessa quisere, que partan. Et desta terra son aladannos: Iohan Martinez, filio de Martin Anaiaz; & de alia parte Martin Martinez, filio de Martin Dominguez; & de alia parte Gonsaluo Martinez & Domingo Petrez, filio de Petro Sordo, & una terra de Sancto Domingo que tene en fronte. Et dio por mano el abadessa a frair Iohan que fue de Mazola, que los metesse en la terra & en el rio a medias, a fondos terra; & el metiolos en ello, otro si quomo el abbadessa mando. (Gifford and Hodcroft 1966: 29) In this twelfth-century text, there are fourteen instances of subjunctive with que (e.g. in volitive, anticipatory, purpose and relative clauses), but only three instances of indicative with que (all relative clauses). Now consider La cantilène de Sainte Eulalie, a ninth century French text: 201
(5-9) OFr.
Buona pulcella fut Eulalia, Bel auret corps, bellezour anima. Voldrent la veintre li Deo inimi, Voldrent la faire diaule servir. Elle no’nt eskoltet les mals conselliers Qu’elle Deo raneiet chi maent sus en ciel. Ne por or ned argent ne paramenz Por manatce regiel ne preiement; Niule cose non la pouret omque pleier, La polle sempre non amast lo Deo menestier. E por o fut presentede Maximiien, Chi rex eret a cels dis soure pagiens. Et li enortet, dont lei nonque chielt, Qued elle fuiet lo nom christiien Ell’ent aduret lo suon element. Melz sostendreiet les empedementz Qu’elle perdesse sa virginitét. Por os furet morte a grand honestét. Enz enl fou la getterent, com arde tost. Elle colpes non auret, por o nos coist. A czo nos voldret concreidre li rex pagiens; Ad une spede li roveret tolir lo chief La domnizelle celle kose non contredist: Volt lo seule lazsier, si ruovet Krist. In figure de colomb volat a ciel. Tuit oram que por nos degnet preier Qued auuisset de nos Christus mercit
202
Post la mort et a lui nos laist venir Par souue clementia. In this text, the preponderance of subjunctive with que is even clearer, since all five examples of que are found with verbs in the subjunctive. For this reason, it is my belief that speakers came to associate que with the subjunctive due to a natural cognitive process. Pons Bordería (2003), in his work on the modal value of que, also refers to this apparent coreferentiality.
He states: “La presencia del
subjuntivo y la de que nos parecen manifestaciones de un mismo fenómeno” (Pons Bordería 2003: 538-539). With regard to French, Ritchie (1907) states: “Nous avons déjà vu qu’il y avait peut-être en ancien français une tendence à considérer que comme partie intégrante du mode subjonctif” (Ritchie 1907: 160). This underlying process is the most important in the synthesis of two elements, even though it is not typically outwardly visible. I believe there is a phenomenon in Spanish which does, in fact, point to a merging, or transfer, of meaning between the subjunctive mood and que. The Romance subordinator que appears to have acquired modal value, and now functions as a marker of modality with several functions, including reiteration (e.g. disagreement, protest, surprise), emphasis, hearsay and non-informative value (see Porroche Ballesteros 2000 and Pons Bordería 2003).
Pons Bordería
labels que a modality marker or “modalizador” (2003: 543), and Porroche Ballesteros defines the underlying meaning as “la voluntad, por parte del
203
hablante, de realizar un comentario a propósito de otro enunciado que constituye la parte fundamental de su comunicación” (Porroche Ballesteros 2000: 4).
5.5.2. Obligatoriness As explained in detail in section 5.4.4, I believe obligatoriness is the only categorical indicator of synthesis, because it is not restricted by certain word order or branching patterns. Therefore, this section will focus on the extent to which the synthesis of the Romance subjunctive and the subordinator que can be justified. I will present obligatoriness from both directions; that is to say, I will show an increase in que in the presence of the subjunctive and an increase in subjunctive in the presence of que.
5.5.2.1. Use of que as Attraction to Subjunctive In this section, I will illustrate an increase in the use of que, which I believe to be attributable to the presence of the subjunctive mood. In other words, que begins to appear in non-subordinate contexts, no longer serving as a connector. Using Old Spanish data from Jensen and Lathrop (1973) and Old French data from Jensen (1974), I have isolated this particular phenomenon.
In Old Spanish,
subjunctive in independent clauses typically appeared alone (as was common in Latin, as well):
204
(5-10) OSp. Dios God
salve
a
nuestros
amigos
save-3SG-PRES-SUBJ
PTC
our
friends
‘may God save our friends’ (Hita, L.B.A. 3038 in Jensen and Lathrop 1973: 15) (5-11) OSp. ¡quemada burn-PP-F
seas,
alcahueta
falsa!
be-2SG-PRES-SUBJ
procuress
false-F
‘may you be burned, false procuress’ (Celest. 92,23 in Jensen and Lathrop 1973: 16) (5-12) OSp. ame love-3SG-PRES-SUBJ
e
honra
and
honor
a
su
fama
PTC
your-SG
reputation
‘love your reputation and honor’ (Talav., Corv. 207 in Jensen and Lathrop 1973: 16) This usage occurs in volitive constructions, with the above examples expressing wishes, curses and commands. Although not exceedingly common, que began to appear in independent clauses, as illustrated in the following examples: (5-13) OSp. ¡que that
Dios
me
la
mantenga!
God
me-DAT
her-ACC
keep-3SG-PRES-SUBJ
‘may God keep her’ (Hita, L.B.A. 939a in Jensen and Lathrop 1973: 15)
205
(5-14) OSp. ¡que that
Santillán
la
confonda!
Santillán
her-ACC
confound-3SG-PRES-SUBJ
‘may Santillán confound her!’ (Hita, L.B.A. 963b in Jensen and Lathrop 1973: 16) (5-15) OSp. que that
ella
ruegue
al
criador
she
beg-3SG-PRES-SUBJ
PTC—the
creator
‘she must beg the creator’ (Egip. 1443 in Jensen and Lathrop 1973: 16) These sentences express the same wishes, curses and commands, yet with the addition of que. French exhibits the same trajectory with clauses of this type. In Old French, the subjunctive in independent clauses could appear with or without que, although clauses without que were more common: (5-16) OFr.
repos
aient
en
paradis
rest-OBL
have-3PL-PRES-SUBJ
in
paradise-OBL
‘may they have rest in paradise’ (Vill., Test. 231 in Jensen 1974: 16) (5-17) OFr.
Dieus
confonde
ton
God-NOM
confound-3SG-PRES-SUBJ
your-OBL
chief! leader-OBL
‘may God confound your leader’ (Nîmes 736 in Jensen 1974: 16)
206
(5-18) OFr.
s’il
ne
la
scet,
if—he-NOM
NEG
it-ACC
know-3SG-PRES-IND
voise
l’aprendre
go-3SG-PRES-SUBJ
it-ACC—learn-INF
‘if he doesn’t know it, he should go learn it’ (Vill., Test. 38 in Jensen 1974: 16) The use of subjunctive alone was prevalent until the fourteenth century, yet many examples of que with subjunctive can be found: (5-19) OFr.
que
Dieu
beneiçon
Vous
that
God-NOM
grace-OBL
you-DAT
doint give-3SG-PRES-SUBJ
‘may God give you grace’ (Miracles 3,1 in Jensen 1974: 16) (5-20) OFr.
que
Dex
le
confonde
that
God-NOM
him-ACC
confound-3SG-PRES-SUBJ
‘may God confound him’ (Chrét., Chev. 3220 in Jensen 1974: 17) (5-21) OFr.
s’il
est
si
hardiz
if—he-NOM
be-3SG-PRES-IND
so
bold-NOM
qu’il
i
veigne!
that—he-NOM
LOC-PRON
come-3SG-PRES-SUBJ
‘if he is so bold, let him come!’ (Passion 1355 in Jensen 1974: 18)
207
By the Classical French period (i.e. seventeenth and eighteenth centuries), the construction with que was very well established and it is now extremely widespread—if not obligatory—in Modern French (Jensen 1974: 15-18). This trend is true of both French and Spanish, as que in independent clauses in now the norm,55 as illustrated by the examples below: (5-22) Sp.
¡que
se
vaya
that
3SG-REFL-PRON
go-3SG-PRES-SUBJ
inmediatamente! immediately
(5-23) Fr.
qu'il
parte
tout de suite!
that—he-NOM
leave-3SG-PRES-SUBJ
right away
‘that he go immediately’ (‘he should go immediately’) I have found one additional trend which supports this increase in obligatoriness in spoken (i.e. informal) French.
Anderson (1993) set out to
analyze the frequent absence of the complementizer in French spoken in Paris. Both corpora used in this study revealed a striking pattern.
Que is used
consistently when the clause is felt to be subordinate both in terms of syntax and semantics/pragmatics. However, que is omitted 30%-35% of the time when the clause is merely syntactically subordinate, which Anderson refers to as “fausse subordination” (1993: 10). Consider the contrast between examples (5-24) and
55
The exception to this is, of course, fixed expressions which preserve the older structure (e.g. Sp. ¡Viva el rey!, Fr. Vive le roi! ‘long live the King’).
208
(5-25), in which an asterix represents the absent complementizer, and (5-26) and (5-27), in which the complementizer is present: (5-24) Fr.
je
me
souviens
I-NOM
1SG-REFL-PRON
remember-1SG-PRES-IND
* *
on
avait
des
one-NOM
have-3SG-IMPERF-IND
INDEF-ART-PL
des
profs
au
INDEF-ART-PL
teachers
at—DEF-ART-SG
lycée
qui . . .
High School
who
‘I remember that we had some teachers in High School who . . .’ (Anderson 1993: 8) (5-25) Fr.
on
sait
par example
one-NOM
know-3SG-PRES-IND
for example
* *
cette
norme . . .
c’est
pas
this
norm
it-NOM—be-3SG-PRES-IND
NEG
parce que . . . because
‘one knows for example that this rule . . . isn’t because . . .’ (Anderson 1993: 8) (5-26) Fr.
il
faut
quand même
it-NOM
be necessary-3SG-PRES-IND
even so
qu’elles
fassent
des
that—they-NOM-F
do-3PL-PRES-SUBJ
INDEF-ART-PL
sports . . . sports
209
‘it’s still necessary that they play sports’ (Anderson 1993: 10) (5-27) Fr.
ils
veulent
pas
que
they-NOM
want-3PL-PRES-IND
NEG
that
j’aie
une
moto
I-NOM—have-1SG-PRES-SUBJ
a
motorcycle
‘they don’t want me to have a motorcycle’ (Anderson 1993: 11) Anderson has thus concluded that que is only obligatory when the main clause semantically modifies the subordinate clause, as is the case in clauses containing verbs in the subjunctive. This distributional pattern reflects precisely the level of synthesis I have proposed in this chapter. I view these related developments as a sign of synthesis between the subjunctive and the grammaticalized element que, which ultimately results in a high level of obligatoriness. I find the comments of Jensen (1974) valuable, as they indirectly support this claim. Although he does not use the term synthesis, he asserts that “que is a morpheme which, in addition to verbal endings, announces a subjunctive” (Jensen 1974: 15).
Furthermore, he addresses the
process in the mind of the speaker, as he states: “This has brought about a situation where que is felt to be an inherent morphological part of the subjunctive, as shown in many normative French grammars which list the subjunctive forms
210
preceded by a que” (Jensen 1974: 15).56 This observation is consistent with my perspective on criteria for synthesis. If we view synthesis exclusively as a matter of inflection, we are ignoring cases such as these, where a distinct unit appears to serve a morphological function.
5.5.2.2. Use of Subjunctive as Attraction to que The process involving an increase in obligatoriness has also ocurred in the opposite direction; that is to say, there has been an extension of subjunctive usage in the presence of que. This spread of subjunctive is a phenomenon which has complicated the most recent debates on the reduction in subjunctive usage. It is my belief that what—on the surface—appears to be an increase in subjunctive is, in fact, a sign of synthesis between these two non-contiguous elements. I believe subjunctive usage after expressions such as Sp. el hecho de que / Fr. le fait que and Sp. después de que / Fr. après que are indicative of this gradual process of synthesis. In Latin, an expression similar to the fact that was formed consistently by quod + indicative (cf. Bennett 1910: 124-126). In Spanish and French, although there is substantial variation, the subjunctive appears to be favored. Consider the following modern examples: 56
This refers to the majority of—if not all—French grammars, where the subjunctive paradigm of the verb ‘to be’ is presented as: que je sois, que tu sois, qu’il soit, que nous soyons, que vous soyez, qu’ils soient. This type of presentation is found occasionally though not in the majority of Spanish grammars.
211
(5-28) Sp.
el
hecho
de
que
en
los
cuatro
el
the
fact
of
that
in
the
four
the
régimen
democrático
se
regime
democratic
3SG-REFL-PRON
impusiera
facilitaba
impose-3SG-IMPERF-SUBJ
facilitate-3SG-IMPERF-IND
la
tarea
the
task
‘The fact that in the four (countries) the democratic regime was imposed facilitated the task’ (Krakusin and Cedeño 1992: 1290) (5-29) Fr.
les
experts
approuvent
le
fait
the
experts
approve-3PL-PRES-IND
the
fact
que
cette
loi
mette
sur
that
this
law
place-3SG-PRES-SUBJ
on
un
pied
d’égalité
les
couples
a
foot
of—equality
DEF-ART-PL
couples
mariés,
divorcés,
séparés
ou
marry-PP
divorce-PP
separate-PP
or
vivant
en
concubinage
live-PRES-PART
in
cohabitation
‘the experts approve of the fact that this law places on an equal footing those couples who are married, divorced, separated or living together’ (Blumenthal 1998: 149) Subjunctive usage here is often justified quite convincingly on pragmatic grounds. Nevertheless, this type of construction was expressed exclusively in indicative mood in Latin, and therefore, the shift is significant. 212
Another instance of subjunctive mood in previously indicative contexts (without semantic motivation) occurs after Sp. después de que and Fr. après que ‘after’. These clauses in Latin were only expressed via the present subjunctive if the action had not yet occurred. In Modern Spanish and French, however, we find examples of subjunctive after después de que and après que even when the action is completed: (5-30) Sp.
retornó
a
casa
después de
que
return-3SG-PRET-IND
to
house
after
that
su
tía
se
fuese
his
aunt
3SG-REFL-PRON
go-3SG-IMPERF-SUBJ
‘he returned home after his aunt had left’ (Hengeveld and Wanders 1997: 266) (5-31) Fr.
on
aurait
oublié
de
fermer
one
have-3SG-COND
forget-PP
about
close-INF
après
que
ça
soit
after
that
DEM-PRON-SG
be-3SG-PRES-SUBJ
fini finish-PP
‘we must have forgotten to turn it off after it was finished’ (Canut and Ledegen 1998: 31) Not unlike other unexpected uses of the subjunctive, this tendency is often attributed to analogy, in this case, with Sp. antes de que and Fr. avant que ‘before’ (cf. Canut and Ledegen 1998: 38-39). While it is definitely feasible to argue that the Romance preference for subjunctive resulted from analogy to a 213
similar temporal clause, I believe it is equally feasible to interpret it as a result of increased association between que and subjunctive. Finally, a very interesting phenomenon is found in French, which supports my argument for synthesis. Cases where there is coordination in conditional clauses exhibit a very interesting mood distributional pattern.
The first
conjunction is si (+ indicative) and the second is que (+ subjunctive). Consider the following examples from Le bon usage: (5-32) Fr.
chaque
larve
royale,
si
l’on
each
larva
royal
if
one
changeait
sa
nourriture
et
change-3SG-IMPERF-IND
his
food
and
qu’on
reduisît
sa
cellule,
that—one
reduce-3SG-IMPERF-SUBJ
his
cell
serait
transformée
en
ouvrière
be-3SG-COND
transform-PP
in
worker
‘every royal larva, if one changed her food and if one reduced her cell, would be transformed into a worker (bee)’ (Grevisse 1993: 1673) (5-33) Fr.
si
mon
nez
ne
vous
plaît
If
my
nose
NEG
you-DAT
please-3SG-PRES-IND
pas,
ou
que
la
coupe
de
mon
NEG
or
that
the
cut
of
my
corsage
ne
vous
paraisse
pas
blouse
NEG
you-DAT
seem-3SG-PRES-SUBJ
NEG
conforme à
la
mode
in accordance with
the
fashion
214
‘if my nose doesn’t please you, or if the cut of my blouse doesn’t seem to be in style’ (Grevisse 1993: 1674) According to Grevisse (1993), in the seventeenth century, indicative was still possible after this conditional que, yet today, subjunctive mood is the norm. It is, in this context, automatically conditioned by the presence of que. In conclusion, non-inherited cases of Romance subjunctive usage should not be seen, in my opinion, as a strengthening of mood as a category. On the other hand, relying on surface analogy alone seems far too superficial. Instead, this “spread” can be seen as an increase in obligatoriness of subjunctive mood in the presence of the generalized subordinator que. That does not imply that there are not additional semantic, pragmatic, or even external factors at work here. It simply means that there may have been an underlying association or increase in “tightness” in the mind of the speaker which facilitated these developments.
5.5.3. Displaceability, Linearity and Separability Now that we have examined semantic synthesis and the criteria of obligatoriness, we are left with Schwegler’s morpho-syntactic criteria of displaceability, linearity and separability. As explained in section 5.4.4, these may indicate levels of syntheticity, but the inverse is not true; in other words, their absence does not necessarily indicate a lack of synthesis. For instance, in terms of displaceability
215
(i.e. whether or not the order of elements is variable), we could conclude that que + subjunctive (with que always appearing first) has undergone synthesis. This claim would, of course, be completely illogical, as subordinating elements such as this have consistently been preverbal throughout all Latin and Romance periods. Word order in this regard is relatively rigid and, additionally, one must consider that the function of the subordinator is to introduce the subordinate clause. These syntactic restrictions render any assessment of synthesis based on displaceability invalid. In terms of linearity (i.e. whether the two elements are linear or fused), the sequence que + subjunctive appears to be analytic. Again, this criteria is simply not appropriate here. The linear—as opposed to fused—nature of the sequence is not a sign of analyticity but rather of Romance word-order constraints (e.g. preverbal pronouns and negation) as well as the right-branching nature of the construction.57 In other words, these criteria only assess syntheticity accurately in languages with flexible word order and left-branching structures.
This
observation is consistent with the perspectives of those scholars involved in the development of these criteria (e.g. Tauli 1945), as they concentrated primarily on older Indo-European languages.
57
This is one of the fundamental characteristics of R-B languages: a substantial amount of forms traditionally labelled as analytic. However, since preverbal inflection is disfavored crosslinguistically (perhaps due to perceptual issues), it is to be expected that right-branching forms will not fuse. Bauer, for instance, asserts that “‘synthetic’ parallels left-branching and ‘analytic’ parallels right-branching” (Bauer 2006: 301).
216
Separability (i.e. whether two items are separate or contiguous) also relies heavily on word order and, therefore, we cannot always expect to see a loss of separability as a result of synthesis.
However, in spite of relatively rigid
Romance word order, a trend toward increased contiguity can be identified. In coordinative subordinate clauses (e.g. I hope that he works hard and saves money), it is common to repeat the subordinating conjunction, as shown in the following examples: (5-34) OFr.
e
cumandad
que
léal
and
command-3SG-PRET-IND
that
fair-OBL
justise
tenissent . . .
e
que
justice-OBL
have-3PL-IMPERF-SUBJ
and
that
il
en doctrinassent
la
they-NOM
indoctrinate-3PL-IMPERF-SUBJ
the-OBL
gent people-OBL
‘and he commanded that they have fair justice . . . and indoctrinate the people’ (Quat. Liv. R. 3,340 in Ritchie 1907: 160) (5-35) OFr.
crent
que
Daires
think-3PL-PRES-IND
that
Daires-NOM
ait
lor
consence
E
que
have-3SG-PRES-SUBJ
their-OBL
consent-OBL
and
that
par
eus
iço
comence
by
them-OBL
DEM-PRON-NOM
begin-3SG-PRES-SUBJ
217
‘they think that Daires has their consent and that because of them it begins’ (Thèbes 8135 in Ritchie 1907: 160) While it is true that this repetition also occurs with verbs in the indicative, Ritchie (1907) uncovered an intriguing trend: the repetition of que is twice as common with subjunctive verbs. One might find this surprising, arguing that less repetition of que would be expected with subjunctive, since the subjunctive itself is capable of indicating subordination (Ritchie 1907: 160). From my perspective, however, this is not at all unexpected, as it is precisely the type of increase in “tightness” one would expect among elements undergoing synthesis.
5.6. Conclusion Traditionally, analyticity refers to phonological, morphological, and syntactic autonomy of morphemes, while syntheticity refers to the phonological, morphological and syntactic interdependency of morphemes. However, in Analyticity and Syntheticity (1990), Armin Schwegler argues that many constructions deemed to be analytic are actually quite cohesive, thus drawing our attention to the possible synthesis of non-contiguous elements. Although a lack of syntactic flexibility in Modern Romance languages often prevents surface-level restructuring, I have argued that semantic synthesis between que and the subjunctive has, indeed, ocurred. As a result of the grammaticalization process,
218
all three of Schwegler’s (1990) pre-conditions—relevance, frequency and generality—had already been met in early Romance. Furthermore, a gradual increase in obligatoriness as well as a reduction in separability can be identified in medieval data. Finally, it has been shown that these trends have continued to develop, as obligatoriness, in particular, is nearly categorical in Modern Spanish and French.
219
CHAPTER 6 DESTABILIZATION AND REFUNCTIONALIZATION OF ROMANCE MOOD 6.1. Introduction In the previous three chapters, I explored the interrelated morpho-syntactic developments which I believe to be at the core of changes in Romance mood distribution.
In this chapter, I will argue that the shift in branching, the
grammaticalization of subordinating elements, along with partial synthesis of que and the subjunctive did, in a cumulative fashion, render the inherited Proto-IndoEuropean indicative / subjunctive contrast ineffective. This destabilization or loss of motivation has subsequently resulted in several Romance phenomena aimed at re-motivating the system. The purpose of this chapter, therefore, is to bridge the gap between the morpho-syntactic changes analyzed in chapters three, four and five of this dissertation and the majority of existing research on the subjunctive.
I will
attempt to relate the major developments in Latin and Romance subordination to the reductive tendencies and mood variation often cited in the literature. First, I will assess each of the morpho-syntactic changes, presenting arguments for the ways in which they may have had debilitating effects on mood. Within this context, I will then analyze several well-attested changes in mood as signs of refunctionalization. My aim is to illustrate that what has occurred throughout the 220
history of Romance languages is not merely a random loss, or even a loss related to external interference or phonology, but rather a system-congruent response or repair mechanism necessitated by the substantial changes in subordination techniques discussed in the previous chapters of this dissertation.
6.2. Destabilizing Effects of Typological Drift Typological drift has, without a doubt, had a substantial impact on Romance mood. We have seen that an increase in right-branching structures can be linked to an increase in the frequency of conjunctions, thus facilitating the rapid grammaticalization of subordinating conjunctions, which has ultimately resulted in partial synthesis between que and the subjunctive. However, each of these developments, individually, has had substantial effects on mood. In this section, therefore, I will present the various ways in which these changes in morphosyntax have had an impact on mood distribution.
6.2.1. Shift to Right-Branching The significance of diachronic developments related to changes in branching tendencies is, in my view, severely underestimated. While the direct results are often explored in depth (e.g. the development of prepositions and auxiliary verbs), the broad range of far-reaching indirect results tends to be overlooked. With
221
regard to mood, there are two main ways in which an increase in right-branching structures has affected Romance developments. As illustrated in detail in chapter three, an increase in post-posed subordinate clauses entailed an increase in finite clauses with subordinating conjunctions. The significance of this development is that clauses which were typically expressed via non-finite—and thus, non-modal—constructions would, as a result of this shift to finite clauses, need to be assigned a mood. Below are several examples of infinitival constructions after verba declarandi, verba sentiendi, and verba affectuum: (6-1)
Lt.
hominem
catum
eum
man-ACC
cunning-ACC
DEM-PRON-ACC
esse
declaramus
be-PRES-INF
declare-1PL-PRES-IND
‘we declare that that man is cunning’ (Pl., Pseud. 682 in Bennett 1910: 368) (6-2)
Lt.
poetam
audivi
scripsisse
poet-ACC
hear-1SG-PERF-IND
write-PERF-INF
mulieres
duas
peiores
esse
women-ACC
two-ACC-F
worse-ACC
be-PRES-INF
quam
unam
than
one-ACC-F
‘I heard that the poet wrote that two women are worse than one’ (Pl., Curc. 591 in Bennett 1910: 371)
222
(6-3)
Lt.
sensi
filio
meo
te
perceive-1SG-PERF-IND
son-DAT
my-DAT
you-ACC
esse
amicum
be-PRES-INF
friend-ACC
‘I thought you were a friend to my son’ (Pl., Capt. 140 in Bennett 1910: 376) (6-4)
Lt.
crucior
lapidem
non
torment-1SG-PASS-PRES-IND
stone-ACC
NEG
habere
me
have-PRES-INF
me-ACC
‘I am tormented that I don’t have a stone’ (Pl., Capt. 600 in Bennett 1910: 377) In cases such as these, there has not been a shift in mood between Latin and modern Romance languages, but rather a shift from non-finite to finite constructions. Speakers, when faced with the decision, may have (1) chosen the mood most closely associated with subordinate clauses (i.e. subjunctive), (2) selected mood based on the conjunction being used, or (3) assigned a mood on semantic grounds alone, thereby attempting to fit these uses into the existing modal values. Consider the following Latin examples, in which the subjunctive is selected following the most common complementizers of the period (i.e. quia and quod): (6-5) Lt.
scio
enim,
quia
valde
know-1SG-PRES-IND
for
that
intensely
223
me
bene
ames
me-ACC
very
love-2SG-PRES-SUBJ
‘for I know that you love me very intensely’ (Scaev., Dig. 44,7,61,1 in Raible 1992: 318) (6-6)
Lt.
scire
itaque
nos
principe
in
know-PRES-INF
therefore
us-ACC
first-ABL
in
loco
oportet . . .
quod
place-ABL
be proper-3SG-PRES-IND
that
ad
imaginem
speciemque
towards
image-ACC
shape-ACC—and
mundi
formam
hominis . . .
world-GEN
form-ACC
man-GEN
Deus
ille . . .
perfecerit
God-NOM
DEM-PRON-NOM
execute-3SG-PERF-SUBJ
‘therefore, one ought to know in the first place that God created the form of man according to the shape and image of the world’ (Firm., Math. 3. proem. 1 in Raible 1992: 318) In other words, mood in these cases is not inherited. Instead, these examples represent a response to the need to integrate new finite subordinate clauses into an existing verbal system. This development, resulting in unexpected—and perhaps somewhat arbitrary—contexts for the subjunctive may have led to confusion regarding mood distributional patterns, thus serving as an impetus for reorganization. Furthermore, it is not surprising that we find variation among
224
Romance languages in these clause-types,58 which represent the existence of multiple solutions to the same “problem.” The second correlate to a shift toward right-branching relates to innovative Romance paradigms. The development of right-branching future and conditional forms (i.e. Sp. voy a cantar / Fr. je vais chanter and Sp. iba a cantar / Fr. j’allais chanter), has resulted in the following indicative systems in Spanish and French: Table 26: Modern Spanish Indicative System SPANISH INDICATIVE SYSTEM
non-past axis
past axis
anterior
simultaneous
posterior
he cantado
canto
voy a cantar (cantaré)
había cantado
cantaba /canté
iba a cantar (cantaría)
Table 27: Modern French Indicative System FRENCH INDICATIVE SYSTEM
non-past axis
past axis
anterior
simultaneous
posterior
j’ai chanté
je chante
je vais chanter (je chanterai)
j’avais chanté
je chantais
j’allais chanté (je chanterais)
Cross-linguistically, imperfect past, future and conditional paradigms commonly have modal nuances (cf. Fleischman 1982 and Palmer 1986: chapter 6). However, the co-existence of two rival forms (one right-branching and one left-
58
Consider Italian, for example, where penso che (‘I think that’) is followed by the subjunctive.
225
branching) can be said to have facilitated a sharp increase in modal value among the older left-branching future and conditional.59
While the newer right-
branching forms in modern Romance languages tend to be purely temporal, their older left-branching equivalents are able to gain modal value. As a result, these originally indicative forms have come to usurp several previous functions of the subjunctive. Beginning with the future tense, the older, left-branching form (< Lt. cantare habeo) dates back to Classical Latin (although not widespread until Late Latin), while the newer, right-branching form acquired temporal value in the 13th century and was generalized in the 15th century (cf. Fleischman 1982: 52 & 82). It is during this period that there is an evident increase in modality of the older form, as shown below in an example from the late 16th century: (6-7)
Sp.
esta
noche
os
pondreys
this
night
2PL-REFL-PRON
put-2PL-FUT-IND
debaxo do
les
balcones
desta
sala
under
the
balconies
of—this
room
‘tonight you’ll wait under the balconies of this room’ (Fleischman 1982: 129) Below are additional examples of this increased modality among the leftbranching future forms: 59
This is not surprising, when one considers economical tendencies in language. In the presence of two synonomous forms, it is common for one to disappear or for a “split” to occur, whereby the forms are clearly distinguished with regard to function. In this way, speakers are able to maximize the expressivity of linguistic tools available to them.
226
(6-8)
Fr.
Vous
éteindrez
en
sortant
you-PL-NOM
turn off-2PL-FUT-IND
upon
leave-GER
‘turn off the light as you go out’ (6-9)
Sp.
¿Serán
las cinco,
no?
be-3PL-FUT-IND
five o’clock
NEG
‘It must be five o’clock, don’t you think? (6-10) Fr.
Georges
n’est
pas
venu
George
NEG—be-3SG-PRES-IND
NEG
come-PP
ce
matin.
Il
aura
this
morning
he-NOM
have-3SG-FUT-IND
oublié
notre
rendez-vous.
forget-PP
our
appointment
‘George didn’t come this morning. He probably forgot our appointment’ (6-11) Sp.
¿Que
lo
daré
yo?
that
it-ACC
give-1SG-FUT-IND
I-NOM
‘What! I’m the one who should give it?!’ (6-12) Fr.
Quoi?
J’accepterai,
moi,
what
I-NOM—accept-1SG-FUT-IND
I-NOM
qu’on
me
remplace?
that—one
me-ACC
replace-3SG-PRES-SUBJ
‘What? I should agree to be replaced?’ (Fleischman 1982: 129-133) Examples (6-7) and (6-8) illustrate the volitional future, (6-9) and (6-10) are examples of the future of likelihood, and the future of indignation is shown in (6-11) and (6-12). Of further interest is the fact that in certain dialects, this trend
227
has been developed even further. In the Andean region of Colombia and Ecuador, for example, the left-branching future has actually become part of the imperative paradigm, including a post-verbal clitic position.
Consider the following
examples: (6-13) Sp.
darámelos give-3SG-FUT-IND—me-DAT—them-ACC
‘I’m asking you to give them to me’ (6-14) Sp.
démelos give-3SG-IMP—me-DAT—them-ACC
‘Give them to me!’ (6-15) Sp.
apuraráse hurry-3SG-FUT-IND—3SG-REFL-PRON
‘I’m asking you to hurry’ (6-16) Sp.
apúrese hurry-3SG-IMP—3SG-REFL-PRON
‘Hurry!’ (Niño-Murcia 1992: 707) The forms shown in (6-13) and (6-15) are used to make requests or ask favors, while the corresponding standard imperative forms in (6-14) and (6-16) are considered rude and offensive. In fact, of all the available alternatives, the leftbranching future is used to make requests in Quito, Ecuador in at least sixty percent of all cases (Niño-Murcia 1992: 708). Although this phenomenon is typically considered to be a result of extensive language contact with Quechua, it
228
is important to note that it is also consistent with modal usage attested in Old Spanish. Another interesting phenomenon occurs in the Leonés dialect in Spain, where the left-branching future is now restricted to subordinate clauses (Bybee et al. 1994: 235). In summary, the extension of modal value of this future tense has resulted in several instances of displacement of the subjunctive mood. Another verbal paradigm that commonly has modal value is the conditional. This left-branching form (< Lt. cantare habebam) also dates back to Classical Latin, gaining popularity throughout Late Latin. Issues surrounding the development of Lt. cantare habebam are highly debated, both in terms of the form’s relationship to the future paradigm and in terms of its primary function. The most widely-held view is that the original value of Lt. cantare habebam was temporal, developing in order to accommodate the need for a finite expression of future-in-the-past in indirect speech (Harris 1981: 57).
However, although
Fleischman argues that hypothetical modality is a natural “outgrowth” (1982: 65) of future-in-the-past, she also points out that “investigators are not unanimous in viewing the modal (i.e. conditional) function of cantare habebam as a secondary development” (Fleischman 1982: 60). It appears that, from the outset, the form had not only temporal value (i.e. future-in-the-past) but modal value as well, as it was used in apodoses of hypothetical sentences, taking over one of the more common uses of the
229
subjunctive. Consider the following example from St. Augustine, often cited as the first attestation of this usage: (6-17) Lt.
sanare
te
habebat
cure-PRES-INF
you-ACC
have-3SG-IMPERF-IND
Deus . . .
si
fatereris
God-NOM
if
confess-2SG-IMPERF-SUBJ
‘God would cure you if you confessed’ (Harris 1971: 28) This function (in apodoses of hypothetical sentences) has become the norm in Modern Spanish and French, as shown below: (6-18) Sp.
(6-19) Fr.
si
pudiera,
iría
a
if
be able-1SG-IMPERF-SUBJ
go-1SG-COND
to
la
playa
the-SG
beach
si
je
pouvais,
if
I-NOM
be able-1SG-IMPERF-IND
j’irais
a
la
plage
I-NOM—go-1SG-COND
to
the-SG
beach
‘if
I could, I’d go to the beach’
In spite of the fact that this paradigm appears to have had modal value from the outset, the competition, as it were, between the two forms (Sp. sería / Fr. serait and Sp. iba a ser / Fr. allait être) facilitated the increase in modality on the part of the left-branching form, resulting in a functional split. This substantial increase in
230
modality allowed the conditional to usurp many functions of the subjunctive, such as expressing politeness, alleged information, as well as attenuation. In addition to these standard uses for the conditional in Romance languages, there are also several non-standard uses which reveal the way in which this paradigm continues to take over previous subjunctive roles. Consider the following examples: (6-20) Fr.
si
mon
maître
serait
là,
je
if
my
teacher
be-3SG-COND
there
I-NOM
l’appellerais him-ACC—call-1SG-COND
‘if my teacher were there, I’d call him’ (Lanly 1973: 398) (6-21) Fr.
je
cherche
une
maison
I-NOM
look for-1SG-PRES-IND
a
house
qui
aurait
un
jardin
that-NOM
have-3SG-COND
a
garden
‘I’m looking for a house that has a garden’ (Harris 1978b: 182) (6-22) Sp.
si
me
lo
pedirías,
me
if
me-DAT
it-ACC
ask-2SG-COND
me-REFL-PRON
casaría
contigo
marry-1SG-COND
with you
‘if you asked me, I’d marry you’ (Lavandera 1979: 123)
231
(6-23) Sp.
fue
posible
que
vendría60
be-3SG-PRET-IND
possible
that
come-3SG-COND
‘it was possible for him to come’ (Ridruejo 1975: 125) The way in which the newer right-branching (and purely temporal) forms have enabled the older left-branching forms to serve as exponents of modality is quite dramatic and, according to Harris, “one is forced to see a new mood emerging” (1981: 64) since “these paradigms and their compound equivalents may eventually cease altogether to belong to the indicative mood, being reserved entirely to modal usages” (Harris 1978b: 183).61 There is no doubt, therefore, that a shift toward right-branching tendencies in Romance languages has had substantial effects on mood usage. First of all, several widespread non-finite constructions, upon being replaced by finite constructions, had to be integrated into the existing system.
Secondly, the
emergence of new right-branching alternatives to both future-in-the-present and future-in-the-past increased the modal value of the older forms, which have ultimately usurped several important subjunctive roles. 60
According to Ridruejo (1975), the conditional is used in virtually all past subjunctive contexts (e.g. after expressions of will and doubt, in temporal, consecutive and causal clauses, etc.) in La Rioja, Spain. Furthermore, this phenomenon cannot be attributed merely to an influence from Basque, since similar patterns have been attested in Latin America (1975: 123, 126-127). 61 It is necessary to make a distinction between the left-branching future and conditional, and the left-branching aorist (i.e. Sp. pretérito / Fr. passé simple). While the simple past tense also has a right-branching rival (i.e. present perfect), and a similar functional split could thus potentially occur, this temporal form is simply not predisposed to modality. It is specifically the imperfect past that is characterized by this additional overlay of meaning across languages (c.f. Fleischman 1982, Bybee et al. 1994 and Palmer 1986: chapter 6).
232
6.2.2. Grammaticalization of Conjunctions The second development explored in this dissertation (i.e. the grammaticalization of subordinating conjunctions) has also had a destabilizing effect on mood usage in Romance languages.
Although mood selection in Latin was undoubtedly
based—at least in part—on semantic categories, there is evidence that the conjunctions came to play an important role as well. Several conjunctions appear to have had strong ties to a particular mood. Quando and ubi, for instance, were primarily associated with the indicative, while cum, ut and quin had strong ties to the subjunctive. To illustrate this trend, I have counted all instances of cum, ut, ne, quin, si, quod and ubi in Book I of Caesar’s De Bello Gallico.62 The results are as follows: Table 28: Latin Conjunctions and Mood in De Bello Gallico conjunction subjunctive cum ut ne quin si quod ubi
44 82 19 5 40 41 0
indicative 7 10 0 0 2 25 7
Furthermore, it does not seem convincing to argue that mood usage is motivated exclusively on semantic grounds and that the conjunctions merely reflect those semantic categories. One can easily find instances where mood appears to be
62
Since the focus here is on the forms as conjunctions, I have excluded all instances of quod as a relative pronoun.
233
influenced exclusively by the presence of a particular conjunction. Consider the following contrastive examples including temporal conjunctions meaning ‘when’: (6-24) Lt.
ubi
de
eius
adventu
Helvetii
when
of
his-GEN
arrival-ABL
Helvetii-NOM
certiores
facti sunt,
more certain-NOM
make-3PL-PERF-IND-PASS
legatos
ad
eum
mittunt
ambassadors-ACC
to
him-ACC
send-3PL-PRES-IND
‘when the Helvetii were informed of his arrival, they sent ambassadors to him’ (Caes., D.B.G. 1.7) (6-25) Lt.
Pharao,
quando
vidit,
quod
Pharaoh
when
see-3S-PERF-IND
that
filii
Israhel
dimiserant
children-NOM
Israel-GEN
abandon-3PL-PLUPERF-IND
eum . . . him-ACC
‘when Pharaoh saw that the children of Israel had escaped him . . .’ (Pereg. 8.5) (6-26) Lt.
cum
tridui
viam
when
three days-GEN
march-ACC
processisset,
nuntiatum est
proceed-3SG-PLUPERF-SUBJ
announce-3SG-PERF-IND-PASS
ei . . . to him
234
‘when (Caesar) had advanced a three-days’ journey, it was reported to him . . .’ (Caes., D.B.G. 1.38) The three sentences above are surprisingly similar, in terms of both semantic and pragmatic value. All three cases refer to factual events that have already taken place. There is no doubt or subsequence implied, nor does there appear to be any difference in terms of informative value or hesitation on the part of the speaker. In short, there is little to justify the difference in mood selection in these examples except for the presence of cum in example (6-26).
This seemingly
complementary distribution leads me to believe that several Classical Latin conjunctions had indeed acquired substantial modal value.
Furthermore, this
modal value was not inherited, but rather, developed slowly over time.. The claim that certain conjunctions became increasingly associated with a particular mood is supported by evidence from Early Latin, where cum, for example, did not automatically govern subjunctive in causal clauses, as it did during the Classical period (cf. Bennett 1910: 133-135). Meillet and Vendryes note: “en effet, dans la langue classique, le subjonctif tend à s’employer de plus en plus après quom, d’abord quand il s’agit de marquer l’enchaînement des événements, . . . puis d’une façon générale” (1953: 669-670). For this reason, I believe subordinating conjunctions came to play a vital role in mood selection in Classical Latin. Without going so far as to say they
235
determined mood in a purely mechanistic fashion, one must acknowledge the connection between certain conjunctions and the mood of the subordinate finite verb. It is important to note that this development pertains only to a small group of conjunctions. If all conjunctions had been categorically associated with a particular mood, then mood selection would have become entirely automatic. What we do see, however, is a sufficient amount of automaticity to have contributed to mood destabilization. Furthermore, since the grammaticalization process (discussed in chapter four) substantially reduced this system, many of the conjunctions that had come to be seen as exponents of modality were ultimately eliminated from the inventory.
6.2.3. Synthesis of que and Subjunctive As presented in chapter five, the sharp increase in frequency of the grammaticalized subordinator que led to a partial synthesis between que and the subjunctive. I believe this synthesis also contributed to the weakening of the indicative/subjunctive contrast in Romance languages, primarily due to the increase in obligatoriness. First of all, the extension of subjunctive to previously non-subjunctive contexts must surely have disturbed the system by making the motivation behind the mood increasingly opaque.
Speakers of Spanish and
French would have had to make sense of this subjunctive usage, re-assessing the criteria used for mood selection. 236
Secondly, the extended co-existence of que and the subjunctive, both in subordinate and non-subordinate contexts, may have made the sequence que + subjunctive seem redundant, thus facilitating the loss of one element. Consider the following illustration of the development of que and the subjunctive: Stage 1:
verb
+
SUBJ morphology
Stage 2:
CONJ +
verb
+
SUBJ morphology
Stage 3:
SUB
+
verb
+
SUBJ morphology
Stage 4:
SUB
+
verb
The development from stage 1 to stage 2 represents the shift from an archaic subordination technique where the subjunctive stands alone to a subordination technique with both a conjunction and subjunctive morphology (typically in a post-posed subordinate clause). The transition from stage 2 to stage 3 represents the grammaticalization process, whereby que—via the process of specialization— surfaces as a generalized subordinator. Finally, the shift from stage 3 to stage 4 represents the potential loss of subjunctive, which may be motivated by redundancy perceived by speakers.63
63
Despite the strong parallels between this development and the development of negation in French (i.e. ne…pas), there is one major difference. The grammaticalization of pas results in two forms with one identical function: negation. This absolute coreferentiality facilitates synthesis. The grammaticalization of que, on the other hand, does not produce two elements with identical functions, but rather, two elements which exhibit a certain degree of semantic and functional overlap. As a result, the subjunctive may potentially be deemed redundant and begin to lose ground, although not inevitably so.
237
To conclude this section, this dissertation has identified four major causes behind the loss of motivation for mood distinctions in Latin: 1)
the replacement of non-finite (and thus, non-modal) constructions by finite subordinate clauses (beginning in Late Latin and continuing through the medieval period)
2)
the increase in modal value of older left-branching temporal paradigms (beginning in Late Latin and still ongoing)
3)
the loss of many modally charged conjunctions (throughout Late Latin and early Romance)
4)
the expansion of subjunctive usage due to semantic synthesis and/or formal analogy (since the medieval period)
In my view, these four developments—all rooted in the shift from left to rightbranching structures in Romance languages—resulted in a loss of motivation and contributed to the destabilization of the indicative/subjunctive mood contrast. These factors, together, help to explain why mood in Latin became unstable. The way in which the modal system has been reorganized, thus compensating for this instability, will be the focus of the following section.
6.3. Refunctionalization of Romance Mood Contrast The destabilization of Romance mood distinctions triggered by typological drift did not lead to a complete loss of the subjunctive, and for obvious reasons. The 238
morphological expression of modality was characteristic of Romance languages and, in early Romance, there was still a viable paradigm via which to express it. Therefore, the radical elimination of the subjunctive would have been unlikely.64 The destabilization did, however, lead to the reorganization of mood, a process which does not appear to have ocurred in a random fashion. Instead, it has revealed the existence of a basic or prototypical subjunctive value, where variation has always been—and continues to be—minimal. Furthermore, the categories exhibiting the greatest amount of variation do so for justifiable reasons based on the speakers’ attempt to reorganize a cumbersome system in order to make contrasts more apparent and meaningful. The discussion that follows will, accordingly, be divided into two sub-sections, one dealing with the stable or prototypical subjunctive context and the other dealing with five of the most unstable contexts.
6.3.1. Stable (Prototypical) Subjunctive The claim that the subjunctive has a basic or prototypical meaning (i.e. volition) is not a controversial one, as it can be supported on both synchronic and diachronic grounds. Expressions of volition generally exhibit the least amount of synchronic variation, with the subjunctive consistently chosen in these clauses across 64
This is not to say that morphological mood cannot disappear completely, as we do see evidence for this in English and, to a lesser extent, in French. A complete loss is possible over the course of a very long period of time during which paradigmatic distinctions are lost and alternative expressions of modality become systematic.
239
Romance languages and dialects. Furthermore, the category of volition exhibits the least amount of diachronic variation, as it “represents the strongest domain of the subjunctive mood, and practically no changes have occurred here” (Jensen 1974: 121).
In terms of inheritance, the original modal force of the Indo-
European subjunctive was will and/or futurity,65 and other uses can be seen as extensions of this most basic one (cf. Bennett 1910: 145-152). In Latin, volition was often expressed via non-finite forms (i.e. the Accusative-Infinitive), as shown below: (6-27) Lt.
impera
suovetaurilia
order-2SG-IMP
suovetaurilia-ACC
circumagi lead around-PASS-PRES-INF
‘order the suovetaurilia to be led around’ (Cato, R.R. 141.1) (6-28) Lt.
argentum
iubebo
iam
intus
silver-ACC
order-1SG-FUT-IND
now
inside
ecferri
foras
bring-PASS-PRES-INF
outside
‘I’ll order the silver that’s now inside to be brought outside’ (Pl., Bacch. 95 in Bennett 1910: 380)
65
The notion of futurity is not to be confused with a future paradigm (which did not exist in PIE). Intead, it refers to possible or anticipated events. To elaborate, Meillet and Vendryes explain that the Indo-European subjunctive (along with the optative) expresses “une pensée tournée vers l’avenir” (Meillet and Vendryes 1953: 192).
240
Finite constructions also existed, and when this alternative was chosen, the subjunctive was used categorically, preceded by the conjunction ut (or ut non/ne in the negative): (6-29) Lt.
rogare
ut
eius
voluntate
beg-PRES-INF
that
with his-GEN
will-ABL
id
sibi
facere
this-ACC
REFL-PRON-DAT
do-PRES-INF
liceat be allowed-3SG-PRES-SUBJ
‘they requested that they be allowed to do so with his permission’ (Caes., D.B.G. 1.7) (6-30) Lt.
Allobrogibus
imperavit,
ut
Allobroges-DAT
order-3SG-PERF-IND
that
eis
frumenti
copiam
for them-DAT
grain-GEN
abundance-ACC
facerent make-3PL-IMPERF-SUBJ
‘he commanded the Allobroges to supply them with a sufficient amount of grain’ (Caes., D.B.G. 1.28) In the medieval period, finite right-branching subordination techniques became the norm (thus ousting the AcI), and the subjunctive mood maintained its stronghold: (6-31) OSp. mis my
fasañas
ruégote
que
deeds
beg-1SG-PRES-IND—you-DAT
that
241
bien
las
mires
well
them-ACC
look-2SG-PRES-SUBJ
‘I beg you to look well at my deeds’ (Hita, L.B.A. 908d in Jensen and Lathrop 1973: 43) (6-32) OFr.
ço
ne
vuelt
il
que
DEM-PRON
NEG
want-3SG-PRES-IND
he-NOM
that
sa
medre
le
sachet
his-NOM
mother-NOM
it-ACC
know-3S-PRES-SUBJ
‘he does not want his mother to know’ (Alex., 249 in Jensen 1974: 33) In Modern Spanish and French, the subjunctive is still found in this context, with variation being extremely minimal: (6-33) Sp.
deseo
que
estudies
want-1SG-PRES-IND
that
study-2SG-PRES-SUBJ
el
derecho
DEF-ART-SG
law
‘I want you to study law’ (Bello 1988 [1847]: 362) (6-34) Fr.
Jean
a
besoin
que
John
have-3SG-PRES-IND
need
that
nous
l’aidions
we-NOM
him-ACC—help-1PL-PRES-SUBJ
‘John needs us to help him’ (Grevisse 1993: 1611)
242
The only main difference regarding modern usage is that the grammaticalized subordinator que has replaced Latin ut. In all other respects, the examples above in (6-33) and (6-34) are unquestionably similar to the Latin examples in (6-29) and (6-30). In short, aside from a moderate increase in finite clauses (at the expense of the Latin AcI construction), the usage in this category has been extremely stable. One could easily argue that this is the fundamental meaning behind the subjunctive mood, both in terms of inheritance and modern-day speaker perceptions.
Most other contexts, on the other hand, have not
experienced this amount of stability, as will be shown below.
6.3.2. Unstable Subjunctive Unlike the volitive subjunctive, most other subjunctive categories have undergone changes (some minor and some substantial) between Latin and modern Romance periods. I have selected five of the most variable contexts, whose development I will trace in an attempt to demonstrate the way in which they became unstable and how they were consequently reanalyzed, thus remotivating the system. The five contexts in question are: (1) with verbs of saying and verbs of opinion, (2) in result clauses, (3) in causal clauses, (4) in adversative clauses, and (5) with expressions of emotion.
243
6.3.2.1. With Verbs of Saying and Verbs of Opinion The first category that has undergone a substantial amount of change is subordinate clauses accompanying verbs of saying and verbs of opinion in the main clause. The unstable nature of this category relates to the fundamental shift In Latin, verba
from left-branching to right-branching subordinate clauses.
declarandi (e.g. dico ‘say’, affirmo ‘affirm’, narro ‘tell’) and verba sentiendi (e.g. puto ‘think’, judico ‘judge’, scio ‘know’) were accompanied by AccusativeInfinitive constructions, as shown below: (6-35) Lt.
eorum
una
pars,
quam
these-GEN
one-NOM
part-NOM
which-ACC
Gallos
obtinere
dictum
Gauls-ACC
hold-PRES-INF
say-PERF-PART-N-SG
est . . . be-3SG-PRES-IND
‘one part of this country, which we have said the Gauls occupy’ (Caes., D.B.G. 1.1) (6-36) Lt.
satis
esse
causae
arbitrabatur
enough-ACC be-PRES-INF cause-GEN think-3SG-IMPERF-IND
‘he thought there was sufficient reason . . .’ (Caes., D.B.G. 1.19) In Late Latin, however, there was a gradual shift toward finite forms and the subjunctive appears to have been quite widespread in these subordinate clauses. Consider the following example from Gregory of Tours: 244
(6-37) Lt.
dicunt
alii,
quod
ad
ipsam
say-3PL-PRES-IND
some-NOM
that
to
this-ACC
ripam . . .
sint
reversi
shore-ACC
be-3PL-PRES-SUBJ
return-PERF-PART-NOM
‘some say that they (the Israelites) retuned to this shore’ (GdT, H.F. 1.10) It is not entirely surprising that speakers would have chosen the subjunctive mood, as it shares with the infinitive its role as a marker of subordination. However, this mechanistic approach to mood selection would not have been motivated on semantic grounds. Therefore, as a repair mechanism attempting to re-motivate mood selection, a strong contrast was established between indicative and subjunctive in the medieval period, with the subjunctive being used for instances of nonassertion (specifically, volition and doubt), and indicative for assertion.
In
examples (6-38) and (6-39) below, the subordinate clauses are assertive and, therefore, the indicative is chosen: (6-38) OSp. byen well
creo
que
de
ángeles
believe-1SG-PRES-IND
that
of
angels
fué
tal
cosa
enbiada
be-3SG-PRET-IND
such
thing
send-PP
‘I truly believe that such a thing was sent from angels’ (Hita, L.B.A. 1265c in Jensen and Lathrop 1973: 59) (6-39) OFr.
il
croit
que
el
he-NOM
believe-3SG-PRES-IND
that
she-NOM
245
est
ainsi
feble
be-3SG-PRES-IND
thus
weak-NOM
‘he believes that she is that weak’ (Quinze 64 in Jensen 1974: 56) However, when there is volition involved, the subjunctive is the norm, as shown in the following examples: (6-40) OSp. dezidles say-2PL-IMP—them-DAT
el
rastro
the-SG
trail
que
prendan
that
take-3PL-PRES-SUBJ
‘tell them to pick up our trail’ (Cid 389 in Jensen and Lathrop 1973: 46) (6-41) OFr.
l’amirail
dites
que
the—Admiral-OBL
say-2PL-IMP
that
s’ost
i
seit
his—army-NOM
LOC-PRON
be-3SG-PRES-SUBJ
menede lead-PP-NOM
‘tell the Admiral to lead his army there’ (Rol. 2760 in Jensen 1974: 52) In the above examples, the volitive value is apparent, since the actions in the subordinate clauses are desired actions that have not yet occurred. Similarly, subjunctive mood is also chosen when there is an element of doubt, as shown below:
246
(6-42) OSp. no NEG
paresçe
que
ayas
seem-3SG-PRES-IND
that
have-2SG-PRES-SUBJ
quinze
años
fifteeen
years
‘it doesn’t seem like you are fifteen years old’ (Celest. 141,10 in Jensen and Lathrop 1973: 60) (6-43) OFr.
car for
il
n’est
pas
it-NOM
NEG—be-3SG-PRES-IND
NEG
acertené
qu’ils
aient
ascertain-PP-NOM
that—they-NOM
have-3PL-PRES-SUBJ
rien
fait
de
mal
nothing-OBL
do-PP-OBL
of
bad-OBL
‘it’s not certain that they didn’t do anything wrong’ (Quinze 171 in Jensen 1974: 61) These developments can be seen as a refunctionalization of mood paradigms available to the speaker. Subordinate clauses in this particular context (i.e. with verbs of saying and verbs of opinion) were reorganized in a way that would be both system congruent and semantically motivated. During the medieval period, the distribution of indicative and subjunctive moods had become reliably determined by the dichotomies of volition/non-volition and certainty/uncertainty.
6.3.2.2. In Result Clauses Clauses of result—often referred to as consecutive clauses—were extremely common in Latin. In spite of the fact that they involve a wide range of semantic
247
nuances, all were expressed via the same construction: the conjunction ut (or ut non / quin in the negative) followed by the subjunctive mood. Consider the following examples from Caesar: (6-44) Lt.
a
Sequanis
impetrat
ut
per
from
Sequani-ABL
obtain-3SG-PRES-IND
that
through
fines
suos
Helvetios
ire
borders-ACC
their-ACC
Helvetii-ACC
go-PRES-INF
patiantur allow-3PL-PRES-SUBJ
‘they got the Sequani to allow the Helvetii to go through their country’ (Caes., D.B.G. 1.9) (6-45) Lt.
confecerant,
ut
flumen
accomplish-3PL-PLUPERF-IND
that
river-ACC
transirent cross-3PL-IMPERF-SUBJ
‘they had accomplished crossing the river’ (Caes., D.B.G. 1.13) (6-46) Lt.
fiebat,
ut
et
minus
late
happen-3SG-IMPERF-IND
that
and
less
widely
vagarentur
et
minus
facile
roam-3PL-IMPERF-SUBJ
and
less
easily
finitimis
bellum
inferre
neighbors-DAT
war-ACC
bring-PRES-INF
possent be able-3PL-IMPERF-SUBJ
248
‘it came to pass that they wandered less widely and could less easily bring war upon the neighboring clans’ (Caes., D.B.G. 1.2) While the notion of intention or goal may arguably be present in examples (6-44) and (6-45), in (6-46) there is no discernable expression of volition, but merely fact. For this reason, the categorical use of subjunctive in this context may have been seen as unmotivated, and in Old Spanish and Old French, a new distinction emerged, with subjunctive reserved for modal values such as volition. Consider the following sentences in which the verbs in the result clauses appear in the indicative: (6-47) OSp. mi my
tribulaçion
me
tiene
tan
tribulation
me-ACC
have-3SG-PRES-IND
so
penado
Que
con
amargura
aquesto
pained
that
with
bitterness
this
he
fablado
have-1SG-PRES-IND
speak-PP
‘my tribulation has me so pained, that I have told this with bitterness’ (Ayala, Pal. 931c in Jensen and Lathrop 1973: 73) (6-48) OFr.
elle
fait
tout
si
she-NOM
do-3SG-PRES-IND
everything-OBL
so
gracieusement,
Que
nul
n’y
graciously
that
no one-NOM
NEG—LOC-PRON
scet
trouver
amendement
know-3SG-PRES-IND
find-INF
amendment-OBL
249
‘she does everything so graciously that no one knows where to find amendment’ (Ch.O., Poésies 26,13 in Jensen 1974: 85) Unlike the above examples which are purely assertive/factive, the following examples express volition or intention on the part of the subject: (6-49) OSp. fablad speak-2PL-IMP
tanto
é
tal
cosa,
que
so much
and
such
thing
that
non
vos
arrepintades
NEG
2PL-REFL-PRON
repent-2PL-PRES-SUBJ
‘speak in such a way that you don’t regret it’ (Hita, L.B.A. 721b in Jensen and Lathrop 1973: 74) (6-50) OFr.
feites
ma
main
si
make-2PL-IMP
my-OBL
hand-OBL
so
fort,
Qu’a un
seul
cop
strong-OBL
that—at—one
single-OBL
blow-OBL
reçoive
mort
receive-3SG-PRES-SUBJ
death-OBL
‘make my hand so strong that he dies with a single blow’ (Piramus 869 in Jensen 1974: 85) The examples above convey clear notions of intention or goal and, therefore, motivate the subjunctive. I would like to point out, however, that although they are traditionally categorized as result clauses in Latin, these instances are in fact quite similar to purpose clauses (i.e. ‘so that’), which have required subjunctive mood consistently throughout their history.
250
Again, there has been a destabilization in Latin which has, subsequently, been resolved in the daughter languages. In this case, although the subjunctive was originally an extension of the potential subjunctive (cf. Bennett 1910: 296), it was no longer functioning in this way and had become a mechanistic response triggered categorically by the conjunction ut. The solution found in both Spanish and French is in keeping with an attempt to restore the semantic motivation behind the subjunctive mood. This distinction continues to exist today and, as a result, what was originally a single group (i.e. result clauses) is now further categorized so as to ensure semantic motivation behind the process of mood selection.
6.3.2.3. In Causal Clauses The causal clause is another circumstantial clause of interest in relation to mood selection. In Latin, there were several subordinating conjunctions that could be used to express cause (primarily quod, quia, quoniam, quando and cum). In terms of motivation for mood choice, the main criteria used was that of Oratio Recta (direct discourse), which called for the indicative, versus Oratio Obliqua (indirect discourse), which called for the subjunctive. However, while the conjunctions quod, quia and quoniam did follow this trend, quando was used exclusively with the indicative and cum only with the subjunctive (Gildersleeve and Lodge 1997: 338-339, 370-371; Leiper 1913: 16). This, in my view, is one of the destabilizing 251
factors regarding mood selection in causal clauses. Since mood was, in part, determined by the subordinating conjunction, the subsequent loss of these conjunctions may have created problems. Compare the following four Latin examples, in which only cum is followed by the subjunctive: (6-51) Lt.
quod
Gallia
sub
septentrionibus…
since
Gaul-NOM
under
the north-ABL
posita est…
ne
pabuli
quidem
be situated-3SG-PRES-IND
NEG
fodder-GEN
even
satis
magna
copia
sufficiently
great-NOM
abundance-NOM
suppetebat be present-3SG-IMPERF-IND ‘since Gaul
lies toward the north, not even a sufficient quantity of food was at hand’ (Caes., D.B.G. 1.16) (6-52) Lt.
quia
non
quaesisti
regna
because
NEG
ask-2SG-PERF-IND
kingdoms-ACC
mundi
nec
divitias
eius,
world-GEN
nor
wealth-ACC
DEM-PRON-GEN-SG
sed
postolasti
sapientiam,
ideo
but
ask-2SG-PERF-IND
wisdom-ACC
therefore
accepias
eam
accept-2SG-PRES-SUBJ
this-ACC
‘since you didn’t ask for the kingdoms of the world or for yourself, but you asked for wisdom, take then this thing’ (GdT, H.F. 1.13)
252
(6-53) Lt.
et
quoniam
de
hoc
maris
and
since
about
this-ABL
sea-GEN
transitu
plures
multa
dixerunt…
transit-ABL
many-NOM
much-ACC
say-3PL-PERF-IND
‘and since so much has been written by so many (authors) about this passage on the (Red) Sea…’ (GdT, H.F. 1.10) (6-54) Lt.
cum
ipse
eum
locum
because
himself-NOM
that-ACC
place-ACC
amicitiae
apud
eum
teneret,
friendship-GEN
with
him-ACC
hold-3SG-IMPERF-SUBJ
neminem
existimaturum
non
sua
no one-ACC
think-FUT-INF
NEG
his-ABL
voluntate
factum
will-ABL
be done-FUT-INF
‘since he (Divitiacus) was Caesar’s intimate friend, no one would think it (the punishment against Dumnorix) was done without his consent’ (Caes., D.B.G. 1.20) There does not seem to be any semantic or pragmatic justification for the mood contrast in the above examples. The subjunctive in (6-54) appears to be triggered by the conjunction cum. In Old Spanish and Old French, the indicative was used in factive predicates such as these: (6-55) OSp. e and
pesauales
mucho
weigh-3SG-IMPERF-IND—them-DAT
much
253
por que
tanto
tardauan
los
because
so much
be late-3PL-IMPERF-IND
those
de
la
hueste
from
the
army
‘and it weighed heavily on their minds because those from the army were so late’ (Troyana 7,22 in Jensen and Lathrop 1973: 69) (6-56) OFr.
et
molt
de
gent
le
and
many-NOM
of
people-OBL
him-ACC
sivoient,
pour chou qu’il
follow-3PL-IMPERF-IND
because—he-NOM
estoit
si
preudons
be-3SG-IMPERF-IND
so
honest-NOM
‘and many people followed him because he was so honest’ (Conq. 1,12 in Jensen 1974: 79) However, causal clauses followed by OSp. como and OFr. comme are found mainly in the subjunctive, following in the tradition of Latin cum. Jensen points out that this conjunction was not only a direct imitation of Latin cum, but is also found predominantly in the writing of learned writers such as Mondeville, Christine de Pisan and Alain Chartier (Jensen 1974: 79). Consider the examples below: (6-57) OSp. y
como
sea
natural
a
estos
and
since
be-3SG-PRES-SUBJ
natural
to
these
no
fazer
lo
que
prometen, . . .
NEG
do-INF
it-ACC
which
promise-3PL-PRES-IND
254
‘and since it is natural for them not to do what they promise’ (Celest. 269,6 in Jensen and Lathrop 1973: 69) (6-58) OFr.
comme
toutes
choses
ça
jus
since
all-NOM
things-NOM
here
down
soient
falibles,
seulle
vertu,
be-3PL-PRES-SUBJ
fallible-NOM
only-NOM
virtue-NOM
dist
Tulles,
est
en
say-3SG-PRET-IND
Tulles-NOM
be-3SG-PRES-IND
in
la
puissance
d’elle
meismes
the-OBL
power-OBL
of—her-OBL
self-OBL
‘since all things down here are fallible, only virtue, said Tulles, is in her own power’ (Paix 63 in Jensen 1984: 79) This anomaly, however, has disappeared from the modern daughter languages, in which cause, when purely factual, is consistently expressed via the indicative.
6.3.2.4. In Adversative Clauses Adversative clauses in Latin, not unlike causal clauses, display an interesting distribution of conjunctions and corresponding moods.
The adversative
conjunctions (i.e. ‘although’) along with the moods with which they combine are shown below:
255
Table 29: Latin Adversative Conjunctions and Corresponding Moods Conjunction Mood quamquam etsi etiamsi tametsi quamvis ut cum
consistently indicative usually indicative
consistently subjunctive
Compare the following two examples, both from Caesar: (6-59) Lt.
nam
etsi
sine
ullo
periculo
for
although
without
any-ABL
peril-ABL
legionis
delectae
cum
equitatu
proelium
legion-GEN
chosen-GEN
with
cavalry-ABL
fight-ACC
fore
videbat,
tamen . . .
be-FUT-INF
see-3SG-IMPERF-IND
nevertheless
‘although he saw that a fight with the enemy’s cavalry would be without peril to his chosen legion, nevertheless . . .’ (Caes., D.B.G. 1.46) (6-60) Lt.
cum although
ea
ita
sint,
these-NOM
thus
be-3PL-PRES-SUBJ
tamen . . .
sese
cum
eis
nevertheless
REFL-PRON-ACC
with
them-ABL
pacem
esse facturum
peace-ACC
make-FUT-INF
‘although these things were so, nevertheless he would make peace with them’ (Caes., D.B.G. 1.14)
256
This strong association between specific conjunctions and mood must have presented problems once these conjunctions disappeared. It would seem that, yet again, the speakers of the daughter languages attempted to re-establish a semantic motivation behind mood selection. After adversative conjunctions in Old Spanish (e.g. aunque, maguera que) and Old French (e.g. combien que, quoique66), the indicative gained ground in purely factual instances free of doubt or volition: (6-61) OSp. Dios God
pecado, sin
del
çielo
non
crio
of—the
sky
NEG
create-3SG-PRET-IND
mager que although
homnes
assentado
men
seat-PP
es
en
todos
be-3SG-PRES-IND
in
all
‘God in the heavens did non create sin, even though it is found in all men’ (Egip. 41 in Jensen and Lathrop 1973: 77) (6-62) OFr.
Au meins
j’en
at least
I-NOM—PTV-PRON
feray do-1SG-FUT-IND
mon
povoir,
Combien que
je
my-OBL
power-OBL
although
I-NOM
congnois
et
realize-1SG-PRES-IND
and
sçay
Que
know-1SG-PRES-IND
that
mon
langage
trouveray
Tout
my-OBL
language-OBL
find-1SG-FUT-IND
all-OBL
66
It is interesting to note that French quoique was not strongly grammaticalized until the Classical period. During the Old French period, quoique was still seen as an indefinite relative (‘however much’) and was therefore consistently followed by the subjunctive (cf. Jensen 1974: 93-94).
257
enroillié rust-PP-OBL
‘at least I will make an effort, even though I realize and know that I will find my language all rusty’ (Ch.O., Poésies 119,8 in Jensen 1974: 95) The above examples indicate that the speaker believes the subordinate proposition. In the following examples, however, this is not the case and there is therefore sufficient motivation for subjunctive: (6-63) OSp. mas but
avnque
fuesse
doliente,
although
be-3SG-IMPERF-SUBJ
suffering
podria
sanar
be able-3SG-COND
heal-INF
‘but even if he were suffering, he could recover’ (Celest. 50,15 in Jensen and Lathrop 1973: 76) (6-64) OFr.
se
siet
bien
loing
3SG-REFL-PRON
sit-3SG-PRES-IND
well
far
du
feu,
combien qu’il
from—the
fire-OBL
although—he-NOM
ait
grant
froit
have-3SG-PRES-SUBJ
great-OBL
cold-OBL
‘he (the good man) sits very far from the fire, even if he is very cold’ (Quinze 50 in Jensen 1974: 95) The subtle difference in modality between (6-61)-(6-62) and (6-63)-(6-64) is reflected in the translations provided, where I believe the most faithful English
258
translation for adversative clauses containing an indicative is ‘even though’, while subjunctive clauses are translated with ‘even if’. The clauses containing the subjunctive mood express situations which the speaker does not view as factual, as the person in (6-63) may not be suffering and the person referred to in (6-64) may not be cold. Although this semantic and/or pragmatic distinction—at the discretion of the speaker—remains strong in Spanish,67 further simplification has taken place in French, with the subjunctive being the norm for all adversative clauses. However, Grevisse acknowledges that there has always been fluctuation in this context and although the subjunctive is now considered the norm, the indicative is quite frequent in spoken French (Grevisse 1993: 1662). To conclude this section, adversative clauses have undergone a period of confusion likely due to a lack of motivation resulting from the grammaticalization and subsequent breakdown of the Latin conjunction system. Both tendencies seen here—a distinction based on notions of fact / non-fact, and a trend towards simplification—are ways in which speakers of the daughter languages have been able to reorganize the system.
67
In Modern Spanish, however, the distinction has been shown to be more closely related to the notion of assertion rather than factivity, a trend which will be discussed on pp. 265-267.
259
6.3.2.5. With Expressions of Emotion Subordinate clauses with expressions of emotion have experienced, perhaps, the greatest amount of variation (both synchronic and diachronic) in terms of mood. In Latin, emotion was expressed via impersonal verbs (e.g. iuvat ‘it delights’, miseret ‘it moves to pity’, piget ‘it vexes’) and personal verbs (e.g. miseresco ‘I pity’, gaudeo ‘I rejoice’, irascor ‘I am angry’). These verba affectuum were combined with both finite and non-finite subordinate constructions, and in a somewhat arbitrary fashion. The non-finite AcI construction is shown below: (6-65) Lt.
venire
tu
me
gaudes
come-PRES-INF
you-NOM
me-ACC
rejoice-2SG-PRES-IND
‘you are glad that I’m coming’ (Pl., Bacch. 185 in Bennett 1910: 378) However, verbs of rejoicing and grieving could be combined with either an AcI or an indicative clause (Ferrell 1999: 70). Compare the following two synonymous examples: (6-66) Lt.
paenitet
nos
haec
repent-3SG-PRES-IND
us-ACC
DEM-PRON-PL
fecisse do-PERF-INF
(6-67) Lt.
paenitet
nos
quod
haec
repent-3SG-PRES-IND
us-ACC
that
DEM-PRON-PL
fecimus do-1PL-PERF-IND
260
‘we are sorry that we acted so’ (Mountford 1984: 175 in Ferrell 1999: 70) Without a doubt, the most prevalent construction was the Accusative Infinitive construction, as evidenced by Cuzzolin’s (1994) work on verbs of emotion. His analysis of Cicero’s Epistulae yielded the following results: Table 30: AcI versus Quod/Quia with Verbs of Emotion VERBA AFFECTUUM accusative + infinitive quod / quia + finite verb dolere ‘pain’
36
5
gaudere ‘rejoice’
77
8
laetari ‘rejoice’
28
5
mirari ‘admire’
47
6
queri ‘lament’
12
8 (Cuzzolin 1994: 203)
Yet another variable, broadening the range of possible structures used with verbs of emotion, is that clauses expressing fear were followed, without exception, by the sequence ut (or negative ne) + the subjunctive, as shown below: (6-68) Lt.
ne
eius
supplicio
Divitiaci
NEG
his-GEN
punishment-ABL
Divitiacus-GEN
animum
offenderet,
soul-ACC
offend-3SG-IMPERF-SUBJ
verebatur fear-3SG-IMPERF-IND
‘he was afraid that by his (Dumnorix’s) punishment he would offend Divitiacus’ (Caes., D.B.G. 1.19) 261
Fear clauses in Latin are of particular interest because they retain the older paratactic construction, in which there are two independent clauses and the subjunctive is optative in nature (cf. Bennett 1910: 252). In other words, the literal translation of the example above would be: ‘he was afraid; let his punishment of Dumnorix not offend Divitiacus.’ The synchronically unmotivated negative particle in this case is merely a vestige of this previous paratactic stage. As hypotaxis prevailed, the original optative value of the clause was no longer present. Nevertheless, the subjunctive in fear clauses remained throughout all periods of Latin, again as a seemingly conditioned response to its environment (i.e. preceded by ut/ne). This state of affairs—characterized by a largely unmotivated distribution between AcI, indicative and subjunctive—can be seen as a sufficient impetus for the reorganization that has taken place in the daughter languages. In medieval Spanish and French texts, mood distribution after expressions of emotion gained semantic motivation, with subjunctive being used in cases of volition or doubt, and indicative elsewhere. The following two examples illustrate indicative mood in statements of fact where there is neither volition nor doubt: (6-69) OSp. alegre happy
era
mio
Çid
e
todos
be-3SG-IMPERF-IND
my
Cid
and
all
sos
vasallos,
que
Dios
les
his-PL
vassals
that
God
them-DAT
ovo
merçed
have-3SG-PRET-IND
mercy
262
‘my Cid was happy, along with all of his vassals, that God had mercy on them’ (Cid 1739 in Jensen and Lathrop 1973: 56) (6-70) OFr.
quel
dolor
que
li
Franceis
what
pain-NOM
that
the-NOM-PL
French-NOM-PL
nel
sevent
NEG—it-ACC
know-3PL-PRES-IND
‘how painful that the French don’t know’ (Rol. 716 in Jensen 1974: 46) In the examples that follow, on the other hand, there is a perceived motivation for subjunctive mood: (6-71) OSp. gozome, rejoice-1SG-PRES-IND—1SG-REFL-PRON
Parmeno,
que
Parmeno
that
ayas
limpiado
las
turbias
have-2SG-PRES-SUBJ
clean-PP
the
cloudy
telas
de
tus
ojos
films
from
your
eyes
‘I’m happy, Parmeno, that you (seem to) have cleaned the cloudy films from your eyes’ (Celest. 58,20 in Jensen and Lathrop 1973: 55) (6-72) OFr.
`
il
n’a
cure
he-NOM
NEG—have-3SG-PRES-IND
care-OBL
qu’an
le
that—INDEF-PRON-NOM
him-ACC
conoisse know-3SG-PRES-SUBJ
263
‘he doesn’t care that (if) someone knows him’ (Chrét., Chev. 6043 in Jensen 1974: 47) There is arguably a moderate amount of doubt in both of the above examples. In (6-71), Parmeno may or may not have come to his senses, while (6-72) is quasihypothetical. For this reason, I have chosen to include the terms ‘seem to’ and ‘if’ in my translations in order to more accurately convey this non-factive modality. Expressions of emotion have undergone further simplification in both Spanish and French, where categorical subjunctive usage is now the standard, as shown below: (6-73) Sp.
sentimos
mucho
que
la
carta
no
feel-1PL-PRES-IND
much
that
the
letter
NEG
haya
llegado
a
tiempo
have-3SG-PRES-SUBJ
arrive-PP
at
time
‘we’re very sorry that the letter didn’t arrive on time’ (Porto Dapena 1991: 108) (6-74) Fr.
je
m’étonne
I-NOM
1SG-REFL-PRON—surprise-1SG-PRES-IND
qu’il
revienne
déjà
that—he-NOM
return-3SG-PRES-SUBJ
already
‘I’m surprised that he’s returning already’ (Grevisse 1993: 1612)
264
It is often argued—and convincingly so—that this usage can be justified in terms of assertability (cf. Terrell and Hooper 1974, Klein 1975, Lunn 1989 and 1995, and Mejías-Bikandi 1994).68 This is yet additional evidence for simplification, since it demonstrates an attempt to unify the meanings of the subjunctive mood. Instead of thinking in terms of two or more distinct categories (e.g. volition and doubt) speakers may have expanded the field to include all instances of nonassertion (e.g. volition, doubt, comment clauses, etc.), thus relying on one single determining factor for mood selection. However, it is important to point out that there is a documented trend in Modern Spanish, whereby speakers of certain dialects prefer indicative in factive predicates such as those found after expressions of emotion (cf. Goldin 1974, García and Terrell 1977, Lantolf 1978, Studerus 1981, and Guitart 1982).69 The following example illustrates this preference, which contradicts standard usage: (6-75) Sp.
¡Qué
extraño
que
no
nos
how
strange
that
NEG
us-ACC
llamaron! call-3PL-PRET-IND
‘How strange that they didn’t call us!’ (García and Terrell 1977: 221) Taking the full development of these constructions into account, it would be unwise to consider this a random instance of mood loss. Instead, it seems to be 68 69
A detailed account of this explanation is found in section 2.2.3. A detailed account of this phenomenon is found in section 2.3.1.3.
265
evidence of different criteria used in the process of mood selection. Speakers who prefer indicative in these subordinate clauses are simply relying more heavily on non-factivity as opposed to non-assertion as their criteria for mood selection. This issue, dealing with the scope of the subjunctive value, can be illustrated as follows: Illustration 2: Scope of Subjunctive Mood non-assertion
non-factivity
volition
Volition is, as discussed earlier, the basic or prototypical subjunctive meaning, and it lies at the center of the subjunctive’s modal force. The broader category of non-factivity justifies the use of subjunctive after doubt, for instance, while the even broader category of non-assertion accounts for subjunctive after expressions of emotion, in adversative clauses, after Sp. el hecho de que / Fr. le fait que, etc. It is not surprising, then, that instances of reduction tend to occur in an inward fashion; that is to say, by limiting the scope of the subjunctive mood.
266
For
example, in García and Terrell’s (1977) study, native Spanish speakers from Mexico accepted the indicative with the following frequency:70 Subjective Comment Doubt Commands
48% 25% 8% (García and Terrell 1977: 216)
This reflects precisely the scope illustrated above, as a narrowing of scope will result, first, in the exclusion of subjunctive in comment clauses and secondly after expressions of doubt.
The implication here is that, while the future of the
subjunctive mood in Romance languages cannot be known for certain, if the mood is maintained at all in Romance languages, it will undoubtedly continue to appear in volitive clauses.
6.3.3. General Observations To conclude this section on refunctionalization of mood, I would like to summarize my perspective on the major changes that have occurred with regard to mood distribution in Romance languages. The two main changes from Latin to Romance are a change in structure (i.e. finite versus non-finite subordination) and a change in mood selection criteria.
The brief descriptions presented below
70
While it is true that Mexican speakers of Spanish may accept indicative more often than speakers from other countries, the hierarchy of commands, doubt and subjective comment appears to be a universal one.
267
reflect the complexity of criteria in Latin, along with the resultant simplification in Romance: Table 31: Subordinate Structures from Latin to Modern Romance Choice of structure (non-finite versus finite subordinate clause) Latin:
the choice is somewhat arbitrary, with infinitives being used primarily after verba declarandi, verba sentiendi, verba affectuum, and verba voluntatis
Medieval Sp. & Fr.:
the period is characterized by a reduction in Latin non-finite constructions
Modern Sp. & Fr.:
the choice is based primarily on syntactic criteria, with infinitives used primarily if the subjects of both clauses are coreferential
Table 32: Mood Distribution from Latin to Modern Romance Choice of mood (indicative versus subjunctive) Latin:
mood selection is based on semantics (i.e. volition), pragmatics (i.e. indirect discourse), and syntax (i.e. specific conjunctions)
Medieval Sp. & Fr.:
mood selection is becoming less automatic, with an overall increase in motivation as well as unification of semantic / pragmatic criteria
Modern Sp. & Fr.:
mood selection is based primarily on pragmatics [± assertive] or semantics [± factive], depending on language and/or dialect spoken
It goes without saying that these characterizations are not categorical in nature. Nevertheless, a strong trend toward simplification is evident and the contrast between the Latin and Romance systems is quite dramatic. There is both a quantitative and qualitative difference in the criteria in these two systems, as there 268
are not only fewer criteria in Romance languages but, in addition, the simplification of criteria promotes an increase in semantic and/or pragmatic motivation. Furthermore, it is possible to view the development of mood usage as pendular, whereby a gradual loss of motivation triggers the refunctionalization of a paradigm, including rule simplification and an increase in motivation. This is not meant to imply that the motivations are the same in a qualitative sense, since the motivation behind the PIE subjunctive was clearly not of the same nature as that found in modern Spanish and French. Nevertheless, the quantitative losses and gains in motivation can, in my view, be accurately characterized as pendular. I have illustrated this concept below:
Illustration 3: Pendular Development of Mood Motivation (PIE)
(Latin)
semantic motivation
>
loss of motivation
(FSp./FFr.)71
(Sp./Fr.) >
renewed motivation
>
loss of motivation?
↑ breakdown due to typological drift (i.e. change in branching and grammaticalization of conjunctions) 71
FSp. and FFr. represent Future Spanish and Future French, respectively, and are used here to refer to the potential for yet another decrease in motivation and second process of reanalysis or refunctionalization.
269
The developments documented for the medieval period indicate a restructuring of mood distributional patterns in order to re-motivate a viable paradigmatic opposition available to speakers. It is possible that, if the indicative/subjunctive contrast is felt, once again, to be unmotivated, there may be another period of refunctionalization similar to the one that occurred in the early Romance period. This assertion is, of course, purely speculative, though supported by an attested diachronic trend. It is thus my belief that another refunctionalization phase in the future is likely for those dialects in which the subjunctive paradigm is strong. For those in which this is not the case, refunctionalization may be difficult, and complete paradigmatic loss is a likely alternative.
6.4. Conclusion I have argued that the already complex and multifaceted category of mood, inherited from Proto-Indo-European, was destabilized due to typological drift. In specific,
the
shift
toward
right-branching
structures,
along
with
the
grammaticalization of subordinating conjunctions and specialization of the subordinator que resulted in a mood contrast that was no longer felt to be sufficiently motivated. As a consequence, the Romance subjunctive experienced a period of refunctionalization, as a result of which speakers of Romance languages were able to, once again, exploit a viable paradigmatic distinction to its fullest potential.
Furthermore, this refunctionalization process can be 270
characterized in terms of simplification, since the number of criteria involved in mood selection appears to have been reduced. This perspective is significant in that it illustrates the extent to which seemingly unrelated morpho-syntactic changes can undermine the stability of a system, thus encouraging its reorganization. If we were to ignore these issues and merely analyze changes in mood in isolation, we would potentially come to the erroneous conclusion that the subjunctive has merely been lost or reduced. Consider the following examples, where a subjunctive in Latin—used categorically after causal cum—is now expressed via the indicative in Spanish and French: (6-76) Lt.
Aedui,
cum
se
suaque
Aedui-NOM
since
themselves-ACC
theirs-ACC—and
ab
eis
defendere
non
from
them-ABL
defend-PRES-INF
NEG
possent,
legatos
ad
be able-3PL-IMPERF-SUBJ
ambassadors-ACC
to
Caesarem
mittunt
Caesar-ACC
send-3PL-PRES-IND
‘The Aedui, since they couldn’t defend themselves and their property from them (the Helvetii), sent ambassadors to Caesar’ (Caes., D.B.G. 1.11) (6-77) Sp.
ya que
no
podían
since
NEG
be able-3PL-IMPERF-IND
271
defenderse . . . defend-INF—3PL-REFL-PRON
‘since they couldn’t defend themselves’ (6-78) Fr.
puisqu’ils
ne
pouvaient
since—they-NOM
NEG
be able-3PL-IMPERF-IND
se
défendre . . .
3PL-REFL-PRON
defend-INF
pas NEG
‘since they couldn’t defend themselves’ Consider also the following examples of certainty (i.e. there is no doubt that…), which were expressed via the subjunctive in Latin but now take the indicative in Spanish and French: (6-79) Lt.
nec
dubium
enim
est,
quod
NEG
doubtful-NOM
for
be-3SG-PRES-IND
that
hic
primus
homo
Adam,
antequam
this-NOM
first-NOM
man-NOM
Adam
before
peccaret,
tipum
Redemptoris
sin-3SG-IMPERF-SUBJ
type-ACC
savior-GEN
domini
praetulisset
lord-GEN
display-3SG-PLUPERF-SUBJ
‘for there is no doubt at all that, before he committed sin, this first man Adam was similar to our Lord and Savior’ (GdT, H.F. 1.1) (6-80) Sp.
no
hay
duda
que . . .
NEG
there is
doubt
that
era
similar
be-3SG-IMPERF-IND
similar
‘there is no doubt that . . . he was similar’ 272
(6-81) Fr.
il n’y a
pas
de
doute
que . . .
NEG-there is
NEG
of
doubt
that
il
était
similaire
he-NOM
be-3SG-IMPERF-IND
similar
‘there is no doubt that . . . he was similar’ A decontextualized comparison of these changes in mood selection would lead one to conclude that there has simply been a gradual loss of subjunctive in Romance languages. However, now that we have taken into account the entire development beginning with a shift toward right-branching structures, we can see that this represents, instead, a type of therapeutic change, aimed at re-motivating the subjunctive / indicative contrast.
273
CHAPTER 7 CONCLUSION AND IMPLICATIONS 7.1. Background Research on Romance mood has been primarily synchronic, with the main goal of discovering the meaning(s) and/or function(s) behind the subjunctive. Syntactic analyses have argued that the subjunctive is automatically generated and appears in certain surface structures. They are not able to do so, however, without making reference to certain semantic and/or pragmatic traits (e.g. reservation).
The
majority of synchronic studies have, therefore, been semantic in nature, defining the subjunctive as a means of expressing notions such as evaluation, potentiality, attitude, comparison and debate.
Pragmatic analyses, which have become
increasingly prevalent in recent years, have argued that the subjunctive relates to factors such as non-assertion, thematic organization and reformulation. These studies have, without a doubt, been successful in terms of offering detailed descriptions of the ways in which speakers of Romance languages select subjunctive and indicative moods, along with providing useful terminology. On the other hand, certain aspects of this synchronic work have been problematic, perhaps due to the goal of the research itself. The expectation appears to be that there is a stable, underlying meaning for the subjunctive which simply needs to be identified. This expectation assumes that the subjunctive is a univalent paradigm conditioned by a rule (whether it be syntactic, semantic or 274
pragmatic). The reality—which lies outside the scope of synchronic research—is that Romance languages have inherited a subjunctive paradigm that represents a conflation of two moods (i.e. the PIE optative and subjunctive) and which, due to typological changes in Latin morpho-syntax, became increasingly complex and, ultimately, unmotivated. Diachronic research on mood is scarce, in spite of the fact that, in my view, there is a great amount of potential in this area. Linguists often refer to the apparent paradigmatic and syntagmatic loss of the subjunctive. Unfortunately, not only has there been an insufficient amount of diachronic research on Romance mood, but the meager amount that has been done is quite limited in scope. Since many scholars focus on only one language, significant typological trends often remain unidentified. This non-comparative approach may lead to a tendency to rely disproportionately on external causes for change. Furthermore, the focus is often on changes occurring within a very small window of time, thus ignoring essential early Latin developments. Without looking at other related languages and/or time periods, it is easy to interpret a phenomenon as “foreign” to the language, thus attributing it to external influence. Only a very small percentage of research on Romance mood has been both comparative and diachronic, while at the same time focusing exclusively on language internal developments. Harris (1974) has argued that hypotaxis itself has rendered the subjunctive redundant, while Raible (1992) has pointed to the 275
significance of word order—namely, the shift from OV to VO—in the creation of redundancy. These arguments, though significant, seem to represent pieces of an undoubtedly larger puzzle. With this in mind, my goal for this dissertation was to investigate the role of typological drift—specifically in terms of its impact on subordination techniques—in the development of the Romance subjunctive. From my perspective, many important issues in Romance mood can be explained in terms of typological drift, whereby instances of “mood loss” are actually signs of therapeutic change.
7.2. Recapitulation of Hypotheses and Findings Given that my primary aim was to address what I believe to be the most underrepresented aspect of Romance mood, the focus of this dissertation has not been the subjunctive itself, but rather, the underlying developments in morphosyntax which necessitated changes in mood distribution.
My hypotheses, as
presented in chapter one, are as follows: 1)
Due to the tendency toward branching congruency, the typological shift from left-branching to right-branching structures led to an increase in subordinating conjunctions in Latin.
2)
This increase in frequency accelerated the development of subordinating conjunctions in Latin, a process which can be analyzed effectively within the grammaticalization framework. 276
3)
As a result of this grammaticalization process, the subordinator que and the subjunctive were able to undergo partial synthesis.
4)
These morpho-syntactic changes resulted in a significant amount of instability in mood distribution, thus triggering the refunctionalization of the subjunctive/indicative contrast in the Romance languages.
7.2.1. Hypothesis #1 Inspired by the work of Raible (1992), I set out to investigate the way in which changes in word order may have led to an increase in explicit subordinating conjunctions. However, instead of basing my analysis on the distinction between OV and VO word order, I have focused on the distinction between left-branching and right-branching structures. In this way, I have been able to analyze Latin data from much earlier periods, which are primarily OV, yet do exhibit both SC+MC and MC+SC branching patterns. Furthermore, while Raible (1992) has argued that the sequences SC* MC and MC *SC are prevalent due to an optimal (i.e. medial) location for linking devices, I have instead motivated the preference based on the notion of branching congruency, whereby right-branching markers of subordination (i.e. pre-posed conjunctions) are more consistent with rightbranching (i.e. post-posed) subordinate clauses. In chapter three, I analyzed subordinate clauses in three Latin texts— Senatus Consultum de Bacchanalibus (Old Latin), Caesar’s De Bello Gallico 277
(Classical Latin) and Peregrinatio Egeriae (Late Latin)—with the expectation that a preference for branching congruency would be seen. In other words, I anticipated a high level of correlation between left-branching (i.e. pre-posed) subordinate clauses and left-branching subordination techniques (i.e. subjunctive alone, infinitive, participle), and a high level of correlation between rightbranching (i.e. post-posed) subordinate clauses and right-branching subordination techniques (i.e. explicit subordinating conjunctions). illustrate
left-branching
congruency
and
The examples below
right-branching
congruency,
respectively: L-B sentence structure + L-B subordination technique: (7-1)
Lt.
nam
[[ mihi
credat]
for
me-DAT
believe-3SG-PRES-SUBJ
volo]
affectio
vestra
want-1SG-PRES-IND
affection-NOM
your-NOM-SG
‘for I hope that you (your affection) will believe me’ (Pereg. 7.3) R-B sentence structure + R-B subordination technique: (7-2)
Lt.
[rogavimus
presbyteros,
[ut
beg-1PL-PERF-IND
priests-ACC
that
ibi
fieret
oblatio]]
there
be made-3SG-IMPERF-SUBJ
offering-NOM
‘we asked the priests to make the offering there’ (Pereg. 4.8)
278
The data analyzed in chapter three confirmed my hypothesis, as shown in the following table, in which I have combined the results from all three texts: Table 33: Summary of Subordination Patterns in Latin Data morphology (L-B) conjunction (R-B) subord. + main (L-B)
83 / 118 (70%)
35 / 118 (30%)
main + subord. (R-B)
23 / 108 (21%)
85 / 108 (79%)
In spite of the tremendous amount of variation in Latin word order, my analysis has indeed revealed a strong preference for branching congruency. Therefore, one can conclude that an increase in post-posed subordinate clauses would indeed entail an increase in explicit subordinating conjunctions, with the aim of maintaining branching congruency. The table below illustrates this trend: Table 34: L-B and R-B Subordinate Structures (Table 13) LEFT-BRANCHING RIGHT-BRANCHING
[[faciendam]
censuere]
S.C. Bacch. 25 [[habere]
Pereg. 4.8
arbitrabantur]
[quia videbantur]]
[scire
Caes., D.B.G. 1.2 [[rebus comparatis]
[ut fieret]]
[rogavimus
Pereg. 3.8
diem dicunt]
[fuit Moyses
Caes., D.B.G. 1.6
[cum ascendisset]]
Pereg. 3.7
279
These explicit subordinating conjunctions, due to their increasingly vital role in Latin
syntax,
subsequently
underwent
a
fairly
extensive
process
of
grammaticalization.
7.2.2. Hypothesis #2 In chapter four, I then focused specifically on a development not addressed by either Harris (1974, 1978b) or Raible (1992), who instead propose a direct link between the increase in explicit subordinating conjunctions and redundancy (and subsequent loss) of the subjunctive. In my view, however, the development of these conjunctions provides a necessary bridge between the two developments and thus provides additional motivation for claims regarding redundancy. Therefore, I have conducted an analysis of Latin conjunctions via the grammaticalization framework, with the primary goal of revealing a fundamental difference in the nature of Latin and Romance subordinating elements. In Latin, there was a vast inventory of subordinating conjunctions simultaneously undergoing grammaticalization. Due to the way in which forms tend to compete during the grammaticalization process, I chose to analyze six conjunctions, all based on the PIE relative pronoun (< *kw): cum, quando, quoniam, quia and quod, along with Romance que. The general development of these conjunctions can be illustrated as follows:
280
Table 35: Grammaticalization Cline for Subordinating Elements (Table 16) A B C relative > temporal < > causal > complementizer pronoun conjunction conjunction ACC form
>
quom
< >
quom
ACC form + DEM stem
>
quando
< >
quando
ACC form + jam ‘now’
>
quoniam
< >
quoniam
>
quoniam
ACC N PL form
>
quia
< >
quia
>
quia
< >
quod
>
quod
>
que
ACC N SG form hybrid form
In terms of generalization or extension of semantic meaning, all six forms gained wider applicability, expressing notions such as time and cause. Four of these forms (quoniam, quia, quod and que) also became so generalized that they could be used as semantically “empty” elements, as shown below in two examples from Peregrinatio Egeriae: (7-3)
Lt.
solum
scio,
quia
postmodum
only
know-1SG-PRES-IND
that
afterwards
puer
Abrahae,
ut
peteret
boy-NOM
Abraham-GEN
in order to
seek-3SG-IMP-SUBJ
Rebeccam…
in
Charra
venerit
Rebecca-ACC
in
Charrae
come-3SG-PERF-SUBJ
281
only that afterwards Abraham's servant came to Charrae to seek Rebecca’ ‘I know
(Pereg. 20.9) (7-4)
Lt.
Pharao,
quando
vidit,
quod
Pharaoh
when
see-3SG-PERF-IND
that
filii
Israhel
dimiserant
children-NOM
Israel-GEN
abandon-3PL-PLUPERF-IND
eum, . . .
isset
cum
omni
him-ACC
go-3SG-PLUPERF-SUBJ
with
all-ABL
exercitu
suo
intra
Ramesse
army-ABL
his-ABL
inside
Rameses-ACC
‘when Pharaoh saw that the children of Israel had escaped him, . . . he went with all his army into Rameses’ (Pereg. 8.5) The process of decategorialization was evidenced both on syntactic grounds (i.e. the development from a relative pronoun to a complementizer) and morphological grounds (i.e. the loss of nominal inflection). Specialization, or the narrowing down of an inventory of competing forms, was also seen in this case. In late Latin, this specialization process resulted in a single complementizer: quod. In early Romance, however, competition occurred between quod and a newer form (i.e. que), with que surfacing as the winner. Finally, there was a process of renewal, whereby the restructuring of the entire conjunction system took place, building new forms based on the highly-grammaticalized Romance element: que (e.g. Sp. porque, para que, aunque, después de que and Fr. parce
282
que, pour que, quoique, après que, etc.).
Furthermore, the morphological
regularity resulting from this process has arguably played a significant role in further developments.
7.2.3. Hypothesis #3 In chapter five, I argued that the increasing frequency and coexistence of que and the subjunctive has led to a partial synthesis between the two elements, which I believe has ultimately resulted in a perceived loss of motivation. In contrast to the notion of syntheticity and analyticity as a dichotomy, I agree with Schwegler’s (1990) position that they represent two ends of a spectrum and that two linguistic elements may undergo a process of synthesis even if they are non-contiguous (e.g. ne…pas). While I argued against the validity of several of his morphosyntactic criteria, I was nevertheless able to demonstrate a high level of relevance (due to frequency) between que and the subjunctive, which points to semantic synthesis. I also identified a trend toward increased contiguity, as seen in the repetition of the subordinator que before the subjunctive verb. Finally, I claimed that synthesis is most strongly evidenced by an increase in obligatoriness, as illustrated via the following examples:
283
(7-5)
OSp. ¡que that
Dios
me
la
mantenga!
God
me-DAT
her-ACC
keep-3SG-PRES-SUBJ
‘may God keep her’ (Hita, L.B.A. 939a in Jensen and Lathrop 1973: 15) (7-6)
OFr.
s’il
est
si
hardiz
if—he-NOM
be-3SG-PRES-IND
so
bold-NOM
qu’il
i
veigne!
that—he-NOM
LOC-PRON
come-3SG-PRES-SUBJ
‘if he is so bold, let him come!’ (Passion 1355 in Jensen 1974: 18) This increase in obligatoriness leads (as in the case of ne…pas) to a certain degree of semantic and functional redundancy, represented below in stage three of the complete development: Stage 1:
verb
+
SUBJ morphology
Stage 2:
CONJ +
verb
+
SUBJ morphology
Stage 3:
SUB
verb
+
SUBJ morphology
+
7.2.4. Hypothesis #4 In chapter six, I argued that all three of the above issues in morpho-syntax had a destabilizing effect on mood contrast in Romance languages. A shift toward right-branching structures led to an increase in finite subordinate clauses at the expense of non-finite (i.e. non-modal) forms. These new structures, many of which had not previously been associated with a particular mood, now needed to 284
be integrated into the system.
In addition, new right-branching future and
conditional forms led to an increase in modality among older left-branching equivalents, which consequently usurped several previous subjunctive functions. The grammaticalization of subordinating elements led to further destabilization, as mood selection appears to have been partly conditioned by specific Latin conjunctions, which then disappeared as a result of the specialization process. The process of synthesis, which increased relevance between que and the subjunctive, may have led to an increase in non-modal subjunctive usage, in addition to the perception of redundancy.
In short, these issues led to the
refunctionalization of Romance mood, whereby the subjunctive was re-motivated based on semantic and/or pragmatic notions. The following Latin examples show contexts which previously required subjunctive but now call for indicative in Modern Spanish and French: (7-7)
Lt.
dicunt
alii,
quod
ad
ipsam
say-3PL-PRES-IND
some-NOM
that
to
this-ACC
ripam . . .
sint
reversi
shore-ACC
be-3PL-PRES-SUBJ
return-PERF-PART-NOM
‘some say that they (the Israelites) returned to this shore’ (GdT, H. F. 1.10) (7-8)
Lt.
fiebat,
ut
et
minus
late
happen-3SG-IMPERF-IND
that
and
less
widely
vagarentur
et
minus
facile
roam-3PL-IMPERF-SUBJ
and
less
easily
285
finitimis
bellum
inferre
neighbors-DAT
war-ACC
bring-PRES-INF
possent be able-3PL-IMPERF-SUBJ
‘it came to pass that they wandered less widely and could less easily bring war upon the neighboring clans’ (Caes., D.B.G. 1.2) (7-9) Lt.
cum
ipse
eum
locum
because
himself-NOM
that-ACC
place-ACC
amicitiae
apud
eum
teneret,
friendship-GEN
with
him-ACC
hold-3SG-IMPERF-SUBJ
neminem
existimaturum
non
sua
no one-ACC
think-FUT-INF
NEG
his-ABL
voluntate
factum
will-ABL
be done-FUT-INF
‘since he (Divitiacus) was Caesar’s intimate friend, no one would think it (the punishment against Dumnorix) was done without his consent’ (Caes., D.B.G. 1.20) In isolation, these would appear to be cases of subjunctive loss. However, the shift can be justified in terms of therapeutic change aimed at re-motivating a viable paradigmatic distinction. I have illustrated this series of developments, which is potentially pendular in nature, in the following diagram:
286
Illustration 4: Pendular Development of Mood Motivation (Illustration 3) (PIE)
(Latin)
semantic motivation
>
loss of motivation
(OSp./OFr.) >
renewed motivation
(FSp./FFr.) >
loss of motivation?
↑ breakdown due to typological drift (i.e. change in branching and grammaticalization of conjunctions)
7.3. Concluding Remarks I believe that this dissertation has offered a new perspective on Romance mood, one in which developments are contextualized rather than examined in isolation. It is common for researchers—when interested in a morphological feature or paradigm such as the subjunctive—to focus exclusively on the paradigm in question. Unfortunately, an approach that fails to address the broader typological context including changes in syntagmatic relationships is likely to overlook certain fundamental characteristics of a linguistic feature or diachronic development. In the following sections, I will present a brief summary of my conclusions regarding Romance mood, as well as general implications for the field of diachronic linguistics.
287
7.3.1. General Conclusions regarding Romance Mood One of the important conclusions to be drawn from this dissertation relates to the misleading nature of superficial comparisons based solely on a single surface structure.
A cursory look at the following complex sentences in Latin and
Modern Romance reveals an identical pattern: (7-10) Lt.
volo
ut
venias
(7-11) Sp.
quiero
que
vengas
(7-12) Fr.
je veux
que
tu viennes
want-1SG-PRES-IND
CONJ/SUB
come-2SG-PRES-SUBJ
‘I want you to come’ Therefore, if we were to investigate the matter no further, we would believe changes in mood selection (e.g. subjunctive > indicative) to be completely independent developments.
My analysis has instead underlined the essential
differences between Latin and Romance complex sentences, due primarily to a shift toward finite, right-branching forms followed by the complete restructuring of the Romance conjunction system resulting in a single generalized subordinator. These changes culminated in the destabilization of mood distributional patterns and, as a result, instances of subjunctive “loss” are accounted for in terms of therapeutic change characterized by the refunctionalization of mood contrast. The strong emphasis on Latin in this dissertation also enables me to view changes in mood distribution in a different light, by drawing attention to the
288
importance of the linguist’s perspective regarding language change. Instead of asking Why is the subjunctive less frequent in Romance languages than in Latin?, we should be asking precisely the opposite: Why was the subjunctive so pervasive in Latin? The underlying assumption behind the first of these questions reveals the traditionally-held view that Latin was an ideal, highly-motivated system, with all subsequent changes representing language decay. My analysis has revealed, however, that subjunctive usage in Latin had become not only inconsistent, but also somewhat conditioned by its syntactic environment and was, therefore, lacking the semantic motivation linguists generally ascribe to the paradigm. A third conclusion I have drawn from my research goes against most claims made regarding the Romance subjunctive. It is commonly argued that changes in mood have been most prevalent in main clauses, with subordinate clauses having remained relatively stable. However, as a result of my analysis, I have come to believe just the opposite. In main clauses, two basic changes have occurred: (1) the subjunctive remains but is now preceded by que due to synthesis (e.g. in optative and jussive clauses), and (2) the conditional has replaced the subjunctive (e.g. in deliberative and potential clauses). While it is true that a new paradigm has usurped several previously subjunctive functions, this new paradigm (i.e. conditional) is virtually synonymous in terms of modality. What we are dealing with, then, is simply a new exponent for subjunctive mood in these cases, and the modality of these clauses has not changed at all. In subordinate 289
clauses, on the other hand, the refunctionalization of mood contrast has led to instances of indicative replacing subjunctive (e.g. in causal clauses) and subjunctive replacing indicative (e.g. with expressions of emotion).
In these
cases, speakers have not merely substituted an equivalent form to express a constant modal value but, rather, have completely reversed the modal expression. In short, contrary to existing descriptions, I view the modal changes in main clauses merely as variations in the surface representation, while changes in subordinate clauses represent much more significant developments.
7.3.2. Implications for Diachronic Romance Linguistics At this final stage, I would like to address potential implications of this project for the field of diachronic linguistics. First and foremost, while a narrow focus may facilitate detailed descriptions and analyses in synchronic research, a broad perspective is often better suited for diachronic studies. With regard to this assertion, I believe this dissertation has highlighted several important notions in diachronic linguistics: 1)
Major developments such as changes in word order can have far-reaching consequences with regard to seemingly unrelated linguistic features.
2)
The grammaticalization process—though not itself goal oriented—can have substantial effects on the language system due to its reductive tendencies. 290
3)
The role of system congruency should not be underestimated, as it often serves to explain patterns or directions of change.
4)
While not entirely dismissing the potential influence of reductive phonology and external interference, morpho-syntactic change is often therapeutic in nature, thus representing a highly-motivated repair mechanism.
By keeping these tendencies in mind, diachronic linguists may be more receptive to potentially explanatory factors previously not taken into consideration, in addition to uncovering universal language tendencies. I would also like to stress the significance of the notion of drift, a concept which has, at least indirectly, motivated the investigation pursued in this dissertation. The term drift—introduced by Sapir (1921)—presents a challenge to linguists, specifically due to its ambiguous meaning. Whether one follows the strongest definition (where drift implies a predetermined, programmed trajectory) or the weakest definition (where drift refers merely to a slow, unconscious development shared among related languages), I believe the concept has much to offer the field of diachronic linguistics. First of all, drift points to a common bond between genetically related languages and, in this way, it places emphasis on similarities among languages rather than differences.
Unfortunately, a
disproportionate amount of attention is generally given to contrastive analyses, as “the favoring of divergence (split, fragmentation, and the like) over convergence 291
or coalescence has almost become an occupational hazard with comparatists of all persuasions” (Malkiel 1981: 545). Furthermore, the notion of drift implies a certain hierarchy, which I believe to be of fundamental importance, as it encourages us to distinguish between major and minor changes. This distinction is not in any way meant to imply that minor changes are (or should be) of less interest to linguists. It does, however, provide us with a frame of reference, whereby minor changes may be interpreted as direct or indirect correlates to the major developments related to typological drift.
292
BIBLIOGRAPHY Abouda, Lotfi. 2002. “Négation, intérrogation et alternance indicatif-subjonctif.” Journal of French Language Studies 12.1: 1-22. Adams, J. N. 1977. “A Typological Approach to Latin Word Order.” Indogermanische Forschungen 81: 70-99. Anderson, Hanne Leth. 1993. “Les complétives non introduites en français parlé.” Subordinations. Eds. Claude Muller and Daniel Roulland. Rennes: Presses Universitaires de Rennes, pp. 5-14. Arteaga, Deborah. 1993. “Syntax and Mood Choice in French.” Language Quarterly 31.3-4: 196-216. Ayres-Bennett. 1996. A History of the French Language through Texts. London and New York: Routledge. Bauer, Brigitte L. M. 1995. Development of SVO Patterning in Latin and French. Diachronic and Psycholinguistic Perspectives. New York and Oxford: Oxford University Press. ----. 2003. “The Adverbial Formation in mente in Vulgar and Late Latin. A Problem in Grammaticalization.” Latin tardif et latin vulgaire VI. Eds. Heikki Solin, Martti Leiwo and Hilla-Halla-aho. Hildesheim: Olms, pp. 439-457. ----. 2006. “Synthetic and Analytic in Romance. The Importance of Varieties.” Historical Romance Linguistics. Retrospective and Perspectives. Eds. Randall Gess and Deborah Arteaga. Amsterdam: John Benjamins, pp. 287-304. Bello, Andrés. 1988 [1847]. Gramática de la lengua castellana destinada al uso de los americanos. Madrid: Arco Libros. Bennett, Charles E. 1910. Syntax of Early Latin. Volume 1: The Verb. Boston: Allyn and Bacon. ----. 2000. New Latin Grammar. Reprint of the 1908 Allyn and Bacon edition. Wauconda, Illinois: Bolchazy-Carducci Publishers, Inc.
293
Bergen, John J. 1978. “One Rule for the Spanish Subjunctive.” Hispania 61.2 (May): 218-234. Bhat, D. N. S. 1999. Prominence of Tense, Aspect and Mood. Studies in Language Companion Series 49. Amsterdam: John Benjamins. Bichakjian, Bernard. 1982. “La genèse de la subordination de l’Indo-Européen au Français.” Mélanges de Linguistique, de littérature et de philologie médievales, offerts à J. R. Smeets. Eds. Q. I. M. Mok, I. Spiele and P. E. R. Verhuyck. Leiden: Instituut Frans, pp. 5-19. Blumenthal, Peter. 1998. “Le fait que. Origine et combinatoire.” Travaux de linguistique 36: 147-160. Bolinger, Dwight. 1974. “One Subjunctive or Two?” Hispania 57 (Sept): 462471. ----. 1975. “Again—One or Two Subjunctives?” Hispania 59.1 (Mar): 41-49. Boysen, Gerhard. 1966. “L’emploi du subjonctif dans l’histoire des langues romanes.” Bulletin des jeunes romanistes 13: 19-33. Bybee, Joan, et. al. 1994. The Evolution of Grammar. Tense, Aspect, and Modality in the Languages of the World. Chicago: University of Chicago Press. Canut, Cécile and Gudrun Ledegen. 1998. “Après que…ou la fluctuation des modes en français parlé.” Langage et société 85: 25-53. Comrie, Bernard. 1989. Language Universals and Linguistic Typology. Chicago: University of Chicago Press. Connors, Kathleen. 1978. “The Meaning of the French Subjunctive.” Linguistics 211 (Sept): 45-56. Cristofaro, Sonia. 1998. “Grammaticalization and Clause Linkage Strategies. A Typological Approach with Particular Reference to Ancient Greek.” The Limits of Grammaticalization. Eds. Anna Giacalone Ramat and Paul J. Hopper. Amsterdam: John Benjamins, pp. 59-88.
294
Cuzzolin, Pierluigi. 1994. “On Sentential Complementation after verba affectuum.” Linguistic Studies on Latin. Ed. József Herman. Amsterdam: John Benjamins, pp. 201-210. Danchev, Andrei. 1988. “Language Contact and Language Change.” Folia Linguistica 22.1-2: 37-53. Diessel, Holger. 2001. “The Ordering Distribution of Main and Adverbial Clauses. A Typological Study.” Language 77.3: 433-455. Donaire, María Luisa. 2001. Subjuntivo y polifonía (español, francés). Madrid: Arrecife. ----. 2003. “Les sélecteurs du subjonctif, un domaine sémantique défini?” Theleme. Revista complutense de estudios franceses. Special issue: 121135. Epstein, Richard. 1994. “The Development of the Definite Article in French.” Perspectives on Grammaticalization. Ed. William Pagliuca. Amsterdam: John Benjamins, pp. 63-80. Farkas, Donka F. 1992. “On the Semantics of Subjunctive Complements.” Romance Languages and Modern Linguistic Theory. Papers from the 20th Linguistic Symposium on Romance Languages, Ottowa, 10-14 April 1990. Eds. Paul Hirschbuhler and Konrad Koerner. Amsterdam: John Benjamins, pp. 69-104. Ferrell, Julia E. 1999. The Development of Subjunctive Use with Expressions of Emotion from Latin to Spanish. Dissertation: University of Michigan. Fischer, Olga and Anette Rosenbach. 2000. “Introduction.” Pathways of Change. Grammaticalization in English. Eds. Olga Fischer, Anette Rosenbach and Dieter Stein. Amsterdam: John Benjamins, pp. 1-37. Fleischman, Suzanne. 1982. The Future in Thought and Language. Diachronic Evidence from Romance. Cambridge: Cambridge University Press. Floyd, Mary Beth. 1978. “Verb Usage in Southwest Spanish: A Review.” The Bilingual Review 5:1-2 (Jan-Aug): 76-90.
295
Foster, David William. 1973. “The Spanish Subjunctive as a Non-Semantic Category.” The English Language Journal 4.4: 191-200. Frank, Tenney. 1904. Attraction of Mood in Early Latin. Lancaster, Pa: The New Era Printing Company. García, Mary Ellen and Tracy Terrell. 1977. “Is the Use of Mood in Spanish Subject to Variable Constraints.” Studies in Romance Linguistics. Ed. Michio Peter Hadiwara. Rowley, Mass.: Newbury House, pp. 214-226. Gifford, D. J. and F. W. Hodcroft. 1966. Textos lingüísticos del medioevo español. 2nd ed. Oxford: The Dolphin Book Company. Gildersleeve, B. L. and G. Lodge. 1997. Gildersleeve’s Latin Grammar. 3rd ed. Wauconda, IL: Bolchazy-Carducci Publishers, Inc. Givón, Talmy. 1991. “The Evolution of Dependent Clause Morpho-syntax in Biblical Hebrew.” Approaches to Grammaticalization. Eds. Elizabeth Traugott and Bernd Heine. Amsterdam: John Benjamins. Goldin, Mark G. 1974. “A Psychological Perspective of the Spanish Subjunctive.” Hispania 57:2: 295-301. Greenberg, Joseph H. 1960. “A Quantitative Approach to the Morphological Typology of Language.” International Journal of American Linguistics 26.3: 178-194. ---. 1963. “Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements.” Universals of Language. Ed. Joseph H. Greenberg. Cambridge and London: The M.I.T. Press, pp. 73-113. Grevisse, Maurice. 1993. Le bon usage. Grammaire française. 13th ed. Paris: Duculot. Guitart, Jorge M. 1982. “On the Use of the Spanish Subjunctive among SpanishEnglish Bilinguals.” Word 33.1-2 (April-August): 59-67. Hahn, E. Adelaide. 1953. Subjunctive and Optative. Their Origins as Futures. New York: American Philological Association.
296
Handford, S.A. 1947. The Latin Subjunctive. Its Usage and Development from Plautus to Tacitus. London: Methuen & Co. Ltd. Harris, Martin. 1971. “The History of the Conditional Complex from Latin to Spanish. Some Structural Considerations.” Archivum Linguisticum 2: 2533. ----. 1974. “The Subjunctive Mood as a Changing Category in Romance.” Historical Linguistics II. Theory and Description in Phonology. Eds. S. C. Dik and J. G. Kooij. Amsterdam: North-Holland, pp. 169-188. ----. 1978a. “The Inter-relationship between Phonological and Grammatical Change.” Recent Developments in Historical Phonology. Ed. Jacek Fisiak. The Hague: Mouton, pp. 159-172. ----. 1978b. The Evolution of French Syntax. A Comparative Approach. London and New York: Longman. ----. 1981. “On the Conditional Mood in French.” Folia Linguistica Historica 2.1: 55-69. Haspelmath, Martin. 1999. “Why is Grammaticalization Irreversible?” Linguistics 37.6: 1043-1068. Heine,
Bernd, Ulrike Claudi, and Friederike Hünnemeyer. 1991. Grammaticalization. A Conceptual Framework. Chicago and London: The University of Chicago Press.
Heine, Bernd and Tania Kuteva. 2002. World Lexicon of Grammaticalization. Cambridge: Cambridge University Press. ----. 2005. Language Contact and Grammatical Change. Cambridge: Cambridge University Press. Hengeveld, Kees and Gerry Wanders. 1997. “On the Use of Subjunctive and Indicative Verb-Forms in Adverbial Clauses.” Language Change and Functional Explanations. Ed. Jadranka Gvozdanovic. Berlin: Mouton, pp. 249-271.
297
Hensey, Fritz. 1973. “Grammatical Variables in Southwestern American Spanish.” Linguistics 108: 5-26. ----. 1976. “Toward a Grammatical Analysis of Southwestern Spanish.” Studies in Southwest Spanish. Eds. J. Donald Bowen and Jacob Ornstein. Rowley, Mass.: Newbury House, pp. 29-44. Herman, József. 1963. La formation du système des conjonctions de subordination. Berlin: Akademie-Verlag. ----. 1989. “Accusativus cum infinitivo et subordonnée à quod, quia en latin tardif. Nouvelles remarques sur un vieux problème.” Subordination and other Topics in Latin. Proceedings of the Third Colloquium on Latin Linguistics. Ed. Gualtiero Calboli. Amsterdam: John Benjamins, pp. 133-152. ----. 2000. Vulgar Latin. Trans. Roger Wright. University Park, PA: Pennsylvania State University Press. Hewson, John and Vit Bubenik. 1997. Tense and Aspect in Indo-European Languages. Theory, Typology, Diachrony. Amsterdam: John Benjamins. Hopper, Paul J. and Elizabeth Traugott. 1993. Grammaticalization. Cambridge: Cambridge University Press. Jensen, Frede. 1974. The Syntax of the Old French Subjunctive. The Hague and Paris: Mouton. Jensen, Frede and Thomas A. Lathrop. 1973. The Syntax of the Old Spanish Subjunctive. The Hague and Paris: Mouton. Journoud, Sylvain. 1971. “Déclin du subjonctif.” The French Review 44.3 (Feb): 549-551. Justus, Carol F. 1980. “Typological Symmetries and Asymmetries in Hittite and IE Complementation.” Linguistic Reconstruction and Indo-European Syntax. Proceedings of the Colloquium of the ‘Indogermanische Gesellschaft.’ Eds. Paolo Ramat et al. Amsterdam: John Benjamins, pp. 183-206.
298
Kelemen, J. 1970. “Sprachtypology und Sprachstatistik.” Theoretical Problems of Typology and the Northern Eurasian Languages. Eds. L. Dezsö and P. Hajdú. Amsterdam: Grüner, pp. 53-63. Kilroe, Patricia. 1994. “The Grammaticalization of French à.” Perspectives on Grammaticalization. Ed. William Pagliuca. Amsterdam: John Benjamins, pp. 49-61. Klausenburger, Jurgen. 2000. Grammaticalization. Studies in Latin and Romance Morphosyntax. Amsterdam: John Benjamins. Klein, Flora. 1975. “Pragmatic Constraints on Distribution. The Spanish Subjunctive.” Papers from the Regional Meetings, Chicago Linguistic Society 11: 353-365. Krakusin, Margarita and Aristófanes Cedeño. 1992. “Selección del modo después de el hecho de que.” Hispania 75.5 (Dec): 1289-1293. Krupa, Victor. 1965. “On Quantification of Typology.” Linguistics 24: 29-37. Lanly, André. 1973. “Sur le conditionnel français et roman (à propos d’une remarque de M. Imbs).” Travaux de linguistique et de littérature 1.1: 391399. Lantolf, James P. 1978. “The Variable Constraints on Mood in Puerto RicanAmerican Spanish.” Contemporary Studies in Romance Linguistics. Ed. Margarita Suñer. Washington, DC: Georgetown University Press, pp. 193217. Lavandera, Beatriz. 1979. “Análisis semántico de variación en tiempos verbales. Oraciones condicionales del español.” Anuario de letras 17: 113-136. Le Bidois, G. and R. Le Bidois. 1967. Syntaxe du français moderne. Ses fondaments historiques et psychologiques. Paris: Editions A. et J. Picard & Cie. Lehmann, Christian. 1989. “Latin Subordination in Typological Perspective.” Subordination and other Topics in Latin: Proceedings of the Third Colloquium on Latin Linguistics. Ed. Gualtiero Calboli. Amsterdam: John Benjamins, pp. 153-179.
299
----. 1995. Thoughts on Grammaticalization. München and Newcastle: Lincom Europa. Lehmann, Winfred P. 1973. “A Structural Principle of Language and its Implications.” Language 49.1: 47-66. ---. 1974. Proto-Indo-European Syntax. Austin: University of Texas Press. ----. 1980. “The Reconstruction of Non-Simple Sentences in Proto-IndoEuropean.” Linguistic Reconstruction and Indo-European Syntax. Proceedings of the Colloquium of the ‘Indogermanische Gesellschaft.’ Eds. Paolo Ramat et al. Amsterdam: John Benjamins, pp. 113-144. ----. 1993. Theoretical Bases of Indo-European Linguistics. London and New York: Routledge. Leiper, M.A. 1913. Latin Subordinate Clause Syntax. New York, Cincinnati, Chicago: American Book Company. Levin, Harry and Peter Garrett. 1990. “Sentence Structure and Formality.” Language in Society 19: 511-520. Lewis, Charlton T. and Charles Short. 1879. A Latin Dictionary. Founded on Andrews’ Edition of Freund’s Latin Dictionary. Oxford: Clarendon Press. Ligatto, Dolores. 2002. “Discourse Criteria in the Selection of Mood in Spanish. Concessive Clauses.” Hispania 85.1 (Mar): 137-149. Lunn, Patricia V. 1989. “Spanish Mood and the Prototype of Assertability.” Linguistics 27.4 (302): 687-702. ----. 1995. “The Evaluative Function of the Spanish Subjunctive.” Modality in Grammar and Discourse. Eds. Joan Bybee and Suzanne Fleischman. Amsterdam: John Benjamins, pp. 429-449. Luraghi, Silvia. 1990. Old Hittite Sentence Structure. New York: Routledge. Makamina, Jonas Bena. 2003. “La problématique de la terminologie grammaticale. La nomenclature des modes verbaux.” Revue romane 38.1: 125-150.
300
Malkiel, Yakov. 1981. “Drift, Slope, and Slant. Background of, and Variation upon, a Sapirian Theme.” Language 57.3: 535-570. Matthews, Peter. 1997. The Concise Oxford Dictionary of Linguistics. Oxford and New York: Oxford University Press. McCray, Stanley. 1988. Advanced Principles of Historical Linguistics. New York: Peter Lang. Meillet, Antoine. 1958 [1912]. Linguistique historique et linguistique générale. Paris: Champion. Meillet, A. and J. Vendryes. 1953. Traité de grammaire comparée des langues classiques. Paris: Champion. Mejías-Bikandi, Errapel. 1994. “Assertion and Speaker’s Intention: A Pragmatically Based Account of Mood in Spanish.” Hispania 77.4 (Dec): 892-902. Mellet, Sylvie. 1995. “Quando, Quia, Quod, Quoniam. Analyse énonciative et syntaxique des conjonctions de cause en latin.” Études de syntaxe latine offertes à Marius Lavency. Ed. Dominique Longrée. Louvain-La-Neuve, Belgium: Peeters, pp. 211-228. Moreno Sánchez, Ana Isabel. 1995. “Sobre el empleo de quod, quia y quoniam con los verbos de lengua y entendimiento en las cartas de Pablo de la Vulgata.” Analecta Malacitana 18.1: 75-90. Mountford, Sir James, ed. 1984. Bradley’s Arnold Latin Prose Composition. London: British Classical Press. Niño-Murcia, Mercedes. 1992. “El futuro sintético en el español norandino. Caso de mandato atenuado.” Hispania 75.3: 705-713. Nowikow, Wiaczeslaw. 1995. “El concepto de valoración en el subjuntivo español.” Moenia 1: 203-217. Palmer, F. R. 1986. Mood and Modality. Cambridge: Cambridge University Press. Parker, Frank. 1980. “Typology and Word Order Change.” Linguistics 18: 269288. 301
Perrochat, Paul. 1926. “Sur un principe d’ordre des mots. La place du verbe dans la subordonnée.” Revue des études latines 4: 50-60. Pinkster, Harm. 1990. Latin Syntax and Semantics. London and New York: Routledge. Pons Bordería, Salvador. 2003. “Que inicial átono como marca de modalidad.” Estudios de lingüística 17: 531-545. Porroche Ballesteros, Margarita. 2000. “Algunos aspectos del uso de que en el español conversacional.” Círculo de lingüística aplicada a la comunicación 3: http://www.ucm.es/info/circulo/no3/porroche.htm. Porto Dapena. 1991. Del indicativo al subjuntivo. Madrid: Arco. Pulgram, Ernst. 1963. “Synthetic and Analytic Morphological Constructs.” Festschrift Alwin Kuhn zum 60. Geburtstag. Weltoffene Romanistik. Eds. Guntram Plangg and Eberhard Tiefenthaler. Innsbruck: Sprachwissenschaftliches Institut der Leopold-Franzens-Universität Innsbruck, pp. 35-42. Quer, Joseph. 2001. “Interpreting Mood.” Probus 13.1: 81-111. Raible, Wolfgang. 1992. “The Pitfalls of Subordination. Subject and Object Clauses between Latin and Romance.” Historical Philology. Greek, Latin, and Romance. Papers in Honor of Oswald Szemerenyi II. Eds. Bela Brogyanyi and Reiner Lipp. Amsterdam: John Benjamins, pp. 299-337. Ramat, Anna Giacalone. 1998. “Testing the Boundaries of Grammaticalization.” The Limits of Grammaticalization. Eds. Anna Giacalone Ramat and Paul J. Hopper. Amsterdam: John Benjamins, pp. 107-127. Ridruejo, Emilio. 1975. “Cantaría por cantara, en La Rioja.” Berceo 89: 123134. Ritchie, R. L. Graeme. 1907. Recherches sur la syntaxe de la conjonction ‘que’ dans l’ancien français depuis les origines de la langue jusqu’au commencement du XIIIe siècle. Paris: Champion. Roberts, Ian and Anna Roussou. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. 302
Sabanééva, Margarita. 1993. “Verbal Modality and Connected Problems. Attempt at a New Formula.” La linguistique 29.1: 55-65. Sánchez, Rosaura. 1972. “Nuestra circunstancia lingüística.” El grito 6.1: 45-74. Sapir, Edward. 1949 [1921]. Language. An Introduction to the Study of Speech. New York: Harcourt, Brace & World, Inc. Schlegel, August Wilhelm. 1971 [1818]. Observations sur la langue et la littérature provençales. Tübingen: Tübinger Beiträge zur Linguistik. Schlieben-Lange, Brigitte. 1992. “The History of Subordinating Conjunctions in some Romance Languages.” Internal and External Factors in Syntactic Change. Eds. Marinel Gerritsen and Dieter Stein. Berlin: Mouton, pp. 341-354. Schwegler, Armin. 1990. Analyticity and Syntheticity. A Diachronic Perspective with Special Reference to Romance Languages. Berlin: Mouton. Shawl, James R. 1975. “Syntactic Aspects of the Spanish Subjunctive.” Hispania 58 (May): 323-329. Silva-Corvalán, Carmen. 1994. “The Gradual Loss of Mood Distinctions in Los Angeles Spanish.” Language Variation and Change 6: 255-272. Slavíčková, Eleonora. 1968. “Towards a Typological Evolution of Related Languages.” Travaux linguistiques de Prague, 3. University, AL: University of Alabama Press. Solé, Yolanda. 1977. “Continuidad/descontinuidad idiomática en el español tejano.” The Bilingual Review 4.3: 189-199. Sonnenschein, Edward A. 1910. The Unity of the Latin Subjunctive. A Quest. London: John Murray. Soutet, Olivier. 2000. Le subjonctif en français. Paris: Ophrys. Studerus, Lenard H. 1981. “A Spanish Twilight Zone. Mood, Syntax, and Past Temporal Reference.” Hispania 64.1 (Mar): 97-103.
303
----. 1995. “Some Unresolved Issues in Spanish Mood Use.” Hispania 78.1 (Mar): 94-104. Szemerényi, Oswald J. L. 1996. Introduction to Indo-European Linguistics. Oxford: Oxford University Press. Tauli, Valter. 1945. “Morphological Analysis and Synthesis.” Acta Linguistica (Hafniensia) 5.2: 80-85. Terrell, Tracy and Joan B. Hooper. 1974. “A semantically based analysis of mood in Spanish.” Hispania 57: 484-494. Tesnière, Lucien. 1932. “Synthétisme et analytisme.” Charisteria Guilelmo Mathesio Quinquagenario. A Discipulis et Circuli Linguistici Pragensis Sodalibus Oblata. Prague: Cercle Linguistique de Prague, pp. 62-64. Togeby, Knud. 1980. “Romance Historical Morphology.” Trends in Romance Linguistics and Philology. Volume 1: Romance Comparative and Historical Linguistics. Eds. Rebecca Posner and John N. Green. The Hague, Paris and New York: Mouton, pp. 105-155. Touratier, Christian. 1977. “Valeur et fonctionnement du subjonctif latin.” Revue des études latines. 55: 370-406. Van Gelderen, Elly. 2004. Grammaticalization as Economy. Amsterdam: John Benjamins. Van Vliet, Edward R. 1983. “The Disappearance of the French passé simple. A Morphological and Sociolinguistic Study.” Word 34.2: 89-113. Vennemann, Theo. 1975. “An Explanation of Drift.” Word Order and Word Order Change. Ed. Charles N. Li. Austin: Texas University Press, pp. 269-305. Villalta, Elisabeth. 2000. “A Comparative Semantics for the Subjunctive Mood in Spanish.” Romance Syntax, Semantics and L2 Acquisition. Selected Papers from the 30th Linguistic Symposium on Romance Languages. Eds. Joaquim Camps and Caroline R. Wiltshire. Amsterdam: John Benjamins Publishing Company, pp. 227-242.
304
Watkins, Calvert. 1964. “Preliminaries to the Reconstruction of Indo-European Sentence Structure.” Proceedings of the 9th International Congress of Linguists. Ed. Horace G. Lunt. The Hague: Mouton, pp. 1035-1042. ----.. 2000. The American Heritage Dictionary of Indo-European Roots. Boston and New York: Houghton Mifflin. Wheelock, Frederic M. 2000. Wheelock’s Latin. 6th ed. New York: HarperCollins.
305
VITA
Melissa Dae Murphy was born in Brooklyn, New York, the daughter of John William Murphy and Odette Denise Bogart. She majored in Spanish and French at Florida Atlantic University and, during her years as an undergraduate student, she was employed at Broward Community College as a Spanish, French and ESL tutor. In 1997, she received her B.A. and then began the graduate program in Spanish Literature at Florida International University, where she also worked as a Graduate Assistant and Teaching Assistant. Upon receiving her M.A. in 2000, she entered the Ph.D. program in Romance Linguistics at the University of Texas, where she has twice received the “Diplôme d’honneur” from the Department of French and Italian and has taught various language courses in the Department of Spanish & Portuguese.
Permanent Address: 10320 Boulder Lane #1724, Austin, Texas 78726 This dissertation was typed by the author.
306
