Modality and Mood in Romance: Modal interpretation, mood selection, and mood alternation 9783110234336, 9783110234343

Table of contents :
Frontmatter
Table of contents
Modality and Mood in Romance: An introduction
Modal readings of sentence connectives in German and Portuguese
‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison
Modal non-finite relatives in Romance
Epistemic modality and evidentiality and their definition on a deictic basis
Where mood, modality and illocution meet: the morphosyntax of Romance conjectures
Modality, context change potential and mood selection in European Portuguese
On the (un)stability of mood distribution in Romance
Grammaticalization and language comparison in the Romance mood system
Principles of mood change in evaluative contexts: the case of French
Backmatter

Citation preview

Linguistische Arbeiten

533

Herausgegeben von Klaus von Heusinger, Gereon Mller, Ingo Plag, Beatrice Primus, Elisabeth Stark und Richard Wiese

Modality and Mood in Romance Modal interpretation, mood selection, and mood alternation Edited by Martin G. Becker and Eva-Maria Remberger

De Gruyter

ISBN 978-3-11-023433-6 e-ISBN 978-3-11-023434-3 ISSN 0344-6727 Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet ber http://dnb.d-nb.de abrufbar.  2010 Walter de Gruyter GmbH & Co. KG, Berlin/New York Druck: Hubert & Co, GmbH & Co. KG, Gçttingen

¥ Gedruckt auf surefreiem Papier Printed in Germany www.degruyter.com

Table of contents Martin G. Becker & Eva-Maria Remberger Modality and Mood in Romance: An introduction . . . . . . . . . . . . .

1

I. Modal interpretation Hardarik Blühdorn & Tinka Reichmann Modal readings of sentence connectives in German and Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

15

Sarah Dessì Schmid ‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

39

Ion Giurgea & Elena Soare Modal non-finite relatives in Romance . . . . . . . . . . . . . . . . . . . . . . . .

67

Gerda Haßler Epistemic modality and evidentiality and their determination on a deictic basis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

95

Mario Squartini Where mood, modality and illocution meet: the morphosyntax of Romance conjectures . . . . . . . . . . . . . . . . . . . . 109

II. Mood selection and mood alternation Rui Marques Modality, context change potential and mood selection in European Portuguese . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Josep Quer On the (un)stability of mood distribution in Romance . . . . . . . . . . 163

VI

Jan Lindschouw Grammaticalization and language comparison in the Romance mood system . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

181

Martin G. Becker Principles of mood change in evaluative contexts: the case of French . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209

Appendix Name index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

235

Subject index

239

..............................................

Martin G. Becker & Eva-Maria Remberger

Modality and Mood in Romance: An introduction

1. Mood and Modality – Modality and Mood This book is a selection of revised and further elaborated papers given at the session “Mood and Modality in Romance / Modus und Modalität in den romanischen Sprachen” at the 30th German Romanistentag 2007 in Vienna. It offers a deeper insight into the current linguistic debate on modality and mood and their categorial and semantic status, as well as their mutual relationship. Echoing the title of Palmer’s well-known monograph on the subject (cf. Palmer 1986, ²2001), but reversing its emphasis, our title takes modality as its starting point. By modality we mean the abstract semantico-functional category, which includes the huge variety of phenomena under discussion throughout our workshop, while mood as a grammatical form can be considered a morphological surface phenomenon. In fact, the latter is often used as a particular morphological device for coding modality. There can be no doubt that modality is one of the core notions of contemporary linguistic theory, given that it has provided some basic new insights into the semantic contribution of verbal categories in different languages. In this context the seminal works of Giorgi & Pianesi (1997) as well as Quer (1998) are particularly noteworthy, since they have paved the way for a new comprehensive account of the semantics of tense and mood in Romance (and other languages). As the current debate reveals, modality is undeniably an extremely productive notion, which has given rise to a series of further investigations and has contributed to an ongoing process of refinement and sophistication of the descriptive and analytical tools in the field of modal semantics (an overview of basic concepts is provided by von Fintel 2005). But, what is more, linguists share the notion of modality with philosophers, who offer valuable insights and well-defined concepts, and these in turn facilitate a more precise and formalised account of modal aspects in natural languages. In this context we can quote some recent philosphical work on epistemic modals and the problem of relativism (e. g. Egan 2005) as well as on the analysis of future contingents based on the notion of relative truth (MacFarlane 2003). The most recent studies on modal semantics lead us to the crossroads between temporality and modality. Condoravdi’s contributions to the temporal interpretation of modals (Condoravdi 2002, Kaufmann/Condoravdi/Harizanov 2006) have been particularly insightful, as they provide an exemplary analysis

2

Martin G. Becker & Eva-Maria Remberger

of the interplay between tense and modality, enriching our descriptive inventory with some appealing concepts as well as a coherent formalisation apparatus. Condoravdi’s approach has been a source of inspiration for researchers in the domain of Romance linguistics: Laca (2006, 2008a and b) and Demirdache & Uribe-Etxebarria (2006), for example, opened up an interesting new line of research in their recent works by focussing on the interplay between modals and the verbal categories of tense and aspect in French and Spanish. Another recent line of research attempted to explore the intricate field of epistemicity, inquiring in particular into the relationship to and interaction with the notion of evidentiality (De Haan 1999, Cornillie 2007, Abraham & Leiss 2009). Last, but not least, Portner (2009) has published a first synthesis of the different modal-semantic approaches which have been formulated over the last few decades (beginning with Kratzer’s inspiring contributions, cf. Kratzer 1977 et seq.). This ‘anglosaxon’ tradition of research in the domain of modality has not remained uncontested. Gosselin (2010) offers an alternative comprehensive view of modality and proposes a multidimensional model of analysis. This brief sketch of some recent activities in the field of modality research shows the relevance of the subjects presented under the binomial title “modality and mood” in this collection of papers: Modality and mood as the central referential concepts involve several thematic domains and a plurality of aspects, which are discussed in great detail here on the basis of Romance data. Both domains and their particular aspects will be mirrored in the structure of this volume. The first part of the book, Modal interpretation, is mainly dedicated to modality and discusses the syntax and semantics, and in part also the pragmatics, of various modal expressions in Romance. The arguments presented in the papers concern several types of modality expressed by various linguistic means, such as certain verbal categories, particular syntactic constructions or specific lexical and functional elements. The types of modality and the criteria for their definition, delimitation and differentiation, in particular the relationship between epistemicity and evidentiality, will also have a role to play. The second part of this publication, Mood selection and mood alternation, focusses on different aspects of mood. On the one hand, mood selection and mood distribution are analysed in a synchronic perspective: mood can be selected by certain types of elements or in particular contexts, e. g. verb classes (epistemic, factive, volitional etc.), conjunctions, operators, polarity items etc. Furthermore, mood and other grammatical phenomena (word order, subject obviation, control etc.) clearly interact at several levels. On the other hand, comparative and diachronic explorations of mood focus on mood alternation and mood change: the discussions cover the distribution of mood in a com-

Modality and Mood in Romance: An introduction

3

parative view, mood alternation and variation in interpretation (e. g. speakers’ attitudes or models, the opposition between realis/irrealis) as well as the the diachrony of mood systems together with principles and mechanisms of mood change in the Romance languages (grammaticalisation, convergence and divergence in the development of mood systems). All these topics are explored on the basis of mainly Romance data. The Romance languages analysed in this volume are Portuguese, Italian, Spanish, Romanian, French, and Catalan, i. e. the major Romance languages. All contributions also approach their subject under a theoretical point of view. In what follows, the structure of this volume will be further exemplified by characterising each of the two parts and the contributions that they contain in more detail.

2. Modal interpretation This section comprises five contributions that develop different perspectives on mood. Nevertheless, they form two thematic groups: Blühdorn & Reichmann, Dessì Schmid and Giurgea & Soare discuss certain functional elements and constructions, which can be used to achieve a modal interpretation. Sentence connectives, the verbal category of the imperfect and so-called “modal non-finite relatives” (expressed in Romanian by the ‘supine’) are thoroughly discussed and exemplified by a wide range of examples from Romance (as well as German and English). The articles by Haßler and Squartini instead concentrate more on the semantics and pragmatics of modal interpretation, such as the demarcation line between epistemicity and evidentiality and several subdomains of both. Whereas Squartini offers a PanRomance view, Haßler concentrates on specifically Spanish data. Squartini’s contribution also serves as a linking article between part one and part two of this volume, since he discusses interface phenomena between mood and modality. In the next paragraphs, each contribution on modal interpretation in Romance will be introduced and characterised by a short description. Hardarik Blühdorn and Tinka Reichmann investigate what they call sentence connectives, i. e. elements usually characterised as belonging to the complementiser system. The fact that these elements can establish connections between sentences on different levels makes them extremely interesting for an interpretation within the field of modality. While connectives operating on a temporal (or circumstantial, in the sense of Kratzer 1991) level do not concern anything beyond the state of affairs expressed by the sentence, connectives operating at the epistemic and deontic-illocutionary level do indeed encode

4

Martin G. Becker & Eva-Maria Remberger

modality (Speech Act Modality in terms of Sweetser 1990). With the help of examples from German and Portuguese, the modal use of sentence connectives is illustrated and their description in Portuguese grammar books is discussed. Several tests which might help to eliminate context-dependent ambiguities between the afore-mentioned levels are presented (these tests concern: paraphrases, scope of negation and modal adverbs, the factor of tense, and cleft constructions). The interesting conclusion of the paper is that sentence connectives work in a similar way in German and Portuguese and that obvious differences, e. g. in the use of porque/weil (for the epistemic use of the latter, cf. Keller 1995), must either be attributed to lexical idiosyncrasies or to language-specific syntactic properties, as for example the extensive use of the (personal and impersonal) infinitive in Portuguese. The article by Sarah Dessì Schmid discusses the various interpretive possibilities offered by the imperfect in Romance, particularly in Spanish and Italian. This is a frequently discussed topic in the interpretation of verbal categories (cf. Bertinetto 1986, 1997, Bazzanella 1990). Besides the canonical temporal (‘before utterance time’) and aspectual (‘imperfective’) reading of the imperfect, the paper elaborates on the different modal values that can be associated with the imperfect in Spanish and Italian. The author’s analysis is concerned with the question of whether it is the temporal or the aspectual semantic feature of the imperfect which is mainly responsible for the movement towards modal interpretations. The paper also sheds light on the distribution, and differences in distribution, of the modal readings in Italian and Spanish respectively. In response to the first question, the author’s conclusion is that the construction of a modal implicature (by a metonymical link, cf. Detges 2001) of the imperfect depends more on its aspectual meaning than on its temporal meaning. As far as the second issue is concerned, the author shows that the modal use of the imperfect appears very early in Spanish and Italian documents, although to a lesser extent than in the modern varieties, and that there are in fact some differences in distribution with respect to the older as well as to the modern stages of the two languages. Ion Giurgea and Elena Soare analyse a construction they call “modal nonfinite relatives” in Romanian, French, and Italian (with some reference to English, German and Latin). This construction, which is seen as parallel to other non-finite relatives built on participles, has a special morphological form in Romanian, consisting of the element de + supine, a non-finite form distinct from the infinitive. French and Italian use the element à (French) or da (Italian) followed by the infinitive. After a brief presentation of the modal interpretation that these constructions can have (deontic necessity and teleological possibility), the article discusses in detail the syntactic analyses relevant to this construction. The theoretical framework for the analysis they

Modality and Mood in Romance: An introduction

5

propose is recent minimalist theory (cf. Chomsky 1995 et seq.), which offers an elaborate means of treating the topic from several syntactic perspectives. The construction under discussion is interpreted as a reduced relative (following Bhatt 1990, 2002) and as such the authors label it as a participial, notwithstanding its morphological form (introductory element + infinitive or supine). The syntactic structure of these reduced relatives (or participials) is then compared to similar structures in English, from which they differ, and with so-called tough-constructions (cf. also Kayne 1995 for French), with which they seem to have more properties in common. The final conclusions are that in Romanian, French and Italian the modal participials discussed in the article are all passive participials introduced by a modal head (labelled Mood, a functional head proposed by several authors, cf. Rivero 1987, Terzi 1993, Avram 1999), but that Romanian differs in that the modal head de subcategorises a vP complement whereas Italian da and French à select a TP. The contribution by Gerda Haßler offers some new insights into the question of whether epistemic modality and evidentiality belong to the same field or whether a strict boundary must instead be drawn between them (cf. Chafe & Nichols 1986, Cornillie 2007, De Haan 1999, Dendale & Tasmowski 2001, Plungian 2001, Squartini 2004). The author concludes that there are indeed demarcation problems between evidentiality and epistemic modality, at least as far as Romance (here mainly Spanish) is concerned. She starts with the claim that evidentiality is not only a grammatical category in languages like Tuyuca, which have grammatical evidential markers, but also in the European languages, since they do not rely solely on lexical elements to make reference to an information source, but have other grammatical means at their disposal (this is a view contrary to Aikhenvald 2004). The main findings of this article are that evidentiality is primarily concerned with perspectivisation and deixis: the latter classification of deictic categories following the proposal by Frawley (1992) is necessary to establish a relationship between the speaker, the hearer and the speaker’s source of knowledge (and the common knowledge of both). So whereas evidentiality and epistemic modality interact on the one hand, only evidential marking is a deictic property concerned with the pragmatic stance of the speaker, while epistemic modality contributes to the epistemic stance of the speaker without a need for extra-linguistic reference. Mario Squartini’s article aims at bridging the gap between the relationship of modality and mood by providing evidence for a more fine-grained division between different subtypes of epistemic modality. Starting from the treatment of modality by Lyons (1977) and Palmer (1986) and the problem of the Interrogative and the Dubitative (cf. also Hengeveld 2004), this paper shows that the domain of epistemicity can be divided into at least three subdomains,

6

Martin G. Becker & Eva-Maria Remberger

i. e. the dubitative, the inferential (also characterised as “anti-dubitative”) and the conjectural mood. He analyses several examples from French, Italian, and Catalan (and Spanish), where the conditional, the subjunctive, the future and the modal verb devoir/dovere/deure are used, but with different functions as far as the epistemic subdomains identified above are concerned. In at least some Romance languages these subdomains can be morphologically marked by specialised modals or specialised moods, e. g. by the conditional for the dubitative in French questions. Squartini also recognises that certain epistemic adverbs can be restricted to either the conjectural or the dubitative subdomain: French probablement, for example, is incompatible with the dubitative subdomain (again in questions), whereas peut-être is perfectly acceptable either in conjectural or dubitative contexts. Contrary to that, Spanish aparentemente, requiring an inferential context, cannot combine with the subjunctive, which seems to mark conjectural mood. Squartini comes to the conclusion that epistemicity can be interpreted as a continuum between the dubitative [+dubitative] and inferential mood [-dubitative] with an intermediate domain of the conjectural mood [+conjectural], and that the Romance languages under discussion offer different distributions for the formfunction correlations regarding the epistemic-evidential domain.

3. Mood selection and mood alternation This section comprises four contributions which focus on mood selection and mood alternation from both a synchronic and a diachronic perspective. Their common aim is to argue that mood alternation is not random, identifying the relevant principles that account for mood selection in the different Romance languages. Though different aspects are adduced as motivating factors of mood selection – such as, for example, intensionality and polarity (Stowell 1993) as well as (relativised) veridicality (Giannakidou 1999) – all contributions give particular emphasis to the crucial interaction between the morphological category of mood and the notional category of modality. Depending on the theory that they adopt, the authors elaborate the basic insight concerning the relationship between mood and modality within the framework of Heim’s CCP(=context-change-potential)-semantics (Heim 1992), Quer’s model theoretic approach (Quer 1998) or the combination of the theory of assertion (see e. g. Haverkate 2002, Korzen 2003) and a theory of markedness (Andersen 2001). The articles do not only reflect the state of the art as regards the different ways of dealing with the relationship between mood and modality, but also develop an interesting panorama as to the variability – the

Modality and Mood in Romance: An introduction

7

convergences and divergences – of mood selection from a diachronic, comparative and variational point of view. In the following paragraphs the contributions of the second part will be presented in more detail. Rui Marques’ contribution aims at casting some new light on the semantic and pragmatic principles of mood selection in European Portuguese. This main concern is all the more important as the most prominent approaches – the assertion/non-assertion hypothesis defended by Hooper (1975) as well as the veridicality hypothesis put forward by Palmer (1986) and Bell (1990) – fall short of accounting for all types of mood alternation that can be observed in Romance languages, and especially in Portuguese. In his contribution Marques identifies three basic factors which help to explain the selection of mood in Portuguese: the notion of relative veridicality in the sense of Giannakidou (1999) (the truth of a clause according to the epistemic model of the matrix subject), a given type of modality (e. g. doxastic, bouletic, evaluative modality) and the context change potential of a clause (i. e. the set of possible worlds available at each point of the conversation). Given these three parameters, it can be shown that the indicative occurs exclusively in epistemic veridical environments and has the potential to remove all but those possible worlds from the context set where the complement clause is verified. By contrast, the subjunctive does not occur in epistemic veridical environments and may provide a change in the ordering of the context set, bringing into consideration possible worlds that were less accessible prior to the utterance. Finally, a more detailed scrutiny of mood oppositions in negated causative and temporal clauses shows that the infinitive – in contrast to the indicative and the subjunctive – leaves the context set unchanged, due to its inability to update the context with new information. Jan Lindschouw’s article deals with the evolution of mood distribution in concessive clauses in French and Spanish. Focussing on this particular domain, Lindschouw tries to elucidate two related questions: does the system of formal and functional mood distribution in Modern Spanish bear a resemblance to that of Renaissance French? And can the French mood system be considered more innovative than that of Spanish? In his study the author embraces the theory of assertion defended by Hooper (1975), Haverkate (2002) and Korzen (2003) (among others) according to which the indicative is the assertive mood and the subjunctive the non-assertive mood (including the values “irrealis” as well as “presupposed”). The corpus-based study provides some interesting insights into convergent and divergent developments in the domain of mood: the French 16th century system allowed for a certain degree of free alternation, with the subjunctive being the dominant and unmarked form of the system. In Modern French, the subjunctive has undergone a process of ‘grammaticalization’ (in the terminology of the author), since its use has become

8

Martin G. Becker & Eva-Maria Remberger

obligatory and specialised. Furthermore its irrealis value has gradually disappeared over the course of time and its functional content is now restricted to presuppositional contexts. By contrast, in Modern Spanish the distribution of the two modal forms is just the opposite: even though aunque (‘although’) displays a comparable pattern of free mood alternation, the indicative turns out to be the dominant and unmarked category of the Spanish system. Another difference, on the semantic-pragmatic level, should also be taken into account: in Renaissance French the subjunctive favours the presupposition value whereas in Modern Spanish no such preferences can be detected. From his careful study Lindschouw draws two conclusions: Spanish can be characterised as a very conservative Romance language and French as one of the most innovative. However, we should be cautious when drawing parallels between different stages of languages and making undue predictions on these grounds about their future developments. What really matters for the evolution of languages is primarily their internal structure – the relation between the form and the content of linguistic signs and their markedness relation. Josep Quer addresses the problem of superficial variability in mood choice which emerges intralinguistically (as diachronic or diatopic variation) and crosslinguistically among Romance languages. Relying mainly on data from Catalan and Spanish, the author shows that the alleged variation in mood selection is not a result of free optionality but always has an interpretive effect. These interpretive effects are particularly relevant in contexts determined by factive-emotive and epistemic predicates, verbs of speech and communication and by indirect interrogatives. The factors which may be relevant for mood variation can involve scopal effects with intensional operators like negation (in epistemic contexts), the opposition between episodicity and genericity (with factive-emotive predicates) and inferences as to the truth-conditional status of the complement clause (for instance, in indirect questions or with verbs like parecer). The relevance of these factors shows that mood is a meaningful category which overtly marks information about the models in which propositions are to be interpreted. Finally, Quer addresses the question of why mood seems to be fixed with some predicates, but not with others. To answer this question, he invokes economy conditions imposed by the language faculty. Predicates creating the core domains of the subjunctive (such as volitional, directive and implicative predicates) link the interpretation of the embedded clause to a shifted model, whereas epistemic and factive-emotive predicates can link the embedded proposition to more than one model. A related problem is dealt with in Martin Becker’s contribution. The article sketches the different stages of mood selection from Old to Modern French and tries to unveil the underlying principles which can explain the shift from a preference for the indicative to an almost generalised subjunctive selection

Modality and Mood in Romance: An introduction

9

pattern in the domain of evaluative predicates. It is argued that the particular conceptual-semantic structure inherent to evaluative predicates has been a decisive prerequisite for the reanalysis of the relevant principles of mood selection on the part of the speakers. Throughout the article four basic tendencies are outlined in order to explain the relevant changes in their correct chronology: First, speakers narrowed evaluative predicates with a deontic modal base to the domain of strong intensional verbs and their mood selection properties. Second, they strengthened the intensional potential of evaluative predicates in modal contexts. Third, they began to highlight the classificational potential of the matrix predicates at the expense of the truth-conditional status of their complements. Finally, the generalisation of the subjunctive was enforced in the second half of the 17th century by linguistic authorities who imposed the subjunctive for all evaluative contexts by prescriptive influence. The conventionalisation of mood selection in view of a particular class of predicates stripped mood of its motivation and converted it into a subcategorisation feature of the individual predicates. In the end, the subjunctive in evaluative contexts was confined to the role of being simply an element of grammatical well-formedness encoded in evaluative predicates.

4. Further remarks As the collection of the contributions offered in this volume shows, the Romance languages are particularly appropriate for the study of modality and mood and of their intertwined relationship. Of course, many solutions to specific problems proposed here require further investigation from an even broader comparative viewpoint, i. e. one which would also take into account data from languages outside the Romance family. We will leave this to future research and are pleased to be able to offer this volume to a readership of experts in Romance linguistics as well as of researchers in the field of mood and modality. Every edited volume is based on a lot of work, and not only on the part of the editors. Therefore, we first of all wish to thank Florian Scheib for his help in preparing the manuscript, maintaining contact with the authors and straightening out several inconsistencies which came to light during the editing process. Furthermore we express our thanks to Janina Reinhardt and Daniela Giller who helped with the final preparation of the volume. We would also like to thank the editors of Linguistische Arbeiten, especially our main contact there, Elisabeth Stark, for including this volume in the series. And of course, thanks must also go to our authors themselves who were so kind to offer us

10

Martin G. Becker & Eva-Maria Remberger

their contributions and all supported us with helpful comments on the way to the final publication of this volume.

References Abraham, Werner & Leiss, Elisabeth (2009): Modalität – Epistemik und Evidentialität bei Modalverb, Adverb, Modalpartikel und Modus. – Tübingen: Stauffenburg (Studien zur deutschen Grammatik 77). Aikhenvald, Alexandra Y. (2004): Evidentiality. – Oxford: Oxford University Press. Andersen, Henning (2001): “Markedness and the theory of linguistic change.” – In: id. (ed.): Actualization. Linguistic Change in Progress, 21–57. Amsterdam/ Philadelphia: Benjamins. Avram, Larisa (1999): Auxiliaries and the Structure of Language. – Bucharest: Editura UniversităĠii Bucureúti. Bazzanella, Carla (1990): “‘Modal’ uses of the Italian Indicativo Imperfetto in a pragmatic perspective.” – In: Journal of Pragmatics 14, 439–457. Bell, Anthony (1990): “El modo en español: consideración de algunas propuestas recientes.” – In: Ignacio Bosque (ed.): Indicativo y subjunctivo, 81–105. Madrid: Taurus. Bertinetto, Pier Marco (1986): Tempo, aspetto e azione nel verbo italiano. Il sistema dell’indicativo. – Firenze: Accademia della Crusca. – (1997): Il dominio tempo-aspettuale: demarcazioni, intersezioni, contrasti. – Torino: Rosenberg & Sellier. Bhatt, Rajesh (1999): Covert Modality in Non-Finite Contexts. – Diss., University of Pennsylvania. – (2002): “The Raising Analysis of Relative Clauses: Evidence from Adjectival Modification.” – In: Natural Language Semantics 10, 43–90. Bosque, Ignacio (ed.) (1990): Indicativo y subjunctivo. – Madrid: Taurus. Chafe, Wallace & Nichols, Johanna (eds.) (1986): Evidentiality: the linguistic coding of epistemology in language. – Norwood, N.J.: Ablex. Chomsky, Noam (1995): The Minimalist Program. – Cambridge, Mass.: MIT Press. – (2000): “Minimalist inquiries: the framework.” – In: Roger Martin, David Michaels & Juan Uriagereka (eds.): Step by step: essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. – (2001): “Derivation by phase.” – In: Michael Kenstowicz (ed.): Ken Hale: a life in language, 1–52. Cambridge, Mass.: MIT Press. – (2005): “On Phases.” – Ms., MIT. Condoravdi, Cleo (2002): “Temporal Interpretations of Modals. Modals for the Present and for the Past.” – In: David Beaver et al. (eds.): Standford Papers on Semantics, 59–87. Standford: CSLI Publications.

Modality and Mood in Romance: An introduction

11

Cornillie, Bert (2007): Evidentiality and Epistemic Modality in Spanish (Semi-)Auxiliaries. A Cognitive-Functional Approach. – Berlin/New York: de Gruyter. De Haan, Ferdinand (1999): “Evidentiality and Epistemic Modality: Setting Boundaries.” – In: Southwest Journal of Linguistics 18, 83–101. Demirdache, Hamida & Uribe-Etxebarria, Miriam (2006): “Morfosintaxis e interpretación temporal de los verbales modales.” Ms., Universidad de Nantes & Universidad del País Vasco. Dendale, Patrick & Tasmowski, Liliane (eds.) (2001): Evidentiality, 339–464. – Amsterdam: Elsevier (Journal of Pragmatics 33). Detges, Ulrich (2001): Grammatikalisierung. Eine kognitiv-pragmatische Theorie dargestellt am Beispiel romanischer und anderer Sprachen. – Professorial Diss., Universität Tübingen. Egan, Andy (2005): “Epistemic Modals, Relativism, and Assertion.” – In: Philosophical Studies 133 (1), 1–22. Farkas, Donka (1992): “On the semantics of subjunctive complements.” – In: Paul Hirschbühler & Konrad Koerner (eds.): Romance Languages and Modern Linguistic Theory, 71–104. Amsterdam: Benjamins. Fintel, Kai von (2005): “Modality and Language.” Article to appear in: The Encyclopedia of Philosophy – Second Edition edited by Donald M. Borchert (most recent version online at http://mit.edu/fintel/www.modality.pdf). Frawley, William (1992): Linguistic semantics. – Hillsdale: Erlbaum. Giannakidou, Anastasia (1999): “Affective dependencies.” – In: Linguistics and Philosophy 22 (4), 367–421. Giorgi, Alessandra & Pianesi, Fabio (1997): Tense and Aspect: From Semantics to Morphosyntax. – Oxford: Oxford University Press. Gosselin, Laurent (2010): Les modalités en français. La validation des représentations. – Amsterdam: Rodopi (coll. Etudes Chronos). Haverkate, Henk (2002): The Syntax, Semantics and Pragmatics of Spanish Mood. – Amsterdam/Philadelphia: Benjamins. Heim, Irene (1992): “Presuppositional Projection and the Semantics of Attitude Verbs.” – In: Journal of Semantics 9 (3), 183–221. Hengeveld, Kees (2004): “Illocution, Mood and Modality.” – In: Geerd BooƋ, Christian Lehmann, Joachim Mugdan & Stavros Skopeteas (eds.): Morphology: An International Handbook on Inflection and Word-Formation. Vol. 2, 1190–1202. Berlin: Mouton de Gruyter. Hooper, Joan B. (1975): “On assertive predicates.” – In: John P. Kimball (ed.): Syntax and Semantics. Vol. 4, 91–124. New York: Academic Press. Hopper, Paul & Traugott, Elisabeth (1993/22003): Grammaticalization. – Cambridge: Cambridge University Press. Kaufmann, Stefan, Condoravdi, Cleo & Harizanov, Valentina (2006): “Formal approaches to modality.” – In: William Frawley (ed.): The expression of modality, 71–106. Berlin: Mouton de Gruyter. Kayne, Richard S. (1975): French Syntax. The Transformational Cycle. – Cambridge, Mass.: MIT Press.

12

Martin G. Becker & Eva-Maria Remberger

Keller, Rudi (1995): “The epistemic weil.” – In: Dieter Stein & Susan Wright (eds.): Subjectivity and subjectivisation. Linguistic perspectives, 16–30. Cambridge: Cambridge University Press. Korzen, Hanne (2003): “Subjonctif, indicatif et assertion ou: Comment expliquer le mode dans les subordonnées complétives?” – In: Merete Birkelund, Gerhard Boysen & Poul Søren Kjærsgaard (eds.): Aspects de la modalité, 113–129. Tübingen: Niemeyer. Kratzer, Angelika (1977): “What ‘must’ and ‘can’ must and can mean.” – In: Linguistics and Philosophy 1 (1), 337–355. – (1978): Semantik der Rede. – Königstein: Scriptor. – (1981): “The notional category of modality.” – In: Hans-Jürgen Eikmeyer & Hannes Rieser (eds.): Words, Worlds, and Contexts, 38–74. Berlin: de Gruyter. – (1991): “Modality”. – In: Arnim von Stechow & Dieter Wunderlich (eds.): Semantics, 639–650. Berlin: de Gruyter. Laca, Brenda (2006): “Tiempo, aspecto y la interpretación de los verbos modales en español.” – In: Lingüística (ALFAL) 17, 9–43. – (2008a): “Temporalidad y modalidad.” – In: Miguel Casas Gómez & Ana Isabel Rodríguez-Piñeiro Alcalá (eds.): X. Jornadas de Lingüística, 109–136. Cádiz: Universidad de Cádiz. – (2008b): “On modal tenses and tensed modals.” – In: Cahiers Chronos. [halshs.archives-ouvertes.fr/docs/00/32/33/40/PDF/ChronosTEX.MT_TM.pdf] MacFarlane, John (2003): “Future contingents and relative truth.” – In: The Philosophical Quarterly 53 (212), 321–336. Palmer, F. R. (1986): Mood and Modality. – Cambridge: Cambridge University Press. – (²2001): Mood and Modality. – Cambridge: Cambridge University Press. Plungian, Vladimir A. (2001): “The place of evidentiality within the universal grammatical space.” – In: Journal of Pragmatics 33, 349–357. Portner, Paul (2009): Modality. – Oxford: Oxford University Press. Quer, Josep (1998): Mood at the Inferface. – The Hague: HAG. – (2006): “Subjunctives.” – In: Martin Everaert & Henk van Riemsdijk (eds.): The Blackwell Companion to Syntax. Vol. IV, 660–684. Oxford: Blackwell. Rivero, María Luisa (1987): “La teoría de las barreras y las completivas del rumano.” – In: Violeta Demonte & Marina Fernández Lagunilla (eds.): Sintaxis de las lenguas Románicas, 329–354. Madrid: Textos Universitarios. Squartini, Mario (2004): “Disentangling evidentiality and epistemic modality in Romance.” – In: Lingua 114, 873–895. Stowell, Tim (1993): “Syntax of Tense.” – Ms., UCLA. Sweetser, Eve (1990): From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure. – Cambridge: Cambridge University Press. Terzi, Arhonto (1993): Pro in Finite Clauses. A Study of the Inflectional Heads of the Balkan Languages. – Ph.D. Diss., City University of New York.

I.

Modal interpretation

Hardarik Blühdorn & Tinka Reichmann

Modal readings of sentence connectives in German and Portuguese

1. Introduction: Temporal and modal linking of sentences This paper1 deals with an aspect of the semantics of sentence connectives, exemplified by German and Portuguese. Following Pasch, Brauße, Breindl & Waßner (2003) and Blühdorn, Breindl & Waßner (2004), we use the term sentence connective for certain elements traditionally classified as conjunctions, adverbs and particles, which share the semantic function of encoding relations between sentences. The semantic relata of sentence connectives are the meanings of the connected sentences or clauses. Since the influential work by Sweetser (1990: 76ff.) it has been widely known that one and the same connective can encode relations of different kinds. As early as the end of the nineteenth century we find the distinction between three types of causal connections in the handbook of German grammar by Blatz (1896/1970: 708, 1128ff.): Der Grund kann 1. ein realer sein (ein Sachgrund, d.h. die Ursache einer Thatsache), z.B. Wallenstein hatte rasch ein großes Heer versammelt, denn die Zahl der unbeschäftigten Abenteurer war bedeutend, – oder 2. ein logischer (Erkenntnisgrund, d.h. die Begründung einer ausgesprochenen Ansicht), z.B. Das Gemeine muß man nicht rügen; denn das bleibt sich ewig gleich (Goethe), – oder 3. ein moralischer (Beweggrund zu einer Handlungsweise), z.B. Ich muß ihn hassen; denn er hat mich schwer beleidigt. (Blatz 1896/1970: 708)2

In the present paper we define factual reasons (Sachgründe) as states of affairs whose factuality results in the factuality of other states of affairs (see Blühdorn 1

2

We are grateful to the editors of this volume and to the reviewer for their valuable comments and suggestions, as well as to Daniel Gutzmann for an important bibliographical indication. The reason can be 1. a real one (a factual reason, i. e. the cause of a fact), for instance Wallenstein had rounded up a big army quickly, because the number of unengaged adventurers was considerable, – or 2. a logical one (an epistemic reason, i. e. the justification of an enunciated opinion), for example Common things are not to be criticised, because they will always remain the same (Goethe), – or 3. a moral reason (motivation for a behaviour), e. g., I must hate him, because he has greatly insulted me. (Translation H.B. and T.R.)

16

Hardarik Blühdorn & Tinka Reichmann

2003: 17ff.; 2005: 317; 2006: 265f.). In the following example, the factuality of watering the parsley is presented as the cause of the factuality of its luxuriance: (1)

A salsa ficou viçosa porque a reguei todos os dias.

(Mateus et al. 2003: 108)

(2)

Die Petersilie ist riesig geworden, weil ich sie jeden Tag gegossen habe. ‘The parsley has become huge because I watered it every single day.’

By epistemic reasons (Erkenntnisgründe) we mean propositions whose truth leads to the conclusion that other propositions are also true (Keller 1995: 22ff.). In the following example, it is asserted to be true that knowing the main cause of a certain illness can help to prevent it. The truth of this proposition is used as an argument to support the truth of the second assertion, namely that the question of the cause is what is most important: (3)

Por que ocorre a doença? Essa é a questão mais importante, porque, conhecendo a causa principal da doença, você poderá evitá-la.3

(4)

Warum kommt es zu der Krankheit? Das ist die wichtigste Frage, denn wenn Sie die Hauptursache der Krankheit kennen, können Sie sie vermeiden. ‘Why does the illness happen? This is the most important question, because if you know the main cause of the illness, you can avoid it.’

Blatz did not offer a more indepth discussion of the concept of motivation or “moral reason” (Beweggrund). Sweetser (1990: 56ff., 76ff.) shows that a distinction must be drawn between reported motivations that induce the actions of discourse figures and motivations that determine the communicative behaviour of the speaker and his/her partner in a linguistic interaction. In terms of grammar, the former are ordinary factual reasons, whereas the latter have a special semantic and grammatical status called speech-act modality by Sweetser (1990: 69ff.). In the following example, an action of the speaker (the request not to work too much) is justified with a “moral reason” (the aversion to work of the people from the state of Bahia): (5)

Não trabalhem demais, porque isso baiano não gosta.

(6)

Arbeitet nicht zu viel, denn das mögen die Leute aus Bahia nicht. ‘Don’t work too much because people from Bahia don’t like it.’

(Tom Jobim)

Not only in causal connections but in sentence connections in general it makes a difference whether the connected semantic relata are states of affairs, 3

The examples in this paper, whenever no other source is indicated, were taken from Internet sites in Portuguese found by the search engine Google. We thank Selma Meireles for valuable help and comments.

Modal readings of sentence connectives in German and Portuguese

17

propositions or speech acts. Depending on the category of the semantic relata, the resulting complex sentences have different grammatical properties (see Blühdorn 2005: 317ff.; 2006: 265ff.; 2008a). Before going into more detail, we will provide some additional explanations of the basic concepts to be used in this paper. States of affairs are temporal objects (see Blühdorn 2003: 18; 2004: 186). They have temporal extensions and are linked to each other by temporal relations. Sentence connections whose semantic relata are states of affairs often display a characteristic distribution of verbal tenses (consecutio temporum) (see Lehmann 1988: 205). Connectives which link states of affairs interact semantically with temporal expressions of all kinds. The relation encoded by the connective (similarity, situating, condition, cause; for more details of this typology see Blühdorn 2003: 20; 2008b) concerns the factuality or non-factuality of the connected states of affairs in the dimension of time. Extending traditional terminology, connections between states of affairs can therefore be called temporal connections. Propositions are logical or epistemic objects (see Lyons 1977: 443ff.). They are logically extended (i. e., they have areas and degrees of logical validity), and they are linked to each other by epistemic (logical) relations (implication, equivalence, complementarity etc.) (see Blühdorn 2008b). Sentence connections whose semantic relata are propositions often display a characteristic distribution of verbal moods (e. g., the requirement of subjunctive mood in a subordinate clause). Connectives which link propositions interact semantically with epistemic expressions of all kinds (epistemic sentence adverbs, modal particles, modal auxiliaries etc.). The relation encoded by the connective (similarity, situating, condition, cause; see Blühdorn 2003: 20; 2008b) concerns the truth or falsity of the connected propositions. Connections between propositions have therefore been called epistemic connections (see Sweetser 1990: 76ff.; Keller 1995). Speech acts are ethical, i. e., deontic objects. They take place in contexts of social norms of conduct and individual options and goals in social interaction. They are deontically extended (i. e., they have areas and degrees of permittedness, forbiddenness etc.), and they are linked to each other by deontic relations (compatibility, conflict, means, purpose etc.) (see Blühdorn 2008b). Sentence connections whose semantic relata are speech acts often display a characteristic distribution of sentence moods (declarative vs. interrogative vs. imperative). Connectives which link speech acts interact semantically with deontic expressions of all kinds (deontic sentence adverbs, modal particles, modal auxiliaries etc). The relation encoded by the connective (similarity, situating, condition, cause; see Blühdorn 2003: 20; 2008b) concerns the desirability or non-desirability of the connected speech acts. Connections

18

Hardarik Blühdorn & Tinka Reichmann

between speech acts can therefore not only be called illocutionary connections (see Sweetser 1990: 76ff.), but also deontic connections (see Truckenbrodt 2006: 263ff.; Blühdorn 2008b). The truth values of propositions and the values of desirability of speech acts negotiated in epistemic and deontic-illocutionary sentence connections are calculated basically in terms of necessity and possibility (see Kratzer 1991). They are grammatically expressed by means of sentence and verb moods, modal verbs and auxiliaries, modal adverbs and particles and other similar linguistic means. It is therefore reasonable to subsume epistemic and deonticillocutionary connections under the general label of modal connections (see Keller 1995: 23f.), as opposed to temporal (or “circumstantial”, see Kratzer 1991: 640) connections, in which the factuality or non-factuality of states of affairs in temporal contexts is discussed. The semantic structure of a declarative sentence can be described by the general formula in (7): (7)

(i, d, A (e, v, P (t, f, S)))

This formula can be paraphrased as in (8): (8)

A speaker performs a speech act A in an interactional context i. For this context, a value of desirability d is assigned to the speech act. The speech act consists in asserting a proposition P, to which a truth value v (for verity) is assigned for the epistemic context e. The proposition describes a state of affairs S, to which a value of factuality f is assigned for the temporal context t.

The formula can be applied to example (9) as shown in (10): (9)

Hm, Maria ist vielleicht gerade nicht zu Hause. ‘Hm, maybe Maria is not at home right now.’

(10) The speaker utters the sign string Maria ist vielleicht gerade nicht zu Hause. The initial interjection hm evaluates this utterance (A) in relation to the given interactional context (i), e. g. as an intentional (“desired” – d) reply to the question why Maria does not answer the phone. By making the utterance, the speaker asserts the proposition (P) ‘Maria is not at home right now’, but in the given epistemic context (e) s/he shows uncertainty with respect to the truth of this proposition. The epistemic sentence adverb vielleicht [maybe] indicates that it is a possibility considered by the speaker, but that s/he is not quite sure about it (v). The proposition describes the state of affairs (S) MARIA IS AT HOME and assigns to it the value “non-factual” (f) for the temporal context in which the utterance is made (right now – t).

Sentences can be linked with other sentences at the three levels of their semantic structure: at the temporal, the epistemic and the deontic-illocutionary

Modal readings of sentence connectives in German and Portuguese

19

level (see Lehmann 1988: 189ff.). The sentences they are linked to (the “anchor” sentences, as we will call them) provide the respective temporal, epistemic or deontic-illocutionary context (t, e or i) to which a value is assigned. As an illustration we consider three variants of example (9): (9a) Hm, Maria ist vielleicht gerade nicht zu Hause, weil sie einkaufen gegangen ist. ‘Hm, maybe Maria is not at home right now because she has gone shopping.’ (9b) Hm, Maria ist vielleicht gerade nicht zu Hause, denn um diese Zeit hat sie dienstags Sport. ‘Hm, maybe Maria is not at home right now, since she does sports at this time on Tuesdays.’ (9c) Hm, Maria ist vielleicht gerade nicht zu Hause, obgleich ich mir da nicht so sicher bin. ‘Hm, maybe Maria is not at home right now, although I am not quite sure about this.’

One possible interpretation of (9a) is that the speaker has observed that the state of affairs (S) MARIA IS AT HOME is not factual at the time of utterance. The cause of this might be the factuality of the state of affairs MARIA HAS GONE SHOPPING. In this reading, the subordinate “anchor” clause weil sie einkaufen gegangen ist provides the temporal context (t) for the “anchored” clause. This context is another state of affairs whose factuality may have caused the nonfactuality of Maria’s being at home. The most probable reading of (9b) is that the speaker knows that the proposition ‘Maria does sports at this time on Tuesdays’ is true and that this knowledge leads him/her to conclude that the proposition (P) ‘Maria is not at home right now’ may also be true. The “anchor” clause denn um diese Zeit hat sie dienstags Sport provides the epistemic context (e) for the conclusion encoded by the “anchored” clause. In (9c) the speaker utters the supposition that Maria is not at home at the time of utterance. S/he also refers to a circumstance which diminishes the desirability of this speech act (A), but without entirely blocking it: s/he is not sure about the truth of the supposition. This circumstance, encoded by the “anchor” clause obgleich ich mir da nicht so sicher bin, provides the deontic context (i) for the speech act of uttering the supposition, performed by means of the “anchored” clause. In all three examples the semantic relations between the “anchored” clause and its respective “anchor” clauses are encoded by connectives. In German, as well as in Portuguese, the same semantic relations can also be encoded by verb lexemes, by certain verb forms (participles, gerund forms), by nouns and by other linguistic means. Those other means will not be considered in this paper, which concentrates exclusively on connectives.

20

Hardarik Blühdorn & Tinka Reichmann

2. Modal use of connectives in grammars of Portuguese Sweetser’s distinction of different levels of sentence connection has, as yet, not been fully integrated into the grammars of German (some of the few authors who mention it are Zifonun et al. 1997: 2296f. and Duden 2005: 1084ff.), nor does it play a significant role in the grammars of Portuguese. Bechara (1999: 493) states, similarly to Blatz (1896/1970), that causal subordinate clauses can introduce “a causa, o motivo, a razão do pensamento” [the cause, the motive, the reason of the thought], but does not explain these concepts in more detail and does not provide separate examples. Peres & Mascarenhas (2006: 165f.) cite a few other traditional grammars of Portuguese which make the same distinction of three types of causal connections, among them Barros (1961). The Nomenclatura Gramatical Brasileira of 1959 and the Nomenclatura Gramatical Portuguesa of 1967 include in their typology of sentence connections two classes which are conceptually related to Sweetser’s epistemic and speech-act connections: conclusive and explicative connections (see Cunha & Cintra 2009: 595, 611f.; Peres & Mascarenhas 2006: 158ff.). But traditional Portuguese grammarians do not seem to have been interested in exploring in depth the semantic peculiarities of those connections. The main focus of interest used to be on their syntactic behaviour on the scale between coordination and subordination. The fact that conclusive and explicative connections link propositions or speech acts, while other classes of connections link states of affairs, passed unobserved until very recently or was at least not taken as a starting point for further investigation. This is also true for Gärtner (1998), who does not distinguish between connections of states of affairs, propositions or speech acts, even though a considerable part of his grammar is dedicated to sentence connections. In his chapter on complex sentences (1998: 337–582) he focuses exclusively on relations between states of affairs (temporal objects). Consequently he investigates in detail the sequence of tenses in the connected clauses (also see ibid.: 449ff.), whereas modality is treated in less depth. Mateus et al. (2003), when dealing with the semantic relata of sentence connections, always speak of propositions and not of states of affairs, or of speech acts. For causal connections (2003: 108) they distinguish between temporal cause-effect relations, inferential evidence-conclusion relations, and relations between acts and their motives, but the examples they give in order to illustrate these types of relations are not very helpful: (11) A salsa ficou viçosa porque a reguei todos os dias. ‘The parsley has become huge because I watered it every single day.’

Modal readings of sentence connectives in German and Portuguese

21

(12) Visto que choveu nas alturas certas, a colheita deste ano é excepcional. ‘Given that / As it rained at the right times, the harvest this year is outstanding.’ (13) Como os alunos estavam cheios de trabalho esta semana, adiei o teste. ‘As the students had so much work to do this week I postponed the test.’

Example (11) clearly shows a temporal relation between states of affairs. In example (12), an epistemic reading (connection of propositions) does not seem plausible. From knowing that it rained at the right times it is not possible to conclude that the harvest is outstanding. A temporal reading of (12), as paraphrased in (12a), seems much more probable: (12a) This year an outstanding harvest has occurred, and this state of affairs was caused by another state of affairs, namely that it rained at the right times.

Modal readings, in contrast, are preferable for the variants in (12b) and (12c): (12b) Visto que choveu nas alturas certas, a colheita deste ano deve ser excepcional. ‘Given that / As it rained at the right times, the harvest this year must be outstanding.’ (12c) Visto que não choveu nas alturas certas, a colheita deste ano é mesmo excepcional. ‘Considering that it didn’t rain at the right times, the harvest is actually outstanding.’

In (12b) the modal auxiliary construction deve ser indicates that a conclusion is drawn (epistemic connection). In (12c) the speech act of evaluating the harvest is contrasted with conditions which do not really seem to support the selected positive value (deontic-illocutionary concessive connection). Example (13) is intended by Mateus et al. to illustrate the justification of an act by a motive. The relevant act is the postponing of the test by the speaker. This act, however, is only reported and not performed in the example. Therefore the causal connection cannot be interpreted as a justification of a speech act. Instead, a temporal reading as in (13a) must be chosen: (13a) The speaker knew that the students had too much work to do, and this state of affairs made him put into practice another state of affairs, namely the postponing of the test.

In another section, under the heading conexões inferenciais [inferential connections], Mateus et al. (2003: 97) deal with sentence connections which present the truth of a proposition as evidence for or an argument in favour of the truth of another proposition. The other proposition, then, is a conclusion drawn from the first. They give three examples, two of which do not, in fact,

22

Hardarik Blühdorn & Tinka Reichmann

allow inferential (epistemic) readings. For the third one an inferential reading is possible, but not preferred: (14) Estava mau tempo e por isso decidimos ficar em casa. ‘The weather was bad and therefore we decided to stay at home.’ (15) Chegámos atrasados, pois está um trânsito infernal. ‘We arrived late because the traffic is terrible.’ (16) O João está constipadíssimo e portanto não vem à festa. ‘John has a bad cold and therefore he will not come to the party.’

In the first two examples the connections are clearly temporal. From observing the weather or the road traffic, it is not possible to draw conclusions about one’s own staying at home or arriving late in the past. The weather and the traffic in these examples are not to be interpreted as evidences for the truth of any propositions, but simply as states of affairs which caused other states of affairs to happen, in this case staying at home and arriving late (similar observations in Peres & Mascarenhas 2006: 151f.). In the third example an epistemic reading is possible: from knowledge about the health of a person it is possible to make predictions of whether this person will come to a party. Without any context, however, a temporal reading is preferable also for this example, since John having a cold is probably the state of affairs which prevents his going to the party from happening. Sweetser’s distinction of the different levels of connection is explicitly taken up and elaborated for Portuguese by Neves (2000: 804ff.). In her chapter on causal connections she distinguishes between the linking of states of affairs (estados de coisas), of propositions (proposições) and of speech acts (atos de fala). Here are two of her examples: (17) Não deve ter havido nada porque seria a primeira pessoa a tomar conhecimento disto. ‘There should not have been anything (it must have been nothing), because I would have been the first person to know about it.’ (18) Vamos cantar pra Santa Clara uma reza pra ela não deixar chover hoje de noite. Você canta comigo, porque Santa Clara gosta muito de crianças. ‘Let us sing a prayer for Saint Clara not to give us rain tonight. You sing along with me, because Saint Clara loves children.’

In (17) the knowledge of the speaker that s/he would have been the first person to be informed is presented as an argument for supposing that nothing serious happened. This is a clear case of an epistemic connection between two propositions. One proposition is considered to be true, and that leads to the conclusion that the other one is also true. A plausible interpretation of example

Modal readings of sentence connectives in German and Portuguese

23

(18) is that the speaker asks her child to sing along with her and that Saint Clara’s sympathy for children serves as a justification or motive for this speech act. Here the clauses are linked at the deontic-illocutionary level. References to Sweetser’s distinction of the levels of connection can also be encountered in some of the papers in volumes VII (Neves 1999) and VIII (Abaurre & Rodrigues 2002) of the publications resulting from the research project Gramática do Português Falado [Grammar of Spoken Portuguese], e. g., in the papers by Camacho on additive, by Pezatti on alternative, and by Neves on causal, conditional and concessive sentence connections (all in Neves 1999: 351–591) as well as in Pezatti’s (2002) article on conclusive sentence connections. Another recent work which makes explicit reference to Sweetser is the article on sentence connections in Portuguese by Peres & Mascarenhas (2006). But the typology of connections proposed by those authors (2006: 146ff.) follows the tradition of Portuguese grammar handbooks in focussing on predominantly syntactic criteria and arranging the classes of connections on a gradient between juxtaposition and subordination. The semantic category of the connected relata does not play a central role in that approach.

3. Tests for distinguishing between temporal and modal readings of connectives The discussion of examples (12) to (16) from Mateus et al. (2003) has shown that it is not always easy to distinguish whether two clauses are linked at the temporal or at a modal level. Many sentence connections have more than one reading. Thus example (19) can be interpreted as a temporal, an epistemic or a deontic-illocutionary connection. The different readings can be distinguished by translating the example into German, as in (19a–c): (19) Desde que ele está sujeito a ser consultado sobre todo e qualquer assunto, deve ter uma base sólida de conhecimento em geral. (Neves 2000: 803) ‘Since he might be consulted on all and every subject, he must have a solid general knowledge.’ (19a) Seit es ihm passieren kann, dass er über alle möglichen Themen befragt wird, benötigt er ein solides Allgemeinwissen. ‘Since he was given a position where he might be consulted on all and every subject, he needs a solid general knowledge.’

24

Hardarik Blühdorn & Tinka Reichmann

(19b) Da es ihm passieren kann, dass er über alle möglichen Themen befragt wird, wird er wohl ein solides Allgemeinwissen haben. ‘Since/as he might be consulted on all and every subject, I suppose he has a solid general knowledge.’ (19c) Da es ihm passieren kann, dass er über alle möglichen Themen befragt wird, verlange ich von ihm, dass er ein solides Allgemeinwissen hat. ‘Since/as he might be consulted on all and every subject, I want him to have a solid general knowledge.’

The German connective seit, in contrast to since in English and desde que in Portuguese, can only encode temporal connections. Thus (19a) requires a temporal reading: the beginning of the time interval in which it is factual that the person referred to can be consulted on all and every subject coincides with the beginning of the time interval in which it is factual that this person needs a solid general knowledge. In this case, the modal auxiliary construction deve ter in (19) has been interpreted in the sense of Kratzer’s (1991: 640) “circumstantial modality” and has therefore been translated by the verb benötigen [to need]. The translations in (19b) and (19c) use the conjunction da [as] instead of seit. Both are possible equivalents of desde que, but seit and da are specialized on different types of connections and, therefore, reveal semantic differences which remain implicit with desde que (and with since). The conjunction da (see Blühdorn 2006: 259ff., 272ff.; 2009) is hardly ever used for temporally situating connections in contemporary German. In the great majority of cases it encodes conditional or causal connections and, more specifically, modal ones (for details of the typology of relations see Blühdorn 2003: 13ff.). (19b) and (19c) have identical translations of the subordinate clause, but differ in the translation of the main clause. In (19b) deve ter has been translated by a modal auxiliary construction. Werden [will] together with the epistemic particle wohl [possibly] suggests an epistemic interpretation: the speaker knows that the person referred to can be consulted on all and every subject, and this information leads him/her to suppose that the person has a solid general knowledge. In (19c) deve ter has been translated by the performative main verb verlangen [to demand]. A modal auxiliary construction with müssen [must], perhaps with the additional deontic particle unbedingt [definitely], would also be appropriate to achieve the same deontic reading: the fact that the person referred to can be consulted on all and every subject is the motive which makes the speaker demand that this person should have a solid general knowledge. As we have seen, desde que in (19) is ambiguous. It can be read as a connective at the temporal, the epistemic or the deontic-illocutionary level.

Modal readings of sentence connectives in German and Portuguese

25

Which reading is chosen in each individual case depends strongly on the context of use. Within a narrative context such as an overview of a person’s professional career, a temporal reading might be most suitable. In a context in which the knowledge of a person is being discussed, an epistemic reading might be preferable. And if the context is about the speaker’s expectations with respect to a person’s capacities, a deontic-illocutionary reading might be most appropriate. While many sentence connections in empirical data are ambiguous as to the semantic level of linking, many others do not allow for more than one reading – partly for contextual and partly for grammatical reasons. It is therefore important to discuss how sentence connections can be tested with respect to possible readings. The availability of reliable test procedures is an indispensable requirement for integrating the distinction between temporal and modal uses of sentence connectives into the grammaticography of Portuguese and German. In what follows, we will briefly present four test procedures which can be used for this purpose (for more details see Blühdorn 2005: 317ff.; 2006: 265ff.; 2008a). 3.1. Paraphrasing In the previous sections we used paraphrases which contained the key terms state of affairs/proposition/speech act, cause/evidence/motive, and factuality/ truth/desirability in order to distinguish between the different readings of sentence connections. Thus, clauses which are temporally connected can be embedded into the formula it is (not) the case that S or S has (not) happened, without bringing about a change of descriptive meaning (in the sense of Lyons 1977: 50f., 197). If modally connected clauses are embedded into these formulas, their descriptive meaning as well as their speech-act meaning will change significantly. Clauses linked by an epistemic connection can be embedded into the formula the speaker believes (does not believe) that P or P is (not) true; clauses linked by a deontic-illocutionary connection can be embedded into the formula it is (not) desirable that A or the speaker wants (does not want) A. Such paraphrases make explicit part of the speech-act meaning of the sentences, without changing their descriptive meaning. (20a) is a paraphrase for example (19) which forces a temporal reading, (20b) is a paraphrase for the same connection that forces an epistemic interpretation, and (20c) is a paraphrase that forces a deontic-illocutionary interpretation:

26

Hardarik Blühdorn & Tinka Reichmann

(20a) Desde que é o caso que ele está sujeito a ser consultado sobre todo e qualquer assunto, é também o caso que ele necessita de uma base sólida de conhecimento em geral. ‘Since it is the case that he might be consulted … it is also the case that he needs…’ (20b) Desde que é verdade que ele está sujeito a ser consultado sobre todo e qualquer assunto, pode-se supor que ele tem uma base sólida de conhecimento em geral. ‘As it is true that he might be consulted …, it can be supposed that he has…’ (20c) Desde que ele está sujeito a ser consultado sobre todo e qualquer assunto, é desejável (o locutor exige) que ele tenha uma base sólida de conhecimento em geral. ‘As he might be consulted …, he is required to have / the speaker wants him to have…’

3.2. Scope of negation and modal adverbs A second important test concerns the scope of semantic operators. Temporal connectives can enter into the scope of negation and of modal sentence adverbs, while modal connectives must remain outside the scope of such operators. For illustration, let us look at example (21): (21) A salsa não ficou tão viçosa(,) porque a reguei todos os dias. ‘The parsley has not become so huge(,) because I watered it every single day.’

There are three different readings available for (21). The first one is that the speaker watered the parsley every single day, that the parsley has become huge, but that the former was not the cause of the latter: (21a) (não porque (reguei (a salsa))S (ficou tão viçosa (a salsa))S)P4 (not because (I watered (the parsley))S (has become so huge (the parsley))S)P

In this reading the connective is within the scope of negation. The connected relata must be states of affairs (S), i. e., the connection is temporal. The sentence as a whole is one proposition (P) (see Peres & Mascarenhas 2006: 117). It is denied that the fact of watering was the cause of the occurrence of the luxuriance. If this reading is intended, no comma between the connected clauses is used in written Portuguese (see Luft 1996: 63f.). In spoken Portuguese the whole sentence is realized as one intonation phrase. An 4

In order to illustrate scope relations, we use simple formulas which represent the hierarchy of operators and operands by bracketing and linear sequence, but which remain as close as possible to natural surface structure in all other respects. The subscripts S and P indicate the semantic levels of states of affairs and propositions.

Modal readings of sentence connectives in German and Portuguese

27

(ascending) secondary stress can fall, e. g., on the negative particle não in the main clause; the (descending) focal stress can be put on the verb reguei in the subordinate clause. The second interpretation of (21) is that the speaker watered the parsley every single day, that the parsley has not become so huge and that the excess of water was the cause of the failure to become huge: (21b) (porque (reguei (a salsa))S (não ficou tão viçosa (a salsa))S)P (because (I watered (the parsley))S (has not become so huge (the parsley))S)P

In (21b) the connective is outside the scope of negation. Negation only has scope over the main clause. In this reading, the connected relata can, again, be states of affairs, i. e., the connection can be established at the temporal level. Then the sentence as a whole is one proposition. It is asserted that the fact of watering was the cause of the parsley’s not becoming huge. If this reading is intended, there is a tendency to separate the connected clauses by a comma in written Portuguese. In spoken Portuguese the whole sentence can still form one single intonation phrase. In this case, however, the secondary stress will not be placed on the negative particle não, but rather on the predicative viçosa, and the focal stress will tend to fall on dias. A third interpretation of (21) is that the speaker, who knows that s/he watered the parsley every day, concludes that it can therefore not have become very huge. In this reading, the knowledge of the excessive watering serves as evidence which supports the speaker’s conclusion: (21c) porque (sei que (reguei (a salsa))S)P (concluo que (não ficou tão viçosa (a salsa))S)P because (I know that (I watered (the parsley))S)P (I conclude that (has not become so huge (the parsley))S)P

In (21c) the connective is not only outside the scope of negation, but also outside the connected propositions, i. e., the connected relata are propositions and not states of affairs. The connection is therefore at the epistemic level. If this reading is intended, school grammars of Portuguese require a comma between the connected clauses (see Luft 1996: 63). In spoken Portuguese the clauses are realized as separate intonation phrases. In the first intonation phrase, focal stress can be placed on ficou, in the second one it can fall on dias. The discussion has shown that in temporal connections the connective can be within the scope of negation. This is not possible in epistemic, nor in deontic-illocutionary connections. Similar observations can be made for modal sentence adverbs such as talvez in (22). For such adverbs, only the first and third readings are available, i. e., in a temporal connection a modal sentence adverb cannot be within the scope of

28

Hardarik Blühdorn & Tinka Reichmann

the connective. For lack of space we confine ourselves here to giving the formulas for the different readings: (22) Talvez a salsa tenha ficado tão viçosa(,) porque a reguei todos os dias. ‘Maybe the parsley has become so huge(,) because I watered it every single day.’ (22a) (talvez porque (reguei (a salsa))S (tenha ficado tão viçosa (a salsa))S)P (maybe because (I watered (the parsley))S (has become so huge (the parsley))S)P (22b) #(porque (reguei (a salsa))S (talvez tenha ficado tão viçosa (a salsa))S)P5 #(because (I watered (the parsley))S (maybe has become so huge (the parsley))S)P (22c) porque ((reguei (a salsa))S)P (talvez (tenha ficado tão viçosa (a salsa))S)P because ((I watered (the parsley))S)P (maybe (has become so huge (the parsley))S)P

In written Portuguese a comma is used between connected clauses if the main clause contains a negative element or a modal sentence adverb which does not have scope over the connective. No comma is used if the connective is within the scope of negation or of a modal sentence adverb. Thus, modally connected sentences usually have a comma in written Portuguese, while temporally connected ones do not (see Luft 1996: 63f.). We can use example (17), which contains a negative particle in the main clause, as an additional illustration of the scope test: (17) Não deve ter havido nada porque seria a primeira pessoa a tomar conhecimento disto. ‘There should not have been anything/nothing, because I would have been the first person to know about it.’

If a temporal reading is available for this example, then the connective should be able to fall within the scope of negation: (17a) (deve não porque (seria (a primeira pessoa a x))S (ter havido (nada))S)P (should not because (I would have been (the first person to x))S (have been (nothing))S)P

In such a reading of (17), the interpreter would have to presuppose that nothing happened. The question under discussion would be whether the fact that nothing happened was caused by the fact that if anything had happened, then the speaker would have been the first to know about it. (17) would then be a means of uttering the belief that there is no such relation of causality. Such a 5

We use the # symbol to mark expressions that are semantically deviant. (22b) is deviant because a modal sentence adverb such as talvez cannot be part of a description of a state of affairs.

Modal readings of sentence connectives in German and Portuguese

29

reading of (17) is theoretically possible, but it seems very unlikely that a reallife context could be found in which it is appropriate. In order to indicate that such a reading is intended, the speaker would have to realize the whole sentence as one intonation phrase. An ascending secondary stress would have to be put on the negative particle não, and the focal stress on conhecimento. Thinking of real-life situations, it seems to be much more probable that (17) would occur in a context in which the question under discussion is how the speaker comes to suppose that nothing has happened. In this case the connective cannot be within the scope of negation. Consequently, only a modal reading is available: (17c) porque ((seria (a primeira pessoa a x))S)P (deve (não ter havido (nada))S)P as ((I would have been (the first person to x))S)P (it should (have been (nothing))S)P

In order to guarantee this reading, according to traditional grammar, (17) should have been written with a comma. For structural reasons, the scope test can only be applied to connections in which the subordinate clause follows the main clause. Connections in which it precedes the main clause such as (19) must first be reordered as in (23a/b): (23a) Ele não deve ter uma base sólida de conhecimento em geral, desde que / visto que ele está sujeito a ser consultado sobre todo e qualquer assunto. ‘He does not have to have a solid general knowledge, since/as he might be consulted on all and every subject.’ (23b) Talvez ele deva ter uma base sólida de conhecimento em geral, desde que / visto que ele está sujeito a ser consultado sobre todo e qualquer assunto. ‘Maybe he must have a solid general knowledge, since/as he might be consulted on all and every subject.’

3.3. Verb tenses and temporal adjuncts In temporal connections, all temporal elements contained in the clauses, such as verb tenses, temporal adverbs, and temporal prepositions, are semantically related to each other and build up a coherent chronological system (see Lehmann 1988: 205). In modal connections, the verb tenses and other temporal elements of each clause relate independently to the time of utterance. For illustration, let us first have a look at the tenses in example (24): (24) Visto que choveu nas alturas certas, a colheita deste ano é excepcional. [chover-perfect preterit] [ser-simple present] ‘Given that / As it rained at the right times, the harvest this year is outstanding.’

30

Hardarik Blühdorn & Tinka Reichmann

A temporal interpretation of (24) depends on the assumption that it rained in a first moment and that the harvest became outstanding at a later point in time. The tenses in the clauses are distributed correspondingly: pretérito perfeito [past tense] in the subordinate clause and present tense in the main clause. An inverted chronology of the states of affairs, indicated by a different distribution of tenses, as e. g. in (24a), would not be compatible with our knowledge of the world: (24a) Visto que vai chover nas alturas certas, a colheita deste ano é excepcional. [chover-future] [ser-simple present] ‘Given that / As it will rain at the right times, the harvest this year is outstanding.’

A temporal reading of (24a) is inadequate, but depending on the context, an epistemic or a deontic-illocutionary reading may be available. In (25), the same distribution of tenses as in (24a) provides a chronology which is compatible with a temporal reading of the connection: (25) Já que vai chover durante o fim-de-semana, passamos por um posto de gasolina para meter água no reservatório do limpa-parabrisas. . [chover-future] [passar-simple present] ‘As it will rain at the weekend, we go to a petrol station to put some water into the windscreen-wiper reservoir.’

A possible interpretation of (25) is that filling up the water reservoir is meant as a preparation for a rainy weekend. In this example, an inverted chronology, indicated by a tense distribution as in (24), would also be compatible with a temporal reading. But then a purpose interpretation of the connection is quite improbable. In (25a) the rainy weather rather appears to be the cause that leads to filling up the water reservoir: (25a) Já que choveu durante o fim-de-semana, passamos por um posto de gasolina para meter água no reservatório do limpa-parabrisas. [chover-perfect preterit] [passar-simple present] ‘As it rained at the weekend, we go to a petrol station to put some water into the windscreen-wiper reservoir.’

These examples show how the reconstruction of the chronology of states of affairs affects the interpretation of temporal connections. As long as a coherent chronology can be construed in accordance with general world knowledge, temporal readings of connections are possible. Otherwise they are excluded. (24b) is a variant of (24) for which an epistemic reading of the connection is preferable:

Modal readings of sentence connectives in German and Portuguese

31

(24b) Visto que choveu nas alturas certas, a colheita deste ano deve ser excepcional. [chover-perfect preterit] [modal auxiliary dever-simple present] ‘As it rained at the right times, the harvest this year must be outstanding.’

In this variant, the tenses in the clauses are not related to each other. They relate independently to the time of utterance. In conditional and causal connections, the tense of the verb in the main clause indicates the moment in which the speaker draws the relevant conclusion. It is typically a present tense form. Past tenses only occur together with deictic shifts such as in reported speech. The tense of the verb in the subordinate clause is more flexible. It encodes the temporal relation between the time of utterance and the source of the evidence which makes the speaker draw the conclusion. Similar observations hold for adverbial adjuncts of time. In temporal connections, they not only influence the interpretation of the clause of which they are a constituent, but also the interpretation of the sentence as a whole: (26) Como os alunos estavam cheios de trabalho esta semana, adiei o teste. ‘As the students had so much work to do this week I postponed the test.’ (27) Foi isso que me veio à mente esta manhã, enquanto eu caminhava. ‘This was what came to my mind this morning while I was out for a walk.’

Example (26) suggests that the test was postponed in the same week in which the students had so much work to do, and (27) means that the walk happened on the same morning on which the idea came to the mind of the speaker. In modal connections, in contrast, temporal adverbs can only influence the interpretation of clauses of which they are constituents: (28) Como os alunos estavam cheios de trabalho no mês passado, devem estar bem cansados. ‘As the students had so much work to do last month they must be very tired.’ (29) Uma das vítimas morreu esta manhã, enquanto a outra faleceu na noite de sextafeira. ‘One of the victims died this morning, while the other one perished on Friday night.’

In (28), the supposition that the students are tired is not made in the month before the moment of the utterance, but at the moment of the utterance itself. In (29) both clauses have a temporal adjunct of their own. Here the connective enquanto [while] cannot be understood in the sense of a temporal overlap of the states of affairs described. It must be read in the modal sense of comparing the propositions with respect to their truth.

32

Hardarik Blühdorn & Tinka Reichmann

3.4. Cleft constructions Subordinate clauses linked by a temporal connection have the syntactic status of adverbial adjuncts. In many cases they can be separated from the rest of the main clause by a cleft construction in which the subordinate clause is leftdislocated and the rest of the main clause takes the form of a que[that]-clause. The two parts are linked by a form of the verb ser [to be] (see Reichmann 2005: 25–35): (30) É porque a reguei todos os dias que a salsa ficou viçosa. ‘It is because I watered it every single day that the parsley has become huge.’ (31) É desde que está sujeito a ser consultado sobre todo e qualquer assunto que ele precisa de uma base sólida de conhecimento. ‘It is since he might be consulted on all and every subject that he must have a solid general knowledge.’ (32) Foi enquanto eu caminhava esta manhã que isso me veio à mente. ‘It was while I was out for a walk this morning that this came to my mind.’

In contrast, subordinate clauses linked to the main clause by a modal connection cannot be separated by cleft constructions (see Kortmann 1996: 29): (33) #Foi enquanto uma das vítimas faleceu na noite de sexta-feira que a outra morreu esta manhã. ‘It was while one of the victims perished on Friday night that the other one died this morning.’ (34) *É porque isso baiano não gosta que não trabalhem demais. ‘It is because people from Bahia don’t like it that don’t work too much.’ (35) É porque Santa Clara gosta muito de crianças que você canta comigo. ‘It is because Saint Clara loves children that you sing along with me.’ (36) #É porque, conhecendo a causa da doença, é possível evitá-la que essa é a questão mais importante. ‘It is because, if you know the cause of the illness, you can avoid it that this is the most important question.’ (37) #É porque eu seria a primeira pessoa a tomar conhecimento disto que não deve ter havido nada. ‘It is because I would have been the first person to know about it that it must have been nothing.’

(33) is semantically deviant. In the cleft construction only a temporal reading would be available, but a temporal reading of the connective enquanto [while] (which indicates a relation of simultaneity) would be incompatible with the

Modal readings of sentence connectives in German and Portuguese

33

temporal adjuncts in the connected clauses and is therefore blocked. (34) is syntactically deviant because commands cannot be embedded in que-clauses. For the same reason você canta comigo [you sing along with me] in (35) cannot be interpreted as a request, but only as a description of a state of affairs. While the original sentence (18) allows a deontic-illocutionary reading of the connection, the cleft construction in (35) can only be read temporally. Examples (36) and (37) are semantically deviant in the above wording, but they can be reformulated in such a way that temporal readings of the connections become possible: (36a) É porque, conhecendo a causa da doença, é possível evitá-la que eu considero essa questão tão importante. ‘It is because, if you know the cause of the illness, you can avoid it that I regard this question as so important.’ (37a) É porque eu seria a primeira pessoa a tomar conhecimento disto que eu acredito que não houve nada. ‘It is because I would have been the first person to know about it that I believe it was nothing.’

Unlike in the original sentences (3) and (17), no direct conclusions are drawn in (36a) and (37a). Instead, the conclusions and the evidence on which they were based are reported. In this case the clauses are connected at the temporal level (similar examples in Peres & Mascarenhas 2006: 153f.), which can be recognised, among other things, by the fact that the tenses are now related to each other and can be changed together: (36b) Era porque, conhecendo a causa da doença, é possível evitá-la que eu considerava essa questão a mais importante. ‘It was because, if you know the cause of the illness, you can avoid it that I regarded this question as so important.’ (37b) Foi porque eu seria a primeira pessoa a tomar conhecimento disto que eu acreditei que não houve nada. ‘It was because I would have been the first person to know about it that I believed it had been nothing.’

“Anchor” clauses of modal connections are generally excluded from leftdislocation in cleft constructions. If a cleft construction is possible without a change of descriptive or speech-act meaning, then it is clear that the clauses are connected temporally. But not all temporal connections are equally compatible with cleft constructions:

34

Hardarik Blühdorn & Tinka Reichmann

(38) ?Foi como os alunos estavam cheios de trabalho esta semana que adiei o teste. ‘It was as the students had so much work to do this week that I postponed the test.’

The conjunction como [since], besides encoding a semantic relation that can be interpreted as cause – consequence, signals that the state of affairs described in the subordinate clause is already known to the addressee. This property does not fit well with the information structure established by the cleft construction. In a cleft construction, the left-dislocated expression is typically marked as focal information, the rest of the sentence being background information. Leftdislocation of a clause introduced by como is therefore odd in most cases. 3.5. Interpretation of test results Not all the tests lead to equally reliable results in all cases: – The paraphrase test is, in principle, quite reliable for connections at all three levels, but as it is carried out, at least partially, in everyday language, it is not independent of individual preferences of wording and can therefore not always eliminate differences of judgement between individual interpreters. – The scope test can unambiguously exclude temporal and modal readings of connections. A connective which is within the scope of negation and/or of an epistemic operator can only be interpreted temporally. A connective which has scope over a modal operator (e. g., an epistemic or deontic modal auxiliary, a modal sentence adverb or a modal particle) can only be interpreted at a modal level. But a connective which has scope over negation still allows temporal and modal readings. – The tense test can unambiguously attest and exclude temporal readings. If tenses and temporal adjuncts interact semantically beyond the boundaries of the connected clauses, then the connection is temporal. If tenses and temporal adjuncts do not interact between the clauses but relate independently of each other to the time of utterance, then the connection can only be interpreted at a modal level. – The cleft test can unambiguously attest temporal and exclude modal readings of connections. If a cleft construction is possible without a change of descriptive or speech-act meaning, the connection must be temporal. If it is not possible, the connection will typically be modal. In order to reliably distinguish between temporal and modal readings of empirical sentence connections, all four test procedures should be carried out if possible.

Modal readings of sentence connectives in German and Portuguese

35

4. Further perspectives: Semantic specialisation of linking constructions In German, most connectives can be used for establishing connections at all three levels (see Blühdorn 2003: 15ff.; 2004a: 129; 2004b: 191f., 203ff.; 2006: 265; 2008a). For Portuguese, Neves (2000: 804f.) observes that connections with porque [because] usually have epistemic and/or deontic-illocutionary, but only rarely temporal readings. This observation can be confirmed by examining empirical data. But why should this be the case? The German equivalent of porque, the subordinating conjunction weil, is used significantly more frequently for encoding temporal connections than for encoding modal connections (see Frohning 2007: 130ff.). How can this difference between weil and porque be explained? Unlike German, Portuguese makes extensive use of (impersonal and personal) infinitive clauses embedded by prepositions, as in the following examples: (39) Três funcionários de empresa são presos por fazer cópias piratas de DVDs. ‘Three employees are arrested for making (impersonal infinitive) illegal copies of DVDs.’ (40) Estudantes paraenses recebem computadores por terem chegado à semifinal do jogo. ‘Students of the state of Pará receive computers for having reached (personal infinitive) the semi-finals of the game.’

Constructions of this kind always suggest temporal readings in Portuguese (see Peres & Mascarenhas 2006: 162f.). Example (39) can only mean that the factuality of the copying was the cause of the factuality of the arrest, and (40) means that the factuality of reaching the semi-finals led to the factuality of receiving the computers. Modal readings are excluded in these examples. Portuguese infinitive clauses are unmarked for mood (indicative vs. subjunctive vs. imperative) (see Cunha & Cintra 2009: 496). This means that they do not contribute actively to the overall modal semantics of the sentence. They simply fit into the modal structure provided by the main clause. In finite clauses, mood is an important and pervasive grammatical device for encoding epistemicity and deonticity. Even though infinitive clauses are not necessarily totally devoid of epistemic and deontic information, they are clearly less appropriate than finite clauses for encoding independent propositions and/or speech-acts. It seems natural, then, that infinitive clauses (if available) are preferred for encoding temporal connections and finite clauses for encoding modal

36

Hardarik Blühdorn & Tinka Reichmann

connections. Some of the sentence connectives of Portuguese are arranged in lexical pairs, each consisting of a preposition and a corresponding conjunction: por/porque,para/para que, sem/sem que, desde/desde que, até/até que, antes de/antes que, depois de/depois que etc. We leave aside for future empirical research the question of whether all these pairs manifest the same division of labour as por and porque.

References Abaurre, Maria Bernadete M. & Rodrigues, Angela C. S. (eds.) (2002): Gramática do Português Falado. Vol. VIII: Novos estudos descritivos. – Campinas: Editora da Unicamp. Barros, Tomás de (141961): Gramática Portuguesa. – Porto: Editora Educação Nacional. Bechara, Evanildo (371999): Moderna Gramática Portuguesa. – Rio de Janeiro: Lucerna. Blatz, Friedrich (31896/1970): Neuhochdeutsche Grammatik mit Berücksichtigung der historischen Entwickelung der deutschen Sprache. – Karlsruhe: Lang (Reprint Hildesheim: Olms). Blühdorn, Hardarik (2003): “Zur Semantik der Konjunktion als. Paradigmatische und syntagmatische Aspekte.” – In: Elke Hentschel (ed.): Particulae Collectae. Festschrift für Harald Weydt zum 65. Geburtstag. Linguistik online 13, 11–53. (URL: http://www.linguistik-online.com/13_01/bluehdorn.pdf [02/09/2009]) – (2004a): “Temporalkonnektoren: Einleitung.” – In: Hardarik Blühdorn, Eva Breindl & Ulrich H. Waßner (eds.) (2004): Brücken schlagen. Grundlagen der Konnektorensemantik, 125–136. Berlin: de Gruyter. – (2004b): “Die Konjunktionen nachdem und bevor.” – In: Hardarik Blühdorn, Eva Breindl & Ulrich H. Waßner (eds.) (2004): Brücken schlagen. Grundlagen der Konnektorensemantik, 185–211. Berlin: de Gruyter. – (2005): “Zur Semantik kausaler Satzverbindungen: Integration, Fokussierung, Definitheit und modale Umgebung.” – In: Studi Linguistici e Filologici Online. Rivista Telematica del Dipartimento di Linguistica dell’Università di Pisa (SLiFO) 3 (2), 311–338. (URL: http://www.humnet.unipi.it/slifo/2005vol2/Bluhdorn3.2.pdf [02/09/2009]) – (2006): “Kausale Satzverknüpfungen im Deutschen.” – In: Pandaemonium Germanicum. Revista de Estudos Germanísticos (Universidade de São Paulo) 10, 253–282. (URL: http://www.fflch.usp.br/dlm/alemao/pandaemoniumgermanicum/site/images/ pdf/ed2006/Kausale_Satzverknpfungen_im_Deutschen.pdf [02/09/2010]) – (2008a): “Epistemische Lesarten von Satzkonnektoren. Wie sie zustande kommen und wie man sie erkennt.” – In: Inge Pohl (ed.): Semantik und Pragmatik – Schnittstellen, 217–251. Frankfurt a. M.: Lang.

Modal readings of sentence connectives in German and Portuguese

37

– (2008b): “On the syntax and semantics of sentence connectives.” – Ms., Mannheim: Institut für Deutsche Sprache. (URL: http://www.ids-mannheim.de/gra/texte/blu_connectives.pdf [02/09/2009]) – (2009): “Verknüpfungs-Eigenschaften deutscher Kausal-Konnektoren zwischen syntaktischer Hierarchie und Linearität.” – Ms., Mannheim: Institut für Deutsche Sprache. (URL: http://www.ids-mannheim.de/gra/texte/blu_verknuepfungseigenschaften.pdf [02/09/2010]) Blühdorn, Hardarik, Breindl, Eva & Waßner, Ulrich H. (eds.) (2004): Brücken schlagen. Grundlagen der Konnektorensemantik. Berlin: de Gruyter. Cunha, Celso F. da & Cintra, Luís F. Lindley (52009): Nova Gramática do Português Contemporâneo. – Rio de Janeiro: Lexikon. Duden (72005): Die Grammatik. – Mannheim: Dudenverlag. Frohning, Dagmar (2007): Kausalmarker zwischen Pragmatik und Kognition. Korpusbasierte Analysen zur Variation im Deutschen. – Tübingen: Niemeyer. Gärtner, Eberhard (1998): Grammatik der portugiesischen Sprache. – Tübingen: Niemeyer. Keller, Rudi (1995): “The epistemic weil.” – In: Dieter Stein & Susan Wright (eds.): Subjectivity and subjectivisation. Linguistic perspectives, 16–30. Cambridge: Cambridge University Press. Kortmann, Bernd (1996): Adverbial Subordination. A Typology and History of Adverbial Subordinators Based on European Languages. – Berlin: de Gruyter. Kratzer, Angelika (1991): “Modality.” – In: Arnim von Stechow & Dieter Wunderlich (eds.): Semantics. An International Handbook of Contemporary Research, 639–650. Berlin: de Gruyter. Lehmann, Christian (1988): “Towards a typology of clause linkage.” – In: John Haiman & Sandra Thompson (eds.): Clause Combining in Grammar and Discourse, 181– 225. Amsterdam: Benjamins. Luft, Celso Pedro (1996): A vírgula. Considerações sobre o seu ensino e o seu emprego. – São Paulo: Ática. Lyons, John (1977): Semantics. 2 vols. – Cambridge: Cambridge University Press. Mateus, Maria Helena, Brito, Ana Maria, Duarte, Inês et al. (52003): Gramática da Língua Portuguesa. – Lisboa: Caminho. Neves, Maria Helena de Moura (ed.) (1999): Gramática do Português Falado. Vol. VII: Novos Estudos. – Campinas: Editora da Unicamp. – (2000): Gramática de Usos do Português. – São Paulo: Ed. UNESP. – (2006): Texto e gramática. – São Paulo: Contexto. Pasch, Renate, Brauße, Ursula, Breindl, Eva & Waßner, Ulrich H. (2003): Handbuch der deutschen Konnektoren. Linguistische Grundlagen der Beschreibung und syntaktische Merkmale der deutschen Satzverknüpfer (Konjunktionen, Satzadverbien und Partikeln). – Berlin: de Gruyter. Peres, João Andrade & Mascarenhas, Salvador (2006): “Notes on sentential connections (predominantly) in Portuguese.” – In: Journal of Portuguese Linguistics 5, 113–169.

38

Hardarik Blühdorn & Tinka Reichmann

Pezatti, Erotilde Goreti (2002): “As construções conclusivas no português falado.” – In: Maria Bernadete M. Abaurre & Angela C.S. Rodrigues (eds.) (2002): Gramática do Português Falado. Vol. VIII: Novos estudos descritivos, 185–225. Campinas: Editora da Unicamp. Reichmann, Tinka (2005): Satzspaltung und Informationsstruktur im Portugiesischen und im Deutschen. Ein Beitrag zur Kontrastiven Linguistik und Übersetzungswissenschaft. – Frankfurt a. M.: Lang. – (2007): “A clivagem no português: Critérios de classificação e métodos de tradução.” – In: Yves Gambier, Miriam Shlesinger & Radegundis Stolze (eds.): Doubts and Directions in Translation Studies. Selected contributions from the EST Congress, Lisbon 2004, 253–266. Amsterdam: Benjamins. Sweetser, Eve (1990): From etymology to pragmatics. Metaphorical and cultural aspects of semantic structure. – Cambridge: Cambridge University Press. Truckenbrodt, Hubert (2006): “On the semantic motivation of syntactical verb movement to C in German.” – In: Theoretical Linguistics 32, 257–306. Zifonun, Gisela, Hoffmann, Ludger, Strecker, Bruno et al. (1997): Grammatik der deutschen Sprache. 3 vols. – Berlin/New York: de Gruyter.

Sarah Dessì Schmid

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison1

1. Introduction (1)

Tra i Tempi dell’Indicativo, l’Imperfetto […] è certo quello che presenta la maggior duttilità di significato e la più ampia varietà d’impiego. ‘Amongst the indicative tenses, the imperfect […] is certainly the one that presents the greatest flexibility of meaning and the widest variety of use.’ (Bertinetto 1986: 345)

One can doubtless agree with Bertinetto on this point concerning Italian. However – with at most a few restrictions and distinctions – one could also say the same of the other Romance languages.1 Amongst the “flexible meanings” and “various uses” of the imperfect, there are those which are considered ‘typical’ or ‘canonical’, that is to say ‘normal’, and others which are not. In a short introductory paragraph, I will outline the temporal and aspectual (i. e. ‘normal’) meanings and uses of the imperfect in Spanish and Italian, in order to then investigate further those referred to by the common label “valori modali dell’imperfetto” (modal values of the imperfect) (Bertinetto 1986: 368ff.) or “imperfectos modalizados” (modalized imperfects) (García Fernández 2004: 90ff. following Fernández Ramírez 1986). These are uses in which the imperfect can express various ‘degrees’ of modality – from nonfactuality to counter-factuality – and are precisely those which are called “atypical uses” or “non-canonical uses” (Bazzanella 1990: 439) in Italian linguistics and “usos derivados” (derived uses) in opposition to “usos básicos” (basic uses) (Pérez Saldanya 2004: 222) in Spanish linguistics. The possibility of using the imperfect not only in its temporal and aspectual, but also in its “atypical” modal meanings is interpreted as the result of a modal implicature and attests to the existence of a semantic link between the notion of past tense and the notion of remoteness from reality. Along with Detges (2001: 142ff.), I take that link to be metonymical, but I will focus – differently from Detges – on its aspectual (and particularly imperfective) rather than its temporal nature. 1

For a panoramic description of the imperfect, cf. Coseriu (1976) (where the imperfect is considered the key figure in his general analysis of the verbal system of Romance languages) and Dauses (1981). For an analysis of some differences in use of the imperfect in Spanish and Italian cf. Squartini (2004).

40

Sarah Dessì Schmid

I will present and contrast the phenomena in both languages and then examine the semantic mechanisms underlying such usage in the context of recent cognitive Romance research. In doing so, I will also address certain other questions: a) Is the phenomenon of transfer from temporal to modal usage generally characteristic of non-present tenses, or are there certain specifics for the imperfect, and if so, what are they? b) Do these uses truly have a parallel distribution in Italian and Spanish? And finally c) is it appropriate to call certain uses “non-canonical” or “derived”? A few side-glances on diachrony will back my analysis.

2. The imperfect: temporal and aspectual functions The functions generally attributed to the imperfect are mostly classified as follows (cf. e. g. Bazzanella 1990: 439): i. expression of the duration of an event in the past; ii. expression of iterativity/habituality of events in the past; iii. expression of the simultaneity of two events in the past; iv. and finally the ‘background function’, which illustrates the background of states of affairs described (in the past). Both the temporal and aspectual properties of the imperfect can be identified in these functions. In fact, the questions and interpretative positions in research on the imperfect revolve around the “hipótesis temporal” (temporal hypothesis) and the “hipótesis aspectual” (aspectual hypothesis) (cf. García Fernández 2004). They can be summarized in the central question: Should the imperfect be ascribed a mainly temporal or an aspectual nature? In traditional grammars, the imperfect of the Romance languages is considered to be a tense. It is therefore classified based mainly on its temporal meanings (as a tense of the indicative mood) and described and analyzed as such. The temporal properties of the imperfect are quite extensive, as becomes apparent in a summary of everything the imperfect can express on the temporal axis (cf. Bertinetto 1986: 353ff.):

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

41

i. simultaneity of two events in the past; ii. presence in the past; iii. future in the past.2 However, to declare the imperfect to be a ‘pure’ tense or to focus on its predominantly or exclusively temporal nature, as the proponents of the “hipótesis temporal”3 do, leaves too many problems unsolved – especially if one uses the traditional system of Reichenbach (cf. 1947: 287ff.), who defines tense as the expression of the temporal relation between three points (and these three only): point of speech (S) – point of reference (R) – point of the event (E): (2)

el sistema de Reichenbach no puede diferenciar, en términos temporales, entre pretérito imperfecto y pretérito perfecto simple. Ello quiere decir que en este sistema es necesario recurrir a otra categoría gramatical para diferenciar entre ambas formas. ‘Reichenbach’s system cannot differentiate between pretérito imperfecto and pretérito perfecto simple in temporal terms. This means that in this system it is necessary to resort to a different grammatical category in order to distinguish between the two forms.’ (García Fernández 2004: 30)4

Even if the problem of exact differentiation or, more generally, of the relationship between perfective and imperfective forms is approached differently and resolved in part in more complex temporally-based systems,5 every purely temporal interpretation of the imperfect leaves many other aspects unexplained: for example why the Romance languages show such a redundancy of past tenses; or why the imperfect, regarded in its diachronic development, mainly has an aspectual function in older texts. It is thus not

2

3 4

5

However in this last function, certain traits are displayed which cannot be explained merely by temporal meanings; cf. Bertinetto (1986), Bazzanella (1990), Squartini (2004) and (2005). This usage is primarily found in spoken Italian. Cf. e. g. Bello (1841) and (1847), Bull (1960), Rojo (1974), Veiga (1988), (1992), (2004) and Rojo & Veiga (1999), Weinrich (62001), cf. also Coseriu (1976). Though Reichenbach introduced diacritics for aspectual differentiation, these do not sufficiently resolve the questions concerning the relationship between perfective and imperfective uses. Literature on this subject is very extensive; alongside the authors quoted above, I would like to mention a few exemplary ‘classics’ such as Comrie (1985) and Klein (1994); for a general introduction to tense linguistics cf. Vater (1994) and Bonomi & Zucchi (2001).

42

Sarah Dessì Schmid

surprising that more and more studies are dedicated to the aspectual traits of the imperfect, which of course does not negate its temporal uses.6 From the aspectual point of view, the imperfect is traditionally defined as a fundamentally imperfective form.7 Its imperfective aspectuality is generally divided into three basic uses: the ‘progressive’, the ‘habitual’ and the ‘continuative’. These together are then put in opposition to the Romance perfective forms (the pretérito perfecto simple and pretérito perfecto compuesto in Spanish and the passato remoto and passato prossimo in Italian).8 What the aspectual – imperfective – characterization of the imperfect expresses is semantic indeterminacy: (3)

Il fatto che l’IPF si presenti come un Tempo fondamentalmente imperfettivo signfica, […] che esso si caratterizza per la maniera assolutamente indeterminata attraverso cui il processo viene esibito. ‘The fact that the imperfect presents itself as a fundamentally imperfective tense means […] that it is characterized in an absolutely undetermined manner by means of which process is expressed.’ (Bertinetto 1986: 346)

‘Semantic indeterminacy’ means that there are no cognitive spatiotemporal limitations on the state of affairs expressed in the imperfect, or at least that these limitations are not focused. This absence of any kind of limitation on the one hand makes the process seem non-concluded, and on the other hand makes it appear as the background (on the text level). What is salient, relevant for the here and now of the speech situation and thus in the foreground must display traits of definiteness, of determination9; it generally exhibits a higher degree of integration into the present communication situation and is more strongly emotionally marked. Such definiteness in particular is however absent from states of affairs expressed in the imperfect (cf. also Hopper 1979 on this topic).

6

7 8

9

Bertinetto (1997) and many others emphasize the necessity of integrating both categories – for the verbal system in general, not only for the imperfect – by speaking of a ‘temporal-aspectual system’. I would like to leave aside the imperfetto narrativo in this article. The terminology used in Bertinetto (1986), however, is different from that of the ‘traditional’ grammars: contrary to the “heterogeneous and insufficient” (cf. 1986: 17) traditional tense terminology, he uses the terms perfetto semplice and perfetto composto for the two perfective forms. Such definiteness or determination e. g. pertains to a state of affairs in the spatiotemporal structure of which the beginning (point) and/or end are focused: “uscì di casa alle nove” (He went out at nine), “Marta me odió durante años” (Marta hated me for years), and so on. For a more general analysis on an onomasiological level of aspectuality in Romance languages, cf. Dessì Schmid (in press).

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

43

Examining Old Italian and Old Spanish, one mainly finds aspectual uses of the imperfect, especially in contrast to the simple perfect form; here the imperfect fully shows its imperfective character.10 We are actually looking at a very flexible use of the aspectual opposition between perfective and imperfective, which gives the text profile great plasticity. This is a use which would be considered quite daring today: the linguistic possibilities of aspectual alternation in modern Italian have become much more rigid than in earlier stages of the language, and by the same token have become less flexible in Spanish. However, the greater freedom in the aspectual use of the imperfecto/pretérito perfecto simple pair still constitutes one of the distinctive differences between the Italian and Spanish aspect systems, as Squartini (2004) remarks. In his analysis of the aspectual pair imperfetto/perfetto semplice in Old Italian, Bertinetto (1986: 396) emphasizes the “extraordinary agility” with which these different aspectual perspectives can be combined. He remarks that in the Volgari, the perfetto semplice was always accompanied by the imperfetto in an attributive function, both in copulative as well as in existential uses, with the latter expressing a characteristic trait of an individual or a situation. That means that the Volgari11 distinguished rather systematically between the ‘complexive’ perception of a state of affairs – which in its globality was realised as perfective – and the ‘tangential’ perception, so to speak, of the same, which existed within a precise temporal framework and was thus realised as imperfective. In the following examples given by Bertinetto (1986: 396), the perfetto semplice is used in a ‘presentative’ function, to introduce a new state of affairs into the discourse and mark the temporal distance separating it from the moment of speech, whereas the imperfetto is used strictly to indicate the notion of simultaneity; in particular, we can see on the one hand the perfetto semplice in (4) and (5) in copulative use or function and in (6) in existential function, while on the other hand we find the imperfetto in attributive function in all these examples, where it is used to designate a characteristic trait of an individual:

10

11

For Italian cf. e. g. Ambrosini (1960/61) and (2000), Dauses (1981), Bertinetto (1986), Squartini (2005), Tekavþiü (1972); for Old Spanish e. g. Cano Aguilar (1988), Dauses (1981), Lathrop (2002), Penny (2000) and (2001). A similar option already existed in Latin, cf. James (1982: 379ff.). Concerning Latin syntax and semantics in general cf. Kiss (1982), Kühner & Stegmann (1994 [1912]), and Pinkster (1988).

44

Sarah Dessì Schmid

(4)

Agamenon era [imp.] bello huomo e di bello tempo, molto fiero e molto savio, e ffue [perf. sempl.] il più riccho e poderoso d’avere e d’amici che fosse in tutta la Grecia. ‘Agamenon was a handsome man, in his prime, very proud and very wise, and he was the richest and most powerful in wealth and friends in all of Greece.’ (Istorietta troiana)

(5)

Maestro Antonio da Ferrara fu [perf. sempl.] uno valentissimo uomo quasi poeta, e avea [imp.] dell’uomo di corte; ma molto era [imp.] vizioso e peccatore. ‘Master Antonio da Ferrara was a brilliant man, nearly a poet, possessing the qualities of a man of court; but he was very depraved and sinful.’ (F. Sacchetti, Trecentonovelle)

(6)

Nelle parti di Grecia ebbe [perf. sempl.] un signore che portava [imp.] corona di re e avea [imp.] grande reame. ‘In those parts of Greece, there was a man who carried the royal crown and who had a great empire.’ (Novellino)

Cano Aguilar (1988: 160) allows for a similar conclusion in his examination of the verbal system of castellano medieval and especially of the opposition between pretérito and imperfecto and its suspension. In Spanish too, though to a lesser extent than in Italian, we find firstly that this oppositional pair displays an aspectual flexibility that was once much greater; secondly, that the imperfect is characterised by its fundamental imperfectivity and thirdly, that it is frequently used as “tiempo ‘narrativo’ absoluto” (absolute narrative tense) in the epic language. There are examples in Old Spanish too, which are analogous to the Italian cases dicussed by Bertinetto, and which similarly cannot be used in Modern Spanish: (7)

E sabet que fue [pret.] Gayo omne muy grand de cuerpo, et de color amariello; pero el cuerpo era [imp.] feo e auie [imp.] la ceruiz et las piernas muy delgadas, e las […]. ‘You have to know that Gayo was a man of tall size and of yellow color; but his body was ugly and had a slender neck and very slim legs, and the […].’ (Prim. Crón. Gral. Menéndez Pidal 1977: 116/a/6–9, my emphasis)

The development of the use of the imperfect in the Romance languages clearly demonstrates how thin the line is between flexibility and instability in a language system: Though the flexibility of the aspectual verbal system allowed a particularly lively variation of style on the text level, on the system level it soon proved to be destabilizing (according to Ambrosini (1960/61: 96) and Tekavþiü (1972: 512) already in Old Italian): according to Bertinetto (1986: 389), that is the reason why the system gradually regularized itself, e. g. by generalizing the so-called ‘absolute imperfect’ and drastically reducing the

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

45

freedom of coordinating contrasting aspectual valences within the same sentence. This regularization would also explain the increase in the use of the imperfect, which was far rarer in earlier stages of the language.12 The semantic value of indeterminacy can be found at the centre of the more aspectually oriented interpretations of the imperfect in general. In my opinion, it is in this value that the interpretative key to the so-called ‘modal’ uses of the imperfect lies.

3.

‘Modal uses’ of the imperfect

3.1. A general presentation of the phenomena Part of the “flexibility of meaning” of the imperfect and its “wide variety of use” is its capacity beyond that of other tenses to assume textual meanings that are more modal than temporal: generally, the collective term “modal values” (or “uses”) of the imperfect is employed to designate (8)

una serie de usos del imperfecto etiquetados de diferentes maneras cuyo denominador común es el de la modalización. ‘a series of different uses of the imperfect that are labeled differently and the common denominator of which is that of modalization.’ (García Fernández 2004: 90)

Hence these are phenomena for the description and interpretation of which – according to Bertinetto’s (1986) litotic definition – temporal and aspectual parameters are not sufficient. These various phenomena are often cited but seldom systematically analyzed.13 This is actually difficult to explain when considering that the modal use of the imperfect is documented early on14 in the history of Spanish and Italian –

12 13

14

Cf. Weinrich (62001: 159). On this matter Bazzanella (1990: 447) speaks of “extension of the IPF with past time reference”. Bertinetto’s work, as well as articles and short segments in very few grammars are an exception (cf. particularly Bazzanella 1990, Fernández Ramírez 1986, and García Fernández 2004). For the ‘early usage’ of the imperfect, cf. Ambrosini (2000: 556, but also Ambrosini 1960/61), Cano Aguilar (1988: 262) and D’Achille (1990: 302), who finds the first document (in the apodosis) already in Guido Fava and further documents from 1250 on throughout the entire history of the Italian language.

46

Sarah Dessì Schmid

and of Romance languages in general – and that a continual increase in this use can be observed.15 As in modern Italian, in Old Italian the imperfect could not only be used in an aspectual (imperfective) sense, but also in a modal sense (cf. Squartini 2005: 7). In such cases, the imperfect expressed varying degrees of nonfactuality, which could go as far as counter-factuality. This can clearly be seen in the following example from Squartini, where the forms of the imperfect have a counter-factual value (whereas normally the condizionale passato (past conditional) is used in this context): (9)

Tu non la dovei punire né non convenia (sic) ad te punirla di ciò, ma altre [qualcun’altro] la dovea e potea punire senza tua perversità. ‘You shouldn’t have punished her, and it was not appropriate for you to punish her for that, but somebody else should have punished her without your perversity.’ (Brunetto Latini, Rettorica: 135, 9–11, my emphasis)

We are looking at regular means of expression of the Italian language: multiple examples of a hypothetic imperfect in periodi ipotetici dell’irrealtà (counterfactual conditional sentences) – both in the protasis and the apodosis – can in fact be found throughout the entire history of the Italian language, as the following examples (from Sabatini 1985: 167) show: (10) Braccio cercò di ottenere il regno di Napoli e se non era [imp.] rotto e morto all’Aquila, gli riusciva [imp.]. ‘Braccio tried to gain dominion over Naples, and if he had not failed and died in Aquila, he would have succeeded.’ (Machiavelli, Istorie Fiorentine)

15

Some linguists (Bazzanella 1990, who refers to Sabatini 1985 and Berruto 1987, as well as Mazzoleni 1992) say that this “expansion” of the imperfetto indicativo suggests a process of simplification in the Italian verbal system (similar propositions are maintained for Spanish). The argument of simplification however can prove to be difficult in a theory of language change: first of all it targets the language as a system (more than the ‘local’ communicative success of the speaker) and thus implies – or at least can imply – the rather problematic assumption that the speakers wish to change the language; secondly one has to ask more specifically: what is it that is simplified? Does this ‘replacement’ of the subjunctive mood (or the conditional respectively) by the imperfect of the indicative in the analyzed contexts really mean a simplification of the Italian language system in general? The role of the subjunctive (as of the conditional) in the Italian modal system is not actually jeopardized in doing so.

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

47

(11) Se Lucia non faceva [imp.] quel segno, la risposta sarebbe probabilmente stata diversa.16 ‘If Lucia had not made this sign, probably the answer would have been another.’ (Manzoni, Promessi Sposi, cap. 3)

Old Italian also already used the imperfect to describe states of affairs that are non-factual, either by their intrinsic nature (e. g. dreams) or as a result of what traditional aspect research considers to be the problematic intersection between aspect and Aktionsart (which is the case e. g. for the imperfetto di conato (imperfect of attempt)): (12) Un’interpretazione modale può anche sorgere dall’incrocio tra aspetto progressivo e predicati telici (25). In questi casi l’imperfetto esprime il mancato raggiungimento del fine intrinseco nei predicati telici sottolineando così il carattere non fattuale della situazione. […] (25) Levossi questa femina et aiutollo che periva in una vile [piccola] fossatella d’acqua per poca e per cattiva provedenza [sprovvedutezza]. (Novellino, 38, 19f.) ‘A modal interpretation can also arise from crossing the progressive aspect with telic predicates (25). In these cases, the imperfect expresses the failure to reach the end inherent to the telic predicates underlining the non-factual character of the situation. […] (25) This woman got up and helped him as he was about to die in a small puddle of water due to his lack of prudence.’ (Squartini 2005: 8)

Something similar can be observed in Spanish: the use of the imperfect in the apodosis of the condicionales in the earlier stages of the language too clearly reveals modal qualities, even if this breach of the structure of conditional sentences happens only occasionally and creates mixed constructions (with the protasis in subjuntivo and the apodosis in indicativo) (cf. Marcos Marín 1980: 408). Cano Aguilar (1988: 160), from whom I quote the following example, also puts forward multiple examples demonstrating the modal use of the imperfect already in Medieval Castilian; contexts of politeness and conditional structures are notably concerned: (13) Yo que esto vos gané bien mereçía [imp.] calças. ‘I who acquired this for you might indeed deserve shoes.’ (14) Si yo loca non fuesse, non te deuia [imp.] amar. ‘If I wasn’t insane, I shouldn’t love you.’

16

(PCid) (Prim. Crón. Gral.)

The use of the hypothetic imperfect also emerges in Old French (competing with the double subjunctive imperfect) and gradually takes hold to the point that in Modern French, it is the normal structure, apart from purely archaic vestiges.

48

Sarah Dessì Schmid

In Old Spanish we can also find examples of counterfactual conditional use, as example (14) shows, but non-factual conditional constructions are more typical (cf. Cano Aguilar, 1988: 262). We will see that – in contrast to Italian – in Spanish, and particularly Modern Spanish, it is more a matter of condicionales potenciales (potential, non-factual conditional sentences) than condicionales irreales (counterfactual conditional sentences), which unambiguously depends on whether the protasis employs the imperfect of the indicative or not. I will return to this point later. 3.2. On the classification of the ‘modal uses’ of the imperfect There is no unanimous opinion concerning the designation of the various phenomena subsumed by the common label “modal uses (or values) of the imperfect”; there is also no generally accepted classification. I will now display the most extensive classifications in table form for comparison17 (for Italian Bazzanella 199018, for Spanish García Fernández 200419), and in doing so I will demonstrate that the seemingly manifold differences between the two languages can be reduced to one: the use of the hypothetic imperfect, which is already well documented in the earlier stages of the language – as we have just seen – and on which I will then concentrate.

17 18

19

Needless to say, the juxtaposed categories are not always entirely equivalent. Bazzanella’s classification – which in turn refers to Bertinetto (1986) and Conte (1984) – distinguishes between modal uses of the imperfect “with past time reference” and uses “with non-past time reference”. I do not make this distinction, as it is not relevant for this study; I have also modified certain elements of the classification and put forward my own examples where there is no further indication. García Fernández’ classification – which in turn builds upon that of Fernández Ramirez (1986: 4, 269ff.) – is rendered here including examples (the listing of types and the numeration of the examples are adapted to this text as above).

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

49

Table 1: ‘Modal uses’ of the imperfect: two standard classifications.

Bazzanella (1990)

García Fernández (2004)

a) Imperfetto onirico or imperfetto fantastico, used to recount a dream or to identify a fantasy world: (15) Ieri notte ho sognato cose strane. Camminavo per una strada piena di fiori rossi, mentre ad un tratto incontravo Giulia che mi chiedeva di raccoglierne uno per lei… [Yesterday night I dreamt strange things. I walked on a road full of red flowers, when all of a sudden I met Giulia, who asked me to pick one for her…]

b) Imperfetto di conato or imminenziale, used with reference to an attempted or impending action: (16) Mi vedi per caso perché già me ne andavo a casa. [You see me by chance, because I was just about to go home.]

c) Imperfetto ipotetico, used in counter-factual contexts (also in ‘periodi ipotetici dell’irrealtà’): (17) Senza quell’incidente arrivavamo da Giulia in tempo. [If that incident hadn’t happened we would have arrived at Giulia’s house in time.] (18) Se lo sapevo te lo dicevo. [If I had known, I would have told you.]

d) Imperfetto ludico, used by children in order to explain the distribution of roles in a game or in order to identify a fantasy world: (19) Io ero il capitano e tu il marinaio. [Say I am/was the captain and you the sailor.]

i) Imperfecto lúdico (24) Oye, ¡qué divertido! Tú eres el que iba remando, la mar estaba muy revuelta. (R. Sánchez Ferlosio, El Jarama) [What fun! Say you’re the one who goes rowing, and say the sea was very troubled.]

50

e) Imperfetto di cortesia or attenuativo or di modestia, used in assertions and questions expressing a desire in the context of a strategy of courtesy: (20) A. Cosa desiderava? B. Volevo parlarle solo cinque minuti. Venivo a chiederle un favore. [A. What did you want / What would you like? B. I only wanted to talk to you for five minutes. I came to ask you a favor.]

f) Imperfetto epistemico-doxastico, used – mainly in spoken Italian – to illustrate a future action and allude to one’s preceding knowledge thereof: (21) Chi c’era stasera alla festa di Leo? [Who was coming to Leo’s party (again)?]

g) Imperfetto potenziale, used to express assumptions: (22) Carlo doveva arrivare alle tre; non so cosa sia successo. [Carlo should have arrived at three, I don’t know what happened.]

Sarah Dessì Schmid j) Imperfecto de cortesía (25) ¿y usted qué deseaba joven? […] Pues servidor venía, porque […] (Arniches, La gentuza) [And you, young fellow, what did you want? […] Then the servant came, because […]] k) Imperfecto desiderativo de ocurrencia or inspiración (26) De qué buena gana me bebía un vaso, con este calor. (Arniches, La pobre niña) [I would really love to drink a glass of water in this heat.]

l) Imperfecto en oraciones exclamativas (27) ¡Ya lo decía yo! ¡Sí tenía que ser! (Pardo Bazán, Dulce sueño) [I already said it! It had to be!] m) Imperfecto de exclamación or de excusa (28) ¡Ay, sí, hijo, que no me acordaba! Ya sabemos que estáis de enhorabuena. (Arniches, La flor del barrio) [Oh yes, my son, I didn’t remember! Indeed we know that you have reason to be happy.] n) Imperfecto de sorpresa (29) ¡Ah! ¿Pero estudiaba usted para sacerdote? (Arniches, La pobre niña) [Ah! But weren’t you studying to be a priest?]

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

51

h) Imperfetto di pianificazione, used – in spoken Italian – to express a future action that is planned but still up for discussion: (23) A. Non puoi farlo domani? B. Domani andavo in biblioteca. (in Bazzanella 1990: 446) [A. Can’t you do it tomorrow? B. I was going to go to the library tomorrow.]

When comparing the two classifications, one may at first criticize a certain lack of abstraction or generalization primarily – but not solely – in the second:20 The item n) imperfecto de sorpresa as well as the items l) en oraciones exclamativas (in exclamative sentences) and m) de exclamación (of exclamation) or de escusa (of excuse) could indeed be ascribed to epistemic (or potential) uses; 21 items j) de cortesía and k) desiderativo de ocurrencia are closely connected as well.22 Both classifications feature the ludic imperfect and that of courtesy, which are highly conventionalized phenomena in both languages, even if they are not explained in the same way. A striking difference is the lack of the imperfetto onirico, the imperfetto di conato and the imperfetto ipotetico in the Spanish classification, though examples can easily be found. However, the analysis of the imperfect in Fernández Ramírez’ Gramática, on which that of García Fernández (2004: 90ff.) is based, exhibits distinctive and interesting differences compared to this one: on the one hand, the listed uses of the imperfect are not only not marked as ‘atypical’, but not even specifically as ‘modal’;23 on the other hand, it contains those important categories whose absence was striking in comparison to the Italian listing. These are the imperfecto de conato and the imperfecto hipotético, although the latter has a different name and indeed functions somewhat differently to its

20 21

22

23

For an extremely detailed subdivision of types of modal uses of the imperfect, cf. Mazzolena (1992: 180). In the Italian classification, these are listed under types f) and g). Various examples can be found in Spanish. Reyes (1990: 17) gives an interesting explanation for this, compatible with that of Bazzanella. However, in this respect, one would also have to criticize the striking absence of category k) or the more general conception of e) Imperfetto di cortesia or attenuativo or di modestia as ‘wish category’ in the Italian classification. They are simply considered as examples of the “valores generales del imperfecto”, though it must be conceded that the classification seems neither very systematic nor abstracting. The imperfect is defined as categorically imperfective – like in Bertinetto – and thus as strongly polysemous. Cf. Fernández Ramirez (1986: 269ff.).

52

Sarah Dessì Schmid

Italian equivalent, namely non-counterfactually, as was already the case in earlier stages of the language: (30) El ‘imperfecto de figuración’ es un imperfecto irreal cercano en parte al desiderativo […]. Se encuentra en la apódosis de las oraciones condicionales si el tiempo de la prótasis es un imperfecto o pluscuamperfecto de subjuntivo: ‘Si hubiese otra taberna abierta a estas horas, te podías despedir de mí como cliente.’ (I. Aldecoa, En el kilómetro 400, en Libro de lecturas, 37) […] ‘Si yo pudiera, mocito, este trato se cerraba.’ (García Lorca, Poesía Española. Antología por Gerardo Diego, 313) ‘The ‘imperfecto de figuración’ is an imperfect expressing the irrealis and in part near to its desiderative function […]. It appears in the apodosis of conditional clauses if the tense of the protasis is an imperfect or the past perfect of the subjunctive: ‘If there was another open tavern at this late hour, you could bid farewell to me as a client.’ […] ‘If I could, fellow, this deal would be closed.’’ (Fernández Ramírez 1986: 276)

The presence of this so-called imperfecto de figuración (imperfect of figuration) proves to be particularly interesting for our discussion, and its scarcely transparent appellation adds a further challenging piece to our puzzle.24 A consultation of electronic corpora and of native speakers does indeed seem to support this analysis of the classifications: most notably in spoken Spanish, the indicative imperfect can still be found in the apodosis of (potential) conditional sentences, even if this structure (si + imperfecto de subjuntivo + imperfecto de indicativo) is rarer than the standard structure (si + imperfecto de subjuntivo + condicional) and is perceived as strongly marked as language of immediacy25 (nähesprachlich):

24

25

In the Gramática of Alarcos Llorach (101999: 160ff., in particular 163–4) we also find a list of examples of potential-epistemic uses of the imperfect as well as those “de conato”, “de cortesía, precaución o deferencia” (of politeness, precaution or compliancy) and hypothetic use of the imperfect, which has no special designation at all: “De modo análogo se emplea cantabas en lugar del pospretérito en la apódosis de las construcciones condicionales (§ 449): Me daba por contento, si apruebas en setiembre (en lugar de Me daría por contento, si aprobases en setiembre)” (Alarcos Llorach 101999: 164) [Accordingly, cantabas is employed instead of the pospretérito in the apodosis of conditional constructions (§ 449): I would be content if you finished in September (instead of the normative structure conditional-subjunctive)]. Concerning the theory of communicative immediacy and distance, cf. Koch & Oesterreicher (1990) and (2007).

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

53

(31) b. Si mi hermana llegara, se metiera en su habitación, se olvidara de todo […] yo llegara, me metiera en mi habitación y me olvidara de todo, ¡se me venía todo abajo!, de verdad. (Esgueva y Cantarero, 1981: 365) La aparición del imperfecto del indicativo […] en la apódosis de este esquema constituye un uso no normativo, pero se encuentra notablemente extendido en registros coloquiales del español, de modo que podría considerarse un esquema propio del estándar informal. ‘b. If my sister came, she would lock herself in her room, she would forget about everything […] I would come home, I would lock myself in my room and I would forget about everything, everything would really come down on me. The appearance of the imperfect of the indicative […] in the apodosis of the structure represents a non-normative use, but it is significantly extended in colloquial registers of the Spanish language, so that it could be considered a pattern of the informal standard.’ (Montolío 1999: 3669)

The frequent appearance of the imperfect indicative in the apodosis of (potential) conditional sentences, to the point of becoming the second most common pattern for expressing potentiality, constitutes a first point of divergence from Italian, where indicative forms prevail in the spoken language.26 Concerning the general use of this structure, Montolio (cf. 1999: 3669) adds that, although various grammarians explain the appearance of the imperfecto in this context from different theoretical perspectives, all agree in accepting that the indicative form is more assertive and implies a greater commitment of the speaker to the realization of the apodosis in the case of accomplishment of the protasis (as we see in the example b. quoted above, hence also the comment de verdad). In her opinion it does not seem that this pattern replaces the standard one (imperfectocondicional), but that the speaker treats it as an additional syntactic variation available for expressing the belief in a greater probability of fulfilling the proposition in the apodosis if the protasis should be realized. Let us recapitulate the main differences in the use of the hypothetic imperfect in the two languages: i. While in Old and Modern Italian the use of the hypothetic imperfect in conditional sentences and contexts expresses only counter-factual interpretations (Old Italian: examples (9), (10) and (11); Modern Italian: examples (17) and (18)), in Spanish (both Old and Modern) it functions primarily in non-factual states of affairs as ‘condicionales potenciales’. More drastically, the use of the imperfect can be considered as a syntactic alternative to the standard form of the conditional, thus expressing a higher 26

For Italian, cf. the by now ‘classic’ article of Sabatini (1985). However, the abundance of other structures in the dialects must also not be forgotten.

54

Sarah Dessì Schmid

degree of probability (see ex. b. in quotation 31). This is certainly related to the second difference: ii. While in Old and Modern Italian the imperfect can appear both in the protasis and in the apodosis (and thus can express both the conditional and the subjunctive moods; ex. (10), (11), (18)), it appears only in the apodosis of Old and Modern Spanish (ex. (14) and ex. in quotation 30). The presence of the imperfect (indicative) instead of the subjunctive in the protasis of Italian conditional sentences is what causes the increase in the degree of irreality (to the point of counter-factuality) in the meaning.27 The greater reluctance to use the indicative instead of the subjunctive in Spanish is also related to the difficult history of the alternation of -ra and -se in Spanish ‘oraciones condicionales’28 and to the differences in the emergence and development of the Italian and Spanish modal systems.29 This difference in the factual value of conditional sentences in the two analyzed languages is confirmed by the study of possible combinations of opposition between imperfetto dell’indicativo/condizionale/congiuntivo (imperfect indicative – conditional/subjunctive) in conditional contexts in standard and colloquial Italian and Spanish. The analysis of certain phenomena of the substandard varieties, however, which shouldn’t even exist in modern Spanish according to the classifications seen above, does not offer similar confirmation: there is in fact also a substandard option of expressing the potentiality of the protasis through an imperfect of the indicative, which – according to Montolio – is a widely-used construction among groups of low sociolinguistic status: (32) El hecho de que si tenía comience en algunos registros de la lengua actual a sustituir a si tuviera parece corroborar la hipótesis que interpreta estos procesos de sustitución como casos de conmutación de categoría modal, de indicativo por subjuntivo […], y no de la temporal, ya que no parece tener mucho sentido colocar una forma dislocada (un pasado que pasa a expresar futuridad: tenía) en lugar de otra que ya lo estaba (el imperfecto de subjuntivo estuviera), sobre todo si se considera que con ambas formas de pasado se logra idéntico efecto significativo, la expresión de no realidad, como muestran los ejemplos: Si me ocurría, no me daba por vencida; Si me decían eso, me aguantaba.

27 28 29

Concerning this matter, remember that the unreal imperfect (only!) in the protasis of the French conditional sentence structure is taken for granted. Literature on this subject is quite extensive, cf. e. g. Marcos Marín (1980: 402–416), Veiga (1996), Becker (2008). On the development of the French, Italian, and Spanish modal systems cf. Becker (2006).

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

55

‘The fact that si tenía starts to replace the form si tuviera in some registers of current language seems to prove the hypothesis interpreting these substitution processes as cases of modal category switching from indicative to subjunctive […], not temporal category switching, because it does not seem to make much sense to employ a dislocated form (i. e. a past tense expressing future – tenía) instead of another that already expressed this meaning (the imperfecto de subjuntivo – estuviera), especially if we consider that with both past tense forms the same meaning is achieved, namely the expression of unreality, as the examples show: If it happened to me, I would not believe it. If they told me that, I would have to restrain myself.’ (Montolío 1999: 3669)

The first ‘cracks’ in the compact brickwork of the conjunctive protasis thus seem to exist in Spanish too.30 Nonetheless, a distinct difference in the number of these ‘cracks’ and their collocation in linguistic varieties undoubtedly remains. It is imperative to stress the following to counter against any too strongly norm-oriented opinion, such as one often finds in traditional grammaticography: if an innovation takes place – in the form of the emergence of a new fact of speech – and this fact of speech is effectively adopted, the variety in which this fact of speech originated is cognitively of little relevance. Keeping that in mind, even phenomena of the lowest substandard that have reached the point of adoption/diffusion must be taken seriously in linguistic analysis and cannot be treated as irrelevant for linguistic consideration and interpretation only because they are not phenomena of the ‘standard’. The following table attempts to give a complete comparative view – taking into account the substandard phenomena too – of the main differences in the use of the hypothetic imperfect in the two languages: Table 2: ‘Hypothetic’ imperfect in Italian and Spanish. IMPF

in the PROTASIS

IMPF

in the APODOSIS

NON-FACTUAL

Old and Mod. Spanish (ex. in quotation 30) Mod. Span. colloquial (ex. b. in quotation 31)

30

Cf. the radical position of Pérez Saldanya (2004) on this subject and the moderate concession of Fernández Ramírez (1986: 276).

56

Sarah Dessì Schmid

COUNTER-FACTUAL

Old Italian (ex. (10), (11))

Old Italian (ex. (9))

Mod. Italian colloquial (ex. (18))

Mod. Italian colloquial (ex. (17), (18)) Old Spanish (ex. (14))

Mod. Spanish substandard (ex. in quotation 32)

Mod. Spanish substandard in combination with imperfect in the apodosis (ex. in quotation 32)

This leads me to an explanation of the modal uses of the imperfect and subsequently to an illustration of the semantic mechanisms which permit them and the pragmatic forces which direct them. Before reaching this final point, I must however briefly cite a possible objection to the thematic limitation of this article, which could also pose a theoretical objection: Why is this paper concerned only with the imperfect if other tenses may also assume textual meanings exhibiting a more modal than temporal character? 3.3. ‘Modalization’ of the imperfect or of non-present tenses in general? It has frequently been observed in linguistic research31 that non-present tenses in general – and most notably past tenses – are particularly eligible for modal use: (33) There is clearly a universal semantic link between the notion of past tense and the notion of remoteness from reality. (James 1982: 396)

This is said to be the case because the deictic distance from the here and now of the moment of speech is interpreted as modal and epistemic distance from reality and topicality; a reading which interprets this “semantic link” as metaphorical. Detges’ research on grammaticalization offers a different and

31

Cf. e. g. Coseriu (1976), Reyes (1990), King (1992), Fleischmann (1982), (1989) and (1995), Detges (2001), Comrie (1976) and (1986), Ultan (1978).

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

57

more persuasive explanation of this association:32 non-present tenses seem to be particularly likely to construct modal implicatures that strongly mitigate the assertive validity of present events (cf. Detges 2001: 142). The modal uses of the imperfect analyzed above are a perfect example of this. However, similar cases of modalization can also be found for the future33 and pluperfect tenses:34 (34) Serán las tres de la tarde/madrugada. ‘It will be around three a.m./p.m.’ (35) Senza quell’incidente eravamo già arrivati da Leo. ‘If that incident hadn’t happened we would have arrived at Leo’s house in time.’

Detges interprets this procedure, this semantic link, as metonymy. According to him, the semantic foundation for these implicatures is: (36) das Weltwissen, daß Sachverhalte der Vergangenheit, die sich unter Umständen bis in die Gegenwart hinein fortsetzen, dort eben nicht mit letzter Sicherheit bzw. nur unter bestimmten Bedingungen der Fall sind. ‘the encyclopedic knowledge that states of affairs in the past, which possibly can continue up to the present, are then not absolutely certain or rather they are only the case under certain circumstances.’ (Detges 2001: 142)

If means of encoding certain temporal categories frequently also take on modal functions cross-linguistically, we are usually dealing with grammatical polysemy, as there is a synchronically comprehensible associative relation between the temporal and modal functions, which is due to metonymy (cf. Detges 2001: 144). Detges’ theory, however, also deals mainly with the

32 33

34

Cf. Detges (2001) also for a precise critique of grammar development hypotheses based on metaphor or ‘bleaching’. Detges adds that what has been said of the past tenses is similarly valid for the future tense, as states of affairs taking place in the future exist only in the expectation of the speaker and never possess factual character: „Aus diesem Grund ist es wahrscheinlich, daß die Sprecher vieler Sprachen die grammatische Kategorie zur Bezeichnung der Zukunft nicht nur dazu verwenden, das zukünftige Eintreten eines Sachverhaltes zu asserieren, sondern auch, um zu implikatieren, daß sie das gegenwärtige Zutreffen eines Sachverhaltes vermuten.” (Detges 2001: 142) [For this reason, it is probable that the speakers of many languages not only use the grammatical category for indicating the future in order to assert the future taking place of a state of affairs, but also to implicate that they assume the present taking place of a state of affairs.] Frequently, such implicatures have ‘merely’ conversational status. The Romance languages however exhibit several conventionalized cases of this phenomenon.

58

Sarah Dessì Schmid

relation between temporal and modal (or epistemic) meanings of non-present verbal forms. So I return to the question: Why are we analyzing only the imperfect tense here? If we examine ‘phenomena of modalization’ in all non-present tenses more closely, the following points are indeed striking: i.

From a merely quantitative point of view, most cases of modal usage do indeed resort to the imperfect tense, i. e. the imperfective tense par excellence. In contrast, perfective tenses (which are also past and thus nonpresent tenses) are hardly ever employed with a modal meaning.

ii. The modal uses of the pluperfect are exactly parallel to certain typical applications of the imperfect and can thus often be explained in the same way (cf. Bertinetto 1986 and also Montolío 1999, who in turn quotes Klein-Andreu 1986). iii. The use of future tenses is, particularly in Italian, mainly limited to the epistemic value that is primarily found in connection with stative verbs. This moreover reveals an indirect connection to the imperfective aspect, for “statistically (and semantically), the imperfective aspect correlates with the nature of ‘states’” (Berretta 1993: 212, my transl.), and the imperfect appears particularly frequently with such verbs as essere, avere, sapere etc.35 How can it be explained that compared to other non-present tenses, the imperfect seems to play a different and considerably more decisive role in this process? If it is really to be assumed that the imperfect presents a great flexibility of meaning and a wide variety of use, the reasons behind this must also be closely examined. 3.4. Semantic motivations for ‘modalization’: a metonymical bridge My thesis is that the gap between non-present tenses and their modal use is bridged by aspectual (and particularly imperfective) components more than by temporal ones.36 This would adequately explain why the imperfect – as the 35

36

This can at least be said of Italian, given that Spanish (unlike Italian) in some cases accepts the combination of stative verbs and the perfective form, as Squartini (2004) demonstrates. Or in the case of the future tense, ‘indirectly’ by means of the relation between imperfectivity and stativity. By aspectuality I here mean the function by which the type of progression and the distribution of a state of affairs in time is categorized. Aspectuality ‘means’ that complex of information pertaining to the

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

59

imperfective tense per antonomasiam – seems so exceedingly suited for such use. If we take a closer look at Detges’ theory in the passage quoted above on the semantic foundation of implicatures leading to modalization, it seems to be clear that the semantically relevant issue motivating the implicature is not that we are dealing with events in the past in general, but that the events are presented as aspectually indetermined: they display traits of indeterminacy and inconclusion (“which possibly can continue up to the present”) and are considered to be backgrounded. Their beginning and end cannot be asserted (since the event has been expressed as indetermined) but only supposed. The imperfective aspect expresses background events characterized precisely by a “lower degree of assertiveness” (Hopper 1979: 216). The opinion that the interpretative key to modalization processes may lay in the imperfective aspect is one I share with several scholars, who underline the cross-linguistic tendency to associate specifically the imperfective aspect in the past with irreality.37 I would like to further pursue this intuition on the basis of the examination of the ‘modal uses’ of the imperfect.38 If one explains the definitions of “modalized imperfects” analyzed above precisely, i. e. if one approaches the question of what “common denominator of modalization” (García Fernández 2004: 90, my transl.) means, one could agree with Bazzanella (1990: 455)39 that the various uses of the imperfect can be given a common designation, as they have primarily two things in common: i. semantic indeterminacy (namely of an aspectual nature); ii. the fact that they indicate a transfer from the real world to one created by the speaker. I however want to explicitly ask the question: How are these two elements connected? What relation in the mind of the speaker associates the aspectual

37 38

39

particular temporal structure of a state of affairs, i. e. the possibility of its ‘determination’. Cf. e. g. Hopper (1979), James (1982), Bertinetto (1986), Bazzanella (1990) and Mazzoleni (1992). On this matter, cf. Bazzanella, in particular: “Since the imperfective aspect does not allow the speaker either to define the event temporally or to define the conclusion, indeterminacy will result as the main feature (see Bertinetto (1996: 158)), and therefore motivate the ‘modal’ uses of IPF” (Bazzanella 1990: 448). Cf. also Bertinetto (1986).

60

Sarah Dessì Schmid

semantic value of indeterminacy with that of modality? How does this ‘transfer to another world’ come about and how does it take place?40 In using the imperfect, the speaker makes a very economic linguistic decision. We are indeed dealing with a form which is extremely open to the construction of modal implicatures leading to the mitigation of the assertive validity of present states of affairs. Let us analyze the steps of this ‘transfer’ in detail: states of affairs presented in the imperfect tense usually display an imperfective aspectual character. The result is that in the presentation of their temporal structure, neither beginning nor end is focused. It is in this fundamental and aspectual semantic characteristic of imperfect forms that we find the basis for these modal implicatures: they consist in the speaker’s encyclopedic knowledge that indetermined (non-concluded and backgrounded) states of affairs have a low degree of assertiveness, i. e. in the moment of speech they are not absolutely certain, or rather they are only the case under certain circumstances. The semantic relation behind this ‘transfer’ therefore connects elements situated within the same frame and not – as it would be in the case of a metaphorical relation based on similarity/contrast – in two different frames: due to his experience in and concept of the world, the speaker is aware that inconclusion, background function and uncertain incidence are closely conceptually linked in the frame of indetermined states of affairs. Thus, it is by means of the semantic mechanism of metonymy (as a figure-ground effect within a frame as well as between a frame and its components (Koch 1996: 40)), based on the association principle of contiguity, that the constitutive elements of one frame (expressed in the imperfect) are held together, and one or the other is brought out as the figure (in contrast to the ground). If the irrealis modality (the marking of a non-real context on the part of the speaker) is brought out as the figure and the other designats remain in the background in the imperfect form, we refer to this as modal use of the imperfect. There is a comprehensible associative relation between the elements common to the modal uses of the imperfect, between indeterminacy and irreality (of the world created by the speaker), thus between the aspectual and 40

The approach presented here is based on Detges (2001) as regards the interpretation of the cognitive mechanisms behind the modal uses of the imperfect (these mechanisms being of metonymic and not metaphorical nature), but differs from his account concerning the main element which establishes the link to modality (this element being of aspectual and not temporal nature). So I interpret the bridge between non-present tenses and modality to be metonymic as does Detges, but in contrast to him believe that bridge to be constituted by the aspectual characteristics of the verb.

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

61

modal meanings of the imperfect. We are therefore dealing with grammatical/functional polysemy41, which came about to enable economical and efficient communication and leads to communicative success by the means of (at first conversational and then in some cases conventionalized) implicatures. The conceptual path leading to modality is paved aspectually, i. e. imperfectively.

4. Conclusion In dealing with modal uses of the Italian and the Spanish imperfect and by comparing these with its temporal and aspectual uses, this paper has aimed to portray the imperfect as a form particularly likely to be employed in constructing modal implicatures which mitigate the assertive validity of present states of affairs, and to show that the modal uses of the imperfect are an ideal example of this. In explaining why the imperfect – as the imperfective tense par excellence – seems so exceedingly suited for such modal uses, I have interpreted the semantic link between the different meanings/uses of the imperfect as a metonymical one and shown how this metonymical bridge to modality is determined by aspectual (imperfective) meaning components more than by temporal ones. Let us now recapitulate the following facts: i. the ‘modal’ functions, as well as all the others traditionally ascribed to the imperfect, are closely related to its aspectual – imperfective/indetermined – meaning components (and thus are one of the functional meanings of the polysemous imperfect); ii. they appear very early on in Spanish and Italian documents and are present throughout the history of these languages; iii. the frequency of their use in the two languages is continuously increasing – albeit to different extents and with greater or lesser diastratic markedness. Bearing these facts in mind, one need not hesitate to classify these modal functions as a further common basic use of the imperfect. This use can be

41

For such a definition of polysemy, and concerning the role of metonymy as a semantic mechanism, cf. Koch (1996), (1999), (2001), (2004), Blank (1997) and (2001).

62

Sarah Dessì Schmid

considered equally ‘typical’ and ‘basic’ as the others and more than just a common label. Thus, we have a further function to add to the four functions generally attributed to the imperfect discussed at the beginning of this paper (in section 2): the function of expressing a hypothetical world that can display varying degrees of reality up to the point of counter-factuality. This is thus one more of the flexible and manifold faculties of the imperfect, for: (1)

Tra i Tempi dell’Indicativo, l’Imperfetto […] è certo quello che presenta la maggior duttilità di significato e la più ampia varietà d’impiego. ‘Amongst the indicative tenses, the imperfect […] is certainly the one that presents the greatest flexibility of meaning and the widest variety of use.’ (Bertinetto 1986: 345)

References Alarcos Llorrach, Emilio (101999): Gramática de la lengua española. – Madrid: Espasa Calpe (Real Academia Española, colección Nebrija y Bello). Ambrosini, Riccardo (1960/61): “L’uso dei tempi storici nell’italiano antico.” – In: L’Italia dialettale 24 (1), 13–124. – (2000): “Sulla sintassi del verbo nella prosa toscana del Dugento – ovvero – Tempo e aspetto nell’italiano antico.” – In: Lingua e Stile 35 (4), 547–571. Bazzanella, Carla (1990): “‘Modal’ uses of the Italian Indicativo Imperfetto in a pragmatic perspective.” – In: Journal of Pragmatics 14, 439–457. Becker, Martin (2006): Welten in Sprache – zur Entwicklung der Kategorie ‘Modus’ in romanischen Sprachen. – Professorial Diss., Universität Stuttgart. – (2008): “From temporal to modal: divergent fates of the Latin synthetic pluperfect in Spanish and Portuguese.” – In: Ulrich Detges & Richard Waltereit (eds.): The Paradox of Grammatical Change. Perspectives from Romance, 147–179. Amsterdam/Philadelphia: Benjamins. Bello, Andrés (1841/1951): “Análisis ideológica de los tiempos de la conjugación castellana.” – In: Obras Completas: Estudios gramaticales, 1–67. Caracas: Ministerio de Educación. – (1847/1981): Gramática de la lengua castellana destinada al uso de los americanos (ed. by Ramón Trujillo). – Santa Cruz de Tenerife: Instituto Universitario de Lingüística Andrés Bello. Berretta, Monica (1993): “Morfologia.” – In: Alberto A. Sobrero (ed.) (41999a): Introduzione all’italiano contemporaneo. Vol. 2, 193–245. Roma/Bari: Laterza. Berruto, Gaetano (1987): Sociolinguistica dell’italiano contemporaneo. – Roma: La Nuova Italia Scientifica.

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

63

Bertinetto, Pier Marco (1986): Tempo, aspetto e azione nel verbo italiano. Il sistema dell’indicativo. – Firenze: Accademia della Crusca. – (1997): Il dominio tempo-aspettuale: demarcazioni, intersezioni, contrasti. – Torino: Rosenberg & Sellier. Blank, Andreas (1997): Prinzipien des lexikalischen Bedeutungswandels am Beispiel der romanischen Sprachen. – Tübingen: Niemeyer. – (2001): Einführung in die lexikalische Semantik für Romanisten. – Tübingen: Niemeyer. Bonomi, Andrea & Zucchi, Alessandro (2001): Tempo e linguaggio. Introduzione alla semantica del tempo e dell’aspetto verbale. – Milano: Mondadori. Bosque Muñoz, Ignacio & Demonte Barreto, Violeta (eds.) (1999): Gramática descriptiva de la lengua española. Vol. 2–3. – Madrid: Espasa Calpe (Real Academia Española, colección Nebrija y Bello). Bull, William E. (1960): Time, Tense and the Verb: a Study in Theoretical and Applied Linguistics, with particular Attention to Spanish. – Berkeley/Los Angeles: University of California Press. Bybee, Joan L. & Dahl, Östen (1989): “The creation of tense and aspect systems in the languages of the world.” – In: Studies in Language 13 (1), 51–103. Cano Aguilar, Rafael (1988): El español a través de los tiempos. – Madrid: Arco Libros. Comrie, Bernard (1976): Aspect. – Cambridge: Cambridge University Press. – (1985): Tense. – Cambridge: Cambridge University Press. – (1986): “Conditionals: A Typology.” – In: Elizabeth C. Traugott et al. (eds.): On Conditionals, 77–99. Cambridge: Cambridge University Press. Conte, Maria-Elisabeth (1984): “Deixis am Phantasma. Una forma di riferimento nei testi.” – In: Lorenzo Coveri (ed.): Linguistica testuale, 187–203. Roma: Bulzoni. Coseriu, Eugenio (1976): Das romanische Verbalsystem (ed. by Hansbert Bertsch). – Tübingen: Narr. D’Achille, Paolo (1990): Sintassi del parlato e tradizione scritta della lingua italiana: analisi di testi dalle origini al secolo XVIII. – Roma: Bonacci. Dauses, August (1981): Das Imperfekt in den romanischen Sprachen. Seine Bedeutung im Verhältnis zum Perfekt. – Wiesbaden: Steiner. Dessì Schmid, Sarah (in press): „Inquietudine terminologica e categoriale: per un approccio onomasiologico al sistema aspettuale dell’italiano“ – In: Maria Selig & Gerhard Bernhard (eds.): Akten des Marburger Italianistentags 2008. Bern/Berlin/Bruxelles/Frankfurt a. M.: Peter Lang. Detges, Ulrich (2001): Grammatikalisierung. Eine kognitiv-pragmatische Theorie dargestellt am Beispiel romanischer und anderer Sprachen. – Professorial Diss., Universität Tübingen. Fernández Ramírez, Salvador (1986): Gramática española. Vol. 4: El verbo y la oración. – Madrid: Arco Libros. Fleischmann, Suzanne (1982): The Future in Thought and Language: Diachronic evidence from Romance. – Cambridge: Cambridge University Press. – (1989): “Temporal Distance: a basic linguistic Metaphor.” – In: Studies in Language 13, 1–50.

64

Sarah Dessì Schmid

– (1995): “Imperfective and Irrealis.” – In: Joan Bybee & Suzanne Fleischmann (eds.): Modality in Grammar and Discourse, 519–551. Amsterdam/Philadelphia: Benjamins. García Fernández, Luis (2004): “El préterito imperfecto: repaso histórico y bibliográfico.” – In: Luis García Fernández & Bruno Camus Bergareche (eds.): El pretérito imperfecto, 13–95. Madrid: Gredos. García Fernández, Luis & Camus Bergareche, Bruno (eds.) (2004): El pretérito imperfecto. – Madrid: Gredos. Hopper, Paul J. (1979): “Some observations on the typology of focus and aspect in narrative language.” – In: Studies in Language 3 (1), 37–64. James, Deborah (1982): “Past Tense and the Hypothetical: A Cross-Linguistic Study.” – In: Studies in Language 6 (3), 375–403. King, Larry D. (1992): The Semantic Structure of Spanish: meaning and grammatical form. – Amsterdam: Benjamins. Kiss, Sándor (1982): Tendances évolutives de la syntaxe verbale en latin tardif. – Debrecen: Kossuth Lajos Tudományegyetem. Klein, Wolfgang (1994): Time in language. – London: Routledge. Klein-Andreu, Flora (1986): “Speaker-Based and Reference-Based Factors in Language: Non-Past Conditionals Sentences in Spanish.” – In: Osvaldo Jaeggli & Carmen Silva-Corvalán (eds.): Studies in Romance Linguistics, 99–119. Dordrecht: Foris. Koch, Peter (1996): “Der Beitrag der Prototypentheorie zur Historischen Semantik: eine kritische Bestandaufnahme.” – In: Romanistisches Jahrbuch 46, 27–46. – (1999): “On the Cognitive Bases of Metonymy and Certain Types of Word Formation.” – In: Klaus-Uwe Panter & Günter Radden (eds.): Metonymy in Language and Thought, 139–167. Amsterdam/Philadelphia: Benjamins. – (2001): “Metonymy. Unity in diversity.” – In: Journal of Historical Pragmatics 2 (2), 201–244. – (2003): “Qu’est-ce que le cognitif?” – In: Peter Blumenthal & Jean-Emmanuel Tyvaert (eds.): La cognition dans le temps, 85–100. Tübingen: Niemeyer. – (2004): “Metonymy between pragmatics, reference, and diachrony.” – In: metaphorik.de 7, 6–54. (URL: http://www.metaphorik.de/07/koch.pdf [02/09/2009]) Koch, Peter & Oesterreicher, Wulf (1985): “Sprache der Nähe – Sprache der Distanz. Mündlichkeit und Schriftlichkeit im Spannungsfeld von Sprachtheorie und Sprachgeschichte.” – In: Romanistisches Jahrbuch 36, 15–43. – (1990): Gesprochene Sprache in der Romania: Französisch, Italienisch, Spanisch. – Tübingen: Niemeyer [Spanish edition (2007): Lengua hablada en la Romania: Español, Francés, Italiano. – Madrid: Gredos]. Kühner, Raphael & Stegmann, Carl (1912/1994): Ausführliche Grammatik der lateinischen Sprache. Vol. 3. – Darmstadt: Wissenschaftliche Buchgesellschaft. Lathrop, Thomas A. (2002): Curso de gramática histórica española. – Barcelona: Ariel. Marcos Marín, Francisco (1980): Curso de Gramática española. – Madrid: Cincel.

‘Modal uses’ of the Italian imperfetto and the Spanish imperfecto: a comparison

65

Mazzoleni, Marco (1992): “‘Se lo sapevo non ci venivo’: l’imperfetto indicativo ipotetico nell’italiano contemporaneo.” – In: Bruno Moretti & Dario Petrini (eds.): Linee di tendenza dell’italiano contemporaneo. Atti del XXV Congresso della SLI, 171–190. Roma: Bulzoni. – (22001): “Ipotetiche e concessive.” – In: Lorenzo Renzi, Giampaolo Salvi & Anna Cardinaletti (eds.): Grande grammatica di consultazione. Vol. 2, 751–817. Bologna: Il Mulino. Menéndez Pidal, Ramón (ed.) (1977): Primera crónica general de España. – Madrid: Gredos. Montolío, Estrella (1999): “Las construcciones condicionales.” – In: Ignacio Bosque Muñoz & Violeta Demonte Barreto (eds.): Gramática descriptiva de la lengua española. Vol. 3, 3644–3737. Madrid: Espasa Calpe. Penny, Ralph (2000): Variation and Change in Spanish. – Cambridge: Cambridge University Press. – (2001): Gramática histórica del español. – Barcelona: Ariel. Pérez Saldanya, Manuel (2004): “Los tiempos verbales: dificultades teóricas y terminológicas.” – In: Luis García Fernández & Bruno Camus Bergareche (eds.): El pretérito imperfecto, 194–228. Madrid: Gredos. Pinkster, Harm (1988): Lateinische Syntax und Semantik (rev. and expanded version). – Tübingen: Franke. Reichenbach, Hans (1947): Elements of Symbolic Logic. – New York/London: Free Press/Collier-Macmillan. Reyes, Graciela (1990): “Tiempo, modo, aspecto e intertextualidad.” – In: Revista Española de Lingüística 20, 17–53. Rojo, Guillermo (1974): “La temporalidad verbal en español.” – In: Verba 1, 68–149. Rojo, Guillermo & Veiga, Alexandre (1999): “El tiempo verbal. Los tiempos simples.” – In: Ignacio Bosque Muñoz & Violeta Demonte Barreto (eds.): Gramática descriptiva de la lengua española. Vol. 2, 2876–2934. Madrid: Espasa Calpe. Sabatini, Francesco (1985): “L’‘italiano dell’uso medio’: una realtà tra le varietà linguistiche italiane.” – In: Günter Holtus & Edgar Radtke (eds.): Gesprochenes Italienisch in Geschichte und Gegenwart, 154–184. Tübingen: Narr. Squartini, Mario (2004): “La compatibilidad aspectual de los predicados estativos intrínsecamente delimitados.” – In: Luis García Fernández & Bruno Camus Bergareche (eds.) (2004): El pretérito imperfecto, 317–345. Madrid: Gredos. – (2005): „Il verbo.“ – In: Giampaolo Salvi & Lorenzo Renzi (eds.): Grammatica dell’italiano antico (URL: http://geocities.com/gpsalvi/konyv/ [02/09/2009]). Tekavþiü, Pavao (1972/74): Grammatica storica dell’italiano. Vol. 3. – Bologna: Il Mulino. Ultan, Russell (1978): “The Nature of Future Tenses.” – In: Joseph H. Greenberg, Charles A. Ferguson & Edith A. Moravcsik (eds.): Universal of Human Language. Vol. 3: Word Structure, 83–123. Stanford, Calif.: Stanford University Press. Vater, Heinz (1994): Einführung in die Zeit-Linguistik. – Hürth-Efferen: Gabel. Veiga, Alexandre (1988): “Planteamientos básicos para un análisis funcional de las categorías verbal en español.” – In: Gerd Wotjak & Alexandre Veiga (eds.): La

66

Sarah Dessì Schmid

descripción del verbo español, 237–257. Santiago de Compostela: Universidad de Santiago. – (1992): “La no independencia funcional del aspecto en el sistema verbal español.” – In: Español Actual 57, 65–80. – (1996): La forma verbal española cantara en su diacronía. – Santiago de Compostela: Universidad de Santiago. – (2004): “La forma verbal “cantaba” y la estructura modo-temporal del sistema verbal español.” – In: Luis García Fernández & Bruno Camus Bergareche (eds.): El pretérito imperfecto, 96–193. Madrid: Gredos. Weinrich, Harald (62001): Tempus. Besprochene und erzählte Welt (new ed.). – Müchen: Beck.

Ion Giurgea & Elena Soare

Modal non-finite relatives in Romance

1. Introduction This paper investigates non-finite modal relatives in Romanian and two other Romance languages (French and Italian). These constructions, which express deontic necessity or teleological possibility (section 2) and are only object-gap constructions, will be argued to be reduced relatives in Bhatt’s (1999) terms, i. e. participials unlike their English counterparts (section 3). After presenting a general analysis of reduced relatives (section 3.2), we will propose a syntactic analysis of the Romance modal participials (section 3.3). We will argue then that there are other environments in which this structure appears (section 4), in particular Tough Constructions (section 4.3).

2.

Modal non-finite relatives in Romance: the forms and their meanings

2.1. The forms Like English, Romance languages have modal non-finite relatives involving relativization of the object. They are built by an introductory element – French à, Italian da and Romanian de – (see section 3.3 for an analysis of this element) and a non-finite form, the infinitive in all Romance languages except Romanian, and the so-called ‘supine’ in Romanian: (1a) CărĠile de citit sînt pe masă. (Rom.) books-the of read-SUP are on table (1b) Les livres à lire sont sur la table. (Fra.) the books to read-INF are on the table (1c) I libri da leggere sono sul tavolo. (Ita.) the books of read-INF are on-the table

The Romanian ‘supine’ is formally identical to the past participle, except in terms of agreement (it lacks agreement marks). It is usually preceded by de, and more rarely by the prepositions la ‘to’, pentru ‘for’. The same stem is used

68

Ion Giurgea & Elena Soare

as a base for a productive nominalization (the ‘nominal supine’; see Soare 2002). 2.2. Meanings Generally, modal non-finite relatives in Romance express deontic necessity (see the possible combination with adverbs such as no matter what, at all costs): (2a) livres à lire à tout prix (Fra.) (2b) cărĠi de citit neapărat (Rom.) books to read at all costs

A less frequent reading is teleological potentiality (something that you can do if you want to achieve a certain goal). The goal is often unexpressed but we can always paraphrase these sentences with sentences containing an expressed goal (purpose clauses or conditionals + bouletic verbs). In the following example, “something to read” can be paraphrased as “something that you can read if you are in the mood for reading / if you want to have a good time”: (3a) Ai adus ceva de citit? (Rom.) have-2.SG brought something to read (3b) Tu as apporté quelque chose à lire? (Fra.) you have-2.SG brought something to read ‘Did you bring something to read?’

In the following Spanish example, the goal is overtly expressed in the parenthetical clause: (4)

Puede dar miedo pensar que los libros de leer – que se leen por el placer de leer libros – puedan utilizarse en la escuela para enseñar valores. ‘It can be frightening to think that the books to read – the one that are read for the pleasure of reading books – can be used in school to teach values.’

We can also find teleological necessity and intermediate cases between teleological necessity and possibility, like in the English be worth. Since the interpretation is either deontic or teleological, the highest role in the argument list of the verb is always an agent (obligations, permissions and goal-oriented modality only apply to actions – events which are controlled by a conscious being). Even if this role is not syntactically projected – see the next section, where we propose that these constructions are passive – it is semantically active, as in normal passives. The ungrammaticality of the examples in (5) is due to the fact that an agent is missing; this is particularly

69

Modal non-finite relatives in Romance

clear in (5b); note that the corresponding construction with the supine is grammatical when the external argument bears an agent role. (5a) *flori de ofilit (Rom.) flowers to wither (5b) ploaia spală maúina rain-the washes car-the

ե

*maúină de spălat (de către ploaie) car to wash by rain

The epistemic reading is unavailable, as expected in a non-finite structure. Ability and stereotypical readings are also not found. If negation is present, in French it scopes under the deontic necessity: (6)

C’est une chose à ne pas dire à tout le monde. this is a thing to not tell to all the people ‘This is something not to tell to everybody.’

The teleological possibility reading is blocked with negation. In Romanian, the use of negation (which has the form of participial negation, see the next section) triggers a change of meaning: the construction expresses inability / circumstantial impossibility: (7)

carte de necitit book de ne-read-SUP ‘unreadable book’

Marginally, it expresses deontic necessity over negation: (8)

lucru de nemărturisit thing to un-confess-SUP ‘undisclosable thing’

In spite of the special meaning, the condition of the existence of an Agent role is observed: (9)

*floare de neofilit flower to un-wilt-SUP

In sum, the type of modality found in these constructions involves an agent and has a circumstantial modal base.

70

Ion Giurgea & Elena Soare

3. Syntactic analysis 3.1. Romance modal non-finite relatives are reduced relatives (i. e. participials) Unlike their English counterparts, it can be shown that Romance modal nonfinite relatives are reduced relatives according to Bhatt’s (1999) criteria: (i) they can appear in postcopular position (cf., for this criterion, Embick 1997, Iatridou et al. 1999) (see (10)) and (ii) relativization is strictly local (see (11)): (10a) Ces livres sont à lire jusqu’à mardi. (Fra.) these books are to read until Tuesday (10b) Questi libri sono da leggere fino a martedì. (Ita.) (10c) Aceste cărĠi sunt de citit până marĠi. (Rom.) (10d) *These books are to read until Tuesday. (Eng.) (11a) *un livre à dire à tes enfants de lire (Fra.) a book to tell to your children to read (11b) *o carte de zis copiilor să citească (Rom.) a book of tell-SUP children-the.DAT SUBJ read.3.PL (11c) a book to tell your children to read (Eng.)

On the other hand, infinitival relatives involving relativization of the subject, which in English do qualify as reduced structures (see (12): (12b) shows the predicative use, (12c) the strict locality condition), are restricted in Romance (see (13)); in some languages they are absent (Romanian), in others, they are probably subcategorized by ordinals, superlatives and only, and are not modal (see French, ex. 14):1 (12a) a man to fix the sink (12b) The wine is to complement the cheese. (12c) *the boy [Opi [it would be fun [ti to dance]]]

(Bhatt 1999: 11, Bsp. 7) (Bhatt 1999: 10, Bsp. 6c)

(13a) *un homme à réparer l’évier (Fra.) a man to fix the sink (13b) *un om de reparat chiuveta (Rom.) a man of fix-SUP sink-the 1

Note that in this case the past infinitive is allowed (see (14a, 14c), which is impossible in modal infinitival relatives. Since this type of infinitival is not modal, it falls outside the scope of this paper.

Modal non-finite relatives in Romance

71

(14a) le premier homme à avoir résolu ce problème (Fra.) the first man to have solved this problem (14b) la seule personne à porter ce genre de robe à cette soirée the only person to wear this kind of dress at this party (14c) le plus jeune pianiste à avoir interprété ce concert the most young pianist to have performed this concert

A further difference is that the periphrastic passive is not allowed in Romance, as opposed to English: (15a) a book to be read (15b) *un livre à être lu (Fra.) a book to be read

Note that according to the distributional properties that define it, the term reduced relative is equivalent to the term participial (reduced relatives thus defined have the distribution and interpretation of participles). Romance reduced relatives differ morphologically from what is traditionally called a “participle” (in languages with adjectival agreement) by the fact that they do not show agreement. However, we consider the term “participial” to be more appropriate than the term “reduced relative” because relatives are normally inside a nominal projection – or constitute one, in the case of free relatives –, while reduced relatives may appear predicatively, as we have seen. In the following, no distinction is made between the two terms. 3.2. The structure of reduced relatives Bhatt (1999) proposes the generalization that reduced relatives are only based on the relativization of the external argument. Since we have seen that Romance modal non-finite relatives are reduced relatives based on the relativization of the object, we propose a reformulation of Bhatt’s generalization in terms of case marking: (16) Reduced relatives involve “relativization” of the non-case-marked argument

Another instance of reduced relatives based on an internal argument which is not case marked is passive participles: (17a) the cake [made by John] (17b) The cake was [made by John].

72

Ion Giurgea & Elena Soare

Modulo this reformulation, we will adopt the core of Bhatt’s analysis of reduced relatives, whose central assumptions are: (i) the relativized position is not case-marked; (ii) relatives are internally headed (the NP raises from the reduced relatives to its surface position); (iii) reduced relatives do not contain a C (contra Kayne 1994). The lack of C is meant to account for the fact that the relativization chain is not an A-bar chain (since the base position is not casemarked), as opposed to full relatives. Bhatt only concentrates on the attributive use, and considers that the base position is not case marked because it is occupied by an NP (the “pivot” of the relative) and NPs, contrary to DPs, do not need case. We would like to propose instead a unified analysis of predicative and attributive reduced relatives. Note that some of the arguments which Bhatt provides in favour of movement for attributive reduced relatives also hold for the predicative use (see idiom chunks in (18)): (18a) the headway made (18b) What headway has been made!

So, for the predicative use, we propose that the subject DP raises from the object position in order to be case-marked. Then it seems reasonable to assume that the same motivation determines the movement of the pivot in the attributive structure. We may thus explain how this movement is possible without a +wh C: while in the case of full relatives the pivot has a +wh or +rel determiner which makes it an active goal for the attractive property of the C probe2, in reduced relatives, since there is no +wh determiner, what makes the pivot active is an unvalued Case. As for the Probe, we may assume that it simply has a +N feature. For the predicative use, the object probably first raises to SpecPred (if predicative constructions are defined by the presence of a Pred head, see Bowers 1993), from where it can further raise to SpecTP in order to be case-marked or receive case from outside in ECM constructions3. This can be illustrated by the following bracketing structures, where (19a) 2

3

In minimalist terms, dependencies like Move and Agree are encoded in terms of relations between an element which has an unvalued feature, called a probe, and an element in the c-command domain of the probe on which that feature is valued, called a goal; the goal must also have an unvalued feature which has a valued counterpart on the probe, in which case it is said to be active. See Chomsky (1995, 2000). ECM (i. e. Exceptional Case Marking) constructions are constructions traditionally labelled as Accusative-Infinitive constructions. For instance, in English, the verb consider may take an infinitival complement with an accusative subject, like in I consider him to be the best. In the classical Government & Binding literature, it is assumed that the subject of the infinitive exceptionally receives accusative case from the main verb, hence the name of the construction (cf. Chomsky 1981).

Modal non-finite relatives in Romance

73

represents the copular predicative construction, and (19b) the ECM construction: (19a) [ DPi T [vcop [PredP ti Pred [vp/AspP .. v [V ti]]]]] [TP[the book] [[vcop+T was] [tvcop [ tthe book Pred [vP/AspP sold tthe book]]]]] (19b) [ v .. [PredP DPi Pred [vp/AspP .. v [V ti]]]] [v' make [PredP [the book] [ Pred [vP/AspP sold tthe book]]]]

For the attributive structure, we adopt Bhatt’s view that the movement involved is linked to the transformation of the clausal projection into a nominal one. This transformation characterizes not only reduced relatives, but also full relatives. The main idea is that the projection obtained after the raising of the pivot becomes nominal before combining with the determiner. Bhatt proposed two implementations of this idea (the first in Bhatt 1999, the second in Bhatt 2002), which try to solve a drawback of the standard analysis of raising relatives, whose most recent and full-fledged variant can be found in Bianchi (1999). The problem which Bhatt addresses concerns the selectional properties of D: in the standard analysis, one must assume that D, as well as other functional nominal items merged above relatives, does not always select for a +N projection, but may sometimes select for a CP. Bhatt solves this problem by considering that raising relatives become, during the derivation, nominal projections. In the 1999 analysis, he uses the idea of ‘projecting movement’, initially proposed by Iatridou, Anagnostopoulou and Izvorski in various drafts of their 2001 paper for free relatives. The idea is that raising relatives instantiate a structure whereby a selected item, the NP outer specifier of the relative C, gives its label to the object formed by its merger with the CP. This analysis is represented, for full relatives, in the tree below, where we introduced a feature +proj [projecting] signalling label inheritance from the specifier, for the second specifier of the relative C:

74

Ion Giurgea & Elena Soare

(20) the book (that) I bought DP 3 D NP ! 3 the NP CP (+rel={+NP,+EPP,+proj}) ! 3 book DP C(+wh,+rel) 2 3 D NP C(+Ifin,+wh, +rel) IP ! ! ! 6 Ø+wh book that I bought [Ø+wh book]

For reduced relatives, we may assume that the +proj feature is borne by a Pred head. As we said, the pivot in this case is made eligible for raising not by virtue of a +wh feature, as classically assumed in the literature on relatives, but by virtue of having unvalued case (not being case-marked). After raising, it will be in a position where it will be able to inherit case from the determiner merged above it (roughly the same proposal concerning case is made in Kayne 1994 and Bianchi 1999): (21) the book read DP V D ! the

NP V N Pred {+N +proj} ! V book Pred IP 5 read [book]

In his 2002 paper, Bhatt replaces projecting movement, which is not a standard device in the current minimalist syntax (although the possibility is recognized by Chomsky 2005), with a nominalizer head. Instead of having an outer specifier in (20), we would have a Nom head selecting a +wh CP, attracting an NP in its specifier and marking the projection as nominal. Bhatt does not develop this analysis for reduced relatives, but we may assume that the Nom head selects a PredP in their case, as shown below:

Modal non-finite relatives in Romance

75

(22) the headway made DP V D ! the

NomP V NP Nom' ! V headway Nom PredP ! V Ø NP Pred ! V headway Pred vP/AspP ! 5 Ø made headway

We will not decide here between these two analyses, which are largely equivalent. What is important is that a raising analysis for reduced relatives is feasible. This analysis directly accounts for the fact that the relativized position is not case-marked. As for the strict locality, it is arguably due to the same fact: the Pred head cannot probe until it finds a +wh element; since it only has a +N feature it will be sensitive to the closest element +N. If this element is not active (being, for instance, case marked), an intervention effect will arise. 3.3. The structural analysis of Romance modal reduced relatives (participials) Having established that Romance modal non-finite relatives are reduced relatives and that in reduced relatives the relativized position is not casemarked, it follows that Romance modal non-finite relatives (“modal participials”), which, as we have seen, always involve relativization of the object, are passive. This immediately accounts for the fact that they do not allow a form infinitive be + past participle (see (15b) above): since they are already passive, they cannot undergo further passivization. This idea is problematic given the fact that modal non-finite relatives do not have explicit passive morphology: the infinitive used in French and Italian as well as the Romanian “supine” are active in other contexts. However, the passive analysis is supported by the possibility of using agentive PPs in French and (at least for some speakers) in Romanian: (23a) livre à lire par tous (Fra.) book to read by all ‘a book everybody should/must read’

76

Ion Giurgea & Elena Soare

(23b) Sunt multe lucruri de rezolvat de către ministerul Agriculturii. (Rom.) are many things de solve-SUP by ministry-the agriculture-the.GEN ‘There are many things to be solved by the Ministry of Agriculture.’ (www.amosnews.ro/PrintArticle201911.phtml [02/09/2009])

There is another context, in French, where infinitival morphology is associated with passive properties: the causative faire + Inf. constructions, in which either the subject or the object receive accusative, but never both, suggesting that this case is assigned by faire. Moreover, in the faire par-construction, the subject is realized as a PP-agent, like in (standard) passives4: (24a) J’ai fait manger Marie. I have done eat-INF Marie (24b) J’ai fait manger la pomme *(à/par) Marie. I have done eat-INF the apple (to/by) Marie

This shows that infinitival inflection is not incompatible with a passive v (the infinitival in (24), as in other restructuring construction, presumably represents a mere vP). So far we have established that modal participials contain a passive v, at the bottom, and do not have a C, but merge with a Pred head, at the top. Other facts will help us to establish what other functional projections they contain (between Pred and v). In French and Italian, clitics and sentential negation may appear before the infinitive. In Romanian, the supine inflection does not allow either clitics or sentential negation – see (26) – while the infinitive, which is not used in modal participials, allows for both, except in the one restructuring context, with a putea ‘can’ – see (27): (25a) une chose à lui transmettre à tout prix (Fra.) a thing to him.DAT communicate at all costs (25b) una cosa da trasmettergli (Ita.) a thing to communicate-him.DAT (25c) une chose à ne pas avouer (Fra.) a thing to not confess

4

See Kayne (1975), Rouveret & Vergnaud (1980) for the similarities between passive and causative structures. Some researchers assume that causatives are bi-clausal structures (like for instance Guasti 1996); more recently, Homer et al. (2009) argue that they are actually monoclausal. An analysis of infinitives in causative structures, though, falls outside the scope of this paper.

Modal non-finite relatives in Romance

77

(25d) una cosa da non ammettere (Ita.) a thing to not admit (26a) *lucru de-i spus (Rom.) thing of him.DAT tell-SUP (26b) *lucru de nu spus (Rom.) thing of not say-SUP (27a) înainte de a-i spune before of to-him.DAT do-INF (27b) pentru a nu rata for to not fail (27c) N-am cu cine îl lăsa not-have.1 with whom him.ACC leave

The Romanian supine only allows the prefixal negation ne-, also found with past participles. As we have shown in 2.2, the meaning also changes when this negation is used: it may be circumstantial, not only deontic (compare (7) and (8) above):5 (28a) lucru neînĠeles thing un-understood (28b) lucru de neînĠeles thing of un-understand-SUP

We consider that the possibility of having clitics and sentential negation is indicative of a Tense (or I) head. So, French and Italian modal participials have a non-finite Tense, while Romanian does not. A further problem concerns the status of the introductory element – Fra. à, Italian da, Spanish de, Romanian de. This element cannot be C, given the assumption that reduced relatives in general do not contain a C. Could it be the realization of the Pred head? We will present arguments against this view in the next section: there, we will show that the structure we are examining, which we called a “modal participial”, is also found in Tough Constructions.6 5

6

The use of the negative form is not equally acceptable with all verbs, which seems to indicate that the construction is at least partially lexicalized: (i) ? întâmplare de nepovestit incident of un-report-SUP The ‘Tough Construction’ is a construction in which an adjective which otherwise may take a clause as its single argument, takes as a complement a non-finite clause containing a gap and has a subject which is interpreted as the filler of this gap: e. g. it is easy to please John (the impersonal use) vs. John is easy to please (the

78

Ion Giurgea & Elena Soare

However, there it does not combine with a Pred head, but is selected by the tough-adjective. Since the introductory element is also present in that environment, it cannot be Pred: (29) livre difficile à lire (Fra.) book difficult to read

Should we consider it a part of the inflection, realizing a modal inflection, perhaps in a scattered form (à/de …-er, where -er is the infinitival inflection)? This does not seem acceptable either because of the position of negation and clitics: in Romance the negation normally appears before I, while clitics either appear before I or after a suffixal I (like in Italian gerunds and infinitivals). But in this case, clitics appear between the introductory element and the infinitival in French and Romanian, and the negation appears between the introductory element and the infinitival in all three languages. This behavior recalls that of the so-called Romanian “mood particles”, subjunctive să and infinitival a: they too precede clitics and negation; moreover, subjunctive să can co-occur with a complementizer, which shows that it is not (always) a C: (30a) Vreau ca Ion să-i vorbească. want-1.SG that Ion să him/her.DAT talk-SUBJ.3.SG ‘I want Ion to talk to him/her.’ (30b) să nu veniĠi să not come-2.PL (30c) a nu-l asculta to not him.ACC listen-INF

Several authors proposed that these particles realize a functional head Mood which is placed above Tense (Rivero 1987 and Terzi 1993 for să, Avram 1999 for să and a). Adopting this proposal for the introductory elements of modal participials has the advantage of explaining their modal properties, in addition to the distributional similarities we have just shown. For Romanian, we assume that de is also a Mood head (which accounts for the modal meaning) but does not select a TP. Since the supine does not have any aspectual oppositions, we may assume that Mood directly selects a vP.

‘Tough Construction’). An influential line of research, which we adopt here, has analyzed this construction as relying on movement (see Rosenbaum 1967, Postal 1971, Berman 1973, Bayer 1990, Sportiche 2002, 2006, Hicks 2004, Giurgea & Soare 2006).

Modal non-finite relatives in Romance

79

To conclude, we propose the following structures for modal participials: (31a) (Pred) [Mood [Tense(non-finite) [v (pass.) .. ]]] (Fra. & Ita.) à lire (31b) (Pred) [Mood [v (pass.) .. ]] (Rom.) de citit

The idea that Romance modal participials are based on a passive v faces several possible objections, which will now be addressed. First, why don’t these forms show agreement, as passive (past) participles do in these languages (cf. (32) and (33))? Note that agreement in passive modal participles is found in some languages, for instance in Latin (cf. (34)). (32a) cărĠi de citit(*e) (Rom.) books to read(-AGR) (32b) cărĠi citite books read-AGR (33a) livres à lire (Fra.) books to read-INF (33b) livres lus books read-AGR (34) libri legendi (Lat.) books-NOM.PL read-MODPART.MASC.NOM.PL

We take the lack of agreement to be an indicator that the functional structure of the two participles is not identical, overt agreement characterizing a head which is not found in modal participials – perhaps an Asp head (past participles do not qualify as IPs according to our criteria, since they do not allow clitics). The lack of agreement in modal participials is a morphological peculiarity which has a historical explanation (modal participials originate from preposition + nominalized infinitive structures of the type “books for reading”). A more important problem is the absence of modal participials with unaccusatives: it is known that past participles in Romance may also be built on unaccusatives. Why then modal participles cannot do so? (35a) om căzut (Rom.) man fallen (35b) *om de căzut man of fall-SUP

80

Ion Giurgea & Elena Soare

(35'a) personne tombée (Fra.) person fallen (35'b) *personne à tomber person to fall

This situation suggests that modal participials always have an unexpressed agent, necessarily distinct from the object, to which the deontic modality refers. We have shown in 2.2 that modal participials always presuppose the existence of an agent. We may assume then that the semantic entry of Mood refers to the implicit agentive argument of passives. Unaccusatives lack this argument role; their only argument is a deep object. Therefore they do not provide the argument required by the modal head. But there are some data which suggest that the agent may be syntactically projected (even if it is null). Modal participials allow some secondary predicates referring to the agent, such as ‘only’, unlike passive past participles: (36a) problemă de rezolvat singur (Rom.) problem of solve-SUP alone (36b) problème à résoudre seul (Fra.) (36c) problema da risolvere da soli (Ita.) problem to solve da alone-M.PL (37a) *Problemele au fost rezolvate singur. (Rom.) problems-the have been solved alone (37b) *Les problèmes ont été résolus seul. (Fra.) (37c) *I problemi sono stati resoluti da soli. (Ita.)

This construction indicates the existence of a PRO agent. But passive verbal forms do not allow this kind of subject, as shown in (37). We claim that what introduces an agent PRO is not the lexical verb itself, but the Mood head. It has long been observed that the subject of deontics receives a theta-role from the deontic and controls the subject of its verbal complement (see Ross 1967, Zubizarreta 1982, Roberts 1985, Hacquard 2006, Avram 1999): the obligation, permission or teleological modality expressed by the deontic is attributed to an agent. To conclude, we propose the following structures (where (38a) is illustrated by examples (33a), (36b, c) and (38b) by examples (32a) and (36a)): (38a) (Pred) [MoodP PRO [Mood [Tense(non-finite) [v (pass.) .. ]]] (Fra. & Ita.) [MoodP PRO [Mood' à [T lire [vP tv tlivres]]]] (38b) (Pred) [MoodP PRO [Mood [v (pass.) .. ]]] (Rom.) [MoodP PRO [Mood' de [vP citit tcărĠi]]]

Modal non-finite relatives in Romance

81

This analysis helps us to solve a third problem, which has not yet been mentioned: the fact that structures with an overt PP agent, as in (23), are not at all common: in Romanian in particular, the presence of a PP agent is marginal for some speakers. The structures in (38) offer a straightforward explanation for this vacillation: for some speakers PRO is always projected, for others not. If PRO is projected, a PP agent cannot appear because it will violate the principle C7: (39) *lucruri [MoodP PROi de [făcut [PP de către elevi]i] things of do-SUP by pupils

Further evidence for the possibility of a modal element introducing an argument is offered by the deontic necessity construction of Latin and other Indo-European languages, where the deontic participle is clearly passive, since it agrees with the object, but the agent may also be expressed as a dative: (40a) Mihi est liber legendus. me.DAT is book(MASC)-NOM read-MODPART.MASC.SG.NOM ‘I have to read a book.’ (40b) liber mihi legendus book(MASC)-NOM me.DAT read-MODPART.MASC.SG.NOM ‘a book I should read’

The form of the agent cannot be explained by a passive v, because passive v is not associated with a dative agent, but to a PP agent introduced by ab: (41) Liber a me legitur. book from me read-PRES.3.SG.PASS ‘The book is read by me.’

So it must be the element which introduces modality that assigns the agent dative case. Moreover, the agent could not have originated inside the passive vP and be case-marked via raising to SpecMood, because passive vPs do not project a non-case-marked agent8. The agent is either introduced by a PP or is 7

8

Principle C of the Binding Theory states that a ‘referential expression’ (a full nominal expression, as opposed to personal and reflexive pronouns) cannot be bound (i. e., cannot be co-indexed with a c-commanding nominal expression). In (39), PRO would bind elevi and violate this principle. See for instance Chomsky (1981) for an overview of the Binding Theory, a module of the grammar which accounts for the referential relations between nominal expressions (anaphors, pronouns and referential expressions). If the agent had originated as a non-case-marked argument in the passive vP, it would have been case-marked by T in contexts in which this low deontic Mood was not present, because it is closer to T than the object. But in passives it is the object

82

Ion Giurgea & Elena Soare

absent from the syntactic representation (see (37) as an argument against the presence of a null PRO in passives). We must conclude then that in Latin the modal head present in modal passive participles comes with an argument position for the agent, to which it assigns dative case. The difference between Romance and Latin is that Romance deontic Mood has lost the ability to case-mark the agent. Therefore only (arbitrary) PRO is allowed.

4.

Other occurrences of modal participials

4.1. Romanian copular construction based on the impersonal passive In Romanian, intransitive verbs may appear in the supine (preceded by de) following the copula, with a deontic necessity meaning: (42) E de vorbit cu profesorul. is of talk-SUP with professor-the ‘We/one should talk to the professor.’

As we have already mentioned, the Romanian verbal supine appears in other environments than the modal participial (for instance, it can be active). However, the supine in (42) has several properties which support the idea that it represents a modal participle: the modal meaning (deontic necessity), the fact that it combines with the copula and the fact that objects are not allowed which is case-marked by T, thus becoming the subject. This argument is based on the following standard assumptions in the current generative analyses of argument realization alternations: (i) the object originates in the same position in active and passive constructions, but in passives there is no accusative assignment (see Baker 1988, Chomsky 1981, Grimshaw 1990), (ii) a case assigner assigns case to the closest non-case-marked nominal in its c-command domain (where for Y, Z ccommanded by X, Y is closer to X than Z if Y asymmetrically c-commands Z) (see Chomsky 1995), (iii) nominative case is assigned by T (see Chomsky 1981, 1995). Since Chomsky (1995), the source of accusative case is considered to be a functional head v immediately above V, which may be identified with the category of Voice: when v is active, it assigns accusative and takes an external argument which is not case-marked and will become the subject; in passives, either v is absent or it lacks the accusative assignment property and introduces the agent as an optional PP. Object-to-subject raising is also found in unaccusatives, but in their case there is no agent in the argument structure of the verb, so the object is the only candidate for case-marking by T.

Modal non-finite relatives in Romance

83

(see Avram 1999). Compare (43a) to (43b), where we see a verbal supine selected by a verb (the verb a se apuca ‘to start’). In (43b) de probably represents a preposition, as shown by the fact that this verb may also be followed by de + DP (see 43c). Note that in this case an object may appear: (43a) *E de spălat geamurile. is of wash-SUP panes-the (43b) M-am apucat [de spălat geamurile]. me.ACC have-1.SG seized of wash-SUP panes-the ‘I started washing the (window) panes.’ (43c) M-am apucat de treabă. me.ACC have-1.SG seized of work ‘I started the work / I set to work.’

If (42) is an instance of a modal participial, what is its object? We propose that it is a null cognate object. The existence of a null cognate object has been proposed by Dobrovie-Sorin (1998) for Romanian impersonal reflexive unergatives, exemplified in (44) below. Having established that Romanian does not have a nominative se like Italian and Spanish, Dobrovie-Sorin analyzes Romanian intransitive impersonal reflexives as impersonal reflexive passives based on the passivization of a cognate object: (44) Se vorbeúte cu profesorul. REFL talks with teacher-the ‘People talk to the teacher / Somebody talks to the teacher.’ Øi se vorbeúte ti

4.2. Small clause in the complement position of have In the complement position of have, a Small Clause based on the modal participle may appear in an ECM configuration. (45a) Am multe lucruri de făcut. (Rom.) have-1.SG many things of do-SUP (45b) J’ai beaucoup de choses à faire. (Fra.) I have many of things to do (45c) Ho molte cose da fare. (Ita.) have-1.SG many things to do ‘I have many things to do.’

84

Ion Giurgea & Elena Soare

The interpretation shows that we do not have to deal with an attributive modal participle: (45) does not imply a possession relation between the subject of have and the object of the participle. So we assign it the following structure: (46) [vP am .. [PredP [DP multe lucruri] [Pred [MoodP de [vP făcut [DP multe lucruri]]]]]]

This construction only expresses deontic necessity (cf. also (47a)): if in the construction have + DP + modal participle the participle expresses teleological possibility, the interpretation shows that we are dealing with the possessive have (there is a possession relation between the subject of have and the object of the participle, cf. (47b)) and the modal is simply attributive (it is part of the DP): (47a) am o carte de citit (deontic necessity)

ե

am o carte

(47b) am o carte de citit (teleological potentiality) ‘I have a book to read.’

ী

am o carte ‘I have a book.’

4.3. Tough-constructions The non-finite complement in Romance Tough Constructions [TCs] is arguably also a modal participial. Tough-adjectives9 are characterized by the alternation pattern exemplified in (48). (48a) Il est difficile de lire ces livres. it is hard to read those books.

Impersonal use

(48b) Ces livres sont difficiles à lire. these books are hard to read.

TC(predicative)

(48c) des livres difficiles à lire books hard to read

TC (attributive)

(Fra.)

In TCs, the external argument (subject/head noun) of the tough-adjective is interpreted as the object of the embedded infinitival. The infinitival appears to be the complement of the tough-adjective and the entire constructions Tough + Infinitival functions as an attribute or a predicate. However, there are several similarities between the verbal complement in TCs and modal nonfinite relatives which led to the proposal that they both represent the same verbal structure (see Kayne 1972, Sportiche 2002, 2006, Soare 2002, Soare & Dobrovie-Sorin 2002, Giurgea & Soare 2006): (i) The non-finite form 9

In French, a list of these adjectives is provided by Léger (2006): aisé, difficile, dur, évident (n’est pas…), facile, impossible, malaisé, pénible, simple.

Modal non-finite relatives in Romance

85

in TC is the same as that used in modal reduced relatives (as noted for French by Kayne 1972): (49a) Ces livres sont difficiles à lire. (Fra.) these books are hard à read-INF (49) livres à lire books à read-INF (50a) Questi libri sono difficili da leggere. (Ita.) these books are hard da read-INF (50b) libri da leggere books da read-INF (51a) Aceste cărĠi sunt greu de citit. (Rom.) these books are hard de read-SUP (51b) cărĠi de citit books de read-SUP

Note that this form is often different from the form used in clausal complements of evaluative control adjectives of the type clever, kind (which do case-license the object) (in French and Italian, the introductory element is different – French de vs. à, cf. (52), Italian di vs. da, cf. (53); in Romanian, the form itself is different – supine in TC and modal relatives, cf. (54b), vs. subjunctive in control clauses, cf. (54a)): (52a) Vous avez été gentil de le faire. (Fra.) you have been kind de it do-INF (52b) Cela est difficile à faire. this is difficult à do-INF (53a) Lei è gentile di averlo fatto. (Ita.) you are kind de have-INF-it done (53b) Questo e difficile da fare. this is hard da do-INF (54a) AĠi fost amabil să o faceĠi. (Rom.) you-have been kind SUBJ it do-2.PL (54b) Asta e greu de făcut. this is hard de do-SUP

86

Ion Giurgea & Elena Soare

In French and Italian, in the impersonal use of tough-adjectives (cf. (55a)), the clausal complement has the same form as with control adjectives (cf. (56a)), a form that is different from the one used in TCs (cf. 55b)):10 (55a) Il est difficile de soutenir cette analyse. (Fra.) it is difficult de defend-INF this analysis (55b) Cette analyse est difficile à soutenir. this analysis is difficult à defend-INF (56) Il a été gentil de soutenir cette analyse. he has been kind de defend-INF this analysis

(ii) The gap in the non-finite clause only corresponds to the object: (57a) ces livres / *ces personnes sont difficiles à lire (Fra.) questi libri / *queste persone sono difficili da leggere (Ita.) aceste cărĠi / *aceste persoane sînt greu de citit (Rom.) these books / these persons are hard to read (57b) des livres / *des personnes à lire (Fra.) libri / *persone da leggere (Ita.) cărĠi / *persoane de citit (Rom.) books / persons to read

(iii) No periphrastic passive is allowed: (58a) *un livre difficile à être lu (Fra.) *un libro difficile da essere letto (Ita.) *o carte greu de a fi citită (Rom.) a book difficult to be read (58b) *un livre à être lu (Fra.) *un libro da essere letto (Ita.) *o carte de a fi citită (Rom.) a book to be read

(iv) The dependency is strictly local in Romance, while in English it is unbounded: (59a) ??un livre difficile à convaincre tes enfants de lire sans mémoriser (Fra.) a book difficult to convince your children to read without memorize 10

In Romanian, both the subjunctive and the supine may be used: (i) a. e greu să răspundem la aceste întrebări is hard SUBJ answer-1.PL to these questions b. e greu de răspuns la aceste întrebări is hard de answer-SUP to these questions

Modal non-finite relatives in Romance

87

(59b) *o carte greu de convins pe elevi să (o) citească fără a reĠine (Rom.) a book difficult to convince OBJ pupils SUBJ it read without memorize (59c) a book hard to convince your children to read without memorize (Eng.)

(v) In Romanian, agent PPs are allowed for some speakers (cf. (23)): (60a) Japonia este greu de înĠeles de către cineva care nu locuieúte acolo. Japan is hard de understand-SUP by somebody who not lives there ‘Japan is difficult to understand for somebody who doesn’t live there.’ (www.businessmagazin.ro/arta-si-societate/lifestyle/soc-culturalinjaponia4073249 [02/09/2009]) (60b) PreĠurile […] par greu de “înĠeles” de către mulĠi dintre români. prices-the seem hard de understand-SUP by many among Romanians ‘The prices […] seem to be difficult to understand for many Romanians.’ (www.stiriauto.ro/articol/888/Romanii-se-orienteaza-catre-masini-din import [02/09/2009])

In order to explain the similarities observed between modal participials and the non-finite complement in TCs, we assume that TCs contain the same structure as modal participials, except for the Pred head which in their case appears above the tough-adjective.11 We adopt a raising analysis for TCs (for various arguments in favour of a raising analysis of TCs, see Berman 1973, Ruwet 1991, Goh 2000, Sportiche 2002, 2006, Hicks 2004; for arguments specific to the Romance family, see Giurgea & Soare 2006). (61a) [PredP DP [ Pred [AP tough [ Mood [T(non-fin.) [vpass ..tDP.. ]]]]]] (Fra. & Ita.) livres difficiles à lire ‘books tough to read’ (61b) [PredP DP [ Pred [AP tough [ Mood [vpass ..tDP.. ]]]] (Rom.) cărĠi greu de citit ‘books tough to read’

The main problem with this analysis is the meaning: with tough-adjectives the Mood head does not have the modal meaning found in modal participles. We believe however that even in this case there is a link between this head and modality: even if it does not introduce modality itself, it is selected by an element having a modal meaning, namely the tough-adjective. The toughadjective introduces quantification over possible worlds. There are two 11

Both the participial relative and the adjective-and-infinitive construction (i. e., the TC) are predicative (PredP), but obviously they have different structures. In TCs, the whole construction adjective + non-finite clause appears in PredP positions (i. e. after the copula and adnominally), so we must assume that the Pred head is inserted above the tough-adjective.

88

Ion Giurgea & Elena Soare

readings: a psychological one, in which an action is experienced as being difficult, and a purely modal one, stating that the worlds in which the action takes place are few. In the first reading (the psychological one), ‘tough’ takes an experiencer, and states that in all the possible worlds (with a circumstantial modal base and a stereotypical accessibility relation) where the experiencer attempts to do a certain action, this action will be felt to be difficult by him: (62) [[tough1]]=ȜF Ȝx Ȝw ∀w' ∈ fCIRC(w) ((w'≤ stereotyp w ∧ ∃e (attempt(e,x,F(x),w'))) → ∀e ((attempt (e, x, F(x),w') → tough-for (e, x,w'))) F=Ȝx Ȝw [∃e (read (e)(y)(x)(w))]

In the other reading, ‘tough’ does not take an experiencer, but is purely modal, stating that few of the worlds in which the agent attempts to perform a certain act are worlds in which this act takes place: (63) [[tough2]]= ȜF Ȝx Ȝw FEW w', w'∈fCIRC(w) ∧ w'≤ stereotyp w ∧ ∃e (attempt(e,x,F(x),w'))) (F(x)(w´))

We conclude that like other inflectional elements linked to modality, such as the subjunctive, the Mood head of modal participials is not always associated with a particular modal meaning. We may assume that when it is not selected by a modal element such as the tough-adjective, there is an abstract modal operator of deontic necessity or teleological possibility, which either selects it or is inserted in its specifier (in which case the agent PRO may be inserted in an outer specifier). In 2.2 we have seen that the deontic reading may be lost with prefixal negation in Romanian. This may indicate that the modal operator combines with the negation, which supports the idea that the modal operator is inserted in a different position than Mood0. In standard Romanian, TCs have a peculiar property which may indicate a different structure: unlike in the other Romance languages, the tough-adjective does not agree with the subject/head noun: (64a) Aceste ipoteze sunt greu/*grele de acceptat. (Rom.) these hypotheses(FEM) are difficult-MASC.SG/difficult-FEM.PL to accept (64b) Ces hypothèses sont faciles à admettre. (Fra.) these hypotheses(FEM) are difficult-FEM.PL to accept (64c) Queste ipotesi sono difficili da accettare. (Ita.) these hypotheses are difficult-PL to accept

There are two possible analyses for non-agreeing TCs: (i) We may assume that the clause is the external argument of the adjective (as proposed in Giurgea & Soare 2006). The difference between Romanian

Modal non-finite relatives in Romance

89

and the other Romance languages would be that in Romanian toughadjectives do not have two selectional patterns, with and without an internal clausal complement. The supine would occupy SpecPred, and the fact that it appears on the right of the adjective would be due to a linearization rule requiring extraposition for clausal subjects of predication (probably due to their heaviness). A problem is that the subject/head noun should raise directly from the object position of the clausal subject (should be extracted from SpecPred): (65) [aceste ipoteze] [sunt [PredP [Pred [AP greu ]] [MoodP de acceptat aceste ipoteze]]

(ii) We may assume that the tough-word occupies the specifier of MoodP. In this case it would no longer be an adjective, but a modal adverb, which would explain the lack of agreement. A similar analysis was proposed in Soare (2002). This analysis appears somehow more costly, because toughadjectives do not have a uniform syntactic behavior, as they are sometimes predicates, and sometimes modal operators. However, it is independently needed in other languages, namely in German and Dutch. These languages allow prenominal TCs (cf. (67)), although prenominal phrases must end in a head – west Germanic languages obey Emonds’ (1976) Head-FinalFilter in the nominal domain; as shown by Haider (2004), this rule is characteristic of head-initial structures. Since in German and Dutch noun phrases are head-initial, they obey the Head-Final-Filter (which forbids a pre-head modifier to end in anything but the head, cf. (66)). This suggests that the verb is not an argument of the adjective, but rather the head of the string tough + verbal projection. (66a) eine [viel größere (*als ich dachte)] Summe (Deu.) ‘a much bigger (*than I thought) sum’ (66b) ein [unzufriedener (*damit)] Syntaktiker ‘an unsatisfied (*it-with) syntactician’ (66c) ein [viel schöneres (*als ich dachte)] (neues) Haus ‘a [much-more beautiful (*than I thought)] (new) house’ (67) ein schwer zu lesendes Buch (Deu.) a tough to read-AGR book

Moreover, it is the verb and not the adjective which agrees, as shown in (67) above. In German, although in predicative, cf. (68a), and postnominal positions, cf. (68b), the tough-adjective combines with the infinitive, in prenominal position the verb takes the form of the present participle: infinitive + d + agreement, cf. (68c):

90

Ion Giurgea & Elena Soare

(68a) Dieser Philosoph ist schwer zu verstehen. this philosopher is tough to understand (68b) ein Buch, schwer zu lesen a book tough to read (68c) ein schwer zu lesendes Buch a tough to read-AGR book

This suggests that the verb is the lexical head of the TC, and that the toughadjective is probably in a Specifier position (Petra Sleeman, p. c.). Note that the combination zu + infinitive is only found in non-agreeing environments (in German adjectives and participles agree only in prenominal position). So, it is reasonable to assume that we have to deal with the same structure and -nd- is an allomorph used before the agreement markers. Note that the same pattern zu ...-nd-agreement / zu ...-n is found in modal participials, which are passive like in Romance (see Haider 1984, 1986): (69a) Das Buch ist zu lesen. the book is to read-n (69b) ein zu lesendes Buch a to read-nd-NEUT.SG.NOM book(NEUT)

The fact that a passive modal participial is also used in TCs gives indirect support to the analysis we have proposed for Romance.

5. Conclusion In this paper, we hope to have shown that Romance modal non-finite relatives are reduced relatives, in which the relativized position (which is the object position) is not case-marked; i. e., they are passive structures which have the syntactic behaviour of participials (therefore we labelled them ‘modal participials’). We analyzed the introductory element of modal participials as a functional head with modal properties, labelled Mood, which takes a non-finite TP complement in French and Italian and a vP complement in Romanian. We argued that this kind of Mood is also present in other syntactic environments – Tough Constructions, small clause complements of have and, in Romanian, the post-copular position in an impersonal construction. We proposed that in Romanian tough-adjectives in Tough Constructions are adverbials occupying the specifier position of Mood, which explains the lack of agreement. This analysis probably also holds for German and Dutch.

Modal non-finite relatives in Romance

91

References Avram, Larisa (1999): Auxiliaries and the Structure of Language. – Bucharest: Editura UniversităĠii Bucureúti. Baker, Mark C. (1988): Incorporation: A theory of grammatical function changing. – Chicago: University of Chicago Press. Bayer, Samuel (1990): “Tough movement as function composition.” – In: The Proceedings of the Ninth West Coast Conference on Formal Linguistics, 29–42. Stanford, Calif.: Stanford Linguistics Association. Berman, Arlene (1973): “A constraint on tough-movement”. – In: Papers from the Ninth Regional Meeting of the Chicago Linguistic Society, 34–43. Chicago: Chicago Linguistic Society. Bhatt, Rajesh (1999): Covert Modality in Non-Finite Contexts. – Diss., University of Pennsylvania. – (2002): “The Raising Analysis of Relative Clauses: Evidence from Adjectival Modification.” – In: Natural Language Semantics 10, 43–90. Bianchi, Valentina (1999): Consequences of Antisymmetry: Headed Relative Clauses. – Berlin: Mouton de Gruyter. Bowers, John (1993): “The syntax of predication.” – In: Linguistic Inquiry 24, 591– 656. Burzio, Luigi (1986): Italian Syntax. A Government-Binding Approach. – Dordrecht: Kluwer/Reidel. Chomsky, Noam (1981): Lectures on Government and Binding. – Berlin: Mouton de Gruyter. – (1995): The Minimalist Program – Cambridge, Mass.: MIT Press. – (2000): “Minimalist inquiries: the framework.” – In: Roger Martin, David Michaels & Juan Uriagereka (eds.): Step by step: essays on minimalist syntax in honor of Howard Lasnik, 89–155. Cambridge, Mass.: MIT Press. – (2001): “Derivation by phase.” – In: Michael Kenstowicz (ed.): Ken Hale: a life in language, 1–52. Cambridge, Mass.: MIT Press. – (2005): “On Phases.” – Ms., MIT. Dobrovie-Sorin, Carmen (1998): “Impersonal Se Constructions in Romance and the Passivization of Unergatives.” – In: Linguistic Inquiry 29, 399–437. Embick, David (1997): Voice and the interfaces of syntax. – Diss., University of Pennsylvania. Emonds, Joseph E. (1976): A Transformational Approach to English Syntax. – New York: Academic Press. Fischer, Olga (1992): “Syntactic change and borrowing: The case of accusative-andinfinitive construction in English.” – In: Marinel Gerritsen & Dieter Stein (eds.): Internal and external factors in Syntactic Change. Berlin: Mouton de Gruyter (Trends in Linguistics. Studies and Monographs 61).

92

Ion Giurgea & Elena Soare

Giurgea, Ion & Soare, Elena (2006): “When are Adjectives Raisers? Tough to get it.” – In: Proceedings of the 23rd Israel Association for Theoretical Linguistics Conference. (URL: http://linguistics.huji.ac.il/IATL/23/Giurgea-Soare.pdf [02/09/2009]) Goh, Gwang-Yoon (2000): “Is the tough-subject thematic?” – Ms., presented at the LSA Annual Meeting, Chicago. Grimshaw, Jane (1990): Argument Structure. – Cambridge, Mass.: MIT Press. Guasti, Maria Teresa (1996): “Semantic restrictions in Romance causatives and the incorporation approach.” – In: Linguistic Inquiry 27, 294–313. Haider, Hubert (1984): “Was zu haben ist und was zu sein hat – Bemerkungen zum Infinitiv.” – In: Papiere zur Linguistik 22, 47–100. – (1986): “Fehlende Argumente: vom Passiv zu kohärenten Infinitiven.” – In: Linguistische Berichte 101, 3–33. – (2004): “Pre- and postverbal adverbials in VO and OV.” – In: Lingua 114 (6), 779– 807. Hacquard, Valentine (2006): Aspects of Modality. – Ph.D. Diss., MIT. Hicks, Glyn (2004): "So easy to look at, So Hard to Define: Tough movement in the minimalist framework. MA dissertation (unpublished), University of York, September 2003. Homer, Vincent, Ishizuka, Tomoko & Sportiche, Dominique (2009): “The Locality of Clitic Placement and the Analysis of French Causatives.” – Ms., presented at LSRL 39, University of Arizona. Iatridou, Sabine, Anagnostopoulou, Elena & Izvorski, Roumyana (2001): “Some Observations about the Form and Meaning of the Perfect.” – In: Michael Kenstowicz (ed.): Ken Hale: A Life in Language. Cambridge, Mass.: MIT Press. Kayne, Richard S. (1975a): “French Relative que.” – In: Recherches Linguistiques, Université de Paris VIII 2, 40–36 / 3, 27–92. – (1975b): French Syntax. The Transformational Cycle. – Cambridge, Mass.: MIT Press. – (1994): The Antisymmetry of Syntax. – Cambridge, Mass.: MIT Press. Kim, Boomee (1995): “Predication in tough constructions.” – In: Proceedings of the Fourteenth West Coast Conference on Formal Linguistics, 271–285. Stanford, Calif.: Stanford Linguistics Association. Léger, Catherine (2006): Sentential Complementation of Adjectives in French. – Ph.D. Diss., Université du Québec à Montréal. Postal, Paul M. (1971): Cross-over phenomena. – New York: Holt, Rinehart & Winston. Rivero, María Luisa (1987): “La teoría de las barreras y las completivas del rumano.” – In: Violeta Demonte & Marina Fernández Lagunilla (eds.): Sintaxis de las lenguas románicas. 329–354. Madrid: Textos Universitarios. Roberts, Ian G. (1985): “Agreement parameters and the development of English auxiliaries.” – In: Natural Language and Linguistic Theory 3, 21–58. Rosenbaum, Peter S. (1967): The grammar of English predicate complement constructions. – Cambridge, Mass.: MIT Press.

Modal non-finite relatives in Romance

93

Ross, John R. (1969): “Auxiliaries as Main Verbs.” – In: William Todd (ed.): Studies in Philosophical Linguistics (Series I), 77–102. Evanston, Ill.: Great Expectations Press. Rouveret, Alain & Vergnaud, Jean-Roger (1980): “Specifying reference to the subject: French causatives and conditions on representations.” – In: Linguistic Inquiry 11, 97–202. Ruwet, Nicolas (1991): Introduction à la grammaire générative. – Paris: Plon. Soare, Elena (2002): Le supin roumain et la théorie des catégories mixtes. – Ph.D. Diss., Université de Paris VII. Soare, Elena & Dobrovie-Sorin, Carmen (2002): “The Romanian Supine and Adjectival Complementation. Tough Constructions.” – In: Bucharest Working Papers in Linguistics. Syntax and Phonology 4 (1), 75–87. Sportiche, Dominique (2002): “Movement types and triggers.” – In: GLOW Newsletter 48, 116–117. – (2006): “NP Movement: How to Merge and Move in Tough-Constructions.” – Ms., UCLA. Terzi, Arhonto (1993): Pro in Finite Clauses. A Study of the Inflectional Heads of the Balkan Languages. – Ph.D. Diss., City University of New York. Zubizarreta, María Luisa (1982): On the relationship of the lexicon to syntax. – Diss., MIT.

Gerda Haßler

Epistemic modality and evidentiality and their definition on a deictic basis

It has often been pointed out that there is some overlap between epistemic modality and evidentiality (Chafe & Nichols 1986, Cornillie 2007, De Haan 1999, Dendale & Tasmowski 2001, Plungian 2001, Squartini 2004). In this paper I would like to present a number of reflections on the necessity of drawing a boundary between modality and evidentiality. Starting from the typological category of evidentiality – extended here for use in pragmatic studies – I will then explore demarcation problems in Romance languages, which lack grammaticalized forms for expressing evidentiality. The underlying premise of this paper is that evidentiality as a marker of the origin of the speaker’s knowledge is connected to the speaker’s pragmatic stance. Because the perspective of the speaker is thus incorporated into the utterance, it seems appropriate to analyse the applicability of the deictic category. Finally, under the aspect of deixis, I shall attempt a demarcation between evidentiality and modality.

1. Evidentiality as typological category and its extension to include pragmatic functions Typologists assume that there is a universal inventory of grammatical and lexical categories, out of which each language makes a distinct selection. In about one fourth of all of the languages in the world, every utterance must indicate the type of source on which it is based (Aikhenvald 2004: 1). For typologists, the distinction between epistemic modality and evidentiality is unproblematic in so far as both are expressed by different morphological elements, which together can also be combined with a verb. Practically every paper on evidentially makes at least some reference to the evidentiality system in the language of the Tuyuca, which makes a five-way distinction: (1a) the direct personal and visual experience of the speaker, (1b) the perception through hearing, (1c) making deductions from evidence, (1d) learning from the report of another person and (1e) drawing a conclusion based on logic (cf. Barnes 1984):

96

Gerda Haßler

(1a) díiga apé-wi soccer play-3.PERS.PRET.VISUAL ‘He played soccer [I saw it].’ (1b) díiga apé-ti soccer play-3.PERS.PRET.NOT VISUAL ‘He played soccer [I heard it but didn’t see it].’ (1c) díiga apé-yi soccer play-3.PERS.PRET.INFERENCE ‘I have evidence that he played soccer, but I didn’t see it.’ (1d) díiga apé-yigi soccer play-3.PERS.PRET.REPORT ‘I was told that he played soccer.’ (1e) díiga apé-hƭyi soccer play-3. PERS.PRET.CONCLUSION ‘It can be logically assumed that he played soccer.’

The simple assertion that someone played soccer is not possible here without providing the source of this knowledge. But even where marking evidentiality is obligatory, the languages with evidentials differ – e. g. some languages have fewer evidential markers and only differentiate between ‘first hand’ and ‘not first hand’. Typologists view evidentials as linguistic means with the primary characteristic of ‘source of knowledge’, without implying a direct reference to the certainty and responsibility of the speaker or the truth of his assertion. However with this limitation, not only is the category of evidentiality inapplicable to European languages, but it also becomes problematic for everything that goes beyond the determination of elements with evidential core meaning.1 Evidential suffixes also characterise the caution with which a speaker handles the information. In this regard we can list the following maxims, which have a pragmatic character: 1. (Only) one’s own experience is reliable. 2. Unnecessary risks are to be avoided, for example assuming the responsibility for information about which one is not absolutely sure. 1

For further studies on evidentiality cf. Aikhenvald (2004), Aikhenwald & Dixon (2003), Chafe & Nichols (1986), Dendale & Tasmowski (1994, 2001), Hoff (1986), Ifantidou (2001), Lazard (2001), Mushin (2001), Nuyts (2001), Plungian (2001), Willett (1988), Bybee, Perkins & Pagliuca (1994), Haßler (2001, 2003), González Vázquez (2006); and with a consciously chosen contrasting term (médiatisation) Guentchéva (1996).

Epistemic modality and evidentiality and their definition on a deictic basis

97

3. One may not be gullible. 4. Responsibility can only be assumed for what is certain. An example showing that evidential markers assume the pragmatic function of marking the uncertainty of the speaker even in languages with pure evidentials is the use of the suffixes -mi/-shi/-chi in the Huallaga dialect of Quechua, which allows the speaker to assume or reject responsibility for the content of what is said. With -mi the speaker assumes responsibility, with -shi he defers it to someone else, with -chi he indicates that no responsibility can be taken for this type of content. According to Weber (1989: 421) with (2a) the speaker expresses his certainty, with (2b) he refers to a report from someone else and with (2c) to a possibility. (2a) Wañu-nqa-paq-mi. it will die [I assert that]evidential ‘[I assert that] it will die.’ (2b) Wañu-nqa-paq-shi. it will die [I was told] evidential ‘[I was told] it will die.’ (2c) Wañu-nqa-paq-chi. it will die ‘[perhaps] it will die.’

[Perhaps] evidential

The use of the evidentiality marker -shi goes far beyond the expression of second-hand information. It is also used in the so-called narrative past, for which one does not wish to assume responsibility. The Quechua culture places great value on avoiding gullibility, which the use of this evidential marker definitely reveals. A mixture of morphologically conceived evidentials with pragmatic circumstances seems apparent as soon as one goes beyond typological issues to look at the use of these elements.

2. Demarcation problems in Romance languages If we turn to the situation in Romance languages, drawing a boundary between epistemic modality and evidentiality presents problems that are difficult to resolve. Even if we assume that there are linguistic elements which fulfil the original function of marking the source of the speaker’s knowledge, this is contingent with the term epistemic modality. Hence, the term ovviamente

98

Gerda Haßler

‘obviously’ in example (3), an excerpt from the Italian press, is not what the author has seen directly but rather a personal conclusion from what he has heard, to which a possible counterargument is immediately considered with the use of eppure. (3)

Ciampi, ovviamente, non fa riferimenti diretti alle polemiche scatenate dalle dichiarazioni del premier, eppure nelle sue parole si legge un appello all’indipendenza dell’informazione rispetto all’esecutivo. ‘Ciampi obviously does not make any direct reference to the polemic incited by the statements of the prime minister, and yet in his words one can read an appeal to the independence of information with respect to the executive branch.’ (La Repubblica, 19th April 2002)

Removing ovviamente makes the sentence more definite. Based on its original lexical meaning, the adverb would have to underscore the obviousness of the conclusion. Still greater indefiniteness is expressed in sentence (4) where a conclusion oriented on the future is formulated. (4)

Un argomento di cui, ovviamente, si parlerà in studio. ‘An argument which will apparently be discussed in the studio.’

The lexical meaning of ovviamente must have undergone changes for it to be used as the modalisation of a future event that is considered here to be only hypothetical. With this meaning the adverb can only conditionally be traced back to ovvio (‘what comes about with direct necessity and without any possibility of doubt on the level of interpretation or judgement’, Devoto & Oli (1990: 1313)), in so far as the characteristics ‘direct visibility of the proof’ and ‘exclusion of error’ have diminished in importance. Ovviamente is a sentence adverb which is used parenthetically in the two examples. It comes before the conjugated verb form and is separated by pauses from the rest of the sentence. Essentially, when the adverb is used in this way it contains an additional predicate; it can thus be transformed into a main clause, where the sentence in which it appears contains the argument: (3')

È ovvio che Ciampi non fa riferimenti diretti alle polemiche scatenate dalle dichiarazioni del premier. ‘It is obvious that Ciampi did not make a direct reference to the polemic incited by the statements of the prime minister.’

(4')

È ovvio che è un argomento di cui si parlerà in studio. ‘It is obvious that this is an argument which will be discussed in the studio.’

Whereas in the examples with the predicate adjective ovvio the author makes a clear statement about the superficiality and plausibility of his conclusion, in the examples using the adverb ovviamente he merely marks the conclusion as

Epistemic modality and evidentiality and their definition on a deictic basis

99

such. No explicit reference is made to what was behind it or what it was based on. The use of the adverb ovviamente allows the sources of the reported knowledge to remain undefined, but however allows the speaker to characterise them as given. The Spanish adverbs visiblemente ‘visibly’, aparentemente ‘apparently’ and evidentemente ‘evidently, obviously’, which in their direct meaning denote visual access to the content of the proposition, appear to behave in a similar way. However, whereas this access for visiblemente is explicit, aparentemente and evidentemente can also denote conclusions which are not based on visual perception (cf. Haßler 2004): (5)

La suma de estos quebrantos – que se hizo evidente cuando el gobierno debió renegociar su gigantesca deuda externa – ya obligó al Presidente Figueiredo a someterse a una delicada operación de by-pass y lo ha afectado visiblemente en lo físico. ‘The sum of these losses, which became evident when the government had to renegotiate its gigantic foreign debt, had already compelled President Figueiredo to undergo a serious by-pass operation and had visibly affected him physically.’ (Hoy, 25/04–01/05/1984)

(6)

Esta confianza estaba evidentemente fundada en la disposición general de todos los españoles, que guiados por el instinto de la felicidad, que el autor de la naturaleza puso en el corazón de los hombres, sabían que no había otro camino para que se mejorase la suerte de la España, que el de cambiar las instituciones, ni otro medio de conseguirlo que por un alzamiento militar. ‘This trust was evidently based on the general disposition of all Spaniards who – guided by the instinct of happiness which the creator of nature has placed in the human heart – knew that there was no other way for the fate of Spain to turn positively than to change the institutions, and there was no other means to achieve that than by a military uprising.’ (El Imparcial, 1st February 1822)

(7)

Y, de igual manera que en el Estado jurídico se destacan lazos legislativos que descubren y ordenan las relaciones en todo el ámbito nacional, porque a todos protegen y a todos obligan las relaciones aparentemente invisibles de la colectividad, de la misma manera en el ámbito económico nacional hay que descubrir también las interrelaciones económicas entre los distintos sectores, […]. ‘And in the same way that in the lawful state legislative ties dominate which uncover and organise the relationships on the entire national level because they protect everyone and oblige everyone to uphold the apparently invisible relationships of the group, in the same way on the national economic level one must also uncover the economic interrelationships between the different sectors […].’ (Contabilidad Nacional, ABC, 11th July 1958)

As example (7) shows, aparentemente is readily combinable with elements which negate the visibleness (aparentemente invisibles).

100

Gerda Haßler

In example (8) aparentemente is used autonymically in an oral interview and indicates the process of becoming conscious of its use. At first the speaker uses it automatically (David es un niño supernormal, aparentemente) after which she becomes aware that her speech act is not a conclusion, but a simple statement. For this reason she corrects herself with words such as bueno, aparentemente y sin aparentemente, es supernormal: (8)

El segundo se llama David. David es un niño supernormal, aparentemente, bueno, aparentemente y sin aparentemente, es supernormal y tiene ahora pues yo creo que veintiséis veinticinco, bueno, nunca me aclaro con estos dos. ‘The second one is named David. David is apparently a supernormal child, well, apparently and not apparently, he is supernormal, and now I think he’s twentysix or twenty-five, I’m not sure which of the two.’ (1998, Entrevista CSC008, woman, aged 20 years, untranscribed fragments in the text body for the study of Spanish)

These examples also illustrate pragmatisation: Whenever the speaker seems motivated to explicate evidence, a reduction of the degree of evidentiality takes place as well. It is not that the content of the utterance is apparent, not that it is visible, but rather the non-presence, the not-mentioning of the information source which is decisive for the use of the adverbs. Analogously, in the Italian si dice one can ascertain that in all syntactic usages the mentioning of a speech act is not of primary importance. In example (9) the statement gained from hearsay is directly questioned and contrasted with a position in a direct quotation. (9)

Si dice che la quotazione in Italia sia propedeutica a nuove acquisizioni. È vero? “Noi ci guardiamo sempre attorno. Luxottica continua a crescere per vie interne, ma a volte ci sono delle occasioni sul mercato che possono consentire di fare dei grandi salti in avanti. Noi vogliamo essere pronti”. ‘It is said that having quotas in Italy paves the way for new achievements. Is this true? “We always keep our eyes open and look around. Luxottica is continuing to grow internally, but sometimes there are occasions on the markets which allow great leaps forward. We want to be ready.”’ (La Repubblica, 2nd December 2000)

The meanings of these elements are not referentially definable; on the contrary, they are more procedural in character2 and thus are crucial for the processing of the utterance in understanding speech. The instruction arising from the use of markers of the vague origin of the speaker’s knowledge raises difficult questions concerning the relationship between the content of the proposition and its logical value. If a proposition is really evident, it does not require a marker to show that it is. From this simple fact the conclusion can be 2

Ifantidou (2001: 87); Sperber & Wilson (1995) use the term procedural meaning.

Epistemic modality and evidentiality and their definition on a deictic basis

101

drawn, with reference to Grice’s conversation maxims,3 that evidentiality markers can be interpreted as relevant in the sense that they are always claimed to have a limiting effect with respect to the actual evidence. This also applies when the respective words would denote a maximal degree of evidence based on their lexical original lexical meaning. Finally, the use of certain strategies which also serve as evidential markers can even contribute to the blurring of the different kinds of evidentiality. Thus, the French conditionnel can express reported speech, but also a conclusion. In example (10) already used by Guentchéva (1994: 17f.) it is impossible to distinguish whether the statement of the cause of death was concluded from the diagnosis that was given beforehand or if it reports what the doctor said: (10) Les résultats des examens réalisés, notamment à l’hôpital neuro-cardiologique de Lyon, par le docteur T., neuro-cardiologue, et par le professeur V., toxicologue, font état de la présence dans le sang, où le taux d’alcoolémie atteignait 1,8 gramme, d’opiacés, de la morphine en particulier. La cause de la mort serait ainsi une crise cardiaque déclenchée dans un contexte de prise d’opiacés par voie buccale qui ne semble pas devoir être assimilé à une “surdose”. ‘Notably, the results of the examinations carried out in the Neuro-Cardiological Hospital of Lyon by the neurocardiologist Doctor T. and by the toxicologist Professor V. reveal narcotic drugs, in particular morphine, in the blood along with an alcohol concentration of 1.8. The cause of death would thus be a cardiac crisis triggered in the context of the intake of narcotics through the mouth, which does not seem to be linked to an “overdose”.’ (Le Monde, 17th June 1993)

It can only be unequivocally determined in this example that it is not an utterance based on the own experience of the producer of the text and his/her insight into the matter. As we have seen, in the Romance languages there are problems in demarcating epistemic modality and evidentiality as well as some overlap between the different kinds of evidentiality. With the determination of the category of evidentiality as marker of the origin of the speaker’s knowledge evidentiality becomes a category centred on the speaker. It therefore seems justified to include deixis in the definition of evidentiality.

3

Sperber & Wilson (1995: 78): “We take for granted that there is a good match between the strength of our assumptions and the likelihood that they are true. That is, we trust our cognitive mechanisms to strengthen or weaken our assumptions in a way that is epistemologically sound: we trust our representation of the world to be adequate in this respect, as in others. As a result, intuitions about the strength of our assumptions are expressed as intuitions about the degree of confirmation. Such intuitions are assumptions about assumptions, and can be processed as such.”

102

Gerda Haßler

3. Modality, evidentiality and deixis Evidentiality is closely related both to the reference to the source of information on the one hand and on the other to the epistemic stance of the speaker. In this sense it seems appropriate to view evidentiality as a deictic phenomenon, in so far as the evidentiality markers refer to elements outside the linguistic context, namely to the source of information and to the speaker who has access to it. The deictic character of evidentiality has already been pointed out in previous papers (e. g. Jakobson 1957, Schlichter 1986, Frawley 1992, Volkmann 2005). We take deixis to be a specific mechanism of referentialisation which refers to the context of the utterance and establishes a relationship to ego-hic-nunc. If we apply the categories suggested by Frawley (1992) regarding deictic classification of the epistemic area (deictic centre: ego or alter, directionality: coming from X, going toward X), the result is four subcategories: (1) coming from ego (inference), (2) going toward ego (visual perception scale, auditory, other senses), (3) coming from alter (quote, report, rumour…), (4) going to alter (one to all possible conversation partners) (cf. Wachtmeister Bermúdez 2006: 25–26). Scheme 1: Deictic centre of the speaker (ego) and the other communication partner(s) (alter). Source of Information

Strength of Information

ego outgoing incoming

degree of inference necessary > possible degree of perception visual > auditory > other sense perceptions

alter outgoing incoming

degree of outside information quotation > report > rumour > others degrees of participants the other > the remaining

To begin with, a distinction is drawn between the deictic centre of the speaker (ego) and the other communication partner(s) (alter). In all cases there is a gradual progression in the strength of information. While the information perceived by the speaker via his or her own senses is strongest and most secure when it is perceived visually, it is somewhat weaker when perceived by hearing and still weaker when perceived by the other senses. A similar gradation occurs with respect to the communication participants: The

Epistemic modality and evidentiality and their definition on a deictic basis

103

information received from the immediate communication partner (the other) is stronger than the information received from the remaining indirect communication partners. Recognising the deictic character of evidentiality and describing it as a deictic phenomenon helps to determine the place of evidentiality within the language and provides a basis for distinguishing it from epistemic modality. Deixis implies a perspectivation from the standpoint of the speaker, which the hearer, outgoing from his/her own discursive position, has to reinterpret. Epistemic modality and evidentiality act together in the construction of the reliability of the utterances. While the epistemic modality monologically contributes the epistemic stance of the speaker or the author, evidentiality can also require the hearer or reader to produce a reference to the source of information, which means a necessary adaptation with respect to his/her identity, individuality and epistemic stance toward the sources. In the last of these – the epistemic stance – he/she may also differ from the producer of the utterance. First, a model for temporal relationships shall serve as a visualisation which we will then transfer to evidentiality. Temporal deixis can be represented as a scheme with three factors in which the points of reference would be moments, the deictic centre would be the prototype of the speech moment, the directions on the time axis would be oriented on the future or the past; and alongside that there would be a distance between the point of reference and the event time. The Spanish adverb después could be represented as an extension of the time axis beyond the point of reference. The following scheme illustrates the orientation of the time axis toward the future starting from the deictic centre of the speaker. The event time assigned to después is not characterised as having sharp contours, whereas the point of reference as such is precisely determined (cf. Reichenbach 1947: 290; Wachtmeister Bermúdez 2006: 29): Scheme 2: después. Time

Deictic centre (Speech moment)

Point of reference X

Event time

A use of this scheme of the point of reference, of the distance and of the direction on evidentiality must take into consideration the non-discrete

104

Gerda Haßler

character of evidential factors and access to the information source. Besides, pure modal meanings must be excluded in order to adequately describe and classify evidential values. Points of reference here would be the participants in the communication act with or without access to the source of information; their access to information can take place via sensory perception or cognitive operations, such as reasoning. The distance to the information source would be represented by the continuum between own and foreign information and the mode of access to information would be represented by the access direction, which can be universal or limited. Scheme 3: Distance to the information source. mode of access to distance to the information information source cognitive foreign

sensory

own

access direction to information universal

limited

Evidentiality is thus not merely a reference to the source of information reported in an utterance, but also a deictic phenomenon which refers to the speaker and his/her complex relationships to information and its sources. The following example of the imperfect tense in Spanish illustrates its evidential value in the deictic sense: A teacher is teaching mathematics to a student and in solving an equation produces one of the following utterances: (11) … y simplificando nos da entonces 1,5 por la raíz cuadrada de 2, y la raíz cuadrada de 2 era 1,4142, así que el resultado final… ‘[…] and the square root of 2 was 1.4142 […]’ (12) … y simplificando nos da entonces 1,5 por la raíz cuadrada de 2, y la raíz cuadrada de 2 es 1,4142, así que el resultado final… ‘[…] and the square root of 2 is 1.4142 […]’

In utterance (11) the teacher emphasises that both he and the student have access to the source of information, in this case to the numerical value of the square root of 2. In utterance (12) however, the teacher presents this information as new for the student. It has long been known that the temporality of the imperfect and also the German preterite can also depend on deicticals. Let us look at the Spanish translation of the example Käte Hamburger used back in 1957 (cf. Hamburger 1957):

Epistemic modality and evidentiality and their definition on a deictic basis

105

(13) Pero por la mañana tenían que adornar el árbol. Mañana era navidad. ‘But in the morning she had to decorate the tree. Tomorrow was Christmas.’

I believe that the anaphoric character of some of the verb forms provides an explanation for the presence of the imperfect in this sentence (García Fernández & Camus Bergareche 2004: 54). With reference to temporality, the verb form era is coreferential with the deictical mañana, which is contingent on its anaphoric character. Deictic elements tend to assign a value to a verb form which is not their prototypical value. In many cases it is a matter of marking potential or at least not real processes. The aspectual value of imperfectivity approaches uncertainty. With the use of the imperfect the speaker can express that he has uncertain, i. e. second-hand information regarding the matter. For example, sentence (14) could be reformulated using the evidential value of the imperfect as in sentence (15): (14) Juan viene mañana, según me anunciaron. Juan comes tomorrow following me announced ‘Juan is coming tomorrow, according to what was announced to me.’ (15) Juan venía mañana. Juan came tomorrow ‘Juan was coming tomorrow.’

With the use of the imperfect venía the speaker is not referring to the coming of Juan taking place in the past, but to a heard or read utterance, in which the coming was announced. In such cases, which are relatively frequent in spoken Spanish, the speaker shows that he cannot assume full responsibility for the content of the utterance and he pre-empts the ability of the hearer to classify the information as from a foreign source. Another case in which deixis can be used to demarcate evidentiality and modality is the contrast between the consecutive connectors así que, de ahí que and por eso: (16) Me trataron mal, así que me fui. Me [they] treated badly, so that me [I] left ‘They treated me badly so I left.’ (17) Me trataron mal, de ahí que me Me [they] treated badly, of here that me [I] ‘They treated me badly which is why I left.’ (18) Me trataron mal, por eso me Me [they] treated badly, for that me ‘They treated me badly, therefore I left.’

haya ido. had gone.

fui. [I] left.

106

Gerda Haßler

In all three utterances the speaker asserts that the fact that he left is a consequence of his having been treated badly. In sentence (16) he presents the fact that he left as something new for the hearer: ‘They treated me badly so I left.’ and in (17) as information to which the hearer already has access: ‘They treated me badly which is why I left.’ In (18) both interpretations are possible. In conclusion it can be said that evidentiality is not simply a curiosity of a language group which is typologically far removed from the European languages, but rather a fundamental characteristic of languages which finds its place more in perspectivation than in epistemic modality, but which naturally shows some overlap with the latter. Evidentiality is not only a reference to an information source, but is also a deictic phenomenon which refers to the speaker who conceived the utterance and to his complex relationship with the information and its sources. While epistemic modality adds the stance of the text producer to the predication, evidentiality presumes the production of a reference to the source of knowledge by the recipient and thus a judgement of trustworthiness. In this paper I have identified problems of demarcation between evidentiality and modality in the Romance languages. These are particularly evident in the modal use of adverbials which in their original lexical meaning refer to the source of the speaker’s knowledge from direct visual perception/ observation. The development ultimately leading to a reduction of probability when evidential markers are used is known in linguistics as pragmatisation. Evidentiality can be considered a deictic phenomenon insofar as when it occurs, a reference to the speaker and to the speaker’s source of knowledge is made. In cases of evidentiality this extra-linguistic reference can also include the hearer as well as the speaker. In the Romance languages conjunctions may also serve as markers for such evidential meanings.

References Aikhenvald, Alexandra (2004): Evidentiality. – Oxford: Oxford University Press. Aikhenvald, Alexandra & Dixon, Robert M. W. (eds.) (2003): Studies in Evidentiality. – Amsterdam/Philadelphia: Benjamins. Barnes, Janet (1984): “Evidentials in the Tuyuca verb.” – In: International Journal of American Linguistics 50, 255–271. Bybee, Joan, Perkins, Revere & Pagliuca, William (1994): The evolution of grammar. Tense, aspect, and modality in the languages of the world. – Chicago/London: University of Chicago Press. Chafe, Wallace & Nichols, Johanna (eds.) (1986): Evidentiality: the linguistic coding of epistemology in language. – Norwood, N.J.: Ablex.

Epistemic modality and evidentiality and their definition on a deictic basis

107

Cornillie, Bert (2007): Evidentiality and Epistemic Modality in Spanish (Semi-) Auxiliaries. A Cognitive-Functional Approach. – Berlin/New York: de Gruyter De Haan, Ferdinand (1999): “Evidentiality and Epistemic Modality: Setting Boundaries.” – In: Southwest Journal of Linguistics 18, 83–101. Dendale, Patrick & Tasmowski, Liliane (eds.) (1994): Les sources du savoir et leurs marques linguistiques. – Paris: Larousse (= Langue française 102). – (eds.) (2001): Evidentiality. – Amsterdam: Elsevier (= Journal of Pragmatics 33, 339–464). Devoto, Giacomo & Oli, Gian Carlo (1990): Il dizionario della lingua italiana. – Firenze: Le Monnier. Frawley, William (1992): Linguistic semantics. – Hillsdale: Erlbaum. García Fernández, Luís & Camus Bergareche, Bruno (eds.) (2004): El pretérito imperfecto. – Madrid: Gredos. González Vázquez, Mercedes (2006): Las fuentes de la información: Tipología, semántica y pragmática de la evidencialidad. – Vigo: Universidade de Vigo. Guentchéva, Zlatka (1994): “Manifestations de la catégorie du médiatif dans les temps du français.” – In: Langue française 102, 8–23. – (ed.) (1996): L’énonciation médiatisée. – Louvain/Paris: Peeters. Hamburger, Käte (1957): Die Logik der Dichtung. – Stuttgart: Klett. Haßler, Gerda (1998): “Die Markierung der Herkunft des Sprecherwissens – ein grammatisches und pragmatisches Problem.” – In: Udo L. Figge, Franz-Josef Klein & Annette Martínez Moreno (eds.): Grammatische Strukturen und grammatischer Wandel im Französischen. Festschrift für Klaus Hunnius zum 65. Geburtstag, 169– 191. Bonn: Romanistischer Verlag. – (2001): “Kontrastive und typologische Überlegungen zur epistemischen Modalität in den romanischen Sprachen und im Deutschen.” – In: Gerd Wotjak (ed.): Studien zum romanisch-deutschen und innerromanischen Sprachvergleich. Akten der IV. Internationalen Tagung zum romanisch-deutschen und innerromanischen Sprachvergleich (Leipzig, 7.10. –9.10.1999), 169–184. Frankfurt a. M.: Lang. – (2002): “Evidentiality and reported speech in Romance languages.” – In: Tom Güldemann & Manfred von Roncador (eds.): Reported Discourse. A meeting ground for different linguistic domains, 143–172. Amsterdam/Philadelphia: Benjamins (Typological Studies in Language 52). – (2003): “Epistemic modality revisited: evidential functions of lexical and grammatical forms in Romance languages.” – In: Eva Hajiþová, Anna KotČšovcová & JiĜí Mírovský (eds.): Proceedings of the XVII International Congress of Linguists. Prague, July 24–29 2003. CD-ROM. Praha: Maftyzpress. – (2004): “El uso evidencial de adverbios modales.” – In: Juan Cuartero & Gerd Wotjak (eds.): Algunos problemas específicos de la descripción sintácticosemántica, 229–244. Berlin: Frank & Timme. Hoff, Berend Jacob (1986): “Evidentiality in the Carib: Particles, affixes and a variant of Wackernagel’s Law.” – In: Lingua 69, 49–103. Ifantidou, Elly (2001): Evidentials and Relevance. – Amsterdam/Philadelphia: Benjamins.

108

Gerda Haßler

Jakobson, Roman (1957/1971): “Shifters, verbal categories and the Russian verb.” – In: Roman Jakobson: Selected Writings. Vol. 2, 130–147. The Hague: Mouton. Lazard, Gilbert (2001): “On the grammaticalization of evidentality.” In: Journal of Pragmatics 33, 359–367. Mushin, Ilana (2001): Evidentiality and Epistemological Stance. Narrative Retelling. – Amsterdam/Philadelphia: Benjamins. Nuyts, Jan (2001): “Subjectivity as an evidential dimension in epistemic modal expressions.” – In: Journal of Pragmatics 33, 383–400. Plungian, Vladimir A. (2001): “The place of evidentiality within the universal grammatical space.” – In: Journal of Pragmatics 33, 349–357. Reichenbach, Hans (1947): Elements of Symbolic Logic. – New York: Free Press. Reyes, Graciela (1990): “Tiempo, modo, aspecto e intertextualidad.” – In: Revista Española de Lingüística 20, 17–53. – (1994): Los procedimientos de cita: citas encubiertas y ecos. – Madrid: Arco Libros. Schlichter, Alice (1986): “The origins and deictic nature of Wintu evidentials.” – In: Wallace Chafe & Johanna Nichols (eds.): Evidentiality: The Linguistic Coding of Epistemology, 46–59. Norwood, N. J.: Ablex. Sperber, Dan & Wilson, Deirdre (21995): Relevance. Communication and Cognition. – Oxford/Cambridge: Blackwell. Squartini, Mario (2004): “Disentangling evidentiality and epistemic modality in Romance.” – In: Lingua 114, 873–895. Volkmann, Gesina (2005): Weltsicht und Sprache. Epistemische Relativierung am Beispiel des Spanischen. – Tübingen: Narr (Tübinger Beiträge zur Linguistik 481). Wachtmeister Bermúdez, Fernando (2006): Evidencialidad. La codificación lingüística del punto de vista. – Ph.D. Diss., Stockholms universitet. Weber, David J. (1989): A grammar of Huallaga (Huánuco) Quechua. – Berkeley: University of California Press (University of California Publications in Linguistics 112). Willett, Thomas (1988): “A cross-linguistic survey of the grammaticalization of evidentiality.” – In: Studies in Language 12 (1), 51–97.

Mario Squartini

Where mood, modality and illocution meet: the morphosyntax of Romance conjectures*

1. Lyons’ dubitative mood John Lyons’ (1977: 725–849) comprehensive treatment of modality, even though deemed by its own author as a “lengthy, and at times tortuous, argument” (Lyons 1977: 841), has authoritatively influenced more recent reappraisals and was basically echoed in Palmer (1986: 21–23). As mentioned by Bybee et al. (1994: 176), Lyons’ influence on Palmer can clearly be observed in their common definition of modality, but is also particularly apparent in their comparable conceptions of the relationship between modality and moods, which they both take to be a mutual interplay of semantics and morphosyntax. Nevertheless, Palmer (1986) appears generally sceptical about other significant aspects of Lyons’ model, especially as far as the interaction between illocutionary acts and morphosyntactic realizations is concerned. In this respect Palmer, a more empirical data-driven typologist compared to the philosophical approach of Lyons (1977), points out several inconsistencies, in some cases elaborating on issues that Lyons himself presented as debatable. Apparently, one of the thorny empirical questions that both considered a concern is “the problem of Interrogative” (Palmer 1986: 94), which is also effectively summarized by Lyons (1977: 748) when he wonders why there is no evidence supporting the existence of a “distinct mood that stands in the same relation to questions as the imperative does to mands”. In fact, Palmer (1986: 30–31, 78–81) does provide evidence of interrogative moods in some of the world’s languages but these were, at least then,1 too sparse and * 1

I am extremely grateful to Hardarik Blühdorn, Martin Becker and Eva-Maria Remberger for their valuable comments on previous drafts. Due to more accurate typological descriptions (Dryer 2005), we are nowadays in a position to consider the distribution of interrogative verbal morphology in polar questions as less sparsely attested than in Palmer’s (1986) description. However, in comparing the WALS maps the distribution of morphological Imperatives and Interrogatives (van der Auwera & Lejeune 2005 and Dryer 2005) still seem to be skewed in favour of the former. The comparison may possibly be biased by the unclear boundary between question particles and proper interrogative affixes (Dryer 2005: 470), which might lead to the underestimation of the occurrence of inflectional affixes.

110

Mario Squartini

controversial to be compared to the robust evidence supporting the Imperative. In Lyons’ and Palmer’s perspectives this is a very general problem crucially impinging on the controversial relationship between mood and modality. In a sense, the gap between the Imperative and the Interrogative constitutes a more radical imbalance between deontic and epistemic modality. Given that deontic modality has its own prototypical mood – the Imperative, defined by Palmer (1986: 108) as the “unmarked member of the deontic system” – the Interrogative might have been the prototypical representative of epistemic modality (if it had been more robustly documented). This would have confirmed the systematic relation between illocutionary acts (questions) and modal functions, which is what is required in Lyons’ general architecture, heavily dependent on Searle’s speech acts theory. This focus on the connections between mood, modality and illocution is most probably one of the reasons why Lyons (1977: 753–768) insists on separating questions from mands, thus assuming that the former may have a dedicated expression in the interrogative mood. However, Lyons also tries to envisage a more general solution in which not only questions but also epistemically modalized utterances may be represented by the same mood. In passing he suggests that: it would not be unreasonable to expect that what is basically a dubitative mood might be regularly used both for posing questions and expressing doubts and uncertainty. (Lyons 1977: 748, fn. 14)

Lyons is here entertaining the possibility that the epistemic counterpart of the Imperative might not be the Interrogative, but a more comprehensive grammatical category that he calls the ‘dubitative mood’, using a label that can already be found in Jespersen (1924: 321) and occasionally occurs in grammatical descriptions of some languages (see Palmer 1986: 78, 81). Apparently, Jespersen (1924: 321) intended the Dubitative as merely covering weak conjectures (He may be (is perhaps) rich), while in Lyons’ definition this mood is conceived as the grammatical marking of any “dubitative, or epistemically qualified, utterances”, which in his view crucially includes questions. As mentioned above, this would consistently fit in with the general model in which Lyons interprets all the connections between illocutionary marked utterances and grammatical moods: similarly to mands, whose illocutionary force corresponds to the Imperative, the illocutionary force of questions and “dubitative utterances” can be grammatically expressed by a separate mood (the Dubitative).2 Ultimately, this would also bridge the gap 2

It must be remembered that Lyons (1977: 745) defines ‘statements’ as “modally unqualified, or categorical, assertions”, i. e. “statements that are unqualified in terms

Where mood, modality and illocution meet

111

between modality and moods: if the Imperative is the grammatical expression of deontic modality, epistemicity also has its own mood, which is the Dubitative. As it is presented by Lyons, the point is totally speculative and this attracts Palmer’s criticism. According to Palmer (1986: 81), Lyons’ proposal is not only unproven, but also intrinsically circular, especially as far as the mutual relationship between the Dubitative and the Interrogative is concerned: There is no obvious reason to argue that it is the Dubitative used for questions rather than the Interrogative used for expressing doubt, and there is a danger that Lyons’ claim is true only by definition (that a morphological (mood) category must be Dubitative rather than Interrogative).

This point might seem merely terminological, but in fact what is at stake here is also the connection with epistemic modality, and this is why Lyons is firm in claiming that it is the Dubitative which encroaches on the proper domain of the Interrogative rather than vice-versa. Since the Dubitative is intrinsically epistemic by definition, this guarantees a direct connection to epistemicity, while the epistemic nature of questions is the speculative and highly-debated hypothesis on which Lyons builds his argument. Thus, the controversy surrounding the Dubitative, which at first sight might seem a marginal question, actually involves the proper relationship between modality and illocution (Hengeveld 2004: 1191–1192). As Cresti (2002) and Pietrandrea (2005: 19–24) demonstrate when they plead for a neat distinction between modality and illocution, the issue is still the subject of debate, possibly because of the authoritative influence of Lyons’ model, which is often tacitly and uncritically adopted, and despite the efforts made to clarify this point in different functionalist approaches (see in particular the Functional Grammar treatment in Hengeveld 1987, 1988). In this general perspective, the empirical consistency of Lyons’ dubitative mood remains crucial, the basic question being whether Lyons’ argument is solely speculative, as Palmer maintains, or whether the Dubitative can be supported by actual data without risking a circular identification of modality and illocution. This is precisely the issue that will be addressed at the outset of this article: drawing on French and Italian data it will be shown that Romance languages may provide evidence of a dubitative mood that, despite occurring in questions, cannot be confused with an interrogative mood. Obviously, this may respond to Palmer’s objection, but demonstrating the occurrence of a dubitative mood is not tantamount to accepting Lyons’ general idea, i. e. the unification of questions and episteof possibility and necessity”. Therefore, epistemically marked utterances cannot be considered as genuine statements and must be lumped together with other illocutionary types.

112

Mario Squartini

mically modalized utterances under the general umbrella of epistemicity, which might lead to a misleading and oversimplifying representation of the boundaries between modality and illocution. In this respect, the Romance data discussed below will confirm that the illocutionary force of questions should not be confused with epistemic modality, but, at same time, it will be demonstrated that interrogative sentences constitute a special morphosyntactic environment in which given moods may be required in order to express epistemic conjectural meanings. Thus, even though it has neither dedicated illocutionary acts nor special moods, epistemic modality may surface in regular combinations of given morphosyntactic constructions and verb forms (moods), suggesting that the relationship between modality and mood is probably less biunivocal (Palmer 1986: 24) than Lyons assumed, but at the same time admits some morphosyntactic regularities in which sentence types and moods cluster together. In this respect, the relationship between epistemic modality and its morphosyntactic expressions will be proven to be extremely varied in the Romance verb system, giving rise to several combinations and offering the opportunity to describe in better detail the relationships between forms and functions in a significant area of the epistemic domain, which will go beyond interrogative sentences. It will be shown that, apart from the Dubitative, other regular associations of morphosyntactic structures and given moods permit two epistemic subdomains to be distinguished: the inferential and the conjectural, which will eventually provide a more fine-grained classification of epistemic meanings in interrogative as well as in declarative sentences.

2. The dubitative mood in French and Italian As mentioned above, Palmer (1986: 81) criticized Lyons’ argument on the dubitative mood due to its circularity: Why should a form occurring in questions as well as in epistemically modalized declarative sentences be considered a Dubitative and not an Interrogative, the usage of which is extended to epistemic utterances? According to Palmer, Lyons’ conclusion is only speculative and cannot therefore be solved on an empirical basis. On the contrary, the data presented in this section aim to demonstrate that French and Italian provide empirical evidence of what can confidently be considered as a dubitative mood without risking circularity.

Where mood, modality and illocution meet

113

In the French verb system an inflectional form, the Conditional, is required in questions expressing a doubt:3 (1) Max serait-il là? Je vois sa voiture. ‘Is [be:COND] M. there? I see his car.’

(Leeman 2001: 227)

A relevant point here is that French questions admit virtually any verb forms, including the Present Indicative in (2a), and not only the Conditional (2b): (2a) Luc est-il à Paris? ‘L., is he in Paris?’

(Leeman 2001: 225)

(2b) Luc serait-il à Paris? ‘L., is [be:COND] he in Paris?’

Given that French allows other verb forms in interrogative sentences (2a) and the Conditional is restricted to conjectural questions such as (1, 2b), this phenomenon cannot be confused with an interrogative mood. Instead, a viable solution here is considering the Conditional as a proper Dubitative, a form expressing doubts and occurring in questions, which empirically demonstrates Lyons’ basic point. Obviously, this responds to Palmer’s objection on the empirical consistency of the Dubitative, but also demonstrates, contra Lyons (1977), that the illocutionary force of questions and the conjectural meaning of the Dubitative can be kept distinct, since only conjectural contexts, and not any kind of questions, admit the Conditional in French. On top of that, another contradictory empirical point with respect to Lyons’ conclusions is that the French Conditional cannot occur in the declarative counterpart (3) of the question in (1): (3)

??

Max serait là: je vois sa voiture. ‘M. is [be:COND] there: I see his car.’ (Leeman 2001: 227; see also Dendale 1994: 38)

Obviously, this behaviour is not what Lyons would predict for a dubitative mood, which in his view is supposed to cover questions as well as conjectures in declarative sentences. Nonetheless, it must be borne in mind that the French Conditional is prototypically a form occurring in non-factual contexts such as apodoses of conditional clauses:

3

On this usage of the Conditional see also Diller (1977), Haillet (2001), Melis (2001); on its ‘evidential’ interpretation see Tasmowski (2001), Rocci (2007: 130), Squartini (2008: 935–941).

114 (4)

Mario Squartini Si je pouvais, j’irais au cinéma. ‘If I could, I would go [go:COND] to the movies.’

Creating possible worlds, which is the function of the Conditional in (4), is not traditionally considered a prototypical instance of epistemicity, the latter being more properly connected to the expression of judgements (Palmer 1986: 51– 95). However, the behaviour of the French Conditional can be capitalized on to verify empirically the actual scope of what is to be defined as a dubitative mood. Instead of demonstrating a functional relationship between questions in general and epistemic modality, which would confirm Lyons’ speculation in its original form and suggest a significant overlap of illocution and modality, these French data only indicate a connection between conjectural questions and non-factuality. The link between the dubitative mood and non-factuality could be confirmed by additional Romance data, which also demonstrate a special morphosyntactic marking in conjectural questions. Even though nowadays the occurrence of the Conditional in conjectural questions seems to be restricted to French,4 other Romance languages do have different morphological means with a comparable distribution. This is the case of the Italian Subjunctive (Rocci 2007), that occurs in yes/no interrogative sentences (5a) while it is barred as a conjectural form without the rising intonational contour of interrogative sentences (5b): (5a) [Suonano alla porta] Che sia G.? ‘[The doorbell rings] Is [be:SUBJ] it G.?’ (5b) [Suonano alla porta] *(Che) sia G. ‘[The doorbell rings] It is [be:SUBJ] G.’

As also suggested by Rocci (2007: 130) this resembles the behaviour of the French Conditional seen above, which only occurs in interrogative but not in declarative conjectural sentences, as demonstrated by (1) vs. (3). Obviously, the behaviour of the Italian Subjunctive may also be due to independent factors if one considers that the Italian Subjunctive is barred in declarative independent sentences. However, as will be shown in section 4 below, a construction parallel to (5b) is admitted with the Spanish Subjunctive in declarative sentences and the similar behaviour of the French Conditional allows us to speculate that the Italian Subjunctive and the French Conditional can be grouped together in that they both show a special morphosyntactic marking for dubitative questions (for a ‘constructional’ analysis of the Italian 4

Note, however, that the occurrence of the Conditional in conjectural questions appears to be rather varied both diachronically and synchronically (see Squartini 2001, 2004 for details).

Where mood, modality and illocution meet

115

dubitative Subjunctive see also Rocci 2007: 141). Having hypothesized above that the occurrence of the Conditional in dubitative questions can be explained as a result of the intrinsic non-factual meaning of the Conditional, now the question arises of whether a similar explanation could also be extended to the Italian Subjunctive. This raises the more general issue connected to the semantics of the Subjunctive: as is well known, a unified semantic definition of the Romance Subjunctive is difficult to assess and indeed, whether a semantic definition is possible at all is questionable, a purely morphosyntactic interpretation being also viable.5 Nonetheless, it is a fact that the Italian Conditional and Subjunctive share a common semantic mould in the ‘irrealis’ context of conditional clauses, the Conditional normatively occurring in apodoses and the Subjunctive in protases. This confirms that in general the Romance forms used as dubitative, Conditionals and Subjunctives, are all connected to non-factuality. As will be shown below in section 4, this conclusion might be contradicted by the behaviour of the Spanish Subjunctive and indeed, the Subjunctive will prove to be less semantically characterized than the Conditional and therefore more freely combinable with different semantic subdomains of epistemicity, not only with dubitative contexts but also with the conjectural contexts described below. However, before introducing this point, the occurrence of modal forms in interrogative sentences will be examined more thoroughly in the next section, not only for the purposes of helping to establish what can be considered as a dubitative mood, but also as a negative test of what is intrinsically anti-dubitative, which will be dubbed as ‘inferential mood’.

3. The inferential mood in Romance In the previous section it was argued, following Lyons’ proposal, that the occurrence of a given form in interrogative sentences can be considered as a test in order to delimit the scope of the dubitative mood. Now the argument will be reversed and the non-occurrence in interrogative sentences will be used as a negative test: the epistemic forms that cannot occur in interrogative sentences will be considered as intrinsically non-dubitative. Palmer (1979: 62) notes that English epistemic modals cannot easily be found in interrogative sentences, but this seems to be merely a “question of frequency” and “not an inherent restriction”, as Verstraete (2001) also 5

For a comprehensive approach in which semantics interacts with pragmatics suggesting a unified analysis, the reader is referred to Quer (1998, 2001).

116

Mario Squartini

concludes, confirming Palmer’s claim. Nevertheless, some specific incompatibilities of epistemic modals with interrogative sentences have often been pointed out, as is the case of the following contrast (from Roulet 1979: 66; see also Leeman 2001: 226) between two French epistemic modals, the weak epistemic pouvoir ‘can, may’ (6) as opposed to its strong counterpart, devoir ‘must’ (7): (6)

Pierre peut-il être malade? ‘May P. be sick?’

(7)

*Pierre doit-il être malade? ‘Must P. be sick?’

More recently, the issue has been reinterpreted from an interactional and pragmatic perspective, in which similar contrasts have been considered less straightforward than traditionally assumed in more structural descriptions (see Pietrandrea 2005: 130–131 commenting on this point). Nonetheless, it is a fact that, all other things being equal (also from an interactional point of view), significant differences among epistemic verbal forms can be detected if they are placed in interrogative sentences. Consider the contrast in (8–10) showing that in declarative clauses both the Italian epistemic Future and the modal dovere ‘must’ + infinitive can occur, while in conjectural questions only the Future is allowed (see also Rocci 2005: 238, 2007: 133): (8a) [Suonano alla porta.] Sarà il postino. ‘[The doorbell rings.] It will be [be:FUT] the postman.’ (8b) [Suonano alla porta.] Deve essere il postino. ‘[The doorbell rings.] It must be the postman.’ (9a) [Suonano alla porta.] Sarà il postino? ‘[The doorbell rings.] Will it be [be:FUT] the postman?’ (9b) [Suonano alla porta.] ??Deve essere il postino? ‘[The doorbell rings.] Must it be the postman?’

In an interactional perspective, it may be noted that dovere + infinitive becomes grammatical in echo-questions and requests for confirmation expressed in tag-questions (Rocci 2007: 134), and, more generally, whenever the conjecture is attributed to the interlocutor (Verstraete 2001: 1522, ex. (10) from Pietrandrea 2005: 131): (10) [Suonano alla porta.] Secondo te deve essere il postino? ‘[The doorbell rings.] According to you must it be the postman?’

However, the relevant structural point is that an explicit shifting of perspective to the interlocutor is not required by the epistemic Future in (9a), which

Where mood, modality and illocution meet

117

demonstrates that a contrast between the modal dovere ‘must’ and the epistemic Future exists, even if it can be neutralized by shifting the pragmatic perspective. The distributional behaviour of dovere + infinitive, as well as that of its French counterpart (7), can now be interpreted with respect to the conclusions reached above. As opposed to the French Conditional and the Italian Subjunctive, which were proved to be incompatible with declarative inferences, the modal dovere + infinitive turns out to be the prototypical marker of inferences in declarative sentences (8b), while excluding dubitative interpretations in interrogative sentences (9b). Note that contrasting the dubitative Italian Subjunctive and French Conditional described in section 2 with the inferential modals Italian dovere / French devoir ‘must’ implies that the latter are considered on a par with proper moods, which may be controversial. However, as noted by Palmer (1986: 21– 23) the traditional definition of moods as prototypically inflectional may be too strict as it excludes “particles and other formal markers” the function of which is comparable to moods (Palmer 1986: 21). As also observed by Palmer, the traditional definition of moods is crucially based on their morphosyntactic restrictions as subordinate forms (see also Jespersen 1924: 313). In this respect, it can be noted that the boundary between moods and modals may be fuzzier than traditionally assumed, as is shown by the occurrence of dovere + infinitive in subordinate complement clauses dependent on jussive verbs in Old Italian (similar Old Occitan examples are documented in Jensen 1990: 301): (11) Ora ti priego che mi debbia mostrare ed aprire la natura delle dette quat[t]ro virtú principali. ‘Now I beg you to show [I beg that you must:SUBJ show] me and unveil the nature of the aforementioned four main virtues.’ (Bono Giamboni, Il Trattato di Virtù e Vizî, 126.1, 13th c.)

In modern Italian the use of the verb dovere in a subordinate clause as above is ungrammatical and an Infinitive or a Subjunctive would be required. This blurs the boundary between moods and modals demonstrating the possibility that modals develop (grammaticalize?) as proper morphosyntactic moods. Even though the diachronic evolution of the Italian modals has been different, with the modal dovere dismissing morphosyntactic restrictions that used to be applicable in Old Italian, the permeability of the boundary between moods and modals is confirmed by data such as (11), which tentatively allows a comparative analysis of inflectional moods (Conditionals and Subjunctives) and modals.

118

Mario Squartini

The form-function correlations presented above indicate that the Romance epistemic domain can be split up into two basic, and semantically different, subdomains, the dubitative and the inferential, which in some Romance languages can be morphologically marked by specialized moods (the French Conditional and the Italian Subjunctive) or specialized modals (devoir, dovere + infinitive). In comparing the empirical data that support the dubitative and the inferential as regular form-function associations it could be observed that while the occurrence of a specialized marker of dubitativity such as the French Conditional and the Italian Subjunctive seems to be rather peculiar to these languages, the non-occurrence of the epistemic modal ‘must’ in dubitative questions might be more consistently general across languages, being apparently due to the intrinsic strong epistemic meaning of the modal ‘must’. Nonetheless, even in this respect Romance data provide a varied picture, for in Catalan the modal deure ‘must’ + infinitive turns out to be also allowed in dubitative questions: (12) Què deu ser? ‘What is it [must it be]?’

Significantly, Italian does not permit the modal dovere + infinitive in similar contexts, where, instead, the epistemic inflectional Future occurs: (13a) ??Che deve essere? ‘What must it be?’ (13b) Che sarà? ‘What is [be:FUT] it?’

These comparative data are relevant in assessing the internal distribution of the epistemic domain in Romance: they show that the dubitative and the inferential subdomains belong to one and the same functional continuum, for they can be neutralized by the same morphological marker, deure + infinitive in Catalan. On the other hand, the dubitative and the inferential can still be considered as independent, due to the behaviour of given moods (the French Conditional and the Italian Subjunctive) or modals (French and Italian devoir / dovere + infinitive) that are selectively compatible with only one of the two subdomains. The epistemic continuum ranging from the dubitative to the inferential is also demonstrated by languages such as Italian or Spanish, in which the conjectural Future occurs both in interrogative and declarative clauses (see (8a) and (9a) above for relevant examples and Escandell Vidal 1999: 3968 for the occurrence of the Spanish Future in conjectural questions). Interestingly, in normative French such a neutralization seems not to be permitted, as shown by the following contrast presented in Vet (1990: 130–131; see also Schrott 1997: 310–311):

Where mood, modality and illocution meet

119

(14a) Pierre n’est pas là. Il aura manqué le train. ‘P. is not here. He must have missed [miss:FUT.PERF] the train.’ (14b) *Pierre n’est pas là. Est-ce qu’il aura manqué le train? ‘P. is not here. Did he miss [miss:FUT.PERF] the train?’

The contrast in (14) implies a complementary distribution between the Conditional, which is only allowed in questions (see (1) and (3) above), and the Future, which is disallowed in questions. This can be interpreted as a tendency of French to mark not only the distinction between the inferential modal devoir and the dubitative Conditional, but also an intermediate function that is represented by the conjectural Future. Thus, the French Future could be described as more weakly inferential with respect to devoir + infinitive but, at the same time, more strongly assertive than the dubitative Conditional. Note that in this respect French is much more radical than Italian in marking the Dubitative, where a dedicated dubitative marker (che + Subjunctive) exists, but, unlike French, the conjectural Future can also occur in questions, thus avoiding a complementary distribution between the dubitative Subjunctive and the conjectural Future. However, it must be noted that French native speakers seem to be not so rigid in rejecting a conjectural Future in questions, which suggests that the distribution of the Conditional and the Future in colloquial French might be less complementary than traditionally assumed. This might be connected to register variation and needs further research, but it can be tentatively concluded that at least in standard literary French the contrast between the dubitative and the inferential is more conclusively marked than in other Romance languages. Summing up the data presented so far, it can be concluded that different degrees of neutralization in the epistemic domain can be found when comparing different Romance languages. In this respect Catalan appears to be representative of a tendency to neutralize most distinctions, having a single form, originally the strong inferential modal deure ‘must’ + infinitive, that can be used in any inferential-conjectural contexts including dubitative questions. On the contrary, the variety of French described by Vet (1990) displays a system of structural oppositions in which there is a complementary distribution between the Future and the Conditional, the latter being consistently dubitative (conjectural questions and non-factual conditionals) and the former being restricted to conjectural declarative sentences. This can be interpreted as a tendency of French to have a proper conjectural mood, the Future, as opposed to a dubitative mood, the Conditional. From this point of view French can be described as the Romance language in which the epistemic domain exhibits the highest degree of disambiguation among rather subtle functional subdistinctions. As described above, these semantic subdomains can be recognized by

120

Mario Squartini

the regular combination of morphosyntactic structures (interrogative vs. declarative sentences) and moods. Given these results, one might also consider the possibility of distinguishing not only the inferential from the dubitative mood but also the conjectural subdomain as an intermediate area between the dubitative and the inferential subdomains. In this perspective, the dubitative mood could be defined as the expression of conjectures in the special illocutionary context of questions, while a form that expresses conjectures with no special restrictions to questions could be considered as an instance of the conjectural mood. In the following section this hypothesis will be elaborated further, showing that French and Spanish seem to provide evidence of morphosyntactic structures that cover the conjectural area crucially distinguishing it from the inferential subdomain.

4. Conjectural morphosyntax? In describing the peculiarities of the Spanish Subjunctive as opposed to its French counterpart, Schifko (1967: 18) interestingly pointed out the following pair of examples, in which a French original Future is translated into a Spanish Subjunctive, thus demonstrating a special ‘overuse’ of the Spanish Subjunctive in conjectural declarative sentences: (15a) Peut-être seras-tu étonné que je t’écrive ainsi, moi. ‘Perhaps you might be [be:FUT] surprised that I write to you in this way.’ (15b) Quizá te extrañe que sea yo quien te escriba así. ‘Perhaps you might be surprised [surprise:SUBJ] that I write to you in this way.’

In fact, the occurrence of the Spanish subjunctive following the adverb ‘perhaps’ in an independent clause (15b) has repeatedly attracted attention as a rather exceptional phenomenon in the Romance mood system. Its areal restriction to “Western Romance languages” has been clearly underlined since Meyer-Lübke (1890–1899: vol. 3, § 119), but apart from being an areal phenomenon, the Subjunctive in (15b) is also problematic if one considers that Romance languages do permit root subjunctives but generally only when a special illocutionary force is involved such as in optative or jussive sentences (see the generalization in Giorgi & Pianesi 1997: 194). It would be difficult to detect a special illocution in (15), unless we follow Lyons (1977), who suggests considering “epistemically qualified” utterances as a separate illocutionary type, distinct from “modally unqualified” statements (see fn. 1), which, again, would have the disadvantage of confusing modality and

Where mood, modality and illocution meet

121

illocution. However, instead of stressing the peculiarities of the Spanish Subjunctive, in what follows I would like to focus on the similarities between French and Spanish. What is common to (15a) and (15b) is that they are both characterized by a special morphosyntax, which in French is represented by the verb-subject inversion, a phenomenon that, like other similar morphosyntactic structures involving the syntactic domain of the Complementizer (Rizzi & Roberts 1989), can be interpreted as a residual effect of a V2 morphosyntax in modern French (Roberts 1993: 52). In Spanish the occurrence of the Subjunctive is also significantly constrained as far as morphosyntactic restrictions are concerned: a preposed modal adverb (quizá) is required (15a, 16a), as is confirmed by the impossibility of omitting (16b) or even postponing it (16c) (examples from Ridruejo 1999: 3220 and Kovacci 1999: 756): (16a) Quizá venga mañana Pedro. (16b) *Venga mañana Pedro. (16c) *Venga quizá mañana Pedro. ‘Perhaps P. comes [come: SUBJ] tomorrow.’

The parallel drawn between the French verb-subject inversion (‘subject-clitic inversion’ in Rizzi & Roberts 1989: 3) and the restrictions on the Spanish Subjunctive is also confirmed by the list of adverbs admitted in these constructions, a point in which French and Spanish also show striking similarities. Even though the weak epistemic adverb ‘perhaps’ is most generally described as representative of these constructions, various other adverbs, with different degrees of epistemic commitment, can occur. Generally speaking, they are all those dubbed ‘modal adverbs expressing degrees of probability’ along the ‘positive epistemic scale’ by Ramat & Ricca (1998: 226), such as probablement as well as certainement and sans doute in French (Cinque 1999: 177, fn. 53), and probablemente, posiblemente, seguramente in Spanish (Kovacci 1999: 755–758).6 Note that this makes the semantic scope of this phenomenon in declarative sentences quite different from what occurs in interrogative sentences. Consider that epistemic adverbs such as probably are not generally accepted in questions (Bellert 1977: 344–346),7 those meaning ‘perhaps’ being exceptional in this respect (Bellert 1977: 344,

6 7

Note that the compatibility of seguramente, admitted by Kovacci (1999), is explicitly denied by Hengeveld (1989: 139). Not unlike the modal dovere ‘must’ + infinitive (see above section 3), some epistemic adverbs may become grammatical in special interactional contexts (see Ramat & Ricca 1998: 194–195) and examples in Venier (1991: 37–38).

122

Mario Squartini

Nølke 1988: 29–32, Nøjgaard 1992–1995: 239–240).8 Take for instance the contrast mentioned in Blumenthal (1975: 332), who shows that probablement and certainement are not compatible with a conjectural question (17), while peut-être is perfectly compatible in (18), where it collocates with the dubitative Conditional described above in section 2 (see also Nølke 1988: 23):9 (17) *A-t-il probablement (certainement, …) raison? ‘Is he probably (certainly, …) right? (18) Serait-il peut-être …? ‘Will [be: COND] he be perhaps …?

These data show how epistemic adverbs can be sensitive to morphosyntactic constructions with a distribution that parallels the behaviour of the French inflectional forms, the Conditional and the Future described in sections 2–3, which basically confirms the hypothesis that the epistemic domain can be subdivided into different subdomains characterized by special morphosyntactic restrictions. From this point of view, peut-être, being compatible with conjectures in questions as well as in declarative sentences, can be described as neutralizing the distinction between the dubitative and the conjectural subdomains, while probablement is more consistently conjectural and not dubitative (for it is not permitted in questions). More generally speaking, this also demonstrates that the behaviour of lexical adverbs is comparable to inflectional verb forms, thus helping to disambiguate different epistemic subdomains. In particular, the behaviour of the Italian epistemic Future neutralizing the boundary between dubitative and conjectural subdomains is paralleled by the neutralization shown by the adverb peut-être, while the incompatibility of probablement in questions corresponds to the restrictions of the French and Italian modals devoir / dovere + infinitive described above (Rocci 2007: 133–134). As for Spanish, it should be observed that the Subjunctive is not obligatory in the conjectural context in (16), the Future being also permitted with no significant semantic difference, as reported by Ridruejo (1999: 3220): 8

9

In Italian the special nature of forse ‘perhaps’ might also be recognized in some morphosyntactic peculiarities (Cinque 1999: 177, fn. 53; see also Lonzi 1991: 401), such as the possibility of occurring to the left of the complementizer (Forse che ci aiuterà ‘Is it possible that [perhaps that] he will help us?’). This might be another case of cluster of morphosyntactic constraints and semantic specialization in the epistemic domain, which deserves further research. Similar data in Roulet (1979): 66 and Nøjgaard (1992–1995): 226 as well as in Lonzi (1991): 400 for the Italian adverbs probabilmente and forse and Rocci (2007): 134–135 as far as probabilmente is concerned.

Where mood, modality and illocution meet (19) Quizá vendrá mañana Pedro. ‘Perhaps P. will come [come:FUT] tomorrow.’

123

(Ridruejo 1999: 3220)

However, in the perspective of the present article what counts is not the extent to which the alternation between the Subjunctive and the Future in (16) and (19) can be considered free (see also Palmer 1986: 4 and Hengeveld 1989: 139 commenting on these data), but that the alternation exists and that one of the possible forms (the Subjunctive) is morphosyntactically constrained, which supports the hypothesis that the Spanish Subjunctive can be considered as a conjectural mood in contexts such as (16a) and can be opposed to a free epistemic inflectional form, the Future in (19). In a sense, it can be concluded that the latter may have the same function without any morphosyntactic restrictions. As an additional caveat with respect to these conclusions, it must also be observed that, especially in French, the list of adverbs admitted in the verbsubject inversion is not limited to epistemics, the adverb malheureusement ‘unfortunately’ being a relevant exception. As noted by Cinque (1999: 177, fn. 53) further research is needed to define more precisely the limits of this phenomenon. Ultimately, this may imply that the French construction, possibly unlike the Spanish usage of the Subjunctive, cannot be considered as a morphosyntactic structure specifically dedicated to the conjectural domain, but what is relevant for the purposes of the present article is demonstrating that, possibly among other functions, the French verb-subject inversion is sensitive to semantic distinctions internal to the epistemic domain. In this perspective it is worth observing that only some adverbs are permitted the French verbsubject inversion. In particular, the construction is sensitive to a distinction that is common in classifications of ‘modal’ (Ramat & Ricca 1998: 226) or ‘assertive’ adverbs (Borillo 1976: 81–82, Roulet 1979: 63, Nøjgaard 1992– 1995: 203–260), where two different subtypes are traditionally distinguished: those expressing different degrees of probability (properly epistemic adverbs) such as ‘perhaps, possibly, probably’ as opposed to evidential and reportive adverbs (Ramat & Ricca 1998: 226; see also Ramat 1996) that denote the external source of evidence on which the utterance is based. As noted by Cinque (1999: 177, fn. 53; see also Nøjgaard 1992–1995: 221), the French adverb évidemment, which can be interpreted as an evidential marker pointing out an external source of evidence on which the speaker bases his/her reasoning, is not compatible with verb-subject inversion: (20) *Evidemment viendra-t-il. ‘Evidently will he come [come:FUT].’

Similarly Kovacci (1999: 759–760) mentions the adverbs ‘reforzadores’ (evidentemente) or ‘restrictivos del valor de verdad de la aserción’

124

Mario Squartini

(aparentemente) as those not admitted with the Subjunctive in Spanish. Kovacci (1999: 759) provides the following example with respect to the incompatibility of aparentemente with the Subjunctive: (21) No contesta; aparentemente no está/*esté en casa. ‘S/he doesn’t answer; apparently s/he is [be:INDIC/SUBJ] not at home.’

The context in (21) can be defined as inferential, more precisely as what in the typological literature would be considered a circumstantial inference (Anderson 1986: 274), i. e. an inferential process in which the speaker’s own reasoning is based on an external piece of evidence (‘nobody answers’). My interpretation of these data is that the Subjunctive behaves here as a ‘conjectural mood’ and is therefore not compatible with the explicit inferential context in (21). The Subjunctive is only admitted when the epistemic judgement is presented as the product of the speaker’s own reasoning, with the speaker freely interpreting the external data and evaluating the probability of the judgement (which can be properly defined as a conjecture): (22) Quizá no esté en casa. ‘Perhaps s/he is [be:SUBJ] not at home.’

While in Spanish the conjectural meaning is expressed by a proper mood, the Subjunctive, in French the same meaning is expressed by the subject-verb inversion, which cannot be described as a ‘mood’. However, what is more relevant for my purposes is that Spanish and French support the hypothesis that the conjectural meaning, even if not necessarily expressed by a mood, tends to be identified by means of a special morphosyntax. Whatever definition of mood one should follow, the relevant point here is that French and Spanish permit an internal subdivision to be drawn in the epistemic domain. In particular these data demonstrate that conjecturals and inferentials can be kept distinct if a morphosyntactic perspective is adopted. Nonetheless, considering the adverbs which are not admitted in these constructions and following Dendale’s (1994, 2001) treatments of the French epistemic Future and modals, this distinction can also be reinterpreted as an opposition between purely epistemic subjective modality (conjectures) and evidentiality (Aikhenvald 2004). This means that the two morphosyntactic constructions under scrutiny are sensitive to a distinction between modality and evidentiality that cuts across the epistemic domain described above, singling out inferentiality as intrinsically evidential rather than conjectural epistemicity. As correctly noted by Hengeveld (1989: 138), who elaborates on Lyons’ (1977: 797–801) original distinction between objective and subjective

Where mood, modality and illocution meet

125

modality,10 “[i]n the case of subjective modality the speaker is the source”, which might imply that the distinction between conjectures and inferences can be considered as a contrast between an internal source, the speaker’s own conjectures, and an external source. Ultimately, this means that the difference between ‘perhaps’ and ‘apparently’ should not be understood as a contrast between epistemicity and evidentiality, but within evidentiality, in which two different sources (internal vs. external) are distinguished. In Squartini (2008) further Romance data are compiled suggesting a semantic continuity within evidentiality, but this goes beyond the scope of the present article. The basic point here is demonstrating that Romance morphosyntax permits different form-function correlations to be distinguished in the epistemic-evidential domain.

5. Conclusion Elaborating on Lyons’ speculations on the relationship between mood and modality in the epistemic domain, it can be shown that, among the Romance languages, French and Italian provide evidence supporting Lyons’ suggestion of a dubitative mood as an interface phenomenon between mood and modality. Nonetheless, the Romance data examined above has also limited the scope of this phenomenon with respect to Lyons’ comprehensive model: instead of demonstrating a general unification of interrogative and ‘epistemic utterances’ under the vast umbrella of Lyons’ dubitative mood, the Romance Dubitative only covers conjectural questions, which avoids undesirable confusion between illocution and modality. On the other hand, the Romance data demonstrate that epistemicity is much more varied internally than so far assumed, the dubitative only being a single element of a more complex picture. If the whole Romance spectrum of variation is taken into account, epistemicity can be interpreted as a continuum from dubitativity to inferentiality. This semantic area can be subdivided into different subdomains – dubitative, conjectural and inferential – whose formal correlations can surface as morphosyntactically constrained combinations of sentence structures and moods. In order to provide a summarized representation of the multifarious set of data analyzed in this article, two semantic features, [+/-DUBITATIVE] and 10

On the interpretation of Lyons’ (1977: 797–801) original distinction between objective and subjective modality as involving evidentiality see also Venier (1991: 34–35) and Nuyts (2001: 31).

126

Mario Squartini

[+CONJECTURAL], may be chosen to represent the major subdistinctions in the Romance conjectural-inferential domain. As shown in table 1, the feature [+DUBITATIVE] corresponds to the dubitative mood described in section 2 and covers forms that, when expressing conjectures, are restricted to a given morphosyntactic structure (questions). As demonstrated above, this restriction applies to the Italian Subjunctive as well as to the French Conditional. The opposite value of the same feature [-DUBITATIVE] characterizes the French and Italian inferential modals dovere/devoir ‘must’ + infinitive (section 3), which show the opposite morphosyntactic restriction, being only allowed in declarative sentences. Note that epistemic adverbs, such as French probablement ‘probably’, have comparable morphosyntactic restrictions and could be lumped together with the modals dovere/devoir under the same inferential subdomain, here described as [-DUBITATIVE]. The data scrutinized above (sections 3–4) also suggest that an intermediate area can be detected between dubitative and inferential forms. This intermediate domain has been tentatively dubbed ‘conjectural mood’ in section 4 and is represented by the feature [+CONJECTURAL] in table 1. This feature covers the distribution of the Spanish Subjunctive in conjectural declarative sentences with preposed epistemic adverbs (crucially excluding inferential evidential adverbs such as Spa. aparentemente ‘apparently’), which is comparable to the distribution of French epistemic forms in verb-subject inversions. As some descriptions suggest (see section 3 above), the French epistemic Future might also be considered to be characterized by the feature [+CONJECTURAL], but more research needs to be done to assess the sociolinguistic distribution of such a restriction. Table 1 also shows that the semantic area covered by the features [+/DUBITATIVE] and [+CONJECTURAL] can be considered a unitary domain, for a single form (the Catalan modal deure ‘must’ + infinitive) occurs with no restrictions in the conjectural-inferential continuum. A more restricted neutralization applies to the Italian Future, which occurs in questions [+DUBITATIVE] and declarative sentences [+CONJECTURAL], thus being less strongly inferential than the Italian modal dovere ‘must’ + infinitive, for the latter does not occur in questions. As mentioned above, Romance ‘perhaps’adverbs (French peut-être, Italian forse, Spanish quizá, tal vez), which occur not only in declarative sentences but also in questions, are characterized by a distribution comparable to the Italian Future (relevant data in section 4).

Where mood, modality and illocution meet

127

Table 1: Form / function correlations in the Romance epistemic domain. [+DUBITATIVE] Dubitative mood: Fra. Conditional Ita. Subjunctive Ita. [+CONJECTURAL] Conjectural mood: Cat. Future Spa. Subjunctive deure ‘must’ + Fra. verb-subject inversion infinitive Fra. Future (?) [-DUBITATIVE] Inferential mood: Fra. devoir + infinitive Ita. dovere + infinitive

References Aikhenvald, Alexandra Y. (2004): Evidentiality. – Oxford: Oxford University Press. Anderson, Lloyd B. (1986): “Evidentials, paths of change, and mental maps: typologically regular asymmetries.” – In: Wallace Chafe & Johanna Nichols (eds.) (1986): Evidentiality: The linguistic coding of epistemology, 273–312. Norwood, N. J.: Ablex. Bellert, Irene (1977): “On semantic and distributional properties of sentential adverbs.” – In: Linguistic Inquiry 8, 337–350. Blumenthal, Peter (1975): “Zur kommunikativen Funktion von Adverbien und Umstandsbestimmungen im Französischen.” – In: Romanische Forschungen 87, 295–332. Borillo, Andrée (1976): “Les adverbes et la modalisation de l’assertion.” – In: Langue française 30, 74–89. Bybee, Joan L., Perkins, Revere & Pagliuca, William (1994): The evolution of grammar. Tense, aspect, and modality in the languages of the world. – Chicago/London: The University of Chicago Press. Chafe, Wallace & Nichols, Johanna (eds.) (1986): Evidentiality: The linguistic coding of epistemology. – Norwood, N. J.: Ablex. Cinque, Guglielmo (1999): Adverbs and functional heads. A cross-linguistic perspective. – Oxford: Oxford University Press. Cresti, Emanuela (2002): “Modalità e illocuzione.” – In: Gian Luigi Beccaria & Carla Marello (eds.): La parola al testo. Scritti per Bice Mortara Garavelli, 133–145. Alessandria: dell’Orso. Dendale, Patrick (1994): “Devoir épistémique, marqueur modal ou évidentiel?” – In: Patrick Dendale & Liliane Tasmowski (eds.) (1994): Les sources du savoir et leurs marques linguistiques, 24–40. Paris: Larousse (= Langue française 102). – (2001): “Le futur conjectural versus devoir épistémique: différences de valeur et de restrictions d’emploi.” – In: Le français moderne 69, 1–20.

128

Mario Squartini

Dendale, Patrick & Tasmowski, Liliane (eds.) (1994): Les sources du savoir et leurs marques linguistiques. – Paris: Larousse (= Langue française 102). – (eds.) (2001): Le conditionnel en français. – Paris: Klincksieck (Université de Metz, Recherches linguistiques 25). Diller, Anne-Marie (1977): “Le conditionnel, marqueur de dérivation illocutoire.” – In: Semantikos 2, 1–17. Dryer, Matthew S. (2005): “Polar questions.” – In: Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie (eds.): World Atlas of Language Structures, 470–473. Oxford: Oxford University Press. Escandell Vidal, Ma. Victoria (1999): “Los enunciados interrogativos. Aspectos semánticos y pragmáticos.” – In: Ignacio Bosque & Violeta Demonte (eds.): Gramática descriptiva de la lengua española, 3929–3991. Madrid: Espasa Calpe. Giorgi, Alessandra & Pianesi, Fabio (1997): Tense and aspect. From semantics to morphosyntax. – New York/Oxford: Oxford University Press. Guentchéva, Zlatka (ed.) (1996): L’énonciation médiatisée. – Louvain/Paris: Peeters. Haillet, Pierre Patrick (2001): “À propos de l’interrogation totale directe au conditionnel.” – In: Patrick Dendale & Liliane Tasmowski (eds.) (2001): Le conditionnel en français, 295–330. Paris: Klinksieck (Université de Metz, Recherches linguistiques 25). Hengeveld, Kees (1988): “Illocution, mood and modality in a functional grammar of Spanish.” – In: Journal of Semantics 6, 227–269. – (1989): “Layers and operators in Functional Grammar.” – In: Journal of Linguistics 25, 127–157. – (2004): “Illocution, mood, and modality.” – In: Geert BooƋ, Christian Lehmann, Joachim Mugdan & Stavros Skopetas (eds.): Morphologie / Morphology. Ein internationales Handbuch zur Flexion und Wortbildung, 1190–1201. Berlin/New York: de Gruyter. Jensen, Frede (1990): Old French and comparative Gallo-Romance syntax. – Tübingen: Niemeyer. Jespersen, Otto (1924): The Philosophy of Grammar. – London: Allen & Unwin. Kovacci, Ofelia (1999): “El adverbio.” – In: Ignacio Bosque & Violeta Demonte (eds.): Gramática descriptiva de la lengua española, 705–786. Madrid: Espasa Calpe. Leeman, Danielle (2001): “Pourquoi ne peut-on combiner si et le conditionnel?” – In: Patrick Dendale & Liliane Tasmowski (eds.) (2001): Le conditionnel en français, 211–230. Paris: Klinksieck (Université de Metz, Recherches linguistiques 25). Lonzi, Lidia (1991): “Il sintagma avverbiale.” – In: Lorenzo Renzi & Giampaolo Salvi (eds.): Grande grammatica italiana di consultazione. I sintagmi verbale, aggettivale, avverbiale. La subordinazione. Vol. 2, 341–412. Bologna: Il Mulino. Lyons, John (1977): Semantics. – Cambridge: Cambridge University Press. Melis, Ludo (2001): “Hypothèses non temporelles sur le conditionnel comme tiroir de l’indicatif.” – In: Patrick Dendale & Liliane Tasmowski (eds.) (2001): Le conditionnel en français, 67–88. Paris: Klinksieck (Université de Metz, Recherches linguistiques 25). Meyer-Lübke, Wilhelm (1890–1899): Grammatik der romanischen Sprachen. – Leipzig: Reisland.

The morphosyntax of Romance conjectures

129

Nøjgaard, Morten (1992–1995): Les adverbes français. Essai de description fonctionnelle. Copenhagen: Munksgaard (= Historisk-filosofiske Meddelelser 66 (1– 3)). Nølke, Henning (1988): “Peut-être.” – In: Verbum 11, 13–43. Nuyts, Jan (2001): Epistemic modality, Language and Conceptualization. A cognitivepragmatic perspective. – Amsterdam/Philadelphia: Benjamins. Palmer, Frank R. (1979): Modality and the English modals. – London/New York: Longman. – (1986): Mood and modality. – Cambridge: Cambridge University Press. Pietrandrea, Paola (2005): Epistemic modality. Functional properties and the Italian system. – Amsterdam/Philadelphia: Benjamins. Quer, Josep (1998): Mood at the interface. – The Hague: Holland Academic Graphics. – (2001): “Interpreting mood.” – In: Probus 13, 81–111. Ramat, Paolo (1996): “Allegedly, John is ill again: stratégies pour le médiatif.” – In: Zlatka Guentchéva (ed.): L’énonciation médiatisée, 287–298. Louvain/Paris: Peeters. – & Ricca, Davide (1998): “Sentence adverbs in the languages of Europe.” – In: Johan van der Auwera (ed.): Adverbial constructions in the languages of Europe, 187–275. Berlin/New York: Mouton de Gruyter. Ridruejo, Emilio (1999): “Modo y modalidad. El modo en las subordinadas sustantivas.” – In: Ignacio Bosque & Violeta Demonte (eds.): Gramática descriptiva de la lengua española, 3209–3251. Madrid: Espasa Calpe. Rizzi, Luigi & Roberts, Ian G. (1989): “Complex inversion in French.” – In: Probus 1, 1–30. Roberts, Ian G. (1993): Verbs and diachronic syntax. A comparative history of English and French. – Dordrecht: Kluwer. Rocci, Andrea (2005): “On the nature of the epistemic readings of the Italian modal verbs: the relationship between propositionality and inferential discourse relations.” – In: Bart Hollebrandse, Angeliek van Hout & Co Vet (eds.): Crosslinguistic Views on Tense, Aspect and Modality, 229–246. Amsterdam/New York: Rodopi. – (2007): “Epistemic modality and questions in dialogue. The case of Italian interrogative constructions in the subjunctive mood.” – In: Louis de Saussure, Jacques Moeschler & Genoveva Puskas (eds.): Tense, Mood and Aspect. Theoretical and Descriptive Issues, 129–153. Amsterdam/New York: Rodopi (Cahiers Chronos 17). Roulet, Eddy (1979): “Des modalités implicites intégrées en français contemporain.” – In: Cahiers Ferdinand de Saussure 33, 41–76. Schifko, Peter (1967): Subjonctif und subjuntivo. Zum Gebrauch des Konjunktivs im Französischen und Spanischen. – Wien/Stuttgart: Braumüller (Wiener romanistische Arbeiten 6). Schrott, Angela (1997): Futurität im Französischen der Gegenwart. Semantik und Pragmatik der Tempora der Zukunft. – Tübingen: Narr (Romanica Monacensia 50). Squartini, Mario (2001): “Filogenesi e ontogenesi del Futuro italiano.” – In: Archivio glottologico italiano 86, 194–225.

130

Mario Squartini

– (2004): “La relazione semantica tra Futuro e Condizionale nelle lingue romanze.” – In: Revue romane 39, 68–96. – (2008): “Lexical vs. grammatical evidentiality in French and Italian.” – In: Linguistics 46, 917–947. Tasmowski, Liliane (2001): “Questions au conditionnel.” – In: Patrick Dendale & Liliane Tasmowski (eds.) (2001): Le conditionnel en français, 331–343. Paris: Klinksieck. (Université de Metz, Recherches linguistiques 25). Van der Auwera, Johan & Lejeune, Ludo (2005): “The morphological Imperative.” – In: Martin Haspelmath, Matthew Dryer, David Gil & Bernard Comrie (eds.): World Atlas of Language Structures, 286–289. Oxford: Oxford University Press. Venier, Federica (1991): La modalizzazione assertiva. Avverbi modali e verbi parentetici. – Milano: Franco Angeli. Verstraete, Jean-Christophe (2001): “Subjective and objective modality: Interpersonal and ideational functions in the English modal auxiliary system.” – In: Journal of Pragmatics 33, 1505–1528. Vet, Co (1990): “Asymmetries in the use of tense and modality.” – In: Jan Nuyts, A. Machtelt Bolkestein & Co Vet (eds.): Layers and levels of representation in language theory. A functional view, 123–137. Amsterdam/Philadelphia: Benjamins.

II.

Mood selection and mood alternation

Rui Marques

Modality, context change potential and mood selection in European Portuguese

The purpose of this paper is to shed some light on the semantic and pragmatic principles governing mood selection in European Portuguese. The proposal is made that the selection between the indicative and subjunctive moods is conditioned by three factors: veridicality, the kind of attitude expressed towards the proposition and the effect that the assertion of the proposition has on the context set. In some constructions, this last factor also seems to play a role in the choice between a finite mood and the infinitive, suggesting that the distribution of the infinitive is not constrained solely by syntactic factors.

1. Introduction A large part of the literature on the indicative and subjunctive moods has focused on the distribution of these moods in complement clauses. This is not the only context where these moods occur, but it is the context that shows greatest variation in mood selection. Hence, particular attention has been devoted to mood selection in complement clauses, in an attempt to establish which factors govern the selection of a particular mood. In this paper, I will consider data from European Portuguese, trying to describe the factors that seem to be operative in mood selection. Particular attention will be devoted to complement clauses, but some adverbial clauses will also be discussed. I assume that the identified factors may also play a role in the selection of mood in other Romance languages, though I will not present a comprehensive view of all of these. Nevertheless, some data from other Romance languages will occasionally be considered. The paper is organised as follows: The next section provides a general description of the distribution of indicative and subjunctive in European Portuguese. Section 2 presents a general overview of two of the main lines of enquiry that have been explored in the literature, arguing that none of them allows a satisfactory description of the data. In section 3, I will try to show that to a large extent the selection of the indicative or the subjunctive for complement clauses can be accounted for by the assumption that mood is a grammatical category related to the expression of modality. In particular, it

134

Rui Marques

will be argued that the selection of the indicative or the subjunctive follows from the combination of two factors: veridicality and type of modality. In 4.1, constructions will be considered in which both the indicative and the subjunctive can occur, the suggestion being made that to some extent the selection of mood is ruled by the effect that the speaker wants the assertion of the sentence to have on the discourse. This means that, apart from the factors identified in section 3, a comprehensive account of mood selection in Portuguese must also take into account some modelling of discourse. The consideration of this factor may shed some light on a less explored issue in the semantic literature: the selection of an infinitive or of a finite mood, a subject that will be partially considered in section 4.2.

2. Indicative and subjunctive in European Portuguese − a general overview In European Portuguese, both the indicative and the subjunctive moods may occur in independent and subordinated clauses of different types. In complement clauses, the selection of one or another mood is, to a large extent, conditioned by the predicate of the main clause: predicates of the types listed in (1) select the subjunctive, while those listed in (2) select the indicative. Subjunctive governors: (1a) desideratives − e. g. want, hope … (1b) directives − e. g. order, ask, suggest … (1c) predicates of fear − e. g. be scared (1d) epistemic predicates expressing negative commitment − e. g. doubt (1e) negative declaratives − e. g. deny (1f) some factive verbs1: astonish, find strange, like, regret, deplore, forgive … (1g) authorise, allow, forbid … (1h) be enough, cause, avoid, imply, try …

1

In the definition of factive verb, I follow Karttunen (1971), assuming that factive verbs are those which trigger the presupposition that their complement clause is true. That is, those verbs which allow the inference that the complement clause is true, regardless the truth value of the main clause.

Modality, context change potential and mood selection in European Portuguese

135

(1i) be convenient, be urgent … Indicative governors: (2a) declaratives − e. g. say, claim, confess … (2b) fiction verbs (a term I borrow from Farkas (1992)) − e. g. imagine, dream … (2c) commissives − e. g. promise … (2d) some factive verbs − e. g. know, foresee, realise, find out, ignore, notice …

There is a small group of verbs that accept both the indicative and the subjunctive in the complement clause, such as acreditar (‘believe’), supor (‘suppose’), and a few others, with related meanings. In independent and main clauses the indicative occurs in sentences describing reality, otherwise the subjunctive occurs. For instance, in independent clauses, the subjunctive is allowed in sentences expressing desire or similar notions (3a), in imperative sentences (3b) and sentences introduced by talvez (‘maybe’) (3c), while the indicative is the only accepted mood in declarative sentences (4): (3a) Que Deus te ouça! that God you hear ‘God hears you.’ (3b) Saiam! leave-3.PL ‘You leave.’

(3c) Talvez chova! maybe rain-3.SG ‘Maybe it rains.’ (4)

Está a chover. is-3.SG at rain ‘It is raining.’

In adverbial clauses, for the most part, the situation is similar: the indicative is selected for those sentences taken to be true, with the subjunctive being selected otherwise: (5a) A Ana foi morar para Paris quando acabou o curso. the Ana went live to Paris when finished-3.SG.IND the graduation ‘Ana moved to Paris when she graduated.’ (5b) A Ana vai morar para Paris quando acabar o curso. the Ana went live to Paris when finish-3.SG.SUBJ the graduation ‘Ana will move to Paris when she graduates.’

136

Rui Marques

Sentence (5a) allows the inference that the clause in italics is true, contrary to (5b). In most of the other adverbial clauses, the selection between the indicative and the subjunctive follows the same pattern: the indicative is selected if the sentence is taken to be true, otherwise the subjunctive occurs. The only exception is the concessive clause introduced by the connector embora (‘although’), where the subjunctive is selected, despite the fact that the subordinated clause is taken to be true: (6)

Embora esteja a chover, o dia está agradável. although is-3.SG.SUBJ at rain, the day is pleasant ‘Although it is raining, the day is pleasant.’

As for relative clauses, in restrictive relatives both moods may occur. Mood selection in relative clauses deserves a closer look, which I will not attempt in this paper. I will only note that the selection of the indicative or the subjunctive seems to be related to the referential degree of the relativized NP, as suggested by the following examples: (7a) A Ana quer contratar uma secretária que fala Russo. the Ana want hire a secretary that speak-3.SG.IND Russian ‘Ana wants to hire a secretary who speaks Russian.’ (7b) A Ana quer contratar uma secretária que fale Russo. the Ana want hire a secretary that speak-3.SG.SUBJ Russian ‘Ana wants to hire a secretary who speaks Russian.’ (8a) A Ana teve alguém que a ajudou. the Ana had someone who her help-3.SG.IND ‘Ana had someone who helped her.’ (8b) A Ana teve quem a ajudasse. the Ana had who her help-3.SG.SUBJ ‘Ana had help.’

Sentence (7a), with the indicative, allows the inference that there is a particular secretary that Ana wants to hire and she knows which secretary it is. On the contrary, (b) does not tell us that Ana has in mind a particular secretary that she wants to hire, and that secretary might even not exist. As for sentences (8a) and (8b), though both allow the inference that someone helped Ana, in the first one, with the indicative, the relativized NP is referential, pointing to some particular entity of the world, contrary to what is observed in the second one. In this case, the NP in italics does not identify anyone in particular; it only asserts the existence of someone who helped Ana.

Modality, context change potential and mood selection in European Portuguese

137

3. Semantic and pragmatic approaches to mood selection Simplifying somewhat, one might say that most of the semantic and pragmatic approaches to mood selection have followed two main approaches. One of them establishes a link between the selection of mood and the kind of speech act that the clause is primarily related to. The other approach seeks to relate the selection of indicative or subjunctive to the acceptance of the truth of the relevant proposition. These two approaches will be considered in the following subsections. 3.1. The assertion/non-assertion hypothesis One of the most thorough accounts of the selection of indicative or subjunctive in Romance languages assumes that mood selection is related to speech acts, the indicative being the mood of assertion and the subjunctive the mood selected for non assertive contexts. The following arguments have been presented in favour of this hypothesis (cf. e. g. Klein-Andreu 1990, Bybee & Terrell 1990, Ahern 2005): (i) The indicative is selected in declarative clauses, while the subjunctive occurs in main clauses that express order, desire or some related notion, as shown by the following examples: (9a) As crianças estão caladas. the children are-IND quiet ‘The children are quiet.’ (9b) Estejam caladas! be-3.PL.SUBJ quiet (9c) Deus queira! God wishes-SUBJ ‘God willing!’

(ii) The indicative is selected by causal connectives, while the subjunctive is selected in purpose clauses: (10a) A Ana saiu porque estava atrasada. the Ana went out because was-3.SG.IND late ‘Ana went out because she was late.’ (10b) Construíram uma barragem para que o rio não transbordasse. built-3PL a dam for that the river not overflow-SUBJ ‘They built a dam in order to prevent the river from overflowing.’

138

Rui Marques

(iii) In complement clauses, the indicative is selected by declarative predicates (cf. (11a)), while the subjunctive is selected by verbs expressing desire, doubt, order, fear, among other notions (cf. e. g. (11b)): (11a) A Ana disse que o computador está avariado. the Ana said that the computer is-IND broken ‘Ana said that the computer is broken.’ (11b) A Ana quer que o filho acabe o curso. the Ana wants that the son finishes-SUBJ the degree ‘Ana wants her son to finish his degree.’

Moreover, the interpretation of some verbs − e. g. insist or feel − is different, depending on whether the indicative or subjunctive occurs in their complement clause. In fact, as Klein-Andreu (1990), among others, observes, in Spanish the verb feel is interpreted as equivalent to notice if its complement clause is in the indicative mood, while it is interpreted as the equivalent to regret if the subjunctive is selected: (12a) Siento que aprende. (Spa.) feel-1.SG that learns-3.SG.IND ‘I feel that he is learning.’ (12b) Siento que aprenda. feel-1.SG that learns-3.SG.SUBJ ‘I regret that he is learning.’

The assumption that the indicative is selected in assertive clauses and the subjunctive in non-assertive clauses goes back at least to Hooper (1975), who proposes the following classification of sentence complementation predicates: Table 1: Hooper’s classification of predicates. Non-factives think, believe, say, agree Assertives Semi-factives find out, discover, know Non-factives be likely/possible/probable Non-assertives Factives regret, resent, be odd/strange

Both in Spanish (the language Hooper examines) as in most other Romance languages, Hooper’s assertive predicates take the indicative, while the subjunctive is selected by non-assertive predicates. Despite the initial appeal of the hypothesis that the indicative is the mood of assertion and that the subjunctive is used in non-assertive contexts, some criticism has been pointed out in the literature. A major problem for this analysis, as Palmer (1986) observes, is the selection of the indicative in interrogatives, a clearly non-assertive context. Another difficulty is the fact

Modality, context change potential and mood selection in European Portuguese

139

that some epistemic predicates accept both the indicative and the subjunctive in their complement clause. For instance, believe preferably selects the subjunctive in Italian, and the indicative in the other Romance languages. Finally, as stressed by several authors, the hypothesis at stake lacks a semantic basis that underlies the notion of ‘assertion’. Particularly in the case of complement clauses, it is hard to grasp a feature common to all Hooper’s assertive predicates that distinguishes them from the non-assertive predicates. Several authors (cf. Noonan 1985, Bybee & Terrell 1990, Hengeveld 2004, among others) argue that factive predicates such as regret select the subjunctive because their complement clause is presupposed, not asserted. As for the other subjunctive governors (the non-assertive and non-factive predicates in Hooper’s classification), they do not commit the speaker or anybody else to the truth of their complement clause and, thus, they are not assertive predicates. This view falls back on a Stalnakerian concept of ‘assertion’. It explores the idea that a proposition with the indicative mood is presented as conveying new information. However, an approach along these lines fails to explain the selection of the indicative in the sentence in italics of (13) and the selection of subjunctive in examples like (14): (13) Todos sabemos que a Ana reprovou no exame; não precisas de nos lembrar. all know-1.PL that the Ana failed-IND in-the exam; not need-2.SG of us remember ‘We all know that Ana failed the exam; you don’t need to remind us.’ (14) Como as imagens mostram, a barragem não impediu que o rio transbordasse. as the pictures show, the dam not prevented that the river overflew-SUBJ ‘As the pictures show, the dam didn’t prevent the river from overflowing.’

The sentence in italics in (13) is not intended to add new information to the common ground and the sentence in italics in (14) might convey new information. However, the indicative is obligatorily selected in the first case and excluded from the second case, thus contradicting the hypothesis that the indicative is used if new information is conveyed, with the subjunctive selected otherwise. 3.2. The veridicality hypothesis Another line that has been explored in the literature connects the selection of the indicative or subjunctive to the truth value of the proposition. According to this approach, the indicative is selected if the sentence is taken to be true; otherwise the subjunctive is chosen. This is the idea found in traditional grammars, where the link is established between the indicative and reality, on

140

Rui Marques

the one hand, and between the subjunctive and unreality or virtuality, on the other hand. A similar intuition is explored by Palmer (1986), Bell (1990), and others, who relate the selection of mood to the degree of belief being expressed. According to these authors, a high degree of commitment to the truth of the proposition leads to the selection of the indicative; otherwise the subjunctive is used: The indicative is used where the subject shows some positive degree of commitment to the proposition, either total as with ASSERT, or partial as with THINK. Where there is no degree of positive commitment as with BE POSSIBLE, but either noncommitment or negative commitment as with DOUBT (partial negative commitment) or don’t think (total negative commitment), the subjunctive is used. (Palmer 1986: 145)

One fact that favours this hypothesis is shown in the Spanish examples (15a– b), from Palmer (1986), and the Portuguese examples (16a–b): (15a) Tal vez venga mañana. maybe comes-SUBJ.3.SG tomorrow (15b) Tal vez vendrá mañana. maybe comes-IND.3.SG tomorrow ‘Maybe he comes tomorrow.’ (16a) Eu acredito que a Ana ganhe as eleições. I believe that the Ana wins-SUBJ the elections (16b) Eu acredito que a Ana ganha as eleições. I believe that the Ana wins-IND the elections ‘I believe Ana will win the elections.’

Though both the indicative and the subjunctive might occur in these cases, sentences (15a) and (16a), with the indicative, express a higher degree of certainty than (15b) and (16b), where the subjunctive is selected. Therefore, in the examples below, the selection of the subjunctive is odd in sentence (17a), which expresses a high degree of belief, and the selection of the indicative is odd in sentence (18a), which expresses a low degree of belief: (17a) ?Acredito convictamente que haja sobreviventes. believe-1.SG really that is-3.SG.SUBJ survivors. (17b) Acredito convictamente que há sobreviventes. believe-1.SG really that is-3.SG.IND survivors. ‘I really believe that there are survivors.’ (18a) ?É difícil acreditar que alguém sobreviveu. is hard believe that someone survived-IND

Modality, context change potential and mood selection in European Portuguese

141

(18b) É difícil acreditar que alguém tenha sobrevivido. is hard believe that someone has-SUBJ survived ‘It is hard to believe that someone has survived.’

Nevertheless, the proposal that the indicative occurs in sentences taken to be true and the subjunctive in the other cases faces two main problems. The first one, as Farkas (1992) observes, is that it does not account for the selection of the indicative by verbs like dream or imagine. Such verbs select the indicative, as shown in the following sentences, although their complement proposition does not describe reality and they do not allow the inference that someone believes that their complement proposition is true. (19) Imagina que estavas no Brasil. imagine-IMP.2.SG that were-IND.2.SG in-the Brazil ‘Imagine you were in Brazil.’ (20) Ele sonhou que era um astronauta. he dreamed that was-3.SG.IND an astronaut ‘He dreamed he was an astronaut.’

The second problem is that the hypothesis at stake fails to account for the selection of the subjunctive in sentences taken to be true. Such is the case with complement clauses of factive predicates like regret or be a shame (cf. (21)), which in Portuguese and other languages select the subjunctive, complement clauses of some nouns, like the equivalent of fact in French (cf. (22)), and concessive clauses introduced by embora (‘although’) (cf. (23)): (21) É pena que esteja a chover. (Por.) is pity that is-3.SG.SUBJ raining ‘It is a pity that it is raining.’ (22) Le fait qu’il soit malade l’ennuie. (Fra.) the fact that he is-SUBJ ill him bothers ‘The fact that he is ill bothers him.’ (23) Embora esteja doente, ele está a trabalhar. (Por.) although is-SUBJ ill, he is at work ‘Although he is ill, he is working.’

The first problem is solved by the concept of ‘extensional anchoring’, proposed by Farkas, or ‘(relativised) veridicality’, as defined in Giannakidou (1999). According to this view, the indicative is selected if the relevant clause is taken to be true by some entity in a particular possible world, not necessarily the real world. Hence, the verb to dream, for instance, selects the indicative because its complement clause is taken to be true in the world modelling the main subject’s dreams, though it might be false in the real world.

142

Rui Marques

As Farkas (1992) observes, this approach faces problems when French (and other Romance languages, one might add) is considered, given that the subjunctive is selected by some factive predicates, such as (the equivalent of) regret, which she labels ‘factive-emotives’. Since the complement clause of such predicates is taken to be true, the proposal predicts that the indicative will be selected. In Romanian, this prediction is confirmed, but not in the other Romance languages. Farkas suggests that the parametric variation in mood selection shown by these verbs is due to the fact that some languages are more sensitive to their factive character, while other languages give prominence to their emotive feature. Romanian, as well as Modern Greek (cf. Giannakidou 1999) and other languages where these predicates select the indicative, would belong to the first group, while the other Romance languages, where these predicates select (obligatorily, as in Portuguese, or optionally, as in French2) the subjunctive, would belong to the second group. However, even if this idea might account for the selection of the subjunctive by factive-emotive predicates, it cannot explain the selection of the subjunctive by causative predicates in sentences like (14), above, or (24): (24) A rápida intervenção dos bombeiros permitiu que toda a gente fosse salva. the quick intervention of-the firemen allowed that every the people was-SUBJ saved ‘Thanks to the quick intervention of the firemen, everybody was saved.’

In these constructions, where the sentence with the subjunctive is given as true, the selection of the subjunctive cannot be explained in the same way that Farkas suggests for complement clauses of factive emotive predicates. In short, a proposal that links the selection of indicative or subjunctive to the assumption of the truth of a proposition (even if the acceptance of the truth is relativized to possible worlds different from that modelling reality) fails to explain the selection of subjunctive by causative predicates and the fact that in most Romance languages factive-emotive predicates govern the subjunctive. In the next section, I will present a different analysis, following the claim that the grammatical category mood is a means to express modality.

2

See Farkas (1992), Quer (1988) and Quer (this volume) for a clearer description of the selection of mood by factive-emotive predicates in French, Spanish and Catalan.

Modality, context change potential and mood selection in European Portuguese

143

4. Modality and mood selection Taking the concept of modality to refer to the attitude expressed towards a proposition (cf. e. g. Palmer 1986), a relationship can be found between this semantic concept and the choice of indicative or subjunctive in most contexts. To show this, I will consider complement clauses, but the proposed analysis may be extended to constructions of other kinds. Let us begin by considering factive verbs, a class of verbs that has been at the heart of a large debate on the literature on mood. While in Romanian all such verbs govern the indicative, in the other Romance languages, some of them select the indicative while the others select3 the subjunctive. Factive verbs that select the indicative include (the equivalents of) know, foresee, realise, find out, notice and be unaware of (in the sense of not knowing). All these verbs express an attitude of an epistemic nature. As for the factive verbs that select the subjunctive, they express a different kind of attitude, one that corresponds to an evaluation of the state of affairs described by the complement clause. This is the case with factive verbs like (the equivalents of) astonish, like, regret, deplore or forgive and predicative expressions like (the equivalents of) be a pity or be pleased. With this is mind, it is clear that the factive verbs that select the indicative are those which express an epistemic attitude towards the complement clause, that is, factive verbs related to epistemic modality4, while those that select the subjunctive are those related to what Rescher (1968) calls evaluative modality. Considering now the non-factive verbs, one may notice that those that select the indicative express an attitude of belief. This is clearly the case with the equivalents of think or assume, and also of declarative predicates, given that they commit the referent of the main clause subject to the acceptance of the truth of the complement clause. Similarly, commissive predicates (e. g. promise), which also take the indicative, allow the inference that their complement clause will become true according to the main clause subject, otherwise the speech act that such predicates express is infelicitous, in Austin’s sense. Finally, fiction verbs also express an attitude of belief, though the belief is relativised to worlds (possibly) different from the real one. This being so, the conclusion arises that all verbs that select the indicative express an attitude of knowledge or belief, that is, an attitude of epistemic or doxastic nature. On the other hand, a connection between the selection of the subjunctive and a particular kind of modality does not seem to exist. In fact, 3 4

Or accept, see footnote 2. In the broad sense, which includes both epistemic (i.e. related to knowledge) and doxastic (i.e. related to belief) modality.

144

Rui Marques

the subjunctive is associated with a wide variety of attitudes, being selected by bouletic predicates (e. g. want or desire), evaluative predicates (e. g. regret or be pleased) predicates of a deontic nature (e. g. force or forbid), and so on. Thus, contrary to what is generally assumed, the subjunctive does not seem to signal any particular value, contrary to the indicative, which occurs exclusively in contexts of epistemic modality. Nevertheless, it should be noticed that the subjunctive may also occur in epistemic environments. In fact, there is a group of epistemic predicates that may select the indicative or the subjunctive in Portuguese. This is the case with a small group of verbs such as the equivalents of believe, admit or suspect. As shown by (16a–b) above, verbs of this kind select indicative or subjunctive depending on the degree of belief being expressed. Moreover, the negative verbs duvidar (‘doubt’) and negar (‘deny’) also accept the subjunctive in their complement clauses, as shown by the following example: (25) duvido/nego que o Paulo tenha sido assaltado. doubt-1.SG / deny-1.SG that the Paulo has-SUBJ been robbed ‘I doubt/deny that Paulo was robbed.’

These verbs are also epistemic, in the sense that they also express an epistemic (or better, doxastic) attitude relating the subject of the main clause to the complement proposition. This being so, one has to conclude that epistemic modality is not the only issue that leads to the selection of the indicative. Let us consider the verbs doubt and deny. One obvious difference between these verbs, on the one hand, and the majority of the verbs that select the indicative on the other, is that the former do not allow the inference that the subject of the main clause accepts the truth of the complement clause, contrary to most indicative governors. In other words, the verbs doubt and deny express a negative epistemic attitude, contrary to indicative governors such as conclude or say, for example. Other epistemic verbs which do not indicate that the (entity referred by the) subject of the main clause takes the complement proposition to be true are the factive verbs be unaware of (in the sense of not knowing) and forget, which, however, do govern the indicative: (26) a Ana ignora/esqueceu-se de que a sala estava fechada the Ana is unaware of / forgot-herself that the room was- IND closed ‘Ana doesn’t know / forgot that the room was closed.’

Nevertheless, such verbs are factive. They allow for the inference that their complement clause is true according to the speaker. On the contrary, the verbs doubt and deny do not commit anyone to the truth of their complement clause. Therefore, apart from the kind of attitude expressed by the predicate, another factor that plays a role in the selection of the indicative or the subjunctive in

Modality, context change potential and mood selection in European Portuguese

145

complement clauses in Portuguese seems to be Giannakidou’s concept of ‘veridicality’. According to Giannakidou (1999), in Modern Greek the subjunctive is selected by non veridical operators; i. e., those operators that do not allow the inference that the proposition they introduce is true in some epistemic model. On the basis of this notion, one might observe that the verbs ignore and forget are veridical, since they allow the inference that their complement clause is true, according to the speaker, while doubt and deny are non-veridical verbs, since they do not allow the inference that their complement clause is taken to be true by someone. Apart from this difference, both groups of verbs − ignore and forget, on one side; doubt and deny, on the other side − express an epistemic attitude relating the subject of the main clause to the complement proposition. Thus, the selection of the indicative or the subjunctive for complement clauses in Portuguese seems to follow from a combination of two factors: veridicality and epistemic modality. The reason for the selection of one mood over the other can be stated as follows: the indicative is selected for veridical contexts − i. e., for propositions taken to be true by some entity in some possible world − if the attitude towards the complement proposition is of epistemic nature; otherwise, the subjunctive is selected. In other words, if the attitude expressed by the main predicate is the belief in the complement proposition, the indicative is selected. If the predicate expresses some other kind of attitude (e. g. desire, evaluation …) or if the sentence is not taken to be true, the subjunctive is selected. Let us now consider two cases of mood variation in complement clauses of epistemic non-factive verbs. In both cases, the variation can be accounted for by the hypothesis that the indicative is selected in veridical epistemic contexts. The first case is illustrated by the following examples: (27a) Acredito que a Maria está doente.5 believe-1.SG that the Maria is-IND ill (27b) Acredito que a Maria esteja doente. believe-1.SG that the Maria is-SUBJ ill ‘I believe that Maria is ill.’

As mentioned above, in these sentences the selection of one or another mood is related to the degree of belief being expressed. The indicative signals a high degree of belief, the subjunctive a lower degree. In other words, the indicative 5

The group of verbs that allow both the indicative and the subjunctive to occur in their complement clause includes the equivalent of believe and some other verbs with the same core meaning (e. g. admit, suspect, assume).

146

Rui Marques

is selected if the complement clause is taken to be true; otherwise the subjunctive is selected. Hence, the concept of veridicality accounts for this case of mood variation. With the indicative, the inference follows that the relevant proposition is true (according to the subject of the main clause), contrary to what happens if the subjunctive is selected. As expected, if the main verb is negated, absence of belief is expressed and only the subjunctive is allowed: (28) Não acredito que a Maria esteja/*está6 doente. not believe-1.SG that the Maria is-SUBJ / *is-IND ill ‘I don’t believe that Maria is ill.’

However, as observed by Guitart (1984) and Hengeveld (1988), among others, if the subject of the main clause is different from the speaker, both the subjunctive and the indicative may occur: (29a) A Ana não acredita que a Maria esteja doente. the Ana not believes that the Maria is-SUBJ ill (29b) A Ana não acredita que a Maria está doente. The Ana not believe that the Maria is-IND ill ‘Ana doesn’t believe that Maria is ill.’

The same is observed with the verb duvidar (‘doubt’): (30a) A Ana ainda duvida que a Maria esteja doente. the Ana still doubts that the Maria is-SUBJ ill (30b) A Ana ainda duvida que a Maria está doente. the Ana still doubts that the Maria is-IND ill ‘Ana still doubts that Maria is ill.’

The selection of the indicative or the subjunctive in these cases illustrates a second case of mood variation that can be accounted for by the hypothesis that the indicative occurs in epistemic veridical contexts. Mood choice in examples (29) and (30) is not a matter of free choice. The complement clause with the indicative allows for the inference that the speaker accepts the truth of the complement proposition, while the subjunctive in the complement clause does not allow for the inference that the sentence is taken to be true by some entity. Naturally, in (28), the indicative may not occur because a contradiction would arise. On the one hand, the indicative would signal that the speaker takes the complement clause to be true; on the other hand, the sentence expresses the 6

The indicative is accepted if the sentence does not describe absence of belief, but surprise (roughly equivalent to I can’t believe it! Maria is ill!). In this case, the sentence is taken to be true.

Modality, context change potential and mood selection in European Portuguese

147

information that the same entity − the (individual identified by the) subject of the main clause − does not believe that the complement clause is true. No contradiction arises, though, if the sentence refers to a previous state of belief, as in the following examples: (31a) Eu não acreditei que a Maria estava doente. I not believed that the Maria was-IND ill (31b) Eu não acreditei que a Maria estivesse doente. I not believed that the Maria was-SUBJ ill ‘I didn’t believe that Maria was ill.’

Sentence (31a), where the indicative is used, conveys the information that, at the time of utterance, the speaker believes that the complement proposition is true, contrary to his former belief. Sentence (31b), where the subjunctive is used, only states that the speaker did not believe, at some prior moment, that Maria was ill. It does not allow for the inference that at the speech time, the speaker believes that the proposition is true or false. In summary, the sentences that express lack of belief show that when the indicative is selected different states of belief are considered. In examples like (29b) or (30b), the indicative signals the contrast between the state of belief relating the complement clause to the speaker and the one concerning the complement clause and the (entity refereed by the) main subject. In examples like (31a), the indicative signals a contrast between the present and a previous state of belief of the speaker concerning the complement clause, stating that he now believes in something that he did not believe before. In either case, when the indicative occurs in the complement clause, the sentences are taken to be true (according to the speaker) and the attitude expressed towards the proposition is one of (non)belief (i. e. the complement clause occurs in a context of epistemic, or doxastic, modality). To sum up, the analysis of mood selection for complement clauses in Portuguese reveals that the indicative occurs in veridical epistemic contexts. That is, the indicative occurs when the complement clause is taken to be true (either in the real world − cf. e. g. the complement clause of verbs like know − or in some other possible world − cf. e. g. the complement clause of verbs like dream) and the attitude expressed towards this proposition is one of epistemic or doxastic nature. This accounts for the cases of mood variation observed so far: (i) cases where the selection of the indicative or subjunctive depends on different degrees of belief being expressed; (ii) cases where the selection of mood is related to different states of belief (states of belief relating to different individuals or to the same individual at different times). In either case, the indicative is selected if the complement clause is presented as one which some

148

Rui Marques

entity assumes to be true at some point, with the subjunctive being selected otherwise. In the next section, other cases of mood variation will be taken into consideration.

5.

Context change potential and mood selection

5.1. Indicative vs. subjunctive To a large extent, the hypothesis that the selection of the indicative or the subjunctive is related to veridicality and the kind of attitude being expressed towards the proposition accounts for the data from Portuguese (and from other Romance languages). However, apart from the cases considered above, there are other instances of mood variation which, at first sight, cannot be accounted for along the same lines. One of them relates to the epistemic verb pensar (‘to think’), which accepts both the indicative and the subjunctive in its complement clause without any semantic distinction. This is shown in the following sentences: (32a) Pensei que a Maria estava doente. thought-1.SG that the Maria was-IND ill (32b) Pensei que a Maria estivesse doente. thought-1.SG that the Maria was-SUBJ ill ‘I thought that Maria was ill.’

Both (32a) as (32b) may be felicitously asserted in a context where the speaker accepts that Maria is ill, as well as in a context where he accepts that she is not. Hence, in these cases the selection of one mood or another does not seem to be due to the (non) acceptance of the truth of the complement proposition. Another context of variation in mood selection concerns the complement clauses of the verb believe. As has been observed in the literature, in Italian, the equivalent of this verb preferably selects the subjunctive, while in the other Romance languages, it is the indicative that most naturally occurs in its complement clause: (33) Credo che Maria sia malata. (Ita.) believe-1.SG that Maria is-SUBJ ill (34) Je crois que Marie est malade. (Fra.) I believe that Marie is-IND ill

Modality, context change potential and mood selection in European Portuguese

149

These facts show that, apart from the two issues identified as operative in the selection of the indicative or the subjunctive − veridicality and epistemic modality −, some other feature plays a role in mood selection. Mood variation in these cases seems to be accounted for if one invokes Heim’s concept of context change potential (CCP) (cf. Heim 1992). Roughly, the CCP of a sentence is its capability of changing the context set (the set of possible worlds available at each point of the conversation). According to Heim’s framework, by asserting a declarative sentence the speaker gives the instruction to remove from the context set all but those worlds where the complement proposition is verified. For instance, if the sentence it is raining is successfully asserted against a context set c, the result will be a new context set: a subset of c which contains only worlds where it is raining. In the examples under consideration, the selection of the indicative or the subjunctive might be related to the instruction that the speaker is willing to give by asserting the sentence. The hypothesis is that, if the indicative is selected, the instruction will be given to consider only those possible worlds where the relevant proposition is verified. On the other hand, if the subjunctive is selected, the update of the context set would allow for possible worlds where the proposition is false to remain available. To evaluate this hypothesis, let us compare the Italian sentence (33) with (35a), from Wandruszka (1991), where the indicative is necessarily selected: (35a) Creda che sono veramente mortificato. (Ita.) believe-3.SG.IMP that am-1.SG.IND really sorry (35b) *Creda che io sia veramente mortificato. believe-3.SG.IMP that I am-1.SG.SUBJ really sorry ‘Believe me, I am really sorry.’

By asserting (35a), the speaker gives the instruction to discard possible worlds where the complement proposition is not verified. On the contrary, by asserting (33), with the subjunctive, possible worlds where the complement proposition is not verified might remain available in the context set after the (successful) assertion of the sentence. Thus, a phrase like ‘but I might be wrong’ is a possible continuation of (33), but not of (35a). A similar explanation might be proposed for the Portuguese cases (32a–b). By asserting (32a), with the indicative in the complement clause, the speaker is focusing on his former belief that Maria was sick, remaining silent as to whether, at the time of utterance, he accepts the proposition to be true or not. Hence, by asserting the sentence, the speaker gives the instruction to consider (even if temporally, given that discourse might continue with a sentence like but I was wrong) only possible worlds where Maria was sick at a previous time. On the other hand, the speaker might want to say that at the speech time

150

Rui Marques

he admits the sentence to be false, in which case he wants the hearer to also consider worlds where the proposition is not verified. The selection of the indicative or subjunctive seems to be related to the instruction the speaker is giving. When all the possible worlds to be considered are those where the proposition is verified, the indicative is selected. If the instruction is given to also consider worlds where the proposition is not verified, the subjunctive will be preferred. Hence, the selection of the indicative in (32a) as in (35a), and the selection of the subjunctive in (32b) and (33). To sum up, so far three factors have been considered to be relevant for mood selection in complement clauses: (i) the kind of modality that the main predicate is associated with, (ii) veridicality, and (iii) the context change potential of the sentence (i. e., when indicative is selected, the instruction is given to discard, even if temporally, possible worlds where the sentence is not verified). In a simplified manner, the proposal being advocated is that the indicative is selected when the purpose is to express the acceptance of the truth of the proposition, with the subjunctive being selected otherwise. Thus, the subjunctive occurs in non-veridical contexts (such as complement clauses of desire predicates) and in veridical non-epistemic contexts (such as complement clauses of predicates like regret), since in this case, the primary purpose is not to express belief in the relevant proposition. Epistemic non-factive verbs (such as believe or think) show variation in mood selection depending on whether the instruction is given to consider only worlds where the relevant proposition is verified. Moving now to main and adverbial clauses, it might be observed that the indicative occurs only in those contexts where the speaker expresses his belief in the proposition. In fact, the indicative is the mood found in declarative simple sentences, as shown by (36), in causal clauses, as in (37), and in temporal clauses that describe events that occurred, as shown by (38) and (39), or which have atemporal or generic readings, as in (40): (36) Viena é a capital da Áustria. Vienna is-IND the capital of-the Austria ‘Vienna is the capital of Austria.’ (37) Ele caiu porque a Ana o empurrou. he fell because the Ana him pushed-IND ‘He fell because Ana pushed him.’ (38) Ele veio quando o chamaram. he came when him called-3.PL.IND ‘He came when he was called.’

Modality, context change potential and mood selection in European Portuguese

151

(39) Enquanto ele esteve doente, nunca foi chamado. while he was-IND ill, never was-1.SG called ‘While he was ill, he was never called.’ (40) O bebé chora quando tem fome. the baby cries when is-3.SG.IND hungry ‘The baby cries when he’s hungry.’

In these cases, by asserting the sentence, the speaker expresses his belief that the proposition is true, hence the indicative is selected. As for the subjunctive, it occurs in sentences that are not taken to be true, such as (41) – (43), but also in concessive clauses introduced by embora (‘although’), such as (44), a veridical context: (41) Talvez a Ana esteja em casa. maybe the Ana is-SUBJ at home ‘Maybe Ana is at home.’ (42) Ele vem quando/se o chamarem. he comes when/if him call-3.PL.SUBJ ‘He will come when/if he is called.’ (43) Ele saiu antes que começasse a chover. he left before that start-3.SG.SUBJ at rain ‘He left before it would start raining.’ (44) Embora esteja frio, o dia está agradável. although is-3.SG.SUBJ cold, the day is pleasant ‘Although it is cold, the day is pleasant.’

The selection of the subjunctive in (41) – (43) is explained by the concept of veridicality. Since the sentence is not taken to be true, the indicative cannot be selected. The selection of the subjunctive in (44) cannot be due to the same factor, since concessive clauses are a veridical context. The selection of the subjunctive in a case such as this is most naturally explained by a discourse parameter. One hypothesis is to presume that, by asserting the concessive clause, the speaker assumes that this clause describes a state of affairs known prior to the utterance. I think that in a context where the speaker assumes he is giving new information, an adversative clause, that demands the indicative and is semantically equivalent to a concessive clause, is more natural than a concessive clause with embora7. If this is so, the proposal can be put forward 7

A reviewer observes that the English translation of (44) might be uttered as an answer to the question how is the weather?, in which case both the main and the concessive clauses introduce new information. I think that in Portuguese (44) would sound a bit strange as an answer to this question, particularly if the concessive clause

152

Rui Marques

that the selection of the indicative or subjunctive in main and adverbial clauses depends on the context change potential of the sentence. If the purpose is to give the instruction to remove from the context set all but those worlds where the proposition is true, the indicative occurs, otherwise the subjunctive is selected. In the case of concessive clauses introduced by embora, the selection of the indicative will not be justified because the proposition is presupposed; that is, there are no worlds in the context set where the proposition is false and, thus, the instruction to remove such worlds makes no sense.8 5.2. Infinitive vs. finite moods So far, only the indicative and subjunctive moods have been considered. Another major mood of the Romance languages is the infinitive, which has not received the same amount of attention as the subjunctive and the indicative in the semantic literature on mood in the Romance languages. I will not provide a comprehensive analysis of the infinitive in Portuguese, but will concentrate only on two kinds of adverbial clauses which might select the infinitive or a finite mood. In Portuguese, as in other languages, the infinitive may occur in different types of sentences, including the following kinds of adverbial clauses, which will be considered in this paper: clauses introduced by causal connectives, and clauses introduced by the temporal connectives até ‘until’ or antes ‘before’. In the next subsections, the hypothesis will be considered that the selection of the infinitive or of a finite mood in these kinds of adverbial clauses is conditioned by the effect that the assertion of the sentence has on the context set. Specifically, I will try to show that, in the kinds of clauses under investigation, the infinitival clause is not capable of changing the context set, contrary to what happens if a finite mood is chosen.

precedes the main clause. This seems to be a particularity of the concessive conjunction embora. A concessive clause introduced by apesar de, which does not select the subjunctive, would sound natural: (i)

8

apesar de estar frio, o dia está agradável. despite of be.INF cold, the day is plesant ‘Although it is cold, the day is pleasant.’

In colloquial speech, indicative might occur in concessive clauses introduced by embora ‘although’ if the speaker assumes he is giving new information. Such a fact favours this hypothesis.

Modality, context change potential and mood selection in European Portuguese

153

5.2.1. Mood selection by causal connectives In Portuguese, causal connectives may introduce a finite clause with the indicative mood as well as an infinitival clause: (45a) A Ana ficou em casa porque estava doente. the Ana stayed at home because was-3.SG.IND ill (45b) A Ana ficou em casa por estar doente. the Ana stayed at home by be-INF ill ‘Ana stayed home because she was ill.’

These two sentences do not differ in meaning. However, the contexts in which the sentences are appropriately uttered only partially match up. In fact, in a context where it is known that Ana was ill, both (45a) and (45b) might be uttered, but in a context where such information is not available only (45a), with the indicative, seems adequate. In other words, the infinitival clause does not seem to be capable of introducing new information into discourse, contrary to the finite indicative clause. This fact becomes clearer if negative sentences are taken into consideration. In this case, not only the indicative and the infinitive but also the subjunctive may occur in the clause introduced by the causal connective: (46a) A Ana não ficou em casa porque estava doente. the Ana not stayed at home because was-3.SG.IND ill (46b) A Ana não ficou em casa por estar doente. the Ana not stayed at home because be-INF ill ‘Ana didn’t stay home because she was ill.’ (46c) A Ana não ficou em casa porque estivesse doente, ficou em casa porque estava à espera de visitas. the Ana not stayed at home because was-3.SG.SUBJ ill, stayed home because was at the expecting guests ‘Ana didn’t stay home because she was ill, she stayed home because she was expecting guests.’

These sentences differ as to the readings they display, three readings being possible. The first one is the metalinguistic reading of negation, which corresponds to the denial of the related affirmative sentence, previously asserted. Such reading is available in (46a) and (46b), but not in the sentence in italics in (46c), given that in affirmative sentences the clausal connective may not introduce a subjunctive clause. The other possible readings correspond to the internal and external reading of negation. In the internal reading, only the main clause is within the scope of negation. That is, the sentence expresses the information that Ana did not stay home and the reason

154

Rui Marques

for this is that she was ill. Given this reading, a possible continuation of the discourse is the sentence she went to the hospital. The external reading of negation corresponds to denying that the reason why Ana stayed home was her illness. The metalinguistic reading of negation is a special case of external reading, although not the only one. In fact, a negative sentence may have an external reading without the corresponding affirmative sentence having been previously asserted. This non-metalinguistic external reading of negation is barely accepted for (46a), with the indicative mood, but is available in (46b), where the infinitive is used, and is the only possible reading of (46c), where the subjunctive mood is used. As for the internal reading, it is the only reading displayed by (46a), with the indicative, and is also available in (46b), with the infinitive. Thus, the infinitive allows for the external and internal reading, the subjunctive allows for the external reading only and the indicative only displays the internal reading. Apart from this difference concerning the available readings, sentences (46a–c) also show what one might call a pragmatic difference. To see this, let us first observe that the internal reading of negation allows for the inference that the proposition introduced by the causal connective is true, i. e., the causal connective is a veridical operator9 (if p because q is true, then q is true). The external reading of negation, on the contrary, allows neither this inference nor the inference that the relevant proposition is false. With these observations in mind, let us consider the assertion of (46a), (46b) and (46c) in a context where the information that Ana is ill is not available − context A − and the assertion of such sentences in a context where this information is part of the common ground − context B. In context A, any of the sentences (46a–c) might be felicitously asserted. Sentence (46a), where the indicative is used, will be interpreted as having internal negation and (if the sentence is accepted) the context will be updated with the information that Ana was ill. On the contrary, if (46b), where the infinitive is used, or (46c), where the subjunctive occurs, are uttered against context A, only the external reading of negation is available, and the context is not updated with the information that Ana was ill. In context B, (46a), with the indicative, and (46b), with the infinitive, may be felicitously uttered, contrary to (46c), with the subjunctive. In this context, (46a), with the indicative, has an unambiguous internal reading of negation, while (46b), where the infinitive occurs, shows ambiguity between the internal and external reading. The following table summarises the possible readings of sentences with the same form as (46a–c) and the contexts where they might be felicitously 9

See Sanchez Valencia et al. (1993) for the definition of veridical operators.

Modality, context change potential and mood selection in European Portuguese

155

asserted (p stands for the proposition introduced by the causal connective; the mood of this clause is registered in the left column; c stands for the common ground). Notice that the relevant external reading of negation is not the metalinguistic reading. Table 2: Mood selection by causal connectives in negative sentences. não q porque p (not q because p) Internal negation External negation [¬ q because p] ¬ [q because p] p∈c p∉c p∈c p∉c Subjunctive * * * o.k. Infinitive o.k. * o.k. o.k. Indicative o.k. o.k. * *

In the constructions under consideration, the possibility of selecting the subjunctive is constrained by the truth value of p: the subjunctive may be selected only if p is not taken to be true. Thus, the clause corresponding to proposition p may only be uttered in a context where p does not belong to the context set and the external reading of negation is the only one available (since the internal reading allows for the inference that p is true). As for the indicative, it is compatible with the internal reading of negation, the reading that allows for the inference that the proposition introduced by the causal connective is true, but is not compatible with the external non-metalinguistic reading of negation. Finally, the selection of the infinitive is not conditioned by the truth value of p, this mood being compatible with both the external and the internal readings of negation. The only case where the infinitive is ruled out corresponds to the third column of the table: p is not known to be true and the negation has internal reading. In this case, p is given to be true (the causal connective is veridical) and the assertion of p updates the context with that information. The fact that the infinitive might not be selected in this case shows that the infinitive is not capable of updating the context with new information. The indicative turns out to be the only mood capable of introducing into the discourse the information that p is true. In other words, by asserting a causal sentence with the indicative, the speaker can give the instruction to remove from the context set all but those worlds where the sentence is true, while the assertion of an infinitive causal sentence provides no change in the context set.

156

Rui Marques

5.2.2. Mood selection with the temporal connectives até and antes In Portuguese, the prepositions até (‘until’) and antes (‘before’) may introduce an infinitive clause or a clause with the subjunctive mood: (47a) A Ana vai ficar à espera até alguém a chamar. the Ana will stay to-the wait until someone her call-INF (47b) A Ana vai ficar à espera até que alguém a chame. the Ana will stay to-the wait until someone her call-SUBJ ‘Ana will wait until someone calls her.’ (48a) É melhor sair antes de começar a chover. is better leave before of start-INF to rain ‘It is better to leave before the rain starts.’ (48b) É melhor sair antes que comece a chover. is better leave before that start-SUBJ to rain ‘It is better to leave before it starts raining.’

Neither of these sentences allows for the inference that the proposition introduced by after or until is true at utterance time. However, (47a) and (48a), where the infinitive occurs, allow for the inference that the proposition is expected to become true, while (47b) and (48b), where the subjunctive is selected, do not allow for such an inference. In other words, (47a) may be felicitously asserted in a context where it is known or expected that someone will call Ana, but (47b) would seem odd in such a context. Similarly, the assertion of (48a) is natural in a context where it is assumed that it will start raining, a context where the assertion of (48b) would be odd. Conversely, in a context where the state of affairs described by the clause introduced by the temporal connective is not taken for granted, it is natural to assert (47b) or (48b), where the subjunctive occurs, but not the corresponding sentences employing the infinitive. This suggests a connection between the choice of the infinitive or the subjunctive and the truth value of the proposition: if the proposition is expected to become true, the infinitive is selected, otherwise, the subjunctive is preferred. This hypothesis is favoured by the following examples, where the temporal connectives have a retrospective use: (49a) Ele morreu antes de eu ter nascido. he died before of I be-INF born (49b) *Ele morreu antes que eu tivesse nascido. he died before that I be-SUBJ born ‘He died before I was born.’

Modality, context change potential and mood selection in European Portuguese

157

(50a) A Ana ficou calada até a terem chamado. the Ana stayed silent until her have-3.PL.INF called (50b) *A Ana ficou calada até que a tivessem chamado. the Ana stayed silent until her have-3.PL.SUBJ called ‘Ana remained silent until she was called.’

In these cases, the proposition introduced by antes or até is taken to be true and the subjunctive is ruled out, contrary to the infinitive. However, the following sentences lead to the rejection of the hypothesis that the selection of the infinitive or the subjunctive by the temporal connectives under consideration is linked to the truth value of the proposition: (51a) A Ana saiu antes de alguém ter tido tempo para a avisar. the Ana left before of someone have-INF had time to her advise (51b) A Ana saiu antes que alguém tivesse tido tempo para a avisar. the Ana left before that someone have-SUBJ had time to her advise ‘Ana left before anyone had time to advise her.’ (52a) O anterior campeão desistiu da corrida antes de ter chegado ao fim. the former champion gave-up of-the race before of be-INF arrived at-the end (52b) ??/*O anterior campeão desistiu da corrida antes que tivesse chegado ao fim. the former champion gave-up of-the race before that be-SUBJ arrived at-the end ‘The former champion gave up before finishing the race.’

All these sentences allow for the inference that the proposition introduced by the temporal connective is false. Still, the infinitive or the subjunctive may be selected in (51), while in (52) only the infinitive is accepted. Thus, the data show that the subjunctive is ruled out if the proposition is taken to be true (cf. (49b) and (50b)), but is not necessarily accepted if the proposition is taken to be false (cf. (52b)). The infinitive may occur regardless of the truth value of the proposition. Therefore, the selection of the subjunctive or the infinitive in the constructions under investigation is not conditioned by the truth value of the proposition. A more promising hypothesis is that the selection of the infinitive or the subjunctive for the clauses introduced by até and antes is dependent on the information available in the context of the assertion. Specifically, though the infinitive and the subjunctive clauses may have the same truth conditions, the infinitive seems to be selected when the complement clause of antes or até describes a state of affairs that is known to exist or expected to occur; otherwise, the subjunctive is preferable. For instance, in (49a), the infinitival clause describes a state of affairs that is known to exist prior to the utterance. In (52a), the infinitival clause describes a state of affairs that is expected, given the previous clause (if an athlete takes part in a race, it is natural to expect that

158

Rui Marques

he will finish it). In these cases, the infinitive occurs naturally, but the subjunctive is ruled out. Subjunctive clauses are better suited for clauses describing a state of affairs that is not expected prior to the assertion. To exemplify, consider (51a), where the infinitive is used, and (51b), where the subjunctive occurs. The first sentence may felicitously be asserted in a context where it is assumed that someone should have advised Ana, while (51b) may be felicitously uttered only in a context where this assumption is not at hand. For instance, (51a) would be a possible answer to the question why didn’t you advise Ana?, while (51b) would sound odd as an answer to this question. Similarly, (48a) may be uttered in a context where there is evidence that it will start raining, while (48b) may be uttered in a context where such evidence is not available. To sum up, the observed data suggest that subjunctive clauses may introduce the description of a new state of affairs into discourse, while infinitive clauses refer only to states of affairs available at the utterance time (because, for example, they have been previously described or because the proposition that describes them follows from what was said). Using the basic notions of the Context Change Potential Framework, this hypothesis can be roughly described as follows. According to the considered framework, the meaning of a sentence is its context change potential (CCP). […] A CCP is a function from contexts to contexts. Contexts are here identified with states of information, which in turn are construed as sets of possible worlds, and the change effected by the CCP of a sentence consists of updating that information by what the sentence says. (Heim 1992: 185)

If a declarative sentence p is successfully asserted against a context c − a set of possible worlds −, the result will be a new context, which corresponds to c minus those worlds where p is not verified. Thus, the assertion of p against a context c gives the instruction to remove possible worlds − those where p is not verified − from c. In the case of clauses introduced by the connectives até and antes, neither the infinitive nor the subjunctive clause gives the instruction to remove possible worlds from the context. Thus, their context change potential is of a different nature. The hypothesis under scrutiny is that if the sentence is in the infinitive, it provides no change in the context, while if it is in the subjunctive mood, it brings into consideration possible worlds that were less accessible prior to the utterance. If this is so, the considered data suggest that in the analysis of mood selection there is another factor to be taken into account, one that has been used in the description of modality within possible world semantics: ordering sources (cf. e. g. Kratzer 1991, Portner 2009). Kratzer proposes that the meaning of modal verbs, and other modal operators, may be described

Modality, context change potential and mood selection in European Portuguese

159

according to three parameters: (i) modal force; (ii) modal base, and (iii) ordering source. The basis idea of ordering sources is that the relevant possible worlds form a set ordered on the basis of the proximity to the ideal. The ideal may be, for instance, what the law provides, if a deontic ordered source is considered, what is normal or expected to assume, if an epistemic or stereotypical ordering source is considered, what is desired, if a bouletic ordering source is considered, and so on. It seems natural to assume that in the course of each conversation, the set of possible worlds that form the context set − i. e., the possibilities that are alive − is ordered according to what is more natural to assume. Given this, consider a context of a conversation C where a certain topic (say, the present weather in Paris) is being discussed. Imagine that in this context of conversation what is assumed to be the case is compatible with possible worlds where there is life on Mars, as it is compatible with possible worlds where it is raining in Paris at the time. In this case, the context set includes, among others, possible worlds where there is life on Mars and possible worlds where it is raining in Paris at the time. In other words, the existence of life on Mars is a possibility that is alive at the considered context, and the possibility that it is raining in Paris at the time is also a living possibility. However, given the topic and the accepted purpose of the conversation, it is not expected that reference is made to possible worlds where there is life on Mars. On the contrary, it is natural to refer to possible worlds where it is raining in Paris at the time. That is, in the given context of conversation C, possible worlds where there is life on Mars are part of the context set, but reference to such worlds is farther away from the expected continuation of the conversation than reference to possible worlds where it is raining in Paris at the time. Given this, the advocated hypothesis concerning the selection of the infinitive or the subjunctive for clauses introduced by the temporal connectives até and antes is that the subjunctive clause makes reference to a possibility that is farther away from the expected continuation of the conversation. In this sense, the subjunctive clause provides a change in ordering of the context set, by bringing into consideration possible worlds that were distant from what was expected from previous discourse. The infinitival clause, on the contrary, provides no change in the ordering of the context set. In the constructions under consideration, the infinitival clause refers to possibilities that were mentioned by previous discourse or that are expected to follow from what was said.

160

Rui Marques

6. Conclusion On the basis of the data that have been observed, three factors have been identified as responsible for the selection of mood: veridicality − the indicative occurs only in veridical contexts −, kind of modality (modality being understood as the kind of attitude towards the proposition) − the indicative occurs only in epistemic environments −, and the context change potential of the clause − an indicative clause has the potential to remove possible worlds from the context set, a subjunctive clause may change the ordering of the worlds forming the context set, and an infinitive clause provides no change in the context set (in the considered adverbial clauses). In a simplified manner, the proposed hypothesis puts forward the idea that the selection of mood is dependent on the communicative intention: if no change in the context set is intended, the infinitive occurs; if the purpose is to express belief in the relevant proposition, the indicative is selected; if some other kind of attitude is expressed, the subjunctive is chosen. Naturally, this generalisation, principally concerning the infinitive, is only tenable if other contexts where this mood can occur are taken into consideration, particularly complement clauses. Nevertheless, the kinds of adverbial clauses that were considered suggest that the context change potential plays a role in the choice between the infinitive and a finite mood. The observation of the semantic or pragmatic contribution of the infinitive in other kinds of clauses might give rise to a reformulation of the hypothesis outlined in this paper, or it might lead to the conclusion that the context change potential plays a role in the selection of the infinitive together with other factors.

References Ahern, Aoife (2005): “Mood choice and sentence interpretation in Spanish.” − In: Bart Hollebrandse, Angeliek van Hout & Co Vet (eds.): Crosslinguistic Views on Tense, Aspect and Modality, 201–214. Amsterdam: Rodopi (Cahiers Chronos 13). Bell, Anthony (1990): “El modo en español. Consideración de algunas propuestas recientes.” − In: Ignacio Bosque (ed.): Indicativo y subjuntivo, 81–105. Madrid: Taurus. Bybee, Joan & Terrell, Tracy D. (1990): “Análisis semántico del modo en español.” − In: Ignacio Bosque (ed.): Indicativo y subjuntivo, 145–163. Madrid: Taurus.

Modality, context change potential and mood selection in European Portuguese

161

Farkas, Donka (1992): “On the semantics of subjunctive complements.” − In: Paul Hirschbühler & Konrad Koerner (eds.): Romance Languages and Modern Linguistic Theory, 71–104. Amsterdam: Benjamins. Giannakidou, Anastasia (1999): “Affective dependencies.” − In: Linguistics and Philosophy 22 (4), 367–421. Guitart, Jorge M. (1984): “On the use of the Spanish subjunctive among Spanish English bilinguals.” − In: Word 33, 59–67. Heim, Irene (1992): “Presupposition Projection and the Semantics of Attitude Verbs.” − In: Journal of Semantics 9 (3), 183–221. Hengeveld, Kees (1988): “Illocution, Mood and Modality in a Functional Grammar of Spanish.” − In: Journal of Semantics 6 (3–4), 227–269. − (2004): “Illocution, Mood and Modality.” − In: Geert BooƋ, Christian Lehmann & Joachim Mugdan (eds.): Morphology. An International Handbook on Inflection and Word-Formation. Vol. 2, 1190–1202. Berlin: Mouton de Gruyter. Hooper, Joan (1975): “On assertive predicates.” − In: John P. Kimball (ed.): Syntax and Semantics 4, 91–124. New York: Academic Press. Karttunen, Lauri (1971): The Logic of English Predicate Complement Constructions. − Bloomington, Ind.: Indiana University Linguistics Club. Klein-Andreu, Flora (1990): “Restricciones pragmáticas sobre la distribución del subjuntivo en español.” − In: Ignacio Bosque (ed.): Indicativo y subjuntivo, 303−313. Madrid: Taurus. Kratzer, Angelika (1991): “Modality.” − In: Arnim von Stechow & Dieter Wunderlich (eds.): Semantics, 639–650. Berlin: de Gruyter. Marques, Rui (2009): “On the selection of mood in complement clauses.” − In: Lotte Hogeweg, Helen de Hoop & Andrej Malchukov (eds.): Cross-linguistic Semantics of Tense, Aspect and Modality, 179–204. Amsterdam/Philadelphia: John Benjamins. Noonan, Michael (1985): “Complementation.” − In: Timothy Shopen (ed.): Language Typology and Syntactic Description. Vol. II, 42–140. Cambridge: Cambridge University Press. Palmer, Frank R. (1986): Mood and Modality. − Cambridge: Cambridge University Press. Portner, Paul (2009): Modality. − Oxford: Oxford University Press. Quer, Josep (1998): Mood at the Interface. − The Hague: HAG. Rescher, Nicholas (1968): Topics in Philosophical Logic. − Dordrecht: Reidel. Sanchez Valencia, Victor, van der Wouden, Ton & Zwarts, Frans (1993): “Polarity, veridicality, and temporal connectives.” − In: Paul Dekker & Martin Stokhof (eds.): Proceedings of the Ninth Amsterdam Colloquium, 587–606. Amsterdam: ILLC. Villalta, Elisabeth (2008): “Mood and gradability: an investigation of the subjunctive mood in Spanish.” − In: Linguistics and Philosophy 31, 467–522. Wandruszka, Ulrich (1991): “Frasi subordinate al congiuntivo.” − In: Lorenzo Renzi & Giampaolo Salvi: Grande grammatica italiana di consultazione, 415–481. Bologna: Il Mulino.

Josep Quer

On the (un)stability of mood distribution in Romance

1. Introduction Although the study of the distribution of indicative and subjunctive has traditionally occupied a significant place in Romance descriptive and historical linguistics, it has received comparatively little attention in formal approaches, unlike other morphosyntactic verbal categories such as tense or aspect. The main stumbling block for mainstream discussion of the syntactic and semantic properties of mood distribution is probably the difficulty of coping with surface variability in mood choice across certain contexts, intralinguistically and crosslinguistically. In this article I intend to address the problems underlying this apparent partial unstability by suggesting a line of analysis that might lead to a better understanding of the factors that determine mood choice. In order to reach this goal, I will draw an empirical map of mood distribution mainly on the basis of Catalan and Iberian Spanish data, relying on the results of previous studies that I have undertaken (Quer 1998, 2001, 2006). Some diachronic and dialectal data from these languages will be introduced, in order to determine the points of variation attested across those two dimensions. Next, some crosslinguistic variation within Romance will be incorporated into the discussion, in order to try to disentangle the patterns behind some apparent cases of diverging behaviour in mood choice. In the last section I will present the essentials of an analysis that can account for this type of variation, namely the interpretation of mood shift as a shift in the type of model where the proposition is interpreted. The main tenet in this proposal is that alleged variation in mood selection is not real optionality and that the attested differences are in fact smaller, proving to be amenable to the way a specific language or variety grammaticalizes mood distribution at a given stage. Given the limitations of this paper, I will restrict the discussion to mood distribution in argument clauses, although the basic claims can be extended to relative and adverbial clauses (see Quer 1998). This implies that subjunctive use in main clauses will be put aside for ease of exposition, but I assume that those instances are also amenable to the rationale of the analysis for embedded contexts. Beyond this restriction, I am aware of the fact that some less central instances and some recalcitrant cases will remain unaddressed. As indicated, the proposal presented here represents just a sketch of an explanation for the variation attested and a great deal of further work is required in order to work

164

Josep Quer

out the whole range of phenomena revolving around the indicative/subjunctive divide.

2. Basic generalizations In this section I will summarize the main generalizations any comprehensive account of mood distribution should be able to cover in the domain of argument clauses. For the sake of the argument, and following a wellestablished tradition in the approaches to mood, I will take it for granted that it is precisely here where some basic properties of the indicative/subjunctive divide manifest themselves. The core cases of subjunctive selection have been often taken to be those under a volitional predicate like Catalan voler ‘to want’ (see ex. (1)).1 In these complements, the subjunctive clauses display a number of distinctive characteristics, illustrated in (2): the subjunctive is only licensed in the immediately embedded clause (2a), it cannot alternate with the indicative (2b), it displays sequence of tense restrictions (2c)2 and the embedded subject cannot be coreferent with the matrix subject, a phenomenon known as subject disjoint reference effect (2d). (1)

Vull que {acabi/*acaba} la tesi. (Cat.) want-PRS.1.SG that finish-SUB/IND.PRS.3.SG the dissertation ‘I want her/him to finish-SUB/*IND the dissertation.’

(2a) Quieres que creamos que tienes/*tengas razón. (Spa.) ‘You want us to believe-SUB that you are-IND/*SUB right.’ (2b) Quieres que creamos/*creemos que tienes razón. ‘You want us to believe-SUB/*IND that you are-IND right.’ (2c) Quieres que creamos/*creyéramos que tienes razón. ‘You want us to believe-SUB.PRS/*PST that you are-IND right.’ (2d) Quieres que creamos/*creas que tienes razón. ‘You want us/*you to believe-SUB.1.PL/*2.SG that you are-IND right.’ 1

2

It should be pointed out that volitional verbs do not seem to constitute a homogeneous class with respect to the properties mentioned in the text (see for instance esperar ‘to hope’). In the absence of a more detailed account of such differences, I will remain with voler ‘to want’ as the representative item for the class. In this case, no past tense can appear under a main clause present tense, but the reverse combination can be found in appropriate contexts (see Quer 1998: 34).

On the (un)stability of mood distribution in Romance

165

Alongside volitionals, directive predicates taking finite complements display subjunctive clauses with the same properties outlined above: (3)

T’ordena que hi {vagis/*vas}. (Cat.) you-order-3.SG that there go-SUB/IND.2.SG ‘S/he orders you to go-SUB/*IND.’

A case that has received almost little or no attention in the literature is subjunctive selection by causative and implicative predicates such as make, contribute, manage, etc. The particularity of this group of verbs is that their subjunctive arguments display the same set of properties as those of volitionals and directives, but interpretively they imply the truth of their complement and do not convey intensional meaning at face value. (4)

Fas que {marxi/*marxa} abans d’hora. (Cat.) make-2.SG that leave-SUB/IND.PRS.3.SG before of-time ‘You make her/him leave-SUB/*IND earlier.’

Epistemic predicates take the indicative by default in Catalan or Spanish. They only license the subjunctive when combined with a negative operator or a yes/no interrogative operator, as illustrated in (5): (5a) No recorda que en Miquel treballi. (Cat.) not remember-PRS.3.SG that the M. work-SUB.PRS.3.SG ‘S/he does not remember that Miquel works.’ (5b) Recordes que en Miquel treballi? remember-PRS.2.SG that the M. work-SUB.PRS.3.SG ‘Do you remember if Miquel works?’ (5c) *Recordo que en Miquel treballi. remember-PRS.1.SG that the Miquel work-SUB.PRS.3.SG ‘I remember that Miquel works-SUB.’

The subjunctive that appears in such structures differs from the one found under volitionals and directives: it can be licensed in consecutively embedded domains (6a), it can alternate with the indicative – with concomitant interpretation alternations – (6b), it shows no restriction on the sequence of tenses (6c) and no subject disjoint reference effects arise (6d): (6a) No piensa que creas que tienes/tengas razón. (Spa.) ‘S/he does not think you believe-SUB that you are-IND/SUB right.’ (6b) No piensa que creamos/creemos que tienes razón. ‘S/he does not think you believe-SUB/IND that you are-IND right.’

166

Josep Quer

(6c) No piensa que creamos/creyéramos que tienes razón. ‘S/he does not think you believe/d-SUB.PRS/PST that you are-IND right.’ (6d) No piensa que crea realmente que tienes razón. ‘S/hei does not think s/hei believes-SUB that you are-IND right.’

Quer (1998), following Stowell (1993), distinguishes these two types of subjunctives with the labels ‘intensional subjunctive’ for the former and ‘polarity subjunctive’ for the latter. The motivation for the split lies in the fact that the intensional subjunctive is unambiguously and consistently selected by strong intensional predicates3 (in the sense of Farkas 1992), while the polarity subjunctive is in fact licensed by the cooccurrence of an operator with a weak intensional predicate like an epistemic. An interesting case of so-called mood alternation is that represented by verbs of speech and communication in general: under a reading conveying the report of an assertion, they unproblematically take the indicative (7a), but when they combine with subjunctive they report the issuing of an order or a wish (7b). Interestingly, under this second reading, the subjunctive complement displays the properties of the intensional one. This pattern is systematic across verbs of speech and communication in the languages that display indicative/subjunctive contrasts. Unless one appeals to systematic homonymy within this verb class, something along the lines of a covert predicate of influence must be assumed in order to account for the facts. (7a) Diu que ve. (Cat.) say-3.SG that come-IND.PRS.3.SG ‘S/he says that s/he is-IND coming.’ (7b) Diu que surtis. say-3.SG that leave-SUB.PRS.2.SG ‘S/he tells you to leave-SUB.’

An additional interesting piece of evidence in this respect is that interrogative verbs introducing indirect questions never take the subjunctive in Catalan and Spanish, even if the main predicate is negated. For approaches attributing general meanings to the moods, this is quite problematic, as the embedded proposition is not presupposed by the speaker; on the contrary, under an interpretation of subjunctive as the mood of uncertainty, one would expect it to appear here. (8a) No ha preguntat si {volíem/*volguéssim} quedar-nos-hi. (Cat.) not have-3.SG asked if want-IND/SUB.IMP.1.PL to-stay-SE-there ‘They haven’t asked whether we wanted-IND/*SUB to stay.’ 3

A qualification is obviously needed for causative and implicative predicates, which are not intensional on the surface. For a discussion, see Quer (1998: 46–50).

On the (un)stability of mood distribution in Romance

167

(8b) Investiguen què {ha/*hagi} provocat l’explosió. investigate-3.PL what have-IND/SUB.PRS.3.SG caused the-explosion ‘They are investigating what has-IND/*SUB caused the explosion.’

A class of verbs that has received significant attention in the literature is that of factive-emotive or psychological predicates. The usual generalization states that in languages like Catalan or Spanish these predicates select for the subjunctive, while in Romanian they select for the indicative. The choice of the subjunctive is explained as the consequence of an additional interpretation of the subjunctive as conveying old information, here linked to the factive lexical meaning of the matrix verb. However, this has shown to be too rough a distinction based on a few examples that are considered to be prototypical. As argued in Quer (1998, 2001), the episodicity of the main predicate has an effect on the possible mood choices: when the main predication is episodic, both indicative and subjunctive are possible in the subordinate clause (9a, b), whereas a matrix generic predication excludes indicative selection (9c).4 (9a) Em va encantar que tots em fessin una pila de preguntes. (Cat.) me AUX.3.SG delight that all me make-SUB.IMP.3.PL a heap of questions ‘I was delighted that everyone asked a lot of questions.’ (9b) Em va encantar que tots em van fer una pila de preguntes. me AUX.3.SG delight that all me make-IND.IMP.3.PL a heap of questions ‘I was delighted that everyone asked a lot of questions.’ (9c) M’encanta que em {facin/*fan} una pila de preguntes. me-delight-3.SG that me make-SUB/IND.3.PL a heap of questions ‘I’m delighted if they ask-SUB/*IND me a lot of questions.’

Even in the contexts where there is apparent double mood selection, the choice is not entirely free, as the readings yielded are not the same: with embedded subjunctive the embedded proposition is presupposed, i. e. it is presented as old information present in the context of conversation or information that has to be accommodated as background in that context; with the indicative, the embedded proposition is asserted. This contrast becomes very clear with factive-emotive verbs that easily allow for an assertive-like interpretation, like queixar-se ‘to complain’: 4

A reviewer states that this generalization might not be valid for Portuguese, where factive-emotives would allegedly select exclusively for the subjunctive. Without having carried out any serious research on this, a quick net search confirmed the proposed generalization for main predications like gostei muito que ‘I liked it a lot’ or adorei que ‘I loved it that’ in Portuguese. Of course, this would also require identifying possible distinct patterns between European and Brazilian Portuguese. Corpus and elicitation studies on this are obviously needed.

168

Josep Quer

(10a) Es queixa que no li facin cas. SE complain-3.SG that not him/her make-SUB.PRS.3.PL attention ‘S/he complains that they don’t pay-SUB attention to him/her.’ (10b) Es queixa que no li fan cas. SE complain-3.SG that not him/her make-IND.PRS.3.PL attention ‘S/he complains that they don’t pay-IND attention to him/her.’

The opposite pattern also holds true: factive-emotives with a pure evaluative interpretation and no possible assertive reading disfavour the occurrence of indicative and unproblematically allow for a subjunctive argument clause: (11a) És una vergonya que no li facin cas. be-3.SG a shame that not him/her make-SUB.PRS.3.PL attention ‘It’s a shame that they don’t pay-SUB attention to him/her.’ (11b) ??És una vergonya que no li fan cas. be-3.SG a shame that not him/her make-IND.PRS.3.PL attention ‘It’s a shame that they don’t pay-IND attention to him/her.’

The generalizations presented here are meant to constitute a representative sample of the distribution of indicative and subjunctive in embedded argument clauses across predicate classes. Even from such a cursory presentation, it becomes evident that verb meaning is not the only factor determining the choice of mood. What we see instead is that its interaction with other elements like operators, aspect or presupposition can be decisive in determining the mood of the embedded clause. In this sense there is only an apparent surface alternation in mood choice, as the choice of one mood is never equivalent to the other and interpretive distinctions can be established for each case.

3. Intralinguistic variation: some dialectal and diachronic evidence Despite the generality of the map of mood distribution in Catalan and Spanish sketched in the previous section, it must be noted that dialectal and diachronic data are reported which seem to counter it in some respects. Obviously, some aspects of mood distribution have varied over the centuries and varieties of the “same” language have developed slightly varying grammatical systems. Still, the points of variation are not random and I will claim that they are in fact systematic. In the next section we will see that similar variation is attested across Romance varieties, thus confirming the general point of this paper. In Modern Catalan and Spanish, plain epistemics (i. e. those not affected by a negative or an interrogative yes/no operator) take indicative complements

On the (un)stability of mood distribution in Romance

169

and the subjunctive is not an option. The situation was different, however, in earlier stages of those languages: as illustrated in (12) and (13), the predicates pensar ‘to think’ in Catalan and creer ‘to believe’ in Spanish could take subjunctive complements even in the absence of a concomitant licenser. (12) Tu penses que la ànima dels bruts sia espiritual. (Old Cat.) you think-2.PL that the soul of-the uncultivated be-SUB.PRS.3.SG spiritual ‘You think that the soul of the uncultivated people is-SUB spiritual.’ (Bernat Metge, 14th century, apud Par 1923) (13) Nazaret creo que sea. (Old Spa.) Nazareth believe-1.SG that be-SUB.PRS.3.SG ‘I believe it is-SUB Nazareth.’ (Arcipreste de Hita, 14th century, apud Jensen & Lathrop 1973)

Moreover, even in contemporary varieties of Spanish we find licensing of subjunctive under certain predicates which in Iberian Spanish only license the indicative. This is the case with creer ‘to believe’ in Venezuelan Spanish, as reported in the literature (14) and further documented through an internet search (15). (14) Creo que Luís ame a María. (Venezuelan Spa.) believe-1.SG that Luís love-SUB.PRS.3.SG A María ‘I believe that Luís loves-SUB María.’

(D’Introno 1990: 197)

(15) Yo creo que haya vida en otros mundos. I believe-1.SG that there-be-SUB life in other worlds ‘I believe there is-SUB life in other worlds.’5

Nevertheless, synchronic variation cannot be reduced exclusively to dialectal variation. A coarse approach to selection on grounds of verb class might blur more subtle distinctions that are relevant for mood choice and might lead us to rashly conclude that there is real mood optionality. The verb sospechar ‘to suspect’ is a case in point: although the default mood choice seems to be the indicative, it can appear with subjunctive as well. The contrast is subtle, but clear when it gets identified: with the indicative an assertion is made by qualifying it as a suspicion, probably based on inference; with the subjunctive the conjecture component is more salient and no straightforward assertion is

5

URL: http://forums.myspace.com/t/3896436.aspx?fuseaction=forums.viewthread [08/12/2008].

170

Josep Quer

made. This is parallel to the Italian case of ipotizzare ‘to hypothesize’, studied by Giorgi & Pianesi (1997, 2004).6 (16) Sospecho que haya/ha pasado por aquí. (Modern Spa.) suspect-1.SG that have-SUB.PRS.3.SG passed through here ‘I suspect that s/he has-SUB/IND passed through here.’

(Gili Gaya 1969: 135)

Another place where variation appears to arise is with a negative epistemic predicate like negar ‘to deny’ in Spanish. In principle, Iberian Spanish requires the subjunctive with this verb, because the embedded proposition is interpreted in the scope of a lexicalized complex meaning “not to say”. In Mexican Spanish, though, we find examples with the same alleged interpretation, but displaying the indicative. (17) Niegan que se efectuó una violación de los derechos individuales. (Mexican Spa.) deny-3.PL that SE make-PST.IND.3.SG a violation of the rights individual ‘They deny that a violation of individual rights took-IND place.’ (Lope Blanch 1990: 181)

We could connect these facts with other observations made in the literature, like the one by Dunlap (2006), who establishes that in journalistic prose, the temporal connectives después de que and luego de que ‘after’ systematically take subjunctive in Iberian Spanish and indicative in Mexican Spanish. A thorough study of mood distribution in Mexican Spanish remains to be undertaken, as it might well be the case that the system has clearly evolved visà-vis Iberian Spanish. However, it should be noted that example (17) is not excluded in Iberian Spanish either. It simply requires a context where the embedded proposition can be endorsed by the speaker or by another relevant illocutionary agent, and in this sense it is a marked reading, because the sentence is interpreted outside the scope of the negation encoded lexically in the main verb. Contrary to what we established for indirect questions in the modern varieties in section 2, previous stages of both Catalan and Spanish could display the subjunctive:

6

Interestingly, when suponer ‘to suppose’ takes the subjunctive, complementizer deletion is possible, much as in the case of Italian ipotizzare. I leave this issue open for future research.

On the (un)stability of mood distribution in Romance

171

(18) lo qual sercava hon pogués fer penitència e contemplar Déu. (Old Cat.) the which search-IMP.3.SG where can-SUB.IMP.3.SG do penance and contemplate God ‘who was trying to find out where he could-SUB do penance and to contemplate God.’ (Ramon Llull, 13th century) (19) No sé qué pueda haberle ocurrido. not know-1.SG what can-SUB.PRS.3.SG have-him/her happened ‘I don’t know what might-SUB have happened to him/her.’

(Cervantes)

Even in contemporary varieties, certain indirect questions seem to license the occurrence of the subjunctive. A well-known instance is that appearing under no sé ‘I don’t know’. Notice that the decisive factor here might be the appearance of negation in the main clause, together with the first-person attitude holder, which seems to be crucial in licensing the subjunctive: despite the surface similarity with diachronic data, the phenomenon probably has a much more restricted scope in contemporary language. (20) No sé qué te diga. not know-1.SG what you say-SUB.PRS.1.SG ‘I don’t know what to tell-SUB you.’ (Peruvian, Colombian, Asturian, Galician Spanish, apud Suñer 1999: 2185)

As already pointed out in section 2, factive-emotive/psych predicates have also featured prominently in the discussions on mood distribution, particularly because their default mood selection of subjunctive seems to go against the semantic characterization of this mood under volitionals and directives. We observed, though, that the facts are less clear-cut than usually assumed, as episodicity and genericity can play a role, along with the availability of an assertive interpretation of the main predicate (see for example (10)). It has been argued that other varieties of Spanish such as Mexican, for instance, do not select the subjunctive by default for factive-emotives. It might well be the case that the indicative is the default mood with factive-emotives in Mexican Spanish, but until we obtain a better picture of the distribution of indicative and subjunctive as a whole, we cannot determine whether this is just another instance of non-optional mood alternation. (21) Mucho me alegra que no ha caído en el vacío mi escrito. (Iberian Spa.) a-lot me gladden-3.SG that not have-PRS.IND.3.SG in the void my text (Lope Blanch 1990: 181) ‘I’m very glad that my text did-IND not fall flat.’

172

Josep Quer

(22) Estoy muy satisfecha de que supo terminarlo él solo. (Mexican Spa.) be-1.SG very satisfied of that know-PST.IND.3.SG to-finish-it he alone ‘I’m very satisfied he managed-IND to finish it by himself.’ (Lope Blanch 1990: 181)

It should also be noted that the option of indicative under a factive-emotive already existed in Old Spanish, too. The question that remains to be answered is whether the same interpretive contrasts derived from mood choice can be detected as in Modern Spanish. (23) Pesa me que non somos çerteros del logar. (Old Spa.) grieve-3.SG me that not be-IND.PRS.1.PL sure of-the place ‘I regret that we’re-IND not sure about the place.’ (Gonzalo de Berceo, early 13th century)

Beyond the question of variation in mood selection under specific embedding verb classes, other aspects of variation should ideally be addressed. In section 1 it was pointed out that intensional subjunctive (essentially that which appears under volitionals and directives) has as one of its distinctive properties a stricter sequence of tense. So, for instance, a present subjunctive is in principle marked under a matrix past tense. Putting Spanish in dialectal perspective, though, this property does not seem to hold across varieties, and a subjunctive present is acceptable where standard Iberian Spanish would use a past subjunctive by default. The following example from Bolivian Spanish illustrates this: (24) Era preciso que corra tiempo. be-IMP.3.SG necessary that run-PRS.SUB.3.SG time ‘It was necessary that time pass-SUB.’

(Kany 1970: 211)

It is important to observe that this difference with respect to Iberian Spanish should not be immediately attributed to the decline in the productive use of past subjunctive forms, as has been argued to be the case in French. The main aim of this section was to show that apparent variation in mood distribution across dialects and throughout the history of the language should be evaluated in the light of a more fine-grained account of mood selection and licensing. Still, some of the differences, such as the choice of subjunctive under an unmodified epistemic, seem to be genuine cases of variation within a language, and it is arguably due to the slightly different ways different varieties grammaticalize the relevant interpretive notions. In the next section we will briefly see that some of the points of variation emerge in exactly the same places when we look at some other Romance varieties.

On the (un)stability of mood distribution in Romance

173

4. A sample of crosslinguistic variation across Romance varieties Old Catalan and Old Spanish are not the only Romance varieties that license subjunctive under plain epistemics, that is under epistemic predicates that are not affected by a negative or a yes/no operator: standard Italian, unlike French or Catalan, allow for this possibility, as illustrated in (25). (25) Credo che lei sia stanca. (Ita.) believe that she be-SUB.PRS.3.SG tired ‘I think she is-SUB tired.’ (26) *Crec que ella estigui cansada. (Cat.) believe that she be-SUB.PRS.3.SG tired ‘I think she is-SUB tired.’ (27) *Marc croit que le printemps soit arrivé. (Fra.) Marc believe-3.SG that the spring be-SUB.PRS.3.SG arrived ‘Marc believes that spring has-IND arrived.’

Another domain where Italian seems to differ from Modern Catalan and Spanish is in embedded interrogatives, which apparently allow for both moods, but different embedding verbs behave differently in this respect: see sapere ‘to know’ vs. dire ‘to say’ (data from Wandruszka 1991: § 3.5). Wandruszka notes, however, that the choice of indicative in (28) implies that the speaker knows the answer to the embedded question, while this is not the case with the subjunctive. (28) Luigi sapeva chi {fosse/era} il nuovo venuto. (Ita.) Luigi know-IMP.3.SG who be-SUB/IND.IMP.3.SG the new arrived ‘Luigi knew who the newcomer was-SUB/IND.’ (29) Dimmi come si {chiama/*chiami} il tuo direttore tell-IMPER.SG-me how SI call-IND/SUB.PRS.3.SG the your director ‘Tell me what’s the name-IND/*SUB of your director.’

In this connection, Giorgi (2009) notices a very relevant difference in the mood choices of domandare ‘to ask’ as an embedder of interrogatives: as a verb of saying reporting a communicative act, it takes the indicative, but its reflexive variant, domandarsi ‘to wonder’ takes the subjunctive because it becomes an attitude predicate. (30) Gianni mi ha domandato se Maria è incinta. ‘Gianni asked me if Maria is-IND pregnant.’ (31) Gianni si domandava se Maria fosse incinta. ‘Gianni asked himself if Maria was-SUB pregnant.’

174

Josep Quer

One of the recurring generalizations concerning mood choice in Romance states that Romanian, unlike Spanish for instance, uses the indicative with factive-emotive/psych predicates, as in (32): (32) Maria regretă că Paul a plecat. Maria regret-3.SG that Paul has-IND.3.SG left ‘Maria regrets that Paul has-IND left.’

Again, though, this picture may be too simplistic, as we also find factiveemotive predications that surface in the subjunctive (see (33)). Interestingly, when in a generic predication, subjunctive arises. This situation is immediately reminiscent of that observed for Catalan factive-emotives in section 2. (33) Ar fi păcat să pierdem úansă asta. (Rom.) COND.3.SG be pity SUB lose-1.SG chance this ‘It would be a pity to lose-SUB this chance.’

(Farkas 1992: 102)

Let us take these examples of variation in mood choice within Romance as a rather strong indication that the alleged differences in mood distribution across varieties turn out to be too coarse and that the domains where languages seem to differ are actually the same domains where mood choices might have been different in previous stages of their history and in dialectal varieties.

5. Pinning down variation After examining these crucial sets of data, it should have become clear that there is enough evidence to question the claim that real variability in mood choice exists: indicative/subjunctive contrasts should be taken to always yield an interpretive effect, even if it is sometimes hard to identify this effect precisely. An indicative in the scope of a negated epistemic expresses the idea that the truth of the embedded proposition is endorsed by the speaker or by some other illocutionary agent, while the subjunctive is interpreted in the scope of the negated attitude. Verbs of speech and communication in general have been shown to have a double selection pattern, as the choice of indicative or subjunctive in their complements leads to distinct interpretations of the embedding verb. With factive-emotive predicates we have observed that the alleged mood selection is less uniform than claimed and that both indicative and subjunctive can occur in their argument clauses depending on factors like the episodicity or genericity of the main predication, or the possibility of interpreting the embedding predicate as reporting on an assertion, i. e. as some sort of verb of

On the (un)stability of mood distribution in Romance

175

communication. The latter aspect reappears in several other cases of apparent variable mood selection, like with the verb admettre ‘to admit’ in French: if the choice of subjunctive correlates with the concessive reading of the sentence, whereby the embedded proposition is treated as information that can be incorporated to the conversational background, the indicative simply introduces an assertion which happens to be a concession to some relevant illocutionary agent (normally the addressee). (34a) J’admets que vous avez raison. (Fra.) I-admit-1.SG that you have-IND.PRS.2.SG ‘I admit that you are-IND right.’ (34b) J’admets que vous ayez raison. I-admit-1.SG that you have-SUB.PRS.2.SG ‘I admit that you are-SUB right.’

(Grevisse 1993: 1618)

An interesting piece of evidence that relates to this discussion comes from the verb dire ‘to say’ in Italian. In principle this verb displays the two options of mood selection, with the corresponding differences in the interpretation (report of an assertion vs. report of a command or a wish, like a directive). Giorgi & Pianesi (2004: 205) noted that there is an additional possibility, namely the choice of subjunctive without a directive reading, but rather with the reading of an epistemic predicate like credere ‘to believe’. This reading arises as a consequence of an impersonal reading of the matrix subject and non-episodic tense. (35) Dicono (che) sia una stupida. say-PRS.3.PL that be-SUB.3.SG a stupid-FEM ‘They say that she is-SUB stupid.’

By now it should have become evident that embedding verbs can no longer be considered as belonging to a closed semantic class that univocally selects for one mood or the other: lexical meaning must be flexible enough to be able to compositionally trigger the different interpretations identified for the structures at hand. The factors leading to this final outcome are heterogeneous, but some of them can be clearly identified, as we have suggested so far. However, sometimes it seems as if it is the choice of mood itself which determines the final interpretation of the sentence, or even of the embedding predicate. The contrast in (36) illustrates this point: with the verb parecer ‘to seem’ in Spanish taking an embedded indicative, there is weak epistemic commitment to the truth of the embedded proposition on the part of the speaker; with the subjunctive, a counterfactual interpretation obtains. The infelicity of a continuation like “… and in fact it is raining” in (36b) makes

176

Josep Quer

this point clear. Unless other factors can be detected, the contribution of mood in these examples is clearly not vacuous. (36a) Parece que llueve. (Spa.) seem-3.SG that rain-IND.3.SG ‘It seems that it is raining-IND.’ (36b) Parece que llueva. seem-3.SG that rain-SUB.3.SG ‘It looks as if it were raining-SUB.’

A final word of caution about the generalizations established in the literature is in order: even the observation that Italian, unlike Spanish or French, selects the subjunctive under epistemic predicates should be qualified when faced with data like the following, from Wandruszka (1991): in the first case, the future indicative seems to be the crucial element; the second example is reported as belonging to a spontaneous register. (37) Credo che verrà. (Ita.) believe-1.SG that come-FUT-3.SG ‘I believe he will come-IND.’ (38) Credo che ora è possibile diffendermi da tutte le calunnie. believe-1.SG that now be-IND.3.SG possible defend-me of all the slanders ‘I believe that now it’s-IND possible to defend myself against all slanders.’

Be that as it may, we can safely conclude that the alleged variation can and must be reexamined, even if some recalcitrant cases do not seem to be explained by the analyses proposed in this paper. Struggling with those problems will have helped us to enrich our understanding of mood distribution and interpretation in the relevant languages.

6. Meaningless or meaningful moods? After this cursory review of intra- and crosslinguistic variation in mood choice in a significant number of contexts, one may justifiably pose the question of whether it makes sense to try to unveil the underlying meaning of each one of the moods, given the disparity of contexts where they can appear. In previous work I argued that mood as a category does not have a stable meaning, as is usually assumed, but instead it marks overtly information about the models where propositions are to be interpreted: mood shift, i. e. going from indicative to subjunctive or the other way around, signals a change of model for the

On the (un)stability of mood distribution in Romance

177

evaluation of the proposition at hand (Quer 1998, 2001). A model must be understood as a set of worlds anchored to an individual. So for example, a sentence like (39) is interpreted with respect to two models: the main clause is evaluated in the epistemic model of the speaker (ME (speaker)), representing his beliefs, and the embedded clause is interpreted in the bouletic model representing the desires of the individual anchor (MBul(x)), in this case the referent of the main clause subject (Anna). This is visualized in a Discourse Representation Structure in Figure 1. This is the paradigmatic case of shift from indicative to subjunctive signaling a shift in the type of model relevant for interpretation. (39) Anna quiere [que vuelva Joanet]. (Spa.) Anna want-3.SG that return-SUB.3.SG Joanet ‘Anna wants Joanet to come back.’ x p x = Anna want (x, p) p:

y y = Joanet return (y)

MODEL SHIFT

MBul(x) ME(speaker) Fig. 1: DRT representation of model interpretation for example (39).

A secondary type of model shift marked by an embedded subjunctive would involve the shift to the common ground, which represents the propositions taken as true by the interlocutors in a given conversational context. In this sense, the common ground contains backgrounded information, signalled by so-called old information or thematic subjunctive in some descriptions. One of the main questions is why in some cases mood seems to be fixed, such as under volitionals, and in other cases it does not. The answer I offered in those works recasts the long standing observation about lexically determined mood in more theoretical terms and states that fixed mood selection is determined by economy conditions imposed by the faculty of language. Economy considerations determine the division of labour among different components of the language faculty, both internal and external to the grammar understood as the computational system. Encoding of a grammatical

178

Josep Quer

dependency by the computational system (syntax in the narrow sense) is always cheaper (more economical) than resorting to mechanisms that make use of interpretive procedures or of discourse storage knowledge. Given this perspective, if a certain dependency can be encoded by the grammar, it must be encoded in this way, because it is less costly (Reuland 2001). Volitional, directive and causative/implicative predicates obligatorily link the interpretation of the embedded proposition to a shifted model. Since this can be marked in the morphosyntax, it must be by the interface economy requirement just presented. This should explain the consistent selection of subjunctive. Epistemic and factive-emotive predicates can in principle link the embedded proposition to more than one model, thus showing apparent variability in mood selection intralinguistically and crosslinguistically: not being encoded by the computational system, mood shift is a more costly means for the interpretive component. It can be shown that other cases of apparent mood alternation may also be explained in this way. One of the most interesting findings is that crosslinguistic variation in mood distribution arises exactly where intralinguistic variation arises, strongly suggesting that the apparent differences are motivated by the same set of underlying interpretive factors. Of course, synchronic and diachronic varieties do not always grammaticalize such factors in exactly the same fashion, but the constraints in those grammaticalization patterns emerge clearly.

7. Conclusion In this paper I have tried to provide the basis for an explanation of why mood alternation in embedded domains arises where it does, both from a crosslinguistic and diachronic point of view, on the basis of a set of core data that are intended to be representative of the mood distribution map within Romance. This attempt is by no means exhaustive, but it is claimed to draw the guiding lines of a general analysis of mood distribution. The core idea is that there is no such phenomenon as real mood optionality and that variation in mood choice or licensing across varieties or across language stages is ultimately highly constrained by economy considerations imposed by the human faculty of language on the grammar components.

On the (un)stability of mood distribution in Romance

179

References D’Introno, Francesco (1990): Sintaxis transformacional del español. – Madrid: Cátedra. Dunlap, Carolyn (2006): “Dialectal Variation in mood choice in Spanish Journalistic Prose.” – In: Language Variation and Change 18, 35–53. Farkas, Donka (1992): “On the Semantics of Subjunctive Complements.” – In: Paul Hirschbühler & Konrad Koerner (eds.): Romance Languages and Modern Linguistic Theory, 69–104. Amsterdam/Philadelphia: Benjamins. Gili Gaya, Samuel (1961): Curso Superior de Sintaxis Española. Barcelona: Biblograf. Giorgi, Alessandra (2009): “Toward a syntax of the subjunctive mood.” – In: Josep Quer (ed.): Lingua (Special issue Section: The distribution and interpretation of indicative and subjunctive) 119, 1837–1858. Giorgi, Alessandra & Pianesi, Fabio (1997): Tense and Aspect. From Semantics to Morphosyntax. – Oxford: Oxford University Press. – (2004): “Complementizer deletion in Italian.” – In: Luigi Rizzi (ed.): The Structure of CP and IP, 190–210. Oxford: Oxford University Press. Grevisse, Maurice (1993): Le bon usage. Grammaire française. – Paris: Duculot. Jensen, Frede & Lathrop, Thomas A. (1973): The Syntax of Old Spanish Subjunctive. – The Hague: Mouton. Kany, Charles E. (1970): Sintaxis hispanoamericana. – Madrid: Gredos. Lope Blanch, Juan Manuel (1990): “Algunos usos de indicativo por subjuntivo en oraciones subordinadas.” – In: Ignacio Bosque (ed.): Indicativo y subjuntivo, 180– 182. Madrid: Taurus. Par, Anfós (1923): Sintaxi Catalana. – Halle: Niemeyer. Quer, Josep (1998): Mood at the Interface. – The Hague: HAG. – (2001): “Interpreting Mood.” – In: Probus 13, 81–111. – (2006): “Subjunctives.” – In: Martin Everaert & Henk van RiemsdƋk: The Blackwell Companion to Syntax. Vol. IV, 660–684. Oxford: Blackwell. Reuland, Eric (2001): “Primitives of Binding.” – In: Linguistic Inquiry 32, 439–492. Stowell, Tim (1993): “Syntax of Tense.” – Ms., UCLA. Suñer, Margarita (1999): “La subordinación sustantiva. La interrogación indirecta.” – In: Ignacio Bosque & Violeta Demonte (eds.): Gramática descriptiva de la lengua española. Vol. 2, 2149–2195. Madrid: Espasa Calpe. Wandruszka, Ulrich (1991): “Frase subordinate al congiuntivo.” – In: Lorenzo Renzi & Giampaolo Salvi (eds.): Grande grammatica italiana di consultazione. Vol. II, 415– 481. Bologna: Il Mulino.

Jan Lindschouw

Grammaticalization and language comparison in the Romance mood system1

1. Introduction It has generally been acknowledged that Romance languages have evolved linguistically at different rates as compared with their common source, Latin (cf. Delattre 1966, Harris 1974, Lamiroy 1993, 1994, 1999, 2001, 2003, Loengarov 1999, 2006). The aim of this article is to test this hypothesis by comparing the evolution of the mood system observed in concessive clauses in French (both Renaissance French (16th century) and Modern French (20th century) with that of Modern Spanish. This article consists of two parts. The first discusses some important works in this field of research, all concluding that some Romance languages are more innovative than others and consequently more distant from Latin, whereas others are more conservative and share a higher degree of similarity with Latin. This part of the article also deals with the advantages and disadvantages of comparing genetically related languages. In the second part, the hypothesis of language comparison at different rates will be examined on the basis of a text corpus of a wide variety of French and Spanish texts (literary and non-literary) in order to discover whether the mood system of French can be considered more innovative than that of Spanish, and whether the system of Modern Spanish and Renaissance French have any features in common. It would appear in fact that these two stages share certain similarities with regard to the formal and functional distribution of the indicative and the subjunctive. As far as the functional alternation is concerned, reference is made to the theory of assertion (Hooper 1975, Confais 1995, García 1999, Korzen 1999, 2003, Pérez Saldanya 1999, Haverkate 2002, Lindschouw 2006, 2007, 2008, Jensen 2008) according to which the indicative is the assertive mood and the subjunctive the non-assertive mood (see section 3.1). But are these similarities due to a temporal displacement between two genetically linked stages, or are they the result of random coincidence?

1

I would like to thank Inger Mees for her comments on an earlier version of this article.

182

Jan Lindschouw

2. The hypothesis of the evolution of Romance languages at different rates As mentioned in the introduction, the Romance languages are generally considered to form a continuum between more conservative, and normally less grammaticalized, languages and more innovative, and normally more grammaticalized, languages when compared to their common source, Latin; see Table 1, where the languages on the left side are more conservative and those on the right more innovative (Loengarov 2006: 23). It should be noted that the term grammaticalization used in this study does not only refer to the transition from a lexical to a grammatical system or from a grammatical to a more grammatical system (Meillet 1948, Lehmann 1995, Hopper & Traugott 2003), but also to reorganizations within grammatical systems (Heltoft et al. 2005). This change will be referred to as a process of regrammation (Andersen 2006). This notion is a particularly relevant descriptive parameter in accounting for the evolution of the mood system, since the indicative and the subjunctive moods already displayed grammatical function in Latin (Harris 1974), but have developed new grammatical functions in the course of time. However, this definition also applies to other parts of the grammar. Note also that not all of the linguistic categories mentioned below (e. g. the phonological changes) can be described with reference to grammaticalization theory. Nevertheless, all these changes show that Romance languages should be considered to form a continuum between more conservative and more innovative languages. Table 1: continuum of evolution within Romance languages. Sardinian Portuguese

Spanish

Catalan Italian

French

Rhaeto-Romance

The table illustrates that French is one of the more innovative Romance languages, Spanish one of the more conservative (together with Sardinian and Portuguese), while Italian occupies an intermediate position. This distribution has been established on the basis of studies at different levels of analysis. At the phonological level, Delattre (1966) presents a list of 31 cases indicating that Modern Spanish is placed at an evolutionary stage similar to that of Old French. He shows, for instance, that the assimilation of n to m preceding an oral labial consonant exists in Old French and Modern Spanish (en + porter > emporter ‘bring along’ (OF) vs. con padre [kompaðre] ‘with father’, con placer [komplașer] ‘with pleasure’ (MS)) (op. cit.: 197). In addition, he observes that the loss of d between a vowel and r took place in Old French ([padre] > [perԥ] ‘father’), and has also been noted in Modern Spanish, where

Grammaticalization and language comparison in the Romance mood system

183

the pronunciation of the fricative d is about to disappear ([paðre]/[pare] ‘father’) (op. cit. 191). Even though some counter arguments exist (for instance the manifold developments in Spanish triggered by the yod, e. g. palatalizations, losses of consonants, etc.), is seems however that in many aspects Modern Spanish resembles a stage found in Old French. The continuum presented in table 1 has also been confirmed by studies at the morphosyntactic level with respect to the partitive article (Lamiroy 1993), prepositions (Lamiroy 2001), auxiliaries (Lamiroy 1994, 1999) and the socalled non-lexical dative (Lamiroy 2003: 419–422). All these studies conclude that Modern French is much more innovative and thus more grammaticalized than Modern Spanish, which is more comparable to older stages of French. Modern Italian represents an intermediate stage between French and Spanish. As far as the auxiliaries are concerned, Lamiroy (1999: 41–42) shows that Modern Spanish and Italian are less grammaticalized than Modern French in the sense that they are generally less restrictive concerning the selection of complements. For instance, the Spanish verb detenerse ‘to stop’ forms an auxiliary construction with the preposition en ‘in, on’: detenerse en ‘to spend a lot of time on’, which accepts an infinitive complement (1) as well as a nominal complement (3). By contrast, the corresponding French verb s’arrêter ‘to stop’ does not allow an auxiliary construction with a preposition such as dans ‘in’ and consequently does not accept such complements as shown in (2) and (4). (1) (2)

(3) (4)

Max se detuvo mucho en comentar los hechos. ‘Max spent a lot of time commenting on the facts.’

(Lamiroy 1999: 42)

*Max s’arrêta beaucoup dans commenter les faits. ‘Max commenta longuement les faits.’ ‘Max spent a lot of time commenting on the facts.’

(loc. cit.)

Max se detuvo mucho en comentarios inútiles. ‘Max spent a lot of time on making useless comments.’

(loc. cit.)

*Max s’arrêta beaucoup dans des commentaires inutiles. ‘Max fit de longs commentaires inutiles.’ ‘Max spent a lot of time on making useless comments.’

(loc. cit.)

As regards the prepositions, Modern French also seems more grammaticalized than Modern Spanish. According to Lamiroy (2001: 97ff.), the use of prepositions in the verbal complementation, especially before infinitive and complement clauses, is much more restricted in Modern French than Modern Spanish. The French prepositions à and de are confined to the verbal complementation at the expense of the lexically more complex prepositions, whereas Modern Spanish, although favouring a and de, continues to use a wide

184

Jan Lindschouw

variety of more complex prepositions (e. g. para ‘in order to’, con ‘with’, en and por ‘in, inside, into’), just as Old French (e. g. Ana sueña con marcharse vs. Anne rêve de s’en aller ‘Ana/Anne dreams about leaving’). Furthermore, in Modern French à and de, contrary to in Old French, are also very frequently used as infinitive markers, where they do not introduce an indirect object but a direct object (e. g. Max accepte/craint/refuse de partir ‘Max accepts/fears/ refuses to leave’ vs. Max accepte/craint/refuse le voyage ‘Max accepts/fears/ refuses the trip’). By contrast, when a and de are found in Modern Spanish, they normally function as prepositions introducing indirect objects. Lamiroy interprets the multifunctionality of French à and de as a sign of grammaticalization, since they have lost their spatial function found in Old French. In other words, they have become colourless (Spang-Hanssen 1963), and since the choice of prepositions has become restricted to two prepositions, Lamiroy (2001: 98–99) speaks of an obligatorification (cf. Lehmann 1995: 139) or specialization (cf. Hooper & Traugott 2003: 116–118) of the prepositional system in the verbal complementation of French. Furthermore, according to Lamiroy (2001: 99), the fact that à and de in some contexts have changed their status from prepositions to infinitive markers can be analysed as a phenomenon of decategorialization (cf. Hopper & Traugott 2003: 106–111). Note however that Modern French prefers prepositional infinitive clauses in many contexts, where Modern Spanish refrains from using this construction (e. g. prier quelqu’un de faire quelque chose ‘ask someone to do something’ vs. pedir a alguien que haga algo ‘ask someone that one does something’, forcer quelqu’un à faire quelque chose ‘force someone to do something’ vs. forzar a alguien que haga algo ‘force someone that one does something’), but note that the prepositions used in French before the infinitive clause are the highly grammaticalized à and de and not the more complex ones. The continuum has also been developed on the basis of studies of the verb system. Loengarov (1999) refers to it in connection with the alternation between the two past forms, the simple and the composed perfect. His conclusions largely correspond to those of Squartini & Bertinetto (2000) who speak of an aoristic drift of the composed perfect, which is more advanced in innovative than in conservative languages. Loengarov concludes that the latter has almost entirely replaced the former in Modern French, at least in the spoken and informal written forms of the language, while in Modern Spanish the two forms are employed for different purposes, the simple perfect being used for remote past events and the composed perfect for recent past events.2 2

Nevertheless, at the same time a tendency for the composed perfect to be extended to prehodiernal contexts, originally reserved to the simple perfect, has been observed in Modern Spanish, mainly but not exclusively, in the speech of the younger

Grammaticalization and language comparison in the Romance mood system

185

In Modern Italian an intermediate case can be observed, because the two forms are used to designate different past events as in Spanish, but at the same time the composed perfect is gradually replacing the simple perfect when referring to remote past events. This analysis does however not seem entirely convincing. See below for criticism. As to the modal system, the studies of Boysen (1966), Harris (1974) and Loengarov (2006) also confirm the continuum shown in Table 1. Harris (1974) demonstrates that the subjunctive referred to as potential, which only existed in main clauses in Latin in order to indicate reservation and uncertainty, has been replaced in Modern French by the conditional belonging to the indicative paradigm. In Modern Spanish, on the other hand, Harris observes that the conditional still alternates to some extent with the potential subjunctive as in Old French. On the basis of a huge text corpus, Loengarov (2006: 343ff.) has studied the alternation between the indicative and the subjunctive mood in complement clauses embedded in a predicate of opinion, volition and communication in French, Spanish, Italian and Romanian. He observes two opposite tendencies in these languages: either the indicative/subjunctive alternation is semantically-pragmatically motivated or the modal alternation shows a tendency to disappear or is not triggered by a semantic/pragmatic principle. According to Loengarov, the degree to which these tendencies are present in the observed languages is in accordance with the proposed continuum. Thus Modern Spanish is the Romance language that most frequently uses the semantic-pragmatic properties offered by the mood system, whereas Modern French uses them most rarely. As already observed in the above-mentioned cases, Italian occupies an intermediate stage between the two tendencies. This conclusion seems however a little surprising, since the Italian mood system is traditionally considered to be very close to the Late Latin and Old French situation, where mood selection was possible after for instance verbs of belief and thought (cf. Renzi & Salvi 1991: 433ff.). Nevertheless, according to Loengarov (2006: 214–215, 343), this alternation cannot be explained in semantic/pragmatic terms, but is more or less random or can at least be explained as a difference of formality (the subjunctive being more formal than the indicative). Loengarov (op. cit.: 343) shows furthermore that Modern Italian occupies an intermediate stage between conservative Spanish and innovative French if all complement clauses (including verbs of volition and communication) examined in his corpus are taken into consideration. Spanish uses thus most generally the semantic/pragmatic possibilities offered by the generations (Squartini & Bertinetto 2000: 416–417 and the references quoted therein).

186

Jan Lindschouw

alternation between the subjunctive and the indicative, while Modern French uses them most rarely. Italian allows for an alternation between the two moods, but the difference is not semantically/pragmatically motivated. 2.1. Objections against the continuum of evolution The proposed continuum presented in table 1 is however not entirely convincing and is open to some criticism. First of all it seems too general. Loengarov (2006: 23) is aware of this fact, when he states that the place attributed to a language on the continuum is heavily dependent on the language phenomenon taken into consideration. For instance, as far as the use of auxiliaries in composed tenses is concerned, Spanish, Catalan and Portuguese are much more innovative than Italian and French. In fact, French and Italian have two auxiliaries être/essere ‘to be’ and avoir/avere ‘to have’, while Spanish, Catalan and Portuguese have only one auxiliary (haber, haver and ter ‘to have’). In Old Spanish, two auxiliaries haber and ser ‘to be’ coexisted, but ser has gradually been superseded by haber as auxiliary (Lamiroy 1999: 33– 34). A similar evolution seems to have taken place in Catalan (Badia i Margarit 1953: 326) and in Portuguese (Mattoso Câmara 1985: 168) (see also Squartini & Bertinetto 2000: 428). As regards inchoative constructions Spanish (and Italian) also seem more innovative than French. In Old Spanish and Italian the verb comenzar/cominciare ‘begin’ combined with both the infinitive markers a and de, but in Modern Spanish and Italian, a has become the obligatory infinitive marker in both languages. In Modern French, the verb commencer ‘to begin’ combines both with à and de, even though à is the favoured infinitive marker (Siversen forthcoming). Another objection that could be raised against the proposed continuum concerns the intermediate position of Italian. This statement also seems too general, as Italian displays significant differences between Northern and Southern varieties. This is for instance the case as far as the replacement of the simple perfect (passato remoto) by the composed perfect (passato prossimo) in truly past contexts (i. e. in its perfective (aoristic) function) is concerned. This replacement broadly speaking only takes place in Northern Italian, which consequently resembles Modern French, where the replacement of the passé simple by the passé composé has been completed in spoken and informal written French (though not in formal written French such as narrative passages in literature, tales and argumentative texts). By contrast, Southern Italian is much more likely to use the traditional simple perfect in marking perfective

Grammaticalization and language comparison in the Romance mood system

187

past events and thus resembles Modern Spanish to a large extent (Squartini & Bertinetto 2000: 422–426).3 In spite of these important objections, the continuum is to a certain degree plausible according to the hypothesis of the present study, since it shows that Modern Spanish is broadly speaking of a more conservative nature than Modern French, which is in accordance with the conclusions of the majority of works cited above. Recall also that the objective of this article is to test the proposed continuum, and as it appears from the following, the comparison of the mood system between Modern Spanish and Renaissance French shows that we must be extremely careful when comparing genetically related languages and particularly different stages of those languages, since the evolution of one is not necessarily an indication of a future evolution of the other. 2.2. Advantages and disadvantages of language comparison between genetically related languages This section will discuss the advantages and disadvantages of comparing genetically related languages. Beginning with the advantages, one could argue that language comparisons constitute a means of verifying changes that have occurred in genetically related languages. If a change in language x is also observed at an earlier stage of language y, there is a great probability that the observed changes have not 3

This statement is however something of a generalization. Squartini & Bertinetto (2000: 422–426) have carried out an investigation in eleven towns of Italy (three in the North, three in the Centre, three in the South, one in Sicily and one in Sardinia). They divide the functions of the composed perfect in perfectal (indicating past events relevant in present time) and perfective/aoristic (truly past events). They show that as far as the perfectal functions are concerned, the composed perfect is the preferred form in all eleven towns, but as regards the perfective function, the contrast between the North and the remaining geographical areas is significant. Northern speakers generally tend to extend the composed perfect to specifically aoristic contexts (at least in colloquial speech in personal narration and to a less extent in written texts, especially historical narration). Interestingly, this tendency also holds for Sardinian, though this language is in many respects considered the most conservative of Romance languages (cf. table 1). In the South, the composed perfect is only used with perfective value in about 20 % of the occurrences, the simple perfect being used in these contexts. The results from the Centre are quite interesting. As far as personal narration is concerned, there is a tendency to use the composed perfect for perfective uses as in the North, but when it comes to impersonal and historical narration, the composed perfect is even rarer than in the South.

188

Jan Lindschouw

occurred accidentally, but are due to a general evolutionary tendency common to both languages. Another advantage of this approach is that it enables the linguist to predict future changes in the more conservative languages. However, it is also possible to raise objections against this method. First and foremost, it could be claimed that before it is possible to undertake a comparison of the languages, it is essential to take into account the structure of the different languages and their internal distribution of grammatical categories, particularly in order to establish the relationship between the form and content of the linguistic signs, as well as their markedness relations. The second objection concerns the synchronic variation of the specific languages. How can one be sure that the change observed in language y leads to a change in language x and that what seems to be a diachronic change in one language is not simply a synchronic or regional/social variation in the other? Finally, there is a possibility that the similarity between the observed languages is the result of polygenesis, i. e. parallel changes that have taken place spontaneously and independently in genetically related languages.

3. Comparison of the mood system in concessive clauses of French and Spanish In this section, we will examine whether the proposed evolutionary continuum can be applied to the mood system in concessive clauses of French and Spanish. As mentioned in the introduction, it seems that Modern Spanish shares some features with older stages of French, more specifically with Renaissance French (16th century). This part of the article consists of three sections. In the first, the modal value of the indicative and the subjunctive will be presented, as it is believed that the mood system carries a semantic-pragmatic value at every historical step, and is not the result of a formal or ‘ornamental’ opposition with no semantic value, as stated by Bally (1944), Foulet (1968) and Prebensen (2002).4 In the next section, the methodology used to collect data is outlined, and finally the results of the empirical study are discussed.

4

For further discussion, see Lindschouw (2007: 21ff.).

Grammaticalization and language comparison in the Romance mood system

189

3.1. The indicative and the subjunctive, a difference of assertion? It seems that the alternation between the indicative and the subjunctive mood in Romance languages is due to an opposition between assertion and nonassertion; see, for instance, Hooper (1975), Confais (1995), García (1999), Korzen (1999, 2003), Pérez Saldanya (1999), Haverkate (2002), Lindschouw (2006, 2007, 2008), Jensen (2008). The theory of assertion states that the indicative mood presents the information of the clause as asserted (conveying new or rhematic information), while the subjunctive mood presents the information as non-asserted. The notion of non-assertion contains two opposite values: irrealis and presupposed (i. e. known or thematic) information.5 It should be mentioned that the notion of irrealis is used in a wide sense in this study (cf. Givón 1994). It does not only refer to so-called counterfactual hypotheses, as in the Latin tradition, but also to every utterance whose verbal content is either not realized at the moment of utterance or for some reason is questioned by the interlocutors. Irrealis thus refers to epistemic modality such as uncertainty, doubt, etc. This theoretical approach seems particularly relevant for an explanation of the mood alternation of Modern Spanish, as shown in examples (5) – (7), which illustrate mood selection in concessive clauses. (5)

Manolo compra la finca aunque su padre se opone (IND). ‘Manolo is buying the estate even though his father is against it.’

(6)

Manolo comprará la finca aunque su padre se oponga (SUBJ). ‘Manolo will buy the estate even if his father is against it.’

5

Note that two definitions of presupposition have to be distinguished, both regarding presupposition as a pragmatic category. The first is put forward by Levinson (1983: 181ff.) and considers presupposition as an element triggered by the linguistic context. The second is proposed by Chafe (1976: 30–33), who considers presupposition as an element which is part of the interlocutors’ common knowledge. Chafe speaks of givenness, which refers to the speaker’s suppositions about the interlocutors’ knowledge regarding the propositional content. The first definition of presupposition is particularly relevant to the use of mood in complement clauses, where the presupposition interpretation is triggered be a series of verbs (e. g. verbs of emotion and subjective judgement). By contrast, the second definition is relevant to the study of mood in concessive clauses, where the presupposition value seems to depend more on the speakers’ attitude than on the subordinating conjunctions. However, as it is very difficult to get access to speakers’ knowledge, a series of contextual markers has been proposed in order to support the modal interpretation (see below).

190 (7)

Jan Lindschouw Aunque Carlos tenga (SUBJ) 80 años le encanta bailar. ‘Even though Carlos is 80 years old, he loves dancing.’

In (5) the indicative se opone ‘is against’ presents the concessive clause as asserted, the concessive clause occupying the rhematic information of the phrase (Daneš 1974: 109ff., Winter 1982: 81ff., Thompson & Longacre 1985: 206ff.), whereas the subjunctive in (6) and (7) conveys non-assertive information: in (6) the information of the concessive clause is unreal (this analysis is confirmed by the future comprará ‘will buy’ in the main clause), and in (7) the subjunctive tenga ‘has’ presents the information as presupposed. This interpretation is supported by the fact that the concessive clause occupies the thematic position of the phrase (Daneš 1974: 109ff., Winter 1982: 81ff., Thompson & Longacre 1985: 206ff.). The position of concessive clauses with respect to the main clause is only to be considered a contextual parameter that supports the semantic values of the modal forms in order to avoid circularity (see below). This is a relevant descriptive parameter for distinguishing between the modal values assertion (5) and presupposition (7), but it does not have any impact on the irrealis-interpretation (6), which is possible in thematic as well as rhematic clauses. In Lindschouw (2007: 142), a strong correlation between assertion and postposition (rheme) has been observed. This correlation is particularly strong in Modern French, where all observed concessives with the indicative are postposed (op. cit.: 206–208), but it also applies to older stages of French and Modern Spanish. Lindschouw also observes a certain correlation between presupposition and preposition (theme), though this tendency is less pronounced. In Modern French, however, the theory of assertion seems less convincing as a means of explaining the mood alternation. In (8) the subjunctive soit ‘is’ is used for presupposed information, while the indicative in (9), contrary to the predictions of the theory of assertion, presents a hypothetical (unreal) state of affairs, this analysis being confirmed by the future ferons ‘will make’ in the main clause. (8)

Paul est parti bien que/quoique/malgré que Mireille soit (SUBJ) revenue. ‘Paul left even though Mireille has returned.’

(9)

Nous ferons une partie de campagne même s’il pleut (IND). ‘We will go for a picnic, even if it rains.’

It is possible that the differences between Modern French and Spanish are a result of the diachronic evolution in accordance with the evolutionary schema of table 1. It would appear, in fact, that the formal and functional mood distribution of Renaissance French corresponds to that of Modern Spanish (cf. examples (5) – (7)), as can be seen from the following examples.

Grammaticalization and language comparison in the Romance mood system

191

(10) On luy demanda si jamais elle avoit eu affaire à homme; respondit que non jamais, bien que les hommes quelques foys avoient (IND) eu affaire à elle. ‘She was asked whether she had had relations with men; she replied never, even though men sometimes had had relations with her.’ (Rabelais: Tiers livre: 156, Id 381, 1546, cit. BFM) (11) Je suis seur qu’elle ne sera point si farouche qu’elle ne permette bien qu’on la baise et qu’on luy face quelque autre chose, bien qu’au commencement elle face (SUBJ) semblant d’y resister. ‘I am convinced that she will not be so shy that she does not allow someone to kiss her and do other things, even if to begin with she pretends to resist.’ (de Turnèbe: Les Contens: 79, Id 547, 1584, cit. Frantext) (12) Encore que ton aage ne soit (SUBJ) pas achevé, ta vie l’est. ‘Even though you are not finished according to your age, your life is.’ (Montaigne: Essais: t. 1: 96, Id 359, 1592, cit. Frantext)

In (10) the indicative avoient ‘had’ of the concessive clause introduces new information. This interpretation is supported by the presence of the verb respondit ‘answered’ in the main clause, which is an assertive verb ‘par excellence’, and also by the postposition of the concessive clause (Daneš 1974: 109ff., Winter 1982: 81ff., Thompson & Longacre 1985: 206ff.). In (11) the subjunctive face ‘do’ seems to convey unreal information, this analysis being confirmed by the future sera ‘will be’ in the co-text. Finally, in (12) the subjunctive soit ‘is’ presents a presupposed state of affairs, since the concessive occupies the thematic position of the phrase (Daneš 1974: 109ff., Winter 1982: 81ff., Thompson & Longacre 1985: 206ff.) and the speaker presents information that the interlocutor is supposed to know (he speaks about the interlocutor’s age, cf. the possessive articles ton and ta ‘your’ referring to the interlocutor). This state of affairs gives us a basis for considering Modern Spanish a more conservative language than Modern French, sharing some properties with older stages of French; consequently the inability of the theory of assertion to explain the mood alternation of Modern French could be due to the evolution of that language, which is more innovative than Spanish. It is true that the line between presupposition and assertion is sometimes difficult to draw, since both concepts belong to the realis domain. This problem also applies to the opposition between presupposition and irrealis, though they belong to two different conceptual areas. This interpretative difficulty presents a potential risk of circularity as regards the identification of the functional values of the mood paradigm. In order to avoid such circularity, a series of contextual markers that support the interpretation of the mood forms has been revealed, as shown in examples (5) – (12) above. These markers are of a different nature, as they belong to different analytical levels, i. e. the

192

Jan Lindschouw

(morpho-)logical, phrasal and textual level. Markers that support the presuppositional reading of the mood forms are for instance generic statements, factual markers and markers of habits and iterativity (e. g. parfois ‘sometimes’, souvent, fréquemment ‘often’, toujours ‘always’, jamais ‘never’), reference to the interlocutors (by personal and possessive pronouns and/or possessive articles), past contexts, deictic and anaphoric markers (e. g. demonstrative pronouns and articles as well as adverbs), and thematic position of the concessive. All these markers indicate that the information in the concessive clause is supposed to be known by the interlocutors. Irrealis markers include for instance markers of non-referentiality (e. g. relative pronouns with non-specific reference qui ‘whoever’ and indefinite articles), abstract text genres, future and conditional tenses, imperatives and conditional clauses. All these markers indicate that the content of the concessive clause is formally to be considered unreal. Finally, assertion markers are verbs of saying and answering, see (10) above, concessives in the simple perfect tense, stage directions and rhematic position of the concessive. The function of these markers is to present the content of the concessive clause as new information. In Lindschouw (2007: 138–161) a more exhaustive presentation of these markers is outlined. They are placed in a hierarchy according to their relative force as contextual markers too, but this discussion is beyond the scope of the present article. The theory of assertion is naturally not the only theory of mood, which has been proposed during the 20th century, but it seems generally very plausible as an explanation for the alternation between the indicative and the subjunctive, since it does not exclusively reduce this alternation to modality, which is the case for many of the theories which take the subjunctive to be the mood of non-reality or doubt (Brunot 1922, Alarcos Llorach 1994: 153–154), subjectivity (van der Molen 1923: 36–37), prospectivity (Charaudeau 1971), etc., while considering the indicative as the mood of reality or certainty, objectivity, actuality, etc. See Lindschouw (2007: 29–31) for further discussion. As stated by Ridruejo (1999: 3215), mood alternation in Modern Spanish is not only a question of modality, and this is also the case for the other Romance languages. In complement clauses this definition holds for verbs of doubt or uncertainty (e. g. dudar ‘doubt’ and desconocer ‘not to know’), which indicate epistemic modality, and verbs of volition and desire (e. g. querer ‘want’ and desear ‘wish’), which indicate deontic modality. But verbs and expressions of emotion and subjective judgement (e. g. lamentar ‘regret’ and estar contento ‘be happy’) cannot be described in terms of modality, since they express a factual state of affairs. Recall that only the cases categorized as irrealis according to the assertion criterion (e. g. (6) and (11) above) can be described in terms of modality (epistemic modality), but not the

Grammaticalization and language comparison in the Romance mood system

193

cases where the subjunctive express factuality (i. e. presupposition) as in (7), (8) and (12) above. Consequently, a theory which also takes pragmatic and communicative factors into consideration is needed in order to give a full account of the uses of mood in Romance languages, diachronically as well as synchronically. Ridruejo (op. cit.) uses the assertion criterion in his description of mood in completive clauses in Modern Spanish, but his definition of non-assertion is more restricted than the one used in this study. According to him, nonassertion should primarily be understood as the non-affirmation of the truth in the subordinate clause; thus it is used to indicate that the verbal content is not realized (op. cit.: 3219, 3249). Ridruejo considers cases where the propositional content is presupposed (e. g. after verbs and expressions of emotion and subjective judgement) to be counter-examples to the theory of assertion. However, the extended version of the theory of assertion, adopted in this study, also considers cases where the subjunctive presents the verbal content as presupposed as instances of non-assertion, since known information has very little informative value and consequently constitutes the opposite of assertion. Even though Ridruejo’s version of the theory of assertion largely correlates with the one used in this study, it should be noted that his article is restricted to the study of mood in complement clauses, whose modal selection is governed by a different principle to that governing concessive clauses, i. e. the nature of the verb and the presence of modal factors such as negation and interrogation in the matrix clause, while mood selection in concessive clauses instead seems to depend on the speaker’s attitude towards the propositional content. 3.2. Method In order to test the hypothesis that Modern Spanish is more conservative than Modern French, an empirical study was carried out using French and Spanish electronic text corpora (see References for further information). Results were obtained for a number of French and Spanish concessive conjunctions (see section 3.3). For French 200 occurrences of each conjunction were collected from each century (the 16th to the 20th), while for Spanish 200 occurrences of each conjunction were collected from the 20th century (1975–2004). In this article, only the results from Renaissance French, Modern French and Modern Spanish will be presented, but in Lindschouw (2007) all findings are discussed. In order to constitute a representative text corpus, the data cover a huge variety of text genres, from the more formal genres such as argumentative and nonliterary prose and versified texts (poetry and drama) to more informal genres such as theatre plays in prose, dialogues and direct discourse. Literary texts are also taken into account, but it is difficult to know whether to place them at the

194

Jan Lindschouw

formal or informal end of the spectrum, as they often contain features from both. 3.3.

Results

3.3.1. Renaissance French In the Renaissance French corpus, two concessive conjunctions bien que and encore que, both meaning ‘even though/if’, emerge.6 They express two different kinds of concessive relations. Bien que indicates a concessive relation called logic (Morel 1996: 6ff.) or simple (Soutet 1990: 8), i. e. a type of concession which is based on an assertion of the co-existence of two propositions that are normally considered as incompatible. This relation could be formalized as ‘[if p, then normally ~q]’, and is illustrated by (13), where it is implied that if it rains, one is normally not expected to go for a walk, but the concessive relation suspends this expectation. (13) Bien qu’il pleuve, Marie se promène au bord de la mer. ‘Even though it is raining, Marie is walking by the sea.’

The conjunction encore que also contains this implicative structure, but contrary to bien que it modifies this implication by rectifying the content of the main clause (cf. Soutet 1990: 11, 1992: 214, 2000: 98, Morel 1996: 10ff., 25– 26). In (14) the speaker asserts in the first place that Mireille had a lot of fun last night, but subsequently modifies this statement in the concessive clause by questioning this assertion. (14) Mireille s’est beaucoup amusée hier soir, encore qu’elle est/soit rentrée tôt. ‘Mireille had a lot of fun last night, even though she went home early.’

Tables 2 and 3 give the modal distribution of bien que and encore que in the 16th century. They show that both conjunctions allow mood alternation between the indicative and the subjunctive, even though the indicative 6

Both of these conjunctions still exist in Modern French together with même si ‘even if’. As même si does not emerge until the end of the 17th and the beginning of the 18th century, this conjunction will not be considered here, but see Lindschouw (2007) for further information. It is true that the conjunction quoique ‘even though/if’ also emerges in the 16th century, but it has been omitted from the present study because it is derived from the relative subordinator quoi que ‘whatever (that)’, whose mood selection is determined by a different principle than the conjunction quoique, namely the nature of the antecedent. Furthermore, according to Grevisse (1986: 1675) these two constructions were not distinguished until the 18th century.

Grammaticalization and language comparison in the Romance mood system

195

constitutes a subsystem representing only approximately 7 % of the occurrences. The tables also indicate that the subjunctive has two functional interpretations (presupposition and irrealis), as predicted by the theory of assertion (see 3.1), and that the indicative expresses an assertion in most cases. However, in some cases a functional interpretation of the occurrences with respect to the assertion criteria has not been possible. These occurrences will be referred to as ‘unclassifiable’ in the following tables. As it appears from the tables, the indicative mood can also be used with a non-assertive (i. e. presupposed or irrealis) value in a limited number of cases, contrary to the expectations of the theory of assertion. However, these occurrences can be explained as diverging cases. The presupposed value can to a great extent be explained by metric and other stylistic features, whereas the irrealis value is due to the presence of the future and conditional forms, which belong to the indicative paradigm even though they present the content as unreal because of their prospective and/or modal values. Table 2: bien que (16th century). Subjunctive

Value

Number of occurrences

Occurrences: 185 (92,5 %)

Presupposition

158 (85,4 %)

Irrealis

17 (9,2 %)

Assertion

0 (0,0 %)

Unclassifiable

10 (5,4 %)

Indicative

Value

Number of occurrences

Occurrences: 15 (7,5 %)

Presupposition

2 (13,33 %)

Irrealis

5 (33,33 %)

Assertion

8 (53,33 %)

Unclassifiable

0 (0,0 %)

Total occurrences (subj./ind.): 200

196

Jan Lindschouw

Table 3: encore que (16th century). Subjunctive

Value

Number of occurrences

Occurrences: 186 (93,0 %)

Presupposition

122 (65,6 %)

Irrealis

60 (32,3 %)

Assertion

0 (0,0 %)

Unclassifiable

4 (2,1 %)

Indicative

Value

Number of occurrences

Occurrences: 14 (7,0 %)

Presupposition

2 (14,3 %)

Irrealis

1 (7,1 %)

Assertion

11 (78,6 %)

Unclassifiable

0 (0,0 %)

Total occurrences (subj./ind.): 200

3.3.2. Modern French During the centuries that follow, the system undergoes a gradual reduction. In Modern French we have arrived at a system where the alternation between the indicative and the subjunctive has almost ceased to exist. This applies especially to concessive clauses introduced by bien que, as shown in table 4. As we can see from the table, the subjunctive has virtually developed into the obligatory mood, having been observed in 98,5 % of the cases, contrary to the indicative, which has only been observed in 1,5 % of the instances. As far as the functional alternation of the subjunctive is concerned, the presupposition value has almost entirely replaced the irrealis value. Thus in Modern French, bien que has become restricted to the subjunctive used with a presupposed value. In Lindschouw (2007, 2008) the evolution of the subjunctive has been interpreted as a case of grammaticalization, and according to the extended definition proposed in Section 2 (Heltoft et al. 2005) as a case of regrammation (Andersen 2006), since it has undergone a paradigmatic reduction (Lehmann 1995), whereby its functional content has been reduced (desemantisized) (op. cit.: 127) and its variation with the indicative has gradually disappeared (the subjunctive has thus undergone a process of obligatorification (op. cit.: 139) and specialization (Hopper & Traugott 2003: 116–118)). This discussion is however beyond the scope of the present study. It should be noted that even though it is not completely clear from table 5, concessive clauses introduced by encore que also undergo a degree of reduction in their modal system. It is true that this conjunction allows mood

Grammaticalization and language comparison in the Romance mood system

197

alternation even in Modern French to a much greater extent than bien que, but the indicative has been restricted to concessive clauses that appear after the main clause, indicating an autonomous speech act, contrary to what is observed for the 16th century where it could be pre- or postposed. Consequently, in Modern French the two moods no longer allow free alternation as in the 16th century; see table 8. Table 4: bien que (20th century). Subjunctive

Value

Number of occurrences

Occurrences: 197 (98,5 %)

Presupposition

189 (95,94 %)

Irrealis

5 (2,54 %)

Assertion

2 (1,02 %)

Unclassifiable

1 (0,50 %)

Indicative

Value

Number of occurrences

Occurrences: 3 (1,5 %)

Presupposition

0 (0,0 %)

Irrealis

0 (0,0 %)

Assertion

3 (100,0 %)

Unclassifiable

0 (0,0 %)

Subjunctive

Value

Number of occurrences

Occurrences: 176 (88,0 %)

Presupposition

162 (92,04 %)

Irrealis

12 (6,82 %)

Total occurrences (subj./ind.): 200 Table 5: encore que (20th century).

Assertion

0 (0,0 %)

Unclassifiable

2 (1,14 %)

Indicative

Value

Number of occurrences

Occurrences: 24 (12,0 %)

Presupposition

0 (0,0 %)

Total occurrences (subj./ind.): 200

Irrealis

10 (41,7 %)

Assertion

14 (58,3 %)

Unclassifiable

0 (0,0 %)

198

Jan Lindschouw

3.3.3. Modern Spanish In Modern Spanish, the modal distribution has been examined in concessive clauses introduced by aunque ‘even though/if’ and a pesar de que ‘even though’. These conjunctions can by and large be considered as the Spanish counterparts of bien que and encore que, though they cannot be reduced to one single concessive value (except for a pesar de que). They are the most frequent concessive subordinators in Modern Peninsular Spanish. Note that aunque is multifunctional because it is able to express three concessive relations: in (15) it expresses logic concession, based on the underlying structure ‘[if p, then normally ~q]’, since it is implied that if it is late, one is normally not expected to finish the explanation, but this expectation has been cancelled. Example (16) is a case of rectificative concession (García 1999: 3819ff.). In the first place the speaker implies that if one says peculiar things, one is normally not expected to be intelligent, but instead of cancelling this expectation, as in the case of logic concession, it is in fact borne out, implying that maybe María is not as intelligent as one would have thought in the first place. Example (17) is a case of hypothetical concession, since both the protasis and the apodosis contain hypothetical information. By contrast, a pesar de que can only indicate logic concession and thus resembles bien que in French. In (18) it is implied that if one looks foolish when smiling, one is not very attractive, but this implication has been cancelled. (15) Acabaré de explicarte este tema aunque ya sea tarde. ‘I shall finish the explanation of this point even though it is already late.’ (García 1999: 3819) (16) María es una chica muy espabilada y siempre está en todo; ¡aunque a veces nos viene/venga con unas cosas más extrañas […]! ‘María is a very enthusiastic girl, and she always participates in everything; even though she sometimes says the most peculiar things.’ (loc. cit.) (17) Aunque tuviera todo el dinero del mundo, no me casaría con ese pelagatos. ‘Even if he had all the money in the world, I would never marry this silly guy.’ (op. cit.: 3832) (18) Y eso que físicamente me gusta, me parece un hombre muy atractivo a pesar de que, cuando sonríe, a veces se le queda cara de bobo […]. ‘And he still physically pleases me, I think that he is a very attractive man even though sometimes, when he smiles, he looks foolish.’ (Grandes: Los aires difíciles: 279, Id 212, 2002, cit. CREA)

The modal distribution of these conjunctions in Modern Spanish is presented in tables 6 and 7. They reveal that formally the two moods alternate to a great extent, the subjunctive representing 21,5–39,5 %, whereas the indicative con-

Grammaticalization and language comparison in the Romance mood system

199

stitutes 60,5–78,5 % of the occurrences. As regards the functional distribution, the subjunctive allows two non-assertive interpretations (presupposition and irrealis), at least in the case of aunque, while the indicative asserts the content of the concessive clause in the vast majority of cases, as predicted by the theory of assertion, but the indicative also expresses non-assertion in a limited number of cases (see below). Table 6: aunque (1975–2004). Subjunctive

Value

Number of occurrences

Occurences: 79 (39,5 %)

Presupposition

38 (48,1 %)

Irrealis

41 (51,9 %)

Assertion

0 (0,0 %)

Unclassifiable

0 (0,0 %)

Value

Number of occurrences

Indicative Occurrences: 121 (60,5 %)

Total occurrences (subj./ind.): 200

Presupposition

3 (2,48 %)

Irrealis

10 (8,26 %)

Assertion Unclassifiable

108 (89,26 %) 0 (0,0 %)

Table 7: a pesar de que (1975–2004). Subjunctive

Value

Number of occurrences

Occurrences: 43 (21,5 %)

Presupposition

36 (83,7 %)

Irrealis

3 (7,0 %)

Assertion

3 (7,0 %)

Unclassifiable

1 (2,3 %)

Indicative

Value

Number of occurrences

Occurrences: 157 (78,5 %)

Presupposition

8 (5,1 %)

Irrealis

3 (1,9 %)

Assertion

146 (93,0 %)

Unclassifiable

0 (0,0 %)

Total occurrences (subj./ind.): 200

200

Jan Lindschouw

3.3.4. Could Modern Spanish be said to reflect Renaissance French? In this section of the article it will be discussed, on the basis of the data presented in 3.3, whether Modern Spanish could be considered as being similar to Renaissance French. Let us first consider the arguments in favour of this supposition. The most convincing argument regards the alternation between the indicative and the subjunctive following different concessive conjunctions in the two stages of language, and as indicated by tables 2–3 and 6–7, this alternation can be described in terms of assertion and non-assertion. It is true, however, that in Modern Spanish the indicative has the property of indicating non-assertion in a limited number of cases (aunque: presupposition: 2,48 %, irrealis: 8,26 %; a pesar de que: presupposition: 5,1 %, irrealis: 1,9 %), but the indicative of Renaissance French also had this property (bien que: presupposition: 13,33 %, irrealis: 33,33 %; encore que: presupposition: 14,3 %, irrealis: 7,1 %).7 In other words, the distribution in terms of markedness agreement (cf. Andersen 2001) between the two modal forms is thus identical in the two stages of language. The indicative constitutes the unmarked domain, as it is able to indicate assertion as well as non-assertion, while the subjunctive constitutes the marked domain, being capable of expressing non-assertion only, with a few exceptions. As we have seen in 3.3.1, the non-assertive uses of the indicative in Renaissance French can be explained either by metric and other stylistic features (presupposition) or by the fact that the indicative is expressed by future and conditional forms, which belong to the indicative paradigm even though they present the content as unreal (irrealis). This last explanation also holds for the irrealis-uses of the indicative in Modern Spanish, which are all future and conditional forms, belonging to the indicative paradigm (Seco 1954: 61, 69, Ridruejo 1999: 3216). As far as the presuppositional use of the Modern Spanish indicative is concerned, a stylistic explanation can be suggested. As it appears from the following, the subjunctive functions to some extent as a marker of formal style, at least in propositions introduced by a pesar de que, where the indicative used with presuppositional value is most frequent. Thus, by using the subjunctive the speakers signal that they master a cultured variant of the language. By contrast, if the speakers do not want or need to give such a signal, they use the indicative regardless of its semantic value. This 7

Note that the numbers give the impression that the relative frequency of French indicatives with non-assertion value is much higher than in Spanish, but this is because the proportions of the indicative in Renaissance French are considerably lower than in Modern Spanish, which leads to greater statistic divergences, cf. table 2, 3, 6 and 7.

Grammaticalization and language comparison in the Romance mood system

201

explanation is supported by the fact that since a pesar de que favours in general the realis domain (García 1999: 3835), this conjunction only admits the assertive and presuppositional interpretations. As this opposition is very weak, the speakers are not able to differentiate the two values in practice and thus reinterpret the modal system, so that it changes from a system determined by semantic-pragmatic features to a system governed by stylistic features. The third argument in favour of this hypothesis is the position of the concessive clause in the indicative with respect to the main clause. As mentioned in 3.3.2, in Modern French the indicative is restricted to concessive clauses appearing after the main clause, indicating an autonomous speech act, but as shown by table 8, the indicative of Renaissance French appears in preand postposed clauses. As we can see, the same tendency is observed in Modern Spanish. Table 8: position of concessive clauses in the indicative mood with respect to the main clause. Bien que – indicative (16th century)

Aunque – indicative (20th century)

Preposition: 40 %

Preposition: 24,8 %

Postposition: 60 %

Postposition: 75,2 % th

Encore que – indicative (16 century)

A pesar de que – indicative (20th century)

Preposition: 35,7 %

Preposition: 32,5 %

Postposition: 50 %

Postposition: 66,9 %

Unclassifiable: 14,3 %

Unclassifiable: 0,6 %

Finally, as mentioned above, the Spanish conjunction a pesar de que exhibits a certain restriction in the subjunctive as regards its distribution across textual genres. This conjunction seems to be confined to formal genres such as argumentative and non-literary prose and is almost absent in ‘oral’ genres like drama and dialogues. The restriction could be interpreted as a sign that in this context the subjunctive is a stylistic rather than a modal marker, having been neutralized with respect to the indicative. Bien que, the French counterpart of a pesar de que, shows a similar restriction in Renaissance French, being almost exclusively observed in formal genres such as argumentative and non-literary prose as well as versified texts. It could be argued that, as a pesar de que is restricted to certain genres in Modern Spanish, it is probable that it will undergo a future reduction similar to that of bien que in French. However this observation could also be regarded as a counter-argument (see below).

202

Jan Lindschouw

In spite of these apparently convincing arguments, there are also some important counter-arguments to the proposed hypothesis. The most important is that the concessive system of Modern Spanish has a distribution somewhat different from that of Renaissance French. As we have seen in tables 2 and 3, the subjunctive is the dominant modal form in Renaissance French, where the indicative constitutes a subsystem. By contrast, in Modern Spanish almost the opposite distribution of the two moods is observed. The indicative is the dominant mood (representing 60,5 % (aunque) and 78,5 % (a pesar de que) of the occurrences). It is perhaps an exaggeration to say that the subjunctive constitutes a subsystem in Modern Spanish, but its frequency is considerably lower than that of the indicative (representing 39,5 % (aunque) and 21,5 % (a pesar de que) of the occurrences). Secondly, the functional distribution of the two moods also displays major differences when the two language systems are compared, especially that of the subjunctive. In Renaissance French the subjunctive favours the presupposition value, but Modern Spanish does not have a particular preference for any of the non-assertive values of the subjunctive, which are distributed more or less equally, a pesar de que being an exception. Finally, it seems that the French and Spanish systems have the same historical point of departure, the subjunctive being the dominant mood in both systems (cf. Pottier 1970: 191–192, Rivarola 1976: 46–47). Meanwhile, they have evolved in opposite directions, the French system having led to the development of the subjunctive as the obligatory mood, and the Spanish system to the indicative as the obligatory mood. This is particularly the case for a pesar de que. As stated earlier, it is the counterpart of French bien que, but if a pesar de que were really a reflection of the latter, one would have expected the subjunctive with presupposed value to have developed into the obligatory mood. In fact, it is in these clauses that the frequency of the indicative is highest (78,5 %).

4. Conclusion In this article a series of works that consider Modern Romance languages to form a continuum between more conservative and more innovative languages have been discussed. They generally regard Spanish as one of the most conservative languages and French as one of the most innovative, while Italian occupies an intermediate stage. Though some important objections can be raised against the proposed continuum (for instance concerning the intermediate stage of Italian), it seems to a large extent plausible.

Grammaticalization and language comparison in the Romance mood system

203

In this article the hypothesis according to which Modern Spanish is more conservative than Modern French and consequently resembles an older stage of French (i. e. Renaissance French) has been tested with reference to the mood system in concessive clauses on the basis of data taken from electronic text corpora. It can be concluded that even though there exist some apparently convincing arguments in favour of this hypothesis (for instance both Renaissance French and Modern Spanish make use of the alternation between the indicative and the subjunctive, the markedness agreement between the indicative and the subjunctive is identical in both stages of language, the indicative being the unmarked mood and the subjunctive its marked counterpart, and the indicative appears both in pre- and postposed concessive clauses in both stages), one must be careful about considering Modern Spanish as a direct parallel to an older stage of French, since a series of important counter-arguments can be raised against this hypothesis. For instance, even though modal alternation is observed in Renaissance French, it is not employed to the same extent as in Modern Spanish. There already seems to be an indication as to how mood will develop in the following centuries in French. Furthermore, the subjunctive is the favoured mood in Renaissance French, and the indicative constitutes a sub-system, whereas in Modern Spanish almost the opposite distribution of the modal forms is observed. It would thus be more appropriate to consider the two languages as two individual systems having evolved in opposite directions. However, the possibility that Modern Spanish will undergo restrictions in the modal system in the future cannot be excluded, especially in the propositions introduced by a pesar de que, but in that case it is highly probable that the indicative will be the obligatory mood, and not the subjunctive, as one would have expected, if Modern Spanish were a reflection of Renaissance French. This conclusion therefore weakens the hypothesis that Modern Spanish is able to serve as a means of verifying the evolution observed in the more ‘innovative’ language, French, and that Modern French predicts parallel future changes in Spanish. The data show how careful one must be when comparing genetically related languages and particularly different historical stages of those languages, since the evolution of one is not necessarily an indication of a future evolution of the other. First the internal structure of the languages must be studied, synchronically as well as diachronically, in order to establish the relation between the form and the content of the linguistic signs and their markedness relation. Subsequently a comparison of different stages of the languages can be undertaken, in order to see whether they have certain features in common across different historical stages.

204

Jan Lindschouw

References Alarcos Llorach, Emilio (1994/21999): Gramática de la lengua española. – Madrid: Espasa Calpe. Andersen, Henning (2001): “Markedness and the theory of linguistic change.” – In: Henning Andersen (ed.): Actualization. Linguistic Change in Progress, 21–57. Amsterdam/Philadelphia: Benjamins. – (2006): “Grammation, regrammation and degrammation. Tense loss in Russian.” – In: Diachronica 23 (2), 231–258. Badia i Margarit, Antonio (1953): “El subjuntivo de subordinación en las lenguas romances y especialmente en iberorrománico.” – In: Revista de Filología Española 37, 95–129. Bally, Charles (1932/21944): Linguistique générale et linguistique française. – Bern: Francke. Boysen, Gerhard (1966): “L’emploi du subjonctif dans l’histoire des langues romanes.” – In: Bulletin des jeunes romanistes 13, 19–33. Brunot, Ferdinand (1922): La Pensée et la langue. – Paris: Masson. Chafe, Wallace L. (1976): “Givenness, contrastiveness, definiteness, subjects, topics, and point of view.” – In: Charles N. Li (ed.): Subject and Topic, 27–55. New York/San Francisco/London: Academic Press. Charaudeau, Patrick (1971): Cours de linguistique. Description sémantique de quelques systèmes grammaticaux de l’espagnol actuel. – Paris: Centre de Documentation Universitaire. Confais, Jean-Paul (1990/21995): Temps, mode, aspect. Les approches des morphèmes verbaux et leurs problèmes à l’exemple du français et de l’allemand. – Toulouse: Presses Universitaires du Mirail. Daneš, František (1974): “Functional sentence perspective and the organization of the text.” – In: František Daneš (ed.): Papers on Functional Sentence Perspective, 106– 128. The Hague/Paris: Mouton. Delattre, Pierre (1946/21966): “Stages of Old French phonetic changes observed in Modern Spanish.” – In: Pierre Delattre: Studies in French and Comparative Phonetics. Selected Papers in French and English, 175–205. London/The Hague/Paris: Mouton. Foulet, Lucien (1919/31968): Petite syntaxe de l’ancien français. – Paris: Champion. García, Luis F. (1999): “Las construcciones concesivas y adversativas.” – In: Ignacio Bosque & Violeta Demonte (eds.): Gramática descriptiva de la lengua española. Vol. 3, 3805–3878. Madrid: Espasa Calpe (Real Academia Española). Givón, Talmy (1994): “Irrealis and the subjunctive.” – In: Studies in Languages 18 (2), 265–337. Grevisse, Maurice (1986): Le Bon Usage. Grammaire française. – Paris: Duculot. Harris, Martin (1974): “The subjunctive mood as a changing category in Romance.” – In: John M. Anderson & Charles Jones (eds.): Historical Linguistics. Vol. 2, 169– 188. Amsterdam/Oxford: North-Holland Publishing Company.

Grammaticalization and language comparison in the Romance mood system

205

Haverkate, Henk (2002): The Syntax, Semantics and Pragmatics of Spanish Mood. – Amsterdam/Philadelphia: Benjamins. Heltoft, Lars, Nørgård-Sørensen, Jens & Schøsler, Lene (2005): “Grammatikalisering som strukturforandring.” – In: Lars Heltoft, Jens Nørgård-Sørensen & Lene Schøsler (eds.): Grammatikalisering og struktur, 9–30. Copenhagen: Museum Tusculanum Press. Hooper, Joan B. (1975): “On assertive predicates.” – In: John P. Kimball (ed.): Syntax and Semantics. Vol. 4, 91–124. New York: Academic Press. Hopper, Paul J. & Traugott, Elizabeth C. (1993/22003): Grammaticalization. – Cambridge: Cambridge University Press. Jensen, Kjær (2008): Ny spansk grammatik. – Copenhagen: Akademisk Forlag. Korzen, Hanne (1999): “Principper for opstillingen af modus i kompletivsætninger på fransk.” – In: Carl Bache, Lars Heltoft & Michael Herslund (eds.): Ny forskning i grammatik. Vol. 6, 181–203. Odense: Odense Universitetsforlag. – (2003): “Subjonctif, indicatif et assertion ou: Comment expliquer le mode dans les subordonnées complétives?” – In: Merete Birkelund, Gerhard Boysen & Poul S. Kjærsgaard (eds.): Aspects de la modalité, 113–129. Tübingen: Niemeyer. Lamiroy, Béatrice (1993): “La dichotomie synchronie – diachronie et la typologie des langues romanes.” – In: Wolfgang Raible & Wulf Oesterreicher (eds.): Actes du XXe Congrès International de Linguistique et Philologie Romanes. Tome III, Section IV – Typologie des langues romanes, 211–221. München: Saur. – (1994): “Les syntagmes nominaux et la question de l’auxiliarité.” – In: Langages 115, 64–75. – (1999): “Auxiliaires, langues romanes et grammaticalisation.” – In: Langages 135, 33–45. – (2001): “La préposition en français et en espagnol. Une question de grammaticalisation?” – In: Langages 143, 91–105. – (2003): “Grammaticalisation et comparaison de langues.” – In: Verbum 25, 409– 429. Lehmann, Christian (1982/21995): Thoughts on Grammaticalization. – München/ Newcastle: Lincom Europa. Levinson, Stephen C. (1983): Pragmatics. – Cambridge: Cambridge University Press. Lindschouw, Jan (2006): “Grammaticalization, assertion and concession in French and Spanish.” – In: Kerstin Eksell & Thora Vinther (eds.): Change in Verbal Systems. Issues on Explanation, 139–160. Frankfurt a. M.: Lang. – (2007): Etude des modes dans le système concessif en français du 16e au 20e siècle et en espagnol moderne. Evolution, assertion et grammaticalisation. – Copenhagen: University of Copenhagen. – (2008): “L’évolution des modes verbaux dans les propositions ouvertes par bien que et encore que du XVIe au XXe siècle. Un cas de grammaticalisation?” – In: Benjamin Fagard, Sophie Prévost, Bernard Combettes & Olivier Bertrand (eds.): Evolutions en français. Etudes de linguistique diachronique, 249–267. Bern: Lang. Loengarov, Alexander (1999): Passé simple et passé composé: l’évolution des temps du passé du latin aux langues romanes. Étude de grammaire comparée. – Leuven: Katholieke Universiteit Leuven.

206

Jan Lindschouw

– (2006): L’alternance indicatif/subjonctif dans les langues romanes. Motivation sémantico-pragmatique et grammaticalisation. – Leuven: Katholieke Universiteit Leuven. Mattoso Câmara, Joaquim (1972/41985): História e estrutura da língua portuguesa. – Rio de Janeiro: Padrão. Meillet, Antoine (1912/21948): “L’évolution des formes grammaticales.” – In: Antoine Meillet: Linguistique historique et linguistique générale, 130–149. Paris: Champion. Morel, Mary-Annick (1996): La concession en français. – Paris: Ophrys. Pérez Saldanya, Manuel (1999): “El modo en las subordinadas relativas y adverbiales.” – In: Ignacio Bosque & Violeta Demonte (eds.): Gramática descriptiva de la lengua española. Vol. 2, 3253–3322. Madrid: Espasa Calpe (Real Academia Española). Pottier, Bernard (1970): Lingüística moderna y filología hispánica. – Madrid: Gredos. Prebensen, Henrik (2002): “Modus, modalitet og modeller. Om den franske konjunktivs betydning.” – In: Irène Baron, Michael Herslund & Henrik H. Müller: Lingvistiske essays til minde om Finn Sørensen, 97–107. Copenhagen: Samfundslitteratur. Renzi, Lorenzo & Salvi Giampaolo (1991): Grande grammatica italiana di consultazione. I sintagmi verbale, aggettivale, avverbiale. La subordinazione. Vol. 2. – Bologna: Il Mulino. Ridruejo, Emilio (1999): “Modo y modalidad. El modo en las subordinadas sustantivas.” – In: Ignacio Bosque & Violeta Demonte (eds.): Gramática descriptiva de la lengua española. Vol. 2, 3209–3351. Madrid: Espasa Calpe (Real Academia Española). Rivarola, José L. (1976): Las conjunciones concesivas en español medieval y clásico. – Tübingen: Niemeyer. Seco, Rafael (1954): Manual de gramática española. – Madrid: Aguilar. Siversen, Annelise (forthcoming): “Los marcadores de infinitivo en la construcción incoativa con comenzar y cominciare – ¿Un caso de gramaticalización?” – In: Actes du colloque CILPR 2007. Soutet, Olivier (1990): La concession en français des origines au XVIe siècle. Problèmes généraux. Les tours prépositionnels. – Geneva: Droz. – (1992): La concession dans la phrase complexe en français. Des origines au XVIe siècle. – Geneva: Droz. – (2000): Le subjonctif en français. – Paris: Ophrys. Spang-Hanssen, Ebbe (1963): Les prépositions incolores du français moderne. – Copenhagen: Gad. Squartini, Mario & Bertinetto, Pier M. (2000): “The simple and compound past in Romance languages.” – In: Östen Dahl (ed.): Tense and Aspect in the Languages of Europe, 403–439. Berlin: Mouton de Gruyter. Thompson, Sandra A. & Longacre, Robert E. (1985): “Adverbial clauses.” – In: Timothy Shopen: Language Typology and Syntactic Description. Vol. 2, 171–134. Cambridge: Cambridge University Press. Van der Molen, Willem (1923): Le subjonctif, sa valeur psychologique et son emploi dans la langue parlée. – Amsterdam/Zaltbommel: Van de Garde. Winter, Eugene D. (1982): Towards a Contextual Grammar of English. The Clause and its Place in the Definition of Sentence. – London: Allen & Unwin.

Grammaticalization and language comparison in the Romance mood system

207

Text Corpora Base de Français Médiéval (BFM):

URL:http://bfm.ens-lsh.fr [02/09/2009]

Corpus de Referencia del Español Actual (CREA):

URL: http://rae.es [02/09/2009]

Dictionnaire du Moyen Français (DMF): URL: http://atilf.atilf.fr/dmf.htm [02/09/2009] Frantext:

URL: http://atilf.atilf.fr/frantext.htm [02/09/2009]

Martin G. Becker

Principles of mood change in evaluative contexts: the case of French

1. Introduction Evaluative predicates introducing propositional attitudes of a speaker or a matrix subject vary considerably with respect to their mood selection properties. This observation holds on several levels and from several analytical perspectives. First of all, mood alternations occur with single lexical units of one and the same language system. A case in point is the French verb comprendre (‘to understand’), which displays several readings in line with the corresponding mood. The indicative profiles ‘understanding’ as a cognitive activity which results in a change of a person’s epistemic state. On the other hand, the subjunctive mood elaborates ‘understanding’ as an empathetic attitude of the speaker (or less frequently of a matrix subject), hence turns the verb into an evaluative predicate (with the meaning of something like ‘to consider p a natural reaction or feeling’). See the following contrastive example taken from Grevisse (131993: § 1073, 1619): (1a) J’ai compris que je me suis trompé. ‘I understood that I was mistaken.’ (1b) Je comprends qu’il soit mortifié et décontenancé, qu’il n’ait pu supporter une telle situation. ‘I understand that he was mortified and disconcerted (and) that he could not bear such a situation.’

However, alternation does not only show up with particular lexical items within a language system but can also be observed when we compare the selectional properties of different members of the Romance family. A very conspicuous example is mood selection in factual evaluative contexts. Romanian speakers generally opt for the indicative whereas in French and Spanish the subjunctive is the rule as the following examples show: (2)

Se bucura mama că-i veneau neamurile. ‘My mother was happy that relatives came to visit her.’

(3)

Je suis content que Pierre soit venu hier. ‘I am happy that Peter came yesterday.’

(GLR 21966: 290)

210 (4)

Martin G. Becker Me alegro de que hayas venido esta mañana. ‘I am happy that you have come this morning.’

A third perspective, however, has not been taken into consideration until now. Mood alternation also unfolds in the diachronic dynamics of language systems. A particularly interesting case is the development of mood selection in French, given that in this language the evaluative component of the complex mood system underwent a process of gradual change and reorganisation from Old French to the 17th century period of Classical French. In our article we intend to analyse the different stages and principles of mood change in the evaluative domain of the French language. Before we start our in-depth analysis of the diachrony of mood, we need to address two major aspects that will play an important role in our study. First, we must take a closer look at what is actually covered by the term ‘evaluative domain’ and ‘evaluative predicate’ (chap. 2). Then we will provide a short survey of the different approaches concerned with mood selection in the evaluative domain, since we will see throughout our study that their explanatory power is at stake (chap. 3).

2. Evaluative predicates Evaluative predicates characterise the attitude of a speaker or a matrix subject towards a state of affairs to be taken into consideration. To be more precise, we can say that the predicates in question focus on a particular propositional attitude held by some individual. As we will see, the evaluative domain is clearly distinguished from two other basic domains, the bouletic (or volitional) and the epistemic (and/or doxastic) domain. The former deals with individuals’ wishes and intentions (hence ideal worlds), the latter with what they know or believe (more or less epistemically accessible worlds). Evaluative predicates fall into three general sub-classes that are characterised by a particular logical-semantic structure. The emphasis of this article is on the first subgroup, which is composed of the so-called factiveemotive predicates. These predicates, which express a feeling of satisfaction or dissatisfaction with a given state of affairs, generally presuppose the truth of their complement and can therefore be classified as “veridical”. The truth presupposition falls out from the conceptual structure of the relevant verbs (e. g. regretter: ‘to regret’ and se réjouir: ‘to be happy with’ or copular constructions as être content que: ‘to be satisfied with’) as they imply, at least for default contexts, that the state of affairs under evaluation holds at some index of the base world (p = 1 in w0), as in, for instance

Principles of mood change in evaluative contexts: the case of French (5)

211

Nous regrettons que Pierre soit déjà venu. ‘We regret that Peter has already come.’

Incidentally, the truth requirement is not invalidated by examples adduced by Levinson (see e. g. Levinson 32000: 219), among them (6)

Pierre ne regrette pas qu’il ait raté son examen, étant donné qu’il l’a passé. ‘Peter does not regret to have failed his examination given that he passed it.’

In this type of example the predicate placed under the scope of negation yields a metalinguistic interpretation: The use of regretter is inappropriate in this context, due to the fact that the conditions of use inherent to the predicate fail to apply. The predicate presupposes that a state of affairs holds which is qualified as being regrettable. In the quoted example above this is not the case, given that the context cancels out the necessary preconditions required by the proper use of the predicate in question. However, we will see throughout this article that other types of specialised contexts were relevant in diachrony for the gradual reorganisation of the evaluative component of the mood system. In contrast to the first (factive-emotive) group, the predicates of the second group do not necessarily presuppose the truth of their complement. The socalled axiological expressions are typically embedded in copular (‘il est PRED que’) or impersonal constructions (‘il PRED que’, e. g. ‘il convient que p’). They introduce some sort of principle or evaluative norm that can be invoked, such as for example. logic, justice, goodness or convenience. Therefore the predicates imply a scale, which ‘measures’ different states of affairs by ordering them in accordance with their compliance with a principle, a norm or an ideal. The relevant predicates like il est bon, il est logique, il convient etc. refer to states of affairs which may include the real world. However, the veridical status of the complement clause can be superseded by the principle of scalar ordering implied by those predicates which focus on ideal states or, more precisely, possible worlds where particular requirements are fulfilled (e. g. logic or convenience reigns). The third group is composed of predicates like espérer (‘to hope’) and craindre (‘to fear’). These verbs also involve – conceptually – an evaluative component as they classify situations or situation types as desirable or undesirable against the backdrop of an implicit preference scale. However, another – even more prominent – characteristic allows them to be singled out as a proper sub-group in its own right: in view of the fact that these predicates refer to possible future alternatives of how things might develop (seen from the moment of utterance), they also involve a probability scale. This scale ranks different possible outcomes of ongoing developments in the actual world or, to put it slightly differently, future versions of reality as it stands at the time of

212

Martin G. Becker

utterance, according to the expected probability of their coming into existence. Therefore, hope-worlds may be considered as realistic candidates for a future world or they may be shrugged off as highly unlikely fictitious alternatives. At any rate, these probabilistic differences mattered for the speakers and gave rise to systematic mood alternations until the end of the 16th century, thus opposing future and subjunctive verbal categories (see Becker 2006). Throughout our diachronic study we will see that the three subclasses are somehow intertwined as they exert a mutual influence on their mood selection properties, especially in a diachronic perspective.

3. Accounts on mood in evaluative contexts Different analyses have been proposed to account for mood selection of emotive-factive predicates like ‘to regret’, ‘to be happy’, ‘to be sad’ and so on. It is not easy to justify mood choices in the evaluative domain of modern French, given that the relevant predicates presuppose the truth of their complement in default contexts. So if I am glad that Peter came yesterday, I take it for granted (as part of my epistemic model) that Peter really came at some past time index in the real world. Pragmatic accounts (as proposed by Lunn 1989 & 1995, Reyes 1990, Korzen and especially Haverkate 2002) emphasize the low informational status of subjunctive complement clauses. Haverkate, for instance, interprets mood variation in evaluative contexts in terms of fore- vs. backgrounding. According to this approach, the subjunctive “defocalises the content of the embedded proposition” by marking it as background information. However, this explanation overgeneralises the principle of backgrounding, given that background information is not systematically marked by the subjunctive mood in French. It suffices to mention causal clauses introduced by ‘thematic’ conjunctions like comme and puisque, which mark the content of the clause as ‘given’ or ‘shared knowledge’ of the interlocutors. The exclusively thematic status of these conjunctions is confirmed by the fact that they are not compatible with pseudo-clefting constructions which highlight the rhematic information of a clause (as shown in example (7)): (7)

*C’est puisque tu me le demandes que je te le dis. ‘Since you ask me, I tell it to you.’

(Riegel et al. 1994: 508)

As a typical example for the thematic status of puisque, as can be adduced in (8), we see that it always selects for the indicative:

Principles of mood change in evaluative contexts: the case of French (8)

213

Puisque tu sais tout, en quelle année est mort François Villon? ‘Since you know everything, in which year did François Villon die?’ (Riegel et al. 1994: 507)

Semantic accounts of mood alternation, revolving around the notion of relativised veridicality (Farkas 1992, Giannakidou 1998 & 1999), stress the idea that for the indicative to be selected the complement proposition p has to be true with respect to a speaker’s or a matrix subject’s epistemic model, formally: ||p|| = 1 in ME(s). The principle of veridicality captures basic insights into the mood selection of most Romance languages (at least in their contemporary form) and Greek with respect to belief-contexts, but it fails to account for the more complex situation in Italian, Brazilian Portuguese and, in general, in older stages of Romance languages (particularly Old French). In these cases other aspects, especially the question of doxastic accessibility and the evidential status of information come into play (see Becker 2006).1 It may be no accident that the 1

Take the following two Old French examples which illustrate the importance of the principle of doxastic accessibility and of the evidential status for mood selection in sentences introduced by belief-verbs: (a) Et il cuide bien que ce soit fame a qui il parole, mes non est, ainz est li anemis qui le bee a decevoir. ‘And he really believes that it is a woman to whom he speaks, but this is not the case. Rather it is the enemy who intends to deceive him.’ (qgraallcb, La queste del saint Graal, 91.038–91.040) (b) or cuit ie que cist cheualiers est morz qu’il n’ot mais ne entent. ‘However, I think this knight is dead because he neither hears nor understands me.’ (percevalb, Chrétien de Troyes: Perceval, ms. B. 6847–6849) In the first example, the speaker who has recourse to subjunctive morphology underlines that the worlds accessible to him and to the matrix subject with respect to the identity of x (worlds in which x is actually a woman) do not converge or overlap. In (b) the worlds accessible to the speaker and the matrix subject do coincide and the speaker signals that he is able to produce first-hand evidence in order to substantiate the truth of his belief, hence turning it into a piece of information shared by his interlocutors. The same holds for Portuguese. In a sentence like: (c) Ela acredita que as flores sejam a causa da alergia. ‘She believes that the flowers may be the cause of the allergy.’ (Cf. Gärtner 1989: 138) The propositional content may be true in some possible world (accessible to the matrix subject) which does not coincide with what the speaker considers real worlds according to his own epistemic state. A speaker who sides with the matrix subject

214

Martin G. Becker

mother tongue of the aforementioned researchers is sensitive to the principle of veridicality, especially in evaluative contexts. Heim (1992), Giorgi & Pianesi (1997) and Villalta (2005) interpret emotive factives as scalar or comparative predicates. In this way, they downgrade the presuppositional aspect of evaluative predicates and emphasize at the same time their classificatory and ordering potential. The complement proposition may hold in the actual world, but this is not the relevant factor. What really counts is the fact that p represents a general type or class of occurrences (or ‘state of affairs’) that can be classified and ‘measured’ according to an implicit scale of evaluation. Let us take a look at Heim’s illustrative example ‘John is glad that he will teach on Tuesdays’. The copular matrix construction ‘to be glad’ involves two different aspects, which underlie the predicate and must be taken into consideration when describing the worlds of evaluation in an appropriate way: a) John’s attitude has to be interpreted against the backdrop of his own belief system, which is to say in accordance with his doxastic model, whereby the latter must also include the set of possible alternatives to the given situation. b) The evaluative component implies a scalar ranking of the alternatives contained in the doxastic base, according to a principle or a norm; in the given case the principle of preference (or desirability) is at stake. The complement clause of our example singles out a particular alternative highly ranked by the matrix subjects. To be more explicit, in view of the doxastic alternatives John attaches a high value (or rank) on a preference scale to situations (or “worlds”), as described by the complement clause. Therefore we can say that worlds in which John teaches on Tuesday are a) realistic alternatives with respect to John’s beliefs, and b) preferable (or “desirable”) worlds and can, on these grounds, be classified as ‘glad’ or ‘happy’ and the like. Heim’s (1992) analysis can be formalised in the following way (see Villalta 2005: 233): (9)

[[be glad]](p)(a)(w)=1 iff ∀ w’∈ revised Doxa(w): Simw’(p)