227 12 6MB
English Pages 409 Year 2007
Prepositional Infinitives in Romance A usage-based approach to syntactic change
Kim Schulte
2007 Peter Lang
Prepositional Infinitives in Romance A usage-based approach to syntactic change
Kim Schulte
2007 Peter Lang
Contents
Acknowledgements
7
List of gloss abbreviations
9
Preface
11
Brief résumé of chapters
13
Chapter 1 1.1 1.2 1.3 1.4 1.5
Motivation and theoretical parameters Why Romance? The choice of languages examined Theoretical parameters Pragmatic causation in syntax Statistical analysis: reasons and methods
17 17 19 26 48 55
Chapter 2 2.1 2.2 2.3
From Latin to Romance Finite subordination In nitival subordination Non-in nitival nominal dependent clauses in Latin The fate of Latin nominal dependent structures in Romance Conclusion
69 70 79
2.4 2.5 Chapter 3 3.1 3.2 3.3 3.4 Chapter 4 4.1
The in nitive in Spanish and Portuguese today Non-overt subjects of in nitival clauses In nitives with overt subject in Spanish Overt subject marking in Portuguese in nitival clauses The in nitive: nominal or verbal? Diachronic development of the in nitive in Spanish Prepositional and non-prepositional in nitives
87 99 118 121 121 150 170 190 209 209
Contents
6
4.2 4.3 4.4 Chapter 5 5.1 5.2 Chapter 6 6.1 6.2 6.3 6.4 6.5 Chapter 7 7.1 7.2 7.3 7.4 7.5
Prepositional in nitives In nitival clauses and their nite counterparts Discussion of the Spanish diachronic data
212 230 240
Portuguese and Spanish developments compared 253 Similarities and differences in diachronic development 253 OSIs and the in ected in nitive 263 The in nitive in Romanian The present-day situation Synchronic use of prepositional in nitives Diachronic development of the prepositional in nitive Pragmatic causation Conclusion Relevance and reanalysis: prepositional complementizers The emergence of prepositional complementizers Subsequent development of prepositional complementizers Diachronic statistical development of de and a Ongoing grammaticalization of para/pra in Portuguese Conclusion
283 283 303 308 317 327 329 329 334 339 342 344
Conclusion
347
References
351
Appendix A
Spanish texts
363
Appendix B
Portuguese texts
371
Appendix C
Romanian texts
383
Index
399
Acknowledgements
This book is a revised version of my doctoral thesis, submitted to the University of Cambridge in 2004. I am indebted to my supervisor, Chris Pountain, for his encouragement and guidance, but especially for sharing his genuine enthusiasm for linguistic observation with me. I am also grateful to Teresa de Carlos, Rosalina Gulão, Martin Maiden, Peter Matthews, April McMahon, and Francis Nolan for having made my university education a pleasant and interesting experience, to Monika Palm-Coenen, my rst Spanish teacher, and to Dominic Church, Guy Deutscher, Paul Heggarty, Luisa Miceli, Marcus Meyer, Urs Schulte, Alexandra Stavinschi, and Janick Wrona for their personal support and academic . Special thanks to Kai for his indispensable assistance in all computational matters, to Marcus Brüggen for his help with SympaTeX, to Claudio Canaparo for his practical advice and support, and of course to my mother and father. Finally, I also wish to express my gratitude to Trinity College for the vast amount of money ungrudgingly spent on me. Virtus vera nobilitas.
List of gloss abbreviations
Grammatical information in glosses is only provided to the extent that it is helpful to clarify examples.
1 2 3 ABL ACC ACT ADJ AUX CND COMP CONJ DAT DEF DEM DIM DO DOMKR FUT GEN GER GRNV IMP IMPF IND INF INVAR IO NMLZ
1st person 2nd person 3rd person ablative accusative active adjective auxiliary conditional complementizer conjunction dative de nite demonstrative diminutive direct object direct object marker future genitive gerund gerundive imperative imperfect indicative in nitive invariant indirect object nominalization
NOBL NOM NP OBJ OBL PASS PERF PL PLUP PO POL POSS PRED PRET PROG PRON PRPN PRS PST PTCP REFL REL SBJ SBJV SG STAT SUP UNINFL
non-oblique nominative noun phrase object oblique passive perfect plural pluperfect prepositional object polite possessive predicative preterite progressive pronoun proper name present past participle re exive relative marker subject subjunctive singular stative supine unin ected
Preface
While much of modern syntactic analysis is primarily concerned with systematizing grammatical structures into a uni ed framework, this book investigates syntactic structure from a different angle: its aim is to establish the reasons why particular constructions are used, rather than attempting to explain how they can be tted into a speci c theoretical system. Adopting an approach that combines cognitive principles with a largely functionally oriented, construction-based view of grammar, this book offers a detailed comparative study of factors involved in the synchronic choice and diachronic development of a group of constructions in three Romance languages: Spanish, Portuguese and Romanian. The objectives are to identify what it is that causes speakers to use these constructions, to what extent and why their use has become obligatory under certain conditions but remains optional under others, and which diachronic mechanisms have led to their present-day distribution. None of these questions can be answered without looking beyond the syntactic domain. It must be kept in mind that the fundamental purpose of using language is to communicate, and syntactic structures are nothing more than tools to achieve our communicative objectives. Which tool we choose in order to achieve these objectives depends on the exact nature of the communicative situation; the linguistic and extra-linguistic context of an utterance is of crucial importance in understanding why a speaker chooses a particular construction. Pragmatic factors such as relevance, shared knowledge between interlocutors, avoidance of ambiguity or potential misunderstandings, and the principle of economy all interact with the semantic content of the utterance, in uencing the syntactic choices made by the speaker. The same pragmatic factors are also diachronically relevant, as yesterday's syntactic choices are today's syntactic rules: certain pragmatic conditions apply more frequently than others, and as a result the syntactic structure typically chosen under such conditions occurs more frequently than its alternative(s). High usage frequency, in turn, facili-
12
Preface
tates the cognitive entrenchment of the respective construction, causing speakers to use it rst as a default choice, and eventually in an increasingly obligatory way. The impact of usage frequency on the evolution of the structures examined in this book is investigated by means of extensive frequency analyses based on diachronic text corpora, and a detailed comparison of the data obtained from separate analyses of the three languages makes it possible to draw some general conclusions about the observed mechanisms of syntactic change. The data presented in this book lends further support to theories that see usage and cognition at the centre of structural change in language.
Brief résumé of chapters
Chapter 1: Motivation and theoretical parameters The rst chapter begins by explaining why the Romance languages are a valuable eld of study in Historical Linguistics, and why the speci c languages chosen for this study were picked. This is followed by a section discussing some fundamental theoretical parameters and de nitions of terms. The nal two sections illuminate how pragmatics can interact with syntactic change, and how diachronic statistical analysis can reveal such processes, leading on to a discussion of the sources used to provide the data, and of the principles and procedures applied. Chapter 2: From Latin to Romance Chapter 2 provides a survey of the different types of nite and non- nite dependent clauses in Latin, and to what extent they have survived into Romance. Special attention is paid to the link between the syntactic construction chosen for a dependent clause and the semantic as well as pragmatic implications of this choice. In particular, the mechanism of semantic underspeci cation is found to be exploited for the purpose of pragmatic backgrounding. Chapter 3: The in nitive in Spanish and Portuguese today Chapter 3 consists of four main sections. The rst section examines the ways in which a subject is assigned to in nitival clauses. After a presentation of a number of existing syntactic and semantic models, a decision tree model integrating syntactic and pragmatic factors into a single system is presented and exempli ed. It is then discussed how the
14
Brief résumé of chapters
pragmatically most likely pattern, coreferentiality, becomes entrenched as the default pattern of subject assignment, and how the existence of such established patterns are exploited semantically. The second and third sections examine in nitival clauses with overt subjects in present-day Spanish and Portuguese, respectively. Several existing analyses regarding this area of syntax are presented, as well as statistical analyses that suggest that the primary function of these constructions is pragmatic, though a different one in each of the two languages. The nal section discusses the question whether it can be said that the in nitive has become more nominal or more verbal today than it was in Latin. An approach placing the in nitive along a continuum between the two extremes is discussed, and an alternative two-dimensional classi catory model is proposed. This model, which looks at the internal and the external syntax of in nitival clauses as two separate parameters, is then applied to the in nitive in Latin, modern Spanish, and medieval Spanish. Finally, a possible link between the demise of morphological case and an increase in the use of in nitives is discussed. Chapter 4: Diachronic development of the in nitive in Spanish Chapter 4 consists mainly of a statistical analysis of the evolution of the prepositional in nitive in Spanish since the early Middle Ages. In the rst section, usage frequencies of the prepositional in nitive are presented, sorted both by individual preposition and by semantic criteria. This is followed by a statistical comparison of nite and in nitival dependent clauses through time, and by a discussion of the results and possible reasons for the developments observed. Chapter 5: Portuguese and Spanish developments compared The rst part of Chapter 5 compares the diachronic statistical development of a number of individual prepositional in nitives, as well as of this group of constructions as a whole, in Spanish and Portuguese, pointing out similarities and differences in the chronology of developments
Brief résumé of chapters
15
in the two languages. In the light of this data, the likelihood of the previously suggested pragmatic mechanisms being the driving force behind the spread of the prepositional in nitive in both languages is then reassessed and con rmed. The second part of the chapter consists of a detailed diachronic analysis of in nitival clauses with overt subjects. A separate diachronic analysis of these constructions for Spanish and Portuguese is followed by a comparison of those developments, in which it is concluded that their origin is largely similar, but the subsequent development has led to their differing present-day usage patterns. Chapter 6: The in nitive in Romanian Chapter 6 begins with a discussion of the status of the in nitive in Romanian, and of the claim that it plays only a marginal role in modern Romanian syntax. This is followed by a survey comparing the use of the in nitive in Romanian with that in the other Romance languages. Attention then focuses on the prepositional in nitive, both on its synchronic distribution and on its diachronic development. The nal section analyses and evaluates the data and compares the ndings with the corresponding results for Spanish and Portuguese in the previous chapters. The similarities that can be observed are shown to be the result of more frequent pragmatic usage patterns becoming entrenched at the expense of less frequent ones. Chapter 7: Relevance and reanalysis: prepositional complementizers Chapter 7 traces the prepositional complementizers a and de, which often have a purely functional role in modern Romance, back to their origin as a meaningful part of adverbial clauses via a progressive process of reanalysis and semantic bleaching, and offers a tentative explanation for the different results of this process in Ibero- and Daco-Romance. This is supported by the presentation of a similar, ongoing process, the incipient reanalysis or grammaticalization of Portuguese para/pra, which may currently be on a path towards becoming a complementizer.
16
Brief résumé of chapters
Conclusion In the nal chapter, the differences and similarities between the individual languages, presented in the previous chapters, are summarized. The results lead to the conclusion that the pragmatics of language usage have played an important role in the development of the prepositional in nitive, regarding the way it has spread as well as the functional shifts it has experienced through time. Though there are certain differences in the way prepositional in nitives are used in the three languages, it is observed that the pragmatic mechanisms involved in their development are in many ways the same. This raises the question whether we may be dealing with universally valid pragmatic principles. To determine whether this is the case, a more extensive cross-linguistic analysis going beyond the Romance domain is called for.
Chapter 1 Motivation and theoretical parameters
1.1
Why Romance?
This book is a study of a speci c area in the syntax of one branch1 of a particular language family, but many of the observations and conclusions drawn are not necessarily limited to the Romance languages, as they illuminate more general mechanisms in the syntactic and semantic evolution of languages. As Jaeggli (1986: ix) points out, the Romance languages have [in recent years] been investigated in depth to gain insights into issues concerning abstract formal structures, matters of linguistic variation [...] and theories concerning language change , and their study has had an impact on the development of ideas in linguistics. The Romance languages traditionally have a privileged position among the languages subjected to linguistic analysis. This is partly due to the fact that, for cultural reasons, many linguists are well acquainted with certain Romance varieties, particularly French, and with their historical source language, Latin. However, there are also more scienti c reasons that make Romance a particularly rewarding area of research, especially for the linguist with diachronic interests. First, this branch of Indo-Europaen is among the best documented of the world's languages, with the earliest Latin texts dating back to well before the third century B.C., and we have a rich and varied range of documents from many geographically diverse Latinized areas. This does not mean that the available documents provide information about all stages or registers of the language or languages spoken by
1
The Romance languages, a branch of the Indo-Europaen language family, can of course be subdivided into further sub-branches.
18
Chapter 1
the population. In particular, there is a large gap in our knowledge of the period between the rst century A.D. and the appearance of the rst vernacular texts. Though written Latin from every century survives on manuscript (Wright, 1982: ix), they do not represent the precursors of what became the Romance Languages, but a continuation of an older linguistic norm. The exact linguistic situation during this period is subject to a great deal of controversy; for instance, it is not clear when speakers began consciously to distinguish between their vernacular and the formal Latin used in of cial contexts. Indeed, there is evidence that even during the `Classical Latin' period, an evolved form of popular Latin was already in use, either only in some less educated classes of the population, or perhaps in everyone's spoken language. The situation is further complicated by the fact that regional features would have started to develop even during the time of the Roman empire, and there is considerable disagreement among scholars as to whether there was widespread Latin/vernacular bilingualism, and what the exact relationship between Late Latin and Early Romance was. In his article on the highly variable use of the term `Vulgar Latin', Lloyd (1979) presents a wide range of theories on the exact nature of the language linking Classical Latin and the Romance languages; Wright (1982) examines the period between the end of the Roman Empire and the appearance of the rst vernacular documents, concluding that speakers did not, before that point, understand the variety of Latin they spoke as anything other than Latin. That the spoken language of the day differed increasingly from Classical Latin is documented by the fact that the Council of Tours in 813 refers to the common spoken language of the day as RUSTICA ROMANA LINGUA, which is understood to have a linguistic status similar to that of THEOTISCA LINGUA, i.e. German. By the time of the rst preserved documents in local vernacular, i.e. in the 9th century in France2 , in the 12th century in Iberia, and as late as the 16th century in Romania, we are clearly dealing with languages that are different both from Latin and from each other. Though some information about
2
The Serments de Strasbourg (842 A.D.) are the earliest known Romance vernacular text.
Motivation and theoretical parameters
19
the earlier stages of Romance is provided by regularly occurring `mistakes' and glosses in Latin texts, many details of the linguistic evolution during this period can only be reconstructed. But whilst most linguistic reconstruction arrives at a hypothetical protolanguage, the Romance languages offer the advantage that the Classical Latin documents provide ample information where a reconstruction must ultimately lead; its accuracy can thus easily be veri ed. For instance, the absence in Classical Latin of the construction examined in this study, the `prepositional in nitive', means that its emergence must be reconstructed as part of the interim linguistic stages. A further reason why the Romance languages are of particular value in Historical Linguistics is the fact that the documented development of several sister languages can be compared over a period of several centuries between the Middle Ages and today. Beyond the mere observation that branching development has led to the numerous differences between the modern Romance varieties, this allows us to trace back some of these developments to their point of origin. This may allow us to pinpoint the factors involved in causing a change to occur in one variety but not in another, and to identify which conditions may be responsible for similar but independent developments in more than one variety. Such ndings will add to our understanding of linguistic change and its causes on a more general level, and be of assistance in the reconstruction of less well-documented language families.
1.2
The choice of languages examined
A detailed analysis of the evolution of the syntactic structure under investigation cannot, here, be undertaken for all Romance languages and varieties, even if it was possible to de ne a series of discrete varieties. This study will focus mainly on Spanish, Portuguese, and Romanian; the differences and parallels observable between these sample languages can provide particularly valuable insights into the processes involved, for the following reasons.
20
Chapter 1
1.2.1 Spanish and Portuguese: internally motivated divergence According to one of the traditional classi catory systems of the Romance languages, Spanish and Portuguese are representatives of Western Romance3 , and geographically located in the extreme West of the Romance-speaking area in Europe. Not only are the territories in which these two languages are spoken adjacent, but also largely isolated from non-Romance languages from a geographic point of view. Two pre-Latin substrate languages of Iberia, Basque and Ibero-Celtic, have left some phonological, morphological and lexical traces, but no substrate features are known to have entered the Romance syntactic system (Lapesa, 1980: 36 52), so syntactic differences between Spanish and Portuguese cannot be attributed to different substrate in uence. It is unlikely that the varieties (or mixtures of different varieties) spoken by the Roman colonizers varied greatly or in any systematic way between the areas that now constitute Spain and Portugal. Nor did the subsequent impact of linguistic contact with non-Romance languages, in particular with Arabic, have a signi cant differentiating effect across the Iberian Peninsula. On the one hand, the territories of modern-day Spain and Portugal were not separate countries under Moorish rule, and thus subjected to broadly the same cultural in uence. On the other hand, though the Moors invaded Iberia in 711 A.D., had occupied most of it by 718 A.D., and remained a powerful cultural in uence until the end of the 15th century, the linguistic impact of their presence did not go much beyond lexical borrowing. Galmés de Fuentes (1996) identi es a number of syntactic and stylistic Arabisms in medieval Castilian prose, many of which are, however, merely an increased use of structures already present in Old Spanish, such as an increased use of in nitives as true verbal nouns (ibid., 174 179). Some Arabisms, such as para-
3
This classi cation is originally put forward by Wartburg (1950) based on the merger of the Latin vowels I/E and U/O in `Western Romance'. The validity of a genetic classi cation based on a single vowel merger is dubious and much disputed; numerous alternative classi cations based on lexical and syntactic features have been proposed, a number of which are listed in Posner (1996: 196 202). Nevertheless, the similarity between Spanish and Portuguese in comparison with other Romance varieties is largely undisputed.
Motivation and theoretical parameters
21
nomastic constructions of the type burla burlando, he says, entered the popular language and resurfaced in the popular literature of the Golden Age, but los arabismos sintácticos, que de la lengua literaria del XIII no habían pasadao a la coloquial, desaparecen de nitivamente, por lo general, en los siglos de presión latinizante. (Galmés, 1996: 233) Most historical accounts of Spanish and Portuguese (e.g. Green, 1988; Parkinson, 1988) do not note any signi cant lasting syntactic in uence of Arabic, whilst Lapesa (1980: 151 152) mentions that verb- rst word order is more common in Spanish and Portuguese than in other Romance varieties, which may be attributable to Arabic or Hebrew in uence (Crabb, 1955). With a common ancestor language and common external in uence, Spanish and Portuguese have nevertheless developed in markedly different ways in some areas, most obviously perhaps in their phonology, but also in certain areas of morpho-syntax. One of these areas, the differing use of the in nitive, is the topic of this study. As the external conditions and in uences do not differ signi cantly, the motivation for their diverging development must barring random or spontaneous changes be differing patterns of usage in the different speech communities. If, for instance, a structure is used in a potentially ambiguous way more frequently by the members of one of the communities, this will increase the likelihood of its reanalysis, as illustrated in the following example of ambiguity between pluperfect indicative and imperfect subjunctive: Spanish (1)
Me informó
que tal.vez lo hiciera
el día anterior.
me inform.3sg.pret that perhaps it do.3sg.pst.sbjv the day previous.
He informed me that perhaps he did it the previous day. Portuguese (2)
Informou-me
que o zera
talvez no
dia anterior.
inform.3sg.pret-me that it do.3sg.plup.ind perhaps in.the day previous
He informed me that perhaps he had done it the previous day.
22
Chapter 1
In the Spanish example, hiciera is past subjunctive; in the Portuguese sentence, the cognate zera is pluperfect indicative. Both sentences are, nevertheless, grammatically correct and semantically virtually equivalent. Until the end of the Middle Ages, Spanish, like Portuguese to the present day, preserved the pluperfect indicative function of the -ra-paradigm inherited from Latin (FECERAT)4 , but it subsequently came to be used as the imperfect subjunctive.5 There is no compelling structural need for such a shift; on the contrary, it leaves Spanish without a synthetic pluperfect, while the -ra-form takes on a function for which the (largely)6 synonymous -se-form is already available. So what might the motivation for such a shift be? A facilitating factor is de nitely the cross-linguistically attested conceptual link between temporal remoteness and non-assertion/counterfactuality, as found, for instance, in the English backshift of tenses in counterfactual conditions. But if this were the only factor, and if we accept that the source language and the external in uences were the same for both languages, we should expect the same development in Portuguese. The decisive difference lies not in the structure itself, but in its actual usage by members of the two speech communities. The examples above illustrate the potential structural ambiguity between a pluperfect and a past subjunctive, but give no indication how likely it is that such ambiguity will actually arise in real communicative contexts. Occasional ambiguity is unlikely to have a structural effect, but if structure A (the pluperfect) can equally be understood to be structure B (the imperfect subjunctive) in a large number of pragmatic contexts, this can lead language users to perceive a lack of functional opposition, which in turn facilitates full reanalysis. What this implies is that structural differences between Spanish and Portuguese are the consequence of differences in actual language usage, which consists of individual utterances. Each utterance is itself
4 5 6
Penny (1991: 171) Note that the -ra-form retains numerous non-subjunctival uses in different varieties of Spanish; for a full account, see Hermerén (1992). Martinell Gifre (1985) argues that the degree of subjunctivity of the -ra-form is weaker than that of the -se-form.
Motivation and theoretical parameters
23
the result of an individual cognitive process by a speaker, in the course of which he can make certain choices. Such a choice can be arbitrary, but may frequently be determined by social and cultural factors, such as whether the speaker feels that one construction is situationally more appropriate in terms of style, register, expressive force, etc. than the other. These social conventions will unavoidably vary between different linguistic communities, and therefore so will the overall usage frequency. But the more often a certain construction is used in a language, the more likely it is to become entrenched as the typical or default construction, which in the case of originally ambiguous constructions results in reanalysis and syntactic change. The fact that usage-based pragmatic differences are likely to be the main factor responsible for differential developments in Spanish and Portuguese makes the two languages a valuable object of comparative linguistic research, since a wider range of potential causes would unavoidably obscure the role played by each one. 1.2.2 Romanian: evolution in isolation The situation of the third language examined, Romanian7 , stands in marked contrast to that of Spanish and Portuguese described above. In the traditional division of the Romance-speaking world into Eastern and Western Romance8 , Romanian falls into the Eastern Romance category, but it differs particularly strongly from other Romance languages (including Eastern Romance varieties) due to its geographical and long-lasting cultural isolation from the rest of the Romance-speaking area. There are two principle sources for non-Romance features in Romanian. On the one hand, Roman rule in the Dacia province was limited to a period of approximately 150 years, a fact that may be credited for the comparatively high degree of substrate in uence from the local pre-Ro-
7 8
I will be primarily concerned with Daco-Romanian. Any reference to a different Daco-Romance variety will be pointed out as such. cf. Section 1.2, Footnote 3.
24
Chapter 1
man Thracian, Illyrian, Getian and Dacian populations. Substrate in uence is understood to be responsible not only for a considerable number of lexical items (cf. Poghirc, 1969: 327 356; Russu, 1981), but also for phonological, morphological and syntactic features shared with Albanian, such as the morpheme //-ne// on the disjunctive pronouns mine, tine, sine, and the `possessive article' (cf. du Nay, 1996: 64 69). On the other hand, Romanian has been strongly in uenced by areal convergence within the Balkan Sprachbund. Sprachbünde or linguistic convergence areas, as rst identi ed by Trubetzkoy in 1928 (Trubetzkoy, 1930: 17 18)9 , contain languages belonging to more than one family but showing traits in common which are found not to belong to the other members of (at least) one of the families (Emeneau, 1956: 16). The Balkans are one of the most frequently studied Sprachbünde, in which varying degrees of convergence between Greek, Albanian, Slavic, Romance, Turkish and Hungarian have been identi ed. Under these circumstances, it comes as no surprise that many structural changes, among which the area of clausal subordination has acquired a prominent position, have been attributed either to straightforward borrowing or to linguistic convergence (e.g. Joseph, 1983). This has, in fact, been such a dominant theory that scholars like Bari¢ (1961) often feel the need to justify any theory that proposes independent development and language-internally motivated changes. It can, however, be argued that the long period of isolation from other Romance varieties, especially in its formative phase during which an absence of linguistic norms and standards prevailed in the speech of the people, makes Romanian a particularly valuable object of comparative Romance philology. Though in uence from the extremely varied surrounding linguistic area presents us with a wide range of potential factors contributing to structural changes, a central concept in the Sprachbund theory is the idea of mutual in uence and joint parallel development, rather than unidirectional borrowing. This implies that there
9
In fact, Trubetzkoy had previously already coined the Russian term jazykovoj sojuz in 1923, but as the publication was concerned mainly with theological rather than linguistic issues, linguists adopted the corresponding German term, rst used by him in 1928.
Motivation and theoretical parameters
25
is some likelihood of structural shifts being internally motivated, even if they occur in a similar fashion in surrounding areas. If it can be shown that comparable internally motivated changes have taken place in Ibero- and Daco-Romance, particularly if the resulting structures in Romanian are not Sprachbund features shared with other Balkan languages, this would suggest that linguistic drift may be involved. This could either be the type of drift described by Sapir (1921: 150 ff.), in which the linguistic structure inherited from an earlier stage of a language, in this case Latin, favours certain types of change. Alternatively, we may be dealing with a more universal or natural drift, explained by Aitchison (1991: 133 134) as the indirect result of universal mental tendencies that cause similar patterns of change in different, not necessarily related languages. Romanian can thus provide linguistic insights not available from any of the varieties in the dialect continuum between Apulia and the Algarve, in which cross-dialectal in uence and borrowing is much more dif cult to disprove10 . One point to keep in mind in any study comparing Romanian and other Romance varieties is the fact that certain strata of Romanian society were subjected to intense contact with French, and to a lesser degree with Italian, in the 19th and the rst half of the 20th century. Apart from vast amounts of vocabulary, this has allowed some French syntactic patterns to enter the language. In literary register this is indisputable11 , but as the use of French was essentially limited to the educated classes, it is doubtful how much genuine syntactic borrowing into the spoken language of the people took place. It appears more likely that independently evolved cognate structures were already present, and that French in uence merely reinforced or modi ed their usage. In fact, Sørensen (1957: 133) even goes so far as to claim that it is a necessary condition that there should exist in the receiving language certain innate tendencies and possibilities with which the foreign idiom does not
10
11
This is not to say that the same feature cannot arise independently in separate dialects of the continuum. The in ected in nitive in Portuguese and Old Neapolitan (cf. Ledgeway, 2000: 109 14) is just one example. An example of this is the in uence of French on the resurgence of the passive with a , (Mallinson, 1988: 418)
26
Chapter 1
clash .12 It is, however, necessary to take this `re-Romancing tendency' (Mallinson, 1988: 418) into account to avoid premature conclusions about parallel developments in the two languages.
1.3
Theoretical parameters
The approach chosen for this study is a cognitive, construction-based one. Though reference to work conducted within other frameworks will be made at various points, the object and methodology of the present study differs fundamentally from that of many formalist approaches in that its primary aim is not to provide an abstract description of linguistic structures beyond the observable level. Instead, it attempts to provide explanations for syntactic changes by focusing on the effects of observable usage patterns, usage frequency and the role of the language user. 1.3.1 Categories and constructions Whilst grammarians and linguists traditionally attempt to subdivide language into discrete categories, Croft (2004) convincingly proves that uniform syntactic categories can neither be established among different languages, nor within a single language. Syntactic categories are generally understood to consist of elements that have the same syntactic distribution; membership of a particular category can be established by verifying whether or not an element does or does not participate in the same range of constructions or transformations as its fellow category members. However, Croft shows that membership of classes established by this method varies consider-
12
A similar observation is made by Pountain (1994: 121; 1999: 36 38), stating that syntactic in uence from English does not lead to signi cant structural innovation in Spanish, but rather encourages the fuller and more effective use of similar structures already present.
Motivation and theoretical parameters
27
ably, depending on which distributional tests one chooses to consider diagnostic of class membership. A simple Spanish example of this is the distinction between the classes `direct object' and `oblique object'. One distributional test to establish which of these two classes a constituent belongs to is whether or not it is linked to the verb by a preposition. According to this criterion, `el gato' in (3a) and `cien kilos' in (5a) are direct objects, whereas el profesor in (4a) is not. (3)
a.
Pedro acaricia el gato. Peter
b.
strokes
El gato es acariciado por Pedro. The cat
(4)
a.
the cat.
is stroked
by Peter.
Pedro habla con el profesor. Peter
talks
to
the teacher.
b. *El profesor es hablado The teacher
(5)
a.
is talked
por Pedro.13 (to) by Peter.
Pedro pesa cien kilos. Peter
weighs 100 kg.
b. *Cien kilos son pesados por Pedro. 100
kg
are weighed by Peter.
A second test that might be applied to determine membership of the class `direct object' is passivization: direct objects can appear as the subject of a corresponding passive sentence, whereas oblique ones cannot. According to this criterion, `el gato' in (3b) is again classi ed as direct object, and `el profesor' in (4b) once again does not fall into this class. But `cien kilos' in (4b), which the previous test identi ed as a direct object, is not a direct object according to the second test, as it cannot appear as the subject of the passivized sentence. In other words, whether or not something can be classi ed as a direct object depends on the distributional tests we choose to apply; different tests provide con icting classi cations.
13
This sentence becomes no more grammatical in Spanish if the preposition con is incorporated.
28
Chapter 1
One way of dealing with this dif culty is simply to pick one distributional test as the authoritative one. For instance, one might decide that the absence or presence of a preposition is the determining factor for classi cation, and that `cien kilos' is thus a direct object, but that some exceptional feature prohibits it from participating in the passive construction. However, this would arbitrarily give primacy to one distributional test over another: had we decided that participation in the passive construction is the determining factor for classi cation, `cien kilos' would not be considered a direct object. As there is no good reason why one of the distributional patterns should be given primacy over the other, Croft concludes that syntactic categories are construction-speci c, rather than being valid for a language as a whole. According to this approach, (3-5) can be said to represent three different constructions: (I) one that permits a prepositionless argument and also permits this argument to appear as the subject in the corresponding passive sentence (3), (II) one that permits a prepositionless argument but does not permit it to appear as the subject in the corresponding passive sentence (5), and (III) one that permits neither a prepositionless argument, nor its appearance as as the subject in the corresponding passive sentence (4). Each of these three constructions is restricted in terms of the range of verbs14 that can participate in it. Verbs such as acariciar `to stroke', ver `to see', leer `to read' can participate in (I), as can pesar `to weigh' (e.g. `Pedro pesa la fruta.'/`La fruta es pesada por Pedro.' (Peter weighs the fruit./The fruit is weighed by Peter.)) In construction (II), a wide range of verbs can participate, including most transitive verbs. The range of verbs that can participate in (III) is somewhat smaller; it includes pesar `to weigh', medir `to measure' and costar `to cost'. It should be noted that the meaning of each construction goes beyond the sum of its components' meaning; for instance construction (I) with pesar has a somewhat different meaning than construction (III) with pesar. In this section it has been argued that there are no absolute syntactic categories. That is not to say that it is pointless or impossible to
14
The category such as `verb' and `transitive verb' are, of course, themselves construction-speci c. Traditional terminology is used here for the sake of simplicity.
Motivation and theoretical parameters
29
group together certain elements that show certain parallels in their syntactic behaviour across constructions (cf. Section 1.3.6). But rather than de ning membership in a category for the language as a whole on the basis of one randomly chosen distributional test, it is more appropriate to establish classes of elements on the basis of their distribution pattern in a particular construction. 1.3.2 Categories in diachrony As already mentioned in Section 1.2.1, it is largely uncontroversial that most syntactic change is triggered by various types of reanalysis. The principle of reanalysis is such that the semantic contributions of the components in a particular construction, itself conveying a particular meaning, are remapped among these components. Crucially, the construction as a whole does not take on a radically new meaning in this process, nor can any formal changes be observed. In other words, reanalysis is initially only a mental remapping of which components within the construction correspond to which components of its meaning. The reanalysis rst becomes visible in the subsequent process of `actualization', constituted by the use of the construction in novel ways that are compatible with the reanalysed structure, but would be incompatible with the original one. An important point to be made is that change triggered by reanalysis is not an abrupt process that causes a sudden switch from one category to another. Consider the case of the Romance de nite article, derived from the Latin demonstrative ILLE. Its increasing use in Late Latin, in contexts that do not require a demonstrative pronoun, is a clear instance of reanalysis. But rather than suddenly switching from the category `pronoun' to the category `article', it was used in a gradually increasing number of contexts, initially where a demonstrative pronoun might have been appropriate at an earlier stage, then gradually spreading to contexts where Classical Latin would under no circumstances have permitted the demonstrative pronoun. In other words, neither the category `demonstrative pronoun', nor the category `de nite article', as de ned on the basis of modern Romance, would be an appropriate choice for its status during the transitional period.
30
Chapter 1
Other cases of reanalysis show a similar de ance of categorizability. In early French, for instance, the particle pas appears to fall somewhere between the categories of `emphatic particle' and `negative marker' (cf. e.g. Price, 1984: 252; McMahon, 1994: 163). The gradual nature of the increase in the negative force of pas was made possible by the continuing presence of the original negative marker ne, so that varying degrees of negative force could be attributed to each of the elements while the negative force of the complete construction `ne... pas' remained unaffected. Similarly, a+in nitive in the earliest Romanian documents must be situated somewhere on the borderline between nal adjunct and complement (cf. Section 6.3.1, example (314)). What these examples show is that in a diachronic syntactic study of any kind, much care has to be taken with the use of concepts that may be appropriate for synchronic linguistic analysis. Following Croft (2001, 2004), it was argued in Section 1.3.1 that syntactic categories are construction-speci c; this section has provided evidence that, diachronically speaking, constructions themselves are variable entities which are subject to gradual, non-discrete change. A synthesis of these two observations must lead to the conclusion that using a xed set of discrete categories would be methodologically inappropriate for the examination of syntactic change. It would, in fact, defeat the purpose of diachronic linguistic analysis, which is to examine the path along which structures have evolved, if we were to force all evolutionary stages into a single, rigid syntactic framework. Syntactic categories can, however, be a useful descriptive tool if they are not used as a strait-jacket that a language must be forced into at any cost, but are understood merely as cornerstones between which the elements of a language are located and can move around freely; in Section 1.3.6 it will be further discussed in how far the concept of syntactic categories is compatible with the description and analysis of syntactic change. 1.3.3 Dependent clauses In a study that investigates the development of constructions involving clausal subordination, it is important to discuss some different proposals
Motivation and theoretical parameters
31
regarding the nature and categorizability of subordinate clauses. The discussion in this section will centre on whether a discrete subdivision is realistic and useful, especially in a framework that assumes structural changes to be gradual and construction-speci c. The term dependent clause will be used to refer speci cally to subordinate clauses that are dependent on a main clause15 . The term `dependent clause' covers all clauses which, depending on theoretical framework or personal preference, are variably referred to as complement clause or clausal argument on the one hand, as well as those known as adjuncts, adverbial clauses or peripheral elements on the other. Such a categorical distinction between different types of dependent clauses is not made by all linguists; the term complement, from French complément, has traditionally been used in Romance linguistics to refer to both types of dependent clauses16 , but this clashes with the more restricted use of complement in recent syntactic theories. Doubts concerning the validity of a clear distinction between these two types of dependent clauses have also been raised by cognitive linguists over the past years (cf. e.g. Langacker, 1987). As the distinction between complements and adjuncts is usually understood to be semantic as well as syntactic, these two areas will be discussed separately. The semantic complement ↔ adjunct continuum Among the semantic criteria most frequently proposed for the distinction between complements and adjuncts (e.g. Matthews, 1981: 124 25) is the clearly semantic notion of whether the respective element is necessary as an integral participant or part in an event, or whether it provides circumstantial or additional background information to the event. For
15
16
These terms will be used throughout most of the book, but when discussing Generative theories, I will adopt the corresponding labels `embedded clause' and `matrix clause'. Pountain (1998: 395 405), for example, subdivides complement clauses into `object complements', `subject complements' and `prepositional complements', the latter (e.g. `... sin ella decir nada.') largely corresponding to what in other frameworks might be considered an adjunct.
32
Chapter 1
instance, in a sentence such as (6), `su cartera' is a necessary part of the event, whereas the location `en el parque' is a circumstantial dependent that provides additional background information about the event; hence the term peripheral element, which is often used interchangeably with adjunct. (6)
Juan busca su cartera en el parque. John searches his wallet
in the park
John is looking for his wallet in the park. The standard formal semantic analysis of the complement/adjunct distinction is that the complement [su cartera] is a semantic argument of the head [busca], whereas the adjunct [en el parque] is a functor that has the searching event [Juan busca su cartera] as its argument. This explains why a verb such as buscar can semantically require an object that is searched for, but cannot require an adjunct. Instead, it is the adjunct [en el parque] that requires an event as its argument. Langacker (1987: 278 ff.) argues that such unidirectional semantic valence relations are an inadmissible oversimpli cation, because localizable activities such as buscar necessarily require a location, as well as a searcher and an object that is being looked for. This is not the case for all predicates, as the unacceptability of (7) shows. (7)
*Juan heredó mil millones de pesetas en el parque. John inherited 1000 millions
of pesetas in the park
John inherited a billion pesetas in the park. One might say that verbs like heredar `to inherit' and enviudar `to be widowed' are not semantically subcategorized for place, nor for purpose, as shown in (8). (8)
*Juan enviudó
para. . .
John was.widowed in.order.to. . .
John was widowed in order to. . . Langacker (1987: 300) observes that the distinguishing feature between typical `complements' and typical `adjuncts' is their degree of importance in the characterization of the event, which he refers to as
Motivation and theoretical parameters
33
`salience'. Thus, `en el parque' is less salient, i.e. less central to the event, than `su cartera' in (6). As the entities involved in an event can have a varying degree of salience, this implies that, from a semantic perspective, the complement adjunct distinction is also gradient (cf. also Deutscher, 2000: 9; Croft 1988: 108, 2001: 280). What must also be taken into account is the pragmatic level, as the discourse situation is a crucial element in determining how central (salient) or relevant a particular piece of information is for a satisfactory description of an event. If we consider, for instance, a verb such as Spanish leer `to read', the central or relevant additional information will frequently refer to the item or information that is being read, as in (9) or (10). Alternatively, many discourse situations do not require this additional information, and providing it would in fact out the Gricean `maxim of relevance', as in (11). (9)
Estoy leyendo un libro. stat.1sg read.ger a
book
I am reading a book. (10)
Estoy leyendo que el presidente está stat.1sg read.ger that the president
muerto.
stat.3sg dead
I am reading that the president is dead. (11)
½No quiero escuchar música! ½Estoy leyendo! not want.1sg listen
music
stat.3sg read.ger
I don't want to listen to music! I'm reading! But there are, similarly, discourse situations in which it is irrelevant to the interlocutors what is being read, but highly relevant, and thus central to the event, where the reading takes place, as in (12). (12)
¾Por qué no te llevas el libro a casa? Pre ero leer en la biblioteca. *Pre ero leer. *Pre ero leer el libro Corazón tan blanco. Why don't you take the book home? I prefer to read in the library.
34
Chapter 1
* I prefer to read. * I prefer to read the book A Heart so White . In (12), [location] is thus the most central and salient element; in this particular discourse situation it is so crucial that it must obligatorily be mentioned. Mentioning the [patient], on the other hand, is unacceptable due to a lack of situational relevance. Sentences (11-12) illustrate the variability of the salience parameter. The event or activity of reading necessarily involves something being read, but it also necessarily involves a place, a time, etc. Which of these pieces of information are suf ciently relevant to be mentioned is primarily a matter of discourse pragmatics; it would be inappropriate to claim that any particular one of them is a priori more central, necessary or obligatory for the description of the event. It would, however, be an overstatement to claim that the choice of dependents is governed entirely by the discourse context. There are, for instance, transitive verbs like perseguir `to chase', which obligatorily require the [patient] role to be lled, even in contexts where it might not be of any pragmatic relevance. (13)
a. *Están persiguiendo. stat.3pl persecute.ger
They're persecuting. b.
Están persiguiendo a stat.3pl persecute.ger
alguien.
domkr someone
They're persecuting someone. Such apparently arbitrary lexical restrictions can be explained in terms of entrenchment and conventionalization. Depending on the meaning of a verb, a certain type of additional information may be particularly relevant in the majority of contexts in which the verb is used. With perseguir, for instance, it is highly likely that the `chased object or being', i.e. the [patient] will be mentioned, since it is usually of central importance to the act of chasing. Similarly, both the [theme] and [goal] of poner `to put' are generally suf ciently relevant to be mentioned. As a result, these lexically speci c patterns become entrenched and conventionalized, thus turning into syntactic rules or argu-
Motivation and theoretical parameters
35
ment structures. This leads straight into the next section, the discussion of syntactic criteria for the distinction of complements and adjuncts. Complements and adjuncts as syntactic categories It is an undeniable fact that certain verbs require certain semantic roles to be obligatorily lled, as discussed above for perseguir and poner. First of all, it must be pointed out that such rules are synchronic, and by no means cast in stone, but merely the result of the way the individual verb is most frequently used. If, for instance, `following' became a popular sporting activity (like `hunting') in which the object to be followed was highly predictable or irrelevant, it is likely that speakers would at some point start omitting this information, as its predictability would make it redundant in the context of the sporting activity. This pragmatically triggered change in usage could then easily become generalized, leading to a change in the argument structure of the verb and allowing the use of perseguir without an overt direct object. Nevertheless, synchronically perseguir requires the semantic role [patient] to be lled, whereas others, such as [time], need not be. However, as a consequence of the argument against universal categories presented in Section 1.3.1, this has no further implications for the syntactic status of the [patient] role; all it tells us is that Spanish has a construction containing a verb and an obligatory [patient], which perseguir can participate in. The usual way of determining whether an element has the status of complement or adjunct is by means of a number of distributional tests (cf. e.g. Radford 1988: 226 86). In Section 1.3.1, a logical argument against the usefulness of distributional tests for the purpose of syntactic categories was presented, based on the fact that distributional patterns in different constructions do not match, which makes it an essentially arbitrary decision which test should be diagnostic of syntactic category. This is con rmed, for the classi cation of dependent elements, by the following examples.
36
Chapter 1
The `do so' (pro-V-bar) test As brie y mentioned above, the verb poner `to put' requires both the semantic roles [theme] and [goal] to be occupied (14); for the verbs comprar `to buy' and olvidar `to forget', on the other hand, only [theme] is obligatory (15-16). (14)
a.
Pedro pone el libro en la mesa. Peter
puts
the book on the table.
b. *Pedro pone el libro. Peter
(15)
a.
a.
the book in the shop.
buys
the book.
Pedro olvida el libro en la mesa. Peter
b.
buys
Pedro compra el libro. Peter
(16)
the book.
Pedro compra el libro en la tienda. Peter
b.
puts
forgets the book on the table.
Pedro olvida el libro. Peter
forgets the book.
If we were to take obligatoriness as the criterion for complement status, the dividing line would have to be drawn between the complement `en la mesa' in (14) on the one hand, and the adjunct `en la mesa/tienda' in (15-16) on the other. A different test to determine complement status, based on the Generativist assumption that adjuncts are structurally more distant from the verb than complements, predicts that hacerlo `do so' cannot stand for constituents below the V-bar-level. This means that hacerlo can replace either the verb with its complements and adjuncts, or the verb with its complements but without its adjuncts, but not the verb without its complements. The ungrammaticality of (17b), in which hacerlo cannot replace `poner su libro' without `en la mesa', thus con rms the status of `en la mesa' as complement of poner, while the fact that it can replace `comprar su libro' without `en el mercado' in (18b) con rms the status of `en el mercado' as an adjunct of comprar.
Motivation and theoretical parameters
(17)
a.
37
Pedro va a [poner su libro en la mesa]v y Juan va a hacerlov también. Peter will put his book on the table and John will do so, too.
b. *Pedro va a [poner su libro en la mesa]v y Juan va a hacerlo en la silla. Peter will put his book on the table and John will do so on the chair. (18)
a.
Pedro va a [[comprar su libro]v en la tienda]v y Juan va a hacerlov también. Peter will buy his book in the shop and John will do so, too.
b.
Pedro va a [[comprar su libro]v en la tienda]v¯ y Juan va a hacerlov en el mercado. Peter will buy his book in the shop and John will do so in the market.
However, if we consider (19), this test for complement/adjunct status fails to match the results of the test for obligatoriness illustrated in (14-16): according to the pro-V-bar test, en la mesa in (19) is a complement of olvidar, whilst the obligatoriness test categorizes it as an adjunct. (19) * Pedro va a [olvidar su libro en la mesa]v y Juan va a hacerlo en la silla. Peter will forget his book on the table and John will do so on the chair. Once again, it has been shown that syntactic categories are not de nable beyond the level of individual constructions, as their distribution among constructions does not match. Clausal complements The nal example of a distributional mismatch in the area of complements and adjuncts concerns clausal complementation. Clausal comple-
38
Chapter 1
ments are frequently de ned either as clauses which are arguments of a predicate or as clauses that function as subject and/or object of a verb . Despite the fact that `argument' is an essentially semantic notion, whilst `subject' and `object' are purely syntactic categories, there is a tendency to equate the two de nitions. Givón (1993: 515), for instance, claims that sentential complements are propositions functioning in the role of either subject or object argument of the verb. Even more explicitly, Noonan (1985: 42) writes: By complementation we mean the syntactic situation which arises when a notional sentence or a predication is an argument of a predicate. For our purposes, a predication can be viewed as an argument of a predicate if it functions as the subject or object of that predicate. Noonan's examples further imply that object complements function as direct objects of verbs. Leaving subject complements aside for the time being, the general assumption that a complement `functions as' the (direct) object of the main verb is accurate for sentences such as (20-22). Direct object NP: (20)
Quiero [un café]DO . want.1sg a
coffee
I want a coffee. Finite DO complement: (21)
Quiero [que tomes
un café]DO .
want.1sg that take.2sg.sbjv un café
I want you to have a coffee. In nitival DO complement: (22)
Quiero [tomar un café]DO . want.1sg take.inf un café
I want to have a coffee. There is, however, a class of main verbs that participate in a less straightforward complementation pattern. Consider example (23).
Motivation and theoretical parameters
(23)
a.
Pedro cuenta que has Peter
tells
39
fracasado.
that you.have failed.
b. *Pedro cuenta tu fracaso. Peter
c.
tells
your failure.
Pedro cuenta de/sobre tu fracaso. Peter
tells
of/about
your failure.
If `functioning as' is to be understood as `having the same distribution as', then we face a serious problem explaining why the clausal complement does not require the preposition that the corresponding NP does. Ascribing `dummy status' to the preposition does not resolve the issue, as it nevertheless constitutes an additional element that cannot be accounted for if object NPs and complement clauses are considered to occur in the same syntactic environment. The reason for this mismatch is a con ation of the semantic and the syntactic level. The sentences in (23) are semantically, but not syntactically equivalent. Deutscher (2000: 7 13) draws a clear distinction between an object, a syntactic surface category, recognizable (with varying degrees of con dence across languages) by factors such as case marking, word order, or passivization , and an argument, which is a (semantic) category which denotes intimacy in the relation between an element and its predicate. If we understand complements only to be the same type of argument as the corresponding nominals, but not the same type of object, this can account for the absence of the oblique marker (i.e. the preposition) in the examples above. While different semantic case-roles are explicitly marked by prepositions such as de and sobre, the corresponding nite complements with que do not always require this overt marking; they merely mark an intimate link between the predicate and its clausal argument. 17 (Deutscher, 2000: 10) This constitutes yet another distributional mismatch: clausal complements, so-called `object clauses', occur in a far wider range of syntactic environments than canonical NP direct objects. Whilst the construction in (20-21) assigns them to the same category, thus making
17
With reference to the corresponding English that.
40
Chapter 1
them interchangeable, the two different constructions in (23) require a clear distinction between dependent clauses and object NPs. 1.3.4 Prepositional constructions In line with the argument that categories are de ned in terms of the constructions they participate in, presented in Sections 1.3.1 and 1.3.2 following Croft (2004), I will limit my de nition of the term `preposition' to its role in relation to the constructions under investigation, dependent clauses.18 I will adopt a broad formal de nition of the term `preposition' as a component of Spanish constructions involving a dependent clause; the same de nition also holds for Portuguese and Romanian. In general terms, a preposition can be de ned as the element [prep] in constructions of the generic type shown in (24). (24)
[[main clause] [prep] [infinitival dependent clause]]
A construction of this type is presented in (25), in which the [infinitival dependent clause (idc) represents the purpose of the [main clause (mc)]. (25)
[[mc] [para] [idc]] [[Trabaja] [para] [ganar dinero]] He works to earn money.
That the overall meaning of this construction depends partly on the lexical meaning of the preposition itself is clearly visible from the fact that para has nal meaning in other constructions as well. But at the
18
This does not in anyway imply a denial of the fact that the same lexical elements also participate in other constructions, notably in connection with NPs. However, there is no uniform class of prepositions shared by both construction types, which is evidenced by the fact that certain `NP-prepositions' available for the nominal construction are not available for the prepositional in nitive construction; for Spanish this group includes detrás `behind', sobre `over', durante `during', etc.
Motivation and theoretical parameters
41
same time, it is the construction as a whole that carries the following relational meaning: {[idc] stands in relation to [mc] in the way speci ed by [para]} How much the preposition's lexical meaning contributes to the meaning of the construction as a whole is a construction-speci c parameter19 ; this is illustrated by the construction in (26). (26)
[[mc] [a] [idc]] [[Aprende] [a] [leer]] He learns to read.
In contrast to the construction with [para], this construction carries the meaning: {[idc] stands in relation to [mc] in a way speci ed by [mc]}, not ∗{[idc] stands in relation to [mc] in the way speci ed by [a]}.
It is not dif cult to see that the relation expressed in [[mc] [a] [idc]] corresponds closely to what might be termed a typical `complement relation'20 , while [[mc] [para] [idc]] corresponds to an `adjunct relation'. The advantage of this construction-based model, however, is that it can also account for the intermediate stages between the two extremes, as the following analysis of the diachronic development of the construction [[mc] [a] [idc]] exempli es.
19
20
That the lexical meaning of the preposition is not necessarily linked to the meaning of the construction is corroborated by the cognate construction in Old Romanian, where the preposition had disappeared from the language as a whole, but occurred in the in nitival construction in a range of variably complement- and adjunct-like usages (cf. Section 6.3.1). However, according to Givón (1984: 519), the presence of a complementizer of this kind nevertheless tends to constitute a lesser degree of integration between the main clause and its complement than there would be in the absence of a complementizer.
42
Chapter 1
Originally, the meaning of the construction [[mc] [a] [idc]] can be assumed to have been: {[idc] stands in relation to [mc] in a way speci ed by [a] (in its nal meaning).} But this meaning gradually evolved to nally arrive at its present-day meaning: {[idc] stands in relation to [mc] in a way speci ed by [mc]} This is due to a process of reanalysis, described in more detail in Chapter 7, during which the semantic contribution of a to the meaning of the construction gradually diminished. During the transitional stage, the semantic contribution of the construction as a whole can be assumed to have been neither the original one, nor the present-day one, but rather a combination of the two, something like: {[idc] stands in relation to [mc] (a)in a way speci ed by [a] (in its nal meaning) and/or21 (b)in a way speci ed by [mc]} The weighting between the components (a) and (b) gradually shifted from predominantly (a) to predominantly (b), until (a) was nally lost altogether. This process will be discussed in more detail in Sections 4.4 and 7.1. The fact that it has been possible for the construction with [a] to evolve independently, acquiring an overall meaning that differs from that of other similar constructions such as the one with [para], indicates that we are, indeed, dealing with two separate constructions. Whilst these constructions do, in the modern language, represent opposite ends of the semantic complement adjunct continuum, the construction-based model proposed here provides a method of analysis that can account for less clear-cut cases as well as gradual transition between the two extremes of the continuum. A currently incipient case of this kind, the
21
The juxtaposition of and and or represents the gradient nature of the weighting between these components: even when one component is the predominant one (or), the other may also apply to a lesser extent (and).
Motivation and theoretical parameters
43
Portuguese construction with [para], will be presented in Section 7.4. The fact that the preposition a has, in this construction, lost its lexical content and become a purely functional element justi es giving it a separate label; as the construction it forms part of expresses a prototypical semantic complement relation, I will use the term prepositional complementizer, retaining the element `prepositional' to indicate that it is nevertheless located on the continuum shared with other prepositional dependent clauses.
1.3.5
Compound prepositions and conjunctions
Compound prepositions The term `preposition' will, in this study, be understood to include etymologically monomorphemic prepositions as well as `compound prepositions' such as Spanish `en vez de' `insted of', Portuguese `apesar de' `despite' or Romanian `în loc de' `instead of'. Though it is uncontroversial that such `compound prepositions' contain an element that corresponds etymologically to a noun, they have become fully grammaticalized prepositions. A range of features of these `compound prepositions' supports this analysis. Most importantly for this study, `compound prepositions' occur in the same construction type as the etymologically monomorphemic ones. For instance, the constructions in (27) are entirely parallel, both in form and in the meaning their constructional meaning. (27)
a.
[[mc] [sin] [idc]] [[Trabaja] [sin] [ganar dinero]] He works without earning money.
b.
[[mc] [en vez de] [idc]] [[Trabaja] [en vez de] [dormir]] He works instead of sleeping.
The meaning in both cases is:
44
Chapter 1
{[idc] stands in relation to [mc] in the way speci ed by [prep].} Regarding the semantic status of the noun, its literal meaning is no longer present: neither does the notion `despite' contain a component of weight, regret or worry (pesar), nor does `instead of' necessarily involve a turn, place, or location (vez, lugar). Syntactically, too, such `compound prepositions' are fully grammaticalized. Their distribution is equivalent to that of non-compound prepositions, and they consist of invariable morphological sequences with no possibility of other lexical or morphological elements being inserted. Furthermore, if the etymologically separate elements were still to be understood as separate syntactic constituents, the lack of an article would constitute a syntactic anomaly in all three languages.22 The fact that they retain an orthographic form that to some extent re ects their etymological origin must thus not be understood as indication of non-prepositional syntactic status. Compound conjunctions In the statistical section of this study, comparisons will be drawn between the diachronic development of prepositional in nitives and the corresponding nite dependent clauses. The constructions these nite clauses occur in are similar to the prepositional construction(s) introduced in Section 1.3.4 and can, correspondingly, be described generically as in (28). (28)
[[main clause] [conjunction] [finite dependent clause]]
Parallel to the case of the prepositional construction, all elements that can occupy the [conj] position can be de ned as conjunctions; again, this is a construction-speci c classi cation. The majority of conjunctions are, from an etymological point of view, `compound conjunctions', e.g. sin que `without', antes de que `before'; also porque `because' and aunque `though'. The meaning of the constructions that
22
Normally, a noun that has a (prepositional) modi er must also have an article.
Motivation and theoretical parameters
45
these conjunctions participate in is similar to that of constructions with the semantically (and often morphologically) related prepositions. It is perhaps worth mentioning that the construction with the conjunction que is similar to constructions with prepositional complementizers such as a in that the semantic relationship between main and dependent clause is largely determined by the main verb. 1.3.6 Arguments for the continued use of traditional category labels Having rejected the principle of uniform syntactic categories, synchronically among different constructions as well as within individual constructions through the course of their development, the traditional category labels are nevertheless useful. This is because in many cases, a correspondence among distributional patterns can be identi ed for a comparatively large group of elements that do show a correspondence in their behaviour across constructions and through time; these elements can be considered prototypical elements of a traditional `cross-constructional' syntactic category. A prototypical member of such a traditional category would be one that passes a certain number of distributional tests. Which speci c tests we choose to include in the list will remain essentially arbitrary, but we would be guided by the aim to group together a relatively large number of elements that show a relatively high degree of cross-constructional correspondence. A prototypical Spanish direct object complement, for instance, might be de ned as one that is obligatory, must appear in a position right-adjacent to the verb, and can function as the subject of a passive sentence; semantic criteria can also be included, so we might add the requirement for it to have the semantic role [theme]. This selection of criteria is based purely on the fact that all of them are satis ed in a relatively large number of cases in a speci c language23 ; the more criteria we pick, the smaller the number of prototypical cases we will be left with.
23
Certain distributional correspondence patterns are also cross-linguistically common, but by no means universal.
46
Chapter 1
Having de ned a prototypical member of such a cross-constructional category, it is also possible to identify less typical cases that satisfy some, but not all of the criteria selected. If a suf cient number of criteria are satis ed, we might still include these non-prototypical cases in the cross-constructional category. However, there will also be cases that share certain features with the prototypical members of one cross-constructional category, but a similar number of features with the prototypical members of another or several other cross-constructional categories. It must be emphasized that such a classi catory system contains several arbitrary elements and therefore has no analytic value in itself; it is nothing more than a convenient shorthand way of saying that a group of elements shares a certain number of features which happen to co-occur quite often; non-prototypical members can further be de ned in terms of the speci c features they do not share with the prototypical members. As long as the syntactic elements we are analyzing generally match the syntactic prototypes we have de ned, using the established category labels is convenient and largely unproblematic, especially if they are de ned on the basis of what is common and typical in the analyzed language. Applying the same category labels to diachronic syntactic analysis, however, can prove more problematic. If the category labels we try to use are de ned on the basis of typical combinations of features and distributions of the modern language, this must not necessarily match the most typical combinations at an earlier stage of the same language. Indeed, it is the very object of diachronic syntactic analysis to identify and examine those elements that have changed, which includes changes in individual distribution patterns. A second dif culty is that syntactic change is a gradual process, and it is not necessarily clear whether a criterion for membership in a cross-constructional category is satis ed or not. In the case of semantic criteria, this is uncontroversial, as semantic shift is in most cases gradient; recall the case of French pas in Section 1.3.2, in which the particle pas took on a gradually increasing degree of negative force over the centuries. That the semantic distinction between traditional categories
Motivation and theoretical parameters
47
such as complement and adjunct is also gradient was discussed in detail in Section 1.3.3. Similar problems arise when checking whether a speci c distributional criterion is satis ed. A very practical dif culty is the fact that native speaker judgements are unavailable for earlier stages of the language. Though it is possible to determine whether a syntactic structure was grammatical at the time on the basis of the documents we have, the absence of a structure from the available corpus cannot be understood as proof that the construction was ungrammatical at the time; the strongest legitimate conclusion would be that it was not particularly commonly used. This leads straight to a second, more fundamental issue: even if a few sporadic instances of a structure can be found, does this make it a fully integrated part of the language? Throughout the statistical sections of this study, it will become apparent that syntactic structures do not appear suddenly as a fully- edged part of the language, but enter it gradually, their frequency increasing over time. This permits an analysis by which the grammaticality of a structure is gradient, its rise in frequency mirroring an increasing degree of grammaticality. Though rejected by some theoretical frameworks, the gradient nature of grammaticality is a well-known fact that can frequently be observed when asking informants for native speaker judgements about a structure that occurs only sporadically; the answer in such cases will often be something like: One might say that, but one wouldn't normally. Though the non-discrete nature of linguistic change adds to the dif culty of applying traditional category labels, these problems are not necessarily insurmountable; where a criterion for membership in a particular cross-constructional category is only partially satis ed, such cases can be understood to be less prototypical than other cases that fully satisfy the same criterion. It may therefore be concluded that traditional category labels, though essentially arbitrary, can be usefully applied in the description of language change, as long as we keep in mind that membership in these categories is not a binary property. This is particularly important when examining changes in which an element shifts from being a prototypical member of one category to being a prototypical member of a different one; in such cases, there is bound to be a stage of ambiguity somewhere
48
Chapter 1
along the way, for which neither of the two labels would be adequate. Such a situation will be presented in Chapter 7.
1.4
Pragmatic causation in syntax
Pragmatics is perhaps the subdiscipline of linguistics most closely related to semantics, and is usually conceived as a branch of semantics concerned with the meanings that sentences have in particular contexts in which they are uttered (Matthews, 1997: 290). At the same time, however, it is perceived as opposed to truth-conditional semantics based on Frege's (1892) Principles of Compositionality, which constitutes the core of much of modern semantics24 . It was originally conceived by Morris (1971: 43 54) as opposed to syntax as well as semantics, covering the eld of relations between signs and their interpretants. A discussion of the intricate subdivision of language into such categories is not the aim of this study. It will be assumed throughout that the meaning of any utterance is the compositional product of the semantic content of the individual words it contains, the syntactic structures chosen to combine them, and the (linguistic and extra-linguistic) pragmatic context of the utterance. The focus will be on the interface between meaning and syntax. It is assumed that the objective of any normal utterance is to convey information as precisely as necessary but also as economically as possible, in line with the Gricean Maxims of Conversation (Grice, 1989). To attain this objective, the language user can make syntactic choices, and these choices are in uenced by the semantic content as well as the pragmatic context of the individual utterance. An important factor in this process is language users' expectation: everyone has certain patterns of expectations concerning the way in which the elements of a sentence are likely to interact; these expecta-
24
Formal semantics ... is generally taken to be complementary with pragmatics, ... the study of that part of meaning which is not purely truth conditional... (Lyons, 1981: 170 171)
Motivation and theoretical parameters
49
tion patterns are a re ection of our knowledge and experience of the way the things represented by the words interact in the real world. Expectation is, thus, the product of word and construction meaning plus the individual speaker's experience. The relative predictability of how entities in the real world tend to interact allows the language user to develop his own expectation patterns, but such expectation patterns cannot themselves be understood as part of semantics, as they are merely a default interpretation that is applied if no contextual indications specify a different pattern of interaction, as illustrated in the following Spanish example. (29)
Las dos chicas se the two girls
miran.
REFL look.3sg
The two girls look at each other. (30)
Las dos chicas se the two girls
miran en el espejo.
REFL look.3sg in the mirror
The two girls look at themselves in the mirror. It can be observed that, according to our knowledge of the real world, the default assumption (29) for mirarse with two agents is one of reciprocity, but that certain contexts can override this default interpretation in favour of a literal re exive interpretation (30). Throughout this book, it will become apparent that default assumptions, conditioned by what is perceived to be the pragmatically most likely relationship between constituents, are a crucial factor synchronically in the choice and analysis of syntactic structures, and consequently also diachronically in syntactic change. 1.4.1 Synchrony: pragmatic factors in syntactic choice Syntactic choice is an omnipresent feature of language, as a speaker is obliged to choose a particular structure for every sentence he utters. In the majority of cases, the semantic content of a sentence is directly responsible for determining, or at least severely limiting, the syntactic options a speaker has. A transitive verb like Spanish castigar `to punish', for instance, projects the semantic need for the semantic roles
50
Chapter 1
[Agent] and [Patient] to the syntactic level, at which the corresponding [Subject] and [Object] represent the semantic roles. An alternative projection would be the corresponding passive clause, in which the prepositional phrase represents the agent and the subject represents the patient. (31)
PedroS castiga al Peter
perroDO .
punishes domkr.def dog
Peter punishes the dog. (32)
El perroS es castigado por PedroPO . the dog
is punished
by Peter
The dog is punished by Peter. Banal as this example may be, the point is that while part of the syntactic structure is determined by the semantic content of the sentence, the speaker is left with more than one syntactic option for his sentence. Whilst even strict followers of formal syntactic models generally accept that argument structure and verbal subcategorization are at the interface between semantics and syntax, the corresponding interface between pragmatics and syntax is less widely recognized. This is partly due to the fact that pragmatic meaning is an even less uniform category than straightforward semantics. For the choice between the active and the passive construction in the example above, there are at least two rather different pragmatic factors involved. One is register, the other is topicalization. Colloquial Spanish has a certain resistance to the use of the ser-passive, whilst journalistic register involves a much more frequent use of the structure. Furthermore, `el perro' is topicalized in sentence (32) but not in (31). It is obvious that such notions cannot be integrated into any syntactic theory in which syntax is an autonomous entity; the fact that speakers can choose among structures calls for an analysis of what in uences or determines this choice, be it the linguistic or social context, discourse pragmatics, or expectation. Until not too long ago, pragmatics was often sidelined by linguists because it was regarded as intrinsically dif cult to systematize, due to its dependence on individual contexts and utterances. Indeed, during most of the 20th century, Structuralists as well as Generativists were almost exclusively concerned with understanding and describing the ab-
Motivation and theoretical parameters
51
stract underlying system in language(s), consciously disregarding the role of the speaker in their analysis. In recent years, however, a renewed interest in pragmatic causation within syntax has arisen, and linguists have begun to follow in the footsteps of such scholars as Whitney (1875) and Bréal (1983 [1897]), who believed that language has to be de ned in relation to human thought and action. (Nerlich, 1996: 70 71) In Romance Linguistics, the pragmatic approach to syntax has been revived by such scholars as Manoliu-Manea, Klein-Andreu, Silva-Corvalán, and Pountain, and it has successfully been applied to such complex areas as the choice of mood in the nite complement (Klein, 1975), the various syntactic functions of the re exive verbal structure (Pountain, 2000). Person assignment in in nitival clauses, an area that is of direct relevance to this study, has been examined from a pragmatic point of view by Comrie (1984, 1985) cross-linguistically, and by Pountain (1995, 1998a) for Spanish and Portuguese. 1.4.2 Diachrony: pragmatic factors in syntactic change It is widely accepted that reanalysis, as brie y exempli ed in Section 1.2.1 above, is a major factor in syntactic change, whether seen as a component of grammaticalization (following Meillet, 1912), or as an autonomous mechanism in its own right (cf. e.g. Harris & Campbell, 1995: 61 93, Croft 2000: 117 44). In fact, Harris & Campbell argue that Alteration of syntactic patterns takes place by means of speci c mechanisms of change. We hypothesize that there are only three basic mechanisms: reanalysis, extension, and borrowing. There appears to be less clarity or consensus regarding the reasons for language users to reanalyse some structurally ambiguous constructions, but not others; explanations range from the in uence of language contact to increased functional ef ciency or linguistic optimality (cf. e.g. Langacker, 1977). With the renewed interest in pragmatics and the interface between linguistic performance and underlying structure, the important role of pragmatics as a central factor in language change has also gained widespread recognition. Traugott & Dasher (2002: 1), for instance, show
52
Chapter 1
that there are predictable paths for semantic change [which] recur so often and across totally unrelated languages [because they are] intrinsically bound up with the cognitive and communicative processes by which pragmatic meanings come to be conventionalized and reanalysed... . Similar principles are also understood to be involved in many instances of syntactic change, such as structural reanalysis, as exemplied in Harris & Campbell (1995: 61 96). One way in which pragmatics contributes to the process of reanalysis is by virtue of the fact that, in the real-world context of utterances, certain patterns of interaction or relation between the entities represented by the constituents of a sentence are more likely to occur than others. As these patterns are generally known to the language user, he can, therefore, expect particularly common or likely relational patterns to be the norm. Confronted with structurally ambiguous utterances, it is thus only natural for him to choose the interpretation that most closely represents the expected default situation. This is a speci c form of the mechanism of pragmatic inference (Croft, 2000: 133), in which a contextual parameter that frequently co-occurs with a structure is reanalysed as part of the structure itself, and the former meaning of the structure eventually gets lost. A simple illustration is the currently ongoing process of reanalysis of the verb tener in Spanish. The original meaning of tener is `to have, to hold', but it is currently acquiring the additional function of auxiliary marker of perfectivity when used with the past participle of a growing number of transitive verbs25 . In sentence (33), it is used in its original meaning. (33)
Ahora tengo el coche, comprado y pagado. now
have.1sg the car
bought
and paid
Now I've got the car, bought and paid for. Relatively frequently, for example in the presence of a pronominal object, the verb tener and the past participle will appear next to each other, as in (34).
25
For a detailed synchronic description of its usage, see Harre (1991).
Motivation and theoretical parameters
(34)
53
Ahora lo tengo [comprado]. now
it have.1sg bought
Now I have it, bought. Anyone who knows how the real world functions will realize that if someone has got something that has been bought, the most likely scenario is that he/she has bought it. In the vast majority of communicative situations, tengo in sentence (34) can thus be understood to express perfective aspect without jeopardizing the functionality of the communicative process. Whilst the speaker may intend to say sentence (34), the hearer may inadvertently parse it as (35). (35)
Ahora lo [tengo comprado]. now
it have.1sg bought
Now [I have bought] it. The nal step in this process of reanalysis is for the re-parsed structure to be transferred from such ambiguous contexts to sentences with a structure that does not allow for ambiguity, as in (36). (36)
Ahora [tengo comprado] el coche. now
have.1sg bought
the car
Now [I've bought] the car. In this case, an important precondition for reanalysis is the fact that the language user knows, from his experience of the real world, that the possession of an object usually entails its previous acquisition. But this alone may not be suf cient to trigger reanalysis. A second pragmatic factor that comes into play are the Gricean maxims of quantity and relevance, according to which the hearer will expect any utterance not to contain more information than required, and only relevant information, and he will interpret the utterance to best satisfy these maxims. Thus, sentence (34) is reanalysed as (35) because the latter contains information about the process of buying as well as the status of ownership; in the absence of any contextual cue to the contrary, the hearer must therefore assume that all extractable information is deemed relevant by the speaker. In Chapter 7, it will be shown that the relevance parameter is
54
Chapter 1
a crucial factor in the development of Romance prepositional complementizers. At this point it must be pointed out that we are talking about individual utterances, reanalyzable only where the pragmatics of the available linguistic and extra-linguistic context do not eliminate the structural ambiguity. Syntactic change will, however, not be triggered by any single individual utterance. But if the proportion of contexts in which reanalysis can and does take place is high, this will lead to a situation in which the reanalysed interpretation becomes the default analysis, as mentioned in Section 1.2.1 above and discussed further in 1.5.1. The nal step, the analogical extension of the reanalysed structure to syntactic contexts in which the original usage would not be possible, can be seen as part of a grammaticalization process, with several of the typical hallmarks described by Lehmann (1985: 306 308). Among other things, reanalysed tener is becoming increasingly semantically bleached, turning into a functional aspect marker that can be used in an increasing number of syntactic and semantic contexts. Sentence (36) illustrates how the syntagmatic variability of auxiliary tener is reduced, as it must immediately precede the past participle. In what manner the extension of the perceived equivalence to other contexts proceeds is one of the central questions examined in this study. The data regarding the evolution of the Romance prepositional in nitive suggest a gradual extension, item by item, in a way similar to lexical diffusion of sound changes, as proposed by Sturtevant (1917) and Wang (1969). Crucially, such diffusional extension does not mean that a change must necessarily spread in a regular fashion to all structurally equivalent environments. Such gradual diffusion is linked to the fact that reanalysis is itself not a discrete process. A single structure can, even synchronically, be used in its original and its newly acquired, reanalysed sense. In genuinely ambiguous contexts, (34) and (35) are not distinguishable, either syntactically or pragmatically, and there is, indeed, no reason to assume that the speaker makes a clear choice between the two. This non-discrete nature is also re ected semantically by the fact that, depending on the context, the auxiliary tener can retain a variable degree of the semantic notion `to hold'. `Lo tengo pensado' `I've thought it out', for instance, retains some sense of holding the result of ones thoughts in
Motivation and theoretical parameters
55
ones mind. On the other hand, in `hace como tres meses lo tengo perdido en este maldito cuarto'26 , `it's about three months since I lost it in this damned room', the semantics of the verb perder and our knowledge of the way things interact in the real world makes it obvious that losing something generally rules out still holding it. Summing up, it can be said that syntactic changes of this type are made possible by the fact that a construction is used ambiguously in certain pragmatic contexts. If such ambiguity occurs frequently, this allows for a gradual extension to contexts in which the reanalysed meaning is more central than the original one, and nally to contexts in which the original meaning is pragmatically unlikely or logically impossible. As the process of extension is a gradual one, the contexts it spreads to rst are those that are most similar to the originally ambiguous one. For instance, the fact that tener can, at present, only function as auxiliary with transitive verbs shows how a certain link to its original meaning and argument structure still exists.
1.5
Statistical analysis: reasons and methods
1.5.1 Why statistics? The motivation for a statistical approach Much of this study is based on statistical data. Contemporary syntacticians sometimes argue that statistics are not relevant in syntax, as the primary object of syntactic analysis is to determine the grammaticality or ungrammaticality of constructions, and to identify the underlying structures they are governed by. However, such a narrow approach to syntax overlooks the fact that some structures are used more commonly than others, and that acceptability or grammaticality of a structure can be gradient, as brie y discussed in Section 1.3.6.
26
Sergio Madrigal González: Los amantes, http://www.geocities.com/albisamm/ 1999/v-30/serg-30.html
56
Chapter 1
Furthermore, such a limited approach cannot do justice to the diachronic dimension. It delivers a series of separate synchronic snapshots, but is unable to relate them to one another in a meaningful way. If we accept that pragmatically-based syntactic changes do not occur in a discrete fashion, then the transitional period must be accounted for as a process, not a series of discrete grammars, especially if the aim is to explain, rather than merely describe, the change. A different approach is thus necessary to analyse development through time, which, as described in Section 1.4, is linked to the frequency with which a particular construction is used: more frequent use of a structure means that the structure has become more acceptable or normal for language users; such increased `normality' generally leads to loss of expressive force, to semantic bleaching and grammaticalization. The concept that the quantitative study of language is important for our understanding of linguistic structure and change, as rst proposed by Zipf in his Introduction to Dynamic Philology (1935), was only adopted sporadically by others until the late 1970s. In recent years, however, the renewed interest in pragmatics, paired with the advent of digital technology and the availability of large annotated text corpora, has led to a marked rise in studies of token frequency27 . Wanner (2003) goes so far as to claim that all linguistic research is overtly or implicitly corpus-based, as the utilization of linguistic data without corpus (or questionnaire) is merely anecdotal, producing accidental results. In current theories of language change, frequency plays an important role. In cognitive linguistics, a central notion is that of the `construction' (e.g. Langacker, 1987), essentially a conventionalized, entrenched routine (Croft, 2004). Viewed diachronically, new constructions thus emerge when a particular pattern occurs often enough to become entrenched and conventionalized. In grammaticalization, which by some linguists has been elevated to a theory in its own right (Bybee et al., 1994: 4; Heine, 1997: 6), one of the central principles is an increase in the frequency of the structure undergoing grammaticalization, which is
27
Token frequency is based on the real number of occurrences of a particular feature, conceptually opposed to the number of types of structures the feature is found in.
Motivation and theoretical parameters
57
seen as a cause for semantic and phonological bleaching (cf. e.g. McMahon, 1994: 160 173). One way of applying and testing such theories is by means of synchronic corpus analysis, which can yield valuable information concerning changes in progress, as shown in numerous contributions to a collection of papers on Frequency and the emergence of linguistic structure (Bybee & Hopper, 2001). However, such accounts fall short of revealing long-term developments and can often merely speculate what the eventual outcome of a change in progress will be. Partly due to the technical dif culties involved in compiling the appropriate corpora, statistically-based diachronic studies were, for a long time, comparatively scarce, but this has been changing in recent years. Dealing with pragmatic factors involved in changing patterns of clausal complementation, there are textually-based studies by Deutscher (2000) on the emergence of nite complementation in Akkadian, and by Schøsler (2000) and van Reenen & Schøsler (1993) on in nitival complementation in Old French. These recent advances make a comparative study based on the diachronic development of token frequency particularly timely, as the additional comparative dimension will not only allow for a more informed evaluation and interpretation of the data for individual languages, but also enable us to identify more general, cross-linguistic patterns of development. The usefulness of comparative diachronic statistics Tracing back the usage of a structure can be expected to provide information about the contexts it originally emerged in, as well as the way in which it subsequently evolved in the respective languages. As pointed out in Section 1.4.2, a crucial variable in pragmatically-based change is how frequently utterances containing the structure occur in actual language usage. If the usage of the same structure differs among related languages, tracing it back as far as possible in the individual languages may allow us to establish whether it is likely that there is a common origin, or whether it might be safe to claim that it has emerged independently in
58
Chapter 1
each language. Such polygenesis can be coincidental, or it can be due to the presence of the same factors in all the languages involved. Even if the structure only evolves after the languages are clearly separated, this does not necessarily rule out common causation: as pointed out in Section 1.2.2 above, common inherited structures may predispose the respective languages to change in similar ways, or independent but similar changes in more than one language may be caused or favoured by more general pragmatic and cognitive principles which are not speci c to any one language or even language family. Parallel changes in more than one language, even if they take place independently, can therefore well have a common cause28 . Diachronic statistics, i.e. charting the frequency of a construction through time, is a method to reveal its evolutionary path. The limited amount of textual data from the past makes it impossible to determine the exact point in time at which the construction was rst used. If we do nd a single token, it might nevertheless not be an integral part of most speakers' language at that time. On the other hand, the structure might have already been in use long before it rst appears in a preserved document. If, however, a statistically signi cant change in usage frequency can be detected, this provides information about more general trends in the overall way speakers use the language. These trends can then be compared across languages to reveal similarities and differences in the way the structure has evolved. Where the use of the structure diverges between two languages, usage frequency data allows us to pinpoint at what stage of a structure's development, and in which contexts, speakers of language A began regularly to use it in a way that speakers of language B did not. Parallel developments are likely to be triggered by the presence of similar conditions or principles in both languages; by comparing the paths along which related constructions emerge and gain popularity in each of the two languages, the principles and conditions common to both can be identi ed as the relevant ones. This will be one of the objectives of the comparison of the Spanish and Portuguese data in Chapter 5.
28
Contact between the languages can complicate the analysis; see Section 1.2.2.
Motivation and theoretical parameters
59
A further use of comparative statistical analysis is the fact that it has some predictive potential: if the distribution and development of a structure in language A resembles that found during earlier stages of development in languages B and C, this suggests not only similar causation, but also that the future development in A may proceed along similar lines as it has done in B and C. This situation, it will be seen in Chapter 6, applies to the use of prepositional in nitives in present-day Romanian. 1.5.2 Data selection and statistical procedures The diachronic corpora analysed in this book are made up of the texts listed in appendices A, B and C for Spanish, Portuguese, and Romanian, respectively.29 Though some effort has been made to provide a varied sample of texts, any such selection must be arbitrary to a certain extent. Due to the relatively large number of texts and the length of the period under investigation, it is hoped that any local imbalances will not have a signi cant effect on the general conclusions that can be drawn from the statistical analysis. The corpora do not contain Latin American texts, as including them indiscriminately would risk distorting the data by ignoring split developments between the Old and New World varieties. A separate analysis of the differences between the Iberian varieties and their Latin American counterparts would, nevertheless, be of great interest and a promising object of future research, particularly because any split development could be compared to the earlier split between Spanish and Portuguese. Any approach concerned with the usage frequency of linguistic structures should be based on the greatest possible corpus of data, so that it can be assumed to be largely representative of the language as a whole. Being representative does not merely require a large number of
29
More extensive digitalized diachronic corpora have become available in recent years, most notably Mark Davies's fully annotated Corpus del Español, but were not yet in the public domain at the time the statistical work for this book was begun.
60
Chapter 1
texts or words; the corpus must also be made up of a balanced mixture from a variety of backgrounds, linguistic registers, specialist jargons, and other varieties that make up the language in question, as it is beyond the scope of this study to produce differentiated series of data for such individual subvarieties. The need for a balanced mixture faces us with a number of practical dif culties. Any diachronic study will unavoidably suffer from a lack of varied data for past periods. In particular, it is generally impossible to nd extensive documentation that accurately represents the spoken language of past periods30 ; this is particularly deplorable because spoken language is generally (but not always31 ) the source of innovation. As a rule of thumb, the further we go back, the less likely it is for a text to reveal features of the popular, spoken language: the earliest Romance prose texts tend to be written in formal register, as they are generally legal, theological/religious and historical. Whilst a great deal of work has been done to compile corpora of modern spoken language and other linguistic registers, the absence of equivalent diachronic data leads to an unavoidable discrepancy between contemporary and historical surveys. With this in mind, care is called for in evaluating any apparent sudden structural changes that coincide with the transition from historical textual data to the more comprehensive modern corpora. Many innovations originally arise in the spoken language and later permeate the written registers. On the other hand, structures can remain present in the oral domain for a long period of time without gaining more formal status. In addition, account must be taken of the fact that the presence of a tangible pragmatic context allows speech to be less precise and far more deictically determined. Though invaluable for the analysis of synchronic choices and their causes, spoken corpora must, thus, be taken cum grano salis from a diachronic perspective.
30
31
Baum (2003: 49) goes so far as to say that there can be no authentic documents of (spontaneous) spoken language before the advent of the new media , thus implicitly doubting the reliability of any written representation of the spoken language. Learned features, in the Romance languages generally borrowed from or calqued on Classical Latin, usually enter the language via the written medium, cf. Pountain (1998b)
Motivation and theoretical parameters
61
The limited number of texts available from some earlier periods does not give the linguist a great range of choice. Text type, register, and many other parameters that have a direct effect on stylistic and syntactic choices (linguistic conservatism, incorporation of learned features, etc.) vary considerably. Even in the most comprehensive of corpora such variation cannot be avoided, as certain text types are often limited to, or dominant in, a certain period, but may be rare, obsolete, or still a thing of the future at another time. Particularly problematic in this respect is the fact that there are so few early medieval texts at our disposal that it is necessary to use lyrical material, despite the well-known fact that poetic licence allows the use of atypical structures, with the purpose of maintaining poetic parameters such as rhythm, rhyme, and assonance. Word order is among the greatest sufferers in this respect. It might therefore be argued that texts of such diverse types represent `separate grammars', and basing a statistical analysis on them would be an arti cial con ation of these grammars. But this could be said of any statistical analysis, as variation in grammatical usage among speakers means that any language or variety, viewed beyond the idiolectal level, consists of multiple grammars to some extent. It is, indeed, one of the strengths of the statistical approach that such variation is taken into account, without giving undue weight to structures that are marginal for the language as a whole. Nevertheless, it must be acknowledged that the somewhat unbalanced use of texts from the earlier stages of the languages under investigation might distort the analysis. To minimize the effects of any such potential distortion, two mathematical algorithms, described in detail below, will be applied to the data gathered. Another issue that must be addressed in a study making extensive use of historical literary material is the subdivision of languages into discrete historical periods such as `Old Portuguese', `Golden Age Spanish', or `Post-war Romanian', tacitly implying that these are uniform entities. This is clearly an idealization, as there is social, regional and diachronic variation in any language at any time. At the same time, there is some justi cation for such a classi cation, especially when dealing with literary writing: during culturally comparatively uniform periods, certain stylistic and linguistic norms become the accepted standard, and it becomes almost obligatory for authors to comply with these norms if
62
Chapter 1
they want to gain recognition in society (cf. Baum 2003: 49 50); such norms can remain comparatively unchanged over a period of time. Support for the idea that there is less potential for linguistic change during periods of social and cultural cohesion also comes from anthropological linguistics. According to the punctuated equilibrium model, applied to historical linguistics by Dixon (1997), languages tend to diversify in short periods of intense social upheaval, remaining in a relatively stable state of equilibrium during periods of social stability. However, the diachronic data presented in this study, it will be seen, suggest no such obvious correlation between social or cultural and linguistic stability; on the contrary, the long-term developments investigated here appear to be largely unaffected by changes in society. Variation in content and register between individual texts, combined with the author's personal style and preferences, possible regional elements32 , and learned or foreign in uence33 , has the effect that, even in texts written at virtually the same time, a particular syntactic structure can be far more frequent in one than in the other. As the objective of this study is not the description and characterization of individual texts or authors, but general trends of linguistic development, it is desirable to eliminate the potentially distorting effect of individual texts that can never be a precise representation of the way language was generally used at the time. One way to eliminate such distortion would be to simply not include a text in a graph if its token frequency deviates from that in other texts from roughly the same time period by more than a certain percent-
32
33
The very earliest text in this study, the Cantar de mío Çid, is an example of a text that has been argued to have regional Aragonese features. Menéndez Pidal (1908) believes that it was written in the region of Medinaceli, in the border area between Castile and Aragón. But Smith (1972) claims it is equally likely to have been written in Burgos, by a poeta burgalés . It has also been claimed by some that certain supposedly Aragonese features might be the result of the text being translated from Aragonese. Learned in uence can be expected to be stronger in more formal register, particularly in religious and legal texts; foreign in uence is particularly likely in texts that are translations into Spanish, such as Dichos de Sabios y Philósofos by Jacob Çadique de Uclés translated from Catalan (1402) and De las mujeres ilustres en romance (Zaragoza: 1494) as well as Laberinto de amor (1546), originally written in Tuscan by Boccaccio.
Motivation and theoretical parameters
63
age. The drawbacks of such a method are that any cut-off point would necessarily be arbitrary, and that it would completely ignore the eliminated texts for formal reasons, without taking into consideration that they are, nevertheless, representative of the way the language was used by certain speakers in certain contexts, and that the usage they document is thus equally part of the language as a whole. To avoid discarding any of the gathered data, but to nevertheless generate graphs that are suf ciently smoothed to allow the viewer to visualize the respective construction's long-term development, two different methods of converting the data from individual texts into representations of the general trend of the respective construction's linguistic development are chosen. One standard statistical way of identifying trends is by means of a linear regression line based on the values of a scatter plot. The goal of linear regression is to adjust the values of gradient and of intercept with the y-axis to nd the line that best predicts y from x. This is done by minimizing the sum of the squares of the vertical distances of the points from the regression line. The gradient (m) of the regression and its intersection point with the y-axis (c) can be calculated with the equations P
(yi −y)(xu −x)
m=
i
P
(xu −x)2
and c = y − mx, where
i
P
P
(yi )
y=
i
n
(xi )
and x =
i
n
in which x and y are the mathematical means of the year and the construction's usage frequency, respectively. The disadvantage of using regression lines is that they are by de nition linear, but syntactic development must not necessarily proceed in a linear way. In order not to ignore the fact that the usage of a structure may increase, decrease or remain stable at different times during the investigated period, a second algorithm providing moving averages over shorter time periods is used. The equation for this curve is as follows:
yj0 =
1 9
j+4 P i=j−4
yij
64
Chapter 1
For the beginning and the end of the time scale, where the number of preceding or following texts is insuf cient for the basic equation above, this must be re ned to form averages over correspondingly shorter periods: max[j+4,n]
P
yj0
=
yij
i=min[1,j−4]
max[j+4,n]−min[1,j−4]
The intervals for which this algorithm generates averages are small enough to show any signi cant steps in the curve, but large enough to give it a suf cient degree of smoothness for the viewer to be able to identify a coherent development. In the graphs presenting the development of one single structure on its own, the actual values from the individual texts are shown as points in the diagram; this is generally omitted in comparative diagrams for the sake of clarity. The problem of statistical classi cation A fundamental problem in diachronic statistical analysis is the apparent paradox between having to classify constructions into discrete categories on the one hand, and using the results of this analysis to draw conclusions about changes in language that proceed in an essentially gradual fashion. In order to chart the evolution of a speci c construction, it is necessary to decide, for each individual token, whether or not to count it as an instance of the examined construction or not. Where there is a clear one-to-one correspondence between form and meaning, classi cation is comparatively straight-forward. But meaning is frequently context-dependent; examples (37-40) illustrate how clauses dependent on the verbal phrase tener cuidado `to be careful' can be linked by a number of different prepositions. (37)
Tienen que tener cuidado de poner multiplicadores o have.3pl that have care
divisores
of put.inf numerators
and
comunes a ambos términos de la fracción.34
denominators common
to both
terms
of the fraction
Motivation and theoretical parameters
65
You have to be careful to add the same numerators and denominators in both terms of the fraction. (38)
Los grupos de rock-punk deben tener mucho cuidado con hacer the groups of rock-punk
algo
must
have much
care
with do.inf
35
que pueda ser considerado comercial.
something that can
be considered
commercial
Punk rock groups must be very careful of doing things that might be considered commercial. (39)
Sólo ten
cuidado en no hacer apuestas demasiado
only have.imp care
in not do.inf bets
too
arriesgadas.36 risky
Only be careful not to make too risky bets. (40)
Tenga cuidado para no escoger el mismo color para el have.imp care
fondo
for
not pick.inf the same
colour for
the
y el texto.37
background and the text
Be careful not to pick the same colour for the background and the text. From a truth-conditional perspective38 , it would be dif cult to claim that the choice of preposition in these sentences makes a semantic difference: in all four cases, the respective preposition introduces the matter in which care has to be taken. But merging in nitival clauses with these four prepositions into a single semantic class would nevertheless be a fallacy, as they are frequently in semantic opposition, e.g. in `vuelve
34 35 36 37 38
Lola Archive, Oral Corpus http://www.mtvla.com/news/noticias/articulos_completos/2000/mayo/0518_ deadkennedys.htm http://www.comal.com.mx/horos34.html http://www.aciprensa.com/t-animo.htm Rather than changing the truth-conditional meaning of the dependent clause, the choice of preposition is here determined by pragmatic factors. Para, con and en give a more vivid quality to the dependent clause than the default complementizer de, which is pragmatically neutral. Out of a sample of 667 instances of tener cuidado+de/con/en+in nitive, 90.5% take de, 7.4% en, and 2.4% con.
66
Chapter 1
de/para hacer compras' `she comes back from shopping/to go shopping'. On the other hand, formally equivalent constructions can be used to mean different things, depending on context, or they can be genuinely ambiguous, leaving us with no means of discrete classi cation. A good example of such a structure is de/di/da+in nitive, which in all Romance languages has multiple semantic functions. It can link direct object complements as well as conditional, temporal, causal, and numerous other circumstantial clauses to the main verb, and in many cases neither the construction, nor the pragmatic context, allows an unambiguous classi cation. In the Spanish sentence (41), for example, it is not clear whether the in nitival clause is temporal or conditional; in this context, the distinction is in fact hardly relevant, and it may be doubted whether even the speaker unambiguously means one or the other. (41)
De pensar que a mi niña le prep think.inf that to my girl
pudieran hacer algo
to.her they.could do
something
parecido se me ponen los peso ... pelos de punta.39 similar
refl to.me put
the
hairs prep end
When/if I imagine that they could do something similar to my girl, it makes my hair stand on end. The fact that certain structures cannot be assigned to discrete semantic classes poses a problem for statistical analysis, which by its very nature depends on categorization of the gathered data. There are two simple yet unsatisfactory strategies to overcome this dif culty. One possibility is to count only those instances of the construction that are suf ciently close to one of the semantic prototypes, and discarding all ambiguous tokens. However, keeping in mind that this study is based on the assumption that pragmatically-based syntactic change progresses via a stage of ambiguity, this would mean discarding the tokens carrying the greatest signi cance in this process. An alternative solution, creating a separate class for ambiguous tokens, would also miss
39
Lola Archive, Oral Corpus
Motivation and theoretical parameters
67
the point, as ambiguity is not a discrete category (cf. Section 1.4), and differing degrees of ambiguity are, indeed, a central part of reanalysis. This dilemma, for which there is no simple, wholesale solution, can only be resolved by examining each ambiguous instance of a construction in its context. It can then be decided whether the context provides suf cient additional information to allow us to assign a particular token to one or the other semantic category, or alternatively to consider it genuinely ambiguous, in which case the strategy adopted in this study is to count it as half a token for each of the two respective semantic categories. In this way, a gradual statistical shift from one semantic prototype towards the other can be represented in terms of the two prede ned semantic types. A change in the overall number of tokens of a construction that is undergoing a semantic shift can also provide useful insights. The extension of a construction's use to an increased number of contexts will be re ected by an increased overall frequency of the structure. In particular, this applies to constructions undergoing grammaticalization, as one of the typical features of this process is an increase in obligatori cation (Lehmann, 1985: 307 308), resulting in more frequent usage. In the statistical sections of this book, it will be seen that tracing the overall usage frequency of individual construction types through time allows us to draw conclusions about the causes behind their evolution, in particular by comparing their statistical development, both language-internally and among languages.
Chapter 2 From Latin to Romance
In Classical Latin, the system of complementation and clausal dependency is, in principal, similar to that of modern Romance, in that it includes nite and in nitival clauses as well as other nominal1 dependent structures. Their distribution, however, is somewhat different from that in any modern Romance variety. The fact that the general typology in this area remains similar does not necessarily mean that all present-day structures are inherited, or indeed that all Latin structures have survived in some form or another. On the contrary, as a closer look will reveal, in nitival complementation, for instance, shows a partial or complete loss of its main Latin pattern, the AcI2 , whilst its modern use following a preposition constitutes an innovation. Prepositional in nitives are, in fact, a structure that is patently absent from Classical Latin, but appears to be the driving force behind the overall increase of in nitival dependent clauses. The range of nominal constructions, on the other hand, has undergone dramatic reduction as well as functional shifts, but appears to have escaped complete extinction3 .
1
2 3
I will, generally, treat in nitival structures as distinct from other nominal dependent structures. In how far the Latin and Romance in nitives can be considered non-nominal will be examined in Section 3.4. ACCUSATIVUS CUM INFINITIVO, Accusative and In nitive; see Section 2.2.3 for a more detailed discussion of this phenomenon. It should be noted that Romanian does follow all of these trends, but not always to the same degree.
70
2.1
Chapter 2
Finite subordination
2.1.1 Conjunctional subordination Compared with the developments in the area of non- nite subordination, nite subordination in Latin does not differ greatly from the corresponding systems in modern Romance in its overall structure, though certain differences are worth noting. In addition to the purely functional4 complementizing particles QUOD5 and UT/NE for indicative and subjunctive complements, respectively, there is a range of other hypotactic conjunctions with varying semantic content. The choice of mood6 for the dependent verb V2 is partially predictable from the conjunction, as certain conjunctions entail a certain modality. For instance, the semantics of QUASI `as if' logically require V2 to be irrealis, a modality that generally triggers the use of subjunctive mood. QUONIAM, which expresses an af rmative `because', on the other hand, is closely associated with assertive7 statements. For other conjunctions that require a speci c mood, such as SIMULATQUE+indicative `as soon as', the semantic motivation is less obvious. The majority of conjunctions, however, do not predetermine the mood of V2 , which is instead determined by the semantics and the pragmatic context of the sentence.8 Despite a general correspondence of the indicative mood with assertive statements and of the subjunctive with desiderative, potential, irrealis and interrogative content, this is not an absolute rule. One of the reasons for this lack of straightforward correspondence9 is a gen-
4 5 6
7 8 9
Functional is to be understood as opposed to particles that have semantic content. QUOD can also function as a causal conjunction, and it is not always possible to categorize it unambiguously as one or the other, as shown in sentence (45) below. Verbal mood is generally retained the standard Romance languages, though some (e.g. Romanian, French) have undergone a partial erosion of the corresponding morphological marking. Assertion can be understood as realis through a pragmatic lter: the speaker is convinced of the reality of his statement, or wants to convince the listener of it. Only with these conjunctions does mood actually carry any meaning of its own. Im obliquen Konjunktiv treten ferner (an sich indikativische) Gliedsätze auf... (Bayer & Lindauer, 1974: 210)
From Latin to Romance
71
eral tendency towards analogical levelling, with the subjunctive being the unmarked or default mood for dependent clauses. A similar trend can be observed today, e.g. Spanish `el hecho de que' (`the fact that'), which usually requires use of the subjunctive despite the fact that its complement is, in many cases, very much realis. A further reason for the imperfect correspondence between semantic modality and morphosyntactic mood is the polysemy of UT, which can function as both a desiderative and a consecutive dependency marker. Use of the subjunctive with desiderative UT can be assumed to have spread rst to sentences with consecutive UT, and once consecutive UT-clauses regularly took the subjunctive, to other consecutive conjunctions. This development is supported by the fact that there is an area of overlap between the desiderative and consecutive notions, i.e. wishing for something to happen and causing it to happen. Desiderative UT (42)
ARISIDES NITEBATUR UT COGNOMINE IUSTUS Aristides
SIT
strove
that epithet
just
APPELLATUS
be.3sg.sbjv named
Aristides strove to be given the epithet `The Just'. Consecutive UT (43)
ARISTIDES ADEO EXCELLEBAT ABSTINENTIA, UT Aristides
so.much excelled
COGNOMINE IUSTUS SIT epithet
just
abstinence.abl.sg
APPELLATUS
that 10
was.3sg.sbjv named
Aristides excelled by his integrity to such an extent that he was given the epithet `The Just'.
10
Cit. Ernstberger & Ramersdorfer (1997, L.72)
72
Chapter 2
Other consecutive conjunctions (44)
ARISTIDES ADEO CORRUPTUS ERAT, QUIN Aristides
so.much corrupt
was
COGNOMINE IUSTUS SIT epithet
just
that.not
APPELLATUS
be.3sg.subj named
Aristides was so corrupt that he was not given the epithet `The Just'. A further instance of extension of subjunctive usage is found in complements containing the opinion of someone other than the speaker. (45)
ATHENIENSES SOCRATI Athenians
CRIMINI DEDERUNT,
Socrates.dat.sg crime.dat.sg gave
QUOD ADULESCENTES CORRUMPERET 11 that
adolescents.acc.pl
corrupt.3sg.impf.sbjv
The Athenians accused Socrates of corrupting the young generation. Sentence (45) raises the question what this oblique subjunctive has in common with the core irrealis notion of the subjunctive. Modern Romance usage of the subjunctive with declarative verbs may supply a clue here: it conveys an absence of assertion by a speaker who does not wish to commit himself to the validity of the quoted opinion (46), or even wishes to express his disagreement with it (47). (46)
Cree
que tenga
que regresar con el padre de mi
think.3sg.ind that have.3sg.sbjv that return
with the father of my
hijo...12 son
He thinks he has to return with the father of my son.
11 12
Cit. Bayer & Lindauer (1974: 210) Dr. Love: entre dos amores (http://www.buscamigos.com/drlove/consulta/entre_ dos_amores/home.htm)
From Latin to Romance
(47)
¾Quién dice who
que tenga
que ser como otros dicen?
say.3sg.ind that have.3sg.sbjv that be like
¾Quién dice who
73
que tenga
others say.3pl.ind
que ser transparente?13
say.3sg.ind that have.3sg.sbjv that be transparent
Who says that it has to be like others say? Who says it has to be transparent? The fact that a subjunctival dependent clause introduced by QUOD as in (45) can be either a direct object complement, a causal adjunct, or really something in between the two14 paves the way for a further erosion of the clear pattern QUOD+indicative, thereby facilitating the extension of QUOD to other types of subjunctival clauses and eventually allowing QUOD and its Romance re exes to oust UT/NE completely. The differences with respect to Romance: typological observations The functional merger of the semantically unmarked conjunctions QUOD and UT/NE to QUE15 might be seen as a structural shift in the system of conjunctional dependency marking, but considering that mood is usually marked unambiguously on the verb in most Romance varieties, the merger is merely a reduction of redundant mood marking. French and Romanian are two notable exceptions to this: whilst in French the present subjunctive is differentiated from the indicative only in comparatively few forms (Posner, 1996: 141) as the result of regular phonological developments, which in connection with the absence of mood-speci c subordination particles has led to a de facto
13 14 15
Diario La Hora, section Nacional (http://www.lahora.com.gt/26-01-01/paginas/ nac_2.htm) This further exempli es the impossibility of making binary distinctions between complements and adjuncts. QUE stands for the varying modern re exes of QUOD, que, che, c , probably derived from QUI(D), which came to replace QUOD. It has been argued that Romanian c may be derived from the QU(I)A (Herman, 1963: 166), which, like QUOD, was a causal conjunction in Classical Latin. The etymological origin of the conjunctional dependency marker is, however, not the focus of this discussion.
74
Chapter 2
loss of a consistent overt mood opposition, Romanian mood is also syncretized in large parts of the verbal paradigm, but overt mood distinction is maintained by means of two distinct conjunctions, c for the indicative and s 99% (580)
subjunctive > inflected infinitive > bare infinitive Givón's (1984: 519) observations about semantic and syntactic integration of main and dependent clause, in which the scale [-integrated] < > [+integrated] roughly corresponds to the FINITE < > NON-FINITE scale regarding the complement verb, provides additional
The in nitive in Spanish and Portuguese today
191
support to this scalar model by intertwining morphology, syntax and semantics. Regarding the in nitive, Vincent (1999: 1 2) points out, this scalar model allows for an approach that integrates both the traditional view that the in nitive is essentially nominal69 , and the Generative view that it is essentially verbal or sentential, as in nitival clauses are understood to belong to the category IP, despite the fact that they are headed by an I that is [-T, -AGR]. Vincent proposes a hybrid solution and argues that the thing we label the in nitive can have different properties in different languages and at different historical moments. What we can call the r-form of the verb in Latin and Romance originates as a nominal element, develops verbal properties in Latin and as it passes into Romance comes increasingly to show clausal properties. Shortcomings of the one-dimensional continuum model While this scalar model is, if understood as an idealization, a clear step forward because it replaces the unnecessarily restrictive and unrealistically idealized binary model, it can be argued that the one-dimensional scalar model is fundamentally awed because it attempts to arti cially subsume a variety of features under a single label. Labels such as nominal and verbal, nite and non- nite stand for combinations of several features. [+nominal], for instance, can be characterised by genitival subject, nominal in ectional morphology, by the absence of typical verbal morphology and morphosyntax (e.g. tense, voice, person agreement), by its use in typically nominal syntactic environments, e.g. with prepositions, etc. Similarly, nite is often equated with tensed, but languages such as Latin do have tensed in nitives.70 An alternative de nition, dis-
69
70
L'in nitif sert à exprimer la notion verbale sans plus, sans considération de personne, de nombre ni de mode, le plus souvent aussi sans considération de temps et de voix. Si on quali e souvent l'in nitif de `forme substantive du verbe', on veut dire par là à peu près la même chose... , Sandfeld (1965: 1). ... the terms [tensed and nite] are not equivalent, since e.g. a language may have in nitives which are also in ected for tense. , Matthews (1997: 375).
192
Chapter 3
tinguishing in nitives from nite verb forms on the grounds that the latter are in ected for person and number (Matthews, 1997: 129), showing subject agreement (Koptjevskaja-Tamm, 1993: 256) is similarly awed, as shown by the `in ected in nitive' examined in Section 3.3. In the one-dimensional scalar model, any non-prototypical verb form is located somewhere between the two ends of the scale. However, due to the differential nature of the various criteria and features involved, the exact position must, in the absence of a uniform method of quanti cation, necessarily be subjectively or arbitrarily determined. Not only is it doubtful whether morphological, syntactic, and semantic features can ever be merged into a single `value' that would allow exact positioning along the scale, but such an approach would, in fact, distort the real facts by disregarding the disjunction between the individual features of a particular verb form. An alternative approach, in which no attempt is made to determine an arti cial degree of overall nominality, niteness, etc., but which instead concentrates on each feature individually, can provide valuable insights into the shifts in the area of dependent clauses from Latin to Romance. 3.4.2 Classi cation of the Latin in nitive There appears to be a general consensus that the in nitive has developed from being relatively nominal in Latin to something much less nominal in modern Romance. Brugmann (1888 95) shows that Latin in nitives have their origin in Indo-Europaen nomina actionis, full nouns that describe an action or state. Though he sees the in nitive as the nal result of an evolutionary process, he does recognize that, during the transition, they passed through a number of intermediate stages, and hence it is often hard to say whether any particular form should be called an in nitive in the strict sense of the word. This view is strikingly similar to that expressed by Vincent (1999: 5; cf. Section 3.4.1), except that it refers to a different period and the end points of the scale are not exactly the same; for Brugmann, it appears, there is a continuum between the extremes noun and in nitive.
The in nitive in Spanish and Portuguese today
193
Beardsley (1920: 1) states that in Latin the in nitive showed close similarity to the noun , but he quali es the statement by adding that it was capable of use only in the nominative and accusative cases. 71 Vincent (1999: 5) also implies that the Latin in nitive is more nominal than its modern Romance counterpart, claiming that ... the [Romance] in nitive, [is] a by now purely verbal form. (my emphasis) An objective analysis of the in nitive in Latin and Romance casts some doubt on the view that it has undergone a straight-forward shift along a nominal/verbal continuum. Non-verbal features of the Latin in nitive On the one hand, there are, indeed, certain features that give the Latin in nitive a stronger nominal quality in comparison with its later re exes in Romance. One of these is the syntactic restriction on its use, which allows it to appear only as the subject or object of a sentence. Though this is not a typical feature of Latin nominal elements in general, it does, nevertheless, show a distributional dependency on case structure, which is something typically associated with NPs. Secondly, the fact that the semantic subject of the Latin in nitive does not take nominative case is an indication that the in nitive is not a prototypical verb form. Though the subject in AcI-clauses appears in the accusative (not the genitive, as it does with prototypical nominals), the fact that the in nitive does not assign nominative case to its subject can be interpreted as an indication of reduced verbal character. Verb-like or clausal features of the Latin in nitive Whilst not as verbal as nite verb forms, the Latin in nitive does have clearly verb-like properties. It can appear, on its own as well as within
71
He presumably refers to the fact that it could only function as the subject or direct object of the main clause.
194
Chapter 3
an AcI, as the complement of main verbs that semantically require a sentential object, such as RECUSARE `to refuse' and COGERE `to coerce'. The fact that Latin does not have prepositional in nitives is a further mark of the in nitive's verbal nature, as prepositions typically occur with nouns or NPs. Though lacking morphological person and number agreement, the Latin in nitive does have morphologically in ected forms for tense and voice, both of which are typical verbal in ection classes. A further typically verbal characteristic is the fact that the in nitive cannot be pluralized a feature it shares with nite verb forms72 as well as the gerund, as neither Latin nor Romance verbs are marked for the number of actions or states they represent.73 So far, these observations are in line with locating the Latin in nitive somewhere along the continuum between prototypical noun and verb, though perhaps rather closer to the verbal end than traditional grammars would suggest. 3.4.3 Is the Romance in nitive verbal or nominal? As mentioned in Section 3.4.2, there appears to be a general consensus that the in nitive has become more verbal, and implicitly less nominal, in the course of the development of the Romance languages. This section will take a closer look at the degree to which such a claim can be upheld. This analysis will separately consider the evolution of the in nitive clause's relation to outside elements, in particular the main clause it is dependent on, on the one hand, and its internal structure on the other hand. It will become apparent that we are dealing with two di-
72 73
The so-called plural forms of nite verbs do not pluralize the verb itself, but merely constitute number agreement with the subject. This does not mean that a verb cannot express more than one action; aspectual marking by means of the `imperfect tense' or by auxiliaries such as SOLERE can be used to express repetitive habituality, and in clauses with a plural subject it is frequently the case that the verbal action is executed individually by each member of the subject NP. In such cases, the verb represents a multiple action, but there is no speci c morphology to distinguish single from multiple actions.
The in nitive in Spanish and Portuguese today
195
verging trends, and that it is therefore inappropriate to speak of a single development towards or away from nominality. Increased nominality in relation to the main clause The overall number of contexts in which the in nitive can occur has steadily increased to the present day. Much emphasis is placed on its increased coreferential usage. Harris (1978: 226) states that the in nitive has extended its role in Romance compared to Latin [...], particularly when the subject of the two underlying sentences is identical. This is doubtlessly an important observation, but nevertheless merely a quantitative shift, as coreferential in nitive complements did exist in Latin, too. More innovative is the use with prepositions. The details of this development will be discussed in the following chapters. At this point it is suf cient to point out that the creation of a novel syntactic pattern, the prepositional in nitive, is a clear shift towards nominality, as the prototypical prepositional structure is undoubtedly [preposition+noun]. Pre-empting some results from subsequent sections, it can be observed that the overall number of prepositional in nitives, as well as the number of different prepositions the in nitive can combine with, has seen a continuing rise to the present day, which can be interpreted as an indication that, in relation to its main clause, the in nitive is still in the process of becoming more nominal.74 The noun-like qualities of the in nitive are also pointed out by Raposo (1987: 239), who argues in nitival clauses resemble nominal projections in terms of Case Filter, and that the in nitival morpheme -r nominalizes the grammatical category to which it attaches . This, he claims, is shown by the fact that in nitival clauses cannot appear
74
An interesting issue is whether those prepositional in nitives that have evolved from prototypical adjuncts to complements (cf. Chapter 7) have reversed the trend, becoming less nominal in the process. According to the theoretical parameters presented in Section 1.3, the construction remains the same, the shift being a purely semantic one.
196
Chapter 3
in a position subcategorized by non-Case assigning categories, such as nouns (227). (227) *o receio [chumbar o exame].75 the fear
fail.inf
the exam
the fear [failing the exam]. Instead, a dummy Case-assigning preposition, de, is required for in nitival clauses (228), as it is for NPs (229). (228) o receio de [chumbar o exame].76 the fear
of fail.inf
the exam
the fear of [failing the exam]. (229) o receio de [os exames]77 the fear
of the exams
the fear of [the exams] A further area in which the in nitive's verbal character has decreased is in the area of overt marking of the time relation between the main clause and the in nitive. Whilst Latin in nitives are obligatorily tensed a typically verbal feature modern Romance in nitives are, by default, not overtly tensed. Anteriority and posteriority of the in nitive can be rendered periphrastically by the standard auxiliary constructions78 of the respective languages, but in modern Romance this only occurs as a disambiguating strategy in marked contexts. A reduction of the need for morphological temporal marking can even be observed in Spanish since the Middle Ages: whilst prepositional in nitives marked as anterior by después de were usually marked analytically as past by means of haber, a steady decline of this redundant marking for relative time reference can be observed, and in the modern language only a fraction
75 76 77 78
Raposo, 1987: 237 ibid.: 238 ibid. When used in in nitival clauses, the in ectable verbal element within the auxiliary construction appears in its in nitive form, e.g. the in nitive of `to have' followed by the past participle for anteriority.
The in nitive in Spanish and Portuguese today
197
of prepositional in nitives with después de use the past tense auxiliary construction. The same also applies to marking of voice, which the Latin innitive is obligatorily in ected for, but which is only exceptionally expressed overtly in modern Romance, whilst usually determined pragmatically through contextual cues. Increased verbal character within the dependent clause Within the dependent clause, the situation is rather different, both morphologically and syntactically. In the majority of Romance languages, in nitival clauses have developed an internal structure that, under certain circumstances, allows the subject of the in nitive to appear overtly, and in subject case. This Overt Subject In nitive or OSI is found throughout the Romance-speaking world (cf. Section 3.2), in several varieties with the possibility of morphological agreement between the in nitive and its overt or non-overt subject a fact that casts some doubt on whether the absence of person agreement is a necessary characteristic of the in nitive.79 Both these features morphological person/number agreement and assigning nominative case to its subject are typically verbal. A further verbal characteristic is the case assigned to semantically object-like arguments of the in nitive. If we were dealing with a noun, we would again expect a genitival construction with de, as shown in (230), while the same argument appears as a direct object in the in nitival construction. Compare the de-construction for the object of the noun búsqueda in (230) with the direct object construction for the object of buscar in (231):
79
To claim, on such formal grounds, that the `personal in nitive' is not really an in nitive at all would be misleading, as such a claim would miss the fact that its syntactic behaviour and its distribution is very similar to that of OSI constructions with non-in ected in nitives in those varieties that do not allow morphological agreement of the in nitive.
198
Chapter 3
(230) La búsqueda de [un cohete sucesor] se complica.80 the search
de a
rocket successor refl comlicate.3sg
The search for [a successor rocket] is becoming more complicated. (231) El gobierno se esfuerza
por buscar[lo]DO / [un cohete
the government refl make.effort.3sg for search.inf.it.do
a
rocket
sucesor]DO . successor
The government is making an effort to nd [it] / [a successor rocket]. The degree of nominality in the medieval language Beardsley (1920: 3 12) devotes an entire chapter of his survey of the innitive in Old Spanish to the in nitive as substantive , i.e. functioning as a noun, which is more widely used in Medieval Spanish than in the modern language, but also admits that there is no distinct line between the verb and noun uses of the in nitive, since even in the types called verbal the substantival sense can be felt, and vice versa (Beardsley, 1920: 13). The most obvious distinction between the two uses is presence or absence of an article or adjectival pronoun. Expressions like el morir era malo 81 and en su versi car82 convey an implicit dynamism of the respective event by semantically exploiting the verbal nature of the innitive, whilst sticking to a nominal syntactic structure. Similar examples abound in Medieval Portuguese texts, too:
80 81 82
El País, 10.1.2001 Cit. Beardsley (1920): Libro de Alixandre 1052 ibid.: 232, 2077
The in nitive in Spanish and Portuguese today
(232) pois aquestes olhos meus por el perderan then these
eyes
199
o dormir83
mine for him lose.3pl.pret the sleep.nomlz
Thus these eyes of mine lost the capacity to sleep because of him. A number of these nomina actionis extend their semantic range beyond the action itself to its object. El/o cantar `the singing'> 'the song', el/o comer `the eating'>'the meal', el/o aver `the owning'>'the possessions', el paresçer `the appearing'>'the opinion' exemplify this semantic extension, with some of the secondary meanings so far removed from the original meaning of the verb that they are reanalysed as fully independent lexical items, as can be seen with el/o parecer `the opinion', el/o andar `the walk' and el/o poder `the power' from medieval times to the present day: (233) Levadeiro dos moinhos com poder de acoimar os water.of cer of.the mills
tomasem
que
with power of denounce those that
a agoa.84
take.3pl.sbjv the water
An milling water of cial with the power to name and shame those who take the water. However, even the highly noun-like use of in nitive form in Medieval Spanish and Portuguese has a varying degree of nominality with regards to its arguments. As would be expected with a noun, both its subject and direct object are usually rendered genitivally by means of the preposition de, which indicates that this use of the in nitive is more noun-like than in Latin:
83 84
Nuno Perez Sandeu, Cantiga de Amigo 6, end of 13th century. Vereações no. 1309, . 48
200
Chapter 3
Genitival Subject: (234) El cuydar de los omnes todo es vanidat.85 def care.inf of the men
all
is vanity
The worrying of man is all vain. Genitival Object: (235) el usar de sus leyes...86 def use.inf of their laws
the use of their laws... There are, on the other hand, also instances of direct objects without de, giving the in nitive a more verbal touch despite the fact that it has an article: (236) mando [...] el refazer
los muros de las uillas87
order.3sg.pret def remake.inf the walls
of the towns
He gave orders to rebuild the walls of the towns. This construction, which is halfway between nominal and verbal use, survives into the modern language: (237) Tendrá
la consideración de estudiante el extranjero cuya
have.3sg.fut the consideration
of student
the foreigner
venida a España tenga
como n
único
arrival to Spain
85 86 87
have.3sg.sbjv as
whose
o principal
purpose exclusive or principal
Cit. Beardsley (1920: 11): Libro de Alixandre 968 Cit. Beardsley (1920: 12): Primera Crónica General 103b19 Cit. Beardsley (1920: 12): Primera Crónica General 293b32
The in nitive in Spanish and Portuguese today
el cursar
o ampliar estudios o realizar
def pursue.inf or expand.inf studies
201
trabajos de
or carry.out.inf works
of
investigación o formación.88 investigation
or education
A foreigner whose coming to Spain has as its only or principle objective to pursue or deepen his studies or to carry out research or educational work will be considered a student. The in nitive preceded by a contraction of de nite article and the preposition a deserves special attention in view of its later development. Beardsley rightly includes this construction in the `in nitive as substantive' class, implicitly equating constructions such as (238) with other prepositional constructions such as (239). (238) Al posar [...] todos se esperaban.89 at.the rest.inf
all
refl wait.3pl.impf
When they rested... everyone was waiting. (239) Del minguar
dize...90
of.the diminish.inf says...
About discrediting he says... At the time, this al-construction is no more verbal than any other in nitive with a de nite article, as its genitival subject (al so mandar91 , `on his orders') and object (Al cargar de las archas veriedes gozo tanto.92 `at the loading of the chests') underlines. The subsequent development, in both Spanish and Portuguese, to the present-day situation, in which al/ao+in nitive is a clearly verbal construction (with a particularly high incidence of overt subjects), is exceptional in that it has formally retained the de nite article. Nevertheless, it has to be included in the list of prepositional in nitives examined in their development between the Middle Ages and today in the next chapters.
88 89 90 91 92
El País, 10.1.2001. Beardsley (1920: 5), cit. Milagros de Nuestra Señora 8 ibid., p. 5, from Primera Crónica General 66a10 Cit. Beardsley (1920): Primera Crónica General 66b41 Cit. Beardsley (1920: 12): Çid 170
202
Chapter 3
Though the `nominal in nitive' is not the primary object of this study, it helps illustrate that we are not dealing with a single, uniform shift of the in nitive along the [nominal]< >[verbal] continuum. Overall, it may be most appropriate to classify the in nitive used in this way as a noun, but closer examination has shown that its range, from lexicalized full noun that has lost all its verbal force, to the incipiently verbal use with a non-genitival direct object, spans a considerable spectrum. `Verbal' usage of the in nitive in medieval texts is, in principle (though not in all details, as the following chapters show), similar to that in the modern language. With regards to its `external' relationship to its main verb, it can function as subject, object, or as various types of adjunct. The internal structure of in nitival complements has not changed dramatically since medieval times, either; in a very verb-like way, the in nitive can have direct objects and overt subjects.93 In Portuguese, the in ected in nitive, agreeing morphologically with an overt or implicit subject, is also already well established in the Middle Ages. It is only in terms of distribution and quantity that the usage of in nitival clauses has changed since the Middle Ages; these changes, which will be examined in some detail in the following chapters, offer crucial insights into an ongoing process, which is already well underway in the 12th century. Impact of the demise of the Latin case system The previous section shows that it is impossible to say that the Romance in nitive, as a whole, has become more nominal or more verbal. In relation to the main clause, nominal features have become more pronounced, but clause-internally, a clear increase in verbal characteristics is evident. One factor that favours the increase in nominality is the overall structural shift that Romance has undergone.
93
Some changes have occurred regarding the possible position of overt subjects, cf. Section (3.2.2) and Mensching (2000: 15 38, 101 28)
The in nitive in Spanish and Portuguese today
203
(Classical) Latin syntax depends heavily on overt in ectional marking of the relations between constituents, rather than relying on word order or the consistent use of lexical relation markers94 . Prepositional constructions are no exception to this requirement. IN `in' and SUB `below, under', for instance, have locative meaning with the ablative, but typically95 directional meaning with the accusative. The Latin in nitive, however, does not participate in standard nominal morphology, which makes it impossible to overtly specify its semantic relation to other constituents. As a consequence, it can only be used in place of an NP where its semantic role within the sentence can nevertheless be unambiguously identi ed, as the subject or direct object of certain verbs that make the semantic role of the in nitive highly predictable (cf. Section 2.2). Such a degree of predictability is not provided for other semantic roles, which therefore require unambiguous in ectional marking96 . This general requirement for oblique nominals to be in ected for case can be seen as crucial for the fact that in nitives cannot appear in prepositional phrases in Latin. With the Romance shift away from morphological case marking (Penny, 1991: 101 ff.), nominal morphology conveniently adapted to that of the in nitive: whether direct, indirect, or prepositional object, neither nouns nor in nitives are morphologically in ected for case in most modern Romance varieties, and it appears that this structural shift has, more or less per chance, led to a morphosyntactic convergence of nouns and in nitives, thereby facilitating a greater syntactic overlap between the two in certain areas. The deviant pattern in Romanian, which does have a two-term morphological case distinction97 , supports this structurally based theory. Languages without case in ection, such as Spanish, tend to have a single in nitive form that can function both as the verb of a dependent
94 95 96 97
Classical Latin does make use of lexical relation markers, but the inventory is defective, with many inter-constituent relations requiring morphological in ection. IN with the accusative very frequently has the gurative meaning `against'. A number of prepositions do, however, also assign semantic roles in an unambiguous way. Romanian distinguishes one form for the nominative and accusative, and a second oblique form. This overt distinction applies to all articulated forms as well as non-articulated feminine singular forms.
204
Chapter 3
clause and as a full noun or in nitive as substantive (Beardsley, 1920: 1 12). These two usages of the in nitive can be distinguished by the different syntactic contexts they require. Sentence (240) is an example of the in nitive obrar `to work, to function, to act' used as a full noun. Both the article el and the fact that its semantic subject is expressed genitivally by means of the preposition de98 are clear indicators of nominality: (240) España es hoy un país Spain
is today a
libre. Libre [...] como nunca lo fue
country free
free
like
antes en su historia, libre [...] en [el obrar before in its history,
free
never
it was
de sus
in the work.inf/nmlz of its
partidos políticos].99 parties
political
Today, Spain is a free country. Free as it has never been before in its history, free in [the working of its political parties]. Sentence (241) shows the same in nitive form in a prototypically verbal syntactic structure, with the auxiliary verb poder: (241) Nunca pensé never
que Israel pudiera obrar así.100
think.1sg.pret that Israel could
act.inf so
I never thought that Israel could act like that. In Romanian, this syncretism between verbal and nominal in nitive is resolved morphologically. Verbs have a `short in nitive', corresponding to the verbal character of obrar in sentence (241), and a `long in nitive'101 , which is in all syntactic and morphosyntactic respects a fully nominal nomen actionis that participates in normal (feminine) case morphology.
98
99 100 101
With regard to direct objects, even the more noun-like in nitive does not usually require `genitival' de. Compare `el negar la existencia' but `la negación de la existencia.' El País, 8.2.2001. El País, ibid. `In nitivul scurt' and `in nitivul lung', respectively.
The in nitive in Spanish and Portuguese today
205
Verbal or `short' in nitive: (242) Niciodat nu m-am never
gândit c cineva poate vopsi ou
not refl-have.1sg thought that somebody can
dye.inf eggs
a³a. so
I never thought anybody could dye eggs like that. Nominal or `long' in nitive (non-oblique): (243) Am
adus vopseaua pentru vopsirea
have.1sg fetched dye.def
for
ou lelor.
dye.nmlz.def.nobl eggs.def.gen.pl
I have brought the dye for the dyeing of the eggs. Nominal or `long' in nitive (oblique): (244) Am
adus vopseaua din cauza vopsirii
have.1sg fetched dye.def
ou lelor
out.of reason dye.nmlz.obl eggs.def.gen.pl
care are loc aici. that has place here
I have brought the dye because of the dyeing of the eggs that takes place here. The verbal character of a vopsi in sentence (242) is particularly obvious in the presence of an overt subject in the dependent clause (245), which appears in the nominative: Verbal or `short' in nitive with overt subject: (245) Am
adus vopseaua pentru a.vopsi ³i noi
have.1pl fetched dye.def
for
ouale.
dye.inf also wenom eggs.def
I have brought the dye for us to dye the eggs, too. Like Latin, Romanian has prepositions that require a speci c case.102 And, as in Latin, these prepositions cannot be used to form prepositional
102
Traditionally, these prepositions are subdivided into those requiring the genitive and those requiring the dative (Pop & Moldovan, 1997: 207), but this distinction
206
Chapter 3
in nitives, whilst in Romanian prepositions requiring non-oblique case frequently take an in nitival clause. For example, Romanian does not have causal in nitival clauses, as both causal prepositions, din cauza and din pricina, assign oblique case.103 Other Romance languages with a less developed case structure permit prepositional in nitives, even with cognate prepositional expressions. Compare the grammatical Spanish construction (246) and the corresponding ungrammatical Romanian sentence (247): (246) Fácilmente podríamos pensar que el Dios cruel [...] mató easily
a
we.could
think
that the God cruel
killed
24.000 personas solamente por causa de haber
domkr 24,000 persons
only
for reason of pst.aux.inf
fornicado.104 fornicated
We can easily think that the cruel God killed 24,000 people only because they had fornicated. (247) *Cu u³urinµ putem gândi c Dumnezeu crud a omorât pe with ease
we.can think that God
cruel has killed 105
24.000 de persoane, numai din cauza (de)
a.
24,000 of persons
pst.aux.inf
only
out.of reason
domkr
p c tuit. sinned
We can easily think that the cruel God killed 24,000 people only because they had sinned. The evidence thus suggests that the disappearance of case in ection after prepositions is a prerequisite for the emergence of prepositional in-
103
104 105
is an arti cial one, as there can be no morphosyntactic distinction between the two cases in a prepositional environment. A similar resistance to concessive in nitival clauses can be found with în ciuda, but as discussed in Section 6.3.2, a comparable resistance in other Romance languages suggests a different cause. http://www.goodnews.or.kr/buenasnuevas/sermon/ley.htm It would appear tempting to copy the Spanish prepositional expression by adding de here. In fact din cauza does, very rarely, combine with de+unoblique case, but never with de+in nitive.
The in nitive in Spanish and Portuguese today
207
nitives, and that once the requirement for morphological case marking no longer exists, the hybrid nominal/verbal nature of the in nitive allows it to function syntagmatically like an NP. The disjunction between this external nominality and its internal verbal character is crucial in allowing in nitival clauses to function as a nominal constituent, but simultaneously to impinge upon the domain of nite dependent clauses. With these necessary structural conditions in place, it was possible for a new type of in nitival construction to emerge and spread; the exact path that this development has taken will be investigated in the following chapters.
Chapter 4 Diachronic development of the in nitive in Spanish
4.1
Prepositional and non-prepositional in nitives
Fig. 4.1 shows the overall development of in nitive usage in Spanish since the Middle Ages. Each point represents one text or author (as listed in appendix A), sorted diachronically along the x-axis, with the position along the y-axis re ecting the number of in nitives per 100,000 words of text. The approximation curve, generated by the algorithm described in Section 1.5.2, shows the overall development trend.
per 100000 words
4000 3500 3000 2500 2000 1500 1000 1100
1150
1200
1250
Figure 4.1
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Overall frequency of in nitives in Spanish through time
Up to the beginning of the 16th century, a remarkable stability at around 2250 in nitives per 100,000 words can be observed. At this point, a relatively steady rise sets in, reaching an average of around 2800 per 100,000 words in the 20th century. This represents a signi cant rise of approximately 25%. In this chapter, the development of the various in nitival constructions this graph incorporates will be analysed and compared in some detail. A fundamental distinction can be made between prepositional and non-prepositional in nitives. Figs 4.2 - 4.4 show the differential overall development of these two construction types. A strong increase in the use of prepositional in nitives sets in around 1500, peaking just over a century later. This increase from ap-
prep. infinitives per 100000 words
210
2000 1750 1500 1250 1000 750 500 250 0 1100
Chapter 4
1150
1200
1250
1300
Figure 4.2
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
1900
1950
2000
1950
2000
Spanish prepositional in nitives through time
2500 per 100000 words
2250 2000 1750 1500 1250 1000 750 500 1100
1150
1200
1250
1300
Figure 4.3
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
Spanish non-prepositional in nitives through time
per 100000 words
1600 1400 1200
plain infinitive total prep. inf.
1000 800 600 1100
1150
1200
Figure 4.4
1250
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
Spanish prepositional vs. non-prepositional in nitives through time
Diachronic development of the in nitive in Spanish
211
prox. 600 per 100,000 words to approx. 1550 per 100,000 words represents a rise of nearly 160%. To some extent, this is counterbalanced by reduced usage of non-prepositional in nitives, dropping well below the level of prepositional in nitives during the same period, from approx. 1700 per 100,000 words to approx. 1000 per 100,000 words. Though this fall does not fully cancel out the increase in prepositional in nitives, it does show a fundamental shift in favour of preposition+in nitive. However, the extreme divergence during this period is only temporary. After acquiring almost equal shares towards the end of the 17th century, prepositional in nitive usage subsequently increases relatively steadily up to the beginning of the 20th century, whilst non-prepositional in nitives experience a slight decline up to this point, followed by a sudden rise after 1900, again bringing about an almost equal share of prepositional and non-prepositional in nitives in present-day Spanish. The most important observation, however, is the overall development, as illustrated by the linear regression lines in Figure 4.5.
per 100000 words
1600 1400 1200 1000
plain infinitive total prep. inf. Linear regression, plain infinitive Linear regression, total prep. inf.
800 600 1100
1150
Figure 4.5
1200
1250
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish prepositional vs. non-prepositional in nitives through time: linear regressions
Essentially, in medieval times, the use of the plain in nitive exceeds that of prepositional in nitive by a factor of nearly three, whereas there is no longer a signi cant numeric difference today. If at all, the prepositional in nitive can be said to have been slightly dominant since the mid-16th century. In the following sections, the increase in the usage of prepositional in nitives will be examined in detail.
212
4.2
Chapter 4
Prepositional in nitives
Two different classi catory approaches to prepositional in nitives will be taken. The rst is lexically based, tracing the use of individual prepositions through time. The second is semantically based, which means that prepositions are grouped together depending on their meaning, or more precisely, on the semantic value that the dependent clause they introduce has within its sentence. Both approaches have drawbacks. The formal, morphological approach ignores the fact that individual prepositions can, themselves, be polysemous, as discussed in Section 1.5.2. Is conditional de really the same preposition as de meaning `about' or de meaning `beginning with'? Most obviously, the de as a complementizer clearly differs from the other uses. Diachronic shifts must also be taken into account. Thus por and para1 are diachronically unstable; por gradually cedes its nal meaning to para.2 It must, therefore, be critically considered whether por in the Middle Ages can be seen as the same preposition as por used today. However, it is exactly this kind of semantic overlap and lack of discreteness that causes changes in usage patterns (cf. Section 1.3, Chapter 7) , and it is thus necessary to obtain an overview of their development as a whole. The semantic approach, on the other hand, is limited by the same factors, as it is often not possible to sort prepositional in nitives into discrete semantic categories. Whether en hacerlo `in doinf -it' is temporal or causal3 varies from context to context, but in the majority of cases the meaning is a combination of both, in varying proportions. Thus, any categorization involves arbitrary or subjective decisions as to which is the primary or central meaning in a speci c context. But a semantic classi cation is, nevertheless, of interest for the purpose of evaluating the status of the prepositional in nitive construction through time. In particular, the correlation between semantic bleaching and increased fre-
1 2 3
Pora and pera initially also participate in this complex shift. For a study about the origins of their current distribution, see Riiho (1979). The causal meaning component can acquire a distinctly conditional avour in a context of future reference.
Diachronic development of the in nitive in Spanish
213
quency provides valuable information about the way in which an originally pragmatic mechanism ultimately leads to syntactic change. How this type of shift occurs was discussed in Section 1.5.2. 4.2.1 Semantically underspecified constructions The general trend for all prepositional in nitives is to increase in frequency. One construction suffering a (slight) overall decline is en+innitive (Fig. 4.6), which initially follows the general trend, rising from approx. 62 per 100,000 words to a peak of nearly 100 per 100,000 words, but subsequently falls to below 50 per 100,000 words in present-day Spanish. incidence per 100000 words
150 125 100 75 50 25 0 1100
1150
1200
1250
1300
1350
1400
1450
1500
1550
1600
1650
1700
1750
1800
1850
1900
1950
2000
year
Figure 4.6
Spanish en+in nitive through time
This can be explained by the rise in alternative, semantically more precise constructions which have come to replace en+in nitive. As illustrated above, en+in nitive conveys a somewhat imprecise notion of temporal simultaneity and causality/conditionality. In Section 2.4.5 it has been argued that such underspecification of semantic relationships is typically found when events are pragmatically backgrounded, and that a way of achieving such pragmatic backgrounding is by using gerundial constructions rather than prepositional in nitives. It might thus be expected that the decline of en+in nitive is counterbalanced by an increase in clausal gerunds. As Fig. 4.7 shows, this is not the case. Their development greatly resembles that of en+in nitive, peaking slightly later, but then also declining to a level below that of medieval times.
clausal gerunds per 100000 words
214
Chapter 4
1000 800 600 400 200 0 1100
1150
1200
1250
1300
1350
Figure 4.7
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish clausal gerunds through time
This leads to the conclusion that underspecified pragmatic backgrounding has, to some extent, declined in favour of more semantically precise, more foregrounded clause types, i.e. conjunctional and/or innitival clauses. In the following, it will be examined in how far, and in what way, prepositional in nitives have gained dominance, both in comparison to the underspecified, backgrounded gerund and en+in nitive, and in comparison to nite dependent clauses introduced by the conjunctions that semantically correspond to the respective prepositions. 4.2.2 Decreasing frequency of por + in nitive
incidence per 100000 words
Por+in nitive is also subject to falling usage frequency through time, as shown in Fig. 4.8: 250 200 150 100 50 0 1100 1150
1200
1250 1300
1350
1400
1450 1500
1550
1600 1650
1700
1750 1800
1850
1900
1950 2000
year
Figure 4.8
Spanish por+in nitive through time
As explained in the introductory section of this chapter, por has been subjected to a reduction of its semantic range, as the separate
Diachronic development of the in nitive in Spanish
215
preposition para, a compound of por and a, both of which could carry nal meaning at the time, came into existence in the Middle Ages. A comparaison of por and para (pera, pora)+in nitive (Fig. 4.9) suggests that the fall in por+in nitive is linked to the rise in para+in nitive
per 100000 words
300 para,pera,pora por
250 200 150 100 50 1100
1150
1200
1250
1300
1350
Figure 4.9
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish por+in nitive vs. para+in nitive
Almost until the 15th century, the ratio remains relatively stable at approximately 2.5:1 in favour of por. But from around 1400 onwards, an increasingly steep rise in the frequency of para causes it to draw even with por by the second half of the 16th century, despite a simultaneous rise of por by around 35%. At this point, it can clearly be seen, para has got the upper hand: a sudden drop in the use of por is paralleled by a similarly steep increase in para, both of which subsequently slacken, arriving at a largely stable ratio of approx. 3.5:1 in favour of para, from around 1800 onwards. The detailed study of por and para by Riiho (1979: 235 257) reveals that during the Classical period , i.e. the 16th and 17th century, para gradually takes over the nal domain, particularly in the semantic area he calls nalidad propiamente dicha , i.e. ` nality in the strict sense'. Whilst por is used for dependent clauses with a combination of nal+causal and nal+substitutional meaning to the present day, the purely nal use becomes increasingly rare during this period, and only survives sporadically afterwards.4 This semantic shift is responsible for the sudden change in frequency during the 17th century.
4
One author who uses nal por regularly as late as the end of the 19th century is Emilia Pardo Bazán in Los pazos de Ulloa (1886, 2a ed. Madrid: Cátedra, 1999).
216
Chapter 4
It is important to note that the decline of por+in nitive is, thus, not motivated by wider structural factors, but caused by a local morphological and lexico-semantic shift. The overall trend of the uid por/para/pora-domain is illustrated by Fig. 4.10, which shows their combined frequency:
incidence per 100000 words
700 600 500 400 300 200 100 1100 1150
1200
1250 1300
1350
1400
1450 1500
1550
1600 1650
1700
1750 1800
1850
1900
1950 2000
year
Figure 4.10
Spanish por/para/pora+in nitive
The picture provided by this graph is fairly straightforward. Relative stability before the 15th and after the 18th century contrasts with an increase from approx. 170 per 100,000 words to approx. 350 per 100,000 words, i.e. a rise by more than 100% between the 15th and the 18th century. A drop by approximately 15% in the course of the 20th century can also be observed. This development pattern, it will be seen, is a typical one found with a number of other prepositions. 4.2.3 Increasing frequency of existing prepositional in nitives The majority of prepositional in nitives occur with increasing frequency through time. Two groups can be distinguished: a number of prepositional in nitives rst appear during the time period under investigation; these can be traced from their very beginning. The other group is already present in the earliest documents, and the information gleaned from the rst group may be of assistance in reconstructing their origin. The subset of prepositional in nitives already present in early medieval Spanish consists of por/para, con, de, en, and a. Their development is illustrated in Fig. 4.10 above and in Figs 4.11 - 4.13 below.
incidence per 100000 words
Diachronic development of the in nitive in Spanish
55 50 45 40 35 30 25 20 15 10 5 0 1100
1150
1200
1250
1300
1350
1400
1450
1500
1550
1600
1650
1700
1750
217
1800
1850
1900
1950
2000
1950
2000
1950
2000
year
Spanish con+in nitive
per 100000 words
Figure 4.11
900 800 700 600 500 400 300 200 100 0 1150
1200
1250
1300
1350
1400
1450
1500
Figure 4.12
1550 1600 year
1650
1700
1750
1800
1850
1900
Spanish de+in nitive
incidence per 100000 words
700 600 500 400 300 200 100 0 1100
1150
1200
1250
1300
1350
1400
1450
Figure 4.13
1500
1550 1600 year
1650
1700
Spanish a+in nitive
1750
1800
1850
1900
218
Chapter 4
When comparing the evolutionary patterns of these prepositional in nitives, the most striking observation is the similarity of all the curves, irrespective of whether the prepositional in nitive is a prototypical adjunct (por/para), mostly used as a prototypical complementizer (de), or moving between the two extremes of the continuum (a): after a more or less unchanging period in the earlier part of the Middle Ages, an increase gradually sets in, which intensi es around the year 1500 and peaks round about a hundred years later. Though similar in principle, the development is not simultaneous. por/para+in nitive reaches its zenith between 1550 and 1600, but de only between 1600 and 1650. On the other hand, the gradual rise in frequency begins around 1300 for de, more than a hundred years earlier than for por/para. This is a clear indication that we are not dealing with a wholesale development, but that every preposition follows its own course of development; these individual courses of development do, however, have a surprising resemblance to each other. Their subsequent development con rms this interpretation. After a drop around 1650, a relatively stable period follows, in which por/para experiences a rise of approx. 25% by the end of the 19th century, but then falls back to the level of the 17th century. De also rises (approx. 10%) and falls back to a level below that of 1650 during the 20th century. This trend is far stronger with con+in nitive, with a very marked rise (50%) that takes it above the peak of the `Classical period' in the 19th century, but then falls to a pre-Classical level during the 20th century. Whilst por/para, de, and con behave similarly in principle, a experiences a slight rise instead of a fall in the 20th century. All of this shows that each prepositional in nitive has its own individual, not precisely predictable evolution, but nevertheless conforms to a broad pattern. It should be noted that the common trend is independent of absolute frequency. Fig. 4.14 gives an impression of the differences in frequency between the respective prepositional in nitives. 4.2.4 A special case: al + in nitive A dif cult issue is the temporal construction al+in nitive. It occurs from the earliest medieval texts, as in (248).
Diachronic development of the in nitive in Spanish
219
600 a con de por + para
per 100000 words
500 400 300 200 100 0 1150
1200
1250
1300
1350
1400
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish a, de, por/para, con+in nitive
Figure 4.14
(248) Al exir
1450
de Salón mucho ovo
al leave.inf from prpn
much
buenas aves.5
have.3sg.pret good
birds
When leaving Salón, there were many good augurs. However, the status of this in nitive with article is not clear. Medieval Spanish regularly uses in nitives in a very noun-like way6 , as illustrated by (249), in which the nominal nature of the in nitive is demonstrated by the fact that the patient or object is linked by `genitival' de, a structure typically used for objects of nouns. (249) Al tirar
de la lança...7
al throw.inf/nmlz of the lance
At the throwing of the lance... Both the de nite article contained in al and the fact that the [patient] is linked by de would appear to indicate nominality, and initially this is surely the case. But the construction's increasingly verbal nature, re ected by an incipient rise in frequency from about 1500 onwards, becomes unmistakable in sentences such as (250) from Don Quijote, where the overt subject of the in nitive is not linked by genitival de, but takes the form of a syntactic subject. (250) Al subir
el duque y la duquesa en el teatro...8
al ascend.inf the duke
5 6 7
Çid, 859 Beardsley (1921: 3 12) Çid, 3686
and the duchess
in the theatre
220
Chapter 4
When the duke and the duchess entered the theatre... The question of such in nitives' nominality is discussed in more depth in Section 3.4.3 above. But whilst the in nitive in other prepositional clauses tends to behave largely like a verb in relation to the other constituents of the dependent clause, the overt morphological presence of the article in al (< a+el) appears to block this split between internal verbal character and external nominality for a long time, as can be seen in Figure 4.15, which shows a marked difference to the development of the other prepositional in nitives discussed above. incidence per 100000 words
175 150 125 100 75 50 25 0 1100 1150
1200
1250 1300
1350
1400
1450 1500
1550
1600 1650
1700
1750 1800
1850
1900
1950 2000
year
Figure 4.15
Spanish al+in nitive
The typical rise between 1400 and 1600 is much less signi cant than with other prepositional in nitives; this can be explained by the continued presence of nominal usage during the Classical period, as in (251). (251) ... y al passar
de un arroyo levantó la falda...9
and al pass.inf/nmlz of a
brook
lifted
the skirt
... and as she crossed a brook, she lifted her skirt... This continuing strongly nominal use during the Classical period appears to retard the typical rise of the curve, as only a limited number of verbs regularly appear as `verbal nouns' in this construction. These are mostly verbs of motion (salir, pasar, andar, llegar, cabalgar etc.), verbs
8 9
Don Quijote, 35535 Fructus Sanctorum, Ejemplos de las sagradas escrituras, Ejemplos Cristanos, 20
Diachronic development of the in nitive in Spanish
221
of eating and drinking (comer, cenar, beber etc.), impersonal verbs denoting times of the day (amanecer, anochecer) and some other notions such as pesar `to grieve', parecer `to seem'. Some of these eventually become fossilized in their nominal form and are fully nominal in the modern language. The crucial point, however, is that as long as the al-construction retained its special, more nominal status, it was not freely available to be used with any verb in the way that other prepositional in nitives allowed. Only after al+in nitive becomes fully integrated into the system of prepositional in nitives does the expected increase in frequency set in, accelerating during the 19th century and easing off in the 20th . 4.2.5 The rise of new prepositional in nitives Throughout the documented history of Spanish, new prepositions are created. They are often analytic, consisting of more than one word (cf. Section 1.3.5; a productive pattern is the combination of an existing preposition, followed by a nominal element and de as a marker of prepositionality. The rst preposition to begin forming prepositional in nitives during the documented history of Spanish is, however, morphologically inherited from Latin. Sin+in nitive rst appears sporadically at the beginning of the 13th century, and its frequency rises sharply from 1450 onwards. After a peak and a subsequent drop between 1550 and 1700, a renewed, more gradual increase up to the end of the 19th century and a slightly steeper decline during the 20th century follow. Después de (Fig. 4.18) and antes de+in nitive (Fig. 4.17) follow a similar development. They rst appear considerably later than sin+inf, in the late 15th century. The structural symmetry between these two temporal prepositional in nitives is re ected by the fact that they rst appear around the same time, and their subsequent development is also largely parallel until the beginning of the 20th century. However, up to around 1900 the absolute frequency of antes de+in nitive is only around 50% of that of después de+in nitive; during the 20th century, a rise of the former and a similarly strong drop of the latter leads to approximate numeric parity between the two.
222
Chapter 4
incidence per 100000 words
160 140 120 100 80 60 40 20 0 1100 1150
1200
1250 1300
1350
1400
1450 1500
1550
1600 1650
1700
1750 1800
1850
1900
1950 2000
1800
1850
1900
1950
2000
1800
1850
1900
1950
2000
year
Spanish sin+in nitive
incidence per 100000 words
Figure 4.16
45 40 35 30 25 20 15 10 5 0 1100
1150
1200
1250
1300
1350
1400
1450
1500
1550
1600
1650
1700
1750
year
Figure 4.17
Spanish antes de+in nitive
incidence per 100000 words
60 50 40 30 20 10 0 1100
1150
1200
1250
1300
1350
1400
Figure 4.18
1450
1500
1550 1600 year
1650
1700
1750
Spanish después de+in nitive
Diachronic development of the in nitive in Spanish
223
incidence per 100000 words
Desde and hasta might be expected to develop in a similar way to después/antes de. For hasta+in nitive this is, broadly speaking, the case, as shown in Fig. 4.19, except for a marked decline in the early 20th century. 90 80 70 60 50 40 30 20 10 0 1100
1150
1200
1250
1300
1350
1400
1450
Figure 4.19
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish hasta+in nitive
However, there is clear resistance to desde+in nitive. One isolated instance can be found in Larra (1832), and beyond this it is only to be found in the journalistic and oral corpora of the late 20th century, with a very low frequency of 0.21 per 100,000 words. What is more, not a single in nitival clause with desde has temporal sense.10 It is used exclusively to express conceptual, not temporal, distance, usually in combination with a to indicate the opposite end. A typical example is sentence (252). (252) ... desde poner aborto gratuito y prácticamente obligatorio a from
put.inf abortion free
and practically
obligatory
to
convertirlo en delito...11 convert.inf.it into an
offence...
... from offering free and practically obligatory abortion to turning it into a criminal offence... When compared to those of de, a, por/para, and con above, Figs 4.18 4.19 appear to suggest that there is no intrinsic, xed time scale for the statistical development of prepositional in nitives. Once established,
10 11
A partir de, synonymous with desde in its temporal usage, is also extremely rare; it appears to be more popular in Latin America than in Spain. Oral Corpus, TV programme Hablando Claro
224
Chapter 4
their development becomes synchronized with that of the other, much older prepositional in nitives. This comes as no great surprise, as language users generally do not have diachronic information at their disposal; they are not aware that one construction is many centuries older than another. Absolute frequency does vary considerably among different prepositional in nitives, but as Fig. 4.14 above shows, this is not a crucial factor in their relative statistical development the curves have a similar shape whether we are dealing with a range of 100 400, or of 0 10 instances per 100,000 words. Other compound prepositions follow in the footsteps of después de and antes de. Many types of prepositional in nitive that today appear unremarkable have, in fact, only emerged very recently; others have been in use for longer, but only sporadically before the 20th century. En lugar de and en vez de are virtually synonymous and interchangeable in the modern language, with a similar frequency of around 5 per 100,000 words. But their evolution has not been as similar. In the texts this analysis is based on, en lugar de+in nitive rst appears in the early 15th century in the Libro de Gatos in ve instances, and subsequently with some regularity until 1650. A gap of 150 years follows, before it reappears just before 1800, gradually increasing its frequency to the current level of approx. 5 per 100,000 words. En vez de, on the other hand, is much less popular during the Classical period12 , but rises to almost twice the frequency of en lugar de during the 19th and the rst half of the 20th century. Fig. 4.20 sketches the differential development after 1500. It should be noted that the two prepositional expressions appear to coexist with no apparent danger of one ousting the other, despite their semantic and syntactic similarity. However, the Oral Corpus suggests that particularly in the spoken language, en vez de is twice as popular as en lugar de; in journalistic style, this discrepancy is much less extreme, with around 10% more en vez de than en lugar de. A n de is another prepositional expression found sporadically in texts from the mid-15th century onwards, as shown in Fig. 4.21.
12
Merely one instance in El Buscón, 4 in Calderón de la Barca.
Diachronic development of the in nitive in Spanish
225
per 100000 words
30 25
en lugar de en vez de en lugar de en vez de
20 15 10 5 0 1500
1550
1600
1650
incidence per 100000 words
Figure 4.20 11 10 9 8 7 6 5 4 3 2 1 0 1100
1150
1200
1250
1300
1350
1700
1750 year
1800
1850
1900
1950
2000
Spanish en lugar de/en vez de+in nitive
1400
Figure 4.21
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish a n de+in nitive
As a n de+in nitive is semantically all but synonymous with para+in nitive, its use is primarily stylistically and pragmatically motivated: two characteristics that contrast sharply with para are the vivid way in which a n de `with (the) end of' literally explains the concept of nality, and the fact that it has more morphological weight than para. Not only does it consist of three separate morphemes, but one of these morphemes is derived from a noun, i.e. a `heavy' (and originally stressed) element. These two features make a n de a more expressive, emphatic, but at the same time a somewhat formal, stilted way of introducing a nal clause. The latter is the reason why a n de is virtually absent from the Oral Corpus (0.11 per 100,000 words), whilst journalese embraces it (1.74 per 100,000 words). Other prepositional expressions with the same morphological structure join the inventory of prepositional in nitives at later points in time. Aspectual a punto de+in nitive `about to' (4.22) is rst found at the end of the 16th century, but regular usage only sets in during the 20th century.
incidence per 100000 words
226
Chapter 4
15 12,5 10 7,5 5 2,5 0 1100
1150
1200
1250
1300
1350
1400
Figure 4.22
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish a punto de+in nitive
Concessive a pesar de and its synonymous cognate pese a are very recent additions to the inventory; before the 19th century there are no instances of concessive in nitival clauses. A single instance of a pesar de+in nitive is found in each of Larra (1832), Giner de los Ríos (1875), and Unamuno (1909); only in the journalistic and spoken corpus of the 1990s does the frequency pick up, reaching 1.83 and 0.74 per 100,000 words, respectively. Pese a+in nitive is even later in coming, with no instances at all before the 1990s, but with a surprisingly high frequency13 of 9.16 per 100,000 words in Marías. In the journalistic corpus its frequency is the same as that of a pesar de+in nitive (1.83 per 100,000 words), which indicates that the two stand in direct competition in this register. However, the fact that there is only a single occurrence in the Oral Corpus indicates that pese a+in nitive has not (yet) entered the spoken language. Another recent addition is luego de+in nitive, which appears to be restricted to journalistic register. Whilst not present in any of the other texts, it occurs sporadically, with a frequency of 0.38 per 100,000 words in the 2001 journalistic corpus. The origin of this construction appears to be journalists' tendency to avoid repetition and vary their expressions; luego de is a straight-forward calque on después de, based on the synonymity of the adverbs después and luego meaning `afterwards'. Journalists appear to favour the economy of the prepositional innitive to replace slightly less concise nite clauses. Another prepositional in nitive bene ting from this `economical style' in the 20th century is tras+in nitive (`after'). Whilst tras+noun is a popular prepo-
13
This is exactly the same frequency as that of en vez de and a punto de.
Diachronic development of the in nitive in Spanish
227
sition during the Classical period, the texts in Appendix A from this period contain only a single instance of tras+in nitive, in Calderón de la Barca (1651). But in the 20th century tras begins to be used in innitival clauses, with one instance in Unamuno, a frequency of 18.33 per 100,000 words in Marías (1992) and 16.75 per 100,000 words in the 2001 journalistic corpus these frequencies are similar to those of the synonymous después de+in nitive. However, like pese a+in nitive, tras+in nitive has only entered the spoken language sporadically (0.42 per 100,000 words). 4.2.6 Semantic grouping The impossibility of clearly classifying certain prepositional clause types along semantic lines has been discussed above. The overlap between nal, causal and substitutional sense of por and para has been mentioned, as has the continuum between the full, clearly delimitable senses of a and de and their bleached, weaker, less clearly de ned usages. Other prepositional in nitives do, however, have a clearer, more stable semantic value. Fig. 4.23 compares three such semantic areas. 140 per 100000 words
120 100 temporal prep+inf. concessive prep+inf. exclusive prep+inf. (sin)
80 60 40 20 0 1150
1200
1250
Figure 4.23
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish temporal,concessive and abessive prepositional in nitives
This comparison shows how temporal and abessive14 prepositional in nitives have undergone a very similar, parallel development, even in
14
The preposition used to form abessive prepositional in nitives is sin.
228
Chapter 4
absolute frequency. In stark contrast to this, concessive prepositional in nitives are all but negligible. The concept of temporality is, in Spanish as in most Indo-Europaen languages, tripartite. It can be subdivided into anterior, simultaneous, and posterior. Fig. 4.24 illustrates how these areas of temporality are represented within the larger category of temporal prepositional in nitives. per 100000 words
100 anterior matrix prep+inf. (hasta, antes de) posterior matrix prep+inf. (después de etc.) simultaneous time ref. prep+inf.
80 60 40 20 0
1150
1200
1250
Figure 4.24
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish anterior, simultaneous and posterior prepositional in nitives
Whilst prepositions expressing anteriority and posteriority of the dependent clause in relation to its main clause develop in a relatively similar way, emerging in the 15th century, and remaining at a relatively stable level from the 17th century onwards, temporal prepositional innitives expressing simultaneity between the main and the dependent clause show a very different development after the Classical period, increasing almost vefold within 200 years, and then remaining stable at this level throughout the 20th century. As these clauses are, in their majority (94%), formed with al, the reason provided above in Section 4.2.4 for the somewhat exceptional development of al+in nitive also accounts for the semantic class of simultaneous dependent clauses as a whole. It may, at rst sight, seem surprising that the ratio between anterior/posterior prepositional in nitives on the one hand, and their simultaneous counterparts on the other, can be subject to such drastic change, as the pragmatic context, the real world, does not change. Though it is possible for speakers' interest in expressing anterior, posterior, or simultaneous time reference to change slightly through time if their attitudes and ways of viewing the world change, the necessities imposed by the
Diachronic development of the in nitive in Spanish
229
pragmatics of the real world should be expected to make any drastic shifts impossible. So how can the increase in frequency, not just of al+in nitive, but of every single prepositional in nitive examined here, be explained? There are two possibilities: the prepositional in nitive might either have increased at the expense of paratactic structures, or at the expense of nite subordination. To a certain extent, both explanations hold: in Medieval Spanish, clausal subordination is less frequent than during later periods; the inventory of subordinating conjunctions and prepositions is not yet developed to the same degree as today. Therefore, Old Spanish resorts more to the juxtaposition of main clauses, and to lexically underspeci ed dependent clauses where the pragmatic context makes the overt marking of logical relations redundant. Fig. 4.25 illustrates the overall rise in the use of dependent clauses, and the proportion of in nitival and nite ones.
per 100000 words
9000 8000
total Infinitive overall finite subordination total clausal compements
7000 6000 5000 4000 3000 2000 1150
1200
1250
1300
1350
Figure 4.25
1400
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish in nitival and nite dependent clauses
The frequency of nite and in nitival dependent clauses combined rises by 14% from 1150 to 2000, with a sudden increase at the beginning of the 15th century. During the Classical period, it is up to 7% more frequent than today. The sudden increase around the year 1500 is caused by increases of both nite and in nitival dependent clauses, though the increase of in nitival clauses is the more marked of the two. Even in this very general graph it is visible that, after 1650, a gradually decreasing frequency of nite dependent clauses is offset by a comparable increase in in nitival ones. The following section will examine the interplay between these two clause types in more depth.
230
4.3
Chapter 4
In nitival clauses and their nite counterparts
Fig. 4.26 contrasts the development of the prepositional in nitives discussed in the previous section with that of the semantically (and in most cases morphologically) corresponding hypotactic conjunctions. 2000 per 100000 words
1800
total prep. inf. total conj.
1600 1400 1200 1000 800 600 400 1150
1200
1250
Figure 4.26
1300
1350
1400
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish prepositional in nitives vs. corresponding conjunctions
In the early Middle Ages, nite dependent clauses are around twice as frequent as prepositional in nitives. Until the 16th century, both types experience an overall increase. This is, as discussed above, mainly at the expense of paratactic structures. Rising at a stronger rate than its nite counterpart, the in nitive draws equal between 1550 and 1650, but only around 1800 does the frequency of prepositional in nitives nally exceed that of the corresponding nite clauses. It should be noted that, after 1600, the two clause types behave very much as a counterpoint to each other, which indicates that they are in direct competition, with any gain by one competitor offset by a corresponding loss for the other. This competition, it appears, is still open: in the 20th century, the continued increase of the prepositional in nitive has lost its momentum, and nite dependent clauses have begun to regain some ground. 4.3.1 Prepositional in nitives postdating their nite counterparts An examination of individual pairs of nite and in nitival clauses reveals that there is no single development pattern. Two fundamental classes of conjunction/preposition pairs can be distinguished: those
Diachronic development of the in nitive in Spanish
231
per 100000 words
with the conjunction predating the prepositional in nitive, and those with the prepositional in nitive predating the corresponding conjunction. Antes de/antes de que, después de/después (de) que, and hasta/hasta que (Figs 4.27 - 4.29) are good examples of the former class. In all three cases, the conjunction is already present in the very earliest texts, but the prepositional in nitive rst occurs around or after 1400. A considerable increase in the frequency of the conjunction prior to the rise of the in nitive is reversed as the in nitive gains ground; indeed, the drop in conjunctional usage is around twice as great as the rise in in nitival usage during the Classical period, until around 1650. Following this convergence of the two frequencies, they remain relatively stable in comparison to the pre-Golden Age period. This does not mean that no further changes occur: antes de+in nitive and antes de que experience a simultaneous gradual increase during the 20th century, whilst después de+in nitive appears to be more successful than hasta+in nitive and antes de+in nitive in replacing its nite counterpart. 35 30 25 20 15 10 5 0 1150
ante(s de) ante(s) (de) que
1200
1250
1300
1350
1400
per 100000 words
Figure 4.27
45 40 35 30 25 20 15 10 5 0 1150
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
1900
1950
2000
1900
1950
2000
Spanish antes de que vs. antes de+in nitive
después de después (de) que
1200
1250
1300
1350
Figure 4.28
1400
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
Spanish después (de) que vs. después de+in nitive
232
Chapter 4
per 100000 words
70 60 50 40 30 20 10 0 1150
hasta hasta que
1200
1250
1300
1350
Figure 4.29
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish fasta/hasta que vs. fasta/hasta+in nitive
per 100000 words
Luego de/luego (de) que (Fig. 4.30) and desde/desde que (Fig. 4.31) might arguably also be included in this class. Like hasta que, antes de que and después (de) que, luego que is present from the earliest texts onwards, whilst desde que is rst found in the 15th and 16th century. 9 8 7 6 5 4 3 2 1 0 1150
luego de luego (de) que
1200
1250
1300
1350
1400
per 100000 words
Figure 4.30
16 14 12 10 8 6 4 2 0 1150
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
1900
1950
2000
1900
1950
2000
Spanish luego que vs. luego de+in nitive
desde desde que
1200
1250
1300
1350
1400
Figure 4.31
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
Spanish desde que vs. desde+in nitive
Both graphs show the typical overall rise in frequency, followed by a downturn. In the case of luego (de) que, this downturn during the 17th
Diachronic development of the in nitive in Spanish
233
per 100000 words
century is particularly sudden and abrupt, but nevertheless generally in line with the trend illustrated in Figs 4.27 - 4.29 above. The crucial difference, however, is that this drop does not coincide with a continued rise in the usage of the corresponding prepositional in nitive. Indeed, the corresponding prepositional in nitive luego de+in nitive is patently absent, and as discussed in the previous section only comes into use in the late 20th century. However, the fact that después de+in nitive (Fig. 4.28) occurs more frequently then the corresponding conjunction después de que can account for the absence of luego de+in nitive: Fig. 4.32 shows how the synonymous conjunctions luego que and después (de) que combined compare to después de+in nitive and tras+in nitive combined. These two prepositional in nitives appear to step in for the absent luego de+in nitive, leaving us with an interestingly asymmetrical pattern. 50 45 40 35 30 25 20 15 10 5 0 1150
posterior matrix prep+inf. (después de, luego de, tras) posterior matrix conjunctions (después de que, luego (de) que)
1200
Figure 4.32
1250
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish luego/después (de) que vs. tras/después de/luego de+in nitive
Not only does this overview of conjunctional and prepositional use in temporal dependent clauses with posterior main clause explain the fact that después de+in nitive is atypically more frequent than nite después de que, as well as accounting for the virtual absence of luego de+in nitive, but it also shows that it is not suf cient to look at isolated pairs of prepositional in nitives and corresponding conjunctions to obtain a fuller picture, it is important to take into account not just morphological correspondence, but also semantic correspondences between in nitival and conjunctional dependent clauses. As for desde que, it has already been pointed out in Section 4.2.5 that the corresponding desde+in nitive is extremely rare even today, and never used temporally. Whether it will eventually establish itself, bringing down the frequency of the corresponding nite construction in
234
Chapter 4
the same way as has occurred with antes, después and hasta is dif cult to predict. 4.3.2 Prepositional in nitives predating their nite counterparts
per 100000 words
The clearest example of a prepositional in nitive becoming established prior to the corresponding nite structure is sin + in nitive. Fig. 4.33 shows how this in nitival structure powerfully establishes itself between 1400 and 1600, subsequently continuing its growth more gradually until the late 19th century. Sin que follows, but always remains well below the frequency of the in nitival construction, at approximately 20 25% of its level. 90 80 70 sin 60 sin que 50 40 30 20 10 0 1100 1150 1200 1250 1300 1350 1400 1450 1500 1550 1600 1650 1700 1750 1800 1850 1900 1950 2000 year
Figure 4.33
Spanish sin que vs. sin+in nitive
A less clear, but in principle similar development is found with a n de+in nitive: though the overall frequency of this preposition/conjunction pair is so low that the text corpus this study is based on is not suf ciently large to give reliable average frequencies, it can nevertheless be observed that a n de+in nitive is found as early as 146015 , whereas the rst instance of a n (de) que is slightly later, in 149416 ,
15 16
Siete Edades 121: `a n de ganar aquel velloçino' Mujeres Ilustres, Opis: `el hermano havía jos que los hoviesse de matar, a n que a sus jos el reyno bolviesse.'
Diachronic development of the in nitive in Spanish
235
and that, during the last two centuries, a n de+in nitive is found three times as frequently as a n de que17 . Some other prepositional in nitives predate and outnumber the corresponding conjunctional construction more signi cantly. Ever since their emergence several centuries ago, en lugar de+in nitive, en vez de+in nitive (Fig. 4.20) and a punto de+in nitive (Fig. 4.22) are only very sporadically joined by their nite counterparts. A punto que is found in the late 16th and early 17th century18 , but does not catch on, and only occurs rarely afterwards (below 0.05 per 100,000 words). Only one single example of en lugar de que19 and none of en vez de que appear in the entirety of the texts analysed here. A somewhat exceptional case is concessive a pesar de+in nitive, which starts off in the 18th century in a way similar to a punto de or en lugar/vez de, without a corresponding nite construction. But very recently and suddenly, after 1980, a pesar de que appears with a surprisingly high frequency of approximately 4 per 100,000 words, almost twice the frequency of a pesar de+in nitive. This is a highly idiosyncratic development: not only is it the only prepositional in nitive outperformed by a corresponding nite structure that it predates, but the availability of synonymous aunque as a nite alternative makes this sudden increase in the use of a pesar de que all the more surprising, particularly in comparison with the third concessive construction, pese a, where the frequency of nite pese a que lies far below that of pese a+in nitive; the ratio is, in fact, about 1:20.
17
18 19
A n de que is the modern cognate of older a n que; this tendency to incorporate the overtly prepositional element de can be observed in a number of analytic prepositional expressions, e.g. antes de que and después (de) que. It is a comparatively recent innovation, rst found in the 19th century. 11 instances in Fructus Sanctorum, 5 in Don Quijote Oral Corpus: `yo puedo llegar a un convenio contigo y en lugar de que sea tuya la mitad sea para el Ayuntamiento'
236
Chapter 4
4.3.3 Conjunctional and prepositional clauses present from the earliest texts Both por(que) and para/pora (que) occur in in nitival and nite dependent clauses from the earliest Spanish texts. Despite the semantic overlap discussed above, a very clear contrast in the distribution of nite and non- nite clauses can be identi ed. From the early Middle Ages, para+in nitive is twice as frequent as para que. From the early 15th century onwards, the numeric dominance of the in nitival construction increases, and by the 19th century it is more than four times as frequent as the nite construction.
per 100000 words
300 250
para,pera,pora para que
200 150 100 50 0 1150
1200
1250
1300
1350
1400
Figure 4.34
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish para que vs. para+in nitive
Por, on the other hand, shows an inverse development to that of para. Throughout the history of Spanish, the instances of porque outnumber those of por+in nitive, and the difference begins to increase even before the end of the Middle Ages. From the 17th century onwards, the frequency of por+in nitive drops below its medieval level, and by the year 2000 it is at a mere 40% of its usage in the Middle Ages. Porque also peaks during the Golden Age, in the early 16th century, and the subsequent decrease in frequency temporarily brings it down to the same level as in the Middle Ages, before the 20th century sees a renewed, if gradual, increase. Whilst it is argued above that for the semantic eld of posterior main clause time reference it is necessary to look at all the prepositions/conjunctions involved, choosing an essentially semantically-based approach, por and para suggest that there is also a lexical factor involved, as even during the period in which porque and por are used
per 100000 words
Diachronic development of the in nitive in Spanish 450 400 350 300 250 200 150 100 50 1150
237
por porque
1200
1250
1300
1350
1400
Figure 4.35
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish porque vs. por+in nitive
for nal clauses and thus to some extent interchangeable with para que and para, their nite-in nitival ratio is inverted. 4.3.4 Finite and in nitival clauses by semantic class In certain semantic areas, nite dependent clauses are vastly dominant. It has already been observed that concessive prepositions have only very recently begun to be used with the in nitive, and Fig. 4.36 shows that, in comparison to nite concessive clauses, their share is negligible. Concessive clauses overwhelmingly make use of the conjunction aunque.
per 100000 words
175 150
concessive prep+inf. concessive conjunctions
125 100 75 50 25 0
1150
1200
1250
1300
Figure 4.36
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Concessive nite and in nitival clauses in Spanish
A similar situation is found with conditional clauses, due to the predominance of the archetypical conditional conjunction si as seen in Fig. 4.37. In the temporal domain, it is necessary to differentiate. Overall, temporal clauses are nite more often than in nitival, as seen in Figure 4.38. This predominance of nite dependent clauses is, however, not uniform throughout the temporal domain. Fig. 4.32 above and Fig.
238
Chapter 4
per 100000 words
700 600 500 400 300
conditional prep+inf. conditional conjunctions
200 100 0 1100
1150
1200
1250
1300
per 100000 words
Figure 4.37 400 350 300 250 200 150 100 50 0 1100
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
1950
2000
Conditional nite and in nitival clauses in Spanish
temporal prep+inf. temporal conjunctions
1150
1200
1250
1300
Figure 4.38
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
Temporal nite and in nitival clauses in Spanish
4.39 show that since its emergence in the 15th century, in nitive usage in non-simultaneous temporal clauses have caught up with their nite counterparts in this area.
per 100000 words
100 anterior matrix prep+inf. (hasta, antes de) anterior matrix conjunctions (hasta que, antes de que)
80 60 40 20 0 1100
1150 1200
Figure 4.39
1250
1300 1350
1400 1450
1500 1550 1600 1650 year
1700 1750
1800 1850 1900 1950 2000
Finite and in nitival temporal clauses with anterior main clause in Spanish
The area in which nite subordination continues to dominate is simultaneous time reference, largely due to the high frequency of cuando. Though the incidence of temporal prepositional in nitives with al and al punto de increases steadily, and nite cuando does experience a compa-
Diachronic development of the in nitive in Spanish
239
rable decrease in frequency, the nite construction is still approximately 75% more frequent.
per 100000 words
250 200 150 simultaneous time ref. prep+inf. simultaneous temporal reference conjunctions
100 50 0 1100
1150
Figure 4.40
1200
1250
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Finite and in nitival temporal clauses with simultaneous main clause in Spanish
It must, however, be remembered that a third type of temporal clause has not been taken into consideration: the gerund/participle. It has been discussed in previous chapters that this type of dependent clause is a semantically underspecified construction, used in all kinds of logical relations between main and dependent clause. However, a temporal component is always present. Every gerundial clause, whether the pragmatic context gives it a causal, concessive, or abessive colouring, refers to two simultaneous notions, just as every participial clause is anterior to its main clause. Fig. 4.41 reveals that if gerundial clauses are added to the equation, dependent clauses with simultaneous time reference are, on the whole, far more frequently non- nite than nite. The strong increase of non- nite clauses between 1350 und 1600 is largely due to the increasing tendency to replace the medieval paratactic style. For simultaneous time reference of two clauses, parataxis is very common in medieval texts, as simple juxtaposition will, under normal pragmatic circumstances, be interpreted as implying simultaneous time reference by default. The subsequent steady decline of gerundial usage is due to an increasing trend towards semantic precision, as discussed in Section 4.2.1 above.
240
Chapter 4
per 100000 words
700 600 500 400
gerund + simultaneous prep.inf. simultaneous temporal reference conjunctions
300 200 100 1150
1200
Figure 4.41
4.4
1250
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Finite and non- nite dependent clauses with simultaneous time reference in Spanish
Discussion of the Spanish diachronic data
The most fundamental observation to be made is the overall increase in prepositional in nitives. This increase is not limited to the frequency of existing structures, but also involves a steady increase in the range of prepositions participating in this construction. On the other hand, Figs 4.4 and 4.5 above have shown an almost symmetrical mirror image for non-prepositional in nitives. Though the non-prepositional decline is not quite as great as the prepositional increase (which is explicable by the overall increase in the usage of dependent clauses as shown in Fig. 4.25), it does suggest a direct link between the two, with prepositional in nitives replacing non-prepositional ones. To some extent, there is a straight-forward replacement. In particular a+in nitive, which accounts for between 20% and 30% of prepositional in nitives (see Fig. 4.13 and Fig. 4.14), increasingly replaces certain plain in nitives. One of the signi cant constructions in this respect is ir (a)+in nitive: both in the literal motional meaning and the emerging gurative auxiliary (future) usage of ir, the dependent clause is not necessarily linked by a in Old Spanish, as exempli ed in (253) and (254); this prepositionless construction subsequently becomes increasingly less common, particularly in written texts. Literal use: (253) ... que el martes quisiese ir that the Tuesday wanted
ver
el su mercado.20
go.inf see.inf the his market.
Diachronic development of the in nitive in Spanish
241
... that on Tuesday he wanted to go to see his market. Use for future reference: (254) ... vayámoslos
ferir
en aquel día de cras.21
go.1pl.sbjv.them wound.inf in that
day of tomorrow
... we are going to wound them tomorrow. The original nal meaning of ir a+in nitive, `to go in order to', undergoes a process of grammaticalization through time. This involves several of the typical symptoms, including an entrenchment of the original structure into an increasingly xed morphological unit22 , incipient phonetic reduction 23 , and the replacement of the original semantic content of both elements by a joint functional content (future tense marking, in this case), and an increase in usage frequency. However, it must not be forgotten that, depending on the context, the prepositional in nitive with a also retains its full nal meaning, as in the short Spanish dialogue in (255). (255)
½Hasta luego, me voy! ¾Te vas afinal hacer qué? Voy ableached encontrarme con mi novia. Bye, I'm going! You're going tofinal do what?' I'm going tobleached meet my girlfriend
20 21 22
23
Libro de Buen Amor, 1372 Çid, 676 The analysis of [ir a+in nitive] as a single morphological (verbal) unit is supported by the fact that elements such as temporal adverbs are not normally inserted: `Voy mañana a hacerlo.' can only be understood to have the literal reading `I'm going tomorrow, to do it.' than a grammaticalized temporal one. Especially in rapid colloquial speech, the `a' is becoming increasingly weakened.
242
Chapter 4
per 100000 words
Though it is not logically or syntactically necessary for the purpose of going somewhere to lie in the future24 , our pragmatic knowledge of the real world tells us that there is an overwhelming likelihood of this being the case. Consequently, a+in nitive lends itself to a semantically bleached reanalysis as purely temporal structure. The shift from semantic to functional also explains the absolute numeric increase of this particular structure, which accounts for around 50% of a+in nitive in modern spoken Spanish25 , though considerably less in the written language. (20% in modern literary texts, 12% in journalistic register.) Fig. 4.42 shows how a+in nitive relates to the total of a+in nitive: up to the 20th century, the share of ir a does not account for a signi cant share of the total; this is in part due to the tendency in modern texts to use a less formal register the signi cantly higher frequency in the Oral Corpus shows that ir a can primarily be seen as a feature of the spoken language. 450 400 350 300 250 200 150 100 50 0 1150
a ir a + infinitive
1200
1250
1300
1350
1400
Figure 4.42
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish a+in nitive and ir a+in nitve
Beardsley (1921: 21 86) gives detailed information on other verbs that can take non-prepositional in nitival clauses in the Middle Ages; several of these verbs link their in nitival complement by means of an obligatory preposition in later periods. To name just a few examples, aiudar, enbiar, guiar, mouer and ir can take complementizerless in nitival complements describing goal or purpose. Verbs of starting and
24
25
A sentence such as the following shows that there is no logical need for future reference: I am going so that I wasn't lying when I said that I wouldn't be here at 4 o'clock. Oral Corpus
Diachronic development of the in nitive in Spanish
243
nishing, such as çessar, començar, empeçar, compeçar, as well as psychological verbs such as atreuerse, ensayar, which commonly take a prepositional in nitive with a or de in the modern language, can also occur with a plain in nitive in Old Spanish. The increase in the number of verbs taking prepositional rather than plain in nitival clauses is ongoing. In modern spoken Spanish, verbs such as dejar can take prepositional in nitives, as in (256). (256) Déjamelo a ver.26 letimp.me.it a seeinf
Let me see it. This continuing process, in which direct object in nitives become prepositional in nitives, is a fundamentally pragmatically based phenomenon, resulting from a semantic overlap between semantic roles such as [theme] and [purpose]. The details of this process will be discussed further in Chapter 7. However, this straight-forward replacement of plain in nitives by prepositional ones can only partially account for the increase in prepositional in nitive usage. Despite the fact that ir a is, numerically speaking, the most signi cant instance of this, its overall contribution to the increase in prepositional in nitives is small, as seen in Fig. 4.42. Some main verbs take other prepositions to link an in nitive to which they stand in a direct-object-like semantic relation: apostar por alcanzar algo `to bet on achieving something' empeñarse en hacer `to insist on doing' soñar con hacer `to dream of doing' But whilst a and de frequently tend to form grammaticalized units with various main verbs, this tendency is far weaker with other prepositions. This includes por and para, which nevertheless have a similar frequency as a+in nitive (Fig. 4.14), as well as the whole range of prepositional in nitives that emerge in the course of the subsequent his-
26
Oral Corpus
244
Chapter 4
tory of Spanish. In order to nd an explanation for their development, we have to look further a eld. Some of the factors have already been mentioned in passing above. In the context of the increasing frequency of prepositional in nitives with simultaneous time reference (Fig. 4.41 above), and in Section 4.2.6, it has been observed that the use of dependent clauses, both nite and in nitival, increases at the expense of parataxis. Which type of hypotactic structure is chosen in its place is determined by a combination of syntactic and pragmatic factors.
4.4.1
The central role of subject reference
As discussed in some detail in the previous chapter, subject reference is a crucial factor in the choice of dependent clause type. But whilst non-prepositional in nitival clauses are highly restricted by this constraint, it has been shown that pragmatically assigned control by external NPs is more common in prepositional in nitival clauses. Not only does the pragmatic context play a crucial role in assigning subject reference in prepositional in nitive clauses, but there is also the possibility of including an overt subject in the prepositional clause. However, the predominant pattern for prepositional in nitives is subject coreference with the main clause, found in around 80% of all cases, and it is therefore also the typical, unmarked environment for in nitival clauses. Finite dependent clauses, on the other hand, are generally avoided in the case of coreferentiality in the modern language. Before the emergence of the respective prepositional in nitive, however, the conjunction carries the combined functional load of both coreferential and non-coreferential clauses. Sentences (257 - 259) present a few medieval examples of coreferential nite clauses, as used before the corresponding prepositional in nitives had come into use. (257) To padre comendo ante que muriesse your father advised
e dixo...27
before that die.3sg.pst.sbjv and said
Diachronic development of the in nitive in Spanish
245
Your father gave advice before he died, and he said... (258) Et començé
a leer sus libros fasta que los entendí.28
and begin.1sg.pret to read his books until that them understand.1sg.pret
And I started to read his books until I understood them. (259) Dize Sant Agostín que con la fortuna deue ombre luchar says Saint Augustine that with the fate
sin
que no se dexe
must man
ght
29
vençer por ella.
without that not re let.3sg.sbjv defeat
by it
Saint Augustine says that one has to ght with fate without letting oneself be defeated by it. In Modern Spanish, coreferential clauses of this type obligatorily make use of the available prepositional in nitive constructions. On the other hand, there does appear to be an occasional incipient tendency to avoid coreferential nite clauses, even when no in nitival alternative is yet available. This is achieved by constructing the dependent clause in such a way that the two syntactic subjects do not coincide, usually by means of a passive construction. (260) ... et non me demandes la razón de lo
que te
yo
and not me.io demand.imp the reason of which that you.io I
mandaré
fazer fasta que sea
acabado.30
order.1sg.fut. do.inf until that be.3sg.sbjv nished
... and don't ask me for the reason of what I will order you to do until it is nished. In sentence (261), the impersonal passive in the dependent clause may also be chosen to avoid a coreferential nite clause such as `sin que sufras vengança'.
27 28 29
30
Fazienda de Utramar, p.61(faz) Calila e Dina, p.104. Dichos de Sabios, 269; Though dated 1402, approximately 100 years after sin+innitive is rst found in other texts, the construction does not appear to have entered the language of the person who translated it from Catalan, or perhaps its absence might be a syntactic calque on Catalan. Calila e Dina, p.346.
246
Chapter 4
(261) Porque sepas so.that
que non estorçerás
tomada de ti vengança. taken
sin
que sea
know.2sg.sbjv that not escape.2sg.fut without that be.3sg.sbjv 31
of you revenge
That you may know that you will not escape without revenge being taken on you. Sentence (262) shows a different avoidance strategy, by means of an idiomatic expression equivalent to morir `to die'. (262) Fue [...] enfermo el was
ill
jo de la bibda [...] de enfermedat grant
the son of the widow
fasta que non remaso
of illness
great
en él alma.32
until that not remain.3sg.pret in him soul
The son of the widow was sick with a great illness until no soul remained within him. These examples suggest that there may already have been a certain resistance to coreferential nite subordination, due to the emerging association of the in nitive with coreferentiality and the corresponding nite structures with disjoint subject reference. The fact that conjunctional dependent clauses cede the coreferential domain to the in nitive as soon as, or even slightly before, the respective prepositional in nitive emerges is visible in almost all the graphs comparing nite and in nitival dependent clauses in Section 4.3, where a new prepositional in nitive tends to coincide with a fall in the frequency of the corresponding conjunction. It should, then, be expected that after a transitional phase an equilibrium sets in. But it would be a false premiss to expect that this equilibrium should see an equal number of nite and in nitival clauses. Indeed, it is surprising that such an equinumeric balance is achieved in some cases, as the following section will explain.
31 32
Calila e Dina, p.187. Fazienda de Utramar, p.121.
Diachronic development of the in nitive in Spanish
247
4.4.2 Coreferentiality and pragmatic relevance The ratio between nite and non- nite clauses of a certain type is fundamentally determined by the pragmatics of the contexts they are used in. This is due to the fact that the purpose of using language is, on the whole, to speak about referents in the real world, and about how these referents interact. As the patterns of interaction are, to a certain degree, statistically predictable, so are their linguistic representations. As a result of this, certain relational patterns between the constituents of a sentence can be expected to occur more frequently than others. Two contrasting examples of which relational patterns are more or less likely are nal and concessive clauses. In nal clauses, linked to their main clause by the nal preposition para (Fig. 4.34), chances are that the subject of the main clause (S1 ) will do V1 in order for himself, S1 , to do, have or get something. In some pragmatic contexts, S1 will do V1 in order for a different dependent clause subject (S2 ) to do, have or get something, but this situation is statistically less frequent. In the concessive domain, on the other hand, the opposite applies. In the vast majority of cases, S1 does V1 despite the fact that someone else (S2 ) does something. Again, this is merely statistically more common; in some pragmatic contexts, S1 will do V1 despite also doing V2 , but this occurs less frequently. The degree to which the subject reference patterns of nal and concessive clauses differ can be seen in Fig. 4.43: the ratio of coreferential to non-coreferential clauses, nal as well as concessive, remains remarkably stable over time in both languages, with around 40% of all concessive clauses and around 80% of all nite clauses being coreferential. The stability of this likelihood ratio can be explained by the fact that it is primarily determined by the way in which people and objects interact in the real world, and by whether language users generally deem these interaction patterns relevant enough to mention. This principle of pragmatic relevance can be explained as follows: almost every action undertaken by a human agent has a purpose. However, this purpose is often very predictable from the verb describing the action. For instance, if we talk about somebody eating, we will automat-
% coref. of all final / concessive clauses
248
Chapter 4
100% 80% 60% 40% 20%
Concessive coref. (%) Final coref. (%)
0% 1100
1200
1300
1400
1500
1600
1700
1800
1900
2000
year
Figure 4.43
Percentage of coreferential nal and concessive dependent clauses in Spanish
ically assume that the reason is to satisfy his hunger, if no indications to the contrary are given. Therefore, it is, in most pragmatic situations, not necessary to make explicit reference to the purpose of eating. If, on the other hand, the purpose of eating is not mere satisfaction of hunger, but something less usual, the likelihood of this purpose being put into words is far greater. Even so, however, relevance is determined by the particular situation. Imagine someone eating, not because he is hungry, but in order to get fat. Normally, this non-default purpose cannot be automatically assumed, and thus merits mention. But if the non-default purpose is (a) probably known by the hearer, or (b) not relevant to what the speaker wants to say, then it may still not be expressed. Thus, (a) if the eater is a sumo wrestler, or (b) if the reason for talking about his eating is the fact that the speaker would prefer to have some of the food himself, no matter what the eater's reasons for eating might be, the non-default purpose is not pragmatically relevant. Anything that is pragmatically irrelevant will usually be omitted in the interest of economy. The use of abessive clauses with sin (`without') exempli es the importance of this principle. Every action, event or state of affairs that can be described takes place while certain other things are not happening. For instance, someone may be driving a car without wearing a pink shirt, or he may be driving without the Prime Minister having given a speech that day. But in the majority of contexts, this information is likely to be irrelevant for the speaker's communicative purposes, and will thus not be mentioned. The present-day ratio of sin que to
Diachronic development of the in nitive in Spanish
249
sin+in nitive, as seen in Fig. 4.33 above, which roughly corresponds to the ratio of non-coreferential to coreferential sin-clauses, shows that in this semantic domain, `non-action' by the main clause subject is considered relevant enough to mention far more often than `non-action' by an entity other than the subject of the main clause (S1 ). Whilst the overall distribution between nite clauses and the corresponding prepositional in nitives is, to some extent, a re ection of the number of non-coreferential and coreferential clauses, there are several factors working against this idealized correspondence pattern. One powerful factor is analogical levelling, a common process in language change. Once a semantic or relational concept, such as nality, becomes strongly associated with a particular syntactic structure, there is a tendency towards an increased use of the respective structure, even in contexts that would normally call for the use of the numerically weaker structure. This can, for instance, be observed with para+in nitive, which occasionally appears even in non-coreferential contexts. Such spreading of prepositional in nitives is made possible by two things: pragmatic control and the Overt Subject In nitive construction. Wherever the subject of the dependent clause is not the NP predicted by means of the syntactic default pattern, either the context allows an unambiguous assigning of S2 -reference, or alternatively the subject can be speci ed overtly within the in nitival clause. These two mechanisms of clarifying the identity of S2 are found most commonly with those prepositional in nitives that already have a dominant position in relation to their nite counterparts, in particular para+in nitive and sin+in nitive, which together account for 75% to 85% of all prepositional OSI constructions in present-day Spanish. A similar analogical tendency is visible for semantic relations which favour nite dependent clauses. One such case is the conditional domain, in which the conjunction si is clearly dominant (Fig. 4.37), even in coreferential nite clauses. Despite the gradual emergence of al+in nitive as a non- nite alternative, si remains rmly established as the standard way of rendering conditionality; the special status of si as the marker for conditional clauses par excellence is further reinforced by the special temporal syntax of conditional clauses, and by the fact that si is morphologically unrelated to its potential competitors.
250
Chapter 4
Similarly, aunque has, at least until recently, been so dominant as the typical marker of concession that it is frequently found in coreferential concessive clauses, as in sentence (263). (263) ... la mujer hablaba y se detenía mirando al the woman talked
and refl remained lookin
aunque ella no estuviera though
suelo [...]
at.the oor
desocupada.33
she not be.3sg.pst.sbjv unoccupied
... the woman talked and remained staring at the oor though she had things to do. A further, and somewhat unusual example of the nite structure retaining its dominant position is desde que. Unlike the semantically similar después de+in nitive, desde+in nitive is virtually absent from Spanish before the end of the 20th century, whereas desde que is found frequently from the 16th century onwards. There does not appear to be any particular pragmatic motivation for this, as coreferentiality is common in this type of temporal construction. What is more, hasta+in nitive, the corresponding prepositional in nitive for future time reference, is documented since the 15th century (cf. Fig. 4.19). The reasons for this dominance of nite desde que, even for coreferential clauses, remain to be determined. 4.4.3 Statistical peculiarities during the Golden Age During the `Golden Age' or `Classical period' of Spanish, roughly the 16th and 17th centuries, the majority of graphs presented above show some irregularity and almost erratic variation in comparison with the periods before and afterwards. This is the result of the combination of two separate features of Golden Age Spanish.
33
Marías, p.76.
Diachronic development of the in nitive in Spanish
251
It is generally accepted that the `Classical period' is a period of linguistic transition and change in Spanish34 , and as a result of this, texts written within years of one another will often vary considerably in the degree to which they preserve old structures or introduce the newly evolving ones. Though it cannot be claimed that prepositional in nitives are a novel structure during this period (with the exception of al+in nitive, which starts becoming more frequent round about 1500), texts looking back to the older language tend to avoid the more recently introduced prepositional in nitives. More signi cant for the purpose of this study is the fact that the usage of most prepositional in nitives experiences a temporary peak during the `Golden Age', and subsequently drops back to a lower level around the mid 17th century, from where the overall, more gradual rise continues.35 An explanation for this can be found in the fact that `Golden Age' authors generally favour a style of comparatively complex syntax. Fig. 4.25 above shows that not only in nitival clauses have a peak during this time; nal subordination also suddenly increases by around 10% around the year 1500. Gerundial clauses (Fig. 4.7) are even more affected by this, increasing by 50% before dropping back to their previous level, all during the 150 years between 1500 and 1650. The reasons behind this stylistic preference are primarily socio-cultural and thus beyond the scope of this study, but there also appears to be a more general, linguistic tendency for newly emerged structures to enjoy such popularity among language users that they temporarily `overshoot the aim'36 before dropping back to a more sustainable level. A similar phe-
34
35
36
E.g. Riiho (1979: 235):`En muchos aspectos [...] la lengua clásica puede considerarse como une especie de línea divisoria que separa la fase antigua de la moderna.' Local factors can counteract this pattern, as in the case of para+in nitive, which completes its takeover of the nal domain from por during this time. Once established as the prototypical nal construction, its usage continues to rise at the expense of alternative nite constructions such as a+in nitive, which in turn becomes increasingly associated with its grammaticalized, semantically bleached use as a complementizer. This is a cross-linguistic phenomenon. A recent example from English is the ubiquitous `looking to'+in nitive.
252
Chapter 4
nomenon is visible with en vez de+in nitive (Fig. 4.20) during a much later period. An afterthought to these observations about the statistical peculiarities of `Golden Age' syntax is the possibility that we might, from a syntactic point of view, currently be in the midst of a ` Golden Age'. A number of diagrams presented in this chapter certainly show unexpected changes in curve shape at the beginning of, or during, the 20th century. In view of the recent emergence of several new prepositional in nitives, it does appear that the system may be going through a renewed period of accelerated change, the outcome of which is yet unknown. Judging by the fact that many of the new prepositional in nitive constructions are found rst, and most frequently, in journalistic register texts, it appears likely that an important factor in this process is the increased importance of the mass media in modern society, which by their presence in all walks of life allow their stylistic features to permeate the whole language more easily.
Chapter 5 Portuguese and Spanish developments compared
As discussed in Section 3.1, Spanish and Portuguese in nitival subordination follows roughly the same combination of syntactically and pragmatically based rules, despite the fact that the Portuguese in nitive is more versatile than its Spanish counterpart because it can be morphologically in ected for person (cf. Section 3.3). In this chapter, a closer look will be taken at the differences between the development of prepositional in nitive constructions in the two languages from a diachronic perspective. In the gures of this section, the dotted lines represent the Portuguese structure, the solid line stands for its Spanish counterpart.
Similarities and differences in diachronic development
5.1
5.1.1 A comparison of the overall frequency of prepositional in nitives Fig. 5.1 shows the differential development of the prepositional in nitive in Spanish and Portuguese.
per 100000 words
1600 1400 1200 1000 800 600 1150
1200
1250
Figure 5.1
1300
1350
1400
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
1900
1950
Spanish and Portuguese prepositional in nitives through time
2000
254
Chapter 5
Whilst Spanish sees a steady increase from a comparatively low level up to around 1600, prepositional in nitives are already well established in Portuguese in the early Middle Ages, and their frequency remains relatively stable until the 19th century. During this period of stability in Portuguese, Spanish does not merely catch up; the number of prepositional in nitives rises from around 55% of that in Portuguese during the Middle Ages to around 140% in the Golden Age Spanish; only in the 19th century does Portuguese in nitive usage close this gap, reaching a level similar to that found in Spanish. How this overall development pattern translates to that of the individual prepositional in nitives will be discussed in the following sections. 5.1.2 Similar developments in Spanish and Portuguese Considering the common history and the degree of synchronic similarity between Spanish and Portuguese, one might initially expect syntactic structures shared by both languages to have developed parallely. This is, indeed, the case for a number of prepositional in nitive constructions, applying to structures that have emerged or increased in terms of frequency, but also to structures that have suffered a decline. A good example of this is the temporal domain of anteriority of the main clause, represented largely by ante(s de), as well as Spanish hasta (fasta in Old Spanish) and cognate Portuguese até. Fig. 5.2 illustrates this. The temporary peak in Spanish during the Golden Age has been discussed in the previous section; a renewed split can be observed during the 20th century, with an increase in Portuguese since the mid 19th century. But of course we cannot expect developments in separate languages to run parallel to each other, as we have already seen in the previous chapter that each prepositional in nitive construction has its own dynamics and follows an individual trajectory, though usually within a roughly predictable general pattern. In this particular case, the Portuguese increase is due to an increase in antes de+OSI since the mid 19th century by around 35%, which must be understood as a clear instance of this construction becoming increasingly entrenched at the expense of the corresponding nite clauses.
Portuguese and Spanish developments compared
255
30 per 100000 words
25 20 15 10 5 0 1150
1200
Figure 5.2
1250
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish and Portuguese temporal prepositional in nitives with anterior main clause
The recent decline in Spanish, on the other hand, is entirely due to a strong drop in hasta+in nitive, as seen in Fig. 5.3. A reduction in frequency of a prepositional in nitive construction is somewhat unexpected in view of the overall development pattern and is thus likely to have been triggered by a local factor. A possible explanation is that, in view of the morphosyntactic and semantic correspondence between desde+in nitive and hasta+in nitive, the virtual absence of desde+innitive might have started causing some analogical resistance to hasta+in nitive. 20
per 100000 words
17,5
antes de hasta
15 12,5 10 7,5 5 2,5 0 1150
1200
1250
1300
1350
1400
1450
1500
1550
1600
1650
1700
1750
1800
1850
1900
1950
2000
year
Figure 5.3
Spanish hasta+in nitive vs. antes de+in nitive
Fig. 5.4 shows that this does not occur in the same way in Portuguese, where até continues to enjoy increasing popularity.
per 100000 words
256 22,5 20 17,5 15 12,5 10 7,5 5 2,5 0 1150
Chapter 5
1200
1250
1300
Figure 5.4
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish hasta+in nitive vs. Portuguese até+in nitive
per 100000 words
A further example of very similar development in both languages from its emergence onwards is Sp. al+in nitive/Ptg. ao+in nitive (Fig. 5.5), the frequency and increase of which is almost equal in the two languages from the earliest texts onwards until approximately the end of the 19th century. During the 20th century, the frequency of the construction remains relatively stable in Spanish, whilst in Portuguese it falls to a level clearly below that of Spanish. 100 90 80 70 60 50 40 30 20 10 0 1150
1200
1250
1300
Figure 5.5
1350
1400
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish al+in nitive vs. Portuguese ao+in nitive
The in nitival concessive construction with Sp. a pesar de, pese a/Ptg. apesar de also develop along very similar lines. Comparatively infrequent in both languages (not exceeding 10 per 100,000 words in any text, which with the amount of data processed here is too low to produce a statistically reliable graph), they rst appear in the latter half of the 19th century and subsequently experience an increase in usage frequency in the second half of the 20th century. More surprisingly, perhaps, por undergoes a remarkably similar development in both languages (5.6), despite the semantic shift it undergoes (cf. Section 4.2), and despite the fact that, in its modern causal
Portuguese and Spanish developments compared
257
per 100000 words
meaning, it has a powerful competitor in the gerund; the impact of these factors appears to be similar in both languages, causing the observable gradual decline in the usage frequency of por+in nitive. 180 160 140 120 100 80 60 40 20 1150
1200
1250
1300
1350
Figure 5.6
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Por+in nitive in Spanish and Portuguese
5.1.3 Portuguese prepositional in nitives predating their Spanish counterparts
per 100000 words
A considerable number of prepositional in nitives that are only present to a limited extent or not at all in Medieval Spanish texts are already well-established in Portuguese during the same period. This is particularly signi cant because it includes several of the prepositional in nitives found most frequently in later Spanish texts, such as sin/sem+innitive (Fig. 5.7), después/depois de+in nitive (Fig. 5.8), as well as para+in nitive (Fig. 5.10). 90 80 70 60 50 40 30 20 10 0 1150
1200
1250
1300
Figure 5.7
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
Spanish sin+in nitive vs. Portuguese sem+in nitive
1950
2000
258
Chapter 5
per 100000 words
17,5 15 12,5 10 7,5 5 2,5 0 1150
1200
1250
Figure 5.8
1300
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish después de+in nitive vs. Portuguese depois de+in nitive
per 100000 words
14 12 10 8 6 4 2 0 1150
1200
1250
1300
1350
Figure 5.9
1400
1450
1500
1550
1600 year
1650
1700
1750
1800
1850
1900
1950
2000
1950
2000
Ante(s de)+in nitive in Spanish and Portuguese
per 100000 words
300 250 200 150 100 50 1150
1200
1250
1300
1350
Figure 5.10
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
Para+in nitive in Spanish and Portuguese
Spanish después de+in nitive attains the same frequency as Portuguese depois de+in nitive within a hundred years of its emergence (Fig. 5.8), and sin draws equal with sem by the 16th century (Fig. 5.7). Similarly, temporal ante(s de) only appears in Spanish around 1500, but experiences a dramatic increase within a hundred years (Fig. 5.9). After a parallel drop before 1700, followed by two hundred years of rel-
Portuguese and Spanish developments compared
259
ative stability, by the beginning of the 20th century, antes de+in nitive is once again more frequent in Portuguese than in Spanish, with Spanish catching up towards the end of the same century. Interestingly, the pattern of the Portuguese development predating that in Spanish repeats itself a second time here, which might be seen as an indication that the Portuguese remains the more innovative of the two languages regarding prepositional in nitives. The relative frequency in Spanish and Portuguese of para with its early variants pora and pera, on the other hand, develops in a somewhat different way (Fig. 5.10): after an initial period of far lower frequency than in Portuguese, the Spanish construction with para then increases at a steady rate between the 15th and the 18th century, far exceeding its Portuguese counterpart, and remains at this high level to the present day. The stronger performance in Medieval Portuguese can be explained in part by the fact that para, pera started to take over the nal domain from por earlier in Portuguese than in Spanish.1 The examples in this section show that the semantic shift between por and para is not the only area in which linguistic developments come earlier in Portuguese than in Spanish. As illustrated in Fig. 5.1 above, Portuguese already has a strong prepositional in nitive at the time of the earliest documents, whereas Spanish is still in the midst of developing it. This instance of syntactic conservatism of Old Castilian in comparison to Old Portuguese is somewhat contrary to the traditional view that Portuguese is the more conservative of the two languages2 , but it is in keeping with the also innovative in ected in nitive, both of which t into the picture of a structural framework in which the in nitive plays an important part in the syntax of subordination. A similar development pattern can be identi ed for prepositional in nitives that are currently experiencing increasing popularity in the modern language. A good example is the nal Sp. a n de/Ptg. a m de+in nitive, which has been latently present in both languages since the Middle Ages, but only extremely sporadically. Fig. 5.11 illustrates
1 2
Riiho (1979: 211 213). Portuguese is another so-called conservative language, allegedly owing to its geographical distance from the innovatory centres. , Posner (1996: 327).
260
Chapter 5
per 100000 words
how Portuguese experiences a much more vigorous rise in this structure, whilst in Spanish a similar increase has not (yet) set in. Though the general trend in both languages is for an increasing number of prepositional in nitive constructions to experience an increase in frequency over time, the fact that a n de+in nitive is semantically equivalent to nal para+in nitive means that what distinguishes Spanish and Portuguese in this case is essentially a lexical choice that does not affect the overall increase of purposive in nitival clauses in Spanish, which in their majority make use of para (cf. Fig. 4.34). 11 10 9 8 7 6 5 4 3 2 1 0 1700
1725
1750
Figure 5.11
1775
1800
1825
1850 year
1875
1900
1925
1950
1975
2000
Spanish a n de+in nitive vs. Portuguese a m de+in nitive
5.1.4 Underspeci ed prepositional in nitives Whilst there is no widely used prepositional in nitive that emerges in Portuguese much later than its cognate in Spanish, Portuguese com+in nitive and em+in nitive are far less commonly used than their Spanish counterparts con+in nitive and en+in nitive, as seen in Figs(5.12) and (5.13). Even though com+in nitive is already present in Portuguese at the time when con+in nitive is only just emerging in Spanish, it soon becomes considerably more frequent in Spanish. En/em+in nitive is already present in both languages from the very earliest texts, but it remains, on average, around twice as frequent in Spanish as it is in Portuguese throughout the examined period. As this pattern differs considerably from the one found for most other prepositional in nitives when comparing their development in Spanish and Portuguese, an explanation must be sought by establish-
Portuguese and Spanish developments compared
261
per 100000 words
25 20 15 10 5 0 1150
1200
1250
1300
1350
per 100000 words
Figure 5.12 100 90 80 70 60 50 40 30 20 1150
1200
1250
1300
Figure 5.13
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
1950
2000
Spanish con+in nitive vs. Portuguese com+in nitive
1350
1400
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
Spanish en+in nitive vs. Portuguese em+in nitive
ing what sets these two prepositional in nitives apart from the rest. The feature they share is their comparative imprecision, or semantic underspeci cation. In Section 4.2.1 it was observed that in Spanish, such semantically underspecified prepositional in nitives do not experience an increase in frequency in the same way as the rest, because the central function of the prepositional in nitive is the precise speci cation of the logical relation between the main and dependent clause, with the gerund being chosen for non- nite clauses where no such precision is intended (cf. Section 2.4.3). The even lower incidence of underspecified prepositional in nitives in Portuguese suggests that the association of the prepositional innitive with an overt and clear marking of interclausal relations is even stronger. This analysis is con rmed by the fact that the lower incidence of underspecified prepositional in nitive constructions in Portuguese, such as those with em and com, is counterbalanced by a stronger use of gerundial clauses, as shown in Fig. 5.14.
262
Chapter 5
1000
per 100000 words
900 800
Spanish Portuguese
700 600 500 400 300 200 100
1150
1200
1250
1300
1350
Figure 5.14
1400
1450
1500
1550 year
1600
1650
1700
1750
1800
1850
1900
1950
2000
Gerundial clauses in Spanish and Portuguese
5.1.5 Discussion of the data What the comparison between the diachronic development of different prepositional in nitives in this section has shown is, rst and foremost, that their evolution has, by and large, proceeded in a similar way, irrespectively of whether the respective preposition already participated in this construction type at the time of the rst available documents, or whether it emerged at a later stage. This observation applies to both Spanish and Portuguese. As the degree of cultural exchange or bilingualism during most of the examined period was not suf ciently signi cant to make large-scale syntactic borrowing or convergence a feasible explanation, it must be assumed that these parallel developments occurred independently. This is supported by the fact that several, but not all prepositional in nitives emerged in Portuguese at an earlier date than their Spanish counterparts. Despite the overall similarities in their development, characterized by a development that roughly resembles the S-curve predicted for processes of diffusional change by Kroch (2001), each prepositional in nitive follows its own trajectory. There is no wholesale syntactic shift, as new prepositions continue to join the group that takes in nitives, and it appears that the structure of both languages remains open to the addition of new prepositional in nitives. At the same time, it has been observed that there is a clear resistance to certain types of prepositional in nitives. Whilst semantically underspecified prepositional in nitives have to compete with the gerund, and therefore show a different, more subdued development than
Portuguese and Spanish developments compared
263
percentage
the more semantically speci c constructions, certain semantic types of dependent clauses support the use of the in nitive frequently and at an early stage, whilst others remain largely or completely resistant to this syntactic type. The clear correspondence between Spanish and Portuguese in this respect suggests that the same pragmatic principles described for Spanish in Section 4.4.1 are a crucial factor in encouraging or inhibiting the use of the in nitive in dependent clauses, depending on their type of semantic relation to the main clause: in actual language usage, some clause types, such as purpose, are typically used coreferentially, whilst others, like concessivity, appear more frequently with disjoint subject reference. Evidence that this is true for Portuguese, too, is provided in Fig. 5.15, which largely mirrors the corresponding Spanish distribution seen in Fig. 4.43: on average, around 75% of nal clauses are coreferential, compared to around 30% of concessive clauses. 100 90 80 70 60 50 40 30 20 10 0 1350
1400
1450
1500
1550
1600
1650
1700
1750
1800
1850
1900
1950
2000
year
Figure 5.15
5.2
Percentage of coreferential nal and concessive dependent clauses in Portuguese (dotted = nal, solid = concessive)
OSIs and the in ected in nitive
5.2.1 History of the Portuguese personal in nitive The morphological origin of the Portuguese in ected in nitive has been widely discussed in the literature. The oldest and probably most popular theory, proposed by early Romance philologists such as Diez (1881)
264
Chapter 5
and Meyer-Lübke (1894), is that the in ectional endings are formed on the basis of those found in the future subjunctive paradigm.3 According to Vasconcelos (1900) and Bourciez (1956), this would have been facilitated by the homonymy, in regular verbs, of the in nitive and the future subjunctive in the 1st and 3rd person singular. In a similar vein, Zauner (1921) suggests that both the future subjunctive and the in ected in nitive forms are based on the Latin perfect subjunctive paradigm. An alternative source, the Latin imperfect subjunctive paradigm, was rst suggested by Wernecke (1885) and has found support in Gamillscheg (1913), Rodrigues (1914), Michaëlis (1918), Meier (1950) and Sten (1952). On the other hand, Otto (1888), Michaëlis (1891) and Maurer (1951, 1968) have suggested that the in ectional endings are the result of paradigmatic extension of a suf xed 1st person plural subject pronoun -nos > -mos. To the present day, this question has not been conclusively resolved. The most plausible source, from a usage-based point of view, would appear to be the future subjunctive, as there is a considerable amount of formal and semantic overlap between it and the in nitive, as exempli ed in (264) and (265), which may have facilitated an extension of future subjunctive forms to in nitival contexts. Overlap between future subjunctive and in nitive: (264) a.
Quando chegar when
b.
o meu pai, falar-lhe-ei.
arrive.3sg.fut.sbjv the my
Ao chegar
father, talk.to.him.1sg.fut
o meu pai, falar-lhe-ei.
at.the arrive.inf.3sg the my
father, talk.to.him.1sg.fut
When my father arrives, I'll talk to him. Extension of future subjunctive morphology to the in nitive: (265) a.
Quando chegarem when
3
os meus pais, falar-lhes-ei.
arrive.3pl.fut.sbjv the my
parents, talk.to.them.1sg.fut
Es [...] wird nach dem Muster des Fut. Konj. zu -ar als Inf. bei der 3. Sing. als Subj. gebildet 1. -ar, 2. -ares, Plur. -armos -ardes -arem. (Meyer-Lübke, 1894: 158)
Portuguese and Spanish developments compared
b.
265
Ao chegarem os meus pais, falar-lhes-ei. at.the arrive.inf.3pl the my
parents, talk.to.them.1sg.fut
When my parents arrive, I'll talk to them. Though the emergence of the in ected in nitive predates the earliest documents, there is some evidence that it may have been a recent innovation in Old Galaico-Portuguese, as in ection appears not yet to be obligatory4 in the presence of an overt subject, as seen in (266). (266) Era costume [...] de os cavaleiros andantes seer was custom
of the knights
errant
be.inf.uninfl
5
recebidos em lugares estranhos. received
in
places
strange
It was customary for the errant knights to be received in strange places. The existence of the personal/in ected in nitive in Portuguese could, theoretically, pose a threat to nite subordination, especially since tense and mood of dependent clauses are frequently predictable from the semantic context and from the discourse situation. It has been shown in Section 3.3.2 that there is little danger of the in ected in nitive ousting nite subordination synchronically, and that it more frequently impinges upon the domain of the `plain' unin ected in nitive. It is therefore interesting to determine whether a diachronic tendency for the personal in nitive to gain ground at the expense of the impersonal in nitive. Fig. 5.16 shows that this is not the case, thus disproving an unsubstantiated claim to the contrary by Dias (1959: Ÿ313).
4 5
Cf. also Molho's modern example from Brazilian Portuguese (Section 3.3, Footnote 43), which is, however, probably a modern regional development. A demanda do Santo Graal, Ÿ423, quoted by Maurer (1968: 152).
266
Chapter 5
3000 per 100000 words
2500 2000 1500
personal infinitive impersonal ‘plain’ infinitive
1000 500 0 1250
1350
Figure 5.16
1450
1550
year
1650
1750
1850
1950
Personal and impersonal in nitives in Portuguese
per 100000 words
Whilst a slight, gradual rise in the frequency of impersonal in nitive can be observed since medieval times, the number of personal innitives is far lower, and declines further over the same period. The larger scale of Fig. 5.17 makes it possible to identify a steady drop from the Middle Ages until around 1800; whether the subsequent slight increase constitutes a permanent reversal of the trend remains to be seen. As illustrated by Fig. 5.20, the recent rise can be attributed entirely to a rise in prepositional in nitives with a, which will be discussed further below. 500 450 400 350 300 250 200 150 100 50 0 1250
1350
1450
Figure 5.17
1550
year
1650
1750
1850
1950
Personal in nitives in Portuguese
The decline in the overall usage of personal in nitives is also visible if we look speci cally at prototypical adjunct clauses. Representative of these, Fig. 5.18 charts the development of two typical, well-established in nitival adjunct types with the personal in nitive, with sem and para.
Portuguese and Spanish developments compared
267
per 100000 words
90 80
solid, boxes: para/pora
70
dotted, triangles: sem
60 50 40 30 20 10 0 1350
1450
Figure 5.18
1550
1650
year
1750
1850
1950
Personal in nitives with para/pora and sem in Portuguese
Whilst it has been seen that the impersonal prepositional in nitive adjuncts with para and sem have been increasing in frequency over the centuries, in Portuguese as well as Spanish (Fig. 5.7 and Fig. 5.10), the corresponding adjuncts with the personal in nitive appear to be going the opposite way, experiencing a gentle long-term decrease (Fig. 5.18). This is probably due to the trend towards using nite clauses for genuinely non-coreferential cases, whilst the personal in nitive is used mainly in cases where the subject could potentially be non-coreferential, but is in fact coreferential (cf. Section 3.3.2). Fig. 5.19 shows how, in contrast to sem and para, the concessive adjunct with apesar de is gaining ground. But viewed in perspective, apesar de with the personal in nitive does, nevertheless, not have a major role in the language as a whole: despite the increase illustrated in Fig. 5.19, it occurs approximately a hundred times less frequently than para+personal in nitive, and only approximately one in twenty noncoreferential concessive clauses use the in nitival construction. Similar to what was observed regarding para and sem, there appears to be a tendency to use nite clauses with ainda que or embora for cases of genuine disjoint subject reference, whilst apesar de with the personal in nitive is often chosen to clarify the identity of a coreferent subject, for example if the concessive adjunct is at a distance from the main verb.
268
Chapter 5
0.45 per 100000 words
0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 1450
1550
1650
1750
year
1850
1950
The personal in nitive with apesar de in Portuguese
Figure 5.19
Fig. 5.20, on the other hand, presents an interesting contrast. On the one hand, there is a slight decline of de with the personal in nitive. Since de has prototypical complementizer status in most cases, the decrease of its use with the personal in nitive is not unexpected, as it is primarily adjuncts, not complements, that make use of the possibility of marking a non-coreferential subject (S2 ). 140 per 100000 words
120 100
solid, boxes: ‘de’ dotted, triangles: ‘a’
80 60 40 20 0 1350
1400
1450
1500
1550
1600
1650
1700
1750
1800
1850
1900
1950
2000
year
Figure 5.20
Personal in nitives with a and de in Portuguese
The sudden rapid increase in the use of a with the personal in nitive in the 19th century is more surprising, as a is also a typical complementizer. However, what a closer analysis of the data reveals is that the sudden increase in popularity of [a+personal in nitive] is not due to an increased tendency to use personal in nitives in complement clauses. On the contrary, the construction triggering this frequent use of a with the personal in nitive is the typically peripheral use of a+in nitive that has come to replace the gerund in European Portuguese, and which
Portuguese and Spanish developments compared
269
is employed to provide additional circumstantial, pragmatically backgrounded information, as in (267). (267) Chegaram as meninas de fazer compras na arrived
the girls
cidade, a
from making purchases in.the town
a
falarem alegremente. talk.inf.3pl happily
The girls came back from shopping in town, talking happily. In Section 3.3.2 it was shown that the OSI is more typically used in adjuncts than in complements because its main purpose is to disambiguate the identity of the dependent clause subject. The pragmatically backgrounded `gerund-like' use of a+in nitive is particularly susceptible to potential ambiguity regarding its subject, as it is often located far away from the main verb within the sentence, and is frequently even separated from the main clause by a short pause. The increase of [a+personal innitive] is, therefore, fully in line with the personal in nitive's primary function of increasing pragmatic clarity (cf. Section 3.3.3); it should be noted that in the vast majority of cases, the a-clause is coreferential and would therefore not, strictly speaking, require overt subject marking, as seen in (267). What makes this construction atypical in comparison with other in nitival adjuncts is the fact that it is pragmatically backgrounded, which has been shown to be a domain in which prepositional in nitives have otherwise been losing ground over the centuries (cf. Section 5.1.4). Position of the overt subject It has been seen in Section 3.3.4 that in modern Portuguese, the pre-innitival position is generally the preferred position for overt subjects, whilst the post-in nitival position is reserved for pragmatically marked purposes such as topicalization or contrastive focus. Mensching (2000: 28) notes that the high frequency of preverbal subjects [is] a fact that already seems to have been the case in the oldest medieval texts. An analysis of the diachronic corpus in Appendix B con rms this for pronominal subjects: up to the 17th century, very few instances of
270
Chapter 5
post-in nitival pronominal subjects can be found, while the pre-in nitival subject position is the norm (268). (268) Primeiramente, disse Álvaro Pais, por vós rstly
del-rei,...6
said
prpn
serdes
irmão
prpn for you.pl be.inf.2pol brother
of.the-king
In the rst place, said Álvaro Pais, because you are the king's brother... During the 16th century, a striking discrepancy between the position of pronominal and nominal OSI subjects can be seen. From the complete works of Luis de Camões (1524 1580), Otto (1888: 92 4) lists a total of 37 instances of personal in nitives with overt subjects. A closer analysis of his data reveals that of these, 34 are prepositional adjuncts with de, por, para and sem, while the remaining three are subject clauses of the type `Me atormenta ver eu que...'7 ; the latter all have a post-innitival pronominal subject. In the prepositional OSI clauses, all seven instances of overt pronominal subjects are pre-in nitival, whilst only two of the 27 non-pronominal subjects occupy that position. In other words, there is a very clear tendency for pronominal subjects to occur preverbally and nominal subjects postverbally in OSI clauses. Though not making speci c reference to this distribution pattern, Otto believes that Camões, a renaissance author, is likely to have been strongly in uenced by the syntax of Italian, the language serving as the literary model of the time. An interesting parallel may thus be drawn to the fact that of the Italo-Romance varieties that have an OSI construction, none allows preverbal nominal subjects, but some do allow preverbal pronominal ones (cf. Ledgeway, 2000: 126 30).8 >From the 17th century onwards, however, the proportion of pronominal subjects that occur post-in nitivally also gradually in-
6 7 8
Fernão Lopes: Crónica de El-rei D. João I de Boa Memória., Pt. 1, Chapter 14: O Mestre de Avis compromete-se a matar o Andeiro, (ca. 1430), cf. Appendix B. Cf. Otto (1888: 93) N.B. In Old Italian, however, nouns as well as pronouns could occur pre-in nitivally (Mensching, 2000: 18, 102).
Portuguese and Spanish developments compared
271
creases. A possible explanation for how this position became available is by structural reanalysis of copular clauses such as (269), where the predicative pronoun vós is coreferent with the subject (`... por [vós] serdes vós' ). In the semantically similar sentence (270), vós occupies the post-in nitival position even though it is only the subject, the predicative element being `quem sois'. (269) ... única e puramente por serdes only
and purely
vós.
for be.inf.2pol you.pred
... solely because it's you (and nobody else). (270) ... única e puramente por serdes only
and purely
vós
quem sois.9
for be.inf.2pol you.sbj [who be.2pol].pred
... solely because you are who you are. Such reanalysis, followed by subsequent extension to non-copular contexts, might also have contributed to the fact that a post-in nitival subject generally carries contrastive focus, since the copular construction is in itself a focalizing construction10 . But irrespective of the possible in uence of the copular construction, it must be kept in mind that quite generally, not just in OSI clauses, postverbal position causes topicalization and pragmatic focus. (`Eu vou para a escola' vs. `Vou eu para a escola' ) 5.2.2 The history of the Spanish OSI construction In the very earliest Spanish documents, such as the Çid, the OSI construction is not attested (cf. Körner, 1983: 78). Considering the relatively small number and size of these documents, as well as the low overall proportion of OSI constructions in Spanish throughout its his-
9 10
António Vieira: Sermões Escolhidos, Ÿ10: Exortação e Oração Final (1655). São Paulo: Edameris, 1965. The related construction of the type `para ser ele a fazer' (so that it is him who does) is a very common intensifying/focalizing construction (cf. Section 3.3.4.)
272
Chapter 5
tory (From a diachronic point of view, the particle a, used by the verbs examined in the previous section to connect in nitival clauses, is a grammaticalized preposition (cf. Section 7.2.1). Synchronically, it can be seen either as the default complementizer, or as a morphological marker of the in nitive, similar to to in English. Some accounts of complementation, e.g. Givón's `integration hierarchy' between main and complement clause (Givón, 1984: 519), dis-
33
Activitatea în µar a AOAR, http://www.aoar.ro/pozitii/in2_ro.htm
302
Chapter 6
tinguish `in nitive complements' with the morphological marker on the one hand from `bare stem complements' without it on the other. Absence of the marker tends to coincide with a higher degree of semantic integration. This accurately predicts the situation in Romanian: temporal and modal notions are semantically highly integrated with the complement verb, and this semantic integration is iconically re ected by the absence of the subordinating morpheme and the increased proximity between the two verbs involved. `Pot veni' (`I can come') can thus be understood as an auxiliary+stem construction, as suggested by Mallinson (1988: 411). However, some doubt is cast upon the applicability of Givón's hierarchy, based on a correlation between semantic integration of main and complement verb on the one hand and the complement clause's syntax on the other, by the fact that both the periphrastic future and a putea/a ³ti have corresponding nite counterparts with the subjunctive (cf. Section 6.1.2). According to Givón, subjunctive complements should be far less integrated with their main verb than bare-stem complements, but this does not hold for Romanian. The analysis of a as a general morphological marker of the in nitive is further called into doubt by its absence in the construction with a avea + in nitival indirect question, as discussed in Section (6.1.3) above. Though a cannot, therefore, be understood to be an integral morpheme of the in nitive, perhaps the most compelling argument to analyse it as a largely obligatory component of the in nitive, rather than an independent complementizer, is the fact that a number of main verbs, such as a opri din a+in nitive and a impiedica de a+in nitive have their own, lexically determined prepositional complementizers (cf. Table 6.5). This analysis is further corroborated by the fact that all prepositional in nitives are preceded by the respective preposition and the particle a, as will become evident in the following section.
The in nitive in Romanian
6.2
303
Synchronic use of prepositional in nitives
6.2.1 Finite and non- nite adjuncts Apart from the auxiliary/modal constructions in Section (6.1.2), by far the most common environment for the use of the in nitive in Romanian is in prepositional clauses. As Table 6.6 shows, this construction stands in serious competition with its nite counterpart. Table 6.6
Proportion of nite and in nitival clauses with different prepositions in Romanian
Preposition/Conjunction spre a/sa˘ pentru a/ca sa˘ înainte de a/sa˘ prin a/faptul ca˘ ˘ a˘ a/sa˘ far ˘ pâna˘ (la) a/ce,sa,când în loc de a/sa˘ dupa˘ a/ce
‘in order to’ ‘in order to’ ‘before’ ‘by means of’ ‘without’ ‘until’ ‘instead of’ ‘after’
% inf.
% finite
98.75 90.17 81.33 62.71 39.45 18.85 6.51 0.05
1.25 9.15 18.44 36.46 60.55 81.15 93.49 99.5
% supine 0.67 0.22 8.83
Table 6.6 shows that the proportion of in nitival to nite clauses introduced by morphologically cognate preposition/conjunction pairs varies considerably. For instance, a strong tendency for adjuncts introduced by the typical [purpose] markers such as pentru and spre to be in nitival can be observed. To determine whether this is, in fact, due to the semantic properties of these prepositions, a separate, semantically-based analysis is required, as there is a certain asymmetry between the inventory of prepositions and corresponding conjunctions in Romanian. For example, the nite counterpart of spre a, i.e. the nal conjunction spre s , is extremely rare; on the other hand, the common nal conjunction ca s has no prepositional equivalent, and nal adjuncts can also be introduced by underspecified s on its own. A further construction not to be overlooked in this context is la+supine, which frequently involves nal meaning. The overall distribution of nal clauses in the 17 million word corpus of modern Romanian examined here, as seen in Table 6.7,
304
Chapter 6
shows that the in nitive is used in more than half of all nal dependent clauses. Table 6.7
Final clauses in Romanian
Final construction
Percentage
pentru a+infinitive ˘ ca sa+subjunctive ˘ sa+subjunctive spre a+infinitive ˘ pentru ca sa+subjunctive la+supine pentru+supine ˘ spre sa+subjunctive
45.77 % 25.18 % 15.73 % 7.62 % 4.65 % 0.62 % 0.34 % 0.09 %
total infinitive construction total finite construction total supine construction
53.39 % 45.56 % 1.05 %
6.2.2 Noun complements A further area in which in nitive complements are clearly preferred over their nite counterparts is in noun complements, as seen in Table 6.8. Table 6.8
Noun complements in Romanian
Complement Type noun + de a + infinitive noun + ca˘ + finite verb noun + de + supine noun + sa˘ + finite verb
Percentage 56.78 % 25.79 % 11.01 % 6.42 %
Nouns do not necessarily have a speci c agent or subject, and this would appear to be conducive to the use of in nitival complements sharing the non-speci c or `arbitrary' subject of the head noun, similar to
The in nitive in Romanian
305
(301) above. But while subjectless or `passive' noun complementation is one of the functions of the in nitive in other Romance languages such as Spanish, this domain is covered by the supine in Romanian34 . This is true for lexicalized constructions such as Sp. goma de mascar, Ital. gomma da masticare, Rom. gum de mestecat `chewing gum', as well as new concepts (306). (306) a.
Sp.: En efecto: para él, la casa es `una máquina de in effect
for
vivir' y reconcibe
him the house is a
machine
of
35
la manera de construir.
live.inf and reconceive.3sg the way
of constructing
In effect, for him the house is a `dwelling machine' and he conceives a new way of building. b.
Rom.: Locuinµa
nu este un obiect, o ma³in de
dwelling.place.def not is
an object
a machine of
locuit...36 dwell.sup
The dwelling place isn't an object, a dwelling machine... The Romanian in nitive, in contrast, is typically used coreferentially, as the complement of nouns with a speci c agent, either explicit as genitival agent (307), or pragmatically implied (308). (307) capacitatea sa de a cînta la toate instrumentele37 ability.def
his of to sing.inf at all
instruments.def
his ability to play any instrument
34 36 36 37
For an in-depth survey of the exact usage and limitations of the supine, see Manoliu-Manea (1985: 279 291). http://www.tam.itesm.mx/art/arquit/earqui05.htm Mircea Eliade: Sacrul si profanul, Cap.1, Cosmogonie ³i sacri ciu de construcµie Bucure³ti: Humanitas, 1995. România Literar 3, 24.1.1996; Lecturi. Monica Spiridon: Introducere in metoda lui Paul Cornea.
306
Chapter 6
(308) G sise
o metod ingenioas de a evita
nd.3sg.plup a method ingenious
acest cuvânt.38
of to avoid.inf this
word
He had found an ingenious way of avoiding this word. With nouns that cannot have an agent, thus making coreferentiality impossible, nite clauses (309) are the more common alternative, but innitival complementation is also found (310), particularly when there is a link such as ownership of the head noun by the complement subject. (309) Cred
c a venit timpul s
semneze
cu adev ratu-i
believe.1sg that has come time.def sbjv sign.3sg.sbjv with true.def-his
nume.
39
name
I think the time has come for him to sign with his real name. (310) Am
cel.puµin timpul de a-mi
have.1sg at.least
bea
cafeaua.
40
time.def of to-pron.1sg.dat drink.inf coffee.def
At least I have the time to drink my coffee. However, with this type of head noun, in nitival complementation is comparatively rare; for example, ocazia de a `the occasion to' takes the in nitive in a quarter of cases, and timpul de a `the time to' in only 6% of all tokens. This suggests that there is a link between coreferentiality and the choice of an in nitival clause, even when the agent of the noun does not appear as a syntactic subject. This will be examined in more detail in Section 6.4.1 below.
38 39 40
România Literar 41, 18.10.1995; Centenar Alexandru Rosetti. Toma Pavel: Un incoruptibil al culturii româane. România Literar 34, 26.9.1998; Dorin Tudoran: Pupat toµi piaµa unive'sit µi Riscul de µar ³i riscul de tat . România Literar 7, 19.2.1997; Iordan Chimeµ: Scrisori printre graµii Odysseas Elytis.
The in nitive in Romanian
307
6.2.3 Summary and analysis of the synchronic data The synchronic data presented in this section provides clear evidence that in nitival clauses are an integral part of modern Romanian syntax; whilst in nitival complements, apart from noun complements and the somewhat unclear case of a putea, are a relatively marginal phenomenon and mostly restricted to very formal register, this cannot be said of prepositional in nitival adjuncts. In the following section (6.3), it will indeed be seen that diachronically, the prepositional in nitive is on the increase. Romanian does not differ fundamentally from the majority of Romance varieties regarding the types of constructions that the in nitive can occur in. Examples have been provided to illustrate the various types of shared in nitival constructions. There are two things that do set Romanian apart from most other Romance languages.41 The rst is the fact that a number of verbs and prepositions rule out the use of an in nitival clause, most notably a vrea `to want', and concessive prepositions. However, similar local restrictions are found in other Romance languages, too. For instance, the Spanish prepositions a pesar de `despite' and temporal desde `since' also show a strong resistance to in nitival use (cf. 4.2.5), and it can be argued that decir `to say' cannot take in nitival complements, since the construction decir+in nitive is a seperate, aspectually different verb meaning `to claim'(cf. 3.1.3). The second difference is the optionality of the in nitive construction. In Section 3.1, it has been shown that in the majority of Spanish and Portuguese constructions, speci c reference patterns trigger obligatory replacement of a nite clause with its in nitival counterpart. In Romanian, in contrast, there is (except in certain auxiliary constructions) no syntactically binding context that necessitates the use of the in nitive.
41
However, the southern Italian dialects that were traditionally believed to have lost the in nitive almost completely (cf. Rohlfs, 1922: 219) appear to use the in nitive in much the same way as Romanian (cf. Ledgeway, 1998).
308
Chapter 6
This means that the principle difference between Romanian and its western sister languages is merely one of obligatoriness, and as a result also one of quantity. This is particularly notable with regard to subject and direct object clauses, whilst prepositional adjunct constructions appear to be a more favourable environment for the in nitive. It is these prepositional in nitives that will be examined in more detail in the following sections of this chapter.
6.3
Diachronic development of the prepositional in nitive
It is sometimes suggested (e.g. Close, 1974: 227) that the present-day usage of in nitives is largely the result of foreign in uence, in particular the imitation of similar French structures, by certain in uential 19th century authors42 . Close (1974: 220) argues that by the beginning of the 19th century, [the in nitive] was practically con ned to certain constructions , but that it was revived (not calqued) in other constructions due to the in uence of Western European languages. Whether a fashionable stylistic feature, used by one particular group of authors, is likely to permeate a language that, at the time, was largely that of an illiterate population, is debatable. Furthermore, even authors from the same period who are known not to have been in uenced by Western European style use the corresponding in nitive structures. Whilst the literary and sociological aspects of the issue are beyond the scope of this study, this section will provide statistical information on the development of one type of in nitive construction, the prepositional in nitive, which will, in turn, shed some new light onto the possible motivation for its emergence and evolution.
42
Among these famous authors are Ion Eliade R dulescu, Barbu Paris Mumuleanu, Iana V c rescu, Grigore Alexandrescu, according to Close (1974).
The in nitive in Romanian
309
6.3.1 The situation in Old Romanian From the earliest preserved Old Romanian (i.e. 16th century) texts, the in nitive is normally preceded by a, except with modal/auxiliary verbs and the construction [a avea+indirect wh-question]. A certain variation is found even with the future auxiliary a vrea43 and the deontic modal a putea44 , which occasionally link their in nitive complements with a; with the epistemic/future a avea a, the subordinator is obligatory45 . The partricle a also functions as a semantically bleached in nitival complementizer, both for object complements (311) and subject complements (312). a m nca p nre mea.46
(311) ... c uitaiu
that forget.1sg.pst to eat.inf bread poss.1sg
... that I forgot to eat my bread. (312) În sf nta beserecâ mai bine e a grâi in holy.def church
înµeles decât 10 mie sense
than
cinci cuvinte cu
more good is to speak.inf ve
de cuvinte neînµelese
ten thousand of words
words
with
în limbâ
incomprehensible in language
striinâ.47 foreign
In the holy church it is better to say ve words with sense than ten thousand incomprehensible words in a foreign language. But at the same time, a still retains its nal meaning in other contexts, introducing purpose adjuncts (313).
43 44
45 46 47
E.g. c veµi cu adev r a a a `that you will surely a nd', preface of the Palia de la Or ³tie (1581 1582), as edited by M. Avram in Iordan (1962: 166). E.g. cum s poatâ ³i ei propovedui ³i a spune... `how they, too, can preach and a say', epilogue of the Tîlcul Evanghelilor (1564), as edited by M. Avram in Iordan (1962: 162). E.g. s aib a dare `that he should a give', Sentinµ din 1588, as edited by M. Avram in Iordan (1962: 151). Psaltirea Hurmuzaki, Psalm of David No.101, as edited by F. Dimitrescu in Iordan (1962: 170). Epilogue of the Tetraevanghelul by Hans Benker (1561), as edited by M. Avram in Iordan (1962: 151).
310
Chapter 6
(313) ... ce-mi
era d ruit de Dumnezeu a da
rel-me.dat was given from God
la acest lucru.48
a give.inf to this
work
... which was given to me by God to dedicate to this work. Such in nitival purpose adjuncts are found relatively frequently; in fact, they occur in Old Romanian with almost exactly the same frequency as pentru/spre a+in nitive in the modern language, with approximately 9 instances per 10,000 words. In addition to these clear-cut cases, there are also numerous instances where a clear classi cation of a as either a nal preposition introducing an adjunct, or as the complementizer of a complement with a nal nuance, can be made, as in a înv µat a49 `to be taught (in order) to', or example (314). (314) ... ca.s
e
tare ³i put rnece a µine
so.that be.3sg.sbjv strong and powerful
de acole.
parte de ocen
to hold.inf part
of property
50
of there
... so that he may have the strength and power to keep that part of the property. What is crucial for the further development is that Old Romanian did not have prepositional in nitives apart from the nal construction with a. This is a result of the hybrid function of a as a grammaticalized, semantically bleached marker of the in nitive on the one hand, but as an introducing element for nal adjuncts on the other. The (in most contexts) obligatory nature of the sequence a+in nitive rules out replacing a by a different preposition; at the same time, the fact that a also still functions as a fully semantic preposition appears to make the sequence preposition+a+in nitive unacceptable.51 Therefore, the modern Roma-
48 49 50 51
From the introduction to an edition of Coresi's Evanghelia cu înv µ tur printed in 1580/81, as edited by M. Avram in Iordan (1962: 164 165). ibid. Sentinµ din 1588, as edited by M. Avram in Iordan (1962: 151). Note that this is not a necessary consequence; German um zu or older English for to is acceptable in these languages, even though zu/to retains a separate prepositional function.
per 100,000 words
The in nitive in Romanian
311
250 225 200 de a + infinitive 175 Linear regression, de a + infinitive 150 125 100 75 50 25 0 1560 1580 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 year
Figure 6.1
Romanian de a+in nitive
nian prepositional in nitive construction can only start evolving freely after a has all but lost its prepositional function. 6.3.2 The evolution of the prepositional in nitive In accordance with a retaining its prepositional status in some usages, one of the rst prepositions to take in nitives is de, as it is semantically imprecise or transparent (Simões Frões, 1995; cf. 3.3). Fig. 6.1 shows how de gradually begins to gain signi cance in the 18th century, increasing in frequency rapidly from the 19th century onwards. Even earlier, at the beginning of the 18th century, the rst appearance of abessive f r a+in nitive can be found. Signi cantly, it is from the very beginning not restricted to formal or literary register, but can, for instance, be found in the cook book O lume într-o carte de bucate, in a recipe for cooked bottle gourds (315). (315) Apoi las then
l sa
s
stea
un ceas - doao, pân ce-³i
leave.imp sbjv remains one hour - two,
bine zeama, f r
va
until that-refl.dat will
a-l pune pe foc.52
leave.inf well juice.def without to-it put.inf on
re
Then, leave it to stand for one to two hours, until it is well juiced, without putting it onto the re.
52
O lume într-o carte de bucate
Tigve, Brâncovenesc manuscript.
per 100,000 words
312
Chapter 6
27,5 25 22,5 20 17,5 15 12,5 10 7,5 5 2,5 0
fãrã a+infinitive Linear regression, fãrã a+infinitive
1560 1580 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 year
Figure 6.2
Romanian f r a+in nitive
Like de a+in nitive, abessive f r a also becomes signi cantly more frequent towards the end of the 19th century, as seen in Fig. 6.2. Temporal pân (la) a+in nitive is also found as early as the 18th century (316), used in a slightly different sense than its modern meaning `until'. To the present day, it appears to be a matter of personal preference whether or not authors use pân (la) a+in nitive, but it has, in any case, been available since the 18th century. (316) Am
socotit, pîna îmi
sînt minµile întregi ³i pîna a nu
have.1sg thought until me.dat are minds.def complete and until to not
m
cuprinde de tot sl biciunea, de mi-am
me.acc grip.inf
f cut
of all weakness.def of me.dat-have.1sg make.sup
diiata...53 will.def
I have decided, as long as my mind is sane and as long as weakness hasn't completely taken hold of me, to make my will... In the early 19th century, nal spre a+in nitive also catches on (Fig. 6.3). This is of some signi cance in the sequence of emerging prepositional in nitives, as the use of a new nal preposition in combination with a means that the latter is likely to have lost all or most of its nal force at this stage.
53
Diata Stanc i (1784).
per 100,000 words
The in nitive in Romanian
313
27,5 25 22,5 spre a+infinitive 20 Linear regression, spre a+infinitive 17,5 15 12,5 10 7,5 5 2,5 0 1560 1580 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 year
Figure 6.3
Romanian spre a+in nitive
Following on the heels of spre a+in nitive, nal pentru a+in nitive appears (Fig. 6.4). Once available, it rapidly becomes the most frequent prepositional in nitive, and together with the largely synonymous spre a, it effectively takes on the role of Old Romanian nal a+in nitive; the joint development of these two nal prepositional in nitives can be seen in Fig. 6.5. 100 90
per 100000 words
80 70 60 50
pentru a+infinitive Linear regression, pentru a+infinitive
40 30 20 10 0 1560 1580 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 year
Figure 6.4
Romanian pentru a+in nitive
per 100000 words
314
Chapter 6
110 100 90 80 final prep + infinitive 70 Linear regression, final prep + infinitive 60 50 40 30 20 10 0 1560 1580 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 year
Figure 6.5
Final prepositional in nitives in Romanian
per 100000 words
Instrumental prin a and temporal înainte de a follow suit, gradually gaining popularity from the end of the 19th century onwards. (Figs 6.6 and 6.7, respectively) 7 6 5 4
prin a+infinitive Linear regression, prin a+infinitive
3 2 1 0 1560 1580 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 year
per 100,000 words
Figure 6.6
Romanian prin a+in nitive
20 17,5 15 înainte de a+infinitive Linear regression, înainte de a+infinitive 12,5 10 7,5 5 2,5 0 1560 1580 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 year
Figure 6.7
Romanian înainte de a+in nitive
Interestingly, no counterpart to înainte de a with anterior main clause has, to the present day, gained comparable popularity. Neither
The in nitive in Romanian
315
per 100000 words
of the temporal prepositions dup or în urma (de) commonly introduce an in nitival clause. A development similar to that illustrated above for nal, abessive, instrumental and posterior temporal in nitival adverbial clauses, though on a smaller scale, can also be observed for substitutive în loc de `in place of, instead of'. This is signi cant because it shows that the prepositions taking in nitves are not necessarily monomorphic; however, the development visible in Fig. 6.8 does show a delayed onset of the rise in frequency, which only becomes evident in the mid-20th century. 5,5 5 4,5 4 3,5 3 2,5 2 1,5 1 0,5 0
în loc de a + infinitive Linear regression, în loc de a + infinitive
1560 1580 1600 1620 1640 1660 1680 1700 1720 1740 1760 1780 1800 1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 year
Figure 6.8
Romanian în loc de a+in nitive
To complete this survey, it must be mentioned that no concessive in nitival structure has emerged to the present day, the conjunction de³i covering the entire concessive domain. The concessive preposition în ciuda does not take in nitival clauses. This may be due to the fact that it requires its argument to take genitive case, for which the in nitive has no morphological provisions, as discussed in Section 3.4.3. On the other hand, a similar resistance to concessive in nitival clauses in Spanish and Portuguese, as illustrated in the respective chapters, would suggest a reason that is common to all three languages; this will be discussed further in Section 6.4. The causal domain has similarly remained exclusively nite and gerundial; the preposition din cauza does not take in nitival clauses, and the clausal conjunctions pentru c , indc , deoarece and întrucât do not have prepositional counterparts.
316
Chapter 6
6.3.3 The proportion of in nitival and nite clauses through time
per 100000 words
A question raised by the increase in prepositional in nitives over the last two centuries is how this has affected the frequency of the alternative nite constructions. The diachronic comparison of nite and innitival nal clauses (cf. Table 6.7 above) in Fig. 6.9 shows that there clearly is an interdependence, leading to a convergence of the respective frequencies and to approximate parity between the nite and in nitival construction in the modern language. 225 200 175 150 125 100 75 50 25 1880
Figure 6.9
final prep + infinitive final conjunctions
1890
1900
1910
1920
1930
1940 year
1950
1960
1970
1980
1990
2000
Final conjunction with nite clause vs. nal preposition with in nitive in Romanian
per 100000 words
The comparative development of pentru a+in nitive and its nite counterpart pentru ca (s ), illustrated in Fig. 6.10, shows that pentru as marker of purpose has been gaining popularity in both the nal and the in nitival domain; here, the greater increase in in nitival usage has not caused a decrease of the corresponding nite structure in absolute terms, but is nevertheless likely to be responsable for the comparatively small increase of nite pentru ca (s ). 100 90 80 pentru a + inifinitive 70 pentru ca 60 Linear regression, pentru a+infinitive 50 Linear regression, pentru ca 40 30 20 10 0 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 year
Figure 6.10
Pentru ca (s )+ nite clause vs. pentru a+in nitive in Romanian
The in nitive in Romanian
6.4
317
Pragmatic causation
6.4.1 The subject of Romanian in nitival clauses In the previous chapters, a link between the pragmatic likelihood of subject coreference on the one hand, and the increase in prepositional in nitives on the other, was observed for Spanish and Portuguese. In order to establish whether similar mechanisms are responsible for the developments in Romanian, the rst step must be to take a closer look at typical subject reference patterns in in nitival clauses. In Section 6.2.3, it was observed that unlike the majority of Romance languages, Romanian never requires the use of the in nitive in coreferential dependent clauses, except in auxiliary constructions. On the other hand, these auxiliary constructions and the verbs in Table 6.2 suggest that there is, nevertheless, a connection between coreferentiality and the in nitive, whilst with the causative verbs in Table 6.5 the dependent clause subject is coreferent with the direct object of the main verb. These patterns resemble the pragmatically-based default patterns established for Spanish and Portuguese in Section 3.1, but as in Spanish and Portuguese, they do not apply in all cases. On the one hand, these default patterns are not binding. As seen in sentences (292) and (305), if there is suf cient pragmatic clarity regarding the identity of the subject of the dependent clause, there is no need for PRO to receive the coindexation that would be assigned by default. On the other hand, Romanian allows in nitives to have overt subjects. Mensching (2000: 37) goes so far as to claim that [g]enerally, it seems that whenever speakers accept an in nitive construction, they also accept the fact that it may have a speci ed subject. Though this is technically the case, overt subjects in in nitival direct object complements of verbs such as those in Tables 6.2, 6.4 and 6.5 are virtually always those predicted by the standard assignment pattern. Their function is merely to repeat and emphasize the subject, and they are typically accompanied by an emphatic pronoun însu³i `himself', or by ³i `too, also'. Sentence (317) illustrates the use with an intrinsically coreferential verb; it remains ambiguous whether ³i eu is, in fact, an overt in nitival subject or the postposed subject of the main verb îndr znesc, as these must always be identical.
318
Chapter 6
(317) Când a venit momentul [...] s when has come moment.def
îndr znesc a semna ³i eu
that dare.1sg
to sign.inf also I
o.astfel.de rubric ...54 such.a
rubric
When the time has come that I, too, dare to sign such a rubric... Sentences (318-319) illustrate how causative verbs can also have emphatic overt subjects, but that this subject must obligatorily be coreferent with the direct object of the main verb. (318) M-a
obligat a-l
ajuta ³i eu/∗tu pe
fratele
me.acc-has obliged to-him.acc help.inf also I/∗you domkr brother.def
meu. poss.1sg
He put it upon me that I/∗you, too, should help my brother. (319) M-a
sf tuit a-l
c uta
eu însumi/∗tu însuµi pe
me.acc-has advised to-him.acc search.inf I
preotul satului
ca s
îl
myself/∗you yourself domkr
întreb despre in nitive.
priest.def villagr.def.gen so that him.dat ask.1sg about
in nitives
He advised me that I myself/∗you yourself should nd the village priest to ask him about in nitives. Whilst in the above two examples it is logically, or at least pragmatically, impossible for the complement to have a subject other than the one predictable from the main verb, optionally coreferential verbs such as those in Table 6.4 do not take in nitival complements with overt non-coreferential subjects, either; the association of the in nitive with subject coreference appears to prevail with these verbs. Overt in nitive subjects have a more meaningful role in constructions that lack these relatively strict patterns of subject assignment. Impersonal expressions, for one, do not themselves contain any information as to the identity of the subject of their complement, so there can be a greater pragmatic need to express it overtly, as seen in sentence (320).
54
România Literar 19, 17.5.2000, Aniversare Geo Dumitrescu; Constanµa Buzea: Scrisoarea r t cit
The in nitive in Romanian
(320) E
319
b rb µie ³i curagiu a înfrunta cineva valurile
be.3sg manliness and courage to face.inf someone waves.def
minciunoase.55
lumii
world.def.gen full.of.lies
It is manliness and courage (for one) to face the waves of a world full of lies. Even more commonly, overt subjects specify the subject in complements of nouns, as the noun itself does not normally assign a subject to its complement in a predictable way (321). (321) Obµine de.la rege agrementul
de a forma el
un
obtains from king commission.def of to form.inf he.nom a
guvern...56 government
He receives the king's commission that he should form a government. Prepositional in nitives can also be accompanied by overt subjects, either for emphasis, or because it is not coreferent with the main clause subject (322-323). (322) V
rug m nu mai parcaµi decât la domiciliul dv. pentru a
you.acc ask.1pl not more park.2pol except at home
putea
³i noi
s
your for
to
parc m la domiciliul nostru.57
be.able.inf also we.nom that park.1pl at home
our
Would you please not park anywhere except outside your house, so that we, too, can park outside our house.
55 56 57
România Literar 34, 25.8.2000, Lecturi la zi; Ioana Pavulescu: Riscurile meseriei România Literar 50, 20.12.2000, Cronica ediµiilor; Z. Ornea: Patronul `Universului' Note left behind my windscreen wiper in Bucharest, 1998.
320
Chapter 6
(323) Cu puµin timp înainte de a ajunge eu with little time before
în •ara
Sfânt ,
of to arrive.inf I.nom in land.def holy
Sebastian Costin facuse un infarct.58 prpn
prpn
made
a
heart.attack
Shortly before I arrived in the Holy Land, Sebastian Costin had a heart attack. However, the possibility of having overt subjects does not mean that they occur very frequently. On the contrary, only about one in 400 innitives has an overt subject. Of these, about half are merely emphatic, while the rest actually speci es a subject other than the one that would otherwise have been assigned. Even though the majority of those instances occur with prepositional in nitives, they remain a minute proportion of the total number of prepositional in nitives. As discussed in Section 3.2, a similar tendency for the availability of overt subjects not to be exploited for the purpose of introducing subjects other than those predicted by the normal assignment patterns is also apparent in Spanish. Consequently, by far the most common pattern of subject assignment in prepositional in nitive clauses is coreferentiality. As in Spanish and Portuguese, this link plays an important part in the readiness of certain prepositions to take in nitival clauses. 6.4.2 Romanian in comparison with Spanish and Portuguese The most obvious and striking observation when comparing the Romanian data with those of Spanish and Portuguese is that, on the whole, the prepositional in nitives corresponding to those with the earliest and strongest development in Spanish and Portuguese are also the ones found most frequently in Romanian. As the timing of their rst appearance rules out cultural borrowing or calquing from other Romance languages, it must be assumed that we are dealing with an independent development. A further argument in support of independent development is the fact that Romanian
58
România Literar 10, 18.3.1998; Gheorghe Schwartz: Amurgul unei literaturi
The in nitive in Romanian
321
prepositional in nitives obligatorily involve the particle a interpolated between preposition and in nitive, which gives the construction an appearance rather unlike its French or Italian counterparts. Indeed, in the 19th century, when Romanian was subjected to strong cultural in uence from France, authors did, occasionally, use calqued in nitive constructions without the a, as seen in Sentence (324) taken from Ion Ghica's Scrisori câtre V. Alecsandri, written in 1887. This lends further support to the observation that the normal in nitival construction with a is not a syntactic calque or loan. (324) Porunce³te de-i order.3sg
aduce îndat
un caftan ³i-l
de-him.dat bring.inf immediately a
caftan and-it
îmbrac . put.on.3sg
He gives order to bring him a caftan immediately and puts it on. Having ruled out borrowing as the source of the Romanian prepositional in nitives, the fact that it emerged at a later stage than in the other Romance languages means that we must be dealing with an independent development. A comparison of the similarities in its distribution is thus likely to allow conclusions about cross-linguistic factors facilitating such a parallel development. Synchronic comparison Table 6.9 shows the present-day proportion of prepositional in nitives to their nite counterparts in Romanian, Spanish and Portuguese. On the whole, Table 6.9 shows that the proportion of prepositional innitives is generally lower in Romanian than it is in Spanish and Portuguese. This is, however, not surprising, as the in nitive is never obligatory in the Romanian constructions. It is, therefore, perhaps more sur-
59
I have found a single example of a concessive in nitival construction in Romanian: ... cu ciuda de a nu participat în acest an la Winter Cup `despite not having taken part in the Winter Cup this year', http://www.jurnalul.ro/arhiva/ ian2001/27ian/sport.html
322 Table 6.9
Chapter 6 Proportion of nite and in nitival prepositional clauses in Romanian, Spanish and Portuguese
Preposition/Conjunction
% inf.
% finite
final
Romanian Spanish Portuguese
53.39 % 84.44 % 89.95 %
45.56 % 15.56 % 10.05 %
‘before’
Rom. înainte de a / înainte sa˘ Span. antes de / antes (de) que Port. antes de / antes (de) que
81.33 % 59.50 % 90.77 %
18.44 % 40.5 % 9.23 %
‘without’
˘ a˘ a / far ˘ a˘ sa˘ Rom. far Span. sin / sin que Port. sem / sem que
39.45 % 84.79 % 79.74 %
60.55 % 15.21 % 20.26 %
‘until’
˘ Rom. pâna˘ (la) a / pâna˘ ce/sa/când Span. hasta / hasta que Port. até / até que
18.85 % 28.99 % 48.62 %
81.15 % 71.01 % 51.38 %
‘instead of’
Rom. în loc de a / în loc (ca) sa˘ Span. en vez,lugar de / en vez,lugar (de) que Port. em vez,lugar de / e vez,lugar (de) que
6.51 % 99.39 % 99.89 %
93.49 % 0.61 % 0.11 %
‘after’
Rom. (dupa˘ a), în urma de a / dupa˘ ce Span. después/luego de, tras / después de que Port. depois(de)/segundo/após / depois (de) que
0.05 % 57.42 % 97.74 %
99.5 % 42.58 % 2.26 %
‘concessive’
Rom. (în/cu ciuda de a)/de¸si, cu toate ca˘ Span. a pesar de, pese a / aunque, a pesar de que Port. apesar de / apesar (de), ainda, embora que
0.00 %59 4.61 % 16.92 %
100.00 % 95.39 % 83.08 %
prising that in the nal domain, as well as the temporal domain with anterior main clause (`before'), the prepositional construction dominates in all three languages. Similarly, abessive `without' often takes the innitive, though not in the majority of cases in Romanian. The in nitival construction with `until' is less dominant in all three languages, but it nevertheless accounts for a signi cant proportion of this type of temporal clause. At the other end of the scale, similarities can also be observed: in the concessive domain, the in nitive is far less common in all three languages. Signi cant differences are found in the usage of `instead of' and `after', which commonly take the in nitive in Spanish and Portuguese,
The in nitive in Romanian
323
but not in Romanian. A comparative look at the historical development of these prepositional in nitives provides some clues why this may be the case. Diachronic comparison Parallels between Romanian, Spanish and Portuguese are evident in the development of the more frequently occurring prepositional in nitives: de a+in nitive appears early on in Romanian (Fig. 6.1), and de+in nitive is also present since the earliest Spanish and Portuguese texts (Fig. 4.12). Similarly, f r a+in nitive and înainte de a+in nitive appear early on in Romanian (cf. Figs 6.2 and 6.7), and so do the corresponding constructions in Spanish and Portuguese (cf. Figs 5.7 and 5.2). In the nal domain, the situation is more complex, as the nal preposition itself changes in both Ibero-Romance and Romanian. In Fig. 4.9 it has been shown how para takes over the nal domain from por in Spanish in the 16th and 17th century; this is subsequently joined by Spanish a n de+in nitive (cf. Fig. 4.21) and Portuguese a m de+innitive (cf. Fig. 5.11), the latter having experienced a strong increase since the late 19th century. From a semantic point of view, however, it is important to note that nal prepositional in nitives have been present, and comparatively common, from the very earliest texts. This is also true for Romanian, which has also experienced a replacement of the nal preposition involved in the construction. In Section (6.3.1), it has been illustrated how a still frequently functions as a preposition introducing nal in nitival clauses in the 16th century, and it is argued that its continuing grammaticalization and semantic bleaching leads to the rise of spre a+in nitive and pentru a+in nitive (cf. Figs 6.3 and 6.4). The nal domain can therefore be said to be particularly closely linked to the prepositional in nitive, as this construction is not only the rst, or among the rst, to emerge, but also remains the construction in which in nitival clauses are most dominant compared to nite ones. At rst sight, there appears to be less similarity regarding the substitutional domain (`instead of'). In Spanish and Portuguese, the in nitive is used in these clauses in almost every case (Table 6.9), while in Romanian there is a clear dominance of the nite structure in this do-
324
Chapter 6
main. But the diachronic development in Spanish, as charted in Fig. 4.20 comparing the differential development of en lugar de+in nitive and en vez de+in nitive and con ated into a single graph in Fig. 6.11, shows how the synchronic discrepancy between Romanian and Spanish might be explained from a diachronic perspective.
per 100000 words
30 25 20 15 10 5 0 1100
1150
1200
1250
1300
1350
1400
Figure 6.11
1450
1500
1550 1600 year
1650
1700
1750
1800
1850
1900
1950
2000
Spanish en lugar/vez de+in nitive
Fig. 6.11 shows that it took `two attempts' for the Spanish construction to gain today's degree of acceptance and usage, having occurred in the 16th and the rst half of the 17th century to some extent, but only re-entering the language to reach its current usage frequency relatively recently, around the beginning of the 19th century. The `two attempts' show that in this domain there is little predictability regarding the evolutionary trajectory, and that no clear or natural tendency towards the dominance of either the nite or the in nitival construction can be identi ed in the `substitutional' domain. Since prepositional in nitives have emerged much more recently in Romanian, it is possible that în loc de a+in nitive is yet to gain the popularity that Spanish en vez/lugar de+in nitive has acquired over the last 200 years or so. Concessive prepositional in nitives have also only appeared comparatively recently in Spanish (cf. Section 4.2.5), and their overall share of concessive clauses (Fig. 4.36) remains very small, so here, too, a parallel between Romanian and Spanish can be observed, and it may be speculated that, in time, concessive prepositional in nitives may begin to enter Romanian, too. For `after', on the other hand, no such explanation seems likely. In Figs 4.27 and 4.28, it was shown that Spananish antes de+in nitive and después de+in nitive emerged at roughly the same time, and sub-
The in nitive in Romanian
325
sequently continued to develop along similar lines. The asymmetry in Romanian between the common use of înainte de a+in nitive for anteriority of the main clause on the one hand, and the absence of dup or în urma with the in nitive for posteriority of the main clause on the other, is more reminiscent of the asymmetry between frequently used hasta+in nitive and virtually non-existent desde+in nitive in Spanish, and also the corresponding pairs of prepositions até/*desde in Portuguese and pân (la) a/*de a60 in Romanian; the notion `until' is commonly expressed by prepositional in nitives, whilst `since' is not. Coreferentiality and pragmatic likelihood The degree of similarity between Romanian, Spanish, and Portuguese in terms of the order in which the different prepositional in nitives enter the languages, and in terms of their present-day distribution, suggests that the parallels are unlikely to be coincidental. If contact between the respective languages cannot be held responsible, then the data presented for Romanian in this chapter provides strong support for the pragmatically-based hypothesis proposed for Spanish and Portuguese in Sections 4.4.2 and 5.1.5. Throughout this chapter, it has been shown that the patterns by which the subject of in nitives are assigned in Romanian resemble those found in Spanish and Portuguese: whilst some particular main verbs trigger object control, the default pattern with most verbs and prepositions is subject coreference between main and dependent clause. In all three languages, in nitives are capable of taking their own, overt subject, by which the default pattern of subject assignment may be broken. However, the number of cases in which this happens is comparatively small, and even in nitives with overt subjects regularly conform to the default pattern, the overt subject merely having an emphatic or contrastive function.
60
De a+in nitive is, of course, very common in constructions such as noun complements (cf. 6.2.2), but it is never used to mean `since'.
326
Chapter 6
% coref. of all final / concessive clauses
This close association of prepositional in nitives with coreferentiality plays a central role in their rise and frequency, as the main verb and the prepositional clause are more likely to share the same subject with some prepositions than with others in actual language use. Perhaps the two most extreme examples are the nal construction on the one end of the scale, and the concessive one on the other: in all three languages examined here, nal prepositional clauses are in their majority coreferential, whilst concessive ones share the main clause subject much less frequently. This was illustrated for Spanish in Fig. 4.43 (reproduced in Fig. 6.12 for convenience) and for Portuguese in Fig. 5.15 (6.13); Fig. 6.14 shows that in Romanian, too, nal clauses are usually coreferential, whilst concessive clauses do not, in their majority, share the subject of their main clause. It should, in particular, be noted that the ratio of coreferential to non-coreferential clauses, nal as well as concessive, remains remarkably stable over time in all three languages, with around 40% of all concessive clauses and around 80% of all nite clauses being coreferential. 100% 80% 60% 40% 20%
Concessive coref. (%) Final coref. (%)
0% 1100
1200
1300
1400
1500
1600
1700
1800
1900
2000
year
Figure 6.12
Percentage of coreferential nal and concessive dependent clauses in Spanish
percentage
The in nitive in Romanian
100 90 80 70 60 50 40 30 20 10 0 1350
1400
1450
1500
1550
327
1600
1650
1700
1750
1800
1850
1900
1950
2000
year
% coref. of all final / concessive clauses
Figure 6.13
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1800
Figure 6.14
Percentage of coreferential nal and concessive dependent clauses in Portuguese (dotted = nal, solid = concessive)
Concessive coref. (%) Final coref. (%) 1820
1840
1860
1880
year
1900
1920
1940
1960
1980
2000
Percentage of coreferential nal and concessive dependent clauses in Romanian
Similarly consistent rates of coreferential usage appear to apply for several of the clause types listed in Table 6.9 above, and more detailed classi cation according to this parameter may provide further insights into the relation between usage frequency and a preference or resistance to in nitival structures.
6.5
Conclusion
To conclude this chapter, it is at this point suf cient to observe that in Romanian, as in Spanish and Portuguese, there is a clear connection between the types of dependent clause favouring use of the in nitive
328
Chapter 6
on the one hand, and the likelihood of the respective construction being coreferential on the other. That is not to say that coreferentiality automatically triggers the use of the in nitive, as seen in the case of concessive constructions in all three languages, and all prepositional in nitives in Romanian. Nor does it rule out the use of the in nitive in non-coreferential clauses, if the identity of the subject can be determined by other strategies, such as overt person marking within the dependent clause or unambiguous contextual cues. The link between coreferentiality and prepositional in nitives manifests itself most clearly at a different level: the greater the number of coreferential tokens of a semantic type in actual language usage, the more likely it is for a prepositional construction to emerge at an early stage and become widely used.
Chapter 7 Relevance and reanalysis: prepositional complementizers
7.1
The emergence of prepositional complementizers
As discussed in Section 1.3.3, no clear dividing line can be drawn between what is traditionally referred to as peripheral clause or adjunct on the one hand, and clausal object or complement on the other, but it is nevertheless possible to identify prototypical instances of both. Whilst the previous chapters have largely concentrated on tracing the history of constructions that are relatively clearly classi able as one or the other, this section will take a closer look at the grey area between the two prototypes. In particular, the usage of two prepositions, a and de, which in the modern language frequently function as complementizers of prototypical in nitival complements, will be examined diachronically in order to explain how they acquired this purely syntactic function in certain constructions, shedding their semantic content. The account presented in this chapter is based on the assumption that individual `prepositional in nitives' with a speci c preposition can be subject to reanalysis processes of the kind described in Section 1.3.2, taking on the status of an independent construction (cf. Section 1.3.4) in which semantic and functional status is reassigned among its constituent elements. 7.1.1 The origin of prepositional in nitives In Chapter 2 it was shown that the overall typology of Latin was different from that of Romance: Latin used in ectional morphology as its primary strategy to express the logical and semantic relationships between the constituents of a sentence, but large-scale reduction or loss of this morphology rendered this system non-functional. The Romance lan-
330
Chapter 7
guages developed alternative strategies to clarify sentence-internal constituent relations, among which the use of prepositions gures prominently. Where Latin used a wide range of speci c `nominal' verb forms whose morphological shape gave an indication of the dependent structure's logical relation to the main verb, Romance expresses such relations by exploiting the originally purely physical sense of prepositions, widening their range of usage to cover semantically similar, but more abstract notions. This is already well established in the earliest surviving texts, as shown in the following Spanishexamples from the Primera Crónica General. Abstract use of de