Infinitives at the Syntax-Semantics Interface: A Diachronic Perspective 9783110520583, 9783110518474, 9783110518597, 9783110655827

Citation preview

Łukasz Jędrzejowski, Ulrike Demske (Eds.) Infinitives at the Syntax-Semantics Interface

Trends in Linguistics Studies and Monographs

Editor Volker Gast Editorial Board Walter Bisang Jan Terje Faarlund Hans Henrich Hock Natalia Levshina Heiko Narrog Matthias Schlesewsky Amir Zeldes Niina Ning Zhang Editor responsible for this volume Volker Gast

Volume 306

Infinitives at the Syntax-Semantics Interface A Diachronic Perspective Edited by Łukasz Jędrzejowski Ulrike Demske

ISBN 978-3-11-051847-4 e-ISBN (PDF) 978-3-11-052058-3 e-ISBN (EPUB) 978-3-11-051859-7 ISSN 1861-4302 Library of Congress Cataloging-in-Publication Data A CIP catalog record for this book has been applied for at the Library of Congress. Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available on the Internet at http://dnb.dnb.de. 6 2017 Walter de Gruyter GmbH, Berlin/Boston Typesetting: RoyalStandard, Hong Kong Printing and binding: CPI books GmbH, Leck ♾ Printed on acid-free paper Printed in Germany www.degruyter.com

Preface and acknowledgments This book goes back to the 46th annual meeting of the Societas Linguistica Europaea, September 18–21, 2013. The conference hosted the workshop on infinitives at the syntax-semantics interface from a diachronic perspective, convened and organized by the volume editors, Łukasz Jędrzejowski and Ulrike Demske. We thank the organizers of the SLE conference and the SLE secretary, Bert Cornillie, for their excellent preparation and administration of both conference and workshop. We would like to thank anonymous reviewers, the series editors, in particular Volker Gast, for constructive criticism and helpful suggestions. We are particularly indebted to Ilaria De Cesare, Alexander Klaaßen and Joana Windloff for their meticulous editorial work. Thanks are also due to Birgit Sievert and Julie Miess of de Gruyter for their editorial and technical support. Potsdam, May 2016 Łukasz Jędrzejowski, Ulrike Demske

DOI 10.1515/9783110520583-202

Table of contents Preface and acknowledgments List of contributors

v

ix

1

Łukasz Jędrzejowski and Ulrike Demske Infinitival patterns and their diachronic dynamics: Questions and 1 challenges

I

AcI verbs and restructuring effects

2

Thomas Grano Restructuring at the syntax-semantics interface

3

Isabela Nedelcu and Irina Paraschiv The Romanian infinitive selected by perception and cognitive verbs

4

Jerneja Kavčič A diachronic perspective on the semantics of AcI clauses in Greek

31

II Infinitive structures versus other non-finite and finite patterns 5

6

7

8

Frauke D’hoedt and Hubert Cuyckens Finite, infinitival and verbless complementation: The case of believe, 115 suppose and find Virginia Hill Early Modern Romanian infinitives: Origin and replacement

147

Augustin Speyer Semantic factors for the status of control infinitives in the history of 169 German Éva Dékány Anti-agreeing infinitives in Old Hungarian

193

55

81

viii

9

Table of contents

Rosemarie Lühr The emergence of expressions for purpose relations in older Indo-European 223 languages

Björn Wiemer 10 Main clause infinitival predicates and their equivalents in Slavic: 265 Why they are not instances of insubordination Language index Subject index

339 341

List of contributors Hubert Cuyckens KU Leuven Department of Linguistics Blijde-Inkomststraat 21 PO Box 3308 3000 Leuven Belgium [email protected]

Virginia Hill University of New Brunswick Humanities and Languages Tucker Park, P.O.Box 5050 Saint John, NB Canada E2L 4L5 [email protected]

Éva Dékány Research Institute for Linguistics of the Hungarian Academy of Sciences Benczúr utca 33. 1068 Budapest Hungary [email protected]

Łukasz Jędrzejowski Universität Potsdam Institut für Germanistik Geschichte der deutschen Sprache Am Neuen Palais 10, Haus 5 14469 Potsdam Germany [email protected]

Ulrike Demske Universität Potsdam Institut für Germanistik Geschichte der deutschen Sprache Am Neuen Palais 10, Haus 5 14469 Potsdam Germany [email protected]

Jerneja Kavčič Department of Classics Faculty of Arts University of Ljubljana Aškerčeva 2 SI-1000 Ljubljana Slovenia jerneja.kavcic@ff.uni-lj.si

Frauke D’hoedt KU Leuven Department of Linguistics Blijde-Inkomststraat 21 PO Box 3308 3000 Leuven Belgium [email protected]

Rosemarie Lühr HU Berlin Unter den Linden 6 10099 Berlin Germany [email protected]

Thomas Grano Indiana University Department of Linguistics Memorial Hall 322 Bloomington, IN 47405-7005 USA [email protected]

Isabela Nedelcu “Iorgu Iordan – Al. Rosetti” Institute of Linguistics of the Romanian Academy Calea 13 Septembrie 13 050711, Bucharest Romania [email protected]

x

List of contributors

Irina Nicula Paraschiv “Iorgu Iordan – Al. Rosetti” Institute of Linguistics of the Romanian Academy Calea 13 Septembrie 13 050711, Bucharest Romania [email protected] Augustin Speyer Universität des Saarlandes Fakultät P, FR Germanistik Neuere deutsche Sprachwissenschaft Campus C5 3 66123 Saarbrücken Germany [email protected]

Björn Wiemer Johannes-Gutenberg-Universität Institut für Slavistik, Turkologie und zirkumbaltische Studien (ISTziB) Jakob-Welder-Weg 18 55128 Mainz Germany [email protected]

Łukasz Jędrzejowski and Ulrike Demske

1 Infinitival patterns and their diachronic dynamics: Questions and challenges 1 Motivation The present volume contains contributions dealing with non-finite structures and their diachronic change across languages. It relates to a prominent and widely discussed syntactic topic, which has been attracting more and more attention, particularly in the past two decades. There exists a large number of studies on the infinitival systems of particular language periods, as well as on selected topics across various languages. The first group of studies comprises, among other publications, Robbers (1997) on Afrikaans, ter Beek (2008) and Rutten (1991) on Dutch, Chierchia (1984), Geisler (1995), Mair (1990) and Moulton (2009) on English, Hoekstra (1997) on Frisian, Bech (1955/57), Grosse (2005), Höhle (1978), Kiss (1995), Lee-Schoenfeld (2007) and Wurmbrand (2001) on modern German, Zinn (2008) on Ket, Knyazev (2016) on Russian, Denecke (1880) on Old High German, Pearce (1997) on Old French and Rochette (1988) on modern Romance languages, to name just a few. In the case of cross-linguistics studies, much attention has been paid to control structures and their derivation, see for instance Boeckx et al. (2010), Grano (2015), Landau (2013), Martin (1996), Pires (2006) as well as references cited there for recent discussion. Studies on the development and change of infinitival structures are more scarce, but see Coupé (2015) and IJbema (2002) on Dutch, Joseph (1983) on Balkan languages, Callaway (1913), De Smet (2013), Fanego (1996, 1998), Iyeiri (2010), Los (2005), Rudanko (2000) and Warner (1982) on English, Demske-Neumann (1994) on German, and Mensching (2000) and Schulte (2007) on Romance. The main aim of the present volume is to relate the synchrony and diachrony of infinitival systems in different languages to each other, based on empirical data, and to show to what extent infinitival systems of modern languages can be accounted for by examining changes in earlier language periods. In pursuing this aim, we define the term diachrony in a broad sense here, i.e. as covering not only language change observed between individual language periods of one language, but also comprising theoretical as well as empirical investigations dealing with a single language period, and being relevant to diachronic considerations in a larger sample of languages. Łukasz Jędrzejowski and Ulrike Demske, Universität Potsdam DOI 10.1515/9783110520583-001

2

Łukasz Jędrzejowski and Ulrike Demske

2 Infinitive structures in flux The development of infinitival clauses involves different syntactic dimensions of the core grammar. It not only affects the entire complementation system of a particular language, it also gives rise to changes on other levels of grammar, e.g. lexical semantics and morphology. In this section, we briefly illustrate how the syntactic dimensions interact with each other and to what extent other parts of the grammar may be involved in the processes of the change under investigation. In pursuing this aim, we examine, by way of example, the use of the directive matrix predicate befehlen ‘command’ in Early New High German (1350– 1650), and compare its selectional properties with those that are associated with its Present-day German (1900–) counterpart (for a general overview of infinitive-embedding predicates in Early New High German see Ebert 1976 and Maché & Abraham 2011). As it turns out, these observations are relevant to the Germanic as well as the typological diachrony of infinitival clauses and contribute to a broader typological discussion of clause-linkage in general (cf. Noonan 2007 [1985], Bril 2010, Cristofaro 2003, Dixon & Aikhenvald 2009, Gast & Diessel 2012, Haiman & Thompson 1988, Wiemer & Letuchiy, to appear, among many others). In Neuwe Welt, an Early New High German travel report from 1567, the object control verb befehlen ‘command’ embeds predominantly finite complements. We focus on the first 60 examples in which befehlen selects for a sentential complement here. Table 1 gives an overview of the complement types selected by befehlen: Table 1: The distribution of complement clauses embedded under the object control verb befehlen ‘command’ in Neuwe Welt (1567) Number of examples (percentage)

Complement type 1.

indicative complements

0 (= 0%)

2.

finite complements with a modal verb

3.

subjunctive complements

4.

verb second complements

5.

infinitive complements

In total:

Konjunktiv I Konjunktiv II

17 (= 28%) 0 (= 0%) 9 (= 15%) 24 (= 40%)

bare infinitives zu‑infinitives

0 (= 0%) 10 (= 17%) 60 (= 100%)

Infinitival patterns and their diachronic dynamics: Questions and challenges

3

In 50 (= 83%) out of 60 cases befehlen embeds a finite complement. Remarkably, two groups can be distinguished. The first group contains clauses headed by the complementizer dass ‘that’. The embedded verbal phrase usually contains an infinitive of a lexical verb and an inflected modal verb. The following two examples illustrate this configuration with sollte ‘should’ and wollen ‘want’: (1) Er befahl auch dem Bartholme Dias / dz er mit he command.3SG . PST also the.DAT Bartholme Dias dass he with jm biß zu dē ort Mine genant / fahren sollte him.DAT until to the place Mine call.PTCP go.INF should.3SG ‘He also commanded Bartholme Dias to go with him to the place called Mine’ (Ulrich Schmidl, 1567, Neuwe Welt, p. 3) (2) vñ befahl jm mit ernst / and command.3SG . PST him.DAT with earnestness dz er sich wolte bearbeiten that he REFL want.3SG . PST exert.INF ‘and [he] seriously commanded him to exert himself’ (Ulrich Schmidl, 1567, Neuwe Welt, p. 3) In addition, subjunctive forms could occur instead of modal verbs, too. Notice, however, that only the Konjunktiv II mood is attested in such cases: (3)

Vnd and

befahl command.3SG . PST

das that

er he

auß from

dē the

erstlich first

Holtz / wood

einem a.DAT

zwey two

Ferdinand Lorenzo Ferdinand Lorenzo

Schiff ship

machē make.INF

genant / call.PTCP

liesse let.3SG . SBJV

‘And [he] first commanded to Ferdinand Lorenzo to make two ships from wood’ (Ulrich Schmidl, 1567, Neuwe Welt, p. 3) What the examples given in (1)–(3) have in common is the presence of the complementizer dass ‘that’, the final position of finite verbs as well as the use of modal verbs or subjunctive morphology in the embedded clause. The second group of examples contains complements obeying the verb second rule which are not introduced by a complementizer. Interestingly enough, in all 17 cases (= 28%) it is the modal verb sollte ‘should’ that occupies the second position in the clause. The following example is a case in point:

Łukasz Jędrzejowski and Ulrike Demske

4 (4)

vnd and

der the

Oberst colonel

man one

sollte should.3SG

befahl / command.3SG . PST jnen them.DAT

schellen haddocks

geben give.INF

‘and the colonel commanded to give them haddocks’ (Ulrich Schmidl, 1567, Neuwe Welt, p. 7) In addition to finite complement clauses, befehlen could already select for infinitives in Early New High German, too. In all attested cases the infinitive clause is accompanied by the infinitive marker zu ‘to’: (5)

der the in in

Koenig king sein his

hette have.3SG . SBJV

Schiff ship

zu to

jm him.DAT

befohlen / command.PTCP

ziehen move.INF

‘the king commanded him to move on his ship’ (Ulrich Schmidl, 1567, Neuwe Welt, p. 26) As the example given in (5) shows, the embedded infinitive appears to the right of the matrix verb (= extraposed infinitive). Infinitival complements of befehlen could also occur to the left of the matrix verb, as illustrated in the following example: gleich der Koenig ti zuthun befohlen (6) [Das=selbige]i hatte to.do.INF command.PTCP the.same have.3SG . PST just the King ‘The king commanded straight away to do the same’ (Ulrich Schmidl, 1567, Neuwe Welt, p. 27) Here, the embedded verb tun ‘do’ occurs in the middle field of the matrix clause. Additionally, it selects the noun phrase object das=selbige ‘the same’. Notice, however, that this noun phrase does not occur as a part of the infinitive clause on the surface. Instead, it has been separated from the dependent infinitive, and has been placed in the prefield position of the matrix clause, probably for information structural reasons. Regardless of where infinitives appear relative to the matrix verb, we treat them, similarly to the finite clauses, as sentential complements, in line with work by Haider (2010: 272–353) and Sternefeld (2008: 567–569). We thus assume that befehlen selects infinitive complements which are syntactically equivalent to finite dass-clauses:

Infinitival patterns and their diachronic dynamics: Questions and challenges

5

(7) Er befahl die Sicherheitsvorkehrungen zu verstärken to strengthen.INF he command.3SG . PST the safety.precautions ‘He commanded to strengthen the safety precautions’ (DeReKo, Rhein-Zeitung, 12/4/2011) There are two arguments in favour of this view. First, befehlen allows a negation marker in the embedded clause, even if the matrix verb itself is in the scope of another negation operator. Thus, we can modify (7) as follows: (8) Er befahl nicht, die Sicherheitsvorkehrungen nicht zu verstärken NEG to strengthen he commanded NEG the safety.precautions Second, if infinitive complements of befehlen are expected to behave like finite dass-clauses, it should also be possible to use two distinct adverbial modifications encoding opposite values. In analogy to the negation argument, the following temporal mismatch (gestern ‘yesterday’ versus morgen ‘tomorrow’) provides evidence for the sentential status of both non-finite and finite complements of the predicate befehlen: (9)

Er he

befahl commanded

(morgen) tomorrow

die the

gestern, yesterday Sicherheitsvorkehrungen safety.precautions

(morgen) tomorrow

zu to

verstärken strengthen

If we compare the Early New High German data with corresponding data from Present-day German, various differences and similarities can be observed. The most striking difference pertains to the distribution of finite and non-finite complements. As we have seen above, in Early New High German befehlen predominantly selects for a modalized finite clause. In Present-day German, in turn, finite clauses are still allowed as in (10). (10)

der who

seinen his

Soldaten soldiers.DAT

sie they

sollten should.3PL

sich REFL

in in

befahl, command.3SG . PST China China

einen a

Namen name

machen make.INF

‘(. . .) who commended his soldiers to acquire renown in China’ (DeReKo, Süddeutsche Zeitung, 30/9/2000)

Łukasz Jędrzejowski and Ulrike Demske

6

However, infinitives are the preferred complement type for the object control verb befehlen. A small research query in Das Deutsche Referenzkorpus (= DeReKo) yielded exclusively hits of zu-complements.1 No other complementation patterns were attested in this search. Interestingly enough, although befehlen has not changed semantically, it has developed a strong preference towards infinitival complements, while finite complements took a backseat. One of the main research questions dealing with infinitival clauses concerns the extent to which infinitives differ from and – in case there are differences – compete with their finite counterparts. If the semantics of an infinitive-embedding predicate did not change, the question arises what the driving force behind the preference for a selected complementation type could be. Not much attention has been paid to this issue from a diachronic point of view, but see Smirnova (2011) on German, Martin (2007) on Greek, and Rudanko (2012) and D’hoedt & Cuyckens (this volume), as well as references cited there, on English. What German and Greek have in common is that diachronically speaking, infinitives started to compete with finite complements. While in German infinitives ousted finite clauses, in Greek infinitives dropped out of use completely (see Joseph 1983 and PhilippakiWarburton & Spyropoulos 2004). The disappearance of infinitival complements is not only a hallmark of the Greek complementation system, it has also been well-documented for some dialects of Romance languages, e.g. in Salentino, a southern Italian dialect (cf. Calabrese 1991, 1993) and in selected Slavic languages like in Bulgarian (cf. MacRobert 1980 and Reimann 1994; see also Dobrushina 2012 for Russian and Wiemer, this volume, for a diachronic overview of independent infinitives in Slavic). In other words, in those syntactic environments where most standard Romance and Slavic languages favour the infinitive, Greek and Bulgarian use a finite complement. Compare (11) for a volitional use of thélo ‘want’ and (12) for an aspectual use of arxizo ‘begin’ in Greek: (11)

O the

Kostas Kostas

theli want.3SG

na SBJV. PTC

odhiji drive.3SG

‘Kostas wants (him) to drive’ (Roussou 2009: 1812, ex 3) (12)

Te the

pedhja children

arxisan begin.3PL . PST

na SBJV. PTC

trexun run.3PL

‘The children began to run’ (Roussou 2009: 1816, ex 11) 1 We extracted and analyzed the data from the subcorpora Hannoversche Allgemeine 2009 and Süddeutsche Zeitung 2009.

Infinitival patterns and their diachronic dynamics: Questions and challenges

7

In (11) and (12) the complement clauses are introduced by the subjunctive particle na.2 Diachronically, the finite patterns ousted the infinitival patterns. In German, many clause-embedding predicates began embedding infinitives, at some point putting finite complements in the rear. Based on Demske (2001), Denecke (1880) and Johnk (1979) investigating infinitive complement clauses in Old High German (750–1050), we can identify approx. 250 infinitive-embedding predicates. In Present-day German there are over 1400 predicates licensing infinitives. The question arises how this happened. In-depth studies focusing on one language and covering all historical periods of this language are still missing. Nevertheless, cross-linguistically we can divide languages into two major groups: (i) languages developing infinitive complements and suppressing other complement types (e.g. all West Germanic languages), and (ii) languages reducing the number of infinitive-embedding predicates or losing the possibility to select for non-finite complements in favor of their finite counterparts (e. g. Balkan languages). As for group (i), three issues should be kept in mind. First, it does not necessarily have to be case that other complement types disappear when infinitives start to gain ground; other complement types can still be used (cf. e. g. 10 above), but they are often restricted to marked contexts (cf. e.g. Rohdenburg 1995, 2006, 2014 and D’hoedt & Cuyckens this volume). Second, if a language starts allowing the use of embedded infinitive complements in combination with specific matrix predicates, this does not necessarily mean that all clause-embedding predicates will embed infinitives at some point. Present-day German, for example, has approximately 1800 clause-embedding predicates, but only approximately 1400 of them (= 78%) can embed an infinitive clause.3 Question predicates like fragen ‘ask’ do not allow infinitival complements:4 (13) *Ichi I

fragte ask.1SG . PST

_ i einen a

Kuchen cake

zu to

backen bake.INF

2 The presence of a finite clause entails some consequences for the interpretation of the dropped subject in the embedded clause. As noted by Constantini (2006) and Kempchinsky (2009), subjunctive complement clauses embedded under selected matrix predicates often force an obviation effect, presupposing that the subject in the dependent clause cannot be co-referent with the matrix subject. We leave aside the role of obviation effects here, but the interested reader is referred to Martin (2007), who based on diachronic data from Greek elaborately shows to what extent obviation effects occurring in subjunctive complements are related to the loss of their infinitive counterparts. 3 This information comes from a clause-embedding database set up at the Centre for General Linguistics (ZAS) in Berlin. 4 This restriction does not hold for languages allowing wh-infinitives, cf. Gärtner (2009) for a typological overview. In this case, wh-infinitives can be selected by question predicates.

Łukasz Jędrzejowski and Ulrike Demske

8

A third observation goes along with the generalization that the availability of infinitival complements largely depends on the semantics of a clause-embedding predicate. With respect to the example given in (13) we have stated that German question predicates do not allow infinitives. However, some speakers may interpret fragen as a directive predicate meaning ‘request’. In this case, an infinitive complement can appear triggering an object control interpretation: (14) ?Ichi I

fragte ask.1SG . PST

_ j einen a

ihnj him.ACC

Kuchen cake

zu to

backen bake.INF

‘I asked him to bake a cake’ Note that the Dutch counterpart vragen (lit. ‘ask’) developed the directive meaning of ‘request’ (cf. IJbema 2002: 129, Ter Beek 2008: 85, 153), enabling the embedding of infinitive complements. Thus, we can translate the example given in (14) as follows: (15)

Iki I

vroeg ask.1SG . PST

hemj him

(om) COMP

_ j een a

koek cake

te to

bakken bake.INF

The complementizer om is optional. Broekhuis & Corver (2015) illustrate this variation with the following example: (16)

Jani Jan

vroeg ask.3SG . PST

Mariej Marie

_ j te to

komen come.INF

‘Jan asked (= requested from) Marie to come’ (Broekhuis & Corver 2015: 605, ex 17b’) Broekhuis & Corver (2015: 776) point out that “while it is normally always possible to omit om from infinitival argument clauses, it is not always possible to add it to infinitival argument clauses without om”. IJbema (2002: 129) provides a tentative explanation for the optionality of om: “In the course of time, the meaning of om has generalized so that it can appear in more contexts. (. . .). Om can follow verbs such as denken ‘think’ and treuren ‘mourn’, in which case there is a ‘movement of the mind’ around the object one is thinking about or mourning for. Following these verbs, om refers to the cause of or motivation for an activity. With this sense, om occurs in waarom ‘why’ and omdat ‘because’. The motivation for an activity is usually the wish to obtain a certain object or the wish to reach a certain goal. Om expresses this meaning in connection to verbs such as roepen (roepen om ‘call for’), vragen (vragen om ‘ask for’) and sturen, (. . .). Because om can express that one wishes to reach a goal, om comes into use in connection with te-infinitives to express this meaning.”

Infinitival patterns and their diachronic dynamics: Questions and challenges

9

English patterns with Dutch, as ask can be used as a directive matrix predicate, too: (17)

Ii asked himj _ j to bake a cake

However, ask does not always presuppose object control when embedding infinitives. Bhatt (2006: 116) shows that as soon as ask selects for a wh-infinitive, subject control is possible as well: (18)

a.

Stefan Arnii asked Hafdisj [_*i/j to leave]

(object control)

b.

Stefan Arnii asked Hafdisj [_ i/*j when to leave] (Bhatt 2006: 116, ex 210a,b)

(subject control)

Based on this contrast, Bhatt (2006) concludes that ask+wh,–inf and ask+wh,+inf do not seem to have a unified meaning. On his view, ask+wh,–inf means ‘request’, whereas ask+wh,+inf can be paraphrased as ‘put a question to’. It remains to be examined why the German verb fragen did not acquire the meaning of ‘request’, as its Dutch and English counterparts did. We will return to this issue of whinfinitives below. Similar contrasts can be observed between two different languages with respect to the same meaning of a single clause-embedding predicate. Take modern German and modern Polish as an example. What they have in common is that both languages are members of group (i), i.e. they have developed, or are still developing, infinitive complements in combination with selected matrix predicates. However, modern Polish – unlike modern German – does not exhibit AcI constructions (cf. Dziwirek 2000), nor does it allow infinitives in connection with factive predicates like żałować ‘regret’ (cf. Słodowicz 2008). Instead, finite complements have to be used: (19) *Żałuję, regret.1SG

nie NEG

potrafić can.INF

wysoko high

śpiewać sing.INF

Intended: ‘I regret to be not able to sing high’ (20)

Żałuję, regret.1SG

że that

częściej more.often

tu here

nie NEG

występuję perform.1SG

‘I regret that I don’t perform here more often’ (NKJP, Nasze Miasto Kraków, 20/6/2002)

Łukasz Jędrzejowski and Ulrike Demske

10

If we translate (19) and (20) into modern German, we end up with two grammatical sentences: (21)

(22)

Ich I

bereue regret

es, it

Ich I

bereue regret.1SG

nicht NEG

es, it

hoch high

dass that

ich I

singen sing.INF hier here

zu to

nicht NEG

können can.INF häufiger more.often

auftrete perform.1SG

Thus, it sometimes also varies from language to language whether or not a matrix verb licenses an infinitive complement (see also Schlotthauer et al. 2014 for some contrasts among European languages). What factors control this parametrization remains to be determined. One interesting cross-linguistic attempt has been made in the work by Sabel (2005, 2006, 2015). He proposes the following generalization: (23) The Wh-Infinitive-Generalization (WHIG) If wh-movement may terminate in the SpecCP of an infinitive in a language then this language possesses the option of filling the C-system of this (type of) infinitive with an overt complementizer. (Sabel 2015: 318) In more theory-neutral terms, the WHIG can be paraphrased as follows: As soon as a language allows wh-infinitives, this language is also allowed to have complement infinitives introduced by a complementizer. What we have seen so far is that English admits wh-infinitives (cf. 18b). Against this background, we also expect it to license infinitives with a complementizer. This prediction is confirmed by the next example, where for is usually analyzed as a complementizer: (24) As the disease progresses, it will be increasingly hard for him to breathe as his diaphragm weakens (COCA, Denver Post, 2014) Dutch behaves like English with respect to the WHIG. In the following example the matrix predicate vragen ‘ask’ selects for an infinitive clause introduced by the wh-word hoe ‘how’: (25)

. . . omdat because

Jan Jan

vroeg ask.3SG . PST

hoe how

te to

handelen handle.INF

bij at

‘. . . because Jan asked how to act in situations of danger’ (Ter Beek 2008: 159, ex 15a)

gevaar danger

Infinitival patterns and their diachronic dynamics: Questions and challenges

11

The example given in (15), in turn, contains an infinitive clause occupying the object slot of the matrix predicate, introduced by the complementizer om. German behaves differently. It neither allows wh-infinitives, nor does it have complement clauses introduced by the covert complementizer um, corresponding to Dutch om (but see Leys 1991): (26) *Stefan Stefan

fragte ask.3SG . PST

Saskia, Saskia

wann when

zu to

gehen leave.INF

Intended: ‘Stefan asked Saskia when to leave’ (27) *Er he zu to

befahl command.3SG . PST

um COMP

die the

Sicherheitsvorkehrungen safety.precautions

verstärken strengthen.INF

Intended: ‘He commanded to strengthen the safety precautions’ Sabel discusses examples from other European languages and provides more cross-linguistic evidence for the WHIG. His findings are summarized in Table 2: Table 2: European languages (dis-)allowing wh‑infinitives Languages allowing wh‑infinitives

Languages disallowing wh‑infinitives

Dutch, English, European Portuguese, French, Italian, Polish, Spanish

Danish, German, Norwegian, Swedish

According to Sabel, languages allowing wh-infinitives are supposed to allow a covert complementizer in infinitives that are realized as one of the matrix predicate arguments. We can test this assumption by taking, again, modern Polish as an example: (28)

Człowiek human.being

nie NEG

wiedział know.3SG . PST

gdzie where

uciekać run.away.INF

‘One didn’t know where to run away’ (NKJP, Express Ilustrowany, 28/7/2001) (29)

Wielu many

maluchów toddler.PL

marzy dream

żeby COMP

pójść go.INF

w in

ślady traces

‘Many toddlers dream to follow K. Mitoń’s footsteps’ (NKJP, Gazeta Krakowska, 28/9/2001)

K. Mitonia K. Mitoń.GEN

Łukasz Jędrzejowski and Ulrike Demske

12

The corpus examples in (28) and (29) provide direct evidence for the validity of the WHIG. The dependent wh-infinitive in (28) is embedded under the negated semi-factive predicate wiedzieć ‘know’ and introduced by the wh-phrase gdzie ‘where’. In (29), the embedded infinitive is licensed as an argument of the verb marzyć ‘dream’, and headed by the complementizer żeby (że = ‘that’, by = subjunctive clitic). As in Dutch, the covert complementizer can be omitted without affecting the interpretation of the clause: (29’)

Wielu many

maluchów toddler.PL

marzy dream

pójść go.INF

w in

ślady traces

K. Mitonia K. Mitoń.GEN

‘Many toddlers dream to follow K. Mitoń’s footsteps’ Based on what we have seen so far we are bound to assume, from a diachronic point of view, that the availability of wh-infinitives may have paved the way for the development of covert complementizers in dependent infinitives. To the best of our knowledge, not much attention has been paid to this affinity, though. Sabel (2015: 316) assumes that “before the Middle English period no wh-infinitives are attested, but they are found after the Middle English period, i.e. after for introduces complement clauses as a complementizer.” Fischer et al. (2000: 96) show that embedded infinitival questions are first attested in Middle English: (30)

ant and

nuste NEG .know.3SG . PST

hwet what

seggen say.INF

‘and [he] didn’t know what to say’ Likewise, we find the first instances of wh-infinitives in Middle Polish (1535–1780): (31) nie NEG

wiedzieli czym go zadzierżeć w żywocie iego know.3PL . PST what.INS him.ACC keep.INF in life his

‘they didn’t know how to save his life’ (M. Rej, 1558, Wizerunek własny żywota człowieka poczciwego) What (30) and (31) have in common is that the dependent wh-infinitives are complements of the negated matrix verb know. Interestingly, Gärtner (2009: 25) illustrates for Middle English (for the period between 1225 and 1450) that among 20 infinitival wh-complements 17 examples are complements of know, embedded in a negative environment. It would be interesting to see if in the other languages allowing wh-infinitives, these constructions emerged in combination with the verb

Infinitival patterns and their diachronic dynamics: Questions and challenges

13

know as well, and if so, for what reason this particular environment gave rise to wh-infinitives. Finally, it is worth investigating at what time covert complementizers in complement clauses appeared for the first time. In this connection, we would gain more diachronic insight not only into a better understanding of the WHIG but also into the diachrony of complementizers in general. As pointed out above, the complementizers om in Dutch and żeby in Polish can – but need not – be used in selected infinitival complement clauses. In addition, they can also be used in infinitive adjunct clauses introducing purpose clauses (see also Lühr this volume for older Indo-European languages): (32) Bernard ging naar Amerika om beroemd te worden America COMP famous to become.INF Bernard go.3SG . PST to ‘Bernard went to America in order to become famous’ (Sabel 2015: 316, ex 20) (33) Przyszliśmy, żeby zobaczyć pomnik monument come.1PL . PST COMP see.INF ‘We came to see the monument’ (NKJP, Dziennik Zachodni, 28/6/2004) Although the German complementizer um cannot be used in infinitive complement clauses,5 it is allowed to occur in infinitive purpose clauses. Hence, we can translate the Dutch example into German as follows: (34) Bernard ging nach Amerika, um berühmt zu werden America COMP famous to become.INF Bernard go.3SG . PST to Typologically, such a situation is, to some extent, expected. Schmidtke-Bode (2009: 155) examines structural patterns encoding purposive relations in a sample of 80 languages and makes two interesting observations concerning the prominence of infinitives with respect to adverbial clauses in general (see also Wiemer this volume for the prominent role of infinitives in independent structures). Firstly, in 16 languages (= 20%), purpose clauses are the only semantic type of adverbial clause that is expressed by a non-finite construction. Secondly, in 12 languages (= 15%) purpose clauses, but none of the other adverbial clauses, employ an infinitive or an otherwise highly integrated construction. In addition, Schmidtke-Bode (2009: 158) points out that while “in 62 languages (= 77,5%)

5 Cf. Demske (2011) for the use of infinitival complementizers in complement clauses in Swiss German as well as Pennsylvania German.

14

Łukasz Jędrzejowski and Ulrike Demske

at least one purpose clause construction shares some of its morphosyntactic properties with (certain kinds of) sentential complements, up to being completely identical with them, (. . .) in only 18 languages are purpose and complement constructions fully distinct.” One of the examples he refers to comes from Tzutujil (Mayan). In this language the preposition ch(i) ‘at’, ‘to’, ‘in order to’ introduces both purpose and complement clauses (cf. Dayley 1985: 392). It is not surprising that om in Dutch and żeby in Polish can fulfill both functions, too. It calls for an explanation, however, why this possibility did not prevail in German and has been restricted only to purpose clauses. An answer to this question is, in our view, still missing. Another issue closely related to the dynamics of infinitive clauses refers to the presence or absence, and, if present, the role of infinitival markers, e.g. of zu ‘to’ in German. There is no consensus among researchers regarding the syntactic position occupied by infinitival markers, or their main function, cf. Beukema & den Dikken (1989), Biskup (2014), Christensen (2007), Leys (1985), Salzmann (2013), Wilder (1988), among many others. Different scenarios have also been proposed with respect to the historical development of infinitival markers, cf. e.g. Demske (2001) and Speyer (this volume) for German, Faarlund (2007) for Norwegian and Fischer (1995) for English. These authors convincingly illustrate that the development of infinitival complements headed by an infinitival marker cannot be attributed to a universal grammaticalization path, as has been proposed in Haspelmath (1989) or Abraham (2004). It still remains to be investigated what motivated the presence/absence of infinitival markers in earlier stages of a particular language, and what led to the processes of change resulting in a requirement for modern languages to use infinitival markers in most control environments. As we have seen above, in our small corpus study from Early New High German befehlen selects for infinitival complements accompanied by the infinitival marker zu ‘to’. No bare infinitives could be found. This might suggest that befehlen never embedded bare infinitives, and that it selected for a zu-infinitive already in earlier stages of German. Smirnova (2001), for instance, assumes such a scenario for directive predicates in German. Note, however, that this view is challenged by the occurrence of other directive predicates, e.g. bitten ‘request’, which used to select bare infinitives in earlier stages of German. (35) is from Old High German, (36) from Middle High German and (37) from Early New High German: (35)

thoh though

bát ask.3SG . PST

er he

nan (. . .) him.ACC

thie the

stéina stones

duan do.INF

‘Though he asked him to turn stones into bread’ (Otfrid von Weißenburg, Das Evangelienbuch, II, 4: 44)

zi zu

bróte bread

Infinitival patterns and their diachronic dynamics: Questions and challenges

(36)

diu that

mich (. . .) me.ACC

den the

rîchen rich

der the

edeln noble

herren lords.DAT

künste arts.GEN

15

swaere importance.NOM

künden announce.INF

bat ask.3SG . PST

‘[the story] which importance of noble arts urged me to announce to the rich lords’ (Konrad von Würzburg, 1250–1287, Die Klage der Kunst, 32,8; Speyer, this volume) (37)

bat ask.3SG . PST

sye her.ACC

morgens in.the.morning

herwider again

gon go.INF

‘[he] asked her to go [there] in the morning again’ (Kalenberg, p. 194) Moreover, it remains to be explained why most Slavic languages never developed an infinitival marker. In Polish, for example, rozkazać ‘command’, the counterpart of German befehlen can embed infinitival complements as well, and is even the preferred complement type in modern Polish: (38)

Czesław Piątas Czesław Piątas

rozkazał command.3SG . PST

przerwać stop.INF

ćwiczenia training

‘Czesław Piątas commanded to stop the training’ (NKJP, Dziennik Zachodni, 5/4/2003) An interesting difference between German befehlen and Polish rozkazać is that the former verb requires the presence of the infinitival marker zu ‘to’, whereas the latter predicate can only embed bare infinitives. Note, in addition, that the presence of an infinitival marker has no impact on the syntactic size of the embedded complement. As Polish rozkazać allows two independent verbal negations and distinct adverbial modifications marking opposite values, its infinitive complements ought to be analyzed as sentential complements corresponding to finite clauses: (39)

(40)

Czesław Piątas Czesław Piątas

NEG

Czesław Piątas Czesław Piątas

rozkazał command.3SG . PST

jutro tomorrow

nie

ćwiczenia training

rozkazał command.3SG . PST

nie NEG

wczoraj yesterday

przerywać stop.INF przerwać stop.INF

ćwiczeń training

16

Łukasz Jędrzejowski and Ulrike Demske

Finally, it is worth investigating what role word order properties of infinitival complements play, and how they determine their syntactic status. Considering German befehlen and Polish rozkazać, we assumed that both predicates embed sentential complements. However, verbs taking infinitival complements are usually regarded as falling into two main classes, depending on the resulting clausal structure. There are infinitival patterns exhibiting monoclausal characteristics like pronoun fronting, long inversion or wide scope of negation. On the other hand, infinitival constructions allowing extraposition, the intervention of nonverbal material between matrix verb and infinitive or pied piping point to a biclausal structure (cf. e.g. Wurmbrand 2001 for German, Cardinaletti & Shlonsky 2004, Cinque 2006 and Haegeman 2010 for Italian, Ter Beek 2008 for Dutch and Sabel 1996 as well as Grano 2015, this volume for cross-linguistic generalizations). There is an ongoing debate focusing on present-day languages whether or not both types of constructions are derivationally related, a discussion going back to Evers (1975). For a comprehensive presentation of the most prominent accounts, see Haider (2010). Demske (2015) shows for the history of German that the distinction between monoclausal and biclausal infinitival constructions is well established in Old High German. Coherent infinitival patterns as a particular type of monoclausal infinitival constructions, however, start to arise towards the end of the Early New High German period in the 17th century. In the following section, we briefly summarize the most important diachronic findings of the contributions collected in the present volume and show how they contribute to a diachronic typology of infinitive clauses.

3 The structure of the book This volume consists of two parts, each dedicated to different aspects of infinitival syntax.

3.1 AcI verbs and restructuring effects As introduced above, verbs embedding infinitival complements can be divided into two classes, depending on the resulting clausal structure exhibiting either monoclausal or biclausal characteristics. In the present volume, the contributions by Thomas Grano, Isabela Nedelcu & Irina Paraschiv and Jerneja Kavčič address questions concerning the historical development of monoclausal versus biclausal structures.

Infinitival patterns and their diachronic dynamics: Questions and challenges

17

Grano examines in his contribution, Restructuring at the syntax-semantics interface, the class of verbs selecting infinitival complements at large, focusing on the question of why particular verbs cross-linguistically trigger clause union effects such as clitic placement in Romance (Italian, Spanish) or so-called “long passives” in German. Elaborating on Cinque’s (2006) proposal to treat the verbs in question as functional heads, Grano demonstrates how an approach in terms of a universal hierarchy of functional heads accounts for the cross-linguistic stability of the relevant verbal class in a straightforward way, with (41) exemplifying the analysis of the control verb try as an aspectual head: (41)

a.

John tried to leave

b.

[TP John T(ense) ↑

[AspP tried [vP John to leave]]]

Grano further tackles the question of how the raising syntax of such an approach may encompass even classical control predicates such as try. Introducing a variable for this purpose, the author demonstrates how this variable at the same time prohibits non-restructuring predicates from triggering monoclausal infinitival constructions. Whether a predicate allows for the monoclausal structure, in addition to a biclausal structure, depends on its degree of semantic bleaching: Semantically bleached predicates have been cross-linguistically observed to be ‘unstable’, in terms of their status of restructuring. Nedelcu & Paraschiv are concerned with infinitival complements of (direct and indirect) perception and cognition verbs focusing on varieties of Romanian used between the 16th and the 19th centuries. In their contribution, The Romanian infinitive selected by perception and cognitive verbs, the authors consider the infinitival patterns at issue as monoclausal, pointing out that they admit a diagnostic well known from other Romance languages, i.e. clitic climbing (see also the contribution by Grano): (42)

Li-au

CL . ACC .3SG . M -have.3PL

văzut seen

a cădea fall.INF

[ti]

după off

cal horse

‘They saw him falling off the horse’ (BVS) Examples such as (42) suggest a treatment of infinitival constructions headed by perception verbs and cognition verbs as complex predicates, just like infinitival constructions headed by modals and auxiliaries. Further issues addressed in the paper include the competition between different forms of non-finite complements on the one hand, i.e. infinitives with/without an infinitival marker vs. gerunds,

18

Łukasz Jędrzejowski and Ulrike Demske

and, on the other hand, between non-finite clauses and finite clauses, with the latter comprising indicative and subjunctive verbal forms (for more details see Section 3.2 below). A final question to be dealt with concerns the grammatical status of the particle a being used as a complementizer in early stages of the language, while figuring as an infinitival marker in later stages. Kavčič’s contribution, A diachronic perspective on the semantics of AcI clauses in Greek, is a corpus study on restructuring verbs embedding AcI complements in Ancient Greek, focusing on verbs of saying and thinking. Kavčič provides abundant empirical evidence showing that these predicates allow the embedding of the aorist infinitive as well as an infinitive marked for present or future tense in Classical Greek. It is illustrated that temporal relations of anteriority, simultaneity and posteriority between the matrix verb and the embedded proposition can be expressed by using the respective infinitive form. Expanding on a proposal by Thorley (1989), the author assumes that infinitival clauses tend to denote states in New Testament Greek, including the frequent use of stative verbs in non-finite clauses, and of the perfect marker with nonstative verbs when forms of the aorist are embedded. These observations carry over to concurrent sources of Greek, i.e. the non-literary papyri, all of which are taken to give insights into the development of Post-Classical Greek. Kavčič’s contribution analyzes the changes affecting the temporal and aspectual distinctions in AcI constructions with respect to other changes occurring simultaneously, such as the merger between the aorist and the perfect, the disappearance of the Classical Greek future and the emergence of periphrastic future forms.

3.2 Infinitive structures versus other non-finite and finite patterns Our second and even more prominent research question concerns the extent to which infinitives differ from, and compete with their finite counterparts (see Section 2 above). In the present volume, the contributions by Frauke D’hoedt & Hubert Cuyckens, Virginia Hill, Augustin Speyer, Éva Dékány, Rosemarie Lühr and Björn Wiemer deal with instances of diachronic complementation competition across different languages and show what consequences such competition had for the modern languages. In their corpus-based contribution, following the spirit of previous work (cf. Cuyckens et al. 2014 and Cuyckens & D’hoedt 2015), Finite, infinitival and verbless complementation: The case of ‘believe’, ‘suppose’ and ‘find’, D’hoedt & Cuyckens investigate four complement types embedded under the matrix verbs mentioned in the title of their contribution, in the history of English. The authors

Infinitival patterns and their diachronic dynamics: Questions and challenges

19

focus on that-clauses, zero-complementizer clauses, to-infinitives and small clauses. The examples given in (43) illustrate their use with the predicate believe: (43)

a.

I believe that he is a wise man

b.

I believe he is a wise man

c.

I believe him to be a wise man

d.

I believe him a wise man

The main objective of the study is to offer a diachronic analysis of complement clause variation from Middle English to Present-day English – with a focus on Late Modern English – and to investigate what factors co-determine this variation at the level of usage. In pursuing this aim, D’hoedt & Cuyckens use a logistic regression model, taking the following factors into account: (i) the semantics of the complement-taking predicate, (ii) the temporal relation between the matrix clause and the complement clause, (iii) the structural complexity of the complement clause, (iv) the complexity of the complement clause subject, (v) the voice of the verb in the complement clause, (vi) intervening material between the complement-taking predicate and the complement clause subject, (vii) the period of attestation and (viii) the corpus. In general, both authors state that complementation preferences are to a large extent unique for each complement-taking predicate. One of the main trends refers to the use of toinfinitives and small clauses in connection with believe and suppose. The authors provide empirical evidence showing that the number of both complement types decreases over time. Find, on the other hand, selects predominantly small clauses. With respect to zero-complement clauses, D’hoedt & Cuyckens illustrate that their number increases when embedded under suppose. While D’hoedt & Cuyckens investigate differences between infinitive and indicative complements, the contribution by Hill, Early Modern Romanian infinitives: Origin and replacement, is concerned with the replacement of infinitives by subjunctive sâ-clauses in the history of Romanian. The author observes that the replacement of infinitives by subjunctive complements occurred very late in Romanian, i.e. in the 16th and 17th centuries, whereas other Balkan languages had replaced the infinitive much earlier (beginning around the 10th century). Some scholars working on Balkan and Romance languages take Greek as a starting point of the complementation competition and account for this displacement in terms of geographical distance (cf. e.g Rohlfs 1933). Hill, in turn, examines this issue from a syntactic perspective and arrives at different conclusions. Contrary to previous studies, she distinguishes two infinitive forms in older stages of

20

Łukasz Jędrzejowski and Ulrike Demske

Romanian: (i) the long infinitive with the ending -re and (ii) the short infinitive lacking an ending: (44) sosiră pre cel pămînt cel în carele a lăcuire era that in which.the to live.INF was arrive.3PL . PST on that land ‘they arrived in the land in which they had to live’ (Palia de la Orăştie – 1582; {235}) (45)

le-au them.DAT-has

poruncit ordered

a to

face do.INF

şi and

a to

cinsti rever.INF

Domnedzeu God

‘he ordered them to receive and revere God’ (Palia de la Orăştie – 1582; {5}) These two forms occur in free variation and both of them were replaced by the subjunctive sâ-clause. Based on internal properties of both complement types, Hill associates the infinitival marker a and the subjunctive marker sâ with the same syntactic position, viz. FinP, placed in the extended C-domain as proposed by Rizzi (1997). The complement patterns differ, however, with respect to their productivity on the timeline and their phasehood properties. Whereas a-infinitives are productive in the 16th and 17th centuries, sâ-clauses start to be the preferred complementation type in the middle of the 17th century and become fixed by the 18th century. This change is attributed to the status (weak vs. strong) and the absence/presence of the full CP phase by Hill (cf. Chomsky 2001, 2006). The contribution by Speyer, Semantic factors for the status of control infinitives in the history of German, is devoted to the distinction between AcI constructions and object control constructions on semantic grounds, working on the assumption that there is no structural difference between both constructions. In particular, the author defends the view that the morphological distinction between the bare infinitive in the AcI construction and the zu ‘to’-infinitive in the control construction correlates highly with the aspectual meaning of the infinitive. While the bare infinitive is used to express a punctual meaning (cf. 46), zu ‘to’ + infinitive refers to propositions denoting a durative aspect as well as to actions implied to have taken place at several occasions, as exemplified in (47): (46) Dar nach chomen die tummen magde und baten in vf tuon there after come the dumb maids and asked him.ACC open.INF ‘After that the foolish bridesmaids came and asked him to open (the door)’ (Altdeutsche Predigten 30,35; before end 14th cent.)

Infinitival patterns and their diachronic dynamics: Questions and challenges

21

(47) und klagten in allez und bâten inz wenden and complained him.ACC all and asked him-to turn.INF ‘and they complained to him about everything and asked him to undo it’ (Ottokar aus der Gal: Steirische Reimchronik, v. 63909; c. 1300) According to Speyer, the aspectual difference between the two infinitival forms is not restricted to Middle High German (1050–1350) but still holds for the distribution in modern German, though Speyer assumes a reanalysis of the aspectual meaning in terms of temporality. Besides temporal properties of infinitival clauses and based on Fischer (1995), causativity is addressed as a second parameter driving the distribution of bare infinitives vs. zu ‘to’-infinitives. In contrast to widespread assumptions regarding the analysis of infinitival constructions, both control patterns and AcI patterns are considered to exhibit diagnostics for a biclausal structure in Middle High German. In her paper, Anti-agreeing infinitives in Old Hungarian, Dékány elaborates on non-finite clauses in the history of Hungarian. Her main focus lies on agreement patterns of Old Hungarian infinitives, including so called anti-agreeing infinitives that have not yet been described elsewhere in more detail (but see Bácskai-Atkári & Dékány 2014). According to the author, the anti-agreeing infinitives occur in control, raising and ECM structures, but not in infinitives with a referentially independent dative subject. Furthermore, Dékány observes that Old Hungarian infinitives can be either inflected or uninflected. Inflected infinitives show full phi-feature agreement with their subject(’s controller), as given in (48), or they bear a default third singular ending regardless of what person and number features the infinitival subject has, cf. (49): (48)

ne not

akar-i-atoc want-IMP-2PL

fel-n-etec fear-INF-2PL

‘Do not want to be afraid’ (Munich Codex 42ra) (49)

Ne not

akar-y-atok want-IMP-2PL

feel-ny-e fear-INF-3SG

‘Do not want to be afraid’ ( Jordánszky Codex 55) In those contexts where anti-agreement can occur, it is in free variation with agreeing and uninflected infinitives. Dékány preliminarily assumes that if Old Hungarian anti-agreement is only possible in control, raising and ECM infinitives, the default agreement might be triggered by an A-movement.

Łukasz Jędrzejowski and Ulrike Demske

22

The contribution by Lühr, The emergence of expressions for purpose relations in older Indo-European languages, is devoted to clausal patterns encoding a purposive meaning in Hittite and Vedic. The author mainly elaborates on clausal patterns allowing both a subject/object control interpretation on the one hand, and a purpose reading on the other. Observing a structural competition between finite and non-finite clauses in particular with respect to Vedic, Lühr examines the conditions under which both patterns occur and provides an account for their striking differences. One of her main observations regarding the expression of purpose is that finite patterns are preferred if a subject/object control reading would obtain as well in an infinitive clause. The competition of non-finite clauses with finite clauses is also addressed in the contribution by Wiemer, Main clause infinitival predicates and the equivalents in Slavic – Why they are not instances of insubordination, who mainly focuses on the diachrony of independent clauses in Russian, Polish and Macedonian. (50) is an example from Russian: (50)

Ne NEG

opozda-t’ come-late.PFV. INF

by! PTC . IRR

lit. ‘If only not to be late!’ (‘We/I mustn’t be late!’) (50) consists of the negation element ne, the perfective infinitive verb opozdat’ ‘come late’ and the irrealis particle by, corresponding to the subjunctive mood in most Germanic and Romance languages. The entire clause gives rise to an optative reading. In general, root infinitives have been mainly investigated in Germanic and Romance languages (cf. Deppermann 2007, Fries 1983, Gärtner 2013, Glaser 2002, Grohmann 2000, Reis 2003, Wilder 2013), often in connection with language acquisition (cf. Gretsch 2008, Haegeman 1995, Lasser 2002, Rizzi 1993/94). Murasugi et al. (2010) and Sugiura et al. (2016) also investigate so called root infinitive analogues in Japanese. The language sample chosen by Wiemer, in turn, is interesting in several respects. First, the modal root infinitives developed into different directions in the three languages. In Russian no radical changes have taken place. In Polish, in selected environments infinitives were replaced by finite clauses. The Macedonian case illustrates a scenario in which the grammatical category of infinitive disappeared entirely. Second, besides some aspects that play a decisive role in the structural competition, Wiemer introduces an additional parameter: factivity versus non-factivity. One of his major claims is that whenever an infinitive occurs, it is tightly associated with non-factivity. This association, according to Wiemer, can be traced well back into pre-documented history and is largely due to the provenance of the Slavic infinitive as a nomen actionis in the dative.

Infinitival patterns and their diachronic dynamics: Questions and challenges

23

Abbreviations 1/2/3 – 1st/2nd/3rd person, ACC – accusative, CL – clitic, COM – comparative, COMP – complementizer, DAT – dative, GEN – genitive, IMP – imperative mood, INF – infinitive, INS – instrumental, IRR – irrealis, M – masculine, NEG – negation, NOM – nominative, PASS . AUX – passive auxiliary, PFV – perfective, PL – plural, PST – past tense, PTC – particle, PTCP – participle perfect, REFL – reflexive pronoun, SBJV – subjunctive mood.

References Abraham, Werner. 2004. The grammaticalization of the infinitival preposition – toward a theory of ‘grammaticalizing reanalysis’. Journal of Comparative Germanic Linguistics 7: 111–170. Askedal, John Ole. 1998. Zur Syntax infiniter Verbalformen in den Berthold von Regensburg zugeschriebenen deutschen Predigten. Vorstufe der topologischen Kohärenz-InkohärenzOpposition. In John Ole Askedal (ed.), Historische germanische und deutsche Syntax. Akten des Internationalen Symposiums anlässlich des 100. Geburtstages von Ingerid Dal, Oslo, 27.9.–1.10.1995, 231–259. Frankfurt: Peter Lang. Bácskai-Atkári, Júlia & Éva Dékány. 2014. From non-finite to finite subordination: The history of embedded clauses. In Katalin É. Kiss (ed.), The Evolution of Functional Left Peripheries in Hungarian Syntax, 148–223. Oxford: Oxford University Press. Bech, Gunnar. 1955/57. Studien über das deutsche verbum infinitum. København: Munksgaard. Beukema, Frits & Marcel den Dikken. 1989. The position of the infinitival marker in the Germanic languages. In Dany Jaspers, Wim Klooster, Yvan Putseys & Pieter Seuren (eds.), Sentential Complementation and the Lexicon. Studies in Honour of Wim de Geest, 57–75. Foris: Dordrecht. Bhatt, Rajesh. 2006. Covert Modality in Non-finite Contexts. Berlin: de Gruyter. Biskup, Petr. 2014. For, zu and feature inheritance. Linguistische Arbeitsberichte 92: 423–440. Boeckx, Cedric, Norbert Hornstein & Jairo Nunes. 2010. Control as Movement. Cambridge: Cambridge University Press. Bril, Isabelle. 2010. Clause Linking and Clause Hierarchy. Syntax and Pragmatics. Amsterdam: John Benjamins. Broekhuis, Hans & Norbert Corver. 2015. Syntax of Dutch. Volume 2: Verbs and Verb Phrases. Amsterdam: Amsterdam University Press. Calabrese, Andrea. 1991. The lack of infinitival clauses in Salentino: A synchronic analysis. In Christiane Laeufer & Terrell A. Morgan (eds.), Theoretical Analyses in Romance Linguistics. Selected Papers from the Linguistic Symposium on Romance Languages XIX, Ohio State University, April 21–23, 1989, 267–294. Amsterdam: John Benjamins. Calabrese, Andrea. 1993. Sentential complementation in a language without infinitival clauses: The case of Salentino. In Adriana Belletti (ed.), Syntactic Theory and Dialects of Italy, 28– 98. Torino: Rosenberg & Selliers. Callaway, Morgan. 1913. The Infinitive in Anglo-Saxon. Washington, D.C.: Carnegie Institution of Washington.

24

Łukasz Jędrzejowski and Ulrike Demske

Cardinaletti, Anna & Ur Shlonsky. 2004. Clitic positions and restructuring in Italian. Linguistic Inquiry 35: 519–557. Chierchia, Gennaro. 1984. Topics in the Syntax and Semantics of Infinitives and Gerunds. PhD thesis, University of Massachusetts, Amherst. Chomsky, Noam. 2001. Derivation by phase. In Michael J. Kenstowicz (ed.), Ken Hale: A Life in Language, 1–52. Cambridge, MA: MIT Press. Chomsky, Noam. 2006. On phases. In Robert Freidin, David Michaels, Carlos P. Otero & Maria Luisa Zubizaretta (eds.), Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud, 133–166. Cambridge, MA: MIT Press. Christensen, Ken Ramshøj. 2007. The infinitive marker across Scandinavian. Nordlyd 34: 147– 165. Cinque, Guglielmo. 2006. Restructuring and Functional Heads. Oxford: Oxford University Press. Costantini, Francesco. 2006. Obviation in subjunctive argument clauses and the first-personal interpretation. In Mara Frascarelli (ed.), Phases of Interpretation, 295–320. Berlin: de Gruyter. Coupé, Griet. 2015. Syntactic Extension. The Historical Development of Dutch Verb Clusters. PhD thesis, Radboud Universiteit Nijmegen. Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press. Cuyckens, Hubert, Frauke D’hoedt & Benedikt Szmrecsanyi. 2014. Variability in verb complementation in Late Modern English: Finite vs. non-finite patterns. In Marianne Hundt (ed.), Late Modern English Syntax, 182–203. Cambridge: Cambridge University Press. Cuyckens, Hubert & Frauke D’hoedt. 2015. Variability in clausal verb complementation: The case of admit. In Mikko Höglund, Paul Rickman, Juhani Rudanko & Jukka Havu (eds.), Perspectives on Complementation. Structure, Variation and Boundaries, 77–100. London: Palgrave Macmillan. Dayley, Jon P. 1985. Tzutujil Grammar. Berkeley CA: University of California Press. Demske, Ulrike. 2001. Zur Distribution von Infinitivkomplementen im Althochdeutschen. In Reimar Müller & Marga Reis (eds.), Modalität und Modalverben im Deutschen, 61–86. Hamburg: Buske. Demske, Ulrike. 2011. Finale Marker in Infinitivkonstruktionen – Wandel und Variation. In Helen Christen, Franz Patocka und Evelyn Ziegler (eds.), Struktur, Gebrauch und Wahrnehmung von Dialekt, 20–46. Wien: Präsens. Demske, Ulrike. 2015. Towards coherent infinitival patterns in the history of German. Journal of Historical Linguistics 5: 6–40. Demske-Neumann, Ulrike. 1994. Modales Passiv und ‘Though Movement’ – Zur strukturellen Kausalität eines syntaktischen Wandels im Deutschen und Englischen. Tübingen: Niemeyer. Denecke, Arthur. 1880. Der Gebrauch des Infinitives bei den althochdeutschen Übersetzern des 8. und 9. Jahrhunderts. Leipzig: Pöschel & Trepte. Deppermann, Arnulf. 2007. Grammatik und Semantik aus gesprächsanalytischer Sicht. Berlin: de Gruyter. De Smet, Hendrik. 2013. Spreading Patterns: Diffusional Change in the English System of Complementation. Oxford: Oxford University Press. Dixon, Robert M. W. & Alexandra Y. Aikhenvald. 2009. The Semantics of Clause Linking. A Cross-Linguistic Typology. Oxford: Oxford University Press. Dobrushina, Nina. 2012. Subjunctive complement clauses in Russian. Russian Linguistics 36: 121–156.

Infinitival patterns and their diachronic dynamics: Questions and challenges

25

Dziwirek, Katarzyna. 2000. Why Polish doesn’t like infinitives. Journal of Slavic Linguistics 8: 57–82. Ebert, Robert P. 1976. Infinitival Complement Constructions in Early New High German. Tübingen: Niemeyer. Evers, Arnold. 1975. The Transformational Cycle in Dutch and German. PhD thesis, Universiteit Utrecht. Faarlund, Jan Terje. 2007. Parameterization and change in non-finite complementation. Diachronica 24: 57–80. Fanego, Teresa. 1996. The development of gerunds as objects of subject-control verbs in English (1400–1700). Diachronica 13: 29–62. Fanego, Teresa. 1998. Developments in argument linking in Early Modern English gerund phrases. English Language and Linguistics 2: 87–119. Fischer, Olga. 1995. The distinction between to and bare infinitival complements in late Middle English. Diachronica 12: 1–30. Fischer, Olga, Ans van Kemenade, Willem Koopman & Wim van der Wurff. 2000. The Syntax of Early English. Cambridge: Cambridge University Press. Fries, Norbert. 1983. Syntaktische und semantische Studien zum frei verwendeten Infinitiv. Tübingen: Narr. Gärtner, Hans-Martin. 2009. More on the indefinite-interrogative affinity: The view from embedded non-finite interrogatives. Linguistic Typology 13: 1–37. Gärtner, Hans-Martin. 2013. Infinite Hauptsatzstrukturen. In Jörg Meibauer, Markus Steinbach & Hans Altmann (Hgg.), Satztypen des Deutschen, 202–231. Berlin: de Gruyter. Gast, Volker & Holger Diessel. 2012. Clause Linkage in Cross-Linguistic Perspective. DataDriven Approaches to Cross-Clausal Syntax. Berlin: de Gruyter. Geisler, Christer. 1995. Relative Infinitives in English. PhD thesis, Uppsala Universitet. Glaser, Elvira. 2002. Fein gehackte Pinienkerne zugeben! Zum Infinitiv in Kochrezepten. In David Restle & Dietmar Zaefferer (eds.), Sounds and Systems. Studies in Structure and Change, 165–183. Berlin: de Gruyter. Grano, Thomas. 2015. Control and Restructuring. Oxford: Oxford University Press. Gretsch, Petra. 2008. Functions of finiteness in child language. In Veerle van Geenhoven (ed.), Semantics in Acquisition, 273–302. Grohmann, Kleanthes. 2000. Null modals in Germanic (and Romance): Infinitival exclamatives. Belgian Journal of Linguistics 14: 43–61. Grosse, Julia. 2005. Zu Kohärenz und Kontrolle in infiniten Konstruktionen des Deutschen. Marburg: Tectum Verlag. Haider, Hubert. 2010. The Syntax of German. Cambridge: Cambridge University Press. Haegeman, Liliane. 1995. Root infinitives, tense, and truncated structures in Dutch. Language Acquisition 4: 205–255. Haegeman, Liliane. 2010. Evidential mood, restructuring, and the distribution of functional sembrare. In Paola Benincà & Nicola Munaro (eds.), Mapping the Left Periphery: The Cartography of Syntactic Structures, Volume 5, 297–326. Oxford: Oxford University Press. Haiman, John & Sandra A. Thompson. 1988. Clause Combining in Grammar and Discourse. Amsterdam: John Benjamins. Haspelmath, Martin. 1989. From purposive to infinitive – a universal path of grammaticalization. Folia Linguistica Historica 10: 287–310. Höhle, Tilman. 1978. Lexikalistische Syntax. Die Aktiv-Passiv-Relation und andere Infinitivkonstruktionen im Deutschen. Tübingen: Niemeyer.

26

Łukasz Jędrzejowski and Ulrike Demske

Hoekstra, Jarich. 1997. The Syntax of Infinitives in Frisian. Ljouwert: Fryske Akademy. IJbema, Aniek. 2002. Grammaticalization and Infinitival Complements in Dutch. PhD thesis, Universiteit Leiden. Iyeiri, Yoko. 2010. Verbs of Implicit Negation and their Complements in the History of English. Amsterdam: John Benjamins. Johnk, Linn Dale. 1979. Complementation in Old High German. PhD thesis, The University of Texas at Austin. Joseph, Brian. 1983. The Synchrony and Diachrony of the Balkan Infinitive: A Study in Areal, General and Historical Linguistics. Cambridge: Cambridge University Press. Keinästö, Kari. 1986. Studien zu Infinitivkonstruktionen im mittelhochdeutschen Prosa-Lancelot. Frankfurt am Main: Peter Lang. Kempchinsky, Paula. 2009. What can the subjunctive disjoint reference effect tell us about the subjunctive? Lingua 119: 1788–1810. Kiss, Tibor. 1995. Infinitive Komplementation. Neue Studien zum deutschen Verbum infinitum. Tübingen: Niemeyer. Knyazev, Mikhail. 2016. Licensing Clausal Complements. The Case of Russian čto-Clauses. PhD thesis, Universiteit Utrecht. Landau, Idan. 2013. Control in Generative Grammar. A Research Companion. Cambridge: Cambridge University Press. Lasser, Ingeborg. 2002. The roots of root infinitives: Remarks on infinitival mail clauses in adult and child language. Linguistics 40: 767–796. Lee-Schoenfeld, Vera. 2007. Beyond Coherence. The Syntax of Opacity in German. Amsterdam: John Benjamins. Leys, Odo. 1985. Zur Semantik nicht-satzwertiger Partikelinfinitive im Deutschen und Niederländischen. Leuvense Bijdragen 74: 433–456. Leys, Odo. 1991. Skizze einer kognitiv-semantischen Typologie der deutschen um-Infinitive. Leuvense Bijdragen 80: 167–203. Los, Bettelou. 2005. The Rise of the To-Infinitive. Oxford: Oxford University Press. Maché, Jakob & Werner Abraham. 2011. Infinitivkomplemente im Frühneuhochdeutschen – satzwertig oder nicht? In Oskar Reichmann & Anja Lobenstein-Reichmann (eds.), Frühneuhochdeutsch – Aufgaben und Probleme seiner linguistischen Beschreibung, 235–274. Hildesheim: Olms. MacRobert, Catherine Mary. 1980. The Decline of the Infinitive in Bulgarian. PhD thesis, University of Oxford. Mair, Christian. 1990. Infinitival Complement Clauses in English. A Study of Syntax in Discourse. Cambridge: Cambridge University Press. Martin, Itziar San. 2007. Beyond the infinitive vs. subjunctive rivalry: Surviving changes in mood. In Luis Eguren & Olga Fernández-Soriano (eds.), Conference, Modality, and Focus. Studies on the Syntax-Semantics Interface, 171–190. Amsterdam: John Benjamins. Martin, Roger. 1996. A Minimalist Theory of PRO and Control. PhD thesis, University of Connecticut. Mensching, Guido. 2000. Infinitive Constructions with Specified Subjects. A Syntactic Analysis of the Romance Languages. Oxford: Oxford University Press. Moulton, Keir. 2009. Natural Selection and the Syntax of Clausal Complementation. PhD thesis, University of Massachusetts, Amherst. Monsterberg-Münckenau, Sylvius von. 1885. Der Infinitiv in den Epen Hartmanns von Aue. Breslau: Verlag von Wilhelm Koebner.

Infinitival patterns and their diachronic dynamics: Questions and challenges

27

Murasugi, Keiko, Chisato Fuji & Tomoko Hashimoto. 2010. What’s acquired later in an agglutinative language. Nanzan Linguistics 6: 47–78. Noonan, Michael. 2007 [1985]. Complementation. In Timothy Shopen (ed.), Language Typology and Syntactic Description. Volume 2: Complex Constructions, 52–150. Cambridge: Cambridge University Press. Pearce, Elizabeth. 1990. Parameters in Old French Syntax. Infinitival Complements. Dordrecht: Kluwer. Philippaki-Warburton, Irene & Vassilios Spyropoulos. 2004. A change of mood: The development of the Greek mood system. Linguistics 42: 791–817. Pires, Acrisio. 2006. The Minimalist Syntax of Defective Domains. Gerunds and Infinitives. Amsterdam: John Benjamins. Reimann, Michael. 1994. Studien über den Verfall des Infinitivs im Mittelbulgarischen. Universität Marburg. Reis, Marga. 2003. On the form and interpretation of German wh-infinitives. Journal of Germanic Linguistics 15: 155–201. Rizzi, Luigi. 1993/94. Some notes on linguistic theory and language development: The case of root infinitives. Language Acquistion 3: 371–393. Rizzi, Luigi. 1997. The fine structure oft he left periphery. In Liliane Haegeman (ed.), Elements of Grammar, 281–335. Kluwer: Dordrecht. Robbers, Karin Barbera Maria. 1997. Non-finite Verbal Complements in Afrikaans. A Comparative Approach. The Hague. Rochette, Anne. 1988. Semantic and Syntactic Aspects of Romance Sentential Complementation. PhD thesis, MIT. Rohdenburg, Günter. 1995. On the replacement of finite complement clauses by infinitives in English. English Studies 76: 367–388. Rohdenburg, Günter. 2006. The role of functional constraints in the evolution of the English complementation system. In Christiane Dalton-Puffer, Dieter Kastovsky, Nikolaus Ritt & Herbert Schendl (eds.), Syntax, Style and Grammatical Norms: English from 1500–2000, 143–166. Frankfurt am Main: Peter Lang. Rohdenburg, Günter. 2014. On the changing status of that-clauses. In Marianne Hundt (ed.), Late Modern English Syntax, 155–181. Cambridge University Press. Rohlfs, Gerhard. 1933. Scavi Linguistici nella Magna Grecia. Halle (Saale): Hoepli. Roussou, Anna. 2009. In the mood for control. Lingua 119: 1811–1836. Rudanko, Juhani. 2000. Corpora and Complementation. Lanham, MD: University Press of America. Rudanko, Juhani. 2012. Exploring aspects of the great complement shift, with evidence from the TIME corpus and COCA. In Tertuu Nevalainenu & Elizabeth Closs Traugott (eds.), The Oxford Handbook of the History of English, 222–232. Oxford: Oxford University Press. Rutten, Jean. 1991. Infinitival Complements and Auxiliaries. PhD thesis, Universiteit van Amsterdam. Sabel, Joachim. 1996. Coherent infinitives in German, Polish and Spanish. Folia Linguistica 33: 419–440. Sabel, Joachim. 2005. Infinitivische Frage- und Relativsätze im Deutschen und in anderen europäischen Sprachen. In Jean-Francois Marillier & Claire Rozier (eds.), Der Infinitiv im Deutschen, 83–102. Tübingen: Stauffenburg. Sabel, Joachim. 2006. Impossible infinitival interrogatives and relatives. In Patrick Brandt & Eric Fuß (eds.), Form, Structure, and Grammar. A Festschrift Presented to Günther Grewendorf on Occasion of his 60th Birthday, 242–254. Berlin: Akademie Verlag.

28

Łukasz Jędrzejowski and Ulrike Demske

Sabel, Joachim. 2015. The emergence of the infinitival left periphery. In Ulrike Steindl, Thomas Borer, Huilin Fang, Alfredo García Pardo, Peter Guekguezian, Brian Hsu, Charlie O’Hara & Iris Chuoying Ouyang (eds.), Proceedings of the 32nd West Coast Conference on Formal Linguistics, 313–322. Somerville, MA: Cascadilla Proceedings Project. Salzmann, Martin. 2013. Rule ordering in verb cluster formation. On the extraposition paradox and the placement of the infinitival particle te/zu. Linguistische Arbeitsberichte 90: 65– 121. Schlotthauer, Susan, Gisela Zifonun & Ruxandra Cosma. 2014. Verbale und nominale Infinitive – Strukturelle Eigenschaften und Funktion als Subjekt. In Ruxandra Cosma, Stefan Engelberg, Susan Schlotthauer, Speranța Stănescu & Gisela Zifonun (eds.), Komplexe Argumentstrukturen. Kontrastive Untersuchungen zum Deutschen, Rumänischen und Englischen, 253– 282. Berlin: de Gruyter. Schmidtke-Bode, Karsten. 2009. A Typology of Purpose Clauses. Amsterdam: John Benjamins. Schulte, Kim. 2007. Prepositional Infinitives in Romance. A Usage-based Approach to Syntactic Change. Frankfurt am Main: Peter Lang. Sevdali, Christina. 2013. Ancient Greek infinitives and phases. Syntax 16: 324–361. Słodowicz, Szymon. 2008. Control in Polish Complement Clauses. München: Sagner. Smirnova, Elena. 2011. Zur diachronen Entwicklung deutscher Komplementsatz-Konstruktionen. In Alexander Lasch & Alexander Ziem (eds.), Konstruktionsgrammatik III: Aktuelle Fragen und Lösungsansätze, 77–94. Tübingen: Stauffenburg. Sugiura, Wataru, Tetsuya Sano & Hiroyuki Shimada. 2016. Are there root infinitive analogues in child Japanese? In Laurel Perkins, Rachel Dudley, Juliana Gerard & Kasia Hitczenko (eds.), Proceedings oft he 6th Conference on Generative Approaches to Language Acquisition North America (GALANA 2015), 131–139. Somerville, MA: Cascadilla Proceedings Project. Ter Beek, Janneke. 2008. Restructuring and Infinitival Complements in Dutch. PhD thesis, Rijksuniversiteit Groningen. Thorley, John. 1989. Aktionsart in New Testament Greek: Infinitive and imperative. Novum Testamentum 31: 290–315. Warner, Anthony. 1982. Complementation in Middle English and the Methodology of Historical Syntax. University Park: Pennsylvania State University Press. Wiemer, Björn & Alexander Letuchiy. to appear. Clausal complementation. A status report. Introduction to a special issue of Linguistics. Wilder, Chris. 1988. On the German infinitival marker zu and the analysis of raising constructions. Lingua 76: 115–175. Wilder, Chris. 2013. Why don’t you be an imperative? Talk and handout delivered at the Universität Potsdam, May 14. Wurmbrand, Susanne. 2001. Infinitives: Restructuring and Clause Structure. Berlin: de Gruyter. Zinn, Marina, A. 2008. Infinitive constructions in Ket. In Edward J. Vajda (ed.), Subordination and Coordination Strategies in North Asian Languages, 203–214. Amsterdam: John Benjamins.

I AcI verbs and restructuring effects

Thomas Grano

2 Restructuring at the syntax-semantics interface 1 Introduction A central topic in the analysis of infinitival constructions concerns the treatment of “restructuring” whereby some matrix predicates but not others give rise to biclausal structures that behave monoclausally with respect to particular syntactic tests (see also Jędrzejowski & Demske this volume).1 Here I illustrate the basic phenomenon with two familiar examples. The first is clitic climbing, as found in Italian (Rizzi 1978; Napoli 1981; Burzio 1986; Kayne 1989; Cardinaletti & Shlonsky 2004; Cinque 2006), Romanian (Nedelcu & Paraschiv this volume), Spanish (Aissen & Perlmutter 1976), or Wolof (Torrence 2013). In this phenomenon, certain infinitival constructions allow for a pronominal clitic to be placed in the matrix clause even though it stands for an argument in the complement clause. This is significant because clitic placement is usually clausebound. In the Italian examples in (1), the clitic is found in the “expected” clause-local position. In (2), by contrast, we see that (1a) (the infinitival configuration introduced by cominciare ‘begin’) has a clitic climbing variant wherein the clitic is placed in the matrix clause, whereas (1b) (the infinitival configuration introduced by detestare ‘hate’) has no such variant. (1)

a.

Gianni Gianni

cominciava was.beginning

a to

veder[-]lo. see-it

‘Gianni was beginning to see it.’ b.

Gianni Gianni

detestava hated

veder[-]lo. see-it.

‘Gianni hated to see it.’ 1 Like many terms in syntax, “restructuring” ambiguously refers both to a phenomenon and to an analysis of that phenomenon. (In the case of “restructuring”, the analysis that it refers to is the one associated with Rizzi 1978.) Throughout this paper, I use “restructuring” strictly in the former sense as a descriptive term for any kind of monoclausality or transparency effect, independent of any particular analysis. Similar terms in the literature with partially overlapping usage include “clause union” (Aissen & Perlmutter 1976) and, in German studies, “coherence” (Bech 1955). (See also the penultimate paragraph of section 1 for relevant discussion.) Thomas Grano, Indiana University DOI 10.1515/9783110520583-002

32 (2)

Thomas Grano

a.

Gianni Gianni

lo it

cominciava was.beginning

a to

vedere. see

‘Gianni was beginning to see it.’ b. *Gianni Gianni

lo it

detestava hated

vedere. see.

Intended: ‘Gianni hated to see it.’ The second example comes from German. In the so-called long passive construction (Höhle 1978; Bayer and Kornfilt 1990; Kiss 1995; Wöllstein-Leisten 2001; Wurmbrand 2001; Sabel 2002; Schmid, Bader, and Bayer 2005; Lee-Schoenfeld 2007), the direct object of an infinitival complement clause can sometimes be promoted to matrix subject position via passivization, in an apparent exception the ordinary clause-boundedness of passivization. (3) (taken from Wurmbrand 2001) shows that infinitival constructions introduced by versuchen ‘try’ allow for long passivization but ones introduced by behaupten ‘claim’ do not. (3)

a.

weil because

der the

Wagen1 car

[t1

zu to

reparieren] repair

versucht tried

wurde was

‘because the car was tried to repair’ ≈ ‘because they tried to repair the car’ b. *weil because

der the

Wagen1 car

[t1

zu to

reparieren] repair

behauptet claimed

wurde was

Intended: ‘because the car was claimed to repair’ The point of departure for this paper is a tension noted by Wurmbrand (2001: 6–9). On the one hand, there is a cross-linguistically stable set of “core” trigger verbs for restructuring. In his seminal study of restructuring in Italian, Rizzi (1978) generalized that modal verbs, aspectual verbs, and verbs of motion tend to allow restructuring. Wurmbrand (2001) has further refined and substantiated this generalization based on evidence from Italian, Spanish, German, Dutch, and Japanese. But on the other hand, there are also cross-linguistic idiosyncrasies (verbs that trigger restructuring in some languages but not others) and languageinternal variation (verbs for which some speakers allow restructuring and others do not). Wurmbrand (2001) furthermore observes that in response to this tension, some approaches to restructuring take the stability as central and assign the irregularities a special status (see, among many others, Napoli 1981; Rochette 1988, 1990; Rosen 1989, 1991; Cinque 2006), whereas other approaches take the set of trigger verbs to be idiosyncratic and language-specific and treat cross-

Restructuring at the syntax-semantics interface

33

linguistic regularities as accidental (see, among many others, Aissen & Perlmutter 1976; Kayne 1989; Roberts 1997). Against this backdrop, this paper has two main goals. The first goal is to review Cinque’s (2006) approach to restructuring, together with modifications that I have argued for in earlier work (Grano 2012, 2015), and show how this approach gives us a handle on explaining the observed cross-linguistic regularities in the set of trigger verbs for restructuring. The second goal is to suggest that embedding this approach to restructuring into a diachronic perspective gives us a useful framework for investigating cross-linguistic and language-internal idiosyncrasies in the set of trigger verbs. In a nutshell, the central idea is that a verb restructures by semantically matching and therefore realizing an inflectional head in a highly articulated Cinque-style hierarchy of functional projections. Restructuring configurations are therefore monoclausal, the “embedded verb” actually constituting the main predicate in the structure and the “matrix verb” actually occupying a functional position in the extended projection of that main predicate. On this approach, regularity in the set of restructuring verbs is a consequence of the universality of the projections in the inflectional layer of the clause. But, I suggest, whether or not a verb semantically matches a given inflectional head depends in part on a diachronic process of semantic bleaching whereby lexical meaning is stripped away until the verb approximates the functional meaning associated with some inflectional head. On this view, the observed variation is not actually idiosyncratic but rather is systematic in the sense that it reflects different degrees of semantic bleaching. And indeed, cases of lexical verbs being grammaticalized into functional heads are well documented in the literature: see e.g. Traugott (1989, 1997) on English, Roberts & Roussou (2003) on Romance languages, and van Gelderen (2004) on Germanic languages. One disclaimer on the scope of this paper is in order: An issue that I will not address here has to do with the extent to which more than one species of monoclausality or transparency phenomenon needs to be recognized. Wurmbrand (2001), for example, argues for a three-way distinction internal to German between lexical restructuring, functional restructuring, and reduced non-restructuring. On the cross-linguistic front, Haider (2010: 351–353) argues for a distinction between German coherence (or what Haider calls “clustering”) and Italian restructuring (cf. also Haider 2003). The position I take in this paper is that the cross-linguistic regularity in the way verb meaning predicts the availability of transparency points to a core phenomenon that is universal and unitary and that makes a basic cut, though this does not exclude the possibility for finer grained syntactic distinctions both cross-linguistically and internally to some languages as well. The organization of the rest of the paper is as follows. Section 2 reviews the basic facts that need to be accounted for, drawing on Wurmbrand’s (2001)

34

Thomas Grano

survey. Section 3 reviews Cinque’s (2006) approach to restructuring. Section 4 reviews Grano’s (2012; 2015) proposed modifications to Cinque’s view and shows how this approach captures the observed cross-linguistic regularities. Section 5 suggests how the approach can be embedded into a diachronic perspective in a way that helps us understand the irregularities. Finally, section 6 concludes.

2 Cross-linguistic (in)stability in restructuring Table 1 presents an abridged and adapted version of Wurmbrand’s (2001) survey of restructuring in German, Dutch, Italian, Spanish, and Japanese. The PROPOSITIONAL and FACTIVE categories appear as they do in Wurmbrand’s work. The MODAL, ASPECTUAL and MOTION categories are generalizations I have made based on Wurmbrand’s listing of individual verbs. Many other verbs from Wurmbrand’s survey are omitted from this table, though see Table 2 in section 5 below for a larger sampling of those verbs that exhibit variation in restructuring status. Following Wurmbrand’s conventions, “+” indicates that the relevant predicate participates in restructuring configurations in the language in question, “–” indicates that it does not, and “±” indicates either inter-speaker variation or cases where the predicate has more than one translation and these translations vary from one another with respect to restructuring status. Restructuring diagnostics employed by Wurmbrand in making these classifications are long passives (German), verb raising and the infinitivus pro participio effect2 (Dutch), clitic climbing (Italian and Spanish), and lack of embedded tense marking (Japanese). Although the diagnostics vary from one language to the next, the unifying property is that all of these phenomena can be understood as reflecting monoclausal behavior in a superficially biclausal structure. The first few lines in the table provide cross-linguistic substantiation for Rizzi’s (1978) generalization based on Italian that modal, aspectual and motion verbs trigger restructuring, and to this group we can add want. At the bottom of the table, we see two classes of verbs that are cross-linguistically stable in failing to restructure. The first class is propositional verbs, which, following Landau (2000), are verbs like claim or believe whose complements can be the target of truth- or falsity-ascriptions, as in John claimed to be tall, which is true. (cf. John wanted to be tall, #which is true.) The other class is factive verbs: Verbs like regret whose complements are presupposed to be true (Kiparsky & Kiparsky 2 The infinitivus pro participio effect is a phenomenon wherein a restructuring verb shows up as an infinitive in a syntactic environment where a participial form would be expected. See Wurmbrand (2001) and references therein for more details.

35

Restructuring at the syntax-semantics interface

Table 1: (Non-)restructuring verbs, adapted from Wurmbrand (2001: 342) Predicates

German

Dutch

Italian

Spanish

Japanese

MODAL ASPECTUAL MOTION want try manage/succeed promise, threaten allow, permit decide, choose plan PROPOSITIONAL FACTIVE

+ + + + + + + + − − − −

+ + + + + + − − − − − −

+ + + + ± ± − − − − − −

+ + + + ± ± − ± − − − −

+ + + + ± + − − − − − −

1970). Finally, in the middle of the table, we see that some predicates exhibit both cross-linguistic and language-internal instability in their restructuring status. In the following two sections, in reviewing Cinque’s and Grano’s approach to restructuring, I take the stable classes at the top and bottom of the table as central, and then in section 5, I come back to the unstable classes in the middle of the table.3

3 Restructuring and clausal architecture On the basis of cross-linguistic regularity in the ordering of verbal inflectional affixes, clausal functional heads, and semantically corresponding adverbs, Cinque (1999) proposes a universal and rigidly ordered hierarchy of inflectional-layer functional heads, an abridged version of which is provided in (4). (Cf. also Hill this volume for a cartographic approach to infinitives in Early Modern Romanian.) (4)

Moodspeech act > Moodevaluative > Moodevidential > Modepistemic > Tense > Modvolitional > Aspterminative > Aspcontinuative > Aspprospective > Aspinceptive > Modobligation > Modability > Aspfrustrative > Aspsuccess > Modpermission > Aspconative > Aspcompletive (abridged and synthesized from Cinque 1999, 2006)

3 Sabel (1996) provides additional support for the general cross-linguistic stability of trigger verbs for restructuring, showing that cross-clausal scrambling in German and Polish and clitic climbing in Spanish occur with similar kinds of predicates in all three languages. I thank an anonymous reviewer for drawing my attention to this work.

36

Thomas Grano

On this view, adverb ordering effects like that seen in (5) follow from the fact that adverbs are licensed in the specifier positions of particular functional heads. For example, unfortunately has an evaluative meaning, and so it is licensed by Moodevaluative , while probably has an epistemic meaning, and so it is licensed by Modepistemic . Since these functional heads occur in a fixed order – as evidenced by the morphology of languages like Korean (6) in which evaluative and epistemic meanings are realized by functional morphemes – the adverbs occur in a fixed order as well. (5)

a.

John unfortunately probably left.

b. *John probably unfortunately left. (6)

ku that

say-ka bird-NOM

cwuk-ess-keyss-kwun-a. die-ANT-EPISTEM-EVAL-DECL

‘That bird must have died!’ (Cinque 1999: 53) In later work, Cinque (2006) extends this line of analysis to restructuring verbs, arguing that they also are overt instantiations of functional heads in the inflectional layer of the clause. For example, try realizes a particular kind of aspectual head (namely, CONATIVE aspect, realized as a verbal affix in some languages: see e.g. MacDonald 1990: 304 for an example from the Papua New Guinean language Tauya). Accordingly, a sentence like (7) would have the structure in (8), wherein try realizes an aspectual head in the extended projection of its verbal complement, the highest argument of which raises into the surface subject position, [Spec,TP].4

4 Cinque’s analysis of restructuring is based mostly on data from Italian, though he does mention that restructuring effects are reported for many of the world’s languages, suggesting widespread cross-linguistic applicability of the proposals. Here and in what follows, I use English to exemplify the restructuring configurations. See Grano (2012, 2015) for a detailed investigation of the applicability of this approach to complementation patterns in English, Mandarin Chinese, and modern Greek. There I argue that one restructuring diagnostic for English is failure to accept finite complements (e.g., *John tried that he opened the door), which tracks the distribution of standard restructuring properties in other languages quite well. One outcome of extending the analysis to English involves giving up on the view that all functional heads are rigidly ordered, given the observation that when English restructuring predicates occur together, they can be flexibly ordered, unlike what Cinque claims for Italian (e.g., John began to want to read/John wanted to begin to read).

Restructuring at the syntax-semantics interface

(7)

37

John tried to open the door.

(8)

Cinque’s approach – which is the successor to a number of approaches that take restructuring verbs to be auxiliary-like in one way or another (Napoli 1981; Rochette 1988, 1990; Rosen 1989, 1991) – elegantly ties the restructuring verbs’ semantics to their syntax. Semantically, restructuring verbs all have meanings that signal their functional status. And the syntactic consequence of this functional status is that they are not associated with their own clausal projection; rather, they are part of the single clausal projection lexically headed by the main verb in their complement. This gives us monoclausality, which is the reason for the observed syntactic effects such as clitic climbing and long passivization. Cinque’s approach also leads us to expect widespread cross-linguistic agreement in the set of restructuring predicates: Such agreement follows from the hypothesized universality of the inventory of functional projections. One feature of Cinque’s approach that is worth highlighting is its position that restructuring predicates uniformly instantiate monoclausal structures like (8) regardless of whether overt transparency effects like clitic climbing or long passivization obtain. (See also Cardinaletti & Shlonsky 2004 for an analysis of Italian clitic climbing that works on this basis.) One piece of evidence that Cinque points to for this view is that it derives the distribution of exhaustive and partial control: Wurmbrand (1998) generalized that restructuring predicates require that there be total identity between the controller and the controlled position (EXHAUSTIVE CONTROL in the sense of Landau 2000), whereas nonrestructuring predicates allow the controller to denote a proper subset of the controlled position (PARTIAL CONTROL in the sense of Landau 2000). For example, as seen in (9), hope allows partial control, as evidenced by the compatibility between a singular non-group denoting controller ( John) and an embedded predicate (gather) that requires a collective subject. By contrast, as seen in (10),

38

Thomas Grano

inserting try into the same syntactic configuration yields unacceptability, indicating that try disallows partial control. (9)

John hoped to gather at noon.

(10) *John tried to gather at noon. Since Cinque analyzes restructuring as involving a movement relationship between the higher subject position and the lower subject position, total identity is trivially expected. Partial control, in turn, becomes a fully general property of PRO. Crucially, the cut between exhaustive control and partial control remains stable regardless of whether transparency effects obtain. This stability supports a uniformly monoclausal syntax for restructuring verbs. For more on the relationship between restructuring and the exhaustive/partial control distinction, see Wurmbrand (2002); Barrie (2004); Barrie & Pittman (2004); Costantini (2010); Grano (2012, 2015); White & Grano (2014).

4 Subject orientation and Tense Although Cinque’s approach to restructuring has much to recommend it, it also raises a number of pressing questions, two of which guide the investigation in Grano (2012, 2015). The first concern for Cinque’s approach is that some restructuring predicates such as try pass all standard diagnostics for control, and yet Cinque’s analysis of restructuring entails a raising syntax, as reviewed in the previous section. The second concern has to do with the predictive power of Cinque’s approach. Ideally, we would want to be able to predict the restructuring status of a verb from its semantics: We would look at the verb’s meaning, look at the hierarchy of inflectional projections, and predict that the verb restructures (i.e., is functional, on Cinque’s approach) if and only if there is a functional head that corresponds to its meaning. But against this expectation, we find, for example, that verbs of speech like say resist restructuring, despite the existence of Moodspeech act which would seem a natural home for a verb with this meaning. In what follows, I review Grano’s (2012; 2015) solution to these two concerns.

4.1 Subject orientation as variable binding Traditional diagnostics for distinguishing raising predicates from control predicates are all designed to test whether the predicate in question enters into a thematic dependency with its subject: If it does, the predicate instantiates

Restructuring at the syntax-semantics interface

39

control, and if it does not, the predicate instantiates raising. A raising predicate like seem admits expletive subjects (11a), idiom chunk subjects (11b), and inanimate subjects (11c), and furthermore exhibits synonymy when active/passive alternations in the embedded clause affect which argument is in matrix subject position (11d). All of these tests converge on the view that seem bears no thematic dependency with its subject. A control predicate like try, on the other hand, fails on all these counts (12a–d). Following earlier work, I use the term SUBJECT ORIENTATION to refer in an analytically neutral way to this property of predicates like try. (11)

(12)

a.

It seemed to be important that we go.

b.

The cat seemed to be let out of the bag.

c.

The rock seemed to roll down the hill.

d.

The doctor seemed to examine the patient. = The patient seemed to be examined by the doctor.

a. *It tried to be important that we go. b. *The cat tried to be let out of the bag. c. *The rock tried to roll the down hill. d. The doctor tried to examine the patient. ≠ The patient tried to be examined by the doctor.

In order to reconcile the control behavior of restructuring predicates like try with the raising syntax assigned to them by Cinque (2006), Grano (2012, 2015) proposes a solution based on two previous ideas in the literature. The first idea is that although modal expressions are essentially one-place predicates in that they take a single proposition-denoting argument, modals are nonetheless “anchored” to an individual that they require semantic access to, and one way of modeling anchoring is via variable binding (Hacquard 2010; cf. also Farkas 1992; Giannakidou 1998, 1999; Kratzer 2011). In a sentence like (13a), for example, have expresses obligation, and the most natural construal is one in which it is specifically the subject (John) who bears that obligation. But as seen in (13b), the obligation-bearer in a deontic have sentence need not always be in subject position or anywhere in the sentence at all (Bhatt 1998; Wurmbrand 1999). We can assign a uniform syntax to both kinds of sentences via the proposal that have is a raising predicate that comes along with a variable that can be bound either by an entity in the structure or by an entity salient from context.

40 (13)

Thomas Grano

a.

John has to put fifty chairs in the living room.

b.

There have to be fifty chairs in the living room.

The second idea is that some variables in natural language are DEPENDENT in the sense of Giannakidou (1998): Like reflexive pronouns, they cannot get their value from the context. Putting these two ideas together, the proposal is that restructuring predicates like try incorporate into their meaning a dependent individual variable. Consequently, when the subject raises, that subject obligatorily binds the variable, giving the predicate semantic access to the subject and simulating a control relation. This mechanism is exemplified in (14), which assumes a denotation for try represented schematically in (15). Logically, try takes a propositional argument and an individual argument. But whereas the propositional argument composes with the predicate via functional application, the individual argument enters in via variable binding. (See Heim and Kratzer 1998 for the mechanics of functional application and variable binding, as well as an explication of the general approach to the syntax-semantics interface that I assume here.) (14)

a.

John tried to open the door.

b.

(15)

[[try(x)]] = λp.TRY(x)(p)

(cf. Sharvit 2003; Grano 2011)

4.2 Failed restructuring as failed variable binding If Cinque’s (2006) approach to restructuring is to be successful in predicting whether a given predicate permits restructuring, then it ought to be the case

Restructuring at the syntax-semantics interface

41

that if a predicate has a meaning that is independently known to reside on a functional head somewhere in the inflectional layer of the clause, then it restructures. But as Grano (2012; 2015) points out, this prediction fails in a systematic way. In particular, predicates that have meanings that correspond to inflectionallayer functional heads above Tense in Cinque’s hierarchy do not restructure. There are four heads above Tense in Cinque’s hierarchy: Moodspeech act , Moodevaluative , Moodevidential , and Modepistemic . And yet the most plausible candidates for predicates that semantically match these categories do not restructure. These are, respectively, verbs of speech like say or ask, evaluative or emotive factive verbs like regret, verbs that name sources of evidence such as see (that) or hear (that), and epistemic verbs like believe or know. All four of these predicate types correspond to what Wurmbrand (2001) calls PROPOSITIONAL and FACTIVE in her survey and they fail to restructure in a cross-linguistically robust way. Grano goes on to show that the failure of restructuring above Tense is itself subject to an exception: the Italian verb sembrare ‘seem’ participates in restructuring configurations for some speakers of Italian. This is illustrated with the clitic climbing example in (16). Haegeman (2005, 2006, 2010) argues that for speakers who accept sentences like (16), sembrare realizes Moodevidential . The main source of evidence for this analysis is that even speakers who ordinarily allow restructuring with sembrare do not allow it when the verb is found in the antecedent to a conditional (17a) or in the complement to a factive predicate (17b) – precisely environments in which evidential mood is independently known to be ruled out. (16)

Lo sembrano trovare troppo difficile. ‘They seem to find it too difficult.’ (Haegeman 2010: 302)

(17)

a. *Se lo sembrano trovare troppo difficile, faremo il secondo capitolo. Intended: ‘If they seem to find it too difficult, we’ll do the second chapter.’ b.

??È strano che lo sembrino trovar troppo difficile. Intended: ‘It is odd that they seem to find it too difficult.’ (Haegeman 2010: 306)

Because Moodevidential is above Tense in Cinque’s hierarchy, sembrare presents a counterexample to the generalization that predicates corresponding to heads above Tense fail to restructure. Grano hypothesizes, however, that the property of sembrare responsible for this exceptionality is that it is non-subject-oriented;

42

Thomas Grano

i.e., it is a raising predicate in the traditional sense of not entering into a thematic dependency with its subject. Grano thereby arrives at the descriptive generalization in (18). (18) Restructuring generalization: A verb V restructures just in case V matches the meaning of an inflectional-layer functional head F, and either F is below Tense or V is non-subject-oriented. (adapted from Grano 2012: 112) Finally, Grano shows that the restructuring generalization in (18) in fact follows from the variable binding approach to subject orientation, given the reasonable assumption that subjects can be interpreted no higher than [Spec,TP] (von Fintel & Iatridou 2003). To see this, there are four cases to consider. First, take a subject-oriented predicate that corresponds to a head below Tense, such as try. As illustrated in (19), try realizes an aspectual head, and when the subject raises, it binds the dependent variable associated with try. This state of affairs stands in contrast with a subject-oriented predicate that corresponds to a head above Tense. Take for example claim, which names a kind of speech act. Since Moodspeech act is above Tense, even after the subject raises to [Spec,TP], it is too low to bind the dependent variable associated with claim. This is illustrated in (20). Consequently, restructuring with claim fails, and claim must instead instantiate a full biclausal structure whereby it realizes a lexical verb that takes a full CP complement with a PRO subject, as illustrated in (21). Subject-oriented; below Tense: (19)

x bound by subject (= John)

Restructuring at the syntax-semantics interface

43

Subject-oriented; above Tense: (20)

*

x too high to be bound! (21)

Rounding out the picture, when a predicate is not subject-oriented, then its realization above or below Tense is irrelevant to the availability of restructuring because no variable binding is at stake. This is illustrated in (22) for start (which corresponds to an aspectual head below Tense5) and in (23) for Italian sembrare (which corresponds to Moodevidential , above Tense).

5 Perlmutter (1970) argues that aspectual verbs like start are raising/control ambiguous, which in the framework here would mean that they are ambiguous between being subject-oriented (variable incorporating) and non-subject-oriented (non-variable-incorporating). Although Perlmutter’s conclusion is not indisputable (see e.g. Rochette 1999), all that matters for my purpose here is that start is at least sometimes non-subject-oriented, which as far as I know is uncontroversial. See also Fukuda (2012) for an analysis of Japanese aspectual verbs as functional heads.

44

Thomas Grano

Non-subject-oriented; below Tense: (22)

(No variable to bind) Non-subject-oriented; above Tense: (23)

(No variable to bind) In sum, Grano’s approach accurately derives the generalization that restructuring predicates corresponding to heads below Tense can always restructure whereas restructuring predicates corresponding to heads above Tense can restructure only if they are not subject-oriented.6 It is worth highlighting one feature of 6 An anonymous reviewer points out the following counterexample from German: (i) Der Zeuge will den Mörder gesehen haben. the witness wants the murderer seen have ‘The witness claims to have seen the murderer.’ As the reviewer points out, citing Reis & Sternefeld (2004), German wollen ‘want’ – in addition to its desiderative use – has a reportative or evidential use along the lines of claim, which means that it should correspond to Moodspeech act or Moodevidential , both of which are above Tense in the clause. Crucially, it still behaves like a restructuring predicate on this use, and it is also still subject-oriented in that the subject names the source of the claim. This combination of properties (corresponding to a head above Tense, subject orientation, and restructuring/ functional status) are precisely what is predicted to be unavailable on the theory that I am presenting here. I leave this as an open puzzle.

45

Restructuring at the syntax-semantics interface

Grano’s approach that differentiates it from Cinque’s: Cinque’s approach entails a “static” view of the functional status of restructuring predicates whereas Grano’s approach entails a “dynamic” view. For Cinque, restructuring predicates are specified in the lexicon as functional. For Grano, on the other hand, restructuring predicates are specified in the lexicon as lexical, and a restructuring rule allows (lexical) predicates to be merged into functional positions in the structure as long as they have the right kind of meaning – and as long as variable binding (in the case of a subject-oriented predicate) succeeds. It is the possibility of failure in the output of this restructuring rule (as in cases like (20) above) that endows the “dynamic” approach with its predictive power.

5 Diachronic perspective 5.1 Instability in restructuring status On Grano’s reformulation of Cinque’s approach to restructuring, the availability of restructuring is semantically deterministic in the sense that if a predicate’s meaning is subsumed by that of an inflectional-layer functional category which is below Tense or which is not subject-oriented, it realizes that category. This approach raises a question about predicates like those in Table 2 (which consists of a selection of predicates from Wurmbrand’s survey that partially overlaps with the predicates in Table 1 above) that evidence both cross-linguistic and language-internal instability in their restructuring status. (Following Wurmbrand’s convention, “N/A” indicates either that the predicate in question does not combine with an infinitive or that independently identifiable factors affect their ability to restructure; e.g., Wurmbrand observes that particle verbs in Dutch and German block restructuring.) Table 2: “Unstable” predicates, adapted from Wurmbrand (2001: 342) Predicates

German

Dutch

Italian

Spanish

Japanese

return intend (=want, mean) forget try manage/succeed dare seem promise, threaten allow, permit

N/A + + + + + + + +

N/A + + + + + + − −

± ± ± ± ± ± ± − −

+ ± ± ± ± − − − ±

± N/A + ± + N/A − − −

46

Thomas Grano

Toward a solution, what I would like to suggest is that whether a verb’s meaning is subsumed by a functional category depends in part on a diachronic process of semantic bleaching. This idea rests on the view that grammaticalization is primarily driven by semantic bleaching (see e.g. van Kemenade 1999 and references therein) and that the syntactic consequence of grammaticalization is the reanalysis of a lexical head as a functional head (Roberts & Roussou 1999). We then predict that for those predicates that exhibit instability, the restructuring variants are more semantically bleached than their non-restructuring counterparts. For example, since ‘seem’ restructures in German and Dutch but not in Spanish or Japanese, it should have a more bleached meaning in the former two languages. Preliminary evidence for this state of affairs is found in Cinque’s (2006) observation that instability is often found among non-“core” members of certain semantic classes: Thus Italian volere ‘want’ restructures but the semantically more nuanced (i.e., less bleached) desiderare ‘desire’, amare ‘love’ and preferire ‘prefer’ exhibit instability. The diachronic process that leads to this state of affairs plausibly fits under Ogura’s (1998) proposal that the first stage of grammaticalization is one in which “a high frequency, unmarked word is chosen from among synonyms and/ or functional equivalents and gradually idiomatized” followed by a second stage in which “the meaning of the word comes to be broadened” (p. 309).7 Even if this line of reasoning is on the right track, though, what still needs to be explained is why the instability is itself cross-linguistically stable. In other words, what is the crucial property or set of properties that distinguishes the predicates in Table 2 from modals, aspectual and motion predicates, and ‘want’ – which uniformly restructure – and from propositional and factive predicates – which uniformly fail to restructure?

7 An anonymous reviewer draws my attention to von Fintel (1995), who argues that what characterizes the semantics of functional morphemes is not “bleached” meaning but rather permutation invariance (i.e., insensitivity “to specific facts about the world” [p. 179]) and high semantic type (i.e., beyond the type of entity, situation, or predicate of entity). Correspondingly, von Fintel suggests that grammaticalization involves becoming permutation-invariant and assuming a high type. von Fintel points out that characterizing functional morphemes in terms of having a “blander” meaning is problematic because it does not allow comparison across different types of meanings; e.g., von Fintel observes that there is no clear sense in which not has a meaning that is blander than cat. (Cf. also Reis 2007.) The class of predicates under discussion here (‘want’, ‘desire’, ‘prefer’, etc.), though, all have a similar type, and all fall into von Fintel’s “intermediate category” of lexical morphemes that have a logical meaning (having a high type that involves quantification over possible worlds). I think it is conceivable that within this class, blandness of meaning could play a role in determining functional status. Evidence for this lies in the observation that ‘want’ is routinely grammaticalized as an affix in some languages but the more nuanced ‘desire’, ‘prefer’, etc., are not.

Restructuring at the syntax-semantics interface

47

One trend that emerges from Tables 1 and 2 is that implicative verbs (manage, dare, forget) and try all exhibit some degree of instability, so it may be fruitful to take these as a starting point. When they restructure, Cinque (2006) treats them as aspectual heads (see also Sharvit 2003; Grano 2011 for an aspectual semantics for try): The basic intuition behind this approach is that they contribute to the specification of how much of the event supplied by the VP has been realized: none at all ( forget), at least a preliminary stage (try), or the entire event (dare, manage). But unlike standard aspectual verbs such as begin or stop, these verbs also have another dimension of meaning studied by Karttunen (1971), conventionally implicating negligence ( forget), effort (manage), or bravery (dare), or asserting effort (try). Focusing on these predicates, the suggestion I would like to make is that it is this extra dimension of meaning that creates the partial resistance toward their realization as aspectual categories. How can we make this suggestion more concrete? A promising line of reasoning is that inflectional heads below Tense all share a semantic property that the predicates in Table 2 tend to vacillate on due to some feature of their core meaning. In particular, suppose inflectional heads below Tense uniformly fail to introduce their own eventuality description but rather function as modifiers of pre-existing eventuality descriptions: In type-theoretic terms, they are all type 〈εt, εt〉 (where ε is the type for eventualities), combining with a property of eventualities and returning a new property of eventualities. Such a view in fact follows from a particular conception of the syntactico-semantic architecture of the clause whereby Tense (possibly in conjunction with a dedicated aspectual head just below it) is responsible for converting an eventuality description (i.e., a property of eventualities) into a proposition (i.e., a property of worlds): see e.g. Hacquard (2006). This paves the way for a hypothesis that weaves together Napoli’s (1981) idea that restructuring predicates do not express their own states or actions with Hacquard’s (2013) recent suggestion that we can analyze grammaticalization of modals over time as involving the modal losing its status as a predicate of eventualities. In particular, building on Napoli, aspectual, modal and motion verbs and want all straightforwardly function as modifiers of eventuality descriptions and are hence uniformly eligible to realize the relevant functional positions in the clause (see Hacquard 2010 on modals and Hacquard 2008 on want). Turning to implicative predicates and try, what I would like to suggest is that it is the extra dimension of meaning these predicates have (‘negligence’, ‘effort’, etc.) that causes the instability. If the extra dimension is strong enough, the predicate introduces its own eventuality description and thereby precludes restructuring, but (building on Hacquard) if the extra dimension has been sufficiently diminished through bleaching, the predicate does not introduce its own eventuality description, and restructuring becomes available.

48

Thomas Grano

One piece of support for this suggestion comes from an observation about Italian sembrare ‘seem’, whose variable acceptability with clitic climbing suggests that it is in a transitional stage of grammaticalization. Cinque (2004) and Haegeman (2006) show that even among those speakers who accept clitic climbing with this verb, clitic climbing is ungrammatical in the presence of an experiencer argument, as illustrated in (24). If introducing an experiencer argument involves introducing a relation between an individual and an eventuality, then a predicate should only be able to combine with an experiencer argument if it introduces its own eventuality description. So we can make sense of the facts in (24) via the proposal that clitic climbing precludes lexical status and the presence of an experiencer argument precludes functional status, since functionalization entails the loss of the ability to support such a thematic relation. (24)

a. *Gianni Gianni

non not

ce to-us

lo it

sembra seem

apprezzare appreciate

abbstanza. enough

Intended: ‘It seems to us that Gianni does not appreciate it enough.’ b.

Gianni Gianni

non not

ci to-us

sembra seem

apprezzar[-]lo. appreciate-it

‘It seems to us that Gianni does not appreciate it.’ (Haegeman 2006:485) At this stage, the overall proposal remains a hypothesis, but it makes a prediction that can be tested. What needs to be done to test it is to develop criteria for diagnosing the extent of bleaching of the aforementioned dimensions of meaning associated with implicative verbs and ‘try’, and test whether predicates exhibit a correlation between bleaching and a given language’s or speaker’s tolerance for restructuring.8

8 Another potentially fruitful perspective on the unstable predicates is that some of them may be in a transitional stage from control predicates to raising predicates. Cornillie (2005) for example argues that Spanish prometer ‘promise’ and amenazar ‘threaten’ are both undergoing grammaticalization and concomitantly taking on a raising status. The prediction then is that those speakers for whom these verbs are sufficiently bleached should accept clitic climbing with them. de Haan (2007), focusing on the history of SEEM-verbs in Germanic, similarly suggests that raising syntax is the result of grammaticalization. Sometimes, the result of this grammaticalization is a novel aspectual category (see e.g. Tagliamonte and Lawrence 2000 on the history of the English habitual past used to), a fate awaiting try, manage, dare, and forget, if the reasoning here is on the right track.

Restructuring at the syntax-semantics interface

49

5.2 Connection to related phenomena An anonymous reviewer asks to what extent the approach taken here can account for attested cases of change in verbal meaning, specifically the development of epistemic meaning out of deontic modals in English (Traugott 1989), as well as the development of English promise and threaten from control verbs that name particular kinds of speech acts into verbs that maintain this older meaning but that can also be used as raising verbs with an epistemic and evaluative meaning (Traugott 1997). As for the former phenomenon, Hacquard (2013) suggests an account that makes use of concepts similar to those employed here. In particular, drawing on earlier work (Hacquard 2006, 2010), Hacquard identifies a correlation between clausal height and modal flavor: A modal with root meaning is structurally lower than Tense in the clause whereas a modal with epistemic meaning is structurally higher than Tense in the clause. Hacquard makes sense of this correlation by analyzing modals as event-relative. When they occur below Tense, they are bound by and hence relativized to the VP-event, which gives them root meaning, whereas when they appear above Tense, they are bound by and hence relativized to the speech event, which gives them epistemic meaning. Hacquard suggests that just as restructuring involves the conversion of an individual argument into a dependent variable as in Grano (2012), so the development of epistemic meaning may involve the conversion of an event argument into a dependent variable. In the earliest stage, the modal is a lexical predicate that introduces its own event and is consequently fixed in modal flavor. Then, the event argument gets converted into a dependent variable. As long as it remains below Tense in the clause, though, it will be relativized to the VP event and consequently have root flavor. But now, no longer being a predicate of events, it is free to merge above Tense, and its dependent event variable will be bound by the speech event, thereby yielding epistemic meaning. As for the development of threaten and promise, Traugott identifies three historical stages: In the first stage, they are lexical verbs that name illocutionary acts performed by the entity named by the subject. In the second stage, they take on a non-intentional epistemic meaning along the lines of ‘portend’, and the subject names the source of the promise or threat. Finally, in the third stage, the verbs take on a raising semantics whereby the subject no longer has to name the source of the promise or the threat. Given the existence of the second stage, Traugott concludes that “any hypothesis that there is a necessary correspondence between semantic epistemicity and syntactic verb raising [. . .] is too strong” (p. 191). Traugott goes on to suggest that there is a historical cline from being a predicate with full thematic properties, to a raising predicate, and then

50

Thomas Grano

to a more fully grammaticalized modal. Although the details of this historical path are not fully predicted by what I propose here, the theory here does predict that a morpheme with an epistemic meaning cannot simultaneously be subject-oriented and have functional status. This is consistent with the observed historical trajectory whereby loss of subject orientation precedes full-fledged grammaticalization.

6 Conclusion This paper began with an old question in the restructuring literature: Why do some predicates restructure and others do not? Wurmbrand’s (2001) cross-linguistic empirical work imposes important parameters on the shape that the answer to this question must take, by showing that predicates with certain kinds of meanings are stable in their restructuring status while predicates with other kinds of meanings are unstable. I argued that a useful way of making sense of the stability is via an approach in the spirit of Cinque (2006), where the capacity for restructuring is related to independently verifiable and possibly universal properties of clausal architecture. I also argued that the instability can be studied from the perspective of the Cinque-inspired approach once we admit a diachronic dimension to the analysis, and I offered some preliminary suggestions on how semantic bleaching may modulate the availability of restructuring. Stepping back, I hope to have made the case here that a comprehensive theory of restructuring must weave together at least five dimensions of analysis: syntax, semantics, their interface, cross-linguistic (non-)variation, and the diachronic concepts of bleaching and grammaticalization. If what I say in this paper is on the right track, then we are well on our way to having a theory that does this.

References Aissen, Judith & David Perlmutter. 1976. Clause reduction in Spanish. In Henry Thompson, Kenneth Whistler, Vicki Edge, Jeri Jaeger, Ronya Javkin, Miriam Petruck, Christopher Smeall & Robert D. Van Valin Jr. (eds.), Proceedings of the Second Annual Meeting of the Berkeley Linguistics Society, 1–30. Berkeley, CA: Berkeley Linguistics Society. Barrie, Michael. 2004. Moving toward partial control. In Keir Moulton & Matthew Wolf (eds.), Proceedings of the 34th Meeting of the North East Linguistics Society, vol. I, 133–146. Amherst, MA: GLSA.

Restructuring at the syntax-semantics interface

51

Barrie, Michael & Christine M. Pittman. 2004. Partial control and the movement towards movement. Toronto Working Papers in Linguistics 22:75–92. Bayer, Josef & Jaklin Kornfilt. 1990. Reconstruction effects in German. In Elisabet Engdahl, Mike Reape, Martin Mellor & Richard Cooper (eds.), Edinburgh Working Papers in Cognitive Science, vol. 6: Parametric Variation in Germanic and Romance, 21–42. Dordrecht: Kluwer Academic Publishers. Bech, Gunnar. 1955. Studien zum deutschen Verbum infinitum. Tübingen: Max Niemeyer Verlag. Bhatt, Rajesh. 1998. Obligation and possession. In Heidi Harley (ed.), Papers from the UPenn/ MIT Roundtable on Argument Structure and Aspect, MIT Working Papers in Linguistics (MITWPL) 32, 21–40. Cambridge, MA: MITWPL. Burzio, Luigi. 1986. Italian Syntax. Dordrecht: Reidel. Cardinaletti, Anna & Ur Shlonsky. 2004. Clitic positions and restructuring in Italian. Linguistic Inquiry 35:519–557. Cinque, Guglielmo. 1999. Adverbs and Functional Heads: A Cross-Linguistic Perspective. Oxford: Oxford University Press. Cinque, Guglielmo. 2004. ‘Restructuring’ and functional structure. In Adriana Belletti (ed.), Structures and Beyond: The Cartography of Syntactic Structures, volume 3, 132–191. New York: Oxford University Press. Cinque, Guglielmo. 2006. Restructuring and Functional Heads. Oxford: Oxford University Press. Cornillie, Bert. 2005. A paradigmatic view of Spanish amenazar ‘to threaten’ and prometer ‘to promise’. Folia Linguistica 39:385–415. Costantini, Francesco. 2010. On infinitives and floating quantification. Linguistic Inquiry 41:487– 496. Farkas, Donka. 1992. On the semantics of subjunctive complements. In Paul Hirschbühler & Konrad Koerner (eds.), Romance Languages and Modern Linguistic Theory: Selected Papers from the XX Linguistic Symposium on Romance Languages, University of Ottawa, April 10–14, 1990, 69–104. Amsterdam: Benjamins. von Fintel, Kai. 1995. The formal semantics of grammaticalization. In Jill N. Beckmann (ed.), Proceedings of NELS 25, 175–189. Amherst, MA: GLSA Publications. von Fintel, Kai & Sabine Iatridou. 2003. Epistemic containment. Linguistic Inquiry 34:173–198. Fukuda, Shin. 2012. Aspectual verbs as functional heads: Evidence from Japanese aspectual verbs. Natural Language & Linguistic Theory 30:965–1026. van Gelderen, Elly. 2004. Grammaticalization as Economy. Amsterdam: Benjamins. Giannakidou, Anastasia. 1998. Polarity Sensitivity as (Non)Veridical Dependency. Amsterdam: Benjamins. Giannakidou, Anastasia. 1999. Affective dependencies. Linguistics and Philosophy 22:367–421. Grano, Thomas. 2011. Mental action and event structure in the semantics of try. In Neil Ashton, Anca Chereches & David Lutz (eds.), Semantics and Linguistic Theory (SALT) 21, 426–443. eLanguage. Grano, Thomas. 2012. Control and Restructuring at the Syntax-Semantics Interface. Ph.D. Dissertation, University of Chicago. Grano, Thomas. 2015. Control and Restructuring. Oxford: Oxford University Press. de Haan, Ferdinand. 2007. Raising as grammaticalization: The case of Germanic SEEM-verbs. Rivista di Linguistica 19:129–150. Hacquard, Valentine. 2006. Aspects of Modality. Ph.D. Dissertation, MIT. Hacquard, Valentine. 2008. Restructuring and implicative properties of volere. In Atle Grønn (ed.), Proceedings of SuB 12, 165–179. Oslo: ILOS.

52

Thomas Grano

Hacquard, Valentine. 2010. On the event relativity of modal auxiliaries. Natural Language Semantics 18:79–114. Hacquard, Valentine. 2013. The grammatical category of modality. In Maria Aloni, Michael Franke & Floris Roelofsen (eds.), Proceedings of the 19th Amsterdam Colloquium, 19–26. Haegeman, Liliane. 2005. Functional heads, lexical heads and hybrid categories. In Hans Broekhuis, Norbert Corver, Riny Huybregts, Ursula Kleinhenz & Jan Koster (eds.), Organizing Grammar. Linguistic Studies in Honor of Henk van Riemsdijk, 152–161. Berlin: Mouton de Gruyter. Haegeman, Liliane. 2006. Clitic climbing and the dual status of sembrare. Linguistic Inquiry 37:484–501. Haegeman, Liliane. 2010. Evidential mood, restructuring, and the distribution of functional sembrare. In Paola Benincà & Nicola Munaro (eds.), Mapping the Left Periphery: The Cartography of Syntactic Structures, volume 5, 297–326. Oxford: Oxford University Press. Haider, Hubert. 2003. V-clustering and clause union – causes and effects. In Pieter Seuren & Gerard Kempen (eds.), Verb Constructions in German and Dutch, 91–126. Amsterdam: Benjamins. Haider, Hubert. 2010. The Syntax of German. Cambridge: Cambridge University Press. Heim, Irene & Angelika Kratzer. 1998. Semantics in Generative Grammar. Malden, MA: Blackwell Publishing. Höhle, Tilman. 1978. Lexikalistische Syntax: Die Aktiv-Passiv-Relation und andere Infinitkonstruktionen im Deutschen. Tübingen: Niemeyer. Karttunen, Lauri. 1971. Implicative verbs. Language 47:340–358. Kayne, Richard S. 1989. Null subjects and clitic climbing. In Osvaldo Jaeggli & Kenneth Safir (eds.), The Null Subject Parameter, 239–261. Dordrecht: Kluwer. van Kemenade, Ans. 1999. Functional categories, morphosyntactic change, grammaticalization. Linguistics 37:997–1010. Kiparsky Paul & Carol Kiparsky. 1970. Fact. In Progress in linguistics, ed. M. Bierwisch and K.E. Heidolph, 143–173. Mouton: The Hague. Kiss, Tibor. 1995. Infinitive Komplementation: Neue Studien zum deutschen Verbum infinitum. Tübingen: Niemeyer. Kratzer, Angelika. 2011. What “can” can mean. Paper presented at SALT 21, May 20–22, New Brunswick, NJ. Landau, Idan. 2000. Elements of Control: Structure and Meaning in Infinitival Constructions. Dordrecht: Kluwer. Lee-Schoenfeld, Vera. 2007. Beyond Coherence: The Syntax of Opacity in German. Amsterdam: Benjamins. MacDonald, Lorna. 1990. A Grammar of Tauya. Berlin: Mouton de Gruyter. Napoli, Donna Jo. 1981. Semantic interpretation vs. lexical governance: Clitic climbing in Italian. Language 57:841–887. Ogura, Michiko. 1998. The grammaticalization in Medieval English. In Jacek Fisiak & Marcin Krygier (eds.), Advances in English Historical Linguistics, 293–314. Berlin: Mouton de Gruyter. Perlmutter, David M. 1970. The two verbs begin. In Peter Rosenbaum & Roderick Jacobs (eds.), Readings in English Transformational Grammar, 107–119. Waltham, MA: Ginn-Blaisdell. Reis, Marga. 2007. Modals, so-called semi-modals, and grammaticalization. Interdisciplinary Journal for Germanic Linguistics and Semiotic Analysis 12:1–57. Reis, Marga & Wolfgang Sternefeld. 2004. Review article of ‘Susanne Wurmbrand: Infinitives. Restructuring and clause structure’. Linguistics 42:469–508.

Restructuring at the syntax-semantics interface

53

Rizzi, Luigi. 1978. A restructuring rule in Italian syntax. In Samuel J. Keyser (ed.), Recent Transformational Studies in European Languages, 113–158. Cambridge: MIT Press. Roberts, Ian. 1997. Restructuring, head movement, and locality. Linguistic Inquiry 28:423–460. Roberts, Ian & Anna Roussou. 1999. A formal approach to “grammaticalization”. Linguistics 37:1011–1041. Roberts, Ian & Anna Roussou. 2003. Syntactic Change: A Minimalist Approach to Grammaticalization. Cambridge: Cambridge University Press. Rochette, Anne. 1988. Semantic and Syntactic Aspects of Romance Sentential Complementation. Ph.D. Dissertation, MIT. Rochette, Anne. 1990. On the restructuring classes of verbs in Romance. In Anne-Marie Di Sciullo & Anne Rochette (eds.), Binding in Romance: Essays in Honour of Judith McA’Nulty, 96–128. Ottawa: Canadian Linguistic Association. Rochette, Anne. 1999. The selection properties of aspectual verbs. In Kyle Johnson & Ian Roberts (eds.), Beyond Principles and Parameters, 145–165. Dordrecht: Kluwer. Rosen, Sara Thomas. 1989. Argument Structure and Complex Predicates. Ph.D. Dissertation, Brandeis University. Rosen, Sara Thomas. 1991. Restructuring verbs are light verbs. In Aaron Halpern (ed.), Proceedings of the 9th West Coast Conference on Formal Linguistics (WCCFL9), 477–491. Stanford: CSLI Publications. Sabel, Joachim. 1996. Coherent infinitives in German, Polish, and Spanish. Folia Linguistica 33:419–440. Sabel, Joachim. 2002. Das deutsche Verbum infinitum. Deutsche Sprache 29:148–175. Schmid, Tanja, Markus Bader & Josef Bayer. 2005. Coherence – An experimental approach. In Marga Reis & Stephan Kepser (eds.), Linguistic Evidence, 435–456. Dordrecht: Kluwer Academic Publishers. Sharvit, Yael. 2003. Trying to be progressive: The extensionality of try. Journal of Semantics 20:403–445. Tagliamonte, Sali & Helen Lawrence. 2000. “I used to dance, but I don’t dance now.” The habitual past in English. Journal of English Linguistics 28:324–353. Torrence, Harold. 2013. A promotion analysis of Wolof clefts. Syntax 16:176–215. Traugott, Elizabeth. 1989. On the rise of epistemic meanings in English: An example of subjectification in semantic change. Language 65:31–55. Traugott, Elizabeth. 1997. Subjectification and the development of epistemic meaning: The case of promise and threaten. In Toril Swan & Olaf J. Westvik (eds.), Modality in Germanic Languages: Historical and Comparative Perspectives, 183–210. Berlin: Mouton de Gruyter. White, Aaron Steven & Thomas Grano. 2014. An experimental investigation of partial control. In Urtzi Etxeberria, Anamaria Fălăş, Aritz Irurtzun & Bryan Leferman (eds.), Proceedings of Sinn und Bedeutung 18, 469–486. Bayonne and Vitoria-Gasteiz. Wöllstein-Leisten, Angelika. 2001. Die Syntax der dritten Konstruktion. Tübingen: Stauffenburg. Wurmbrand, Susanne. 1998. Infinitives. Ph.D. Dissertation, MIT. Wurmbrand, Susanne. 1999. Modal verbs must be raising verbs. In Sonya Bird, Andrew Carnie, Jason D. Haugen & Peter Nordquest (eds.), Proceedings of the 18th West Coast Conference on Formal Linguistics (WCCFL) 19, 599–612. Somerville, MA: Cascadilla Press. Wurmbrand, Susi. 2001. Infinitives: Restructuring and Clause Structure. Berlin: Mouton de Gruyter. Wurmbrand, Susi. 2002. Syntactic vs. semantic control. In Jan-Wouter Zwart & Werner Abraham (eds.), Studies in Comparative Germanic Syntax: Proceedings of the 15th Workshop on Comparative Germanic Syntax, 93–127. Amsterdam: Benjamins.

Isabela Nedelcu and Irina Paraschiv

3 The Romanian infinitive selected by perception and cognition verbs1 1 Introduction In the present contribution we aim to present the characteristics of structures in which verbs of perception and cognition select an infinitive form. We will focus especially on the diachronic competition of the infinitive with other verb forms selected by verbs of perception or cognition. The corpus used for our study consists of old Romanian texts, covering the period from the 16th to the 18th century (for the periodisation of old Romanian, see Gheție 1997: 52–53). The historical corpus data are compared to the contemporary language. The quantitative data, reflecting restrictions and preferences in the usage of the infinitive, will allow us to propose a more fine-grained analysis of the infinitive in these structures and will shed more light on the replacement of the infinitive by other verb forms (especially by the subjunctive) in Romanian. By and large, in the context of perception and cognition verbs the infinitive is in competition with the gerund, the subjunctive, and the indicative. As we will show, in certain cases the occurrence of an infinitive can be explained by the influence of the source language, specifically in constructions with verbs of physical perception (which prototypically select an indicative or a gerund form both in the old and in the current language). In other cases, the more frequent occurrence of one form in comparison to another can be explained historically. In Romanian, the infinitive-subjunctive competition is general and depends on factors other than historical: (i) geographical factors – (it has been shown that the infinitive is better preserved in Maramureș and to the north of Crișana (Vulpe 2006 [1963]), which has been attributed by certain researchers to the greater distance from the Greek influence; and (ii) linguistic factors – for example, after obligatory control verbs (like aspectuals) the infinitive has been preserved better, whereas after non-obligatory 1 The research that has given rise to this study has been supported by the Sectorial Operational Programme Human Resources Development (SOP HRD), financed by the European Social Fund and by the Romanian Government under the contract SOP HRD/159/1.5/S/136077. Isabela Nedelcu and Irina Nicula Paraschiv, Institute of Linguistics of the Romanian Academy, Faculty of Letters, University of Bucharest. DOI 10.1515/9783110520583-003

56

Isabela Nedelcu and Irina Paraschiv

control verbs (like the verb vrea ‘want’), it was replaced by the subjunctive, presumably because of subject ambiguity; for the infinitive-subjunctive competition, see Frâncu (1969), Diaconescu (1977), Spătaru-Pralea (2013), Nedelcu (2013), and Hill (2016, this volume). In this study, we will show how the competition between the infinitive and other verb forms manifests itself in the context of verbs of perception and cognition and we will see if there are any other (linguistic) factors responsible for the variability in the frequency of the infinitive in these contexts.

2 Verbs of perception and cognition We distinguish between verbs of physical perception2, such as vedea1 ‘see; perceive with the eyes; have the power of sight’ (cf. (1a+b)), a auzi1 ‘hear’, a simți1 ‘feel’, etc., and cognition verbs, which denote inferences/deductions, conveying indirect perceptions (Nicula 2012: 112) – vedea2 ‘perceive mentally; understand; realize’ (cf. (1c+d)), înţelege ‘understand’, a-și imagina ‘imagine’, ști ‘know’. (1) a. Văd că se apropie de noi. see.IND. PRES .1SG that CL . REFL . ACC .3SG comes.close of us.ACC ‘I can see him / her coming towards us.’ b. Văd că a plouat mult (încă este ud pe jos). see. IND. PRES .1SG that AUX . PERF.3SG rain.PPLE much still is wet on down ‘I can see it has rained a lot (the pavement is still wet).’ c. Din toate astea, văd că el este cel vinovat. from all these see.IND. PRES .1SG that he is that guilty.M . SG ‘From all these, I understand that he is the guilty one.’ d. Nu-l văd făcând așa ceva. not=CL . ACC . M .3SG see.IND. PRES .1SG do.GER such something ‘I cannot imagine him doing such a thing.’ 2 In the present study we will not make a proper distinction between constructions denoting direct physical perception (the direct interpretation of the visual data without any inferential judgement, see (1a) in the main text), and constructions denoting indirect physical perception (the visual data are indirectly processed on the basis of some signs in the discourse universe, see (1b) in the main text). We will treat them together in Section 3. This approach is mainly based on the fact that between the two types of structures (physical direct vs. physical indirect perception) there are no major differences with respect to the subordinating verb forms that they allow. For the physical direct vs. physical indirect perception distinction, as well as for the characteristics of every set of structures, see Nicula (2012: 109–114), Niculescu (2013: 69–70).

The Romanian infinitive selected by perception and cognition verbs

57

The present study will highlight the similarities and differences between the two subclasses of verbs (of physical perception and cognition) in relation to the infinitive and the competing verb forms.

3 The competition between the infinitive and other verb forms with perception verbs The preferential selection of an infinitive or of another verb form depends on the syntactic functioning of the element in question in relation to the matrix verb, as well as on certain semantic or syntactic characteristics of the two verbs, the matrix verb and the subordinate verb. 3.1 As a complement selected by a perception verb, the infinitive can enter in competition with the indicative or the gerund only under special circumstances. In Romanian, the occurrence of the infinitival complement with perception verbs is limited to certain texts and / or contexts (especially in the old language). Generally, like the present-day language, the old language does not allow the occurence of an infinitive after verbs of direct perception, thus differing from other Romance languages (see Reinheimer Rîpeanu 2001: 295–296). The extremely rare occurrence of an infinitive after verbs of direct perception (2a+b) can often be explained by the influence of the language from which the text was translated (where such verbs are frequently followed by the infinitive; see Timotin & Nedelcu 2015 for the comparison of a text translated into Romanian in the 18th century [BVS.1763] with its Italian source from the point of view of the infinitive usage): (2) a. Dumnezeu, pre carele-l văzu Ioan DOM who.DEF = CL . ACC . M .3SG see.PS .3SG John.NOM God întocma a fi cu Tatăl şi unul fiind cu similar A I NF be.INF with Father.DEF and one be.GER with Duhul Svânt. (Ev.1642: 362) Spirit Saint ‘(. . .) the Son, that John saw as one with God the Father and the Holy Spirit.’ b. au

AUX . PERF.3PL

simţit ei a işi den zisul feel.PPLE they A I NF come.out.INF from said

trup un miros foarte frumos şi dulce (BVS.1763: 245) body a smell very beautiful and sweet ‘they felt a beautiful and sweet fragrance coming out of the mentioned body’

58

Isabela Nedelcu and Irina Paraschiv

c. Simțeam a mă feel.IMPERF.1SG A I N F CL . REFL . ACC .1SG sufoca. (figter-for-freedom.blogspot.com; 19/02/2015) choke.INF ‘I felt like choking.’ d. Îi

păstrasem loc, dar o CL . DAT.3SG keep.PLUPERF.1SG place but CL . ACC . F.3SG

văd a se pune în altă bancă see.IND. PRES .1SG A I N F CL . REFL . ACC .3SG sit.INF in other desk cu altă colegă. (www.tpu.ro; 19/02/2015) with other colleague ‘I had kept a seat for her, but instead I see her sitting in another desk with another colleague.’ Both in the old (see (3a+b, d–f )) and in the current language (see (3c,g)) verbs of perception generally select either the gerund (cf. (3a–c)) or the indicative (cf. (3d–g)). In the old language, the indicative is more frequent than the gerund (for frequency rates, see Table 1). (3) a. văzu pre un boiari plângând (MC.1620:125) see.PS .3SG DOM a boyar cry.GER ‘(s)he saw a boyar crying’ b. Ce au nu ai auzit Domnul tău but or not AUX . PERF.2SG hear. PPLE God.DEF your grăindu-ţi ţie? (Ev.1642: 165) tell.GER =CL . DAT.2SG you.DAT ‘Or haven’t you heard God telling you?’ c. Doi polițiști stăteau cu radarul la pândă, two cops stand.IMPERF.3PL with radar.DEF at guard când văd venind în viteză un when see.IND. PRES .3PL come.GER in speed a BMW. (www.bancul-haios.ro; 20.02.2015) BMW ‘Two cops were on the watch, when they saw a BMW speeding up towards them.’

59

The Romanian infinitive selected by perception and cognition verbs

d. vădzuiu cum den apă vineră sus şapte vaci see.PS .1SG that from water come. PS .3PL up seven cows frumoase şi grase (PO.1582: 141) fat and beautiful ‘I saw seven fat, healthy cows come up out of the river’ e. veade că i se îngraşă sees that CL . DAT.3SG CL . REFL . ACC .3PL get.fat.IND. PRES .3PL puii cubs. DEF (FD.1592–1604: 126) ‘he sees that her cubs are getting fatter’ f. previia cum aruncă mulţimea bani în look.IMPERF.3SG that throws crowd.DEF. NOM money in vistiiariu (NT.1648: 199) treasurer ‘he was looking at people throwing money in the treasurer’ g. îi

aud cum vin și CL . ACC .3PL hear.IND. PRES .1SG that come.PRES .3PL and

deodată se aud două-trei suddenly CL . REFL . IMPERS hear.PRES .3PL two-three bubuituri (forum.computergames.ro; 20/02/2015) booms.NOM ‘I hear them coming and suddenly two or three booms can be heard’ Table 1: Distribution of the verbal forms found in competition after perception verbs3 vedea1 ‘see’

auzi1 ‘hear’

a simți1 ‘feel’4

Texts

+ ger

+ ind

+ ger

+ ind

+ ger

+ ind

PH.1500–1510 (full text) CC2.1581 (100 p.) PO.1582 (full text) MC.1620 (full text) SVI.~1670 (100 p.) CLM.1700–1750 (100 p.) Prav.1780 (full text)

1 13 2 8 3 – 1

2 6 26 4 3 2 –

– 1 – 1 2 1 –

– – 2 1 4 2 –

– – – – – – –

– – – – – – –

3 The absence of the infinitive from the table is due to the fact that it was not recorded in the context of the perception verbs considered in the analysis. 4 The verb simți is encountered only once (in CLM.1700–1750), but with a cognitive meaning, ‘have the premonition’:

60

Isabela Nedelcu and Irina Paraschiv

The lower frequency of the gerund after perception verbs can be related to semantic restrictions: Perception verbs cannot be followed by the gerund of a state verb denoting a property which cannot be perceived physically (see Niculescu 2013: 78). (4) a. *L-am

văzut pe Ion fiind îmbrăcat CL . ACC . M .3SG =AUX . PERF.1SG see. PPLE DOM Ion be. GER dress.PPLE . M în togă (apud Niculescu 2013: 78) in toga

b. *L-am

CL . ACC . M .3SG =AUX . PERF.1SG

văzut pe Ion fiind în see. PPLE DOM Ion be.GER in

piața centrală (apud Niculescu 2013: 78) market.DEF central Configurations with a direct perception verb followed by the gerund look like “accusative plus gerund” constructions (Sandfeld & Olsen 1936: 292, Caragiu 1957: 71–74) and correspond to the “accusative plus infinitive” construction found in other Romance languages (Salvi 2011: 368). In this type of construction, Romanian differs from French (cf. (5a)) and Italian (cf. (5b)), both of which use the infinitive and the pseudo-relative construction (PRC), as well as from Spanish, where, in certain contexts, both forms, the infinitive and the gerund, are allowed (cf. (6a+b)); see also Nicula (2013: 250). (5) a. Fr. Je le voir venirInf / Je le vois qui vientPRC ‘I can see him coming.’ b. Ital. Senta cantareInf Gianni / Gianni cantareInf / Sento Gianni che cantaPRC ‘I can hear John singing.’ (apud Maiden and Robustelli 2007: 391–392) (6) a. En el baile vi a un oficial de uniforme caminandoGer con dos botellas de champán.

(i) Simțindu Mihaiu vodă că-i tot vin asupră feel.GER Mihai.NOM voivode that=CL . DAT.3SG repeatedly come.IND. PRES .3PL over Leșii cu Ieremie Vodă [. . .]. Poles with Jeremiah Voivode ‘As Voivode Mihai had the premonition that the Poles ruled by Voivode Jeremiah attacked him repeatedly [. . .].’

The Romanian infinitive selected by perception and cognition verbs

61

b. En el baile vi a un oficial de uniforme caminarInf con dos botellas de champán. (apud Verhaert 2008: 162) ‘At the dance, I saw an official wearing a uniform walking to and fro with two bottles of champagne.’ As shown in the literature (Felser 1999, Guasti 1993: 141–143), in French and Italian configurations with direct perception verbs and the infinitive or the pseudo-relative provide the only possibility to express the direct perception of an event, unlike the structures in which perception verbs take sentential arguments, such as that-subordinates, which regularly express indirect perception. This feature distinguishes the aforementioned Romance languages from Romanian, where the configuration [verb of perception + că ‘that’] can denote both direct perception (cf. (1a)) (with the restriction that the two predications express temporally simultaneous events) and indirect perception (cf. (1b)) (when the two predications do not denote simultaneous events). To sum up, in Romanian a perception verb can select any of the three competing verb forms: the gerund (except for the gerund of state verbs such as fi ‘be’, avea ‘have’), the indicative (introduced by că, cum, or cum că ‘that’), and the infinitive (only sporadically and in translated texts, such as BVS.1763). The subjunctive does not participate in this competition because of its incompatibility with the matrix verb: It is the mood of non-referentiality, of non-factual/possible events (GALR I 2008: 387, GBLR 2010: 235), whereas verbs of physical perceptions denote factual events. 3.2 However, in the old language there is a context in which the infinitive occurs more frequently in competition with the other verb forms (cf. (7a–c)), i.e., when it functions as the subject of a reflexive verb with an impersonal reading. This observation is not strictly related to perception verbs; reflexive-impersonal constructions generally allow the infinitive in the old language (see Frâncu 1969, Spătaru-Pralea 2013, Nedelcu 2013). At this stage, the infinitive occurs sporadically with perception verbs with impersonal reading. The examples in (7) show that such attestations are late. (7) a. ameţire de cap, când să trândăvăsc dizziness of head when CL . REFL . ACC .3PL laze.IND. PRES .3PL ochii omului şi toate să CL . REFL . IMPERS eyes.DEF. NOM man.DEF.GEN and all văd a să învârti (CDicţ.1691–1697: 457) see. IND. PRES .3PL A I N F CL . REFL . ACC .3PL spin.INF ‘dizziness, that is when your eyes become lazy and everything seems to spin’

62

Isabela Nedelcu and Irina Paraschiv

b. căci că într-însele să veade because that in=them.ACC CL . REFL . IMPERS see.IND. PRES .3SG a AINF

zice cevaș (CIst.1700–1750: 147) say.INF something

‘because it seems that there is something mentioned in them’ c. s-au

auzit în tabăra CL . REFL . IMPERS =AUX . PERF. 3PL hear. PPLE in camp.DEF vrăjmaşilor multe sunete de trâmbiţe a face enemies.DEF.GEN many sounds of trumpets A I NF make.INF gâlceavă (BVS.1763: 233) noise ‘in the enemy camp, many sounds of trumpets were heard making noise’

3.3 In old Romanian, the infinitive (cf. (8a)) enters into competition with the subjunctive (cf. (8b)) when functioning as an adjunct of a perception verb. A perception verb like privi ‘look’ can take an infinitive or a subjunctive adjunct; unlike the aforementioned perception verbs (a vedea1, a auzi1, a simți1 ), privi ‘look’ (as well as asculta ‘listen’, gusta ‘taste’, etc.) denotes an intentional event. In the current language only the subjunctive has been preserved in constructions with intentional perception verbs (cf. (8c)).5 (8) a. Domnulu den ceriu previ pre feciorii Lord.DEF from sky look.PS .3SG DOM sons.DEF oamelor a vedea doară iaste cinre who humans.DEF.GEN A I NF see.INF only is se SĂ S U B J

înţeleagă sau se ceaie understand.SUBJ.3SG or SĂ S UB J ask.SUBJ.3SG

Dumnedzău. (PH.1500–1510: 96) God ‘The Lord looked down from heaven on the children of man to see if there is any who understands and seeks God.’ 5 In the old language, the a-infinitive could function as an adjunct, especially in the context of motion verbs. In such contexts, the prepositional feature of the complementizer a is more visible than in the contexts in which the infinitive functions as a complement or a subject: (i) acea fată ce va veni a scoate that girl that AUX . FUT.3SG come.INF A I N F take-out.INF apă (PO.1582: 79) water ‘that girl who will come to get water out’

The Romanian infinitive selected by perception and cognition verbs

63

b. acolo au şăzut cu tabăra de there AUX . PERF.3SG sit.PPLE with camp.DEF and priviia şi o parte şi alta, să look.IMPERF.3SG also one.F part and other.F SĂ S UB J vază ce vor să facă (PIst.~1780: 479) see.SUBJ.3SG what want.PRES .3PL SĂ S U B J do.SUBJ.3PL ‘he laid his camp there and he was following both armies to see what they wanted to do’ c. L-am

CL . ACC . M .3SG = AUX . PERF.1SG

privit să văd cum look.PPLE SĂ S U B J see.SUBJ.1SG how

reacționează. (loveorfame.wordpress.com; 15/02/2015) reacts.PRES .3SG ‘I have looked at him to see how he reacts.’

4 The competition between the infinitive and other verb forms with cognition verbs In the old language, the competition between the infinitive and other verb forms in the context of cognition verbs manifests itself differently depending on the syntactic functioning of the competing forms in relation to the matrix cognition verb. 4.1 In the old language, the infinitival complement occurs more frequently with cognition verbs than with verbs of physical perception (cf. (9); see also Table 2). In the current language, the infinitival complement occurs frequently as well in constructions like (13c+d). This is due to the high stylistic register to which these constructions belong; in these contexts we witness a strong revival tendency of the Romance pattern (see GALR I 2008: 496, Pană Dindelegan 2012: 577–578). (9) a. pre glas şi pre deaget cunoscu pre el a after voice and after finger know.PS .3SG DOM him.ACC A I NF fi om (Ev.1642: 425) be.INF human ‘he could guess he was a human being by the voice and the touch of the finger’ b. Cumu vă gândiţi a vă how CL . REFL . ACC .2PL think.IND. PRES .2PL A I NF CL . REFL . ACC .2PL închina şi a fugi de la mene? (LD.1621–1633: 327) worship.INF and A I NF run.INF from me.ACC ‘How can you even think to worship me, and then abandon me?’

64

Isabela Nedelcu and Irina Paraschiv

c. căci nu gândiia a să părăsi because not think.IMPERF.3SG A I N F CL . REFL .3SG leave.INF de răutăţ şi pentru acéia au şi of misdeeds and for that AUX . PERF.3SG really perit (AD.1722–1725: 37) die.PPLE ‘because he did not plan to stop doing misdeeds and that’s why he died’ d. el socotea şi gândea a mă he calculate.IMPERF.3SG and think.IMPERF.3SG A I N F CL . ACC .1SG avea în slujba lui în toate întâmplările de events.DEF of have.INF in service.DEF his in all oaste (BVS.1763: 109) army ‘he calculated and thought to have me to his service for everything that happens on the battlefield’ Both in the old and in the contemporary language, verbs like vedea, simți, and auzi are also used as cognition predicates (vedea2 ‘imagine’, simți2 ‘have a premonition; consider’, auzi2 ‘imagine (saying); hear from someone else, find out’). This characteristic has syntactic consequences. In the case of the verbs vedea2 and simți2, the non-finite / finite complement cannot generally remain unexpressed (more precisely, the DP and the CP jointly constitute a propositional argument). Obligatoriness makes the cognitive meaning explicit (see examples (10) and (13) below with vedea2 and examples (11a–b) with simți2). However, when they function as verbs of direct perception, they allow the deletion of the finite / non-finite complement, similarly to auzi1, because the DP (the perceived entity) is processed separately from the CP; accordingly, (19a) below combines two different events, which allows for one of them to remain unexpressed [l-au văzut ‘they saw him’] + [au văzut că cădea după cal ‘they saw the event of his falling off the horse’]. (10) Soția mea, Ivy Sherman, vrea un nume franțuzesc [pentru Ivy Sherman wants a name French for wife.DEF my copilul nostru], dar nu știu, parcă îl child.DEF our but not know.IND. PRES .1SG as.if CL . ACC . M .3SG și văd [pe copil] făcând already see.IND. PRES .1SG DOM child do.GER karate. (www.procinema.ro; 20/02/2015) karate ‘My wife, Ivy Sherman, would like a French name (for our kid), but I’m still thinking, I already imagine him doing karate.’

7

2 4

–

–

–

– –

–

–

–

–

1 –

–

1

11

4

–

–

–

PH.1500–1510 (full text) CC2.1581 (100 p.) PO.1582 (full text) MC.1620 (full text) SVI.~1670 (100 p.) CLM.1700–1750 (100 p.) Prav.1780 (full text)

+ i n d

+ g e r

+ i n f

Texts

–

– –

–

–

–

–

+ s u b j

vedea2 ‘imagine; consider’

–

– –

–

–

1

–

+ i n f

–

– –

–

–

–

–

+ g e r

–

– 2

–

1

–

–

+ i n d

cugeta ‘think’

–

– –

–

–

–

1

+ s u b j

1

1 –

–

–

–

–

+ i n f

–

– –

–

–

–

–

+ g e r

4

7 –

1

6

–

7

+ i n d

–

– –

–

–

–

–

+ s u b j

cunoaște ‘know’

Table 2: Distribution of the verbal forms found in competition after cognition verbs

–

1 –

–

–

–

1

+ i n f

–

– –

–

–

–

–

+ g e r

–

– 2

–

1

–

–

+ i n d

gândi ‘think’

–

1 –

–

1

1

–

+ s u b j –

+ g e r + s u b j + i n f

4 – –

+ i n d

– –

1 – – –

– –

5 2 –

13 1 2 4 – –

– – –

4 – 23 – –

–

– –

–

–

–

–

+ g e r

–

1 5

2

2

10

2

+ i n d

–

– –

–

–

–

–

+ s u b j

înțelege ‘understand’

– – 20 – –

1

+ i n f

ști ‘know’

The Romanian infinitive selected by perception and cognition verbs

65

66

Isabela Nedelcu and Irina Paraschiv

(11) a. Dumitraşco-vodă, dacă au simţit pre Antiohie AUX . PERF.3SG feel.PPLE DOM Antioch Dumitrașcu-voivode if Jora hatmanul că-l pâreşte la Moscu, Jora hetman.DEF that=CL . ACC . M .3SG tells.on to Moscow l-au CL . ACC . M .3SG = AUX . PERF.3SG

mazilit. (NL.~1750–1766: 211) overthrow. PPLE

Voivode Dumitrașcu, if he felt that Antioch Jora the hetman was telling on him to Moscow, he overthrew him.’ b. [. . .] a AINF

reuși să renunț la tot manage.INF SĂ S U B J give.up.SUBJ.1SG at everything

ceea ce simțeam a fi o mare which feel.IMPERF.1SG A I NF be.INF a big povară ( justme-al3x.blogspot.com; 20/02/2015) burden ‘[. . .] to manage to give up everything I’ve thought to be a big burden’ The cognition verb a auzi2 ‘imagine (saying); find out’ behaves differently from vedea2 and simți2 with respect to the possibility of leaving its complement unrealized. On the one hand, when it has the meaning ‘imagine’, the verb auzi2 requires the obligatory realization of the complement, similarly to the verbs vedea2 and simți2 (see example (12b) from the contemporary language). On the other hand, when it has the meaning ‘find out’, it allows the omission of the complement (which can be retrieved anaphorically); see example (12a) from the old language. (12) a. Cându auzi că boleaşte, atunce fu până la then be.PS .3SG until to when hear.PS .3SG that is.sick elu. (CC2.1581: 94) him.ACC ‘When he found that he was sick, he went to him.’ b. În timp ce tastez aceste rânduri, îl și CL . ACC . M .3SG like while type.IND. PRES .1SG these lines aud pe șeful nostru, domnul Boerescu, mister.DEF Boerescu hear.IND. PRES .1SG DOM boss.DEF our explicându-mi că iubirea platonică nu explain.GER = CL . DAT.1SG that love.DEF platonic not

The Romanian infinitive selected by perception and cognition verbs

67

înseamnă ce credem noi că means what think.IND. PRES .1PL we that înseamnă. (tabu.realitatea.net; 19/02/2015) means ‘While I am typing this, I can imagine our boss, Mr. Boerescu, explaining to me that platonic love does not mean what we think it does.’ It has been shown that, in the present-day language, the usage of the verb vedea2 ‘consider; imagine’ with the a-infinitive is limited to two contexts, in which the infinitive verb is either the existential / copula verb fi ‘be’ (or one of its synonyms), or the verb avea ‘have’ (or periphrases containing that verb); see Nicula (2012: 130–131) and references cited there. In such constructions, the subject of the infinitive, differently from the subject of the matrix verb, is frequently placed postverbally (cf. (13c)), which in fact is the prototypical position of the subject of the infinitive in Romanian; in other cases, the subject is fronted to a sentence-initial position, beyond the matrix verb (cf. (13d)). In the old language, the infinitive is encountered late and only sporadically in this type of construction (for example, an infinitive after vedea2 occurs only once, in SVI.~1670); see (13a–b). (13) a. Atuncea, văzând chiar Isaac-Paşa de a nu then see.GER himself Isaac-Pasha DE A I N F not mai putea sta împotrivă nepriiatenului au enemy.DEF.GEN AUX . PERF.3SG more can.INF stand.INF against început a fugi cu acei puţini soldaţi. (BVS.1763: 161) soldiers start.PPLE A I NF run.INF with those few ‘At that moment, when Isaac-Pasha himself saw that he could not stand against the enemy, he started running together with the few remaining soldiers.’ b. Ome, văzu-te a fi deplin man.VOC see.IND. PRES .1SG = CL . ACC .2SG A I NF be.INF complete şî cu minte de bătrân. (SVI.~1670: 110) and with mind of old.man ‘Man, I can see that you are complete and with and old man judgement.’

68

Isabela Nedelcu and Irina Paraschiv

c. nu văd a fi acesta [scopul artei]Subj not see.IND. PRES .1SG A I NF be.INF this purpose.DEF art.DEF.GEN în general (www.hermeneia.com; 20/02/2015) in general ‘I don’t consider that this is the general purpose of art’ d. [Napoli]Subj nu văd a avea mari Naples not see.IND. PRES .1SG A I N F have.INF big jucători. (www.fcsteaua.ro; 20/02/2015) players ‘I don’t consider Naples to have important players.’ Cognition verbs also take complements other than infinitives: the indicative (even more frequently than the infinitive) introduced by că ‘that’ (cf. (11a), (12a), (14a)), de ‘whether; if’ (introducing indirect interrogatives) (cf. (14b)), cum ‘that’ (cf. (14c)), cum că ‘that’ (cf. (14d)), or the subjunctive, when the matrix verb is in the affirmative (cf. (15a)) or, more frequently, in the negative (cf. (15b+c); see Table 2 for the competition between the infinitive and other verb forms. (14) a. Că iudeii părea-le-se that Jewish.DEF. PL seem.IMPERF.3SG =CL . DAT.3PL =CL . REFL . IMPERS fiii că sânt şi nu crezură că sons.DEF that be.IND. PRES .3PL and not believe.PS .3PL that vor

să

AUX . FUT.3PL SĂ S U B J

cază den împărăţiia fall. SUBJ. 3PL from kingdom.DEF

ceriului. (CC2.1581: 230) sky.DEF.GEN ‘That the Jewish thought they were the children of God and they didn’t believe that they would fall from heaven.’ b. Şi atuncea vei vedea de veneţiianii vor AUX . FUT.2SG see.INF if Venetians.DEF AUX . FUT.3PL and then putea să te mântuiască den mâinele can.INF SĂ S UB J CL . ACC .2SG save.SUBJ.3PL from hands.DEF meale. (BVS.1763: 197) my ‘Then you will see whether the Venetians can save you from my hands.’

The Romanian infinitive selected by perception and cognition verbs

69

c. Vedea-veți cum nu va putea moartea see.INF = AUX . FUT.2PL how not AUX . FUT.3SG can.INF death.DEF să SĂ S U B J

stea împrotiva învăţăturei stand.SUBJ.3SG against teaching.DEF.GEN

meale. (CazV.1643: 61) my ‘You will see how death cannot stand against my teachings.’ d. amu ştiu eu cum că temi pre now know.IND. PRES .1SG I that fear.IND. PRES .2SG DOM Domnezeu (PO.1582: 71) God ‘now, I know that you fear God’ (15) a. au

AUX . PERF.3SG

purces în călătorie, gândind să begin.PPLE in trip think.GER SĂ S U B J

lovească pe prinţipul Scanderbeg fără hit.SUBJ.3SG DOM prince.DEF Scanderbeg without veaste (BVS.1763: 187) announcement ‘he began the trip, thinking to attack Prince Scanderbeg unexpectedly’ b. Iară fariseii se ruşina, că and Pharisees.DEF CL . REFL . ACC .3PL feel.ashamed.IMPERF.3PL because prepusul şi nebuniia nu ştie să cinstească suspicion.DEF and madness.DEF not knows SĂ S U B J honour.SUBJ.3PL pre carile sânt de folos. (Ev.1642: 249) DOM which.DEF. PL are of use ‘And the Pharisees were ashamed, for mistrust and madness don’t know how to honour what is useful.’ c. nu o crez să o facă not CL . ACC . F.3SG believe.IND. PRES .1SG SĂ S UB J CL . ACC . F.3SG do.SUBJ.3SG nimeni (AD.1722–1725: 161) nobody.NOM ‘I don’t think that somebody is going to do this’ The above data show that cognition verbs select the subjunctive, the infinitive, the indicative (introduced by the complementizers că, cum, cum că ‘that’, de

70

Isabela Nedelcu and Irina Paraschiv

‘whether; if’), and the gerund, but to varying degrees. Note that, in the old language, cognition verbs mostly select the indicative (see Table 2). The other finite form, the subjunctive, is encountered as the complement of a cognition verb more rarely (in the texts investigated, the subjunctive is encountered only once in 100 pages excerpted from CC2.1581, after the verb gândi ‘think’, and twice in 100 pages excerpted from SVI.~1670, after the verbs gândi ‘think’ and ști ‘know’, and twice in Prav.1780, after the verb ști ‘know’). Non-finite forms, the infinitive and the gerund are rarely encountered as complements of a cognition verb. Among the non-finite forms, the gerund is even rarer and occurs later (among the first attestations of the gerund are with the verbs vedea2, auzi2, simți2). 4.2 As in contexts with perception verbs, in the old language the occurrence of an infinitive can be favoured by a reflexive matrix verb with an impersonal reading (cf. (16a–c)). As shown above (see Section 3.2), perception verbs in impersonal constructions can select an infinitive subject. (16) a. adesea s-au văzut mai cu frequently CL . REFL . IMPERS = AUX . PERF.3PL see.PPLE more with norocire izbândele a vini (CII.~1705: 91) luck victories.DEF A I NF come.INF ‘victories were often seen happening’ b. Întristat, jalnic și amar se mournful.M sorrowful.M and bitter.M CL . REFL . IMPERS vede a fi răspunsul morții la om. (AD.1722–1725: 224) sees A I N F be. INF answer.DEF death.DEF.GEN to man ‘The answer to man’s death seems to be mournful, sorrowful and bitter.’ c. când să va cunoaşte a fi when CL . REFL . IMPERS aux.fut.3SG know.INF A I N F be.INF într-acestaş chip, iaste priimit de martur (Prav.1780: 122) in-this way is received as witness ‘and when it becomes clear that he is like his, he will be taken as a witness’ 4.3 Unlike intentional perception verbs (such as privi ‘look’, asculta ‘listen’, gusta ‘taste’; see 3.3 above), which frequently select a purpose adjunct, cognition verbs take this type of adjunct only rarely. The adjunct position is unambiguous only when there is a preposition with a purposive meaning in the construction.

The Romanian infinitive selected by perception and cognition verbs

71

Accordingly, in (17a), the purpose adjunct is signaled by the preposition pentru ‘for’, which is associated with ca să ‘in order to’, whereas in (17b+c), the connector (ca) să (ca followed by the să-subjunctive) introduces a complement. (17) a. vei

AUX . FUT.2SG

cugeta întru dânsa ziua şi think.INF about her.ACC day.DEF and

noaptea, pentru ca să pricepi să in.order.to SĂ S UB J understand.SUBJ.2SG SĂ S U B J night.DEF for faci toate ceale scrise (BB.1688: 153) do.SUBJ.2SG ALL . F. PL CEL . F. PL written ‘you will think of it day and night to understand how to do all that is written’ b. gândi ca să facă războiu şi cu think.PS .3SG that SĂ S U B J do.SUBJ.3SG war also with léşii (ULM.~1725: 140) Pollacks.DEF ‘he thought to challenge Pollacks to war’ c. gândi să-ş întărească think.PS .3SG SĂ S U B J = CL . REFL . DAT.3SG strengthen.SUBJ.3SG lucrurile întăi cu vecinii şi things.DEF first with neighbours.DEF and arate nume bun (ULM.~1725: 126) SĂ S U B J = CL . REFL . DAT.3SG show.SUBJ.3SG name good să-ș

‘he thought first to strengthen relationships with his neighbours and show his good reputation’ In the old language, the infinitive could function as a purpose adjunct by itself (see Note 5 above); later on (beginning particularly with mid-18th century) it is frequently accompanied by a preposition with purpose meaning, such as pentru, spre ‘for’ (Frâncu 1969: 102), and the whole sequence functions as a purpose adjunct. In the absence of a purposive preposition, as in (18), the infinitive is interpretable as a complement (as with the subjunctive in (17b+c)), not as an adjunct.

72

Isabela Nedelcu and Irina Paraschiv

(18) şi când avăm în trup vreo nevoe, noi în and when have.IND. PRES .1PL in body any need we in toate chipurele cugetăm a întări el (CC2.1581: 57) all ways.DEF think.IND. PRES .1PL A I N F strengthen.INF him.ACC ‘when we suffer from any physical need, we consider in each and every way to strengthen our body’ Given the above data, it seems that the competition adjunct infinitive vs. adjunct subjunctive manifests itself late in the old language, when the infinitive comes to be preceded by pentru ‘for’ (or spre ‘for’) in order to convey a purposive meaning. As a consequence of the late emergence of the purpose adjunct infinitive introduced by a preposition, the subjunctive is more frequently used with cognition verbs (see also Lühr this volume).

5 Notes on the analysis of the constructions made up of a perception / cognition verb and the infinitive The syntactic function of the infinitive (as a complement / subject / adjunct) in constructions with perception / cognition verbs is relevant for the syntactic interpretation of the whole configuration and for the competition between the infinitive and the other verb forms. In configurations with a perception / cognition verb, an accusative clitic, and a non-finite complement (cf. (1d), (10), (19a+b)), case is attributed by ECM6: The gerund / the CP a-infinitive form cannot mark the subject with the nominative; accordingly, the subject, generated in the subordinate clause, raises to the matrix verb in order to get case-marked (see also Alboiu and Hill 2013, Niculescu 2013: 75–76, Cinque 1992 for the interpretation of the subordinates of perception verbs). By comparison, in example (2b) above, the postverbal subject un miros is case-marked by the infinitive verb. (19) a. li-au

CL . ACC .3.M . SG =AUX . PERF.3PL

văzut a cădea ti după see. PPLE A I N F fall.INF over

cal (BVS.1763: 151) horse ‘they saw him falling off the horse’ 6 In Romanian ECM does not presuppose strict adjacency between the matrix verb and the subject of its complement. See Niculescu (2013: 76) for a similar observation.

The Romanian infinitive selected by perception and cognition verbs

73

b. când li-ar vedea a să when CL . ACC . M .3SG =AUX .COND.3SG see.INF A I NF CL . REFL . ACC .3SG opri și a nu vrea ti să facă stop. INF and A I NF not want.INF SĂ S UB J do. SUBJ.3SG aceaea (BVS.1763: 193) that ‘if he happened to see him pausing and not wanting to do that’ In the old language, the construction made up of the verb ști ‘know’ (which develops a modal meaning) and the infinitive shows a special behaviour. It allows, albeit unsystematically, the deletion of the complementizer a (compare examples (20a–b), with no complementizer, to examples (21a–c), in which there is an overt complementizer: In (21a), the complementizer is realized under strict adjacency between the infinitive and the matrix verb; in (21b), there are other elements intervening between the infinitive and the matrix verb; in (21c) the overt realization of the complementizer is required by the presence of the clitic7. (20) a. Dumnezeu părintele ştie răscumpăra sângele Fiiului God father.DEF knows redeem.INF blood.DEF son.DEF.GEN Său unuia născut. (CIst.1700–1750: 163) his one.GEN born ‘God knows to redeem the blood of His only begotten Son.’ b. Şi aceastea eu le spuş vrăjitorilor, and these I CL . DAT.3PL tell.PS .1SG magicians.DEF. DAT ce nime nu lei ştiu but nobody not CL . ACC .3PL know.PS .3SG dezlega ti. (PO.1582: 141) decipher.INF ‘And this is what I told the magicians, but there was none who could explain them to me.’ 7 In the present-day language, the structure made up of the verb ști ‘know’ and the bare infinitive is preserved only in dialects. In the old language the verbs cuteza ‘dare’ and vrea ‘want’ select a bare infinitive as well. In the present-day language only the verb putea ‘can’ (almost obligatorily) selects a bare infinitive, except for the cases where the marker a is required by the presence of a clitic: (i) Pot a-l întreba. can.IND. PRES .1SG A I N F = CL . ACC . M .3SG ask.INF ‘I can ask him.’

74

Isabela Nedelcu and Irina Paraschiv

(21) a. Că unul ce să ţine doftor şi because one who CL . REFL . ACC .3SG pretends doctor and nu ştie a vindeca ranele, ce folos va not knows A I N F heal.INF wounds.DEF what use AUX . FUT.3SG să SĂ S U B J

facă celui rănit? (AD.1722–1725: 105) do.SUBJ.3SG CEL . DEF. DAT wounded.M . SG

‘Because the one who pretends to be a doctor but does not know how to heal wounds, is of any use to the wounded one?’ b. aşa ştie fieştecăruia den vrăjmaşilor Lui a so knows everyone.DAT from enemies.DEF. DAT his A I NF răsplăti (CIst.1700–1750: 163) reward.INF ‘and so he knows to reward each of His enemies’ c. mulţi oameni, neştiind a să many people NEG -know.GER A I NF CL . REFL . ACC .3PL feri de o întunecare ca acéia şi privind avoid.INF of a darkness as that and gaze.GER la soare mult, au pierdut vederea în toată at sun much AUX . PERF.3PL lose. PPLE sight.DEF in all viaţa lor (CLM.1700–1750: 175) life.DEF their ‘and many people, not knowing how to protect against such darkness and gazing a lot at the sun, lost their sight once and for all’ The construction made up of the verb ști ‘know’ and the bare infinitive (cf. (20a+b)), similarly to that made up of the verb putea ‘can’ and the bare infinitive (cf. (22)), is monoclausal (for the analysis of the monoclausal structure formed of the verb putea and the infinitive, see Nicolae 2013). In this context, the infinitive complement is neither a CP8 (it does not project a functional domain) nor an IP (it does not allow sentential negation, pronominal clitics or aspectual adverbials such as mai ‘still’). This type of infinitival complement was analysed as a vP / VoiceP (see Nicolae 2014 for the analysis of the structures with a short bare infinitive as well as Jędrzejowski & Demske this volume for a broader context). In constructions with the verb ști ‘know’ and the bare infinitive (as in (20a+b)), 8 In the old language, the verb putea could also take an a-infinitive CP (Dobrovie-Sorin 2000 [1994]: 50).

The Romanian infinitive selected by perception and cognition verbs

75

which exhibit a high degree of cohesion (behaving as complex predicates), the subject is obligatorily controlled and the object raises to the matrix verb (cf. (20b)). In such constructions, the infinitive could be replaced by the subjunctive.9 When the subject is not obligatorily controlled, the verb ști ‘know’ can also select the indicative (cf. (23a+b)). (22) Îl

CL . ACC . M .3SG

pot câștiga. can.IND. PRES .1SG win.INF

‘I can win it.’ (23) a. Știu că voi câștiga. know.IND. PRES .1SG that AUX . FUT.1SG win.INF ‘I know that I will win.’ b. Știu că vei câștiga. know.IND. PRES .1SG that AUX . FUT.2SG win.INF ‘I know that you will win.’ In constructions with a reflexive-impersonal perception / cognition verb taking an infinitive / gerund subject, the subject of the non-finite form can be fronted and mark agreement with the matrix verb. In example (7a) (repeated here for convenience as (24)), the subject toate ‘all’ triggers agreement with the verb. In this type of construction, the infinitive is preferred to the subjunctive. (24) ameţire de cap, când să trândăvăsc ochii dizziness of head when CL . REFL . ACC .3PL laze.IND. PRES .3PL eyes.DEF omului şi toatei să văd a man.DEF.GEN and all(PL) CL . REFL . IMPERS see. IND. PRES .3PL A I N F învârti ti (CDicţ.1691–1697: 457) CL . REFL . ACC .3PL spin.INF să

‘dizziness, that is when man’s eyes become lazy and everything seems to spin’ 9 There is a special construction in which the verb ști ‘know’ in the negative form is followed by the subjunctive; the matrix verb and the subjunctive have different subjects. (i) Eu nu știu să facă Siemens așa I not know.IND. PRES .1SG SĂ S U B J do.SUBJ.3SG Siemens.NOM like ceva. (www.forumtoyota.ro; 19/02/2015) something ‘I haven’t heard of Siemens to do such a thing.’

76

Isabela Nedelcu and Irina Paraschiv

The constructions in which the infinitive functions as an adjunct can be analyzed as instances of control (not as clitic climbing; see example (8a) with an infinitive, repeated here for convenience as (25)). As control is not obligatory, the usage of the subjunctive was favoured with perception verbs as well as with cognition verbs (see examples (8b+c) and (17a)). pre feciorii oamelor (25) Domnului den ceriu previ God.DEF from sky look.PS .3SG DOM sons.DEF man.PL . DEF.GEN a A I NF

vedea PROi doară iaste cinre se înţeleagă sau se see.INF only is who SĂ S U B J understand.SUBJ.3SG or SĂ S UB J

ceaie Dumnedzău. (PH.1500–1510: 96) ask.SUBJ.3SG God ‘God looks down from heaven on the children of man to see if there is any who understands or seeks God.’

6 Final remarks The general problem of the competition between the infinitive and other verb forms in Romanian needs finer-grained distinctions. Besides the factors mentioned in the introduction (historical, geographical, and the influence of translated texts), which favour the replacement of the infinitive with other verb forms, the characteristics of the different semantic classes of verbs that enter into configurations with the infinitive (and other verb forms) must be taken into account. The presentation of the contexts with perception and cognition verbs has shown that the competition of the infinitive with other verb forms, and especially with the subjunctive, is influenced by many additional factors. With perception verbs, the following observations emerge: (i) It is only in configurations with intentional verbs and adjuncts that we witness the infinitive-subjunctive competition; leaving aside this construction, the subjunctive is not allowed because of the incompatibility between the perception verb (which denotes factual events) and the subjunctive (which denotes non-factual / possible events); (ii) The indicative (introduced by different complementizers) is the most frequent candidate, followed by the gerund and, ultimately, by the infinitive (in impersonal-reflexive constructions or by imitation of the source text).

The Romanian infinitive selected by perception and cognition verbs

77

With cognition verbs, the following observations emerge: (i) The indicative is the most frequent form; (ii) The subjunctive is encountered late and sporadically (in the earliest text, it is much more rarely encountered than the infinitive); (iii) The gerund emerges late, in special contexts (especially with the cognition verbs vedea2, auzi2, simți2); (iv) The infinitive has very few occurrences, most frequently with the verb ști ‘know’, of the ones included in the analysis. In constructions with perception verbs, as well as with cognition verbs, the competition between the infinitive and other verb forms manifests itself especially when the infinitive functions as a complement or a subject (of an impersonalreflexive construction). By contrast, in constructions with adjuncts, the competition is less frequent, because the matrix verb must designate an intentional event in order to take a purpose adjunct. There is a historical factor that adds to this semantic restriction: The non-ambiguous infinitival adjunct occurs in the context of a purposive preposition, and this pattern has been frequently recorded since mid-18th century. Both in the old and in the present-day language (where we witness a strong revival of the Romance pattern), the infinitive is more frequent with cognition verbs than with perception verbs. Control and clitic climbing phenomena are decisive in the competition between the infinitive and other verb forms (see also Grano this volume). For example, as shown above, in the old language, the verb ști ‘know’ forms a monoclausal structure with the bare infinitive (in which subject control is obligatory, and the object raises to object), which allowed for the replacement of the infinitive by the subjunctive.

Primary sources AD.1722–1725 – Antim Ivireanul, Didahii, ed. by G. Ştrempel. In: Miron Costin, Opere, Bucharest: Editura Minerva, 1972, 3–210. BB.1688 – Biblia adecă Dumnezeiasca Scriptură a Vechiului şi Noului Testament, tipărită întâia oară la 1688 în timpul lui Şerban Vodă Cantacuzino, Domnul Ţării Româneşti, Bucharest: Editura Institutului Biblic, 1977. BVS.1763 – Vlad Boţulescu de Mălăieşti, Viaţa lui Scanderbeg, ed. by E. Timotin and O. Olar, Bucharest: Univers Enciclopedic Gold, 2013. CazV.1643 – Varlaam, Cazania, ed. by J. Byck, Bucharest: Editura Academiei Române. CC2.1581 – Diaconul Coresi, Carte cu învăţătură, ed. by Sextil Puşcariu, Al. Procopovici, vol. I, Textul, Bucharest: Atelierele Grafice Socec & Co., 1914.

78

Isabela Nedelcu and Irina Paraschiv

CDicţ.1691–1697 – Teodor Corbea, Dictiones Latinae cum Valachica interpretatione, ed. by A.-M. Gherman, vol. I, Cluj-Napoca: Clusium, 2001. CII.~1705 – Dimitrie Cantemir, Istoria ieroglifică, ed. by S. Toma. In D. Cantemir, Opere complete, IV, Bucharest: Editura Junimea, 1974. CIst.1700–1750 – Constantin Cantacuzino, Istoriia Ţărâi Rumâneşti, ed. by O. Dragomir, Bucharest: Editura Academiei Române, 2006. CLM.1700–1750 – Miron Costin, Letopiseţul Ţării Moldovei, ed. by P. P. Panaitescu. In: Miron Costin, Opere, Bucharest: Editura de Stat pentru Literatură și artă, 1958. Ev.1642 – Evanghelie învăţătoare, ed. by A.-M. Gherman, Bucharest: Editura Academiei Române, 2011, 153–480. FD.1592–1604 – Floarea darurilor, in Cele mai vechi cărţi populare în literatura română, X, ed. by Al. Moraru, Bucharest: Editura Minerva, 1996.

Internet LD.1621–1633 – Legenda duminicii [ms. 5032, Manuscrisul de la Ieud]. In Cele mai vechi cărţi populare în literatura română, I, ed. by E. Timotin, Bucharest: Fundaţia Naţională pentru Ştiinţă şi Artă, 2005. MC.1620 – Mihail Moxa, Cronica universală, ed. by G. Mihăilă, Bucharest: Editura Minerva, 1989. NL.~1750–1766 – Ion Neculce, Letopiseţul Ţării Moldovei şi O samă de cuvinte, ed. by I. Iordan, Bucharest: Editura de Stat pentru Literatură și Artă, 1959. NT.1648 – Noul Testament. Tipărit pentru prima dată în limba română la 1648 de către Simion Ştefan, Mitropolitul Transilvaniei, Alba Iulia: Editura Episcopiei Ortodoxe Române a Albei Iuliei, 1988. PH.1500–1510 – Psaltirea Hurmuzaki, ed. by I. Gheţie, M. Teodorescu, Bucharest: Editura Academiei Române, 2005. PIst.~1780 – Radu Popescu, Istoriile domnilor Țării Rumânești, ed. by M. Gregorian. In Cronicari munteni, I, Bucharest, 1961. PO.1582 – Palia de la Orăştie, ed. by V. Pamfil, Bucharest: Editura Academiei Române, 1968. Prav.1780 – Pravilniceasca condică, ed. by the Group for old Romanian law conducted by Acad. A. Rădulescu, Bucharest: Editura Academiei Române, 1957. SVI.~1670 – Varlaam şi Ioasaf, ed. by M. Stanciu Istrate, Reflexe ale medievalităţii europene în cultura română veche: Varlaam şi Ioasaf în cea mai veche versiune a traducerii lui Udrişte Năsturel, Bucharest: Editura Muzeului Național al Literaturii Române, 2013. ULM.~1725 – Grigore Ureche, Letopiseţul Ţării Moldovei, ed. by P. P. Panaitescu, Bucharest: Editura de Stat pentru Literatură și Artă, 1955.

Secondary references Alboiu, Gabriela & Virginia Hill. 2013. The Case of A-bar ECM: Evidence from Romanian. In S. Keine & S. Sloggett (eds.), NELS 42: Proceedings of the 42nd Annual Meeting of the North East Linguistic Society 1, 25–39. Amherst, UMass: GSLA.

The Romanian infinitive selected by perception and cognition verbs

79

Caragiu, Matilda. 1957. Sintaxa gerunziului românesc [The Syntax of the Romanian Gerund]. In Alexandru Graur & Jacques Byck (eds.). Studii de gramatică II, 61–89. Bucharest: Editura Academiei Române. Cinque, Guglielmo. 1996. The Pseudo-Relative and Acc-ing Constructions after Verbs of Perception. In Italian Syntax and Universal Grammar, 244–275. Cambridge: Cambridge University Press. Diaconescu, Ion. 1977. Infinitivul în limba română [The Infinitive in Romanian]. Bucharest: Editura Ştiinţifică şi Enciclopedică. Dobrovie-Sorin, Carmen. 2000. [1994], Sintaxa limbii române. Studii de sintaxă comparată a limbilor romanice. Bucharest: Editura Univers [translation of The Syntax of Romanian. Comparative Studies in Romance. Berlin: Mouton de Gruyter]. Felser, Claudia. 1999. Verbal Complement Clauses. A Minimalist Study of Direct Perception Constructions. Amsterdam: John Benjamins. Frâncu, Constantin. 1969. Cu privire la uniunea lingvistică balcanică. Înlocuirea infinitivului prin construcţii personale în limba română veche [On the Balkan Sprachbund. The Replacement of the Infinitive by Personal Constructions in Old Romanian]. Anuarul de lingvistică şi istorie literară XX, 69–116. GALR 2008 – V. Guţu Romalo (ed.), Gramatica limbii române [The Grammar of Romanian]. Bucharest: Editura Academiei Române. GBLR 2010 – G. Pană Dindelegan, (ed.), Gramatica de bază a limbii române [The Essential Grammar of Romanian]. Bucharest: Univers Enciclopedic Gold. Gheție, Ion (ed.). 1997. Istoria limbii române literare. Epoca veche (1532–1780) [The History of the Literary Romanian Language. The Old Period (1532–1780)]. Bucharest: Editura Academiei Române. Guasti, Maria Teresa. 1993. Causative and Perception Verbs. A Comparative Study. Torino: Rosenberg & Sellier. Maiden, Martin & Cecilia Robustelli. 2007. A Reference Grammar of Modern Italian. London: Hodder Arnold. Nedelcu, Isabela. 2013. Particularităţi sintactice ale limbii române în context romanic. Infinitivul [Syntactic Features of Romanian from a Typological Perspective. The Infinitive]. Bucharest: Editura Muzeului Naţional al Literaturii Române. Nicolae, Alexandru. 2013. Types of Ellipsis. The Interpretation of Structures Containing Ellipsis Sites and Syntactic Licensing of Ellipsis. Bucharest: University of Bucharest PhD thesis. Nicolae, Alexandru. 2014. On the Uniform Syntax of Modal Verbs in Romanian. Paper presented at the Romance Linguistics Seminar. Oxford: University of Oxford, 20 February. Nicula, Irina. 2012. Modalităţi de exprimare a percepţiilor fizice. Verbele de percepţie în limba română [Ways of Expressing Physical Perceptions. The Romanian Verbs of Perception]. Bucharest: Editura Universităţii din Bucureşti. Nicula, Irina. 2013. The gerund (present participle). In G. Pană Dindelegan (ed.), The Grammar of Romanian, 245–253. Oxford: Oxford University Press. Niculescu, Dana Ioana. 2013. Particularități sintactice ale limbii române din perspectivă tipologică. Gerunziul [Syntactic Features of Romanian from a Typological Perspective. The Gerund]. Bucharest: Editura Muzeului Naţional al Literaturii Române. Pană Dindelegan, Gabriela. 2012. Verbul [The Verb]. In Gheorghe Chivu, Gabriela Pană Dindelegan, Adina Dragomirescu, Isabela Nedelcu & Irina Nicula (eds.), Studii de istorie a limbii române. Morfosintaxa limbii literare în secolele al XIX-lea și al XX-lea [Studies on The History of Romanian. The Morphosyntax of the Literary Language in the 19th and 20th centuries], 535–597. Bucharest: Editura Academiei Române.

80

Isabela Nedelcu and Irina Paraschiv

Reinheimer Rîpeanu, Sanda. 2001. Lingvistica romanică. Lexic. Morfologie. Fonetică [Romance Linguistics. Lexicon. Morphology. Phonetics]. Bucharest: All. Salvi, Giampaolo. 2011. Morphosyntactic Persistence. In Martin Maiden, John Charles Smith & Adam Ledgeway (eds.), The Cambridge History of Romance Languages. I. Structures, 318– 381. Cambridge: Cambridge University Press. Sandfeld, K., H. Olsen. 1936. Syntaxe Roumaine. I. Emploi des mots à flexion. Paris: Libraire E. Droz. Spătaru-Pralea, Mădălina. 2013. Concurenţa infinitiv – conjunctiv în limba română [The Infinitive – Subjunctive Competition in Romanian]. Bucharest: Editura Universitară. Timotin, Emanuela & Isabela Nedelcu. 2015. Stability and Innovation in the Use of the Infinitive in an 18th Century Translation from Italian into Romanian. In Gabriela Pană Dindelegan, Rodica Zafiu, Adina Dragomirescu, Irina Nicula, Alexandru Nicolae & Louise Esher (eds.), Diachronic Variation in Romanian, 253–275. Cambridge: Cambridge Scholars Publishing. Verhaert, A. 2008. El gerundio no perifrástico del espanol. Cómo no ser ni demasiado explícito ni demasiado implícito. Amsterdam: Rodopi. Vulpe, Magdalena. 2006 [1963]. Repartiția geografică a construcțiilor cu infinitivul și cu conjunctivul în limba română [The Geographical Distribution of the Infinitive and Subjunctive Constructions in Romanian]. In Opera lingvistică. II. Dialectologie românească. Varia, 193–226. Cluj-Napoca: Clusium.

Jerneja Kavčič

4 A diachronic perspective on the semantics of AcI clauses in Greek 1 Introduction AcI clauses are believed to convey relative temporality in Ancient Greek (AG), with the future infinitive referring to posteriority, the present infinitive to simultaneity, and the aorist infinitive to anteriority. This state of affairs is exceptional if it is assumed that, in Greek, the infinitive encoded aspect (cf. Duhoux 2000: 282; Miller 2002: 34; Binnick 1991: 93). Although this is a widespread opinion, it has also been claimed that AG verb stems encode relative temporality rather than aspect. The view that AG verb stems are aspectual is thus “untenable” according to Rijksbaron (2006: 2); compare also Sevdali (2013: 325).1 Nevertheless, AG displays clear cases of AcI clauses conveying temporal distinctions. In example (1) (Plato, Resp. 277e8), the aorist infinitive refers to anteriority: (1)

há what

phēsi say.3SG

drãsai do.AOR.INF

autòn him.ACC

Hēsíodos Hesiod

‘what Hesiod says that he had done’ [Plato, Resp. 277e8] An example of an AcI clause containing a present infinitive and referring to simultaneity is Hdt. 7.129 (cf. (2)), whereas in Isoc. 12.249 (cf. (3)) the future infinitive refers to posteriority. (2)

nomízei think.3SG

Poseidéōna Poseidon.ACC

tḕn the

gẽn earth

seíein shake.PRS.INF

‘he thinks that Poseidon shakes the earth’ [Hdt. 7.129] (3)

egṑ I

d’ but

hēgoũmai think.1SG

béltistá best

se you.ACC

práxein do.FUT.INF

‘but I think that you will do it best’ [Isoc. 12.249] 1 For a more detailed description of the AG verb system see, for example, Rijksbaron (2006: 1–8) or Schwyzer (1953: 640–644). Jerneja Kavčič, University of Ljubljana DOI 10.1515/9783110520583-004

82

Jerneja Kavčič

There was another infinitive in Classical Greek (CG), the perfect infinitive. Such clauses were rare in the Classical period (fifth and fourth centuries BC), according to Rijksbaron (2006: 98). An example is Plato, HpMi 364c (cf. (4)): (4)

phēmì say.1SG áriston best.ACC

gàr for

Hómēron Homer.ACC

mèn indeed

pepoiēkénai make.PRF.INF

ándra man.ACC

Akhilléa Achilleus.ACC

‘I say that Homer presented Achilleus as the best man’ [Plato, HpMi 364c] These examples indicate that AcI clauses conveyed the same temporal and aspectual properties as their finite counterparts. As a consequence, a recent study uses the term “isomorphism” when referring to the relation between infinitive and finite complement clauses: “Isomorphism [. . .] means that, apart from person and number information and the corresponding morphemes, infinitival constructions employ almost identical morphological means as their finite counterparts to implement the same grammatical function” (Fykias & Katsikadeli 2013: 39). Both modern linguistics and traditional grammaticography have drawn attention to this phenomenon. Additional examples include Jannaris (1968: 569), who calls infinitive clauses “equivalent” with their finite counterparts, whereas Kurzová (1966: 44) draws attention to the variety of syntactic functions of the AG infinitive. This variety does not seem to “correspond with its morphological character” and could be linked to the subsequent retreat of the infinitive from Greek. Yet another example is Miller (2002: 34), who refers to AG when claiming the following: “Infinitives can have tense morphology. When they do they can be used exactly as fully tensed complement clauses.” As stated above, it remains open whether or not one can claim that AG infinitives encoded tense.2 This issue is addressed again below in Section 7.2. Nevertheless, it is clear that at least 2 In addition, an anonymous reviewer rightly commented that infinitive clauses can display tense even if they do not have overt tense morphology; examples include English and Dutch (Stowell 1982; ter Beek 2010). As a consequence, Miller’s statement quoted above should be modified. However, it is less clear whether or not this modification has consequences for the analysis. According to Stowell (1982: 566), temporal properties of ECM infinitives depend on the main verb. This could raise the question of whether or not temporal properties of Greek ECM clauses are also dependent on the main verb. For instance, it could be assumed that temporal properties of AG AcI clauses containing aorist infinitives are related to the main verb (cf. below, Note 16). However, a detailed investigation into this issue lies beyond the scope of this study because it is not clear whether or not AG AcI clauses can be considered ECM infinitives. A recent study (Sevdali 2013) provides strong arguments against this view (cf. Sevdali 2013: 331‒ 332). See also below, Note 11, on the relevance of Modern Dutch data for this analysis.

A diachronic perspective on the semantics of AcI clauses in Greek

83

some AG syntactic phenomena speak in support of structural equivalence between AcI clauses and finite complement clauses. A particularly convincing argument seems to be provided by instances of “hybrid” AG complement clauses with a finite complement clause alternating with an infinitive clause. An example is Lys. 10.15 (cf. (5)), where a finite part of the complement clause (hóti egṑ mèn orthõs légō) is followed by an infinitive clause (toũton dè skaiòn eĩnai).3 (5)

humãs you.ACC egṑ I

pántas all.ACC

eidénai know.PRS.INF

mèn on the one hand

dè on the other hand

orthõs correctly

skaiòn awkward.ACC

hēgoũmai believe.1SG légō, speak.1SG.PRS

hóti that toũton this.ACC

eĩnai be.PRS.INF

‘I believe you all know that I speak correctly, whereas he is awkward’ [Lys. 10.15] As stated above, these phenomena support the assumption that AG finite complement clauses were structurally equivalent to infinitive clauses and – as passages like (5) seem to indicate – that any finite complement clause or even only one part of it could be replaced with an infinitive clause. Nevertheless, this study assumes that semantic distinctions between finite and nonfinite complement clauses might have occurred in at least some periods of AG. It explores the semantics of infinitive clauses, rather than directly addressing the issue of whether or not any finite complement clause (or part of it) could be replaced by an infinitive clause. However, it seems that these semantic properties of infinitive clauses can also shed light on the issue of when a finite complement clause was likely to be replaced with an infinitive clause in AG (see also D’hoedt & Cuyckens this volume). The study focuses on AcI clauses dependent on verbs of saying and thinking (rather than, e.g., AcI clauses dependent on verbs of ordering such as keleúō ‘order’). Sometimes such clauses are also called “declarative infinitive clauses”; see examples (1)–(4).

2 Aims of the study Various attempts have been made to account for the variety of AG complement clauses in semantic or pragmatic terms. For instance, Cristofaro (2008: 576) argues 3 This passage is also quoted in Sevdali (2013: 338) and in Fykias & Katsikadeli (2013: 38); see also below, Section 7.1. For other “hybrid” passages, see Jannaris (1968: 570) and Kühner & Gerth (1904: 357).

84

Jerneja Kavčič

that, in the case of verbs of saying (utterance predicates), an infinitive complement clause or an indicative complement clause introduced with hōs was used when the speaker was not committed to the truth of the complement clause, whereas an indicative complement clause introduced with hoti was used when the speaker was committed to its truth.4 The aim of this study is to provide an alternative approach by first drawing attention to AcI clauses dependent on verbs of saying and thinking in New Testament (NT) Greek as well as in the non-literary papyri of the first century AD. It is assumed that the state of affairs in these documents, which are believed to provide fairly reliable insights into the language of the period, could shed light on the semantic properties of AcI clauses in earlier periods (cf. Horrocks 2010: 149). For instance, it is claimed below (Section 3) that in the Greek of the first century AD there is a strong tendency towards AcI clauses dependent on verbs of saying and thinking containing the infinitive ‘to be’ or a present infinitive of another stative verb. Although it is clear that AcI clauses conveyed relative temporality in the Classical period, with present infinitives referring to simultaneity, it seems reasonable – particularly in the light of the semantic properties of AcI clauses in the first century AD – to ask whether or not earlier periods display tendencies towards the stativity of these clauses as well. Thus Sections 6 through 11 discuss the issue of whether or not, in diachronic terms, Greek AcI clauses containing a present infinitive display an increasing or decreasing tendency towards stativity. It is very well known that, after the Classical period, the Greek infinitive underwent a process of gradual disappearance. Its earliest signs are believed to go back to as early as the late Classical period (late fifth / early fourth centuries BC, cf. Joseph 1983: 49). As a consequence, another question can be raised, i.e., whether or not the semantic properties of AcI clauses in the first century AD are related to this process, thus going back to a significantly earlier period. The process of the infinitive disappearing has usually been described mainly in terms of the infinitive clause being replaced by finite complements (cf. Joseph 1983: 37‒85; Horrocks 2010: 93, 173; Jędrzejowski & Demske this volume; Hill this volume). By examining the aforementioned issues, which concern the aspectual and temporal properties of AcI clauses, this study also aims to explore the process of the infinitive disappearing from the perspective of semantic changes of AcI clauses. In addition, it briefly addresses the issue of whether AG verb stems were aspectual or temporal (cf. above, Section 1). 4 Cf. also: Ruijgh (1999: 229), Scheppers (1999: 281), and Kurzová (1968: 64). In most of these interpretations, an infinitive clause has a more subjective nuance than a finite complement clause.

A diachronic perspective on the semantics of AcI clauses in Greek

85

3 Semantic properties of AcI clauses in the first century AD For reasons already discussed, this section is an overview of the semantic properties of AcI clauses in the Greek of the first century AD, i.e., in the Greek of the New Testament (NT) and in the contemporary non-literary papyri.5 Both the NT and the contemporary non-literary papyri (dating back to late first / early second centuries AD) are clearly characterized by an avoidance of AcI clauses containing an aorist infinitive. This observation goes back to Burton (1898: 53); compare Kavčič (2017: 25). Other properties common to both NT Greek and non-literary papyri AcI clauses include relatively high frequencies of stative present infinitives and perfect infinitives (Fanning 1990: 401; Thorley 1989: 296; Kavčič 2009).6 In AcI clauses, stative present infinitives prevail over non-stative ones in both NT Greek and in the contemporary non-literary papyri, although the most common stative verb ‘to be’ is more common in the former than in the latter. In the NT, it accounts for approximately 47% of all AcI clauses (Kavčič 2009: 155). AcI clauses containing non-stative present infinitives are rare, accounting for at most 7% of all AcI clauses in the NT and 8% of all such clauses in the non-literary papyri; compare Kavčič (2017: 32). Perfect infinitives are significantly more common than aorist infinitives in both NT Greek and in the contemporary non-literary papyri. In the NT, 21% of all AcI clauses contain a perfect infinitive, whereas this share reaches 34.5% in the contemporary non-literary papyri (Kavčič 2017: 35). These data contrast with Rijkbaron’s claim mentioned above that AcI clauses containing perfect infinitives were rare in the Classical period (cf. Section 1). As a consequence, it seems that, in diachronic terms, AcI clauses reveal an increased tendency towards containing perfect infinitives. On the other hand, AcI clauses containing future infinitives are uncommon in both private letters and in the NT (the latter containing only one certain instance of a future infinitive in an AcI clause, the infinitive ésesthai ‘to be’ in Acts 23.30; Kavčič 2017: 47).7 These tendencies have led to the hypothesis that AcI clauses mostly express simultaneous states in NT Greek: “By the 1st century AD the infinitive construc5 The corpus examined consisted of around 150 non-literary papyri dating from the first and early second centuries AD (around 30,000 words); compare Kavčič (2017: 24). 6 It has to be stressed that the perfect was stative in AG, at least in the earliest stages. This issue is explained in greater detail below in Section 4. 7 AcI clauses containing future infinitives are not uncommon in official papyri documents of the first century AD. However, these documents do not seem to reflect the state of affairs in the spoken language.

86

Jerneja Kavčič

tion had in any case lost ground to ὅτι, and though it was far from defunct it was apparently in common usage becoming restricted to statements about a present state” (Thorley 1989: 295).8 If this solution is accepted, the construction could have parallels in English as well as in some other languages.9 Nevertheless, the assumption that AcI clauses are restricted to expressing simultaneous states in the first century AD calls for an additional comment. On the one hand, it can be argued that by this time AcI clauses could no longer convey posteriority. The gradual retreat of the future infinitive can be traced back to the fourth century AD because it appears that the future infinitive ceased to be a productive category in this period (Fykias & Katsikadeli 2013: 40; Markopoulos 2009: 28). The AG synthetic future was gradually replaced in the later stages of Greek by periphrastic verb forms. According to Markopoulos (2009), HellenisticRoman Greek (third century BC – fourth century AD) yields several such forms. However, the AG synthetic future infinitive was disappearing without being replaced with future periphrases (e.g., with the periphrasis méllō + inf., which was the most commonly acceptable future periphrasis of the period; Markopoulos 2009: 49, 85). In other words, the AG synthetic future infinitive does not seem to tend towards being replaced within AcI clauses with innovative periphrastic future forms. As already stated, AcI clauses containing future infinitives are uncommon in the NT and in the contemporary private letters, although they appear to be more common in official documents (cf. Note 7). On the other hand, the state of affairs seems to be more complex in terms of expressing anteriority in AcI clauses. In the first century AD, this assumption is supported by the avoidance of AcI clauses containing aorist infinitives. Nevertheless, it seems that the role of conveying anteriority in AcI clauses could have been adopted by perfect infinitives.

4 The perfect infinitive as an expression of anteriority? As is frequently argued, the system of the AG perfect tense was not completed before the (late) Classical period (cf. Chantraine 1927; Rijksbaron 2006: 37; Duhoux 2000: 429). The most recent approaches to the semantics of the AG 8 Passages such as the following (NT Mk 14.64) indicate that the correct formulation is “simultaneous states” rather than “present states”: (i) hoi dè pántes katékrinan autòn énokhon eĩnai thanátou the but all condemn.3PL.PST him.ACC guilty.ACC be.PRS.INF of.death ‘they all condemned him as worthy of death’ [NT Mk 14.64] 9 English infinitive clauses of this type can contain the verb be or have, or (rarely) some third stative verb (cf. Stockwell, Schacter & Hall Partee 1973: 569–571); for other languages, see Grevisse (1993: 1278), Skytte (1983: 292ff.), and Pollock (1980: 410).

A diachronic perspective on the semantics of AcI clauses in Greek

87

perfect claim that this period saw the emergence of the so-called anterior perfect. Unlike the original resultative perfect, which emphasizes the state resulting from an anterior event, this perfect is usually described as the perfect referring to an anterior event with current relevance (cf. Haspelmath 1992: 190; Haug 2008: 293). In addition, it tends to be transitive; the history of the Greek perfect is thus further characterized by the transition from an intransitive to a transitive aspect/tense (cf. Chantraine 1927; Duhoux 2000: 428; Haspelmath 1992: 194−217; Rijksbaron 2006: 37). As it is sometimes argued, it is not clear whether, at the earliest stages of its emergence, the AG perfect is to be interpreted in terms of tense or aspect. This is a very well-known problem and concerns other languages as well (cf. Comrie 1981: 52). It is believed, however, that the past orientedness of the AG perfect became increasingly prominent in diachronic terms. The AG perfect thus eventually merged with the aorist. In other words, it lost its stative meaning while retaining the past reference, given that it originally implied a past action. It was the reference to that action that was retained. The issue of when the process of the functional merger between the AG aorist and the perfect was completed remains open. Some scholars (mostly McKay 1980, 1981; Porter 1989) place this development in late antiquity (fifth century AD), whereas others (e.g., Wackernagel 1904; Chantraine 1927; Horrocks 2010) find at least sporadic occurrences of the phenomenon considerably earlier, in the Roman-Hellenistic period (third century BC – fourth century AD) or perhaps even the Classical period.10 The development appears to be reflected in Modern Greek active aorists, which functioned as active perfects in AG; for example, in the aorist brḗka, which goes back to the AG active perfect heúrēka. Compare Horrocks (2010: 302) on other cases. As a consequence, it could be assumed that these processes, which took place in the Classical and subsequent periods and led to the functional merger with the aorist, resulted in restructuring the system of expressing temporal distinctions within AcI clauses. In the first century AD, present infinitives are thus simultaneous with the main verb and (the rare) future infinitives are posterior to the main verb, whereas – due to the past orientedness of the AG perfect becoming increasingly prominent in diachronic terms – perfect infinitives are anterior to the main verb. (In earlier periods, on the other hand, anteriority was conveyed by the aorist infinitive.) In terms of the perfect infinitive, Sevdali (2007: 6) draws attention to this possibility, claiming that: “The temporal interpretation

10 See McKay (1980; 1981: 318–320), Porter (1989: 273), Wackernagel (1904), Chantraine (1927: 184ff.), Jannaris (1968: 439–440, 485), Horrocks (2010: 176–178), Blass, Debrunner & Rehkopf (2001: 281), and Sicking & Stork (1996: 129).

88

Jerneja Kavčič

of AG infinitives can come from perfective Aspect (yielding the anteriority effect observed with some cases, especially with the present perfect infinitive).”11 According to Sicking & Stork (1996: 169–170), actions and processes in the active perfect can convey anteriority in the earliest stages of Greek; examples include téthnēka ‘have died’ and léloipa ‘have left’.12 AcI clauses containing such perfect infinitives are far from uncommon in the first century AD. An example from the non-literary papyri is the perfect infinitive eilēphénai ‘to have gotten’ (falsely spelled ailēphénai) in P.Bad. 35, cf. (6):13 (6)

gígnoske know.2SG.IMP Prosdokímou Prosdokimos

mḕ NEG tòn the

ailēphénai get.PRF.INF

me me.ACC

parà from

khalkón bronze

‘know that I have not received bronze from Prosdokimos’ [P.Bad. 35, AD 87] If the role of the perfect infinitive is to convey anteriority, it cannot be claimed that AcI clauses ceased to convey anteriority by the first century AD. Instead, it should be assumed that the perfect infinitive adopted the function of conveying anteriority in AcI clauses and thus replaced the aorist infinitive in this function.

5 A further note on the semantic change of AG AcI clauses As stated in Section 2, this study explores tendencies concerning the stativity of AG AcI clauses. However, Section 4 raised another issue: namely, whether or not perfect infinitives could convey anteriority in the construction investigated. As a consequence, the study directs additional attention to the issue of whether it can 11 A possible parallel can be found in Modern Dutch. According to ter Beek (2010: 43), perfect infinitives in nonfinite complement clauses represent past tense rather than perfective aspect. 12 A process is a non-controlled situation that involves change; an action is a situation that involves control and change (Sicking & Stork 1996: 138). Compare Rijksbaron (2006: 35–38), who uses the terms “terminative verbs” and “stative verbs.” I follow Sicking and Stork (1996) because their terms are more clearly defined. In addition, this approach broadly corresponds to Haspelmath’s claim that the anterior perfect is agent-oriented whereas the resultative perfect is patient-oriented (cf. Haspelmath 1992: 209−215; he uses the term “perfect” instead of the term “anterior perfect”). 13 Other examples include apodedōkénai ‘to have given back’ (P.Oxy. 3274), diērpakénai ‘to have torn’ (SB 5761), peprakénai ‘to have done’ (BGU 584, PMich 263), ēgorakénai ‘to have bought’ (P.Mich. 236), dedaneikénai ‘to have borrowed’ (P.Dura. 18), and so on.

A diachronic perspective on the semantics of AcI clauses in Greek

89

be claimed that, in diachronic terms, AcI clauses ceased to convey anteriority, as well as relative temporality in general. If it holds true that AcI clauses ceased to convey anteriority, then it can also be claimed that AcI clauses tended towards omitting temporal distinctions in general because it appears that the gradual omission of the future infinitive goes back to a relatively early stage (cf. above, Section 3). According to Fykias & Katsikadeli (2013: 40), the state of affairs in CG, with AcI clauses conveying temporal distinctions, is “non-canonical,” which is why it “was in force for a limited time span in the history of Greek” (cf. also below, Section 10). It is assumed that both aspects investigated are closely related to one another because discussing the issue of whether or not perfect infinitives tended towards adopting the function of expressing anteriority in AcI clauses necessarily raises the question of whether the AG perfect is to be interpreted in terms of an aspect or a tense. If it is to be understood as an aspect, then – owing to its stativity – its use in AcI clauses may be understood as a tendency towards the stativity of AcI clauses; if it was a tense, it could convey anteriority in AcI clauses. However, in the case of an ancient language like AG, this issue is particularly difficult to solve because it appears to be controversial even when dealing with modern spoken languages. See for example the discussion about the Modern Greek perfect in Moser (2003). The problems of the relation between the AG perfect and the aorist, as well as the issue of their functional merger, have sometimes been dealt with rather impressionistically (e.g., McKay 1980; Blass, Debrunner & Rehkopf 2001: 281). The study attempts to avoid such judgments by exploring diachronic tendencies in the use of the aorist and the perfect infinitive in AcI clauses. As a consequence, I analyze perfect infinitives in AcI clauses from two different perspectives: from the perspective of their transitivity, and from the perspective of their semantic properties. This is why I adopt the terms “active transitive perfect” (ATP) and the term “active perfect of actions and processes” (APAP; cf. Sicking & Stork 1996: 146; Rijksbaron 2006: 37; Duhoux 2000: 428). The former is used because the increase in the past orientedness of the AG perfect is believed to be associated with an increase in its transitivity; the latter is needed because it appears that some intransitive verbs can also refer to past completed actions (e.g., téthnēka ‘have died’; cf. above, Section 4). In exploring the issue of whether perfect infinitives tended towards conveying anteriority in AcI clauses, this study assumes that: – If the high frequencies of AcI clauses containing perfect infinitives in the first century AD (cf. above, Section 3) are a result of an increasing use of the ATP (rather than, e.g., of passive perfect infinitives) in AcI clauses, it is likely that the ATP tended towards adopting the role of conveying anteriority

90

–

–

Jerneja Kavčič

in AcI clauses. This is because it is believed that there is a correlation between the spread of the ATP in the language as a whole and the functional merger between the aorist and the perfect (cf. above, Section 4). In addition, if an increasing tendency in the use of AcI clauses containing ATP infinitives is accompanied with a decrease (and eventual omission) of AcI clauses containing aorist infinitives, then this could be another indication that ATP infinitives tended towards adopting the role of conveying anteriority (and, thus, perhaps replacing the aorist infinitive) in AcI clauses. As mentioned above, Sicking & Stork (1996: 169–170) argue that (in the earliest stages of AG) active perfect forms of actions and processes (i.e., APAP) can (but need not) refer to anteriority. As a consequence, this study also examines whether or not, in AcI clauses, the APAP displays an increasing or decreasing tendency; if the latter holds true, then this is perhaps another indication of perfect infinitives in AcI clauses adopting the function of conveying anteriority.

6 AcI clauses in the earliest Greek documents It is perhaps a coincidence that the two earliest instances of an AcI clause in Greek contain a stative present infinitive, the infinitive ‘to have’ (e-ke-e); these cases go back to Mycenaean Greek (ca. 1200 BC; cf. Duhoux 2000: 265). A few centuries later, Homeric Greek displays AcI clauses containing aorist, future, and perfect infinitives, in addition to present infinitives. According to Chantraine (1953: 305–307), these infinitives convey temporal distinctions within Homeric AcI clauses, although – except for the future infinitive – their primary function was to encode aspect rather than tense. Thus Homeric AcI clauses largely display the same temporal distinctions as AcI clauses in later periods. Most Homeric AcI clauses contain a present infinitive, the infinitive ‘to be’ (Ionic eĩnai or Aeolic émmenai) being the most common. An example is provided by Hom., Il. 2.248–249, cf. (7): (7)

ou NEG

gàr for

egṑ I

brotòn mortal man.ACC

seo from.you

phēmì say.1SG

állon another.ACC

kheireióteron worse.ACC

émmenai be.PRS.INF

‘for I say that there is no worse man than you’ [Hom., Il. 2.248–249] Non-stative present infinitives also occur (3% of all AcI clauses containing a present infinitive); compare Figure 1. This is the maximum share of non-stative

91

A diachronic perspective on the semantics of AcI clauses in Greek

present infinitives. Some of them are habitual and could be counted as stative (an example is Od. 1.189‒190). Frequencies of other infinitives in Homeric AcI clauses are presented in Table 1. The passages analyzed contain Hom., Il. 1–12, and Od. 1–12. Table 1: Aspect/tense frequencies in AcI clauses (Homer) PRS.INF.

PRF.INF.

AOR.INF.

FUT.INF.

75 41%

8 4%

27 15%

73 40%

As noted later, the frequencies of future and aorist infinitives in Table 1 may seem unexpected from the perspective of the situation in the language as a whole. However, later literary and non-literary texts largely display the same tendencies (cf. Figures 3 and 6). As expected, AcI clauses containing ATP infinitives are not attested in Homer because this development concerns later stages of AG (cf. above, Section 4).

7 From Classical Greek literary texts to the Greek of the first century AD This section examines AcI clauses in a corpus of AG literary texts, going back to the Classical and Hellenistic periods (fifth–second centuries BC). In Figures 1–5, these authors are arranged in chronological order. In a separate section (8) I discuss early Hellenistic non-literary documents (third century BC) because they are believed to provide more reliable insight into the developments of the spoken language.

7.1 AcI clauses containing present infinitives: stativity versus non-stativity As is shown in Figure 1, AG literary texts display a constant tendency towards the stativity of AcI clauses containing present infinitives. In most cases, this tendency is comparable with the state of affairs in Homer, although most (but not all) AG literary texts display somehow lower frequencies of AcI clauses containing the infinitive ‘to be’ or another stative infinitive (ST PRS), with nonstative present infinitives (NST PRS) becoming slightly more common. This applies to the majority of Attic writers (e.g., Sophocles, Euripides, Plato, and Menander).

92

Jerneja Kavčič

Figure 1: Present infinitives in AcI clauses (literary texts)14

Nevertheless, the difference between the earliest Classical and latest Hellenistic text (Herodotus–Polybius) seems to be minimal, with the share of AcI clauses containing non-stative present infinitives ranging from (at most) 20% in the earliest Classical texts (Sophocles [S.]) to 10% (Polybius [Plb.]). The share of non-stative present infinitives is lower in Homer, whose language is a more archaic form of Greek (based on an archaic form of Ionic, with elements of Aeolic dialect). Thus it can be argued that, in diachronic terms, AcI clauses containing present infinitives display a constant rather than increasing tendency towards stativity. The tendency towards stativity of AcI clauses containing a present infinitive also seems to be reflected in “hybrid” complement clauses, or passages in which a finite part of a complement clause alternates with an infinitive clause (cf. passage (5)). It does not seem surprising, in the light of the above findings, that the infinitive clause in the passage quoted is stative (‘that he is awkward’, with the infinitive eĩnai ‘to be’), whereas the finite part is non-stative (hóti egṑ mèn orthõs légō ‘that I speak correctly’). In addition, other “hybrid” cases of complement clauses cited in Jannaris (1968: 570) that contain a present infinitive 14 The texts examined include Herodotus IV; Sophocles, Antigone, Women of Trachis, Philoctetes, Oedipus at Colonus, Ajax, Electra, Oedipus Tyrannus; Thucydides II; Euripides, Children of Heracles, Alcestis, Cyclops, Medea; Aristophanes, Clouds, Acharnians, Knights, Wasps, Lysistrata, Thesmophoriazousai, Ecclesiazousai, Plutus, Frogs; Xenophon, Hellenica I–IV; Plato, Apology, Crito, Cratylus, Sophist; Demosthenes, On the Crown, Against Androtion; Aristotle, The Athenian Constitution, Poetics; Menander (all works and fragments); and Polybius, Histories. The abbreviations of AG authors follow LSJ and stand for: Hom. = Homer, Hdt. = Herodotus, S. = Sophocles, Th. = Thucydides, E. = Euripides, Ar. = Aristophanes, Xen. = Xenophon, Pl. = Plato, D. = Demosthenes, Arist. = Aristotle, Plb. = Polybius.

A diachronic perspective on the semantics of AcI clauses in Greek

93

display the same tendency: The infinitive clause contains eĩnai ‘to be’ (Th. 3.3) or another stative infinitive (boúlesthai ‘to want’, Th. 1.87; nomízein ‘to mean, think’, Xen., Cyr. 1.13; basileúein ‘to be king’ in Hdt. 3.75). Another passage contains a passive aorist infinitive (proapopemphthẽnai ‘to have been sent away’, Th. 3.25; cf. Kühner & Gerth 1904: 357). As far as the period between CG (fifth/fourth centuries BC) and the first century AD is concerned, it could be argued that AcI clauses containing stative present infinitives show a rather weak tendency to gradually prevail over those containing non-stative infinitives because AcI clauses containing non-stative present infinitives amount to 7% in the NT and to 8% in the contemporary nonliterary papyri (Kavčič 2009: 158; Kavčič 2017: 32), whereas the frequency of AcI clauses containing the infinitive eĩnai accounts for approximately 47% of all AcI clauses in the NT (Kavčič 2009: 155). Figure 2 shows the frequencies of AcI clauses containing the infinitive ‘to be’ and non-stative present infinitives (NST PRS) diachronically:

Figure 2: The infinitive “to be” and non-stative present infinitives in AcI clauses diachronically

7.2 Expressions of anteriority in AcI clauses occurring in AG literary texts As stated at the beginning, the aorist infinitive conveys anteriority in AcI clauses. The corpus investigated contains several examples of such use of the aorist infinitive; for example, Xen., Hell. 2.3.35, cf. (8):

94 (8)

Jerneja Kavčič

phēsì say.3SG

gár for

me me.ACC

toùs the

stratēgoùs generals

apokteĩnai kill.AOR.INF

‘for he says that I killed the generals’ [Xen., Hell. 2.3.35] Nevertheless, it seems that such clauses are less frequent than expected, as is claimed on the basis of Figure 3.

Figure 3: Aspect/tense frequencies in ACI clauses (literary texts)

First of all, Figure 3 shows that the aorist infinitive (AOR) was never very frequent in AG AcI clauses, its frequency in most authors being far below that of the present infinitive (PRS).15 As already mentioned, AcI clauses containing perfect infinitives are characterized as “rare” in Rijkbaron (2006: 98). If this view is adopted, AcI clauses containing an aorist infinitive also need to be interpreted as rare. Note that, in most authors, the perfect infinitive (PRF) is more common in AcI clauses or equally common as the aorist infinitive. Because all of the authors investigated are quite consistent in this respect, it could hardly be attributed merely to individual style. The data in Figure 3 contrast with the situation in the language as a whole. According to Duhoux (2000: 156), the perfect (PRF) was significantly less common in AG than the aorist (AOR), whereas the frequency of the aorist indicative was 15 Figure 3 presents AcI clauses containing aorist infinitives that convey anteriority. In addition to these, an aorist infinitive within an AcI clause can also refer to an iterative and/or potential event when accompanied by the particle án. These cases, very rare in my corpus, are not included in the figures because they would have no significant impact on the data. Moreover, they do not seem to be declarative AcI clauses.

A diachronic perspective on the semantics of AcI clauses in Greek

95

higher than the frequency of the present indicative (PRS). Although non-indicative aorist forms were somewhat less frequent than non-indicative present forms, this distinction was not as strong as within AcI clauses, where present infinitives are up to nine times more frequent than aorist infinitives; for example, in Plato (Pl., cf. loc. cit.). AG aspect/tense frequencies are given in Figure 4:

Figure 4: Aspect/tense frequencies in AG (Duhoux 2000: 155)

This is another argument in support of the assumption that AcI clauses containing aorist infinitives were relatively uncommon as early as the Classical period. As argued at the beginning, this tendency becomes particularly evident in the first century AD, when AcI clauses containing aorist infinitives are avoided because neither the NT nor the contemporary non-literary papyri contain convincing instances of aorist infinitives used in AcI clauses (cf. above, Section 3). The low frequencies of AcI clauses containing aorist infinitives could perhaps be accounted for if it is assumed that all AG infinitives – apart from the future infinitive – encoded aspect rather than relative tense. (As stated at the beginning, the issue of whether AG verb stems are aspectual or temporal has not been agreed upon.) If this view is adopted, the aorist infinitive appears to be uncommon in AcI clauses because – unlike the future infinitive – the aorist infinitive was not specialized for expressing temporality. Note that the future infinitive – which is generally believed to encode temporality – is more common in AcI clauses in the authors investigated than the perfect and the aorist infinitive. Judging from Figures 3 and 4, the future infinitive is more common within AcI clauses than is the future in the language as whole.16

16 According to de la Villa (1999: 363), the most common function of the aorist indicative was to convey relative anteriority in both Homer and in CG. Thus the function of the aorist in AcI clauses corresponds with its most common function in finite complement clauses, which is another reason why low frequencies of AcI clauses containing aorist infinitives call for an additional explanation. An explanation could perhaps be found in a comment made by an

96

Jerneja Kavčič

As stated above, it can be assumed that, as a consequence of the emergence and spread of the “anterior perfect,” which also tends to be transitive (cf. above, Section 4), the perfect infinitive may have tended towards adopting the function of conveying anteriority in AcI clauses. As a result, AcI clauses containing perfect infinitives are common in the first century AD, and aorist infinitives are avoided (cf. above, Section 3). Figure 5 sheds light on this issue from the perspective of the frequencies of passive (PASS), intransitive (INTR), and active transitive perfect (ATP) infinitives occurring in AcI clauses (cf. above, Section 5, on the term ATP).

Figure 5: Perfect infinitives in AcI clauses (literary texts)

First, Figure 5 shows that AcI clauses containing ATP infinitives are absent from the earliest works. This absence is expected because the emergence and spread of the ATP go back to later periods (cf. above, Section 4). In terms of other AG authors, the figure does not seem to provide conclusive evidence of diachronic tendencies in the use of the perfect infinitive in AcI clauses. It is not clear, for example, whether the use of the ATP increases or decreases: It has roughly the

anonymous reviewer of the study. It could be assumed that the AG verb system originally marked aspect and developed temporal properties at a later stage (see also Speyer this volume). (This is consistent with what was claimed above in Section 1.) From this perspective, the low frequencies of aorist infinitives in AcI clauses represent the earliest stages in the emergence of the temporal properties of AG infinitives. Perhaps the infinitive was pivotal in this development because – unlike in the case of the indicative – the augment as an explicit tense marker was not available. Thus in the case of the infinitive (and perhaps other verb forms apart from the indicative), the original aspectual properties were reinterpreted as temporal. In addition, if one assumes that AG infinitives encoded aspect rather than tense, the past reference of AcI clauses containing an aorist infinitive in AG could perhaps be accounted for in terms of the temporality of the main verb. This approach follows Stowell (1982) and is commented on in greater detail above in Note 2.

A diachronic perspective on the semantics of AcI clauses in Greek

97

same frequency in Polybius (Plb.), who is believed to represent the official koine of the second century BC (Horrocks 2010: 97), and in a much earlier author like Sophocles (S.), whereas it increases in Xenophon (Xen.). This is a significant observation in itself. First, the figure does not seem to indicate that, in diachronic terms, AcI clauses containing ATP infinitives tend to increase (at least in the period between the fifth and the second centuries BC). As stated, this could be an indication that perfect infinitives tended towards adopting the function of conveying anteriority (cf. above, Section 5). In addition, it does not seem that the spread of the ATP in the language as a whole led to a significant spread of AcI clauses containing perfect infinitives because the number of AcI clauses containing perfect infinitives remains fairly stable in diachronic terms in spite of the increase in the use of the ATP in the language as a whole (cf. above, Section 4).17 It could also be argued that the use of the ATP in AcI clauses was a matter of individual style because significant differences are found between the AG authors investigated. This interpretation of Figure 5 does not seem to support the assumption that (in Classical and Hellenistic authors) the ATP tended towards adopting the function of conveying anteriority – and, thus, perhaps replacing the aorist infinitive in AcI clauses – because such clauses should be understood as a phenomenon of grammar rather than style. The same conclusion can be drawn if diachronic tendencies in the use of perfect infinitives in AcI clauses are analyzed from the perspective of whether AcI clauses containing APAP (cf. above, Section 5) display increasing or decreasing tendencies. In this case as well, AcI clauses display no clear tendency. This issue is discussed in the following section.

8 AcI clauses in the earliest Hellenistic non-literary papyri (third century BC) The previous paragraph showed that, in AG literary texts, AcI clauses display only weak diachronic tendencies in terms of the parameters examined. This raises the question of whether or not the texts examined reflect the state of affairs in the literary language rather than in the spoken language. 17 This seems particularly obvious in the case of Menander (Men.), in which ATP infinitives are literally absent from AcI clauses. On the one hand, these data could be dismissed as insignificant because, in the case of this author, the figures above contain no more than two cases of AcI clauses containing perfect infinitives. It should be noted, however, that this study examines all of Menander’s existing works and fragments.

98

Jerneja Kavčič

The Classical period has so far yielded only a few sources providing insight into what could be considered the spoken language of the period. However, such sources become more frequent in the Hellenistic period (ca. third–first centuries BC) with the emergence of non-literary papyri and ostraca. AcI clauses are frequently attested in early Hellenistic non-literary papyri and ostraca dating back to the third century AD. In the corpus investigated,18 AcI clauses contain examples of all infinitives used in CG. Nevertheless, especially the AcI clauses containing aorist infinitives are slightly less frequent than in earlier authors. An example of such a clause is PSI 4.359 (cf. (9)), dating back to 251 BC:19 (9)

phēsì say.3SG autòn this kaì and

dè but tòn the

toũton this

toĩs the.DAT

Paú[eta(?)] Pauetas (?) ónon donkey

kaì and

tina some.ACC

misthōtòn servant.ACC

sákkous bags

diakhōrẽsai leave.AOR.INF

sullabṑn having.arrested

en in

Philadelpheíai Philadelphia

sunskeuasámenon having.packed.up

paradedōkénai give over.PRF.INF

phulakítais guards.DAT

‘Paveta (?) says that a servant had packed up this donkey and sacks and left, and that he arrested this man in Philadelphia and gave him over to the guards’ [PSI 4.359, 251 BC] 18 The corpus contains all Greek non-literary documents occurring in the Duke Databank of Documentary Papyri (DDbDP) that go back to the third century BC. This is an electronic database of published Latin and Greek documents written on papyri, ostraca, and other materials. AcI clauses occur in approximately 130 documents from the third century BC. 19 Phonological changes in Greek of the Hellenistic and Roman period (ca. third century BC – fourth century AD) also led to confusion between s-stem aorist infinitives (ending in ‑sai) and active future infinitives (ending in ‑sein; cf. Markopoulos 2009: 54; Lucas 2012: 203). As a consequence, an ‑s stem active aorist infinitive such as diakhōrẽsai in passage (10) could theoretically be interpreted as a future infinitive. Nevertheless, this does not seem to be the case because the infinitive diakhōrẽsai clearly refers to anteriority (and thus functions in its inherited role). In addition, Lucas (2012) cites a private letter dating back to a much later period (AD 1) in which the synthetic future is used “in exactly the same way as a synthetic future might have been used in CG” (Lucas 2012: 104). This passage indicates that distinctions between the AG synthetic future and aorist were retained (at least partially) even later than the third century BC. As consequence, it is not very likely that the aorist infinitive diakhōrẽsai in passage (10) above (as well as other s-stem infinitives from this period) was used mistakenly (instead of a future infinitive).

99

A diachronic perspective on the semantics of AcI clauses in Greek

Table 2 presents the frequencies of infinitives occurring in AcI clauses: Table 2: Aspect/tense frequencies in AcI clauses (early Hellenistic papyri) PRS.INF.

PF.INF.

AOR.INF.

FUT.INF.

126 50%

68 27%

8 3%

52 20%

Leaving aside earlier authors such as Menander, in whom the number of AcI clauses seems to be too small to draw any certain conclusions (cf. Note 17), Figure 6 shows that the frequency of AcI clauses containing the aorist (AOR) slightly decreases in the earliest Hellenistic non-literary documents, with the perfect (PRF) infinitive slightly increasing.

Figure 6: Aspect/tense frequencies in AcI clauses diachronically (Homeric–Hellenistic period)

This increase in the frequency of AcI clauses containing perfect infinitives in early Hellenistic papyri again raises the issue of whether or not this phenomenon is related to the perfect infinitive adopting the function of conveying anteriority in AcI clauses (cf. above, Section 5). On the one hand, it could be argued that in passage (9) the original functions of the aorist and the perfect infinitive are still retained: Whereas the aorist infinitive (diakhorẽsai) refers to an anterior act that has no consequences for the present moment, the perfect infinitive (paradedōkénai) refers to an act that has led to a present state, presumably to the servant still being kept by the guards. However, even if this explanation may sound plausible, it is less clear whether

100

Jerneja Kavčič

or not other perfect infinitives (e.g., pepoiēkénai in P.Col. 4.103, cf. (10)) are to be interpreted as containing a reference to a present state: (10)

egṑ I

dè but

ouk NEG

éphēn say.1SG.PST

pepoiēkénai do.PRF.INF

autõn of.that

‘I said I did none of these things’ [P.Col. 4.103, 275‒226 BC] As stated above (Section 5), this study sheds light on the function of perfect infinitives in AcI clauses from the perspective of diachronic tendencies instead of interpreting individual passages. Figure 7 shows that, within AcI clauses containing a perfect infinitive, AcI clauses containing ATP infinitives are more common in early Hellenistic non-literary papyri than in most AG authors (with the exception of Xenophon). Particularly in the light of an assumption that non-literary papyri provide reasonably good insight into the spoken language, these data can be considered highly significant. Note that Figure 6 shows that AcI clauses containing an aorist infinitive (AOR) become less common in the same period. They are less common in the non-literary papyri of the third century BC than in any literary text. As suggested above in Section 5, these two tendencies could be used to support the assumption that, in the early Hellenistic period, perfect infinitives tended towards adopting the function of expressing anteriority in AcI clauses. On the other hand, it has also turned out that, in diachronic terms, AG displays a constant tendency towards stativity of AcI clauses containing present infinitives (cf. above, Section 7.1). As a consequence, the relatively frequent usage of ATP infinitives (in the early Hellenistic papyri as well as in other sources) in AcI clauses could also be accounted for as a tendency towards the stativity of AcI clauses – if it is assumed that perfect infinitives, including ATP infinitives, are stative.

Figure 7: Perfect infinitives in AcI clauses diachronically (Homeric–Hellenistic Greek)

A diachronic perspective on the semantics of AcI clauses in Greek

101

The comparison between early Hellenistic papyri with the language of the first century AD shows that AcI clauses containing ATP infinitives are less common in the former than in the latter, reaching approximately the same frequencies in the first century AD as in the late Classical period (Demosthenes and Aristoteles). The share of AcI clauses containing ATP infinitives amounts to 28% of all AcI clauses containing perfect infinitives in the NT, and approximately 37% of such AcI clauses in contemporary non-literary papyri. (The percentage of the latter partly depends on whether all perfect infinitives are accepted as such or not; Kavčič 2017: 42.) This share is only around 3% in private letters, which are generally believed to reflect the state of affairs in the spoken language to a higher degree than official documents (cf. below, Figure 8). Even if it is assumed that perfect infinitives adopted the function of conveying anteriority in AcI clauses, the comparison between the early Hellenistic non-literary papyri and the language of the first century AD shows that this function was restricted by the first century AD. Diachronic tendencies in frequencies of AcI clauses containing ATP infinitives are shown in Figure 8.

Figure 8: AcI clauses containing ATP infinitives diachronically

As stated above in Section 5, it is assumed that active perfect forms of processes and actions (APAP) can convey anteriority in AG. Figure 9 shows APAP infinitives in AcI clauses.

Figure 9: AcI clauses containing APAP infinitives diachronically

102

Jerneja Kavčič

Figure 9 is similar to other tables showing frequencies of perfect infinitives in AcI clauses in that, regarding the texts investigated in this study, there seems to be no clear diachronic tendency (cf. above, Figures 3 and 5). If it is assumed that the perfect infinitive tended to replace the aorist infinitive in AcI clauses, AcI clauses containing APAP infinitives would be expected to display an increasing tendency (cf. above, Section 5). Although non-literary papyri of the first century AD in general do show such an increase, this does not seem to be a reflection of the spoken language because private letters of the same period display a significantly lower frequency of AcI clauses containing APAP infinitives. Another parameter examined, the stativity of AcI clauses containing present infinitives, shows a rather weak increase, only 9% of such AcI clauses being non-stative. This tendency seems to correspond with an assumption that, between the Classical and Hellenistic periods, AcI clauses containing a present infinitive display a slight tendency towards stativity increasing (cf. above, Section 7.1).

9 The aorist versus the perfect in conveying anteriority in AcI clauses Although some of the above data may seem inconclusive, they do support one conclusion more strongly than others. At some stages, both aorist and perfect infinitives were used in AcI clauses, in both literary and non-literary sources. Even if the issue of the exact meaning of the perfect infinitive is put aside, it appears that eventually only aorist infinitives were omitted from AcI clauses, whereas this was not the case with perfect infinitives. As stated, AcI clauses still contained active perfect infinitives in the first century AD, whereas aorist infinitives had been omitted (cf. above, Section 3). It is difficult to explain why only aorist infinitives should have disappeared from AcI clauses between the third century BC and the first century AD, whereas this was not the case with perfect infinitives, if it is assumed that the there was no distinction between the aorist and the perfect; in other words, that they had functionally merged. This observation supports the assumption that the perfect retained its (predominantly) stative meaning at least by the time of the omission of aorist infinitives from AcI clauses. Because AcI clauses containing aorist infinitives are still attested in the non-literary papyri of the third century BC (which are believed to be a more reliable source than literary authors), whereas such clauses disappear in the first century AD, this process must have been completed between the third century BC and the first century AD (leaving aside

A diachronic perspective on the semantics of AcI clauses in Greek

103

the second-century BC historian Polybius, whose work is literary and does not necessarily reflect the spoken language.) Regardless of the issue of when this process was completed, the assumption that the omission of the aorist infinitive from AcI clauses and (at the same time) the retention of perfect infinitives in AcI clauses both display a tendency towards omitting temporal distinctions and a tendency towards stativity of AcI clauses seems plausible.20 This conclusion is based on the assumption that the basic distinction between the aorist and the perfect infinitive (when used in AcI clauses) concerned the fact that, whereas the former referred to anterior events, the latter could also refer to present states.

10 Posteriority versus anteriority in AcI clauses It was claimed above that earliest signs of the retreat of the AG synthetic future infinitive go back to the fourth century BC, with the AG future infinitive not being replaced within AcI clauses with innovative future periphrastic forms. It is generally believed that the future infinitive retreated quickly from AG (cf. Markopoulos 2009: 53–54). This can be accounted for a tendency towards AcI clauses omitting temporal distinctions, an assumption that is congruous with Fykias & Katsikadeli (2013: 39–40): “[AG] infinitives are exceptional in the sense that they can be shown to be tensed, whereas infinitives in familiar modern language are not. [. . .] There are also good reasons for assuming that this noncanonical state of affairs [. . .] was in force for a relatively limited time span in the history of Greek.”21 There is a distinction between the omission of the future and the aorist infinitive from AcI clauses: Whereas the omission of the latter can be accounted for in terms of the general retreat of the AG synthetic future, this is not the case with the former because the position of the aorist in the AG verb system remained stable. As a consequence, this can be interpreted as another aspect revealing the tendency towards AcI clauses omitting temporal distinctions. 20 A more precise assessment of when the aorist infinitive was finally omitted from AcI clauses is beyond the scope of this study. Because of the morphological merger between the AG synthetic future forms and s-stem aorist forms that took place in the Hellenistic and/or Roman period, this is an intricate task (cf. above, Note 19; Lucas 2012: 103; Markopoulos 2009: 54). 21 Note, however, that English does not necessarily support the claim that infinitives in familiar modern languages are not tensed because English infinitives in have + perfect participle appear to mark tense; for example, in I consider him to have been a good friend at one time. Compare also Note 11.

104

Jerneja Kavčič

11 Insight into early Byzantine Greek Although both the infinitive and the AG perfect underwent a process of disappearance, AcI clauses are still attested in private letters of the late Roman / early Byzantine period (fifth and sixth centuries AD). Moreover, some of these contain perfect infinitives. These documents are highly significant because they are believed to reveal the state of affairs in the language of the period to a higher degree than other sources.22 Most of the AcI clauses occurring in these letters23 contain a present infinitive; an example is the infinitive ‘to be’ (eĩnai) in P.Gen. 4.172, cf. (11): (11)

éleges tell.2SG.PST

mḕ NEG

eĩnai be.PRS.INF

autḕn her.ACC

graũn old woman.ACC

‘you were telling me she was not old’ [P.Gen. 4.172, AD 301‒500] With one potential exception, all AcI clauses containing a present infinitive are stative.24 An AcI clause can also contain a perfect infinitive, whereas the only potential example of an aorist infinitive in an AcI clause is far from certain.25 The perfect infinitives in the corpus investigated are passive (e.g., peplērõsthai ‘to be paid’, ēdikẽsthai ‘to be done wrong’), with only one clear case of an APAP infinitive (sumbebēkénai ‘to have happened’) in P.Oslo. 2.64 (cf. 12): (12)

ḗkousa hear.1SG.PST

gàr for

e[n]taũtha here

lupēròn sad.ACC

prãgma thing.ACC

sumbebēkénai have happened.PRF.INF ‘for I heard that a sad thing has happened here’ [P.Oslo. 2.64, AD 401‒500] 22 This does not seem to apply to official documents. Mihevc (1959: 108) notes in the official non-literary papyri of this period the “clichés,” or AcI clauses containing perfect infinitives, that remain used in official writings although the corresponding perfect indicative forms seem to have already disappeared by then. Such cases can be found as late as in the seventh century AD (e.g., BGU 2.371). 23 In the period investigated, AcI clauses are attested in twelve letters. 24 Cf. P.Oxy. 1.128, P.Oxy. 17.2156, P.Oxy. 56.3867, P.Oxy. 58.3932, SB 24.16165, SB, 20.14118, CPR 25.13. The present infinitive occurring in these AcI clauses include: eĩnai ‘to be’ (two cases), ékhein ‘to have’, hugiaínein ‘to be healthy’, khrẽnai ‘to have to’, anékhesthai ‘to tolerate’, dúnasthai ‘can’, opheílein ‘to owe’. Non-stative is perhaps the infinitive epiphthánnein in P.Oxy 56.3867. 25 This infinitive is found in SB 9616 (proseltheĩn). However, the following lines of the text seem to indicate that the infinitive clause is not declarative. AcI clauses containing perfect infinitives occur in: P.Oxy. 56.3867, P.Oxy. 16.1869, P.Oslo. 2.64, SB 20.14523.

A diachronic perspective on the semantics of AcI clauses in Greek

105

If it is assumed that the aorist and the perfect had functionally merged by this time (cf. above, Section 4), passage (12) above is a very late (and rare) example of an AcI clause referring to an anterior event. As expected, AcI clauses containing the future infinitive are not attested in private letters of this period (cf. above, Sections 3 and 10). Although AcI clauses containing the perfect infinitive are still attested in this period, the evidence indicates that, in diachronic terms, there was a strong tendency toward AcI clauses expressing simultaneous states (and omitting temporal distinctions). Even if it is assumed that the perfect infinitive replaced the aorist infinitive in the function of conveying anteriority, AcI clauses conveying anteriority are very rare in this period. The three central parameters examined in this study are represented in Figure 10, which shows that, apart from the AcI clauses containing stative present infinitives, other parameters examined display decreasing tendencies from at least the first century AD.

Figure 10: AcI clauses containing stative present/aorist/ATP infinitives diachronically

12 Conclusions AG displays a relatively constant tendency towards the stativity of AcI clauses containing a present infinitive. The share of stative present infinitives in AcI clauses ranges between 80% and 90% in most CG authors, slightly increasing in earlier and later texts (Homer, NT Greek, and non-literary papyri of the first century AD). This tendency also sheds light on whether or not, in the earliest stages of Greek, a finite complement clause dependent on verbs of saying and thinking could be replaced by an AcI clause. It seems that such a replacement

106

Jerneja Kavčič

was likely to take place if the complement clause was stative. In terms of the disappearance of the infinitive from Greek, early Byzantine evidence suggests that AcI clauses containing stative present infinitives were the last to be omitted. It has been claimed that, already by the first century AD, AcI clauses were restricted to expressing statements about simultaneous states. Nevertheless, given that the NT and the contemporary non-literary papyri display a very common use of perfect infinitives within AcI clauses, the function of these infinitives needs to be examined before accepting this claim. It is clear that the AG synthetic future infinitive was gradually omitted from AcI clauses after the Classical period. However, the same period also displays a gradual omission of aorist infinitives from AcI clauses. Although AcI clauses containing aorist infinitives conveyed anteriority in the Classical period, this phenomenon does not necessarily mean that anteriority could no longer be expressed within AcI clauses. Instead one has to take into account that the past orientedness of the AG perfect became increasingly prominent in diachronic terms. Because this process concerns the Classical and subsequent periods, which also display gradual omission of aorist infinitives from AcI clauses, this could be an indication that the perfect infinitive substituted for the aorist infinitive in the function of conveying anteriority in AcI clauses. Although a few indications support this assumption in the early Hellenistic period, this cannot be claimed with certainty (it remains unclear whether the AG perfect is to be interpreted as an aspect or as a tense).26 Nevertheless, the diachronic perspective does shed light on one issue. It is difficult to explain why only aorist infinitives should have disappeared from AcI clauses between CG and the first century AD, whereas this was not the case with the perfect infinitive, if it is assumed that the two tenses functionally merged in this period. This observation seems to support the assumption that the perfect, at least the perfect infinitive, retained its (predominantly) stative meaning at the time of the omission of aorist infinitives from AcI clauses. As was claimed above, this process was completed between the third century BC and the first century AD. Thus, at least in the period before the omission of the aorist infinitive from AcI clauses, the use of the perfect infinitive in this construction can be accounted for as another sign of AcI clauses tending towards stativity. It appears that the construction investigated tended towards omitting temporal distinctions in later periods of Greek. Even if it is assumed that some perfect infinitives could convey anteriority, AcI clauses containing such infinitives display decreasing tendencies between the third century BC and the first century AD and become even less common in subsequent periods. 26 For my current opinion on this issue, see Kavčič (2016).

A diachronic perspective on the semantics of AcI clauses in Greek

General abbreviations ACC AcI Acts AG AOR APAP Ar. Arist. ATP CG D. DAT E. Hdt. Hell. Hell. Pap. Hom. HpMi. FUT Il. IMP IMPF IND INF Intr. Isoc. Lys. Men. Mk NEG NST NT Od. PASS PLPF PRF PL

accusative accusativus cum infinitivo Acts (of the Apostles) Ancient Greek aorist active perfect of actions and processes Aristophanes Aristotle active transitive perfect Classical Greek Demosthenes dative Euripides Herodotus Hellenica Hellenistic papyri Homer Hippias Minor future Iliad imperative imperfect indicative infinitive intransitive Isocrates Lysias Menander (Gospel of) Mark negation non-stative New Testament Odyssey passive pluperfect perfect plural

107

108 Pl. Plb. PRS PST Priv.Lett. Resp. Rom. Pap. S. SG ST Th. Xen.

Jerneja Kavčič

Plato Polybius present past private letters Res publica, Republic Roman papyri Sophocles singular stative Thucydides Xenophon

Papyri abbreviations BGU

Aegyptische Urkunden aus den Königlichen Museen zu Berlin: Griechische Urkunden I‒XX, eds. P. Viereck et al., Berlin 1895‒2014.

CPR

Corpus Papyrorum Raineri I‒XXXI, eds. C. Wessely et al., Vienna 1895‒2011.

P.Bad.

Veröffentlichungen aus den badischen Papyrus-Sammlungen I‒VI, eds. W. Spiegelberg et al., Heidelberg 1923‒1938.

P.Col.Zen.

Zenon Papyri: Business Papers of the Third Century B.C. Dealing with Palestine and Egypt I‒II, eds. W. L. Westermann et al., New York 1934‒1940.

P.Dura.

The Excavations at Dura-Europos: Final Report V, Part I: The Parchments and Papyri, eds. C. Bradford Welles, Robert O. Fink, & J. Frank Gilliam, New Haven 1959.

P.Gen.

Les Papyrus de Genève I‒IV, ed. J. Nicole et al., Geneva 1896‒2010.

P.Mich.

Papyri in the University of Michigan Collection I‒XX, eds. C. C. Edgar et al., Ann Arbor 1931‒2011.

P.Oslo.

Papyri Osloenses I‒III, eds. S. Eitrem & L. Amundsen, Oslo 1925‒ 1936.

P.Oxy.

The Oxyrhynchus Papyri I‒ LXXIX, eds. B. P. Grenfell et al., London 1898‒2014.

PSI

Papiri greci e latini I‒XV, eds. G. Vitelli et. al., Florence 1912‒2008.

SB

Sammelbuch gr. Urkunden aus Ägypten I‒XXVIII, eds. F. Preisigke et al., Berlin 1913‒2013.

A diachronic perspective on the semantics of AcI clauses in Greek

109

References Binnick, Robert I. 1991. Time and the Verb: A Guide to Tense and Aspect. Oxford: Oxford University Press. Blass, Friedrich, Albert Debrunner & Friedrich Rehkopf. 2001. Grammatik des neutestamentlichen Griechisch, 18th edn. Göttingen: Vandenhoech & Ruprecht. Burton, Ernest de Witt. 1898. Syntax of the Moods and Tenses in New Testament Greek. Edinburgh: T & T Clark. Chantraine, Pierre. 1927. Histoire du parfait grec. Paris: Honoré Champion. Chantraine, Pierre. 1953. Grammaire homérique, Tome 2: Syntaxe. Paris: Klincksieck. Comrie, Bernard. 1981. Aspect. Cambridge: Cambridge University Press. Cristofaro, Sonia. 2008. A constructionist approach to complementation: Evidence from Ancient Greek. Linguistics 46(3). 571–606. de la Villa, Jesús. 1999. L’indicative du passé dans les propositions complétives. In: Bernard Jacquinod (ed.), Les compétives en grec ancien, 353–365. Saint Étienne: Publications de l’Université de Saint Étienne. Duhoux, Yves. 2000. Le verbe grec ancient: Éléments de morphologie et de syntaxe historiques, 2nd edn. Louvain-la-Neuve: Peeters. Fanning, Buist M. 1990. Verbal Aspect in New Testament Greek. Oxford: Clarendon Press. Fykias, Ioannis & Christina Katsikadeli. 2013. The rise of ‘subordination’ features in the history of Greek and their decline: The ‘indirect speech traits cycle.’ Journal of Historical Linguistics 3(1). 28–48. Grevisse, Maurice. 1993. Le bon usage: Grammaire francaise, 13th edn. Paris: Duculot. Haspelmath, Martin. 1992. From resultative to perfect in Ancient Greek. Función 11−12. 187−224. Haug, Dag. 2008. From resultatives to anteriors in Ancient Greek: On the role of paradigmaticity in semantic change. In Thórhallur Eythórsson (ed.), Grammatical Change and the Linguistic Theory: The Rosendal Papers, 285−305. Amsterdam: John Benjamins. Horrocks, Geoffrey. 2010. Greek: A History of the Language and its Speakers, 2nd edn. Chicester: Wiley-Blackwell. Jannaris, Antonius Nicholas. 1968. An Historical Greek Grammar Chiefly of the Attic Dialect (Reprografischer Nachdruck der Ausgabe London 1897). Hildesheim: Georg Olms. Joseph, Brian D. 1983. The Synchrony and Diachrony of the Balkan Infinitive: A Study in Areal, General and Historical Linguistics. Cambridge: Cambridge University Press. Kavčič, Jerneja. 2009. A feature of Greek infinitive clauses dependent on verbs of saying and thinking. In Kateřina Loudová & Marie Žáková (eds.), Early European Languages in the Eyes of Modern Linguistics, 151–168. Brno: Masaryk University. Kavčič, Jerneja. 2016. The decline of the aorist infinitive in Ancient Greek declarative infinitive clauses. Journal of Greek Linguistics 16(2). 266–311. Kavčič, Jerneja. 2017. Variation in expressing temporal and aspectual distinctions in complement clauses: A study of the Greek non-literary papyri of the Roman period. In Klaas Bentein, Mark Janse & Jorie Soltic (eds.), Variation and Change in Ancient Greek Tense, Aspect and Modality (Amsterdam Studies in Classical Philology), 22–55. Leiden: E. J. Brill. Kühner, Raphael & Bernhard Gerth. 1904. Ausführliche Grammatik der griechischen Sprache, Zweiter Teil: Satzlehre, Band 2. Hannover: Hahnsche Buchhandlung. Kurzová, Helena. 1966. Zum Aussterben des Infinitivs im Griechischen. Les études balkaniques tchécoslovaques 1. 39–50.

110

Jerneja Kavčič

Kurzová, Helena. 1968. Zur Syntaktischen Struktur des Griechischen: Infinitiv und Nebensatz. Amsterdam: Hakkert. LSJ: Liddell, Henry George, Robert Scott & Henry Stuart Jones. 1996. A Greek-English Lexicon, with a revised supplement. Oxford: Clarendon Press. Lucas, Sandra. 2012. Polarizing the Future: The Development of an Aspectual Opposition in the Greek Future Tense. Copenhagen: University of Copenhagen dissertation. Markopoulos, Theodore. 2009. The Future in Greek: From Ancient to Medieval. Oxford: Oxford University Press. McKay, Kenneth L. 1980. On the perfect and other aspects in the Greek non-literary papyri. Bulletin of the Institute of Classical Studies 27. 23–50. McKay, Kenneth L. 1981. On the perfect and other aspects in New Testament Greek. Novum Testamentum 56(4). 289–328. Mihevc, Erika. 1959. La disparition du parfait dans le grec de la basse époque. Razprave SAZU, razred za filološke in literarne vede V. 93–154. Ljubljana: Slovenska akademija znanosti in umetnosti. Miller, Gary D. 2002. Nonfinite Structures in Theory and Change. Oxford: Oxford University Press. Moser, Amalia. 2003. Tense, aspect and the Greek perfect. In Artemis Alexiadou, Monika Rathert & Arnim von Stechow (eds.), Perfect Explorations, 235–252. Berlin: Mouton de Gruyter. Pollock, Jean-Yves. 1980. Verb movement, Universal Grammar, and the structure of IP. Linguistic Inquiry 20(3). 367–424. Porter, Stanley E. 1989. Verbal Aspect in the Greek of the New Testament with Reference to Tense and Mood. New York: Peter Lang. Rijksbaron, Albert. 2006. The Syntax and Semantics of the Verb in Classical Greek, 3rd edn. Chicago: University of Chicago Press. Ruijgh, Cornelius J. 1999. Sur l’emploi complétif de l’infinitif grec. In Bernard Jacquinod (ed.), Les compétives en grec ancien, 215–231. Saint Étienne: Publications de l’Université de Saint Étienne. Scheppers, Frank. 1999. Morphosyntaxe et morphosémantisme des constructions complétives: domains autonomes – domains médiats. In Bernard Jacquinod (ed.), Les compétives en grec ancien, 277–289. Saint Étienne: Publications de l’Université de Saint Étienne. Schwyzer, Eduard. 1953. Griechische Grammatik, Erster Band: Allgemeiner Teil, Lautlehre, Wortbildung, Flexion. 2nd edn. Munich: C. H. Beckʼsche Verlagsbuchhandlung. Sevdali, Christina. 2007. What focus can do for subjects. Proceeding of the 7th International Conference on Greek Linguistics. http://icgl7.icte.uowm.gr/Sevdali.pdf. (accessed 13 November 2014). Sevdali, Christina. 2013. Ancient Greek infinitives and phases. Syntax 16(4). 324–361. Sicking, Christiaan Marie Jan & Peter Stork. 1996. Two Studies in the Semantics of the Verb in Classical Greek (Mnemosyne, Bibliotheca Classica Batava: Supplementum centesimum sexagesimum). Leiden: E. J. Brill. Skytte, Gunver. 1983. La sintassi dell’infinito in italiano moderno. Copenhagen: University of Copenhagen. Stockwell, Robert P., Paul Schachter & Barbara Hall Partee. 1973. The Major Syntactic Structures of English. New York: Holt, Rinehart & Winston. Stowell, Tim. 1982. The tense of the infinitive. Linguistic Inquiry 13(3). 561‒570.

A diachronic perspective on the semantics of AcI clauses in Greek

111

Ter Beek, Janneke. 2010. Two futures in infinitives. In C. Jan-Wouter Zwart & Mark de Wries (eds.), Structures Preserved: Studies in Syntax for Jan Koster, 41−48. Amsterdam: John Benjamins. Thorley, John. 1989. Aktionsart in New Testament Greek: Infinitive and imperative. Novum Testamentum 31. 290–315. Wackernagel, Jakob. 1904. Studien zum griechischen Perfectum. In Programm zur Akademischen Preisverteilung 1904, 3–23. Göttingen. (Reprint in: Jakob Wackernagel, Kleine Schriften, Göttingen: Vandenhoek & Ruprecht, 1953.)

II Infinitive structures versus other non-finite and finite patterns

Frauke D’hoedt and Hubert Cuyckens

5 Finite, infinitival and verbless complementation: The case of believe, suppose and find 1 Introduction A central issue in studies of clausal verb complementation is the so-called “matching problem” (Noonan 2007: 101; De Smet 2013: 19). As Noonan (2007: 101) puts it, “[c]omplementation is basically a matter of matching a particular complement type to a particular complement-taking predicate”. In most cases, one complement-taking predicate (CTP), or matrix verb, combines with one clausal complement type (CC-type). As such, say patterns with the that-clause, plan with the to-infinitive clause, and avoid with the gerundial -ing-clause. In a number of cases, though, a particular CTP may combine with two (or more) CCtypes, as can be seen from the CC-patterns with remember, intend, and believe in (1)–(3): (1)

(2)

(3)

a.

He remembered to thank her for everything.

b.

I remember reading about it in the newspaper. (Examples taken from Declerck 1991: 511)

a.

. . . but my Health is so much mended by them that I intend to stay some time longer.

b.

I intend staying here for the remainder of this month. (Examples taken from Fanego 1996; quoted in De Smet 2013: 27)

a.

I believe that he is a wise man.

b.

I believe him to be a wise man.

Each of the [CTP–CC] combinations in the alternation patterns above can be independently motivated. However, while the variation in (1) can be characterized as functional differentiation, the patterns in (2)–(3) are functionally equivalent. Indeed, the semantics of the CTP (‘bringing to mind’) in (1a) is compatible with Frauke D’hoedt and Hubert Cuyckens, KU Leuven DOI 10.1515/9783110520583-005

116

Frauke D’hoedt and Hubert Cuyckens

the future event/plan of action expressed by the to-infinitive; at the same time, this semantics is also compatible with an anterior event encoded by the gerundial -ing-clause (1b). Given that the to-infinitive and gerundial -ing-clause in (1) with remember encode events bearing a different temporal relation with the CTP (posterior vs. anterior), these two CC-types can be seen as making up a contrastive pair. In alternation pair (2), then, the to-infinitive with intend is motivated in that “intend is a verb of volition and therefore readily combines with a complement type expressing potentiality”; the gerundial -ing-clause is motivated because intend “is semantically compatible with actions controlled by the matrix subject” (De Smet 2013: 27–28). Unlike in (1), these two motivations are not in conflict, and the different CC-types with intend can be seen as functionally equivalent (they serve “no important communicative purpose” [De Smet 2013: 28]). Similarly in (3), while the different CC-types can be said to encode different construals (or can each be linked with a different motivation), they are communicatively equivalent. In Langacker’s (2008: 440) Cognitive Grammar framework, the to-infinitive with intend in (2) can be said to construe an occurrence lying at the end of a temporal path (seen from the temporal vantage point of the CTP) and which can thus be viewed in its entirety, while the gerundial -ing-clause encodes a shifted vantage point to the time of the future occurrence, “thus fictively affording an internal perspective on it” (Langacker 2008: 440). On this account as well, the to-infinitive and gerundial -ing-clause are functionally equivalent because both CC-types express an event pertaining to the future. Further, in Langacker’s (2008: 438–443) terms, the matrix subject in (3a) adopts the epistemic value (in particular, the validity) of the proposition (or grounded process) encoded in the finite that-complement, while in (3b), the matrix subject commits him-/herself to the occurrence of the process encoded by the to-infinitive. The expressions are functionally equivalent, however, because the proposition’s validity hinges on its occurrence (see Langacker 2008: 443). It is the type of variation exemplified by (2) and (3) that the present paper will be concerned with: While the CC-types can each be independently motivated, thus signaling different construals or differences in the language user’s mental grammar, the variants appear to be functionally equivalent. This, then, leads to an indeterminacy in actual usage, at the level of online discourse; in other words, preference for one or the other CC-type is not categorical (with a clearcut differentiation between the domains of operation of CC-types) but rather probabilistic. The purpose of this paper is to offer a corpus-based analysis of probabilistic complement-clause variation from Middle English (ME) to Present-day English (PDE) – with a focus on Late Modern English (LModE) and Present-day English –

Finite, infinitival and verbless complementation

117

as well as examining what factors co-determine this variation at the level of usage or on-line discourse. Earlier studies within this perspective have considered the variation between finite that-clauses and (mainly) non-finite -ing gerundial clauses with the CTPs remember, regret, deny, and admit (see Cuyckens, D’hoedt, & Szmrecsanyi 2014; Cuyckens & D’hoedt 2015). In the present paper, we wish to focus on alternation patterns including the “Small Clause” (SC), a verbless clause which is considered to be clausal because “it has the conventional subject/ predicate geometry and semantic predication relation” (Miller 2002: 136); an example is I believed him intelligent (see also Section 2.2). In particular, then, we will examine CC-variation between the finite that-clause (including the zerocomplementizer clause), ECM/VOSI complementation (Los 2005) containing the to-infinitive, and the Small Clause (SC), with the matrix verbs believe, suppose, and find.1 Examples are in (4): (4)

a.

I believe that he is a wise man. (that-clause)

b.

I believe he is a wise man. (zero-complementizer clause)

c.

I believe him to be a wise man. (to-infinitive; ECM-clause)

d.

I believe him a wise man. (SC)

The remainder of this paper is organized as follows. Section 2 will offer the background to our study and offer a survey of clausal verb complementation (including the verbless Small Clause-complement). Section 3 sets out the goals, while Section 4 details data and methodology, in particular the coding of the data that will feed into our binary logistic regressions. Section 5, then, examines several CC-alternation patterns as well as the factors predicting preferences for particular variants. Section 6 discusses how these results might bear on broader linguistic issues. Section 7 summarizes our findings and offers a conclusion.

2 Setting the scene In this section, we will first present a short, selective survey of research on (variation in) clausal verb complementation. We will then turn to a characterization of 1 Our analysis is not concerned with the question whether sentences such as (4c) might be derived from sentences such as (4a) (through Raising; see Postal 1974 and Bresnan 1976 for a critical review). Rather, we consider the that-clauses and to-infinitive clauses under investigation to have derived from diachronically different sources, but synchronically to make up sets of variants.

118

Frauke D’hoedt and Hubert Cuyckens

Small Clauses, as this is the type of clausal verb complement whose alternation with other CC-types has not yet received close attention.

2.1 A short survey of clausal verb complementation Clausal verb complementation has been an important research topic within generative as well as cognitive-functional linguistic frameworks. Indeed, complementation phenomena have been a concern of generative linguists (Rosenbaum 1967; Bresnan 1970) since Chomsky’s seminal Aspects of the Theory of Syntax (1965), and important work in this domain has continued ever since (Bresnan 1979; Warner 1982; Chomsky 1986; Rizzi 1990; Radford 1997; Felser 1999). From the 1980s onwards, the scope of complementation research was expanded by functional-typological linguists (e.g. Givón 1980; Noonan 2007 [1985]; Dixon 1991). While the generative literature on verb complementation has largely focused on synchronic, syntactic issues, such as the constituent structure of different complement types, synchronic research within the cognitive-functional tradition to date has been semantic in orientation and has been concerned with the question of how complement clauses are distributed over the various complement-taking predicates (see Wierzbicka 1988; Duffley 1992, 1999; Langacker 1991; Achard 1998; Smith 2002; Noonan 2007). However, as De Smet (2013: 20‒ 33) points out, a satisfactory synchronic account of complementation needs to envisage additional, non-semantic determining principles, such as the role of information structure (Noël 2003), the horror aequi principle (Rudanko 2000; Vosberg 2003), the cognitive complexity principle (Rohdenburg 1995), social and regional stratification (Mair 2002, 2003), and register (Mindt 2000). Significantly, though, this synchronic work has tended to neglect the fact that the synchronic matches between the complement-taking predicate and the complement clause can be subject to change over time. It is only in the last twenty years or so that diachronic studies have appeared which present broad accounts of change and variation in complementation patterns in different periods of the history of English (see Fischer 1995; Fanego 1996, 1998; Rudanko 1998, 2006, 2010, 2012; Miller 2002; Los 2005; Rohdenburg 2006; De Smet 2008, 2013), attesting to a distributional reorganization of CCs over time.2 It has thus been shown that new complement types have made their way into the English language (e.g. the gerundial -ing-CC and the for . . . to-infinitive construction are relatively recent), or that long existing types have spread to new CTP-contexts (e.g. the to-infinitive) or, conversely, have become increasingly restricted (e.g. the bare infinitive). 2 Most linguists, also of the generative persuasion, are now agreed that the diachronic perspective is an important facet of a satisfactory account of complementation.

Finite, infinitival and verbless complementation

119

These changes have led to competition between CC-types; in particular, in some CTP-contexts, one CC-type may have been replaced by another (e.g. with verbs of volition, the that-clause has largely been supplanted by the to-infinitive; see Croft 2000, Los 2005 as well as Jędrzejowski & Demske this volume for a diachronic-typological perspective); in other contexts, a situation of variation may have come about. Sometimes, this CC-variation may be categorical in nature, as in (1), or it may be non-categorical or probabilistic, as in (2)–(3). In the latter case, the variation is unstable in that CC-types may co-exist and show varying/ shifting patterns of preference across speakers and over time. While replacement of CC-types as well as CC-variation characterized by functional differentiation have been well documented, the type of competition between CC-types resulting in unstable variation (or changing patterns of preference over time) has received little to moderate attention. The (change in) variation between that-CCs and gerundial -ing-CCs has only been discussed recently in studies by Cuyckens, D’hoedt & Szmrecsanyi (2014), Rohdenburg (2014), and Cuyckens & D’hoedt (2015); fair attention has been given to the variation in Present-day English between the that-CC and the ECM construction (see, for instance, Mair 1990; Dixon 1991; Noël 1997, 2003); then again, the variation between complementation by Small Clauses and its verbal counterparts has largely been left unexplored. It is the aim of this paper to integrate the (non-verbal) SC-complement into the study of non-categorical CC-alternation.

2.2 Small Clauses (SCs) In the formal syntactic literature, small clauses (SCs), as in (5)–(6), are reduced clauses which involve a subject-predicate relation between an NP and a phrasal projection XP (see Aarts 1992; Haegeman & Guéron 1999: 109; Basilico 2003). (5)

We consider [[prompt action]NP [invariably better than quiet reflection]XP]SC . (PPCMBE, 1908)

(6) I will prove [[you]NP [the notoriousest Traitor that ever came to the Bar]XP]SC . (PPCEME2, 1570–1639) The XP-component typically consists of an AP, NP, or PP, as in (7). (7)

a.

XP = AP:

Mike considers Sue intelligent. (Aarts 1992: 21)

b.

XP = NP:

I declared Jo the winner. (ibid.)

c.

XP = PP:

I want the dog out of the house. (ibid.)

120

Frauke D’hoedt and Hubert Cuyckens

Sometimes the XP can be replaced by as or so, as in He was called so or (. . .), as he was called. Although it has been argued that constructions where the predicative relation between the NP and XP is dynamic in orientation should be labeled SCs as well (Citko 2011: 749),3 we will only consider those constructions as SCs where the predicative relation is of a stative nature.4 This includes examples like (8a), but not (8b), where the relation between NP and XP is dynamic: (8)

a.

I consider Mary smart. (Citko 2011: 749)

b.

Mary heard John leave. (ibid.)

So far, SCs have been approached mainly as a unitary, synchronic construction (but see Miller 2002: 164–172), and where they are further analyzed it is with a predominant focus on their internal structure (Williams 1980, 1983; Hoekstra 1988; Citko 2011) and their syntactic classification (Hoekstra 1988; Basilico 2003).5 This has left the status of the SC in the complementation system and its similarities with other complement types largely underexplored.

3 Goals In this paper, we wish to examine non-categorical or probabilistic variation with the verbs believe, suppose, and find from Middle English to Present-day English. The focus will be on the alternation between the following CC-types: that-clause, 3 The following examples taken from Citko (2011: 749) are representative of constructions that have been claimed in the literature to involve SCs (see Citko 2011 for a more exhaustive list): (i) (ii) (iii) (iv) (v) (vi)

Mary is smart. → copular constructions Mary seems smart. → raising verbs Mary pushed the door open. → resultatives Mary turned the TV off. → verb-particle constructions With Mary absent, we had to cancel the meeting. → absolute constructions Mary made John leave. → causatives

4 The rationale behind this is that not all CTPs under discussion allow dynamic SCs, so including dynamic SCs would limit comparability across CTPs in this paper. For instance, while find can occur with a dynamic SC in some contexts (i), suppose in those contexts is ungrammatical (ii). (i) I found him lying on the floor. (ii) *I suppose him lying on the floor. 5 The discussion on SCs’ internal structure mainly concerns the issue whether the [NP–XP] combination functions as a single constituent or not, and whether the NP should be awarded subject-(like) status; the debate on syntactic classification concerns the word order of the SCcomponents and which word classes are allowed in each SC-slot.

Finite, infinitival and verbless complementation

121

zero-complementizer clause (zero-complement, for short), ECM/VOSI-complement (hence referred to as to-infinitive), and SC-complement. Sentences (9)–(11) provide examples of each of these CC-types per verb: (9)

(10)

(11)

believe that: Well, the President believes that women are playing a more a. and more important part in influencing opinion. (Wordbanks Online, Colin Forbes, Cover Story, 1986) b.

zero: The consul received some affront during the dispute; but I believe he succeeded in making an amicable arrangement between the parties. (PPCMBE, Montefiore, 1836)

c.

to-infinitive: We all radiate something curiously intimate when we believe ourselves to be alone. (CLMET, Edward Morgan Forster, Where Angels Fear to Tread, 1879)

d.

SC: I had come to that love affair as a boy without confidence, who believed himself ugly and unworthy, and was pining for a father in a distant city. (Wordbanks Online, Fergal Keane, All of These People, 2005)

suppose that: I suppose that Mr. Ady and his Family set out from a. Greenwich upon Wednesday morning in two Coaches. (PPCEME2, Hoxinden, 1660) b.

zero: I suppose I ’d had maybe four pints, maybe five . . . and a couple of vodkas. (Wordbanks Online, Sean Thomas, The Cheek Perforation Dance, 2002)

c.

to-infinitive: Shall we suppose these men, tho they see, to be blynde? (PPCEME2, Boethel, 1593)

d.

SC: (. . .) some person entered, out of breath and laughing! My anger was greater than my astonishment for a minute. I supposed it one of the maids, (. . .). (CLMET, Emily Brontë, Wuthering Heights, 1818)

find a. that: You may also find that the hours or rates of pay are not as good as those applying to men doing the same job. (Wordbanks Online, Lee Rodwell, The Single Woman’s Survival Guide, 1989)

122

Frauke D’hoedt and Hubert Cuyckens

b.

zero: When he entered after a three-month siege, Cortés found 160 000 had died of smallpox, including the emperor, Montezuma. (Wordbanks Online, Jim Leavesley & George Biro, The Medical Mysteries, 2001)

c.

to-infinitive: Mr. Boswall was positive not to go to excess. I found him to be pretty opinionative, and hasty in his temper. (PPCMBE, Boswell, 1776)

d.

SC: I had not found it easy to become a student again, and greatly missed Africa and the life I had known there. (Wordbanks Online, Eric Fisher, Healing Miracles, 1993)

As our concern is with on-line choices speakers make (i.e. the choices in actual language usage), we will first explore the patterns of CC-preference (in terms of frequency distributions) that emerge from actual usage as observed in the corpora (see Section 4); second, by fitting a logistic regression model, we aim to isolate the factors significantly predicting this CC-variation, examine whether any factors have changed over time, and explore any differences in the CCbehavior with the various CTPs. In particular, the following semantic, structural, and language-external factors will be considered which have been hypothesized to predict/correlate with speakers’ on-line choices: (i) the semantics of the CTP, (ii) the temporal relation between the matrix clause and the complement clause (see also Kavčič this volume and Speyer this volume), (iii) the structural complexity of the complement clause, (iv) the complexity of the CC subject, (v) the voice of the verb in the CC, (vi) intervening material between the CTP and the CC subject, (vii) the period of the attestation and (viii) the corpus (see Section 4). Finally, we will explore how our findings feed back into the existing literature. More specifically, we will consider how they bear on the following general linguistic issues: (i) the CC-specificity of CTPs (Noonan 2007: 101), semantic specialization, Rohdenburg’s (1996) Cognitive Complexity principle, and “the failure of the to-infinitive”.

4 Data and methodology 4.1 Data selection We based our selection of CTPs on Quirk et al.’s list of factual verbs (1985: 1181).6 From this list, we selected the verbs allowing finite and non-finite, to-infinitive complementation (a verb which could not be taken into consideration, for instance, was hope, because it allows a that-CC or a zero-CC (I hope (that) he’s 6 Quirk et al. (1985) also provide a list of suasive and emotive verbs taking finite as well as non-finite complement clauses; these verbs will not be taken into consideration here.

Finite, infinitival and verbless complementation

123

happy), but not a to-infinitive or an SC (*I hope him to be happy)). We then isolated those verbs that, in addition to the finite/to-infinitive alternation, allow for verbless, SC-complementation as well. After an exploratory search, we selected three verbs that occurred sufficiently frequently in order to guarantee statistically significant results; these are: believe, find, and suppose. Relevant data were extracted from the following corpora: – The suite of Penn Corpora of Historical English, which includes the PennHelsinki Parsed Corpus of Middle English, third edition (PPCME3), the PennHelsinki Parsed Corpus of Early Modern English, second edition (PPCEME2), and the Penn Parsed Corpus of Modern British English (PPCMBE); they cover the periods from ME to LModE (i.e. from 1150 to 1914) and comprise a wide variety of written genres. – The Corpus of Late Modern English Texts (CLMET3.0), a collection of texts drawn from the Project Gutenberg and the Oxford Text Archive and spanning the period from 1710 to 1920; data from this corpus complement LModE data from the PENN corpora. The CLMET texts are comparable to the PENN-texts in terms of genre, origin, and register, but show no actual overlap. – Wordbanks Online; we restricted our search to the British Books-component to ensure comparability with the other corpora. As these texts represent PDE-language, they nicely complement the periods covered in the PENN and CLMET3.0-corpora. We extracted all attestations of the matrix verbs and afterwards manually selected relevant hits. Samples were taken from the CLMET3.0 corpus and Wordbanks Online to keep the number of attestations for coding manageable.7 In selecting relevant attestations, care was taken that for each sentence instantiating one of the CC-types, potential correlates exist in the language which instantiate the other CC-types. Indeed, since we are examining non-categorical or probabilistic variation, that is, speaker preferences for one or the other CC-types, potential variants need to be present in the language as well.8 In particular, the following types of attestations were not selected: 7 For the CLMET, the raw data were first manually pruned to retrieve the relevant attestations, and a sample of 1500 hits per verb was taken from these relevant attestations afterwards. For Wordbanks Online, samples of 1000 were taken for believe and suppose and a sample of 1500 was taken for find, all of which were afterwards manually pruned. The reason why samples were taken first in the case of Wordbanks Online was because this corpus yielded too many raw data, so that manual pruning was only manageable after sampling; also, a larger sample was taken for find than for believe and suppose since find occurs more often without complements, so a larger sample needed to be obtained to end up with a number of relevant hits that was comparable to that of suppose and believe. 8 In Noël (1997: 272), it is stated that the to-infinitive complement making up an ECM is incompatible with nonperfective dynamic verbs. While they may not be frequent, examples such as the following are attested:

124 –

–

Frauke D’hoedt and Hubert Cuyckens

CCs describing an event posterior to the time of the CTP; the reason is that ECMs typically do not express posterior temporal relations, as in (. . .) and a mathematician no sooner finds that he shall often have occasion to speak of the same two things together, than he at once creates a term to express them whenever combined (CLMET, John Stuart Mill, 1843);9 passive ECM-clauses of the type He is supposed to leave – these have a deontic interpretation, and have no correlate in the finite CC-types; similarly, ECM-clauses such as He is believed to be innocent – these have an evidential meaning, which can be captured by ‘the community believes that. . . ; there is evidence that. . .’, and which is not present in finites such as We believe (that) he is innocent.

Table 1 below presents an overview of the number of coded attestations per verb for each corpus. Table 1: Number of attestations per verb, per corpus Verb

Corpus

# of Attestations

Believe

PENN corpora CLMET3.0 Wordbanks Online

462 1330 479 Total: 2271

Find

PENN corpora CLMET3.0 Wordbanks Online

1139 1475 465 Total: 3079

Suppose

PENN corpora CLMET3.0 Wordbanks Online

370 1323 449 Total: 2142

Observations were coded as described below. (i)

Most interesting is that Obama is such a clean slate; you could, in your own imagination, believe him to support anything you want. (http://greenarrowguide.blogspot.be/2008_ 02_01_archive.html; last accessed on 21 December 2014) (ii) Take a sword there; and put on any armor you can find to fit you. (CLMET, Bernard Shaw, 1912) (iii) the air of constant gaiety and good spirits (. . .) which she supposed to arise from content as constant (. . .) (CLMET, Ann Radcliffe, 1826) 9 These attestations were removed from the dataset in a second round of pruning, after the sampling and first round of manual pruning described in the previous footnote.

Finite, infinitival and verbless complementation

125

4.2 Coding of the data Each corpus attestation was entered into an Excel table. Each attestation was then coded for a number of factors determining CC-choice and describing characteristics of the CTP and the CC. Our selection of potentially significant factors was informed by the relevant literature (see Section 2; De Smet 2013: 20–33) as well as by the factors employed in our earlier studies (Cuyckens, D’hoedt, & Szmrecsanyi 2014; Cuyckens & D’hoedt 2015); it comprises semantic, structural, and periodization-related factors. 4.2.1 Semantics of the CTP (SEMANTICS )10 In the Oxford English Dictionary, believe followed by a that-, zero-, to-infinitive, or SC-complement is claimed to have the following meaning: 4.a. trans. With that clause, or with simple object and infinitive clause or complement: to consider to be true; to have as an opinion, think.

For the CTP believe, then, two values (senses) are distinguished. In its first sense, labeled as SEMANTICS _ lexical, believe denotes ‘to consider to be true’, ‘having a conviction/having faith in’; in this sense, the CTP retains its full lexical meaning indicating a mental (or, cognitive) process of believing, and it expresses the main point of an utterance (Boye & Harder 2007: 577–578), as in (12): (12)

They believed the Bible was the word of God; and, when they practised its precepts, they found, by the good that came from them, that it was truly so. (CLMET, John Galt, Annals of the Parish, 1821)

In its second sense, SEMANTICS _bleached, believe is weakened to ‘to have as an opinion, to think’ and often expresses the speaker’s epistemic stance, i.e. the speaker’s assessment of his/her degree of certainty with respect to the complement, as in (13). Here, the main clause acquires an inherently, “parenthetical” secondary discourse or usage function (Boye & Harder 2007: 577): (13) The kind de Brz wrote again on the 1 November, ‘I believe, madame, that the King will find her as pleasing and as much to his fancy as all those who have seen her and found her pretty and of clever wit’. (Wordbanks Online, Jane Dunn, Elizabeth and Mary: Cousins, Rivals, Queens, 2003)

10 The labels of the factors, as used in the regression tables, have each time been mentioned in parentheses.

126

Frauke D’hoedt and Hubert Cuyckens

A similar dichotomy is found with suppose, which may have a full lexical meaning of ‘conjecturing, assuming as true”, as in (14a), or a more bleached, epistemic sense, as in (14b): (14)

a.

Upon this the slaves left Syracuse, where we must suppose that the governor was then residing. (PPCMBE, Long, 1866)

b.

I dare say I lost it at the Opera, before we came on here. Ah yes, I suppose it must have been at the Opera. (PPCMBE, Wilde, 1895)

The CTP find, then, may be used as a verb of perception/empirical observation, where the subject may literally find something by seeing, hearing, or stumbling across it (15a), or may “find” something metaphorically through questioning/ examination or by figuring something out (15b); this sense is labeled SEMANTICS _ perception. What characterizes this first sense, as in (15 a+b), is the element of “newness”, as the subject acquires information that was previously unknown to him/her. (15)

a.

Upon arrival at the pool I found the elephant so deep in the mud that he could barely move. (CLMET, Sir Samuel White Baker, The Rifle and the Hound in Ceylon)

b.

I was desired to call on a young Gentlewoman dangerously ill. But I soon found, she needed no Physician for her Soul, being full of Righteousness and Good Works. (PPCMBE, Wesley, 174x)

By contrast, find is also used as mental verb/verb of cognition (SEMANTICS _ mental), and is then similar to ‘consider, think’. In these cases, the complement clause represents a certain opinion on the part of the matrix subject that is obviously not new to him/her, as in (16): (16)

a.

The Mrs Finches were afraid you would find Goodnestone very dull. (PPCMBE, Austen, 180x)

b.

(. . .) if anyone finds that he can see nothing to admire and love in Virgil, it is probably he and not Virgil that needs to be changed. (PPCMBE, Benson, 1908)

With all three verbs, we were confronted with attestations whose reading was ambiguous (that is, between epistemic and lexical for believe and suppose, or between mental and perception for find). These ambiguous cases have not been included in the regression analysis.

Finite, infinitival and verbless complementation

127

4.2.2 The temporal relation between the matrix clause and the complement clause (TEMPORAL _ RELATION ) The temporal relations coded for are simultaneity (TEMPORAL _ RELATION _ simultaneity) and anteriority (TEMPORAL _ RELATION _anteriority) between the time of the CTP and that of the complement clause, as in (17) (as mentioned earlier, relations of posteriority are typically less compatible with ECM-complements): (17)

a.

General Tindall believes that Corolini and his Our Country Party are being funded with Old Kingdom gold, though he cannot definitely prove it. (Wordbanks Online, Garth Nix, Lirael: Daughter of the Clayr, 2001)

b.

(. . .) they suppose themselves to have attained the highest perfection of which their nature is susceptible. (PENN, Froude, 1830)

Typically, SC-complements denote a relation of simultaneity with the matrix clause, although exceptions are attested: (18)

a.

Is it possible to believe our ancestors guilty of so absurd a resolution? (= to believe that our ancestors were guilty) (CLMET, Charlotte Lennox, The Lady’s museum, 1730)

b.

(. . .) it seemed cruel to abandon her in such a distressful situation; to leave her unprotected, in the hands of a husband who believed himself wronged by her. (= who believed that he had been wronged by her) (CLMET, Charlotte Lennox, The Lady’s museum, 1730)

4.2.3 Structural complexity of the CC (COMPLEXIT Y _CC) CCs may display different degrees of complexity, comprising – the NP + XP-combination typical of the SC; – a single verb; – a copula verb + a nominal predicate; – a verb + one or more objects; – a verb + one or more adjuncts; – a verb + (an) object(s) + (an) adjunct(s). This multi-level variable was afterwards recoded to a three-level variable to avoid overfitting the model. This was done as follows:

128

the NP + XP-combination typical of the SC; a single verb; a copula verb + a nominal predicate; a verb + one or more objects; a verb + one or more adjuncts; a verb + (an) object(s) + (an) adjunct(s).

zfflfflfflfflfflfflfflffl}|fflfflfflfflfflfflfflffl{

– – – – – –

Frauke D’hoedt and Hubert Cuyckens

= non-complex

} }

= low-complex = complex

4.2.4 The complexity of the CC subject (SUBJECT _CC)

zfflfflfflffl}|fflfflfflffl{ zffl}|ffl{

Originally, five values were distinguished; however, as with the factor “Structural complexity of the CC”, the multiple levels were recoded to a binary variable (noncomplex vs. complex): – no subject11 – there = non-complex – pronoun – proper name = complex – noun phrase 4.2.5 Voice of the verb in the CC ( VOICE ) Three values are distinguished: – active – passive – copula (this value also covers SCs, as the predicative relation between NP and XP can be paraphrased by a copula) Note that in one regression ( find + SC vs. find + non-SC), we only made use of the data coded as VOICE = “copula” (see Sections 5.2 and 5.3.3). In the other regressions, it turned out to be necessary to turn off the factor VOICE , as SCs categorically code for “copula” and this would therefore skew the results. 4.2.6 Intervening material between the CTP and the CC subject (INTERVENING _ MATERIAL ) Two values are distinguished: – yes – no 11 Attestations are assigned the value “no subject” when an explicit subject is missing or is positioned outside the complement clause, as in (i): (i) And what I believe had happened was that Slattery and the Pringles had intended to abduct Constoupolis (Wordbanks Online, Elizabeth Ferrars, Come and be killed, 1987).

Finite, infinitival and verbless complementation

129

4.2.7 Period (PERIOD ) This factor locates the attestation within three time bands (simplified from the periodization of the corpora):

PENN:

ME EModE LModE

1 (= pre-LModE, 1150–1700)

CLMET:

1710–1780 1780–1850 1850–1920

2 (= LModE, 1700–1920)

Wordbanks:

PDE

3 (= PDE)

4.2.8 Corpus (CORPUS ) Attested examples were taken from three corpora (see above): – PENN – CLMET – Wordbanks Once the full dataset was coded for these predictors, we fitted several binary logistic regression models with fixed effects (Pinheiro & Bates 2000) to probe the multivariate and probabilistic nature of complementation strategy choices (see Section 5). We used the statistical software package R (R Development Core Team 2011) with the following packages: lme4, Hmisc, language, car, party, rms, gmodels, partykit, mlmRev and MASS.

5 A diachronic analysis of complement variation with believe, suppose, and find 5.1 A survey of CC-variation Table 2 presents the distribution of the CC-variants for each verb from ME to PDE.

130

Frauke D’hoedt and Hubert Cuyckens

Table 2: Distribution of CC-variants per verb and per period Verb

Period

To-infinitive-CC

Small Clause

That-CC

Zero-CC

Believe

1 2 3

(20) (113) (31)

9.5% 7.1% 6.5%

(26) (130) (13)

12.4% 8.2% 2.7%

(88) (468) (245)

41.9% 29.6% 51.1%

(76) (871) (190)

36.2% 55.1% 39.7%

Suppose

1 2 3

(39) (99) (6)

19.9% 6.5% 1.3%

(27) (98) (0)

13.8% 6.5% 0%

(60) (344) (78)

30.6% 23.0% 17.4%

(70) (956) (365)

35.7% 63.9% 81.3%

Find

1 2 3

(83) (117) (11)

12.2% 6.0% 2.4%

(415) (1133) (282)

61.2% 58.5% 60.6%

(122) (434) (112)

18.0% 22.4% 24.1%

(58) (252) (60)

8.6% 13.0% 12.9%

The data in the table show divergent complementation preferences for each verb, as well as differing developments per complement pattern. In that sense, the data tie in with Noonan’s observation that across languages “complementation is basically a matter of matching a particular complement type to a particular complement-taking predicate” (Noonan 2007: 101), or in other words, complementation preferences are to a large extent unique for each complement-taking predicate. The main trends that can be observed from this table are the following: – The to-infinitive and SC show a decrease with believe and suppose. – The zero-CC increases tremendously with suppose. – In contrast, the SC with find is very frequent (around 60%). – Overall, the that-CC takes up an important portion of the CC-variation; with find, it even “trumps” the zero-CC.

5.2 Design of the analysis For each of these trends, we fitted a binary logistic regression model to explore the factors significantly predicting CC-choice. To that effect, we regrouped the four variants into binary response variables (finite vs. non-finite CCs). – For believe and suppose, we investigate the variation between non-finite (SC + to-infinitive) vs. finite (that + zero) complementation. – We fit an extra model for suppose analyzing the distribution of zero-complementation vs. “all the rest”/“non-zero”. – For find, we will focus on SC-complementation by looking at binary variation between SC-complements and “all the rest”/“non-SC”. – Finally, we will also analyze the properties of the that-complementation following find in a search for “that” vs. “non-that”.

Finite, infinitival and verbless complementation

131

For the binary logistic regression models discussed in this section, the following steps were taken. We started from a maximal model comprising all potentially important factors (including interaction effects such as SEMANTICS *PERIOD ). Subsequently, those factors lacking significant explanatory power were reduced to avoid overfitting the model (for instance, with suppose, we removed the predictor SUBJECT _CC because it did not turn out to have a significant effect). Also, as the predictor CORPUS consisted of only three levels, it was not inserted as a random effect but as a main effect. This did not turn out significant, so the factor was finally left out as well.

5.3 Non-categorical variation with believe, suppose, and find: An analysis of CC-choice 5.3.1 Non-finite vs. finite CC-variation with believe and suppose Tables 3 and 4 present an overview of the regression output of believe and suppose. Table 3: Regression output for finite vs. non-finite complementation with believe; predicted odds are for non-finite12 Coefficients: Estimate Std. Error z value (Intercept) -0.1184 0.2104 -0.563 Semantics_lexical -0.6589 0.1472 -4.476 Temporal_relation_anteriority -2.5695 0.3141 -8.180 Complexity_CC2_complex -3.6832 1.0080 -3.654 Complexity_CC2_low-complex -0.9574 0.2682 -3.569 Subject_CC2_complex -0.5627 0.1632 -3.447 Intervening_material_yes -3.3421 1.0095 -3.310 Period22 -0.5960 0.2094 -2.847 Period23 -1.1546 0.2557 -4.515 --Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Pr(>|z|) 0.573730 7.59e-06 2.83e-16 0.000258 0.000358 0.000566 0.000931 0.004418 6.32e-06

*** *** *** *** *** *** ** ***

C = 0.80, Residuals p-value = 0.46, no multicollinearity, no interaction effects

12 The “2” following some of the predictors (e.g. “Subject_CC2” or “Period22”) refers to the reclassification of the variables as discussed earlier (such as the fine-grained classification of the CC subjects which was reclassified into a “complex” vs. “non-complex” binary predictor).

132

Frauke D’hoedt and Hubert Cuyckens

Table 4: Regression output for finite vs. non-finite complementation with suppose; predicted odds are for non-finite Coefficients: Estimate (Intercept) -1.48008 Semantics_lexical 0.47139 Temporal_relation_anteriority -2.01462 Complexity_CC2_complex -2.21785 Complexity_CC2_low-complex -1.29571 Subject_CC2_complex 0.01932 Intervening_material_yes -3.09371 Period21 1.06833 Period23 -2.84000 --Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’

Std. Error 0.12064 0.15567 0.30630 0.46284 0.27708 0.16490 1.01190 0.19859 0.51108

z value -12.269 3.028 -6.577 -4.792 -4.676 0.117 -3.057 5.379 -5.557

Pr(>|z|) < 2e-16 0.00246 4.79e-11 1.65e-06 2.92e-06 0.90671 0.00223 7.47e-08 2.75e-08

*** ** *** *** *** ** *** ***

0.05 ‘.’ 0.1 ‘ ’ 1

C = 0.82, Residuals p-value = 0.78, no multicollinearity, no interaction effects

It can be seen from these data that believe and suppose share a number of factors disfavoring non-finite complementation: (i) an anterior relation of the time of the CC with respect to the time of the CTP; (ii) later period (in other words, in Present-day English, finites are increasingly preferred as clausal complement); (iii) the structural complexity of the CC ; (iv) the structural complexity of the CC subject CC; (iv) intervening material. Believe and suppose differ with respect to the factor “SEMANTICS ”: When the CTP believe is used in its strongly lexical sense, non-finite complementation is disfavored; with suppose, it is just the other way around: When used lexically, suppose favors non-finite complementation. A possible explanation is that there are many more zero-CCs with suppose than that-clauses, and that these zero-CCs typically have a CTP expressing the speaker’s epistemic stance (i.e. are typically non-lexical) (see also Jędrzejowski & Demske this volume on the semantics of infinitive-embedding predicates). 5.3.2 Zero-complementation vs. non-zero-complementation with suppose Since finite complementation with suppose is characterized by a high share of zero-CCs (in comparison with believe), an extra regression was fitted examining the factors predicting the zero-CC. The following data and figure represent the regression output for the binary variation between zero-complementation and non-zero-complementation following suppose:

Finite, infinitival and verbless complementation

133

Table 5: Regression output for zero vs. non-zero complementation with suppose; predicted odds for zero-CC Coefficients: Estimate Std. Error (Intercept) 1.5296 0.1041 Semantics_lexical -1.3080 0.1214 Temporal_relation_anteriority 0.1108 0.1210 Complexity_CC2_complex 0.3424 0.1556 Complexity_CC2_low-complex 0.4126 0.1475 Subject_CC2_complex -1.0877 0.1131 Intervening_material_yes -0.6857 0.1907 Period1 0.6917 0.5006 Period3 2.2195 0.4251 Semanticslexical:Period1 -1.6047 0.5439 Semanticslexical:Period3 -1.5164 0.4595 --Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1

z value 14.694 -10.772 0.915 2.200 2.797 -9.619 -3.596 1.382 5.221 -2.950 -3.300

Pr(>|z|) < 2e-16 < 2e-16 0.360040 0.027808 0.005156 < 2e-16 0.000324 0.167080 1.78e-07 0.003177 0.000966

*** *** * ** *** *** *** ** ***

C = 0.79, Residuals p-value = 0.60, no multicollinearity, interaction effects between Semantics*Period

A factor strongly predicting zero-complementation turns out to be SEMANTICS _ lexical, which in the regression output strongly disfavors zero-complementation. In other words, the epistemic (conjecturing) usage of suppose, as in I suppose I’d had maybe four pints, maybe five . . . and a couple of vodkas (10b), strongly predicts zero-complementation. This is in line with the results from our previous model involving suppose, which was seen to favor non-finite complementation in its lexical sense. The association of zero-complementation with epistemic semantics becomes even stronger through time, as can be observed from the interaction effects (SEMANTICS _lexical*PERIOD 1 vs. SEMANTICS _lexical* PERIOD 3): The lexical sense of suppose is more disfavoring of zero-complementation in Present-day English than in earlier periods. Interestingly, the negative effect of SEMANTICS _lexical overrules the positive effect of Present-day English (which, by itself, strongly favors zero-complementation). Other factors shown to be instrumental in predicting zero-complementation are SUBJECT _CC_noncomplex (favoring) INTERVENING _MATERIAL _yes (disfavoring). A different explanatory technique to look at the dataset, whereby various predictors can be seen to “team up” to create linguistic outcomes, is “Conditional inference trees” (for discussion, see Tagliamonte and Baayen 2012). Bernaisch, Gries & Mukherjee (2014: 14) succinctly sum up the essence of conditional inference trees as follows:

134

Frauke D’hoedt and Hubert Cuyckens

Conditional inference trees are a recursive partitioning approach towards classification and regression that attempt to classify/compute predicted outcomes/values on the basis of multiple binary splits of the data. Less technically, a data set is recursively inspected to determine according to which (categorical or numeric) independent variable the data should be split up into two groups to classify/predict best the known outcomes of the dependent variable [. . .] This process of splitting the data up is repeated until no further split that would still sufficiently increase the predictive accuracy can be made, and the final result is a flowchart-like decision tree.

For suppose then, the following conditional inference tree represents the complexity of factors “joining up” to create linguistic outcomes (in this case: variation between zero or non-zero, predicted odds for zero).

Figure 1: Conditional inference tree for zero- vs. non-zero-complementation following suppose (black = favors zero-CC; grey = disfavors zero-CC)

An example of how this model should be interpreted is the following: (1) the first predicting factor is whether the semantics of suppose is epistemic or lexical; (2) if it is lexical, the next step depends on the subject of the CC (complex or non-complex); (3) if the subject is complex, the final aspect that defines the CC-outcome is the period: About 45% of the cases in PDE will occur with a zero-complement, versus only about 25% in pre-LModE and LModE. The other branches of the tree can be interpreted in a similar fashion, resulting in the bottom row of binary proportions. The conditional inference tree clearly parallels the findings of the regression output, as it can for instance be observed that presence of intervening material yields fewer zero-CCs than absence of intervening material (node 13 vs. node 12), or that non-complex subjects in the CC are more favoring of zero-CCs than complex subjects (node 6 vs. node 7). In general, the model confirms our finding that epistemic semantics is a reliable predictor for zero-complementation (nodes 4, 6, 7, and 8).

Finite, infinitival and verbless complementation

135

5.3.3 The variation between SC- and non-SC-complementation with find Given that find showed a surprisingly high proportion of SC-complementation (see Table 2), it might be instructive to oppose this variable to the other alternatives, and examine what factors significantly predict SC-complements. This, however, required limiting our data to a subset where VOICE = “copula” and TEMPORAL _RELATION = “simultaneity”. Indeed, given that SCs consist of an NP and an XP linked through a predicative, stative relation, SC-complements will only alternate with verbal complements containing the copula to be. We were thus left with 2,193 attestations, which was still sufficient to carry out a binary logistic regression analysis. Table 6 presents the regression output for “SC- vs. non-SC-complementation” with find (predicted odds are for SC). Table 6: Regression output for SC- vs. non-SC-complementation with find; predicted odds for SC (subset with VOICE = “copula” and TEMPORAL _RELATION = “simultaneity”) Coefficients: Estimate Std. Error z value (Intercept) -3.60704 0.62834 -5.741 Complexity_CC2_low-complex -8.95903 324.74430 -0.028 Intervening_material_no 4.39456 0.60879 7.219 Semantics_mental 2.32442 0.33434 6.952 Period22 0.23873 0.23241 1.027 Period23 -0.72236 0.36139 -1.999 Subject_CC2_non-complex -0.07225 0.25900 -0.279 Semantics_mental:Period2 -1.22871 0.37762 -3.254 Semantics_mental:Period3 -0.09461 0.53092 -0.178 Period22:Subject_CC2_non-complex 0.47108 0.30223 1.559 Period23:Subject_CC2_non-complex 1.53917 0.45053 3.416 --Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

Pr(>|z|) 9.44e-09 0.977991 5.25e-13 3.59e-12 0.304326 0.045625 0.780291 0.001139 0.858563 0.119071 0.000635

*** *** *** * **

***

C = 0.75, Residuals p-value = 0.39, no multicollinearity, interaction effects between Period*Subject_CC and Period*Semantics

These data prove somewhat more difficult to interpret, since there are a number of significant interaction effects interfering with the main effects. One straightforward main effect is the predictor INTERVENING _ MATERIAL _no, which significantly favors SC-complementation. The factor PERIOD shows varying behavior, as on the one hand PERIOD 3 (Present-day English) negatively correlates with SC-complementation, but in combination with non-complex CC-subjects, the correlation turns positive again. The factor SEMANTICS as well proves difficult to interpret, showing a favoring effect with the value SEMANTICS _mental, but

136

Frauke D’hoedt and Hubert Cuyckens

a disfavoring effect when interacting with PERIOD 2 (LModE). The conditional inference tree (Figure 2) confirms the inconclusive behavior of SEMANTICS and PERIOD as all levels of these predictors allow a relatively high proportion of SC-complementation. In fact, the only truly decisive factor is the presence of intervening material, which strongly disfavors SC-complements:

Figure 2: Conditional inference tree for SC- vs. non-SC-complementation with find (black = favors SC; grey = disfavors SC)

5.3.4 The variation between that- and non-that-complementation with find In Table 2, it was shown that the zero-complement is the most important CC-type with believe and suppose, and even shows a strong increase through time. The prominence of zero with these two verbs thus renders the lower share of zerocomplements following find all the more surprising. This may in part be explained by the high frequency of “find + SCs”, which, as was just discussed, is only strongly dispreferred in the condition INTERVENING _ MATERIAL _yes. However, since SCs typically do not express dynamicity, the three other CC-types are still necessary alternatives; but this raises the question why, in the three other alternatives, the that-complement has always been more prominent than the zero-complement. We will therefore have one final look at what triggers preference for the that-complement, through a search for binary variation between that-

Finite, infinitival and verbless complementation

137

complementation and non-that-complemenation with find. Table 7 presents the regression output. Table 7: Regression output of find + that- vs. find + non-that-complementation; predicted odds are for that-clauses Coefficients: Estimate Std. Error z value Pr(>|z|) (Intercept) -4.4957 0.2839 -15.835 < 2e-16 *** Semantics_perception 1.6530 0.1731 9.551 < 2e-16 *** Temporal_relation_anteriority 2.7053 0.4838 5.592 2.25e-08 *** Complexity_CC2_complex 1.9745 0.2148 9.192 < 2e-16 *** Complexity_CC2_low-complex 1.6717 0.1672 10.000 < 2e-16 *** Intervening_material_yes 3.3574 0.2811 11.945 < 2e-16 *** Period2 0.5962 0.2582 2.309 0.020920 * Period3 0.5635 0.3272 1.722 0.085093 . Subject_CC2_complex 1.0933 0.2980 3.669 0.000243 *** Complexity_CC2_complex: Intervening_material_yes -1.3927 0.6037 -2.307 0.021055 * Complexity_CC2low-complex: Intervening_material_yes -0.6337 0.6205 -1.021 0.307099 Temporal_relation_anteriority: Period2 -1.7036 0.5287 -3.222 0.001273 ** Temporal_relation_anteriority: Period3 -1.9712 0.7229 -2.727 0.006391 ** Period2:Subject_CC2_complex 0.1352 0.3280 0.412 0.680183 Period3:Subject_CC2_complex 0.9979 0.4184 2.385 0.017086 * --Signif. codes: 0 *** 0.001 ** 0.01 * 0.05 . 0.1 1

C = 0.87, Residuals p-value = 0.58, no multicollinearity, interaction effects between COMPLEXIT Y _CC*INTERVENING _MATERIAL , TEMPORAL _RELATION *PERIOD AND PERIOD *SUBJECT _CC

It can be observed that the that-complement owes its prominence to a complex interplay of predicting factors. First, a number of structural factors significantly favor that-complementation, with (i) the presence of intervening material as the lead factor, followed by (ii) (low) complex complement clauses, (iii) complex CC-subjects. Second, the that-complement is also predicted in the conditions SEMANTICS _perception and TEMPORAL _RELATION _anteriority. A third predictor appears to be PERIOD 2, as well as PERIOD 3 when it interacts with a complex CC-subject. The complexity of these contributing factors is also obvious from the conditional inference tree, presented in Figure 3, which is visually complex and contains many nodes (and hence, many steps that need to be taken).

138

Frauke D’hoedt and Hubert Cuyckens

Figure 3: Conditional inference tree for that- vs. non-that-complementation with find (black = favors that-CC; grey = disfavors that-CC)

It can be observed that in general, semantics of perception (nodes 12 to 19) yields more that-complementation, although this favouring effect may be countered when there exists a relation of simultaneity between find and the CC, when this CC find combines with is non-complex CC, and when there is no intervening material between find and the CC (node 12).

6 Discussion The divergent tendencies of complementation variation that are found across the three CTPs believe, suppose, and find support Noonan’s (2007) idea that complement preferences cannot be accounted for by a single generalizing explanation, but may be unique for each CTP. Even when complement patterns show similar behavior across matrix verbs, the underlying reasons may differ greatly. Consider, for instance, non-finite complementation with believe and suppose, where it has been seen that the predicting factors of semantics are actually contradictory: Non-finites with believe are favored with epistemic semantics, whereas non-finites with suppose prefer lexical semantics. However, we will argue that these diverging tendencies are by no means atypical phenomena, but tie in with more general tendencies in language change. First of all, the increase and atypically high share of zero-complements with suppose can be accounted for in terms of Traugott’s (1989) notion of “subjectification”. This is defined as a tendency whereby “meanings tend to become increasingly based in the speaker’s subjective belief/state/attitude toward the proposition”, and it involves, among others, the development of epistemic meanings (Traugott 1989: 35). With the CTP suppose in general, the relative share

Finite, infinitival and verbless complementation

139

of lexical vs. epistemic meanings has shifted drastically over time: from a proportion of 82% lexical vs. 18% epistemic meanings in pre-LModE to 46% lexical vs. 54% epistemic meanings in PDE. In light of the fact that a shift towards epistemic meanings is commonly viewed as instantiating a process of subjectification, and since we have shown that the epistemic meaning of suppose has increasingly favored the zero-CC, it follows that the process of subjectification has facilitated the selection of this particular CC-variant. In other words, the zero-CC can be said to have semantically specialized towards more subjective, epistemic uses of the verb, which explains its high ratio in the later periods. Our findings also conveniently link up with Rohdenburg’s Cognitive Complexity Principle, which states that, “in the case of more or less explicit grammatical options the more explicit one(s) will tend to be favored in cognitively more complex environments” (Rohdenburg 1996: 151). In the case of complementation, the “more explicit grammatical option” can be argued to be finite complementation, since this carries tense/voice distinctions and obligatorily has an expressed subject. Within the set of finite complementation patterns, then, the that-complement is even more explicit than the zero-complement, because of the presence of a complementizer explicitly signaling subordination. In turn, factors contributing to “cognitive complex environments” are, for instance, constituents of sizeable length detaching the head from the dependent, or syntactically more complex constituents. The predictors in our data corresponding to these cognitively complex environments are (i) the presence of intervening material, (ii) complex CC-subjects, and (iii) (low)-complex complement clauses. In addition, (iv) temporal relations of anteriority can be argued to be more cognitively complex as well, as these often require complex predicates and require a shift of temporal domains. Our analysis has revealed several cases where the Cognitive Complexity Principle can be invoked as an explanatory factor: First, non-finite complementation following believe and suppose is significantly predicted by the use of a copula in the CC and by the temporal relation of simultaneity. In contrast, any form of complexity (i.e. complex complement clauses, complex complement subjects, and the presence of intervening material) highly disfavors non-finites. For find as well, intervening material is the most important factor disfavoring the use of SC-complementation. Second, with the CTP suppose, the zero-complement (which is less grammatically explicit than the that-complement) is associated with non-complex CC-subjects, use of the copula in the CC, and absence of intervening material. Third, the that-complement with find is favored by exactly all the above-mentioned complex environments, and shows higher frequencies when used in a sentence with intervening material,

140

Frauke D’hoedt and Hubert Cuyckens

complex CC-subjects, low-complex to complex complements, and temporal relations of anteriority. While we did not carry out regression analyses for thatcomplementation with the other verbs, it may be hypothesized that the same predictors account for the considerable share of this CC-type across the CTPs: The that-complement, in its unique ability to encode cognitively complex structures, has secured itself a position in the complementation system that remains unchallenged by the other variants. Another element that plays a role in complement choice across the different CTPs is the notion of “semantic specialization”, as it can be observed that some of the complement types specialize for (or, show an increasing preference for) a particular meaning of the CTP. In addition to the tendency for zero-complements to increasingly occur with the epistemic meaning of suppose (see Section 5.3.2), find as a perception verb increasingly takes the that-complement as well. As such, with find, the that-complement seems to have found a twofold niche: On the one hand because of its ability to occur in complex environments, and on the other hand because of its semantic specialization with the perception meaning. An additional example of semantic specialization can be observed with the mental uses of find, which favor SC-complements. This may also explain why the zero-complement with find is uncommonly infrequent, especially when compared to its frequencies with suppose and believe: Since the SC- and thatcomplement have each occupied the semantic values of find, the zero-complement was no longer left the potential to expand through semantic specialization of its own.13 13 The idea of semantic specialization is reminiscent of Cuyckens and D’hoedt (2015), where admit in its meaning of ‘wrongdoing’ was typically associated with the gerund-complement. In this case, the correlation between semantics and complementation (or, the predictive value of the CTP’s semantics on the choice on complement) was justified by the notion of Dependent Time Reference (DTR) (Noonan 2007: 103), according to which the time reference of the complement is logically bound to the time reference of the CTP. In the case of admit, then, the wrongful act one admits to is likely to be situated in the past. Thus, the CTP itself specifies the particular temporal domain of the event in the CC (viz. anteriority), which removes the need to explicitly encode this anterior temporal domain again in the complement clause (e.g. by finite verbs with a tense marker), leading to an increased use of the gerund in these contexts (also see De Smet 2012). In contrast, when admit is used as a speech act verb, it does not specify its CC’s temporal domain, so that here, a finite CC detailing the temporal specifications is preferred (Cuyckens & D’hoedt 2015: 95). While the notion of DTR is valuable in itself, it cannot be account for all invoked to explain all cases of CCs’ semantic specialization. With find for instance, neither the perception meaning nor the mental meaning evoke any kind of temporal domain. Both can assert something about an event that is either simultaneous, anterior or posterior to the matrix event. In light of Noonan’s claims, the occurrence of non-finite SC-complements following mental find is rather surprising. In these cases, then, the semantic specialization of the complement types with the matrix verbs appears more random, and less bound by inherent semantics of the matrix verb.

Finite, infinitival and verbless complementation

141

Finally, we would briefly like to elaborate on the complement type that was awarded least attention in this paper, and more specifically on what we consider “the failure of the to-infinitive” (see also Speyer this volume). It has been shown that the to-infinitival ECM-construction after verbs of thinking and declaring did not appear until late in the ME period (Los 2005: 233) and that it moreover owes a lot to the SC-construction, with a minimal alteration in the addition of the string to be (Bock 1931; Warner 1982: 148–149; Los 2005: 255). Although admittedly, the to-infinitive has since prospered in other contexts (e.g. with verbs of volition; see Croft 2000, Los 2005), it has seen a strong decrease with the CTPs discussed in this paper and plays only a minor role in PDE anymore. This may come as a surprise, especially since the to-infinitive has in fact a number of qualities that are lacking in the other CC-types: On the one hand, the presence of a verb – albeit non-finite – offers the to-infinitive much of the potential that finite CCs have: It can denote dynamic situations (in contrast to SC-complements), vary in tense (simultaneity and anteriority) and voice (active and passive), and contain (complex) objects and adjuncts. On the other hand, the absence of an explicit complementizer and the compactness of the verbal infinitive renders the toinfinitive a very concise and economic linguistic tool. However, a closer look at our data shows that the to-infinitive almost always occurs with the copula be and predominantly expresses a temporal relation of simultaneity, which is exactly what the SC-complement can encode as well. As a result, the relation between the SC and the to-infinitive has developed “from friend to foe”, as they are constantly found in competing contexts. In other words, the to-infinitive, at least in the contexts discussed here, does not appear to have exploited its full capacities and now seems to be dying a slow death.

7 Summary In this paper, we have discussed the variation in complementation patterns with the three complement-taking predicates believe, suppose, and find. We have focused on finite that- and zero-complementation, on non-finite to-infinitival ECM-complementation, and on the non-verbal Small Clause-complement. By fitting multiple binary logistic regression models, we have attempted to shed light on their complement preferences with these verbs. Analysis of our data has revealed the following general tendencies. First, the zero-complement takes up an important share of the complementation patterns, and is on the rise with all three verbs. Moreover, as the zero-complement is strongly tied to the epistemic meaning of suppose, its increasing frequency can be accounted for in terms of

142

Frauke D’hoedt and Hubert Cuyckens

Traugott’s (1989) notion of “subjectification”. Second, Rohdenburg’s Cognitive Complexity Principle (1996) can be seen to play a crucial role in accounting for the tendency of SCs, to-infinitives, and even zero-complements to occur in structurally less complex contexts (characterized by non-complex subjects, no intervening material, and expressing simultaneity), whereas the that-complement preferably occurs in structurally complex contexts. Noël (1997: 282) mentions that subjects of the to-infinitival ECM complement of believe typically refer to already mentioned referents, while that-clauses introduce unmentioned referents. This observation nicely ties in with our results in that non-complex subjects, with pronouns as referential items, are associated with to-infinitives as well. Further, some element of “semantic specialization” is of importance as well, in that the zero-complement is strongly associated with epistemic meanings of suppose, and in that find prefers SCs when denoting mental meanings but that-complements when denoting perception. Finally, the to-infinitive appears to be declining strongly with each of the CTPs, which we ascribe to its similarities with the SC, and its failure to exploit its verbal qualities.

Acknowledgments The first author gratefully acknowledges the financial support of the Research Foundation – Flanders. Both authors are very much indebted to Benedikt Szmrecsanyi, Karlien Franco, and Freek Van de Velde for generous help with the R-code.

List of abbreviations CC(-type) = complement-clause (type); CTP = complement-taking predicate; ECM = Exceptional Case Marking, also known as VOSI-construction; EModE = Early Modern English; ME = Middle English; LModE = Late Modern English; PDE = Present-day English; SC = Small Clause; VOSI-construction = Verb+Object/ subject+Infinitive-construction, also known as ECM

Corpora Corpus of Late Modern English Texts (CLMET 3.0) (httpsː//perswww.kuleuven.be/~u0044428/ clmet3_0.htm) Wordbanks online (http://wordbanks.harpercollins.co.uk)

Finite, infinitival and verbless complementation

143

The Penn Corpora of Historical English, including: the Penn-Helsinki Parsed Corpus of Middle English, third edition (PPCME3); the Penn-Helsinki Parsed Corpus of Early Modern English, second edition (PPCEME2); and the Penn Parsed Corpus of Modern British English (PPCMBE). More information on the Penn Corpora of Historical English can be found at http://www.ling.upenn.edu/histcorpora/ The Oxford English Dictionary (http://www.oed.com/)

References Aarts, Bas. 1992. Small Clauses in English: The Nonverbal Types. Berlin: Mouton de Gruyter. Achard, Michel. 1998. Representation of Cognitive Structures: Syntax and Semantics of French Sentential Complements. Berlin: Mouton de Gruyter. Basilico, David. 2003. The topic of small clauses. Linguistic Inquiry 34(1). 1–35. Bernaisch, Tobias, Stefan Th. Gries & Joybrato Mukherjee. 2014. The dative alternation in South Asian Englishes: Modelling predictors and predicting prototypes. English WorldWide 35(1). 7–31. Bock, Hellmut. 1931. Studien zum präpositionalen Infinitiv und Akkusativ mit dem to-Infinitiv. Anglia 55. 114–249. Boye, Kasper & Peter Harder. 2007. Complement-taking predicates. Studies in Language 31(3). 569–609 Bresnan, Joan. 1970. On complementizers: Toward a syntactic theory of complement types. Foundations of Language 6(3). 297–321. Bresan, Joan. 1976. Nonarguments for raising. Linguistic Inquiry 7(3). 485–501. Bresnan, Joan. 1979. Theory of Complementation in English Syntax. New York: Garland. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, MA: MIT Press. Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press. Citko, Barbara. 2011. Small clauses. Linguistics and Language Compass 5(10). 748–763. Croft, William. 2000. Explaining Language Change. London: Longman. Cuyckens, Hubert, Frauke D’hoedt & Benedikt Szmrecsanyi. 2014. Variability in verb complementation in Late Modern English: Finite vs. non-finite patterns. In Marianne Hundt (ed.), Late Modern English Syntax, 182–203. Cambridge: Cambridge University Press. Cuyckens, Hubert & Frauke D’hoedt. 2015. Variability in clausal verb complementation: The case of admit. In Mikko Höglund, Paul Rickman, Juhani Rudanko & Jukka Havu (eds.), Perspectives on Complementation: Structure, Variation and Boundaries, 77–100. Houndsmills: Palgrave Macmillan. Declerck, Renaat. 1991. A Comprehensive Descriptive Grammar of English. Tokyo: Kaitakusha. De Smet, Hendrik. 2008. Diffusional Change in the English System of Complementation. University of Leuven: PhD dissertation. De Smet, Hendrik. 2012. Review of Yoko Iyeiri. Verbs of Implicit Negation and their Complements in the History of English. Amsterdam: John Benjamins. ICAME-Journal 35. 138‒143. De Smet, Hendrik. 2013. Spreading Patterns: Diffusional Change in the English System of Complementation. Oxford: Oxford University Press. Dixon, Robert M.W. 1991. A New Approach to English Grammar on Semantic Principles. Oxford: Oxford University Press.

144

Frauke D’hoedt and Hubert Cuyckens

Duffley, Patrick J. 1992. The English Infinitive. London: Longman. Duffley, Patrick J. 1999. The use of the infinitive and the -ing after verbs denoting the beginning, middle and end of an event. Folia Linguistica 33(3–4). 295–331. Fanego, Teresa. 1996. The development of gerunds as objects of subject-control verbs in English (1400–1700). Diachronica 13(1). 29–62. Fanego, Teresa. 1998. Developments in argument linking in Early Modern English gerund phrases. English Language and Linguistics 2(1). 87–119. Felser, Claudia. 1999. Verbal Complement Clauses. Amsterdam: John Benjamins. Fischer, Olga. 1995. The distinction between to and bare infinitival complements in Late Middle English. Diachronica 12(1). 1–30. Givón, Talmy. 1980. The binding hierarchy and the typology of complements. Studies in Language 4(3). 333–377. Haegeman, Liliane & Jacqueline Guéron. 1999. English Grammar: A Generative Perspective. Oxford: Blackwell Publishers. Hoekstra, Teun. 1988. Small clause results. Lingua 74. 101–139. Langacker, Ronald. W. 1991. Cognitive Grammar, vol 2. Stanford: Stanford University Press. Langacker, Ronald. 2008. Cognitive Grammar: A Basic Introduction. Oxford: Oxford University Press. Los, Bettelou. 2005. The Rise of the To-Infinitive. Oxford: Oxford University Press. Mair, Christian. 1990. Infinitival Complement Clauses in English: A Study of Syntax in Discourse. Cambridge: Cambridge University Press. Mair, Christian. 2002. Three changing patterns of verb complementation in Late Modern English. English Language and Linguistics 6(1). 105–131. Mair, Christian. 2003. Gerundial complements after begin and start: Grammatical and sociolinguistic factors, and how they work against each other. In Günter Rohdenburg & Britta Mondorf (eds.), Determinants of Grammatical Variation in English, 329–345. Berlin: Mouton de Gruyter. Miller, D. Gary. 2002. Nonfinite Structures in Theory and Change. Oxford: Oxford University Press. Mindt, Dieter. 2000. An Empirical Grammar of the English Verb. Berlin: Cornelsen. Noël, Dirk. 1997. The choice between infinitives and that-clauses after believe. English Language and Linguistics 1(2). 271–284. Noël, Dirk. 2003. Is there semantics in all syntax? The case of accusative and infinitive constructions vs. that-clauses. In Günter Rohdenburg & Britta Mondorf (eds.), Determinants of Grammatical Variation in English, 347–377. Berlin: Mouton de Gruyter. Noonan, Michael. 2007 [1985]. Complementation. In Timothy Shopen (ed.), Language Typology and Syntactic Description: Vol. 2, Complex Constructions, 52–150. Cambridge: Cambridge University Press. Pinheiro, José C. & Douglas. M. Bates. 2000. Mixed-Effects Models in S and S-PLUS. New York: Springer. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A Comprehensive Grammar of the English Language. London: Longman. Postal, Paul. 1974. On Raising. Cambridge, MA: MIT Press. Radford, Andrew. 1997. Syntactic Theory and the Structure of English. Cambridge: Cambridge University Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, MA: MIT Press.

Finite, infinitival and verbless complementation

145

Rohdenburg, Günter. 1995. On the replacement of finite complement clauses by infinitives in English. English Studies 76(4). 367–388. Rohdenburg, Günter. 1996. Cognitive complexity and increased grammatical explicitness in English. Cognitive Linguistics 7(2). 149–182. Rohdenburg, Günter. 2006. The role of functional constraints in the evolution of the English complementation system. In Christiane Dalton-Puffer, Dieter Kastovsky, Nikolaus Ritt & Herbert Schendl (eds.), Syntax, Style and Grammatical Norms: English from 1500–2000, 143–166. Bern: Peter Lang. Rohdenburg, Günter. 2014. On the changing status of that-clauses. In Hundt, Marianne (ed.), Late Modern English Syntax, 155–181. Cambridge: Cambridge University Press. Rosenbaum, Peter. 1967. The Grammar of English Predicate Complementation. Cambridge, MA: MIT Press. Rudanko, Juhani. 1998. Change and Continuity in the English Language: Studies on Complementation over the Past Three Hundred Years. Lanham, MD: University Press of America. Rudanko, Juhani. 2000. Corpora and Complementation. Lanham, MD: University Press of America. Rudanko, Juhani. 2006. Watching English grammar change. English Language and Linguistics 10(1). 31–48. Rudanko, Juhani. 2010. Explaining grammatical variation and change: A case study of complementation in American English over three decades. Journal of English Linguistics 38(1). 4–24. Rudanko, Juhani. 2012. Exploring aspects of the great complement shift, with evidence from the TIME corpus and COCA. In Tertuu Nevalainenu & Elizabeth Closs Traugott (eds.), The Oxford Handbook of the History of English, 222–232. Oxford: Oxford University Press. Smith, Michael B. 2002. The semantics of to-infinitival vs. -ing verb complement constructions in English (with J. Escobedo). In Mary Andronis, Christopher Ball, Heidi Helston & Sylvain Neuvel (eds.), The Proceedings from the Main Session of the Chicago Linguistic Society’s Thirty-Seventh Meeting, 549–565. Chicago: Chicago Linguistic Society. Tagliamonte, Sali A. & Harald Baayen. 2012. Models, forests, and trees of York English: Was/ were variation as a case study for statistical practice. Language Variation and Change 24(2). 135–178. Traugott, Elisabeth Closs. 1989. On the rise of epistemic meanings in English. An example of subjectification in semantic change. Language 65(1). 31–55. Vosberg, Uwe. 2003. The role of extractions and horror aequi in the evolution of -ing complements in Modern English. In Günter Rohdenburg & Britta Mondorf (eds.), Determinants of Grammatical Variation in English, 329–345. Berlin: Mouton de Gruyter. Warner, Anthony. 1982. Complementation in Middle English and the Methodology of Historical Syntax. University Park: Pennsylvania State University Press. Wierzbicka, Anna. 1988. The Semantics of Grammar. Amsterdam: John Benjamins. Williams, Edwin. 1980. Predication. Linguistic Inquiry 11(1). 203–238. Williams, Edwin. 1983. Against small clauses. Linguistic Inquiry 14(2). 287–308.

Virginia Hill

6 Early Modern Romanian infinitives: origin and replacement This contribution addresses one of the most frequent statements we find in historical linguistic studies of Balkan languages, namely, that the replacement of the infinitive occurred very late in Romanian (i.e., 16th–17th century, when să-subjunctive clauses emerge; cf. Chivu et al. 1997 and Nedelcu & Paraschiv this volume), whereas other Balkan languages had replaced the infinitive much earlier (e.g., starting around the 10th century in Bulgarian; cf. MacRobert 1980). The late onset of the infinitive replacement in Romanian is then accounted for in terms of geographical distance from the starting point of this phenomenon (i.e., Greek), Romanian being spoken at the periphery of the Balkan Sprachbund (Rohlfs 1933; Sandfeld 1930, a.o.). These conclusions have been drawn mostly on the basis of phonological and morphological considerations of the forms involved. In this paper, I argue, on the basis of Early Modern Romanian data, that a syntactic perspective on the same phenomenon yields different conclusions: First, I point out that a distinction must be drawn between the original infinitive (i.e., the one inherited from Latin), and the subsequent a-infinitive that emerged in Romanian at some point before the first written texts. Second, I show that the replacement has applied to a certain syntactic structure, not to every instantiation of the infinitive, so the Romanian replacement is of a syntactic, not of a morphological type. Finally, I argue that a syntactic approach to the loss of infinitives leads us to a more systematic picture of the changes in the sentential complements to verbs in general. More precisely, we see that what changes during the replacement process is not the structure but the material that spells it out: The lexical and morphological material is recycled within a constant syntactic pattern. Such systematic recycling has predictive power for linguistic reconstruction.

1 The infinitive in Early Modern Romanian texts: an overview The first Romanian written texts date from mid16th century. Since the earliest texts, the infinitive verbs are preceded (with a few systematic exceptions) by the element a ‘to’. In Romance languages, a ‘to’ functions as a complementizer Virginia Hill, University of New Brunswick DOI 10.1515/9783110520583-006

148

Virginia Hill

in front of infinitive verbs, and as a preposition in front of nouns (e.g., Kayne 1994). For Romanian, there is a debate between a complementizer (DobrovieSorin 1994) or a mood marker status (e.g., Alboiu 2002) for a ‘to’ in front of infinitives; this paper opts for the former, for reasons explained further in the analysis. For the time being, the point is that a ‘to’ precedes one of the two infinitive forms: the long infinitive, with the ending -re, as in (1a); or the short infinitive without an ending, as in (1b). (1) a. Sosiră pre cel pământ [în carele [a lăcuire] era] in which.DEF A live.INF was arrived.3PL on that land ‘they arrived in the land in which they had to live’ (PO {235}) b. le-au poruncit [a face] şi [a cinsti Domnedzeu] and A revere God them.DAT =has ordered A do ‘he ordered them to receive and revere God’ (PO {5}) As shown in (2), the long infinitive may also have, attached to the ending -re, a final segment [a], whose origin is uncertain: It could be a free phonetic variation or a replica of the enclitic definite article (Nedelcu 2013). The variations in the endings (i.e., -rea, -re, no ending) occur in free alternation throughout the Early Modern Romanian (i.e., up to the end of the 18th century), as in (2)–(4). (2) au vrut milostivul Dumnedzău [a nu lăsarea acestŭ A not leave.INF this has wanted merciful.DEF Lord pământ făr’de oameni] land without people ‘the merciful Lord willed it that this land not be left without people’ (Ureche 66) (3) s-au REFL =has=

lui DAT

gătit [a stare cu războiŭ împrotiva prepared A wage.INF with war against

Răzvan] Razvan

‘he prepared himself to wage war against Razvan’ (Costin 16) (4) Nicăirea nu nowhere not

i-au cutezat [a-i sta împotrivă] he.DAT =has= dared A =he.DAT stand against

Radul vodă Radu.DEF King ‘There was no place where King Radu would dare to confront him’ (Ureche 147)

Early Modern Romanian infinitives: origin and replacement

149

The same construction may display de in addition to a ‘to’, with the same range of morphological variation on the infinitive, as shown in (5) (Pană Dindelegan 2013: 211–215). In Early Modern Romanian, as in MR, and as in other Romance languages, de has a dual categorization: It can be a non-finite complementizer or a preposition (Hill 2013a, b). (5) a. Şi lăsă [de-a grăirea cu el] And stopped DE -A talk.INF with he.ACC ‘And he stopped talking to him’

(PO {55})

b. Iar turcii, cum au vădzut poarta cetăţii And Turks.DEF as have= seen gate.DEF fort.DEF.GEN deschisă, au lăsat [pre moscali] de-a-i mai opened have= quit DOM Russians DEA =they.ACC more gonire]] ş-au ş-început a intra în cetate. chase.INF and=have and=started A enter in fort ‘And the Turks, as soon as they saw the gate of the fort opened, quit chasing the Russians and started entering the fort.’ (Neculce {380}) c. Au lăsat şi ei [Cetatea Neamţului] de a o mai bate]] have= quit and they fort.DEF Neamt.GEN DE A =it= still beat ‘They’ve also stopped attacking the Neamtu fort’

(Neculce {107})

In addition, since the earliest texts, there is a productive use of the nominalized infinitive, as in (6), where the nominalization is signaled by the enclitic definite article a ‘the’, and it is confirmed by the presence of a qualifying adjective. (6) strigarea lor mare e înaintea Domnului is before God.DEF.GEN shout.INF. DEF their big ‘their big shouting is (put) before God’ (PO {61}) The examples in (1) to (6) illustrate the default use of the infinitive stem in Early Modern Romanian. There are, however, rare examples of verbal infinitives without a ‘to’; instead, they are preceded by de alone, as in (7a, b). (7) a. Aşa stătu nărodul de aducere darure Thus stopped people.DEF DE bring.INF presents ‘Thus, the people stopped bringing presents’ (PO {301}) b. Se

începu cuvânt de zicere REFL = started word DE say.INF ‘A discourse was started, to say: . . .’

(PO {38})

150

Virginia Hill

In (7a), the infinitive is verbal, because it is followed by a direct object DP unmarked for Case. In Early Modern Romanian, that means either Nominative or Accusative. In (7b), the infinitive de zicere (‘DE say.INF ’) generates a non-finite relative clause modifying cuvânt ‘word’. Crucially, in both cases the infinitive is verbal (versus nominal), and it lacks the preverbal a ‘to’. These rare examples can be considered traces of an earlier possibility of generating an infinitive clause solely by means of the complementizer de. This analysis of the examples in (7) seems self-evident, but it differs from the analysis the same examples would receive in current historical studies of Early Modern Romanian. In particular, the current studies consider de as a preposition across the board (Jordan 2009; Nedelcu 2013). Thus, the examples in (7a, b) would be PPs headed by de: in (7a), P-de selects a DP whose noun selects, in turn, a direct object DP with Accusative Case instead of Genitive; (7b) displays an attributive PP (although zicere is never seen with a nominal use elsewhere). Hence, the conclusion would be that such infinitives have a mixed verb/noun status, and that constructions as in (7) are the source of nominalized infinitives (i.e., arising later versus earlier than a-infinitives). In the framework I adopted here this analysis is untenable: First, typologically, de in non-finite clauses is a complementizer (either in selected or in relative clauses) in Romance in general (Rizzi 1997, a.o.); so is infinitive a, with which de may be in complementary distribution (e.g., in French). Second, there is language internal evidence for the complementizer status of de: (i) Semantically, aspectual verbs as in (7) select either a DP, as in (8a), or a clausal complement, as in (8b), but never a PP. In fact, de a- infinitives and a-infinitives are in free distribution in this context, as in (8c, d), which should not be the case if de were a thematically required preposition. (8) a. marea poate cunteni şi vânturile a înceta calm and winds.DEF A stop sea.DEF can ‘[your word] can calm the sea and stop the winds’ (Varlaam C {231v}) b. nu înceta învăţându=i not stopped teaching=they.ACC ‘he did not stop teaching them’ (Coresi EV {454}) c. nu vom înceta a ruga pre Dumnezău not will.1PL stop A pray DOM God ‘we will not stop to pray to God’ (NT 1648 {113}) d. au vrut înceta de-a le aducerea has= wanted stop DE -A they.ACC = bring.INF ‘he wanted to stop bringing them’ (NT 1648 {298r})

Early Modern Romanian infinitives: origin and replacement

151

(ii) The verb want never displays a thematic preposition in Romanian, but it occurs with de a: (8) e. nevrând de a o da not.wanting DE A it= give ‘not wanting to give it’

(DB, IV, 277 apud Frîncu 1969: 91)

(iii) De is already used as a complementizer in de-indicatives (see 9b), which are concurrent with the infinitive clauses; and also, in non-finite relatives, which may involve either infinitive or supine verbs in Early Modern Romanian (Dragomirescu 2013; Pană Dindelegan 2014). Thus, there is a complementizer de in Early Modern Romanian, independently of infinitives. Accordingly, examples as in (7), compared to those in (5) and (8), indicate that the Early Modern Romanian speakers hesitate not with respect to the nominal or verbal status of the infinitive but with respect to the option for the lexical complementizer (is it de, a or de a?). Therefore, if we look at the data in (1) to (8) from a Romance perspective, we can draw the following conclusions: The Latin infinitive ending in -re (e.g., Latin venire ‘to come’) has been preserved in Romanian (see also Fischer 1985) in two ways: as a noun and as a verb. As a noun, it is inflected as a regular DP (e.g., displaying the definite article as in (6)), and may occur embedded under various prepositions, including de (e.g., loc de o săgetare ‘(the) distance of a dart-throwing’; PO {68}).1 On the other hand, a verbal categorization of the infinitive becomes conditioned by the presence of lexical complementizers, such as a or de – which is a general trend in Romance languages. Romanian is peculiar insofar as the variation between a and de did not survive, only a being gen1 Palia de la Orăştie (printed in 1582) provides minimal pairs of prepositional phrases where de alternates with other prepositions or lack of preposition in front of the same infinitive-based noun, as in (i). These examples show that nominal infinitives are independent of de. (i) a. cum să fie voao spre mâncare that SA be.SUBJ. 3 you.DAT. PL for meal ‘that you can have it for meals’ (PO {15}) b. cum să fie ţie mâncare şi lor acolo and them.DAT there that SA be.SUBJ.3 you.DAT. SG meals ‘that you and they have it for meals there’ (PO {29} c. să fie voao de mâncare ca şi veardzele SA be.SUBJ. 3 you.DAT. PL de meal as also cabbages ‘you should use it for meals as you would cabbages’ (PO {34})

152

Virginia Hill

eralized in the front of the clausal infinitive. The complementizer de has been preserved with infinitives only in conjunction with a, as in (5).2

2 Variation in sentential complementation In order to understand the fate of infinitives, one must take into consideration the more general properties of sentential complementation in these texts. More precisely, infinitive complements occur in alternation with de-indicatives and săsubjunctives, as shown in (9). (9) a. au vrut milostivul Dumnezău [a nu lăsarea A not leave.INF has=wanted merciful.DEF God acest this

pământ land

făr’ de without

a-infinitive

oameni] people

‘merciful God did not want to leave this land without people’ (Ureche {5v}) b. au vrut Dumnedzău [de s-au tocmit aşea] DE REFL =has= arranged thus has=wanted God

de-indicative

‘God wanted for this arrangement to be made’ (Ureche 41) c. Bogdan vodă n-au vrut să le dea război să-subjunctive Bogdan King not-has= wanted SA they.DAT =give.SUBJ War ‘King Bogdan did not want to make war with them’

(Ureche 16v)

The competition was not equal: Although all these constructions may be found embedded under the same verb, the productivity differs. De-indicatives are productive with only two classes of verbs (i.e., causatives and aspectuals; see Frâncu 2009; Hill 2013b), whereas a-infinitives are strong competitors to săsubjunctive across the board until mid17th century. The distributional contrast is the first indication that de-indicatives were already in decline at the time of the first written texts, while a-infinitives are strong and productive. The emergence of a ‘to’ (instead of de) as the preferred infinitive complementizer must have happened at a time when de was already established as the complementizer of anaphoric indicative clauses, as in (9b). The failure of de 2 De has been, however, dropped from a-infinitive complements to verbs in Modern Romanian.

Early Modern Romanian infinitives: origin and replacement

153

as an infinitive complementizer must then be related to the fact that it was already in use as a complementizer with another verb form.

3 The replacement of the original infinitive The Latin infinitive does not need a complementizer to function as a sentential complement. However, Latin infinitives are non-phasal in minimalist terms: They cannot check a DP for Nominative Case in the subject position. Furthermore, they often occur in argumental positions, which confers them nominal properties (cf. Miller 2000). Therefore, the Latin infinitive with the ending -re spreads in the Romance area as an ambiguous nominal and verbal category. Accordingly, the inherited infinitive in Romanian must be the same Latin -re form, with no complementizers. This should count as the original infinitive, which learners could classify either as a noun or as a verb, as argued in the previous section.3 The classification as a noun is very productive by the time of the first written texts (see the Appendix in Dragomirescu 2013). At the same time, the classification of the same infinitive form as a verb was facilitated by embedding the infinitive under an element that spelled out the complementizer (C): Any constituent headed by C is a clause. From this point of view, the morphology of the infinitive did not matter: It could have the ending -re or -rea, or it could drop the ending altogether, because the complementizer is sufficient to indicate the verbal category of the infinitive, the clause type and the nonfiniteness. 3 Stan (2013) proposes a different interpretation of the data: The nominal infinitive springs at a later date from the use of the verbal infinitive in contexts as in (5a), by dropping a ‘to’ and having the infinitive embedded only under de (which is always a preposition in her study). There are several problems with this analysis, among which: (i) The statistics counteract this proposal. For example, Codicele Todorescu, dated from 1601, has 16, 585 words, 55 of which are nominalized infinitives, occurring as subjects, objects or embedded under various prepositions. Only five of these nominalized infinitives are under the preposition de, and there is only one occurrence of de a + infinitive in the text. Such data strongly suggest that the existence of nominalized infinitives is independent of the existence of de a + infinitive constructions, which were too few to have had any serious impact on the acquisition process, in addition of being chronologically simultaneous with (instead of preceding) an already productive system of nominalization. (ii) Diachronically, the element discarded in sentential complements as in (5) is de, not a ‘to’. (iii) Theoretically, it is unlikely that an element that has just been fixed in the language (i.e., a as infinitive complementizer) be immediately dropped from the context in which it had been re-analyzed.

154

Virginia Hill

As argued in Section 1, the learners must have experimented with the two candidates for the complementizer function in non-finite clauses in Romance, that is, de and a, and the preference went to a ‘to’. The point is that the emerging a-infinitive is subsequent to the original form inherited from Latin, and possibly subsequent to de-indicatives. Along these lines, the original infinitive has been replaced in Romanian at a very early stage, by de-indicative clauses, probably at the time when such a replacement took place in the other Balkan languages, and it followed the Balkan replacement pattern (i.e., a switch from non-finite to finite verbs; see Hill 2013b). The a-infinitive whose replacement is recorded in the Early Modern Romanian texts is a more recent construction, which arises either in parallel with de-indicatives or some time later, by following the Romance pattern available in the grammar. A discussion of the reanalysis of a in infinitive clauses will follow after the introduction of the theoretical tools I will use for the syntactic assessment.

4 Cartography The theoretical background for this analysis is that change in language means change in the process of first language acquisition: Changes occur when the learner tries to sort out ambiguous cues in the primary linguistic data, which may lead to the re-analysis of the lexical material in the input (see Hale 2007 for an overview of theoretical studies). The formal tools used to assess changes in language come from the cartographic approach to clause structure (especially Rizzi 1997, 2004). For the present case study, the clause segment that involves changes is the left periphery, that is, the Complementizer Phrase (CP), in its interaction with the properties of the inflectional phrase headed by T(ense). In cartography, the CP field is articulated over several projections ranging from ForceP to FinP, as in (10). (10) ForceP > TopP > FocusPcontrast > FinP > NegP > TP > vP FinP selects a certain type of inflection, that is, it restricts the type of morphological forms associated with T. An important observation is that clausal negation, merged as NegP, is the highest item in the inflectional field when present (Zanuttini 1997 for Romance; Alboiu 2002 for Romanian), so its presence allows us to draw a line between the CP and the TP fields. The articulation of the CP field as in (10) does not apply uniformly in the derivation of clauses: The field may be projected up to ForceP or it may be truncated at lower levels (e.g., at FinP). Accordingly, the clause is phasal, or a closed

Early Modern Romanian infinitives: origin and replacement

155

domain, if it projects up to Force, which further allows for the embedded verb to license Nominative subjects, uncontrolled from the matrix. On the other hand, embedded clauses with controlled subjects are non-phasal, usually truncated as FinPs, and allow for clause border transparency.4 Both situations occur in Early Modern Romanian (e.g., see Niculescu 2013; Alboiu & Hill 2013, a.o.). On the basis of this theoretical background, we can now establish a pattern for the left periphery of clauses that replaced the infinitive in Balkan languages. At a descriptive level, the replacement consisted in the switch from the [-finite] to the [+finite] feature of T (i.e., from infinitive to indicative verb forms), and this switch involved the insertion of a particle (also labeled as a mood marker) in front of the verb, on an obligatory basis. Hence, subjunctive verbal strings as in (11) have emerged (Rivero 1994; Mišeska-Tomić 2006, a.o.). (11) Bulgarian da otida ‘SUBJM go.INDIC.3SG’; Greek

na erthis ‘SUBJM come.INDIC.2SG’;

From a formal perspective, the mood markers are in fact complementizers merged in Fin.5 This can be tested through the word order, as shown for Bulgarian (12) and Greek (13). (12) a. Iskam na moreto s Marija da otide, ne s Ana.6 want.1SG to sea.DEF with Maria DA goes.PF not with Ana ‘I want him/her to go to the sea with Maria, not with Ana.’ b. Iskam toj da ne hodi. want.1SG he DA not goes.IMPF ‘I want him to not go.’

4 The theory of feature transfer from C-to-T in Chomsky (2008) capitalizes on the association of phi-features with C, not with T. T only inherits these features which enable it to license a DP subject. For the map in (10), it means that the phi-features are in Force (i.e., CP is full-fledged); hence, the presence or the absence of Force is decisive for the ability of T to license a DP subject and for the phasal status of the clause (see also Alboiu 2010). 5 Current studies have already established that the subjunctive mood markers are in Fin in the Romance languages that adopted the Balkan clause pattern for their subjunctives (Paoli 2003; D’Alessandro & Ledgeway 2010). 6 Data from Olga Mladenova (p.c.)

156

Virginia Hill

(13) a. Thelo sti thalasa me ti Maria na pai aftos, oxi me tin Ana.7 want.1SG to sea with DEF Maria NA goes he not with DEF Ana ‘I want him to go to the sea with Maria, not with Ana.’ b. Thelo aftos na min pai. NA not goes want.1SG he ‘I want him to not go.’ According to the map in (10), the subjunctive marker da or na is in the CP field, because it is higher than negations (cf. (12b); (13b)); it is, however, low in the CP field, because it follows the sequence of Topic > contrastive Focus constituents (cf. (12a); (13a)). Therefore, we can locate the subjunctive marker in Fin, and assign the map in (14) to the Balkan subjunctive clause. (14) ForceP > TopP > FocusP > Fin-da/na > (NegP) > TP This map extends up to the ForceP level because the construction is phasal: That is, the subject of the subjunctive complement is different from the subject of the matrix clause. Accordingly, although Force is not spelled out, its features are checked by the complementizer merged in Fin, through distance Agree.8 In Early Modern Romanian, de-indicative complements, as in (15), also display de in Fin. (15) a. că apucă [[patul mortului]TOP C O N T R de-l= timpină Domnul] DE - it= met.3 Lord.the for got bed.the dead.the.GEN ‘for the Lord got to see the dead man’s bed’ (Coresi EV {385}) b. să tîmplasă de nu ştiè nemic REFL =happened DE not knew.3 nothing ‘it happened that he did not know anything’

(Neculce {242})

In (15a), the matrix verb apuca ‘get to’ has obligatory control and shares the external th-role with the embedded verb.9 Hence, the clausal complement lacks 7 Data from Melita Stavrou (p.c.) 8 Empirical data supporting the checking of Force by the complementizer merged in Fin are provided for subjunctive clauses in D’Allessandro & Ledgeway (2010) and for Romanian subjunctives and indicatives in Hill (2013b). 9 Many studies agree that the CP field is deleted in constructions where obligatory control or NP-movement applies with raising verbs (see overview in Bošković 1997). In cartography, only ForceP needs to be deleted in such constructions. The examples in (15) meet the expectation in cartography, since Fin does not interfere with DP movement to argumental positions under obligatory control or with raising verbs (Alboiu 2002). Thus, (15a, b) counter the statement in Jordan (2009) that the complementizer de would block NP-movement.

Early Modern Romanian infinitives: origin and replacement

157

ForceP. However, it projects up to Top/FocusP, since de is preceded by a constituent with contrastive reading. Thus, according to the map in (10), the word order indicates that the complementizer de is lower than Focus, hence, in Fin. In (15b), de precedes the negation, providing confirmation that de is merged in the CP field. Therefore, the derivational pattern is as in (16), and it replicates the Balkan pattern in (14). (16) ForceP > TopP > (FocusP) > Fin-de > NegP > TP This comparative outlook with Balkan languages indicates that the Early Modern Romanian de-indicative complement is the first Romanian subjunctive that replaced the Latin inherited infinitive, in the same way, and presumably on the same timeline as in other Balkan languages. This gives us the frame of reference for the analysis of a-infinitives, which arises because of the Romance genealogy of the language: Is this construction comparable to the Balkan pattern in (14)/(16)?

5 The analysis of a The standard assumption in historical linguistics is that, etymologically, infinitive a comes from the Latin preposition ad ‘towards’ ‘for’. Hence, Romanian a was a preposition before becoming an infinitive marker. This line of reasoning, adopted in Jordan (2009) and Schulte (2007), entails a re-analysis of the preposition in (17) as the complementizer in (18a, b), followed by the complete loss of the preposition around the 17th century. (17) mearsă în pădure a leamne went in forest for wood ‘he went into the forest for wood’ (Dosoftei apud Rosetti 1968: 175) (18) a. Toma mearse a-i spune lui ce vrea A-he.DAT = tell he.DAT what wants Thomas went ‘Thomas went to tell him what he wants’ (Coresi 131 apud Jordan 2009: 39) b. Şi vine a spăşi un ucenicu necredinciosu and comes A repent an apprentice unfaithful ‘And an unfaithful apprentice comes to repent’ (Coresi 131 apud Jordan 2009: 39)

158

Virginia Hill

In Jordan’s analysis, the infinitives in (18) are adverbial clauses of purpose, due to the meaning of a as a preposition. So ambiguity arises as to the analysis of the string [a + infinitive]: It is either [P + noun] or [mood marker + verb]. Once the [mood marker + verb] analysis prevails, it spreads to the selected contexts (see also Lühr this volume). There are several problems with this analysis. First, prepositions are rare with adverbial CPs in Early Modern Romanian in general (indicatives, subjunctives, gerunds), so it is not clear that a in (18) has an ambiguous status, since complementizers (e.g., că ‘that’) were the default option in these contexts, with no preceding preposition. Second if the reanalysis of a took place in adverbial clauses as in (18), we would expect them to display more long infinitives than short infinitives.10 That is, ambiguity as to the P status of a involves the analysis of the infinitive form as a noun, and only long infinitives qualify for that. What we see instead is that a-long infinitives occur in clausal complements, in alternation with short infinitives. Hence, it is either the case that the emergence of infinitive complements precedes the emergence of infinitive adverbial clauses, or that there is no relation between the distribution of infinitives in adverbial and selected contexts. The third problem concerns the timeline: The reanalysis of P-a as an infinitive mood marker proposed in Jordan (2009) is supposed to have been finalized during the 16th century, when the preposition a starts to disappear. However, the clausal use of a-infinitives has been shown to have taken place centuries before: Diaconescu (1977) points out that a-infinitive clauses exist in the sub-Danubian dialects, so the construction was stabilized before the break of these dialects from the Common Romanian, around the 13th century. The alternative view I suggest is that Latin ad entered into Romanian in two ways: as a preposition and as a complementizer. The former failed, whereas the latter thrived. In other words, there is no ontological relation between a in (17) and a in (18). The arguments for this analysis are as follows: (i) ad had double categorization (P and C) since Classical Latin (Miller 2000); (ii) during the Romanization, ad was instrumental for converting gerunds to infinitives (Fischer 1985); (iii) although there are examples of the preposition a in texts, the noun it selects is never based on an infinitival stem; (iv) in Early Modern Romanian, a and the infinitive ending -re are analyzed as redundant, hence the drop of -re (long infinitives become short infinitives); -re is a marker of non-finiteness, which is a property of verbs, not of nouns; therefore, the equivalence between a and -re entails that a has always been analyzed as a verb related element versus preposition. 10 Short infinitives are considered more recent than long infinitives (Pană Dindelegan 2013).

Early Modern Romanian infinitives: origin and replacement

159

The last point of contention is the status of a as a verb related element in Early Modern Romanian. Most studies consider it an inflectional mark for grammatical mood (see overview in Pană Dindelegan 2013), hence, in the TP field. Cartographic tests, based on the hierarchy in (10), indicate, however, that a is a complementizer in Fin. This is a more accurate result in light of the etymology proposed above: Latin C-a remains as such in Early Modern Romanian, and Fin is the site for encoding the finiteness feature. The cartographic tests are detailed below: (19) a. iară doo părţi a legiei iaste a nu A not and two parts of law is viia live

noi we

cu with

iale them

‘and two parts of the law is that we should not live with them’ (PO {6}) b. au început [cătră împăratul] a pârî tare have=started towards Emperor.DEF A denounce strongly pre Brâncovanul Brincoveanu

DOM

‘they started to tell a lot on Brincoveanu to the Emperor’ (Neculce 279) c. cela ce poate [toate] a face A Do that who can all ‘the one who can do them all’ (CC, 267 apud Frîncu 1969: 17/85) In (19a), a precedes the negation, therefore it is merged in CP field. In similar selected contexts, a within CP is preceded by Topic (19b) or Focus (19c) constituents. Accordingly, a is in Fin, and it checks the features associated with this head, that is, [finiteness] and [modality].11 The earlier texts show a high incidence of de in front of infinitive a, as mentioned in (5). Hill (2013b) argues that both de and a are Fin elements, and that such constructions are derived by splitting Fin into a head that maps the non-finiteness feature (spelled out as de) and a head that maps the modality feature (spelled out as a). The word order in (20) confirms this analysis.

11 The [modality] in Fin maps the semantic modality (e.g., realis/irrealis), not the grammatical mood (e.g., infinitive), the latter being a function of T (D’Alessandro & Ledgeway 2010).

160

Virginia Hill

(20) a. Vom fi cu vină de a nu ne putea lepădarea forgo.INF will.1PL be with blame of A not us= can ‘we’ll have the blame of not being able to forgo [it]’ (Coresi E {29}) b. Noi sântem datori să fim gata [[de oaste] [în toată We are obliged SA be.SUBJ ready of army in any vremea, [cândŭ va= veni cuvântul împăratului]], Emperor.DEF.GEN time when will.3SG come order şi [de bani] de a le darea pururea]. and of money DE A they.DAT = give.INF forever ‘We are obligated to be ready to provide them forever with ARMY, at all times, – as soon as the emperor’s order comes – and with MONEY.’ (Ureche 131) In (20a), both de and a precede the negation, therefore they are both in the CP field. In (20b), Topic and Focus constituents precede de, showing its location in Fin. Free alternation between de a and single a in the same context confirms that the two sets are equivalent for checking the features of Fin. Presumably, de served for fixing the non-finite value of Fin (since de was associated with nonfiniteness in de-indicatives), a function that was eventually assumed by a. From this perspective, the loss of de is predictable because of functional redundancy. Following the mapping in (10), the representation of a-infinitives is as in (21), and of de a infinitives as in (22). (21)

[ForceP [TopP [FocP [FinP(-finite;

(22)

[ForceP [TopP [FocP [FinP1(-finite) de [FinP2(modal) a [NegP nu [TP Vinf . . .]]]]]]]

modal)

a [NegP nu [TP Vinf . . .]]]]]]

The mapping of the infinitive can go up to ForceP because Nominative subjects are possible, as seen in (19a), where the embedded subject is the pronoun noi ‘we’ with the Nominative inflection. The structure in (21) is strikingly similar to the subjunctive Balkan pattern in (14) and (16), which needs an explanation since the verb forms are different. The explanation will be proposed later in Section 6. For the time being, what matters is the definition of a as a complementizer in Fin. This counters the conclusion of current studies where infinitive a is defined as an inflectional head, that spells out the grammatical mood (see Nedelcu 2013 for an overview). The main arguments towards the inflectional status of a are its obligatory occurrence with the infinitive verb and the fact

Early Modern Romanian infinitives: origin and replacement

161

that it allows for NP-movement under raising verbs.12 The restrictions for NPmovement have been discussed above for (15), where I pointed out that complementizers do not block this operation unless they spell out Force. However, both de and a spell out Fin, and thus allow for the truncation and the nonphasal domain needed for the implementation of this operation. As for the obligatory a with the infinitive, a distinction must be drawn between the clausal and non-clausal context in which the infinitive is merged. First, it is not true that a is obligatory in infinitive clauses – see (23). (23) Nu avea [de ce se apuca.] not had.3 on what REFL = lean ‘He had nothing to lean on.’ (Costin 105) There is an infinitive complement in (23) but a is not present. Instead, we see a wh-phrase. Such examples show that the wh-phrase and a have one equivalent function, namely, to spell out the presence of CP, which confers a clausal status to the infinitive. Conversely, if there is no evidence for a CP field, as in (24), the infinitive is analyzed as part of a modal periphrasis (putem da ‘can.1PL give’) or of a complex tense (vrem înderepta ‘will.1PL straighten’). Crucially, both occurrences of the infinitive in (24) are mono-clausal, which contrasts with the situation in (25), where the same modal selects a FinP a-infinitive and generates a bi-clausal structure (see also Jędrzejowski & Demske this volume and Grano this volume). putem grăi sau în (24) au ce but what can.1PL say or in ce chip ne vrem înderepta pre noi? what way us=want.1PL straighten DOM us ‘but what can we say or in what way will we straighten ourselves?’ (PO {156})

12 Motapanyane (1991) is the first to define a and să as mood markers in Romanian, and merges them in the head of a MoodP, within the inflectional domain. The rationale was the position of the pre-verbal subject, which precedes the subjunctive mood marker (see examples (12) and (13)). Since the DP subject does not always go to a Topic position, but it may also be in a pre-verbal argumental position (see also Motapanyane 1994), it follows that this position must be in the inflectional field, since there are no argumental positions in the CP. New developments in cartography provide a solution to this problem by showing that FinP may also be the locus for the subject, since Spec,FinP has the properties of an A-position (see NP-raising in Bošković 2007). This allows for the reconsideration of the status of a (and să) along the lines proposed in the present paper.

162

Virginia Hill

(25) nu le putem [a le cunoaşte cu singur pipăitul] with just touch not they.ACC = can.1PL A they.ACC = know ‘we can’t know them just by touching’ (Costin 1979: 121) These data clearly show that, in the absence of a, the infinitive can be analyzed as a [V] category, but it cannot be analyzed as a clause. Therefore a is not an inflectional mark (i.e., it is divorced from the grammatical mood, the latter being marked on the verb stem) but a complementizer.13 The fact that a is a complementizer and signals the presence of the CP field and, therefore, the clausal status of the infinitive, is of utmost importance for understanding what happened when să-subjunctive clauses emerged and replaced the infinitives around the 16th century: Since să-subjunctives were CPs, they replaced only the CP infinitives, as in (25), not the non-clausal infinitives in (24) (see Hill 2011 for a detailed discussion in this respect). Thus, an inflectional definition of infinitive a would not only contradict the results of cartographic tests, but it would also fail to predict this important generalization for the infinitive replacement in the language.

6 Recycling and reconstruction This section focuses on the property shared by all the sentential complements to control and non-thematic verbs in Early Modern Romanian: They all have the underlying structure represented in (14), (16), (21/22), although the morphological form of the verb may differ. So far, we have seen that the verb can come as an indicative (de-indicatives) or as an infinitive (a-infinitives). By the 17th century, să-subjunctives also emerge (where the verb has either an indicative or a subjunctive form), and their underlying structure matches (14), (16), (21/22), as shown in Hill (2013b). The first observation is, then, that the change does not affect the syntactic structure but the material that spells out the functional projections within this structure. Hence, the question is what triggers the change of the lexical material? Empirical observations show that, in the case of Early Modern Romanian, the switch in the type of complementizer (and of the verb form it selects) is related to the phasal status of the embedded clause. In particular, if the construction starts to be used more in its truncated FinP version rather than in its full-fledged ForceP version, some other type of full-fledged ForceP is adopted. 13 A similar conclusion can be inferred from the definition of a as a mixed C/I element in Dobrovie-Sorin (1994).

Early Modern Romanian infinitives: origin and replacement

163

Let us consider the case of de-indicatives. In Early Modern Romanian, these constructions occur mainly after causative and aspectual verbs as FinP (versus ForceP): That is, the embedded subject is obligatorily controlled, as shown in the causative construction in (26). (26) Că au pus Ştefan vodă de au arat cu leşii pe o culme that has=put Stefan King DE have=ploughed with Poles on a hill ‘For King Stefan ordered them to use the Poles to plough a hill’ (Neculce 117) Constructions as in (26) co-occur with a-infinitives in the language before the spread of să-subjunctives. One way of explaining the co-occurrence is to say that there was a specialization, de-indicatives being preferred for a resultative meaning, whereas a-infinitives were used for other types of realis or irrealis interpretations. This hypothesis, however, would be problematic because: (i) we can substitute a-infinitive to de-indicative in (26), and still obtain a resultative interpretation; (ii) de-indicatives are also complements to ‘desire’ verbs or to ‘order’ verbs: (27) a. au vrut de au făcut have= wanted DE have= done ‘they wanted to do’ (VC, 106 apud Frîncu 1969: 23/91) b. au poruncit de i-au tăiat capul has= ordered DE he.DAT =have= cut head.DEF ‘he ordered that they cut his head’ (Neculce 340) Examples as in (27) indicate that de-indicatives can be complements to verbs with irrealis interpretation, and may even have non-controlled subjects in this context, as in (27b). Thus, when we compare the distribution of de-indicatives and a-infinitives, the key element seems to be the frequency of occurrences: The former is dis-preferred under verbs with optional control, but still strong under verbs with obligatory control, whereas the latter shows a healthy distribution across the board in the early texts. Knowing that the sentential complements to verbs with optional control should be able to switch from FinP to ForceP according to the matrix verb’s requirements, it follows that de-indicatives are less able to do this than a-infinitives. This inability indicates that de-indicatives were worn out in the sense that de stopped being associated with the features of Force, and the construction became systematically non-phasal. Accordingly, deindicatives stopped being an option under verbs with optional control.

164

Virginia Hill

In the first written texts, a-infinitives are routinely found with a ForceP structure, in alternation with FinP. Thus, ForceP infinitives can license lexical subjects, as in (19a) and further in (28a, b), and null uncontrolled subjects, as in (28c). (28) a. Şi Domnul Domnedzeu lăsă a creaşte den allowed A grow from and Lord.DEF God pământ tot pomul frumos earth all tree.DEF beautiful ‘And the Lord God allowed all the beautiful trees to grow from the land’ (PO {16}) b. Nu iaste bine a fi omul sângur Not is good A be man.DEF alone ‘It is not good for a man to be alone’

(BB {FacereaCapII})

c. den fântână obiceaiu era a adăpa oile from fountain custom was A water sheep.DEF ‘it was the custom to give water to the sheep from the fountain’ (PO {96}) While such constructions are easy to find in the early texts, they become very rare in the later texts. For example, I found only two such examples in the Moldavian chronicles, the other occurrences of a-infinitives being non-phasal.14 These are also the texts in which the să-subjunctive appears as well established in the language. We can, then, see a correlation between the weakening of ainfinitives in complement position, whose selecting verbs become restricted to modals, causatives and aspectuals, and the spread of a new type of sentential complement, that is, the subjunctive. This trajectory looks familiar: The deindicatives attested in the early texts show the same pattern of distribution and FinP truncation (i.e., non-phasal status) as a-infinitives in the later texts. The subjunctive, on the other hand, is the preferred ForceP option as sentential complement, and is unrestricted as to the class of the selecting control/raising V.15

14 Infinitives with lexical subjects are still productive in adverbial clauses, but not in sentential complements. 15 MR already displays signs of weakening in the subjunctive: (i) the subjunctive occurs in both realis and irrealis contexts, while they were strictly irrealis in Early Modern Romanian; (ii) while ForceP was null in Early Modern Romanian, it has to be lexical in MR; thus, Early Modern Romanian să, de să, ca să were systematically in Fin, whereas in MR one of them has to move

Early Modern Romanian infinitives: origin and replacement

165

This analysis indicates that the syntactic structure of the sentential complement is constant throughout the history of Romanian: FinP with obligatory control on the embedded subject, and ForceP when control does not apply. Within this structure, Fin is lexical, whereas Force is null in declarative contexts; the features of Force are checked by the same complementizer that merges in Fin through distance Agree. Null Force does not fare well for language acquisition, because its features become easily disconnected from the complementizer in Fin. When this complementizer is reanalyzed as a checking item for Fin only, a new complementizer is used to check Force through distance Agree. The Early Modern Romanian data shows a systematic recycling of the material in Fin along these lines: As soon as one structure is analyzed as mostly truncated to FinP, a new complementizer arises that can be related to both Fin/Force. The replacement is so systematic that we can calculate the cycles backwards in time, and infer that de-indicatives were the first replacement for the inherited Latin -re infinitive, and that a-infinitives emerged in response to the re-analysis of de as checking only the features of Fin (versus Force as well). The representation of the Early Modern Romanian replacement cycles is given in (29), where the star means that the construction is not attested under verbs with optional control. (29)

Replacement cycles + Linguistic Reconstruction

*de-indicative (phasal) → de-indicative (non-phasal) ↑ a-infinitive (phasal) → a-infinitive (non-phasal) ↑ să-subjunctive (phasal) →. . . . This analysis allows us to predict that să-subjunctives will eventually be analyzed as non-phasal only, and that their maintenance as sentential complements will depend on the obligatory spell out of Force (either by să movement from Fin to Force or by the insertion of another complementizer in Force; e.g., ca in Force and să in Fin). If that fails, then we predict the replacement of să-subjunctives by another type of non-finite clause. to Force or else the structure is non-phasal. Evidence come from the word order, among other properties: (i) pentru [mulţimea] ca să nu-L calce pre El EMR for crowd.DEF that SA not=he.ACC step DOM him ‘so that the crowd does not step on him’ (NT 1648 {179} 9) (ii) pentru ca [mulţimea] să nu-L calce pe El MR for that crowd.DEF SA not-he.ACC step DOM him ‘so that the crowd does not step on him’

166

Virginia Hill

7 Conclusions This paper proposed a new perspective on the history of the infinitive clause in Early Modern Romanian, namely: – The Latin infinitive has been inherited as a noun or as a verb. – The replacement of the verbal infinitive took place twice: First, the long -re infinitive has been replaced by de-indicative and, eventually, a-infinitive; second, de-indicative and a-infinitive have been replaced by să-subjunctives. – The replacement was a syntactic event, applying only to bi-clausal structures. – The underlying structure of the sentential complement did not change; the replacement concerns only the spell out of Fin (different complementizers) and of T (different mood forms). – The replacement is cyclical and has predictive power for linguistic reconstruction. Accordingly, Romanian does not qualify for an exceptional status among the Balkan languages, since the first replacement of the infinitive must have taken place more or less on the same timeline as in the other languages in the area.

Primary sources BB

Coresi EV Costin Neculce NT PO Ureche Varlaam C

Chiţimia, I.C. 1988. Biblia, adecă Dumnezeiasca Scriptură (1688). Bucureşti: Editura Institutului Biblic. (www.sfantascriptura.com) – format pdf Re-edited in 1988. Bucureşti: Editura Institutului Biblic şi de Misiune Ortodoxă. Puşcariu, Sextil & Procopovici, Alexie. 1914. Carte cu învăţătură (1581). Bucureşti: Atelierele Grafice Socec & Co. Panaitescu, Petre P. 1979. Miron Costin, Letopiseţul Ţării Moldovei. Bucureşti: Editura Minerva. Iordan, Iorgu. 1955. Ion Neculce, Letopiseţul Ţării Moldovei. Bucharest: Editura de Stat. _____. 1988. Ştefan, Simion, Noul Testament (1648). Alba Iulia: Editura Episcopiei Ortodoxe Române. Pamfil, Viorica. 1968. Palia de la Orăştie 1581–1582. Bucureşti: Editura Academiei. Panaitescu, Petre P. 1958. Grigore Ureche, Letopiseţul Ţării Moldovei. Bucureşti: Editura de Stat. Byck, Jacques. 1964. Varlaam. Cazania. Bucureşti: Editura Academiei.

Early Modern Romanian infinitives: origin and replacement

167

Secondary references Alboiu, Gabriela. 2002. The Features of Movement in Romanian. Bucharest: EUB. Alboiu, Gabriela. 2010. A-Probes, Case, and (In)Visibility. LingBuzz. http://ling.auf.net/lingBuzz/ 001163 (August 23, 2013) Alboiu, Gabriela & Virginia Hill. 2013. Perception verbs and evidential syntax. Revue Roumaine de Linguistique LVIII (3). 275–298. Bošković, Željko. 1997. The Syntax of Nonfinite Complementation: An Economy Approach. Cambridge, Massachusetts: The MIT Press. Bošković, Željko. 2007. On the locality of motivation of Move and Agree. Linguistic Inquiry 38: 589–645. Chivu, Gheorghe et al. (eds.). 1997. Istoria limbii române literare. Epoca veche. [The History of Romanian Literary Language. The Old Period.] Bucharest: Editura Academiei Române. Chomsky, Noam. 2008. On phases. In Robert Freidin, Carlos P. Otero & Maria Luisa Zubizarreta (eds.), Foundational Issues in Linguistic Theory: Essays in Honor of Jean-Roget Vergnaud, 133–167. Cambridge & Massachusetts: The MIT Press. D’Alessandro, Roberta & Adam Ledgeway. 2010. At the C-T boundary: Investigating Abruzzese complementation. Lingua 120. 2040–2060. Diaconescu, Ion. 1977. Infinitivul in limba română. [The Infinitive in Romanian Language]. Bucharest: Editura ştiinţifică şi enciclopedică. Dobrovie-Sorin, Carmen. 1994. The Grammar of Romanian. Berlin: Mouton de Gruyter. Dragomirescu, Adina. 2013. Particularități sintactice ale limbii române în context romanic. Supinul. [Syntactic Peculiarities of Romanian Language in the Romance Context. The Supine]. Bucharest: Editura Muzeului Național al Literaturii Române. Fischer, Iancu. 1985. Latina dunăreană. [Danubian Latin]. Bucharest: Editura Academiei. Frâncu, Constantin. 2009. Gramatica limbii române vechi (1521–1780). [The Grammar of Old Romanian]. Iaşi: Demiurg. Frîncu, Constantin. 1969. Cu privire la “uniunea lingvistică balcanică. Înlocuirea infinitivului prin construcţii personale în limba română veche. [On the linguistic unity in the Balkans. The replacement of the infinitive by finite constructions in old Romanian]. Anuar de lingvistică şi istorie literară 20. 69–116. Hale, Mark. 2007. Historical Linguistics. Toronto: Wiley-Blackwell. Hill, Virginia. 2011. Modal grammaticalization and the pragmatic field: A case study. Diachronica 28(1). 25–53. Hill, Virginia. 2013a. The emergence of the Romanian supine. Journal of Historical Linguistics 3(2). 230–271. Hill, Virginia. 2013b. The emergence of the Romanian subjunctive. The Linguistic Review 30(4). 1–37. Jordan, Maria. 2009. Loss of Infinitival Complementation in Romanian. Diachronic Syntax. Miami & Florida: University of Florida dissertation. Kayne, Richard. 1994. The Antisymmetry of Syntax. Cambridge, Massachusetts: The MIT Press. MacRobert, Catherine Mary. 1980. The Decline of the Infinitive in Bulgarian. Oxford: Somerville College – University of Oxford dissertation. Miller, Gary. 2000. Gerund and gerundive in Latin. Diachronica 17(2). 293–349. Mišeska-Tomić, Olga. 2006. Balkan Sprachbund Morpho-syntactic Features. Dordrecht: Springer.

168

Virginia Hill

Motapanyane, Virginia. 1991. Theoretical Implications of Complementation in Romanian. Geneva: University of Geneva dissertation. Motapanyane, Virginia. 1994. An A-position for Romanian subjects. Linguistic Inquiry, 25(4). 729–734. Nedelcu, Isabela M. 2013. Particularități sintactice ale limbii române în context romanic. Infinitivul. [Syntactic Peculiarities of the Romanian Language in the Romance Context. The Infinitive]. Bucharest: Editura Muzeului Național al Literaturii Române. Niculescu, Dana I. 2013. Particularităţi sintactice ale limbii române din perspectivă tipologică: gerunziul. [Syntactic Peculiarities of Romanian Language from a Typological Perspective. The Gerund]. Bucharest: Editura Muzeului Naţional al Literaturii Române. Pană Dindelegan, Gabriela (ed.). 2013. The Grammar of Romanian. Oxford: Oxford University Press. Pană Dindelegan, Gabriela. 2014. Does Romanian have mixed categories? In Gabriela Pană Dindelegan, Rodica Zafiu, Adina Dragomirescu, Irina Nicula, Alexandru Nicolae and Louise Esher (eds.), Diachronic Variation in Romanian. 197–224. New-Castle: Cambridge Scholar Publishing. Paoli, Sandra. 2003. COMP and the Left-Periphery: Comparative Evidence from Romance. Manchester: University of Manchester dissertation. Rivero, Maria Luisa. 1994. Clause structure and V-movement in the languages of the Balkans. Natural Language and Linguistic Theory 12(1). 63–120. Rizzi, Luigi. 1997. The fine structure of the left periphery. In Liliane Haegeman (ed.), Elements of Grammar, 281–339. Dordrecht: Kluwer. Rizzi, Luigi. 2004. Locality and left periphery. In Adriana Belletti (ed.), Structures and Beyond: The Cartography of syntactic structures, vol. 3, 223–252. Oxford: Oxford University Press. Rosetti, Alexandru. 1968. Istoria limbii române de la origini până în secolul al XVII-lea. [The History of the Romanian Language from Origins to the 17th Century]. Bucharest: Editura pentru literatură. Rohlfs, Gerhard. 1933. Scavi linguistici nella Magna Graecia. [Linguistic Investigations in Magna Graecia]. Roma: Collezione Meridionale Editrice. Sandfeld, Kristian. 1930/1968. Linguistique balcanique: problèmes et résultats. [Balkan Linguistics: Problems and Results]. Paris: Klincksiek. Schulte, Kim. 2007. Prepositional Infinitives in Romance. Bern: Peter Lang. Stan, Camelia, 2013. La nominalizzazione dell’infinito in rumeno – osservazioni diacronicotipologiche. [The nominalization of the infinitive in Romanian – diachronic and typological observations]. Revue roumaine de linguistique LVIII (1). 31–40. Zanuttini, Raffaela. 1997. Negation and Clausal Structure: A Comparative Study of Romance Languages. Oxford: Oxford University Press.

Augustin Speyer

7 Semantic factors for the status of control infinitives in the history of German 1 Introduction In German, as in many other languages such as e.g. English, Latin and Ancient Greek, sentential complements of several verbs can be realized as infinitive constructions. In German, we find many different types of infinitive constructions, such as raising constructions, several control infinitive constructions, etc. In this contribution I concentrate on two types of constructions which share a similarity in that they represent clausal complements of verbs, namely the so-called accusativus cum infinitivo (AcI) construction (cf. (1a)), and control constructions, among which we have to distinguish between subject control infinitive constructions (SCIC; cf. (1b)) and object control infinitive constructions (OCIC; cf (1c)).1 (1)

a.

Wir we.NOM aus out

spürten sensed

Feuer fire

und and

unter under

uns us

Lärm noise].ACC

ein [a

gigantisches giant

Inferno inferno

losbrechen. loose.break.INF(St.1)

‘We felt that a giant inferno of fire and noise broke out beneath us.’ (Source: Merbold 1986: 23)

Acknowledgement: This study profited considerably from the contributions by audience members of two talks upon which it is based, given at the SLE conference 2013 at Split, Croatia, and in the lecture series of the department for German Linguistics at Saarland University in November 2014. I want to thank especially Werner Abraham, Ulrike Demske, Olga Fischer, Livio Gaeta, Łukasz Jędrzejowski, Sergey Kulakov, Christian Ramelli, Philipp Rauth, Ingo Reich, and two anonymous reviewers. I also want to thank my faithful aids at Göttingen University, Marten Santjer and Leonie Zitzmann, for their help in gathering the data. In the end I wish to thank another reviewer and Jonathan Watkins for going over my English. All remaining errors are my own. 1 The abbreviations used in the glosses are: acc = accusative, comp = non-translatable complementizer, dat = dative, gen = genitive, inf = infinitive, nom = nominative, ptc = verbal particle, st1 = status I (bare infinitive), st2 = status II (zu-infinitive). In order to ease readability, case labels are not attached to all elements of a noun phrase, but rather the noun phrase is bracketed in the gloss and the case label attached to the bracket. Augustin Speyer, Saarland University DOI 10.1515/9783110520583-000

170

Augustin Speyer

b.

hoffte. . . hoped einen [a

ichi , I.NOM

stillen silent

PROi

ihm him.DAT

Dank thank].ACC

auf on

diese this

Weise mode

abzustatten. to.render.INF(ST.2)

‘I hoped to thank him in silence this way.’ (Source: Merbold 1986: 9) c.

Für for PROi

den this auf on

Fall. . . case

hatte had

den the

nächsten next

man one.NOM

unsi us.ACC

Neumond . . . new.moon

aufgefordert, requested

zu warten. to.wait:INF(ST.2)

‘In this case one requested from us that we wait until the next new moon.’ (adapted from Merbold 1986: 18) Control constructions received their name from the classical analysis of Chomsky and Lasnik (1977). The logical subject of the infinitive is not represented overtly; it is realized as an empty element PRO that is coreferential with either the subject (in SCICs) or the object. The trees in (2) illustrate this in a simplified form, showing only the verb phrases of the examples in (1), without functional overhead. Tree (2a) represents the AcI construction; trees (2b, c) both represent control constructions. (2)

a.

AcI

Semantic factors for the status of control infinitives in the history of German

b.

SCIC

c.

OCIC

171

Judging from the semantic setup, that is, the grammatical roles of the participants and their distribution over the two clauses involved, AcIs and OCICs are rather similar. In Speyer (2015) it was argued that they are structurally similar, too, so that the tree (2a) is actually identical to the tree in (2c), the only (notational) difference being that the logical subject of the AcI clause is rendered noncommittally by e and not by PRO (which does not, however, exclude the possibility that it might be PRO in the end). This is the difference in deep structure; in surface structure AcIs regularly undergo clause union with the matrix clause, unlike OCICs (Speyer 2015). Regarding their morphological and syntactic properties, however, at least in German there are several differences, as can be seen by comparing (1a) and (1c). The two differences pertain to coherence and status. Following the by now well-established status-model proposed by Bech (1955), in AcIs the infinitive appears in status I (= bare infinitive), whereas in

172

Augustin Speyer

OCICs it usually appears in status II (= zu-infinitive).2 This difference is apparent in (1a,b). The difference with respect to coherence can be illustrated with the examples in (3). Adopting Bech’s definition (1955), coherence in infinitive syntax basically refers to the potential of being extraposed to the right: Obligatorily coherent infinitives cannot be extraposed, while obligatorily incoherent infinitives need to be extraposed and cannot appear in the clause nucleus (the so called middle field). The third group are infinitives that can, but need not, be extraposed (optionally [in]coherent). AcIs are obligatorily coherent, as is demonstrated by the low degree of acceptability of (3a), a version of (1a) in which the infinitive 2 There are exceptions, most notably helfen ‘to help’ and lehren ‘to teach’. Let us first consider helfen. In general, the above statement holds only for infinitives governed by verbs. Infinitives governed by other elements, most notably prepositions, show status II categorically. A reviewer pointed out that there is variation e.g. in the case of the object control verb helfen ‘to help’. Here the status is variable if the infinitive is directly governed by the verb. If it is governed by the correlate dabei ‘at that’, status II is compulsory (i). (i) a. Sie hilft mir die Aufgabe (zu) lösen. she helps me the problem to solve b. Sie hilft mir dabei, die Aufgabe *(zu) lösen. she helps me at-this the problem to solve ‘She helps me to solve the problem.’ (ii) Sie hilft bei der Renovierung. she helps at the renovation ‘She helps with the renovation.’ In my opinion, the problem here is that the infinitive is not governed directly by the verb, but by the prepositional element bei in dabei, which in turn is governed by helfen. Helfen can govern PP-complements with bei (ii). Infinitives governed by prepositions, however, show status II across the board. There is then a syntactically defined domain for compulsory usage of status II, namely if governed by a PP (or probably by everything that is not a verb), but we still can say that there is no syntactically defined domain that holds for infinitives governed by verbs. Turning to lehren, this argument does not hold up. The fact that there are constraints on the usage of status I with lehren (only bare verbs without complements or modifiers are possible: Sie lehrt ihn schwimmen ‘she teaches him how to swim’, but *Sie lehrt ihn im Meer schwimmen ‘she teaches him to swim in the sea’) suggests that the complement of lehren is probably not a normal infinitive in such cases, but that it is rather a verb-verb compound of the type kennenlernen ‘get to know’. A further piece of evidence pointing in this direction is the fact that nothing can intervene between the infinitive and the verb if both are in the right sentence bracket: *weil sie ihn schwimmen drei Wochen lang lehrte ‘because she taught him how to swim for three weeks’. In the version with a zu-infinitive, both the extension with complements and the possibility to place material between the infinitive and lehren, are possible: Sie lehrt ihn, im Meer zu schwimmen ‘she teaches him how to swim in the sea’; weil sie ihn zu schwimmen drei Wochen lang lehrte ‘because she taught him how to swim for three weeks’. The orthography should not concern us, since it is poorly regulated in the case of verbal compounds, cf. Gallmann (2000).

Semantic factors for the status of control infinitives in the history of German

173

has been extraposed.3 OCICs, on the other hand, can be coherent (in Bech’s sense), as in (3b), or incoherent, as in (1c). (3) a. ?*. . . dass wir spürten unter uns ein Inferno that we.NOM sensed under us [a inferno].ACC losbrechen. loose.break.INF(St.1) ‘that we felt that a giant inferno of fire and noise broke out beneath us.’ b. Für den Fall. . . hatte man uns for this case had one.NOM us.ACC auf den nächsten Neumond . . . zu warten aufgefordert. on the next new.moon to.wait.INF(ST.2) requested ‘In this case one requested from us that we wait until the next new moon.’ These differences in form are unexpected, given that both constructions seem to serve the same function, namely that of expressing propositional complements. Upon taking a closer look, one notices some differences that might indicate either a semantic or syntactic distinction. A well-known difference is for instance that AcIs governed by perception verbs can be replaced by a subordinate clause introduced by either wie ‘how’ or dass ‘that’. Note however that the wie-variant is semantically equivalent to the AcI, whereas the dass-variant shows semantic differences in that it does not presuppose simultaneity of the events (cf. Kretschmar 20134 as well as Jędrzejowski & Demske this volume for a broader

3 Coherence is defined in accordance with Bech (1955). It is more or less the property of being able to form a verbal complex with the matrix verb, which in turn means that clause union takes place (Pittner 1991). The distinction in analysis that is mentioned here and discussed at length in Speyer (2001, 2015) pertains to the infinitival clause ‘before’ clause union takes place (adopting a procedural view of syntactic operations). Coherence is defined slightly differently in other publications (e.g. Sternefeld 2006; Haider 2009), so that the statements made in this article might be incongruent with regards to the definitions of coherence used there. 4 The semantic statements Kretschmar (2013) makes about wie-clauses hold for AcIs as well. The reason why sometimes an AcI and sometimes a wie-clause is used might have more so to do with constraints on information density management than with a semantic difference between the two. English shows a similar effect in that bare infinitives after perception verbs are semantically not equivalent to that-clauses that seemingly describe the same event. See e.g. Mittwoch (1990: 104).

174

Augustin Speyer

context). The complementizer wie is devoid of its original meaning ‘how, in what way’ in these cases. OCICs, on the other hand, can only be replaced by a dassclause (cf. (4)).5 (4) a. wir spüren, wie das Inferno unter uns losbricht we.NOM sense COMP [the inferno].NOM under us loose.breaks b. wir spüren, dass das Inferno unter uns losbricht we.NOM sense that [the inferno].NOM under us loose.breaks c. *Man forderte uns auf, wie wir auf Neumond warteten one.NOM requested us.ACC PTC COMP we on new.moon waited d. Man forderte uns auf, dass wir auf Neumond warteten one.NOM requested us.ACC PTC that we on new.moon waited Another difference is that the logical subject of the subordinate proposition has to be expressed overtly after verbs like sehen ‘to see’ (*wir sehen _ die Treppe hinunterkommen ‘we see [someone] walking down the stairs’) but not after verbs like bitten ‘to ask’ (wir bitten _, die Gelegenheit aufzuklären ‘we ask [sc. someone] to clarify the matter’). A look at the history of the OCIC reveals that at least the status assignment has not always been as strict as it is today. In Section 2 of this study, the historical record is presented briefly. In Section 3, an attempt is made to identify a factor determining the variation in status.

2 The history of the status assignment in OCICs The status distinction in infinitival constructions cannot be very old, as the zu-infinitive only developed in the course of Old High German. In OHG, zu ‘to’ 5 Some control verbs like lehren ‘to teach’ can also govern other type of complement clauses, so-called dependent interrogative clauses. There is a superficial similarity of dependent interrogative clauses introduced by wie‚ how’, as in Der Vater lehrte seinen Sohn, wie man fliegt‚ ‘the father taught his son how to fly’, and the complement clauses governed by perception verbs introduced by wie ‘[complementizer]’, as in (4a). Note that the wie in the father-son-example is different from the wie in (4a): In the father-son-example, it is a real interrogative adverb of manner, whereas in (4b) this is not the case. This can be tested by replacing wie with another suitable expression of manner, e.g. die Kunst, mit deren Hilfe. . . ‘the art, with the help of which. . .’. The father-son-example allows for this replacement: Der Vater lehrte seinen Sohn die Kunst, mit deren Hilfe man fliegt ‘the father taught the son the art, with the help of which one can fly’, the AcI-example (4a) does not allow this type of substitution, at least not with the original reading: *Wir spüren die Kunst, mit deren Hilfe das Inferno unter uns losbricht ‘we sense the art, with the help of which the inferno under us breaks loose’.

Semantic factors for the status of control infinitives in the history of German

175

still shows properties of a preposition governing a nominalized infinitive in many cases. This is apparent in cases in which the infinitival form shows a dative ending which is governed by zu. From there, a process of reanalysis, triggered by semantic bleaching – that is, the gradual loss of meaning components – of zu took place. In the case of the zu-infinitive, this path led from an expression denoting purposive meaning to a form expressing realis modality (Haspelmath 1989). In descriptive terms, such processes in which a loss of semantic features triggers reanalysis are often referred to as ‘grammaticalization’ (cf. Haspelmath 1989; Abraham 2004 on the grammaticalization of the zuinfinitive). At the end of this process, the zu-infinitive developed into an infinitival form of its own, with zu being either an exponent of v° (Abraham 2004) or some head in the I-architecture, e.g. T° (see below in Section 3). The ‘new’ zu-infinitive consequently enters into competition with the ‘old’ bare infinitive. This process actually had already begun during the OHG period, as can be seen from usages of the zu-infinitive that cannot be analyzed as preposition-noun-complexes (Demske 2001). From Middle High German onwards, we then see an ongoing process of specialization of the two infinitive forms that finally results in a nearly categorical distribution of contexts in which one form (bare infinitive, status I) or the other (zu-infinitive, status II) is used (so-called Statusrektion). Demske (2001) shows that control infinitives are variable with respect to their status in Old High German, which is what we would expect if the form is only just developing and is still busy finding its role, so to speak.6 Clearly then, there is a noticeable development. In order to investigate this process I looked at the correlation of OCIC with the infinitival status in Middle High German and the first decades of the Early New High German period. The data spans from 1101 to 1500 AD. I selected texts from the digital databases publicly available for these periods, the Mittelhochdeutsche Begriffsdatenbank and the relevant portions of the Bonner Frühneuhochdeutschkorpus, specifically all prose texts from these periods (excluding the Prose-Lancelot) plus some poetic texts, in order to slightly increase the database without bringing in too much of a bias pro poetic texts.7 As these corpora are not parsed I could not 6 Another thing that Demske (2001, 2008) notes is that in OHG the status is not correlated to coherence. This correlation is visible in Modern German (von Stechow 1990) in that obligatorily coherent infinitives are in the first status, although not all infinitives in the second status are optionally incoherent (scheinen ‘to seem’, for instance, is quite restricted in terms of incoherent constructions). 7 I used the categories Kleine Erzählungen/Fabeln/Lehrgedichte; Juridische Texte; Medizinische Texte; Naturwissenschaftliche Schriften; Predigten; Rechtsurkunden; Religiöse Prosa in the MHDBDB and the texts of the first period from the Bonner FnhdC. The references to the texts in the examples of this paper refer to the digital versions that are part of the corpus. As the status of

176

Augustin Speyer

search for infinitives directly, but instead searched for several sample verbs governing an OCIC, namely bitten ‘to ask s.o. to do s.th.’, eischen ‘to induce s.o. to do s.th.’, hindern ‘to hinder s.o. in doing s.th.’, mânen ‘to request/admonish/ remind s.o. to do s.th.’ twingen ‘to force s.o. to do s.th.’, wîsen ‘to induce s.o. to do s.th.’, and then manually extracted those hits where these verbs really do govern an infinitive clause. These verbs combine with infinitives very rarely in Middle High German, so I could find only 92 examples. These examples are distributed over the statuses as shown in Table 1. Table 1: Distribution of infinitives in OCIC to status, whole time range

1101–1500

status I

status II

sum

% status I

62

30

92

67

We can see that more than two thirds of all control infinitives are bare. Two examples of this type are given in (5, 6).8 This is surprising from a modern point of view, where the rate of bare infinitives is negligible, but it continues the Old High German state of affairs. (5)

diu that

mich . . . me.ACC

den [the

rîchen rich

der [[the

edlen noble

herren lords].DAT

künste arts].GEN

swaere importance].NOM

künden announce.INF(St.1)

bat asked

‘[the story,] which [the personification of] Importance of Noble Arts urged me to announce to the rich lords’ (Konrad v. Würzburg: Die Klage der Kunst, 32,8; c.1250–1287) (6)

die them.ACC

bat asked

er he

unde and

mande requested

sîn [his

gebot commandment].ACC

ervollen fulfil.INF(St.1) ‘he asked them and ordered them to obey his commandments’ (Buch von Akkon, 50750f.; c.1300–1320) the infinitive should be irrelevant for verse construction – in the Middle Ages, we still have the principle of ‘freie Senkung’, which means that the poet can insert as many unstressed syllables as he wants as long as the number of stressed syllables is constant – I felt justified in including – yet to a moderate degree – poetic texts. 8 Note that directive matrix predicates are also amongst the predicates that govern bare infinitives (contra Smirnova 2011).

177

Semantic factors for the status of control infinitives in the history of German

Over the 200 years that stretch between 1201 and 1400, there is variation. If we spread Table 1 and report the rates of status I in the 4 centuries separately, we see a clear predisposition (Table 2, Figure 1). Table 2: Distribution of infinitives in OCIC to status, 100-year periods separated

1101–1200 1201–1300 1301–1400 1401–1500

status I

status II

sum

% status I

4 50 8 0

0 9 11 10

4 59 19 10

100 85 42 0

Figure 1: Distribution of infinitives in OCIC to status, 100-year periods separated

We can say with some caution – considering the scarcity of the data – that the status in OCICs developed relatively smoothly from a seemingly categorical status I in the 12th century to a categorical status II in the 15th century. As status II is categorical today, one might be inclined to take at least this part of the development seriously. In any case it is safe to say that the replacement of status I by status II took place during a period of at least 200 years. The question is whether it was simply a replacement following the lines of Kroch’s (1989, 2000) competing grammar model or whether the variation followed some system during the transition phase. In the following I take the stance that the variation was at least influenced by some semantic factors, two of which are discussed.

178

Augustin Speyer

3 Potential reasons for the differentiation 3.1 Aspect and temporality Let us assume that there was some system governing the variation during the transition period. The reason for this assumption is that there is still variation in the infinitival system today and that the modern variation is quite systematic. If it was simply a matter of one infinitive form replacing the other, we would expect this to be extended to all contexts eventually, but obviously this did not happen. The reason why this syntactic change did not develop into a wholesale replacement – as syntactic changes of the kind described by Kroch and explained by the competing grammar model usually do – must be that the ‘old’ status I infinitive managed to preserve some strongholds, so to speak, that is, it found a domain in which it was no longer in competition with the new status II infinitive. In Speyer (2015) it is argued that at least in the subdomain of OCICs there are no hints suggesting that this domain is defined syntactically. Consequently it would then be productive to verify whether there is a semantically defined domain that status I could ‘defend’ against status II, and likewise, whether there is a direct way from this domain to the modern division of labor between the statuses. In English, where a similar process (the to-infinitive taking over some domains of the bare infinitive) occurred, the choice of an infinitive form was clearly governed by semantic considerations in the Middle English period, at which time the system developed (Fischer 1995; see also D’hoedt & Cuyckens this volume). The semantic factors holding for English can be considered as follows: If the infinitival clause is part of the same event described by the matrix clause, a bare infinitive is used; if it is a related, but distinct event, the to-infinitive is used (Fischer 1995: 19–20). In the following we concentrate on the two periods in which there is variation and in which the material is a bit less sparse than in the other periods. Before we do so, a brief note of caution is in order: As it turns out, it will not be possible to pinpoint a single semantic factor, which is partly due to the fact that there is not much material available to properly quantify, thus only allowing us to observe tendencies. Yet, this investigation is intended to induce further investigations into the matter, and to state a few ideas that could guide future research. For instance, the inclusion of metrical texts, which are abundant, might enlarge the database in future studies, but since syntactic studies on metrical texts face serious well-known problems, it is useful to have a treatise at hand in which this problem is avoided, yet at the cost of somewhat provisional results.

179

Semantic factors for the status of control infinitives in the history of German

One semantic category that appears in German verbal morphology every now and then again is the notion of aspect. While there is no systematic marking of aspect in German like there is in e.g. the Slavic languages (and as there was, presumably, in the Proto-Indo-European predecessor of German), occasionally strategies appear that encode aspectual differences (see also Wiemer this volume for the role of aspect in Slavic independent infinitives). This holds for as diverse fields as the selection of case in Old High German (Donhauser 1998), the presence or absence of the prefix ge- in Old High German verb forms (Oubouzar 1974), the rise of the haben/sein-perfect alongside the inherited preterite (Oubouzar 1974; Grønvik 1986; Abraham 1999), and the development of the ‘Rheinische Verlaufsform’ in the course of the last couple of centuries (van Pottelberge 2004; Ramelli 2016). So aspect presents itself as a parameter that is worth examination. In fact, the infinitive was explicitly related to aspect in earlier research (Abraham 2004), which makes the assumption even more plausible. I found indeed a correlation (Tables 3, 4; Figure 2). In both 100-year periods (13th and 14th century), there is a correlation between status and aspect, insofar as status I is concerned with what I describe as ‘punctual’ aspect, following traditional classicist grammar (cf. Leumann, Hofmann & Szantyr 1977: 508), whereas status II is rather related to ‘durative’ aspect, although here the correlation is less clear.9 In the second period the overall portion of status II is bigger, which is reflected in a parallel downswing of the rate of status I in both aspectual classes. Table 3: Distribution of infinitives in OCIC to status in relation to aspect, 13th century 1201–1300

status I

status II

sum

% status I

Durative Punctual

4 45

8 1

12 46

33 98

Table 4: Distribution of infinitives in OCIC to status in relation to aspect, 14th century 1301–1400

status I

status II

sum

% status I

Durative Punctual

1 7

9 2

10 9

10 78

9 These seem to be the correct descriptions for the Proto-Indo-European aspectual distinctions (e.g. Leumann, Hofmann & Szantyr 1977: 508f.; Sihler 1995: 446), so it is not surprising that it is exactly this distinction that pops up in German, being a descendant of Proto-Indo-European.

180

Augustin Speyer

Figure 2: Distribution of status I infinitives in OCIC to status in relation to aspect, 13th/14th century

The aspectual classes are defined as follows: ‘durative’ can be an action that takes a long time to perform (hence the name; s. e.g. Comrie 1976: 41–42). Other cases that fall under that definition are habitual or repeatedly performed actions (cf. (7a)), as well as actions that are implied to have taken place at several occasions (cf. (7b)). (7)

a.

den whom.ACC

man one

zesprigen to.jump.INF(St.2)

twanch forced

so so

lange, long

unz. . . until

‘whom one forced to jump up to that point that. . .’ (Der Stricker: Der Hofhund 73.41f.; c.1220) b.

den this.ACC warheit truth

twinget forces

leit Grief

zu klagen, to.mourn.INF(St.2)

disen. . . this.ACC

die the

zu sagen to.tell.INF(St.2)

‘Grief forces the one to mourn, the other to tell the truth’ (Johannes von Tepl: Der Ackermann aus Böhmen, 33.16; c. 1400) ‘Punctual’ events, on the other hand, are typically short and fleeting (cf. (8a); see e.g. Comrie 1976: 41–42). They can also take a longer time, but only happen once (cf. (8b)). This latter aspect is a complement to the habitual nature of the durative aspect. ‘Durativity’ is probably really not the right way to think about

Semantic factors for the status of control infinitives in the history of German

181

this category; rather, the distinction is whether the action has some internal structure (= durative) or not (= punctual; similar Comrie 1976: 41–42). (8)

a.

Dar there

nach after

in him.ACC

chomen came

die the

tummen dumb

magde maids

und and

baten asked

vf tuon open.INF(St.1)

‘After that the foolish bridesmaids came and asked him to open [the door]’ (Altdeutsche Predigten 30,35; before end 14th cent.) b.

den [the

kunic king].ACC

des [the

grâven count].GEN

man one

dô then

varen travel.INF(St.1)

bât asked

fur before

eine [a

stat city].ACC

‘One asked the king to travel to a city belonging to the count’ (Ottokar aus der Gaal: Steirische Reimchronik, v.35372f.; c. 1300) When dealing with aspect, one always faces the problem which aspectual differentiations to take. ‘Aspect’ does not simply boil down to the well-known perfectiveimperfective opposition; there are numerous ways to specify the internal temporal constituency of an action (cf. e.g. Comrie 1976; Dahl 1985; definition following Comrie 1976: 3). The definitions chosen here are similar to those that are usually used to account for the variation between the past tenses: ‘imperfect’ as opposed to ‘perfect/aorist’ in Latin and Ancient Greek (Sihler 1995: 446). The reason for this choice is that these definitions seem to provide correct descriptions of the aspectual distinctions in Proto-Indo-European (e.g. Leumann, Hofmann, and Szantyr 1977: 508–509; Sihler 1995: 446). It should come as no surprise that it is exactly this distinction that manifests itself in German, especially in its older stages, being a descendant of Proto-Indo-European: To look for inspiration in other Indo-European languages seems to be a natural choice. Another piece of evidence supporting this point of view comes from the verb heizzen ‘to make so. do sth.’. I searched the MHDBDB (same text types as for the control verbs) for heizzen and manually picked out all instances of heizzen which combined with an infinitive clause. I found 330 tokens with heizzen with a bare infinitive and no tokens containing heizzen with a zu-infinitive. It should be noted in this context that the requests expressed in the examples all refer to a singular action, not a repeated or habitual one. The choice of status would follow directly from the aspectual assignment discussed above.

182

Augustin Speyer

Linguistic indicators of aspect have a certain tendency to be reinterpreted as indicators of tense or other temporal values. This happens in many languages (for example in Latin, where the old Indo-European aspectual classes of stative and perfective have been mashed together to form the perfect tense, cf. Sihler 1995: 452), and also in Germanic languages, with German being very much in line with this tendency. In Proto-Germanic, the Indo-European stative aspect was reinterpreted as a past tense. Later, in German, an aspectual construction (the perfect with ‘haben/sein’ + past participle) came to be used as the default past tense in the southern dialects (e.g. Dentler 1998; Abraham 1999). Could we observe a similar development with the infinitival status? It is interesting to note in this context that the status assignment in Modern German is sensitive to temporal values. The AcI is confined to cases of direct perception, that is, simultaneity of events, and is represented with a bare infinitive (status I). By contrast, the ‘classic’ object control verbs entail nonsimultaneity of propositions. Verbs like bitten ‘to ask’, auffordern ‘to request’ or hindern ‘to hinder’ entail that the matrix proposition temporally precedes the proposition expressed in the infinitival clause; verbs like beschuldigen ‘to accuse’ entail that the matrix proposition temporally follows the proposition expressed in the infinitival clause. It is not inconceivable that the punctual aspect developed into temporal simultaneity and durative aspect into temporal dis-simultaneity. A bridge between the punctual aspect and simultaneity is provided by the property of the punctual aspect being used to highlight that an event takes place only once. When we say that the time of the single occasion of the proposition expressed in the infinitive clause coincides with the time of the matrix proposition, we state a necessary condition for direct perception, which entails this coincidence of times. On the other hand, the property of the durative aspect to denote that an action takes place on several occasions, that is, on a number n of occasions, with n > 1, entails that there is no simultaneity with the matrix proposition of at least n – 1 occasions, 1 being one occasion that could coincide temporally with the matrix proposition.10 Using this bridge, the temporality assignment could have developed out of the original aspectual assignment of statuses. Note that in English a similar process can be observed; Fischer (1995) showed for Middle English that the choice between the bare infinitive

10 In this context an observation by Haspelmath (1989: 300) is of relevance, namely, that complement clauses in realis-modality have indeterminate time reference. According to Haspelmath (1989), the zu-infinitive developed out of a verbal noun with purposive function, the last step on the cline of semantic bleaching being realis modality. Based on this point, one could say that the bare infinitive is used with determinate time reference (namely referring to the same time as the matrix event), while the zu-infinitive is not.

Semantic factors for the status of control infinitives in the history of German

183

and the to-infinitive (which is historically the equivalent of the German status II infinitive or zu-infinitive) is governed by a temporality assignment: Simultaneity and direct perception are related to bare infinitives, dis-simultaneity and lack of direct perception to to-infinitives. From this view, we would also get a plausible scenario for the language change that happened in the course of Middle High German / Early New High German, and that led to the morphosyntactic differentiation of AcI and OCIC in the later development of German (see Speyer 2015). The aspectual difference was originally hosted relatively low in the IP-architecture, in Asp°, where the verb moved obligatorily by ‘trolleying’ (cyclic head movement).11 This can be regarded as a remnant of the Proto-Indo-European state of affairs, where verb forms obligatorily exhibited aspect marking. Aspectual distinctions are often reinterpreted as temporal distinctions, which is structurally reflected in the fact that the features that were originally hosted in Asp° are reinterpreted as being hosted by T°. This is what happened in the course of Middle / Early New High German with the infinitives under discussion: While in infinitives in OCIC constructions the feature [+ durative] was reinterpreted as something like [non-default temporality assignment] and consequently was seen as being hosted in T°, in AcI infinitives the feature [- durative] was reinterpreted as the absence of the feature [nondefault temporality assignment], and therefore required no host; hence, the infinitive verb did not project into the functional IP-domain at all. The particle zu was likewise interpreted as a morphological exponent in T° added when the verb is trolleying through T°. This explains the fact that it is relatively close to the verb, closer than verbal particles and so on. A question that may be asked when dealing with language change is why a particular change happened at just the point in time at which we observe it. In the case of the development of the infinitival status system as an exponent of aspectual information, a reviewer raised a very interesting point: Most Slavic languages never developed an infinitival status system. In light of the correlation of the status system with aspect this is not unexpected, as aspect is marked on Slavic verb forms anyway, so an additional exponent in the form of something like zu in the German zu-infinitive is unnecessary. The same could be said for earlier Latin and Ancient Greek, where the aspectual specification of the infinitive was signaled by the stem form from which the infinitive was formed, and where no status system developed, either. In Old High German, aspect still played a greater role in the verbal system than in Modern German (e.g. Schrodt 2004: 104–110), the means of expression most notably being the gi-prefix. If there are other ways of expressing aspectual differences, an infinitival status system 11 Note that Felser (1999) argues in general for a status of infinitival clauses as AspP.

184

Augustin Speyer

would also be redundant and in fact, the origin of the zu-infinitive as a nominal verb form with a purposive or final meaning is still visible in many cases in OHG (see Abraham 2004). The status system begins to develop during OHG times (Demske 2001), but it is probably not yet correlated to aspect. In Middle High German, when the aspectual marking system based on the prefix gi- (MHG ge-) is weakened, the newly developed status system can be ‘recruited’ to express aspectual differences, as it is not redundant any more.

3.2 Thematic roles and causation Another property that distinguishes the classical AcI verbs from the classical OCIC verbs is that in OCIC verbs the event denoted by the infinitive clause has an external causer, whereas in AcI verbs this is not the case. In fact, the entire lexico-semantic structure of both classes of verbs is quite different. The classical AcI verbs are atelic verbs, states in the sense of Vendler (1957) and Dowty (1979). They assign an experiencer and a stimulus role (in the role model of Primus 2012), the stimulus being the direct object of which the remainder of the infinitive clause is predicated.12 The OCIC verbs, on the other hand, are telic verbs, accomplishments in the sense of Dowty (1979). They assign an agent role (the external causer), a patient (or, as it is sometimes – e.g. Wechsler 1997 – called, an affected theme), and a further argument – the control infinitive clause. It is doubtful what role is assigned to it by the matrix verb, if any role is assigned at all. So there are crucial differences in valence from a qualitative point of view, but also from a crude quantitative point: AcI verbs are always bivalent, OCIC verbs are always trivalent. Sentence (9a), consisting of an AcI verb, a subject and a direct object, is a semantically complete statement, while sentence (9b), consisting of an OCIC verb, a subject and a direct object, is not semantically complete. (9)

a.

Julia Julia.NOM

sieht sees

Philipp Philipp.ACC

b. #Julia Julia.NOM

bittet asks

Philipp Philipp.ACC

Similar differences have been shown to play a role in secondary predicates in English (Wechsler 1997), insofar as properties of the object – such as being the

12 I adopt the bi-constituental analysis of the AcI, for which I argued in Speyer (2001; 2015).

Semantic factors for the status of control infinitives in the history of German

185

affected theme – play a role for the selectional properties of the matrix verb in these constructions. There is, thus, independent evidence that linguistic differences can in general be due to the roles assigned by verbs. Looking at historical stages of English, Fischer (1995) showed that causation plays a role for the choice between bare infinitives and to-infinitives too, in that indirect causation is correlated with the to-infinitive. A similar bridge between causation and status assignment could in theory then also have existed in German. Can we conclude from that that status II infinitives came to be associated with the presence of external causers and/or affected themes and/or three argument slots in the lexico-semantic structure of the matrix verbs, whereas status I infinitives came to be associated with the absence of external causers and/or affected themes and/or the presence of only two argument slots? We would expect the generalization about causation and status assignment to hold not only for the AcI / OCIC distinction, but for other types of infinitival constructions as well (such as subject raising infinitives), so let us look at other cases. One problem that the conclusion reached above faces is that in Modern German causative verbs like lassen ‘to make s.o. do s.th.’ or heißen ‘to order s.o. to do s.th.’ are not construed with status II infinitives, but with status I infinitives,13 although the same lexico-semantic properties hold for them that hold for classical OCIC verbs like auffordern ‘to request s.o. to do s.th.’. Then, typical raising verbs like scheinen ‘to seem’ or versprechen ‘to be likely’ (in e.g. es verspricht, ein schöner Tag zu werden ‘it is likely to be a beautiful day’) take status II infinitives, although they do not assign any thematic roles, let alone an external causer or an affected theme. This is not to say that there are no connections between status and semantic roles. But obviously, more work is needed to uncover the exact connection. That there probably is a connection, is suggested by the fact that the sample verbs show clear differences in status assignment (Table 5, Fig. 3). Table 5: Distribution of infinitives in OCIC to status; sample verbs separated Per.3/4

bitten mânen twingen hindern

Per.5/6

Status I

Status II

Status I

Status II

49 (94%) 1 (25%) 0 0

3 (6%) 3 (75%) 3 (100%) 0

8 (73%) 0 0 0

3 (27%) 4 (100%) 3 (100%) 1 (100%)

13 See Reis (1976) on the assumption that infinitives governed by causative verbs and AcIs governed by perception verbs have a different structure.

186

Augustin Speyer

Figure 3: Rate of status I with selected sample verbs

Trying to account for this correlation from the perspective of causation and similar concepts, we see that the degree of involvement is quite different: With twingen ‘to force’ and hindern ‘to hinder’ there is potential physical involvement on the part of the external causer, while with mânen ‘to admonish/request/order’ and bitten ‘to ask’ there is no such physical involvement. This is not the distinction we are looking for, however, as the groups are bitten on the one hand versus the other three verbs on the other hand. We might try looking at them from a speech act point of view. All verbs denote directive speech acts in the sense of Searle (1979), at least (in the case of mânen, which can also mean ‘to remind so. to do sth.’) potentially. Likewise, the degree of personal mental involvement on the part of the speaker in the speech acts of hindering and forcing is probably high as a rule, whereas in the speech act of asking it is not necessarily high. Asking is located relatively low on the scale of directive speech acts. The speech act denoted by mânen, if used directively, is probably stronger than the one denoted by bitten, as mânen can mean ‘to order someone to do something’, which bitten cannot mean. One would have to check in the examples (and not only those few which govern an infinitive) whether mânen involves a higher degree of personal involvement on the part of the causer. For the tokens with infinitive we can make the following statement: In the first period they are mainly conjoined with bitten (10a); the person doing the asking/admonishing, however, regularly occupies a higher social rank than the asked/admonished one, so that it is rather an order than a request. In (10a), for instance, the addressee is Leutolt von Kuenring-Dürnstein, a member of the Austrian lower nobility that rebelled against Duke Albrecht I.14 The tokens in the second period 14 Source: http://geschichte.landesmuseum.net/index.asp?contenturl=http://geschichte. landesmuseum.net/personen/personendetail.asp_id=-648594253.

Semantic factors for the status of control infinitives in the history of German

187

are heterogeneous; whereas in some cases a certain amount of personal involvement is visible, in others this is not the case. In (10b), for instance, we might assume that improving the souls of the people is close to Iohannes’ heart, which is indicated by the fact that he also goes so far as to punish sinners. In (10c), on the other hand, the subject is an abstract entity, so personal involvement is excluded. In cases like (10c), which do not imply personal involvement, the verb mânen has an epistemic meaning (‘to remind s.o. to do s.th.’), so they are not comparable to bitten anyway. (10) a. der herzoge mant und bat den Kuonringaere the duke admonished and asked [the Kuonringer].ACC ze sagen wie im waere da=tz Beheim gelungen to.say.INF(St.2) how him was that=at Bohemia succeeded ‘The duke requested Leutolt to say how he managed to do what he did in Bohemia’ (Ottokar aus der Gaal: Steir. Reimchronik, v. 67474–67477, c.1300) in welen suinden b. er seit och [. . .] etlichen luiten he says also [many people].DAT in which sins guettlich vnd manet sy wa e rend vnd stra xffet sy they were and punishes them amicably ans admonishes si zebihten vnd zeruiwenne them to.confess.INF(St.2) and to.repent.INF(St.2) ‘he also says to many people what sins they have committed, punishes them in an amicable way and admonishes them to confess and repent [their sins]’ (Buch der heiligen Altväter, 98.17, 14th cent.) c. daz er=z tue durch di minne vnd durch die trewe that he=it do through the love and through the faithfulness di in ermante vns armen so hoh speis zv machen which him admonished us poor so high meal to.make.INF(St.2) ‘that he should do it because of the love and the faithfulness which reminds him to give us poor people such a festive meal’ (Benediktinerregel Nassau 12.24, 14th cent.) What is relatively certain, however, is that – regardless of the personal mental or physical involvement of the causer – the causer can be pretty sure that after having performed speech acts of ‘forcing’, ‘hindering’ and ‘requesting’, the

188

Augustin Speyer

action which s/he wanted the addressee to do is going to happen, whereas after having performed a speech act of ‘asking’, the causer cannot be sure whether the proposition is going to happen, as it is still, to an extent, up to the addressee whether s/he will comply with the request of the causer or not. So perhaps this is another factor influencing the choice of a status in the transition period. An argument against this viewpoint, however, is again that verbs like heizzen ‘to order’ also categorically combine with status I infinitives in Middle High German, during the transition period (see Section 3.1). Another problem is that the use of verbs of perception with AcI entails that an actual event is witnessed in real time, which would fall under the category of propositions that are actually happening. We would thus expect status II using the argumentation above in such cases, but we never do.

3.3 Sameness of events In spite of the objections mentioned in the preceding section, an important point raised by Fischer (1995) with respect to English infinitive complement clauses probably holds for German as well, though in a slightly different form: In English the bare infinitive is associated with events that are part of the matrix event (in the following referred to as sameness of events, or short sameness), whereas the to-infinitive is associated with events that are distinct from the matrix event. While in German status assignment is probably not the primary morphosyntactic means to express this distinction, another property of infinitive clauses that German infinitive clauses show, in contrast to English ones, might be associated with sameness or distinctness of the two events, namely coherence. In structural terms, coherence means that infinitive clauses undergo clause union with their respective matrix clause (cf. e.g. Pittner 1991; Haider 2009: 276). AcIs are obligatorily coherent, and they exhibit the semantic properties that characterize sameness of events. OCICs are not obligatorily coherent, and they do not exhibit the semantic properties characterizing sameness of events. The aspectual distinction expressed by the choice of infinitival status is to some extent congruent with the semantic properties of sameness of events, so that obligatorily coherent infinitives denoting a part of the matrix event (sameness) are usually status I because sameness of events implies punctual aspect by virtue of the infinitival event having no internal structure of its own. In the case of verbs that govern an OCIC, sameness of events does not necessarily hold, so they have a tendency to be incoherent. Accordingly, the morphosyntactic distinction between AcI and OCIC is probably really a mere epiphenomenon of semantic properties of the respective infinitival events, sameness and aspect.

Semantic factors for the status of control infinitives in the history of German

189

This strengthens the point that structurally – at least in deep structure before clause union might take place – these constructions are not distinct at all. As status assignment and coherence happen to be largely congruent, it is not surprising that they come to be regarded as being interdependent, which leads to the modern state of affairs in which there is a relatively strong connection between obligatory coherence and status assignment (von Stechow 1990).

4 Conclusions AcI and OCIC constructions are not yet formally distinct in Middle High German. The differences in form – AcI: status I, obligatorily coherent, OCIC: status II, optionally coherent – did not become categorical until the 15th century, probably even later (see Maché & Abraham 2011). Looking at the statuses, we can see that before 1200, there is no status II infinitive in either construction. While AcIs remained rather stable in using status I infinitives (and develop obligatory coherence), in OCICs there is variation between status I and status II for roughly 2 centuries. If we assume that this variation is not simply a matter of one grammar replacing another, but that it is governed, or at least influenced by grammatical, cognitive, or pragmatic factors, we should be able to identify some of these factors. In this study attention was focused primarily on semantic factors. It is indeed possible to identify two candidates. The less likely candidate is the presence or absence of an external causer, presence being correlated with status II, absence with status I. While this might explain, to some extent, the Modern German distribution (abstracting away from causative verbs like lassen in the sense of ‘to order s.o. do s.th.’), it does not seem to be the right generalization for Middle High German, where verbs like bitten ‘to ask’ have a much higher quota of status I than verbs like mânen ‘to request, order’ or twingen ‘to force’, although in all of these verbs there is an external causer present. This last-mentioned fact could be accommodated by a factor derived from the first factor, namely whether the causer thinks that the action demanded in his/her request will happen for sure (which is the case with twingen and mânen), or whether the result of the event is open (which is the case with bitten). Again, from this standpoint we would expect verbs like heizzen ‘to order’, as well as perception verbs, to govern status II infinitives, which they never do. The other, more likely candidate has to do with aspect: Status I is associated with punctual aspect, status II with durative aspect. From this association of status with aspect it is possible to derive the later temporality assignment, status I (= AcI) denoting

190

Augustin Speyer

simultaneity of the matrix event and the subordinate event, while status II (= OCIC) denotes dis-simultaneity.

References Primary sources and corpora Bonner FnhdC: Bonner Frühneuhochdeutschkorpus. URL: http://www.korpora.org/Fnhd (last access: February 28, 2015). Merbold, Ulf. 1986. Flug ins All. Bergisch Gladbach: Lübbe. MHDBDB: Mittelhochdeutsche Begriffsdatenbank. URL: http://www.mhdbdb.sbg.ac.at:8000/ (last access: February 26, 2015). TITUS: Thesaurus indogermanischer Text- und Sprachmaterialien. URL: http://titus.uni-frankfurt. de/ (last access: September 2, 2013).

Reserach literature Abraham, Werner. 1999. Preterite decay as a European areal phenomenon. Folia Linguistica 33. 11–18. Abraham, Werner. 2004. The grammaticalization of the infinitival preposition – toward a theory of ‘grammaticalizing reanalysis’. Journal of Comparative Germanic Linguistics 7. 111–170. Bech, Gunnar. 1955. Studien über das deutsche Verbum infinitum. København: Munksgaard. Chomsky, Noam & Howard Lasnik. 1977. Filters and control. Lingustic Inquiry 8. 425–504. Comrie, Bernard. 1976. Aspect. An Introduction to the Study of Verbal Aspect and Related Problems. Cambridge: Cambridge University Press. Dahl, Östen. 1985. Tense and Aspect Systems. Malden, MA & Oxford: Blackwell. Demske, Ulrike. 2001. Zur Distribution von Infinitivkomplementen im Althochdeutschen. In Reimar Müller & Marga Reis (eds.), Modalität und Modalverben im Deutschen (Linguistische Berichte Sonderheft 9), 61–86. Hamburg: Buske. Demske, Ulrike. 2008. Raising patterns in Old High German. In Thórhallur Eythórsson (ed.), Grammatical Change and Linguistic Theory. The Rosendal Papers, 143–172. Amsterdam & Philadelphia: John Benjamins. Dentler, Sigrid. 1998. Gab es den Präteritumschwund? In John Ole Askedal (ed.), Historische germanische und deutsche Syntax, 133–147. Berlin: Peter Lang. Donhauser, Karin. 1998. Das Genitivproblem und (k)ein Ende? Anmerkungen zur aktuellen Diskussion um die Ursachen des Genitivschwunds. In John Ole Askedal (ed.), Historische germanische und deutsche Syntax, 69–86. Berlin: Peter Lang. Dowty, David. 1979. Word Meaning and Montague Grammar: The Semantics of Verbs and Times in Generative Semantics and in Montague’s PTQ. Dordrecht: Reidel. Felser, Claudia. 1999. Verbal Complement Clauses. A Minimalist Study of Direct Perception Constructions. Amsterdam & Philadelphia: John Benjamins.

Semantic factors for the status of control infinitives in the history of German

191

Fischer, Olga. 1995. The distinction between to and bare infinitival complements in late Middle English. Diachronica 12. 1–30. Gallmann, Peter. 2000. Wortbegriff und Nomen-Verb-Verbindung. Zeitschrift für Sprachwissenschaft 18. 269–304. Grønvik, Ottar. 1986. Über den Ursprung und die Entwicklung der aktiven Perfekt- und Plusquamperfektkonstruktionen des Hochdeutschen und ihre Eigenarten innerhalb des germanischen Sprachraumes. Oslo: Solum. Haider, Hubert. 2009. The Syntax of German. Cambridge: Cambridge University Press. Haspelmath, Martin. 1989. From purposive to infinitive – a universal path of grammaticalization. Folia Linguistica Historica 10. 287–310. Kretschmar, Alexandra. 2013. The witness wins it all. The interplay of evidentiality, factivity and event type: The German conjunction wie ‘how’ following verbs of perception. Zeitschrift für Sprache und Sprachen 45. 32–54. Kroch, Anthony. 1989. Function and grammar in the history of English: Periphrastic do. In Ralph Fasold & Deborah Schiffrin (eds.), Language Change and Variation, 133–172. Amsterdam & Philadelphia: John Benjamins. Kroch, Anthony. 2000. Syntactic change. In Mark Baltin & Chris Collins (eds.): The Handbook of Contemporary Syntactic Theory, 629–739. Malden, MA & Oxford: Blackwell. Leumann, Manu, Johann Baptist Hofmann & Anton Szantyr. 1977. Lateinische Grammatik. vol. 1: Lateinische Laut- und Formenlehre. München: Beck. Maché, Jakob & Werner Abraham. 2011. Infinitivkomplemente im Frühneuhochdeutschen – satzwertig oder nicht? In Oskar Reichmann & Anja Lobenstein-Reichmann (eds.), Frühneuhochdeutsch – Aufgaben und Probleme seiner linguistischen Beschreibung, 235–274. Hildesheim. Olms. Mittwoch, Anita. 1990. On the distribution of bare infinitive complements in English. Journal of Linguistics 26. 103–131. Oubouzar, Erika. 1974. Über die Ausbildung der zusammengesetzten Verbformen im deutschen Verbalsystem. Beiträge zur Geschichte der deutschen Sprache und Literatur (Halle) 95. 5– 96. Pittner, Karin. 1991. AcI-Konstruktionen und Valenz. In Eberhard Klein, Francoise Puradier Duteil & Karl Heinz Wagner (eds.), Betriebslinguistik und Linguistikbetrieb, 245–251. Tübingen: Niemeyer. Primus, Beatrice. 2012. Semantische Rollen. Heidelberg: Winter. Ramelli, Christian. 2016. Über progressive und konservative Rheinfranken. In Speyer, Augustin & Philipp Rauth (eds.): Syntax aus Saarbrücker Sicht 1. Zeitschrift für Dialektologie und Linguistik Beiheft 165. Stuttgart: Steiner, 69-90. Reis, Marga. 1976. Reflexivierung in deutschen A.c.I.-Konstruktionen. Ein transformationsgrammatisches Dilemma. Papiere zur Linguistik 9. 5–82. Schrodt, Richard. 2004. Althochdeutsche Grammatik II. Syntax. Tübingen: Niemeyer. Searle, John R. 1979. A taxonomy of illocutionary acts. In John R. Searle: Expression and Meaning, 1–29. Cambridge: Cambridge University Press. Sihler, Andrew. 1995. New Comparative Grammar of Greek and Latin. Oxford: Oxford University Press. Smirnova, Elena. 2011. Zur diachronen Entwicklung deutscher Komplementsatz-Konstruktionen. In Alexander Lasch & Alexander Ziem (eds.), Konstruktionsgrammatik III. Aktuelle Fragen und Lösungsansätze, 77–93. Tübingen: Stauffenburg.

192

Augustin Speyer

Speyer, Augustin. 2001. Ursprung und Ausbreitung der AcI-Konstruktion im Deutschen. Sprachwissenschaft 26. 145–187. Speyer, Augustin. 2015. AcI and control infinitives: How different are they? A diachronic approach. Journal of Historical Linguistics 5. 41–71. Sternefeld, Wolfgang. 2006. Syntax. Eine morphologisch motivierte generative Beschreibung des Deutschen. Tübingen: Stauffenburg. van Pottelberge, Jeroen. 2004. Der am-Progressiv. Struktur und parallele Entwicklung in den kontinentalwestgermanischen Sprachen. Tübingen: Narr. Vendler, Zeno. 1957. Verbs and times. The Philosophical Review 66. 143–160. von Stechow, Arnim. 1990. Status government and coherence in German. In Günther Grewendorf & Wolfgang Sternefeld (eds.), Scrambling and Barriers, 143–198. Amsterdam & Philadelphia: John Benjamins. Wechsler, Stephen. 1979. Resultative predicates and control. Texas Linguistic Forum 38. 307– 321.

Éva Dékány

8 Anti-agreeing infinitives in Old Hungarian 1 Introduction Infinitives in Modern Hungarian are well-known for having a full f-feature paradigm, that is, an inflection that co-varies with the f-features of the infinitival subject. Representative examples are given in (1).1 (1) a. János-nak ki kell takaríta-ni-a a szobá-t. John-DAT PRT have.to clean-INF-3SG the room-ACC ‘John has to clean the room.’ b. (Nekem) ki kell takaríta-n-om a szobá-t. I.DAT PRT have.to clean-INF-1SG the room-ACC ‘I have to clean the room.’ Inflected infinitives are possible only if the matrix predicate has a Dative noun phrase argument.2 In (2), for instance, the matrix verb lát ‘see’ has no Dative argument, and the f-feature inflection on the infinitive is ruled out. (2) (Én) lát-t-am János-t olvas-ni-(*a). I see-PST-1SG John-ACC read-INF-3SG ‘I saw John read.’ Acknowledgement: I thank Marcel den Dikken for useful discussion and the two anonymous reviewers for their constructive comments, which helped me to improve the manuscript. This material is based upon work supported by the Hungarian Scientific Research Fund under grant OTKA 112057 (Hungarian Generative Diachronic Syntax 2) and a postdoctoral grant of the Hungarian Academy of Sciences. 1 The paper contains the following abbreviations: 1SUBJ.2OBJ: first person subject agreemet fused with second person object agreement, ABL: ablative case, ACC: accusative, ATTR: attributivizer, C.: Codex, CAUS: causative, COND: conditional, DAT: dative, DESID: desiderative, INF: infinitive, PART: participle, PASS: passive, PL: plural, POSS: possessive, PRT: verbal particle, PST: past, Q: question particle, SS: singular subject, SBJ: subjunctive, SG: singular. 2 Pronominal Dative arguments can undergo pro-drop, as in (1b); their referential content can be recovered from the infinitival agreement. Éva Dékány, Research Institute for Linguistics, Hungarian Academy of Sciences DOI 10.1515/9783110520583-008

194

Éva Dékány

Whether there is a correlation between the syntactic position of the Dative noun phrase and the presence of inflection is a debated issue. According to É. Kiss (2002: 216), the judgments show a clean cut: Inflection on the infinitive is i) obligatory if the Dative noun phrase is thematically part of the infinitive, ii) optional if it may belong either to the infinitive or the matrix predicate, and iii) impossible if it unambiguously belongs to the matrix clause. Tóth (2002: 148), on the other hand, reports that the judgments show considerable inter-speaker variation, and there is no clear correlation between the syntactic position of the Dative noun phrase and the presence or lack of inflection on infinitives. Inflected infinitives were already in place in the first surviving coherent Hungarian text, the 50-line Funeral Sermon and Prayer (1192–1195). (3) ki-nec odut hotolm ovdo-ni-a eʃ ket-ni-e who(sg)-DAT given power lose-INF-3SG and bind-INF-3SG ‘who was given power to bind and lose’ (Funeral Sermon and Prayer) The full infinitival paradigm is given in (4).3 (4) a. men-n-em go-INF-1SG (for me) to go (Soproni Virágének)

d. men-n-ẃnk go-INF-1PL (for us) to go (Könyvecse 14r)

b. men-n-ed go-INF-2SG (for you) to go (Horvát C. 122v)

e. men-n-etek go-INF-2PL (for you) to go (Jókai C. 82)

c. men-ni-e go-INF-3SG (for him) to go (Kazinczy C. 27r)

f. mē-ni-ec go-INF-3PL (for them) to go (Vienna C. 301)

However, in Old Hungarian (henceforth OH, 826–1526 A.D.) inflected infinitives have a different distribution than in Modern Hungarian: Inflection is possible on all infinitives, irrespective of whether or not the matrix predicate has a Dative noun phrase argument (see Tóth 2000, 2002, 2011 and the examples in Section 2). Furthermore, OH infinitives can optionally feature a 3SG ending irrespective of the person and number features of the infinitive’s subject. Observe (5), where 3 The written form of the infinitival suffix in Modern Hungarian is -ni. In Old Hungarian, however, orthography is not standardized, and this suffix is written in three different forms: -ni, -ny, and -nÿ. In both Modern and Old Hungarian, the i/y/ÿ vowel is dropped from the suffix if the infinitive bears a first or second person inflection.

Anti-agreeing infinitives in Old Hungarian

195

the expected inflection would be 1SG, and (6), where we would a expect 2PL ending. (5) nem ÿǫ-tt-em hÿ-nÿ-a az ÿgaz-ak-ath not come-PST-1SG call-INF-3SG the righteous-PL-ACC ‘I have not come to call the righteous’ (Könyvecse 29r) (6) akar-na-tok-ee ven-nÿ-e want-COND-2PL-Q take-INF-3SG ‘would you(pl) want to take it?’ (Jókai C. 22) Compare the regularly inflected forms of these infinitives with the same types of subjects and the same matrix verbs: (7) nem iǫ-tt-em hy-n-om igaz-ak-ot not come-PST-1SG call-INF-1SG righteous-PL-ACC ‘I have not come to call the righteous’ (Döbrentei C 205v) (8) ha gǫzǫdelm-eth akkar-tok raitok uen-n-ǫtǫk if victory-ACC want-2PL on.them take-INF-2PL ‘if you want (to take) victory over them’ (Kazinczy C. 66r) The phenomenon whereby a verb bears an invariant, default agreement regardless of the f-features of the subject is known as anti-agreement and is attested in a variety of languages, among them Berber, Bantu, and Celtic languages, as well as Imbabura Quechua and some Italian dialects (see Ouhalla 1993, 2005; Richards 2001; Elouazizi 2005; Schneider-Zioga 2007; Ouali 2008; Baker 2008a, b, among others). However, to the best of my best knowledge, anti-agreement in infinitives has not yet been documented in other languages. In this paper inflected infinitives whose inflection co-varies with the f-features of the subject/controller, e.g. (3) and (4), will be referred to as agreeing infinitives. I will term inflected infinitives like (5) and (6) as anti-agreeing infinitives. ‘Inflected infinitives’ will be a cover-term for agreeing and anti-agreeing infinitives. Infinitives that feature the infinitival suffix without a person-number ending, for instance (2), will be called uninflected infinitives.4 4 Hungarian does not feature an infinitival complementizer at any point in its recorded history. In this sense Old Hungarian uninflected, agreeing, and anti-agreeing infinitives are all ‘bare infinitives’.

196

Éva Dékány

The purpose of this paper is to examine the distribution of OH anti-agreeing infinitives and speculate on how this anti-agreement arises. The paper is structured as follows. In Section 2 I examine the distribution of uninflected, agreeing, and anti-agreeing infinitives in OH. In Section 3 I review and argue against the analysis of anti-agreement put forth in descriptive historical grammars. In Section 4 I show that anti-agreement has a small fossil in Modern Hungarian, but it is a different phenomenon from anti-agreement in OH. Section 5 argues that anti-agreement is default agreement in OH, while Section 6 speculates on why default agreement arises on OH infinitives. Finally Section 7 concludes my discussion.

2 The distribution of uninflected, agreeing, and anti-agreeing infinitives As discussed in the introduction, OH features uninflected, agreeing, and antiagreeing infinitives. Tables 1 through 7 summarize the numbers of the three types of infinitives in OH from seven codices. The Jókai Codex is the first surviving Hungarian codex. The text was translated from Latin around 1370 but the surviving codex is a copy of this text from 1446. The Vienna Codex is from the middle of the 15th century. The infinitives in these codices were counted by Károly (1956).5 Table 1: Infinitives in the Jókai Codex (token number: 23194)

1SG 1PL 2SG 2PL 3SG 3PL

uninflected

agreeing

anti-agreeing

23 2 13 4 211 52

24 5 23 6 37 4

4 Ø 3 2 N/A 13

305

99

22

5 The token numbers were taken from http://omagyarkorpusz.nytud.hu/en-search.html, an online Old Hungarian corpus (see Simon & Sass 2012), and indicate the word count in the OH text without puctuation marks.

Anti-agreeing infinitives in Old Hungarian

197

Table 2: Infinitives in the Vienna Codex (token number: 55294) uninflected 1SG 1PL 2SG 2PL 3SG 3PL

agreeing

Ø Ø Ø 1 200 61

29 9 34 21 3 53

262

149

anti-agreeing Ø Ø Ø Ø N/A 1 1

The Guary Codex is from before 1508, while the Könyvecse was copied in 1521. These codices are normalized and morphologically analyzed and annotated in the online OH corpus, and their infinitives were searched with the corpus query tool. Among infinitives with a 3SG ending, anti-agreeing ones were singled out by manual checking. Table 3: Infinitives in the Guary Codex (token number: 21714) uninflected 1SG 1PL 2SG 2PL 3SG 3PL

agreeing

Ø 1 11 Ø 59 9

7 14 4 Ø 79 8

80

112

anti-agreeing 2 1 4 Ø N/A 8 15

Table 4: Infinitives in the Könyvecse (token number: 9757) uninflected 1SG 1PL 2SG 2PL 3SG 3PL

agreeing

1 1 3 Ø 44 18

4 3 3 Ø 3 5

67

18

anti-agreeing 1 Ø Ø Ø N/A 1 2

The Székelyudvarhely Codex is from between 1526 and 1528, the Kazinczy Codex is from between 1526 and 1541, and the Bod Codex is from the early 16th century. These codices are normalized but not morphologically analyzed or annotated in

198

Éva Dékány

the online OH corpus. Their infinitives were retrieved by searching the uninflected infinitival ending -ni and the inflected infinitival endings for all persons and numbers. Hits that yielded words other than infinitives were filtered out manually. Anti-agreeing infinitives were identified from among infinitives with a 3SG ending manually, and the f-features of the subjects of uninflected infinitives were also identified manually. Table 5: Infinitives in the Székelyudvarhely Codex (token number: 38048)

1SG 1PL 2SG 2PL 3SG 3PL

uninflected

agreeing

anti-agreeing

Ø Ø 1 Ø 8 10

2 Ø 4 Ø 4 7

Ø Ø Ø Ø N/A 3

19

17

3

Table 6: Infinitives in the Kazinczy Codex (token number: 20027)

1SG 1PL 2SG 2PL 3SG 3PL

uninflected

agreeing

anti-agreeing

Ø Ø Ø Ø 42 9

22 6 31 7 121 4

Ø Ø Ø Ø N/A 27

51

191

27

Table 7: Infinitives in the Bod Codex (token number: 10084) uninflected 1SG 1PL 2SG 2PL 3SG 3PL

agreeing

anti-agreeing

Ø Ø Ø Ø 2 1

7 4 13 Ø 99 1

3 2 8 1 N/A 6

3

124

20

Anti-agreeing infinitives in Old Hungarian

199

While anti-agreeing infinitives have not been described in other languages, agreeing infinitives are known to exist in European and Brazilian Portuguese (Raposo 1987; Pires 2001, 2006; Modesto 2007), Sardinian (Jones 1993; Miller 2003), Galician (Longa 1994), Old Neapolitan (Miller 2003; Scida 2004) as well as two Spanish dialects: Old Leonese and Mirandese (Scida 2004);6 see also Lühr (this volume). As pointed out in Miller (2004: 330), in these languages the uninflected infinitive “correlates with PRO subjects in any case”, while the inflected infinitive correlates with “lexical and pro subjects in the nominative”. This correlation does not hold in Old Hungarian, however: Both infinitives with a PRO subject and a lexical/pro subject feature uninflected as well as agreeing infinitives; the two types of infinitives appear to be in free variation (Tóth 2002, 2011). OH infinitives may serve as a subject, as an object, or as an adjunct in the clause. The subject position of infinitives may be occupied by a referentially independent subject (only with monadic matrix predicates), or a trace (ECM infinitives and infinitival complements to raising verbs), or PRO. A referentially independent subject is possible only if the matrix predicate is monadic, i.e. its only argument is the infinitival clause. Such predicates are epistemic modals (e.g. kell ‘must’ in one of its uses), non-directed deontic modals (e.g. kell ‘must’ in another use), and nominal predicates without an Ablative DP (e.g. szemtelenség ‘impertinence’ without such a DP). As these matrix predicates do not have a DP argument, the infinitive cannot have a PRO subject because there is no potential controller for it. Referentially independent subjects bear Dative case. The ECM matrix verbs of OH are permissive verbs. Permissive verbs in OH can take either an Accusative or a Dative permissee (for instance hagy valaki-t ‘let somebody-ACC’ or hagy valaki-nek ‘let somebody-DAT’). ECM structures arise when the permissee is Accusative. The control structure is obligatory if the matrix predicate takes a DP argument that may potentially control the infinitive’s subject (Tóth 2000, 2002, 2011; É. Kiss 2002). The relevant predicates are: subject oriented deontic modals (e.g. kell ‘must’ in a third type of use), evaluative predicates (for instance fontos ‘important’, jólesik ‘feels good’), permissive verbs with a Dative permissee (e.g. hagy valaki-nek ‘let/command somebody-DAT’), and nominal predicates with an Ablative DP (e.g. szemtelenség valaki-től ‘impertinence somebody-ABL’).

6 In the above-mentioned languages, as in (Old) Hungarian, it is the infinitival verb that bears agreement. In Welsh, on the other hand, it is the infinitive marker i that may take an agreement suffix, see Miller (2004).

200

Éva Dékány

2.1 Control infinitives As pointed out above, if the matrix predicate has a noun phrase argument, then the infinitive has to have a PRO subject and a control structure ensues. The controller may play the role of the subject, object, or Dative experiencer in the matrix predication, or it can be an Ablative noun phrase in the matrix clause. The possible combinations (based on Károly 1956 and Tóth 2000, 2002, 2011 but slightly differing from their classification) are shown in Table 8. Table 8: Control in OH infinitives the controller

the infinitive in the clause

subject object Dative experiencer Ablative

subject N/A – Z Z

object Z – Z Z

adjunct Z Z – –

Uninflected infinitives are attested with all types of infinitives except for Ablative control into object infinitives. I take this to be an accidental gap in the data.7 (9) Ne akar-y-atok ty ffel-ny not want-SBJ-2PL you.PL afraid-INF ‘do not be afraid’8 (Jordánszky C. 450)

subject control into an object inf.

(10) Es iǫ-nèc èzec meg-yèzt-èni ǫk-èt and come-3PL these PRT-terrify-INF they-ACC ‘and they are coming to terrify them’ (Vienna C. 296) subject control into an adjunct inf. (11) hal-lak teged-ett zol-nÿ hear-1SUBJ.2OBJ you-ACC speak-INF ‘I can hear you speak’ (Jókai C. 45)

object control

7 Uninflected and agreeing infinitives are in free variation in OH subject, object, and Dative control structures, and the null hypothesis is that this is also the case in Ablative control structures. Ablative control infinitives are very rare in OH. The expansion of the normalized part of the online OH corpus holds out the possibility that more examples of Ablative control will be found later on, and some of these may be object infinitives. 8 The literal translation of this sentence is do not want to be afraid. Here the Old Hungarian text follows the Latin structure nolite timere (not.want.SBJ.2PL afraid.INF) ‘do not want to be afraid’. I thank Barbara Egedi for clarification on this point.

Anti-agreeing infinitives in Old Hungarian

201

(12) az isten-nek any-a-nak alkolmas volt meg hal-ny was PRT die-INF the God-DAT mother-POSS-DAT timely ‘it was timely for God’s mother to die’ (Horvát C. 70v) dative control into a subject inf. (13) hagy-aa az kǫ́ lrw̋ l all-ok-nak a hw̋ command-PST.3SG the around stander-PL-DAT the his zay-a-t arczw̋ l ver-ny mouth-POSS-ACC in.the.face strike-INF ‘he commanded those standing beside him to stike him on the mouth’ (Jordánszky C. 783) dative control into an object inf. (14) zent fferencz-tewl vala zerez-tet-ett ez-t mond-anÿ saint Francis-ABL be.PST procure-PASS-PART this-ACC say-INF ‘this was said by Saint Francis’ (Jókai C. 42) ablative control into a subject inf. Agreeing infinitives, that is, infinitives showing full agreement for the f-features of the subject, are found in all subtypes of control infinitives except for Ablative control into an subject infinitive. I take this to be an accidental gap in the data. (15) ne akar-i-atok fel-n-etèc not want-SBJ-2PL fear-INF-2PL ‘do not be afraid’ (Munich C. 42ra)

subject control into an object inf.

(16) Mert nem iǫ-tt-em hy-n-om igaz-ak-ot because not come-PST-1SG call-INF-1SG righteous-PL-ACC ‘I have not come to call the righteous’ (Döbrentei C. 205v) subject control into an adjunct inf. (17) lat-lac teged-et ekepen all-an-od see-1SUBJ.2OBJ you-ACC this.way stand-INF-2SG ‘I can see you stand this way’

(Miskolc Fragment 2v)

object control

(18) le-gyen alkolmas en-nek-em zol-n-om ty-nek-tek be-SBJ.3SG suitable I-DAT-1SG talk-INF-1SG you-DAT-2PL ‘let me speak freely to you’ (lit: let it be suitable for me to speak to you) (Jordánszky C. 712) dative control into a subject inf.

202

Éva Dékány

(19) hagy-ad en nek-em be tellyeʃeyt-en-em az-t. that-ACC let-SBJ.2SG I DAT-1SG PRT fulfill-INF-1SG ammy-re iev-tt-em that-for come-PST-1SG ‘let me fulfill what I have come for’

(Cornides C. 113v) dative control into an object inf.

(20) erdǫmli isten-tul ǵacorta meg bočat-ni-a PRT forgive-INF-3SG deserve God-ABL often ‘often deserves that God forgive (his sins)’ (Guary C. 104) ablative control into an object inf. Uninflected and agreeing infinitives are in free variation (Tóth 2002, 2011). When the same Latin text has multiple translations in OH, one codex may use an uninflected infinitive and another an agreeing infinitive in the translation of the same sentence, compare (9) with (15) and (21a) with (21b). (21) a. ʃokan kereʃ-nek bel me-n ¯y many seek-3PL in go-INF ‘many seek to enter’

(Jordánszky C. 576)

b. ʃokac bè-men-ni-èc many seek-3PL in-go-INF-3PL kèrèʃ-n c

‘many seek to enter’

(Munich C. 72 rb)

The optionality of the agreement can also be observed within the same codex. (22) s-ew-nÿ nem akar-t uala and-come-INF not want-PST.3SG be.PST ‘and he did not want to come’

(Jókai C. 10)

(23) Elÿas nem akar en haz-am ÿew-nÿ-e Elijah not want I to-1SG come-INF-3SG ‘Elijah does not want to come to me’ (Jókai C. 6) Uninflected and agreeing infinitives can also be coordinated. (24) tud-yak mÿ-tt kel tarta-nÿ-ok Es my-t know-3PL what-ACC must keep-INF-3PL and what-ACC el-tauoz-tatt-nÿ away-leave-CAUS-INF ‘they know what they need to keep and what they need to keep away (from themselves)’ (Jókai C. 119)

Anti-agreeing infinitives in Old Hungarian

203

Finally, anti-agreeing infinitives are found in all types of control infinitives except for Ablative control into subject infinitives. The only anti-agreeing Ablative control example that I am aware of is found with Ablative control into an object infinitive. (25) Ne akar-y-atok feel-ny-e not want-SBJ-2PL fear-INF-3SG ‘do not be afraid’ (Jordánszky C. 55)

subject control into an object inf.

(26) nem ÿǫ-tt-em hÿ-nÿ-a az ÿgaz-ak-ath not come-PST-1SG call-INF-3SG the righteous-PL-ACC ‘I have not come to call the righteous’ (Könyvecse 29v) subject control into an adjunct inf. (27) lat-lak vala fi-a-m giokarta suyr-ni-a cry-INF-3SG see-1SUBJ.2OBJ be.PST son-POSS-1SG often ‘I see you, my son, cry often’ (Weszprém C. 17r) object control into an adjunct inf. (28) ha alkomas en-nÿ-e med aʒok-balal ewangelıumm-ÿ tart-ok-na hold-PL-DAT if suitable eat-INF-3SG all that-from gospel-ATTR ‘if it is expedient for the followers of the gospels to eat of all those’ (Jókai C. 16) dative control into a subject inf. (29) en I

magam-at self-ACC

hat-t-am en zeretǫ-i-m-nec zertelen let-PST-1SG my lover-POSS.PL-1SG-DAT excessively

fogdoʃ-ni-a, es paddle.on-INF-3SG and

zorongat-ni-a press.hard-INF-3SG

‘I let my lovers to paddle on me and press hard on me excessively’ (Guary C. 94) dative control into a complement inf. (30) ʃohha ne laʃ-ʃ-onc te tųl-ed fia-t zųl-ni-e never not see-SBJ-1PL you ABL-2SG son-ACC give.birth-INF-3SG ‘let us never see you give birth to a son’

(Guary C. 103)

ablative control

Anti-agreeing infinitives may be coordinated with agreeing infinitives.9

9 I do not have any examples in which an anti-agreeing infinitive is coordinated with an uninflected infinitive. I take this to be an accidental gap in the data. Since uninflected infinitives may be coordinated with agreeing infinitives, and the latter can be coordinated with anti-agreeing infinitives, it stands to reason to assume that coordinating uninflected and anti-agreeing infinitives was also a possiblitiy.

204

Éva Dékány

(31) kÿ-k ÿgér-ÿk on magok-at megh zeplǫsÿt-enÿ-ek az violate-INF-3PL the who-PL promise-3PL them self-ACC PRT teh zenthsegh-i-d-eth, Es megh fertezet-nÿ-e az violate-INF-3SG the your sacrament-POSS.PL-2SG-ACC and PRT teh zent new-ed-nek haÿléék-a-t your holy name-2SG-DAT house-POSS-ACC ‘who promise themselves to violate thy sanctuary, and defile the dwelling place of thy name’ (Székelyudvarhely C. 30r) Anti-agreement is independent of the linear order of the matrix predicate and the infinitive (32) and the order of the controller and the infinitive (33). (32) a. mely retenetes lezen te-nek-ed a criʃtuʃ-th lat-ny-a how terrible will.be you.sg-DAT-2SG the Christ-ACC see-INF-3SG az itilet-ben the judgment-in ‘how terrible it will be for you to see Christ at the (Last) Judgment’ (Teleki C. 253) b. es nez-ny-e vtalatos ew zem-ek-uel and see-INF-3SG loathsome their eye-3PL-with ‘and it is loathsome for them to see it with their eyes’ (Jókai C. 125) (33) a. nem az mv wezedelmvnk-éérth hy̋ ǵ-ǵv́ k [tǫrtéént-nek not for our destruction-for believe-SBJ.1PL transpired-DAT len-ny̋ -e] ezek-et ḿv ray̋ tvk be-INF-3SG these-ACC us on ‘let us believe that these (scourges of the Lord) have happened not for our destruction’ (Székelyudvarhely C. 26v) b. Jnt-lek teghed-et [keesen zolo-nak len-nÿ-e] warn-1SUBJ.2OBJ you.SG-ACC slow speaking-DAT be-INF-3SG ‘I warn you to be slow to speak’ (Winkler C. 65r) Anti-agreeing infinitives are found in almost all linguistic records (A. Jászó 1992).10 10 The first Hungarian Bible translation, the so-called Hussite Bible, was prepared in the first half of the 15th century. While the original work is lost, the Munich Codex, the Vienna Codex, and the first part of the Apor Codex are copies of this translation. The Munich Codex is one of

Anti-agreeing infinitives in Old Hungarian

205

2.2 ECM infinitives ECM matrix verbs are found with uninflected, agreeing, and anti-agreeing infinitives as well. (34) nem haggÿah az lelk-eth az sǫtethsegek-reh men-nÿ go-INF not let.3SG the soul-ACC the darkness-to ‘he does not let the soul go into darkness’ (Könyvecse 9v) (35) hagÿ engemet egÿeb kewzewnet-ÿt mond-an-om let.2.SBJ me different thanks-ACC say-INF-1SG ‘let me say a different thanks’ (Jókai C. 90) (36) ne(m) engǫmet haǵ-ot vr v˛ zolgalo leaṅ-a-t not me let-PST.3SG Lord his servant maid-POSS-ACC meg PRT

fǫrtǫz-tet-ni-e blemish-PASS-INF-3SG

‘the Lord did not let me, his servant, get blemished’ (Guary C. 114)

2.3 Raising infinitives Raising matrix verbs are also found with uninflected, agreeing, and anti-agreeing infinitives. (37) kèzd-è ǵondol-ni begin-PST.3SG consider-INF ‘began to consider’ (Vienna C. 91) (38) kēzd-ē-nc ʃir-ń-oc begin-PST-3PL weep-INF-3PL ‘they began to weep’ (Vienna C. 2)

the OH codices that contains no anti-agreeing infinitives, and the Vienna Codex has only a single anti-agreeing infinitive (Károly 1956). (The Apor Codex in the OH corpus is neither morphologically analyzed nor normalized; whether this codex contains anti-agreeing infinitives, and if so, how many, is not yet known). As both earlier codices (eg. the Jókai Codex) and later texts (for instance the Székelyudvarhely Codex) feature anti-agreeing infinitives, the almost complete lack of such infinitives in the Munich Codex and the Vienna Codex could be a dialectal feature.

206

Éva Dékány

(39) ezd-e-k en-es erǫʃʃen syr-ny-a start-PST-1SG I-too very.much cry-INF-3SG ‘I, too, started to cry very much’ (Teleki C. 299)

2.4 Infinitives with a non-coreferent Dative subject Infinitives with a Dative-marked lexical subject are selected by epistemic modals, non-directed deontic modals and nominal predicates. Such matrix predicates, however, are much more scarce than those that select for a control, raising, or an ECM infinitive. Among matrix predicates selecting for an infinitive, verbal predicates are by far more common than nominal predicates. Furthermore, modals in these texts are typically deontic and directed.11 Infinitives with a Dative-marked lexical subject may be uninflected (40) or agreeing (41). In (40) kel ‘must’ is a non-directed deontic modal. The Dative DP is inanimate, so it cannot be the recipient of obligation; the obligation is localized in some other individual. Consequently, this DP cannot be interpreted as a Dative experiencer in the matrix clause; it is the overt subject of the infinitive instead. (41) features a nominal predicate. (40) kel-uala [ew ʒerʒet-e-nek nagÿ ʃokaʃʃag-ban terÿed-nÿ] have.to-be.PST his order-POSS-DAT great multitide-in spread-INF ‘his order had to spread among great multitudes of people’ (Jókai C. 13)

11 (ib) below is a rare example of an epistemic modal (the context that provides the epistemic reading, the sentence immediately preceding (ib) in the codex, is given in (ia)). (i) a. az ireǵseg ez-t mond-’a, Sokak-at te-het-z kik-et amaz the jelaousy this-ACC say-PST.3SG many-ACC do-POSS-2SG that-ACC that.other.one nem te-het-i not do-POSS-3SG ‘Jealousy said this: You can do many things that that other one cannot do’ b. Azert nekm kell len-nÿ-e nek-i fell’ebvalo-nak naladnal than.you therefore nekm must be-INF-3SG DAT-3SG senior-DAT ‘therefore he must be senior to you’ (Bod C. 7v) The word glossed as nekm appears to be a mistake that made its way into the text. It resembles nekem, the Dative form of the first person singular pronoun. That, however, would make two Datives in (ib), making the sentence ungrammatical. The context in (ia) makes it clear that the subject of the epistemic kell ‘must’ is the third person singular Dative neki, which is coreferent with the amaz ‘that other one’ demonstrative of the previous clause.

Anti-agreeing infinitives in Old Hungarian

207

(41) Hewsag [nek-thek wylaagh elewth fel kel-n-ethek] world in.front.of up get-INF-2PL vanity DAT-2PL ‘it is vanity for you to stand up in front of the world’ (Festetics C. 85) In my sample, which contains 107 anti-agreeing examples (all the anti-agreeing infinitives of the seven codices in Tables 1 through 7, and gleaning examples from seventeen other codices and two legends12), there are no anti-agreeing infinitives with a monadic matrix predicate.

3 Previous treatments of OH infinitival anti-agreement Previous discussions of OH anti-agreeing infinitives are all couched in the framework of descriptive historical linguistics (Keresztes 1953: 341, Károly 1956: 67–70, A. Jászó 1992: 412, 422, a.o.). Observing that the -a/-e inflection on an infinitive co-occurs with non-3SG subjects, too, these works conclude that -a/-e cannot be a 3SG agreement suffix. They suggest that -a/-e originated as a 3SG agreement but lost its 3SG value, ceased to be an agreement suffix, and amalgamated with the infinitival marker -ni/-ny/-nÿ (see esp. Keresztes 1953: 341 and A. Jászó 1992: 422). This gave rise to a new, monomorphemic infinitival suffix -nia/-nya/nÿa/-nie/-nye/-nÿe, which was used as an alternative variant of the original infinitival suffix -ni/-ny/-nÿ. In other words, in this approach the infinitival form in (42) should be analyzed as in (43), and it is morphologically equivalent to the uninflected infinitive in (44). (42) kÿ-k ÿgér-ÿk on magok-at . . . megh fertezet-nÿ-e violate-INF-3SG who-PL promise-3PL them self-ACC . . . PRT az teh zent new-ed-nek haÿléék-a-t the your holy name-2SG-DAT house-POSS-ACC ‘who promise themselves to violate thy sanctuary’ (Székelyudvarhely C. 30r) (43) megh fertezet-nÿe PRT violate-INF ‘to violate’ 12 One or two examples from each source.

208

Éva Dékány

(44) k-ic iǵer-ic maǵok-at . . . mėg-ferteztèt-ni te who-PL promise-3PL self-ACC . . . PRT-violate-INF your nèu-èd-nc hailak-a-t name-2SG-DAT house-POSS-ACC ‘who promise themselves to violate thy sanctuary’ (Vienna C. 30) A. Jászó (1992: 422) speculates that the loss of the 3SG value of -a/-e was caused by the frequency of 3SG inflected infinitives. Among infinitives with a 3SG subject, uninflected infinitives outnumber agreeing infinitives by far. However, among all agreeing infinitives, those with a 3SG inflection are the most numerous. A. Jászló suggests that this may have played a role in the loss of the 3SG value and the amalgamation into -ni/-ny/-nÿ. The approach of descriptive historical grammars outlined above does not strike me as very promising, however. Firstly, the claim that -nia/-nya/nÿa/-nie/-nye/-nÿe became a morphological alternative of -ni/-ny/-nÿ is simply not backed up by the data. While the uninflected infinitival suffix -ni/-ny/-nÿ could be followed by an agreement inflection (see many examples in Section 2), this is not true for -nia/-nya/nÿa/-nie/-nye/-nÿe. That is, while we find minimal pairs like (45a) and (45b), examples like (46b) do not exist. (45) a. Ne akar-y-atok ty ffel-ny not want-SBJ-2PL youPL afraid-INF ‘do not be afraid’ (Jordánszky C. 450) b. ne akar-i-atok fel-n-etèc not want-SBJ-2PL fear-INF-2PL ‘do not be afraid’ (Munich C. 42ra) (46) a. Ne akar-y-atok feel-ny-e not want-SBJ-2PL fear-INF-3SG ‘do not be afraid’ (Jordánszky C. 55) b. *ne akar-y-atok fel-nye-tec not want-SBJ-2PL fear-INF-2PL ‘do not be afraid’ At best, this shows that -nia/-nya/nÿa/-n.ie/-nye/-nÿe was an indeclinable infinitival suffix as opposed to the declinable -ni/-ny/-nÿ. Note, however, that affixes genuinely reanalyzed as part of their original host do not block further transparent affixation of the same type. Consider the affixation possibilities of the

Anti-agreeing infinitives in Old Hungarian

209

nominal compound városháza ‘city hall’, for instance. Historically (and to some degree even synchronically), városháza was a possessive construction comprising the possessor város ‘city’ and the possessum ház ‘house’ bearing the possessive suffix -a/-e/-ja/-je. (47) város-ház-a city-house-POSS ‘city hall’ Synchronically, however, városháza ‘city hall’ is listed as a lexical unit; the -a in it has no morphemic status. Importantly, when városháza ‘city hall’ is a possessum, it obligatorily takes the transparent possessive suffix, even though historically it already has such an affix:13 (48) az ország legrégebb-i városházá-ja the country oldest-ATTR city.hall-POSS ‘the country’s oldest city hall’ The indeclinability of -nia/-nya/nÿa/-nie/-nye/-nÿe thus could not be derived from the putative fusion of the 3SG -a/e into the infinitival suffix -ni. It is more plausible that -nia/-nya/nÿa/-nie/-nye/-nÿe cannot be followed by an agreement suffix because it is not a single suffix but a sequence of two suffixes: The regular infinitival ending -ni/-ny/-nÿ and a genuine 3SG agreement suffix -a/e. Secondly, this analysis raises an issue about the status of the 3SG cell of the infinitival agreement paradigm. The whole paradigm is shown in (49) (with normalized orthography). (49) a. -(e/o/ö)m 1SG

d. -unk/ünk 1PL

b. -(e/o/ö)d 2SG

e. -(e)tek/-(o)tok/-(ö)tök 2PL

c. -a/e 3SG

f. -(u/ü)k 3PL

13 The lengthening of the last a to á is a regular morphophonological process in the language that need not concern us here.

210

Éva Dékány

If -a/-e was originally a 3SG inflection but it lost its status as an agreement marker, then the 3SG cell of the paradigm must have been replaced by a zero suffix.14 In Modern Hungarian, however, the 3SG infinitival agreement is actually -a/-e. (50) János-nak 8-ra a munkahely-é-n kell len-ni-e. John-DAT 8-by the workplace-POSS-at have.to be-INF-3SG ‘John has to be at work by 8.’ The previous analyses thus must assume that after the early Middle Hungarian period the alternative -nia/-nya/nÿa/-nie/-nye/-nÿe infinitival suffix was lost (as anti-agreement did not survive beyond this period) and at the same time the 3SG infinitival agreement changed from f back to the original -a/-e. This is a highly unappealing theory. It is much more plausible that the 3SG infinitival agreement has always been -a/-e, and OH -nia/-nya/nÿa/-nie/-nye/-nÿe can be decomposed into two suffixes (infinitival ending+3SG agreement).15

4 Infinitival anti-agreement in Modern Hungarian OH infinitival anti-agreement has a small fossil in Modern Hungarian, too. As already mentioned before, agreeing infinitives in Modern Hungarian occur only in the presence of a Dative DP, so anti-agreeing infinitives are also found only in this context. (51a) and (51b) are representative of the Modern Hungarian facts. In (51a) the Dative DP is the overt subject of the infinitive (it features an inanimate noun, therefore it cannot be analyzed as a Dative experiencer in the matrix clause controlling a PRO infinitival subject), while in (51b) the Dative phrase fulfills the role of the experiencer in the main clause (the boys are the recipient of the obligation) and it controls a PRO subject in the infinitive.

14 The alternative is that it was not replaced by a zero suffix, and 3SG subjects simply could not go together with inflected infinitives. This is unlikely, as all other subjects could co-occur with inflected infinitives. 15 Károly (1956: 68) raises the possibility that OH has two different -nia/-nya/nÿa/-nie/-nye/-nÿe endings. The one that co-occurs with a 3SG subject is decomposable into an infinitival marker and a 3SG agreement, while the one that appears in the anti-agreeing examples is an already fused, monomorphemic suffix. Thus for Károly, the infinitival 3SG inflection has remained -a/e throughout the history of Hungarian. The previous criticism, however, applies to this approach as well.

Anti-agreeing infinitives in Old Hungarian

211

(51) a. Jövő év vég-é-re itt ház-ak-nak kell next year end-POSS-by here house-PL-DAT have.to pül-ni-ük/épül-ni-e. built-INF-3PL/built-INF-3SG ‘By the end of next year houses have to be built here.’ (Tóth 2000: 152) b. A fiú-k-nak ki kell vin-ni-ük/?vin-ni-e the boy-PL-DAT out have.to take-INF-3PL/take-INF-3SG a szemet-et. the garbage-ACC ‘The boys have to take out the garbage.’ Such examples are not acceptable for everybody, and those speakers that do not rule them out as ungrammatical often find them degraded compared to full agreement. Modern Hungarian anti-agreement is subject to a strict constraint: It is possible only if the Dative noun phrase is headed by a lexical noun (see Tóth 2000, 2002; É. Kiss 2002; Rákosi & Laczkó 2008).16 Anti-agreement in the context of pronominal Dative noun phrases, whether the pronoun is first, second, or third person, is impossible. (52) and (53) minimally differ from (51a) and (51b) in that the former feature a 3PL Dative pronoun rather than a lexical noun. (52) Nekik át kell men-ni-ük/*men-ni-e a vizsgá-n, hiszen they.DAT PRT have.to go-INF-3PL/go-INF-3SG the exam-on for olyan sok-at készül-t-ek. so much-ACC study-PST-3PL ‘They must pass the exam, for they have studied so much.’ (53) Nekik ki kell vin-ni-ük/*vin-ni-e a szemet-et. they.DAT out have.to take-INF-3PL/take-INF-3SG the garbage-ACC ‘They have to take out the garbage.’ With the exception of É. Kiss (2002: ch. 9.4.3), data like (51) are mentioned in passing but not analyzed in the literature. É. Kiss argues that the Hungarian infinitival ending -ni is a nominalizing suffix, and so the infinitive is a nominal category. The Dative noun phrase in (51a) and (51b) occupies the position of the 16 Of course, for anti-agreement to arise, the lexical noun must be plural.

212

Éva Dékány

possessor. Since Hungarian lexical (but not pronominal) possessors are widely known to exhibit an anti-agreement effect (see Den Dikken 1999; Bartos 1999, 2000; É. Kiss 2002; Dékány 2011, 2015), É. Kiss suggests that what we see in (51a) and (51b) is none other than this well-known phenomenon. The argument goes as follows. Possessors in Hungarian may be either caseless or they may bear Dative case. Dative possessors may be extracted from the DP and function as external possessors. Possessa show agreement for the f-features of pronominal possessors, while they do not agree with lexical possessors. There appears to be one exception to this generalization. With an external (hence Dative) possessor headed by a plural lexical noun, the possessor may appear in two forms: without agreement, as in (54a), or with 3PL agreement (54b).17 (54) a. A fiú-k-nak eláz-ott a kalap-ja. the boy-PL-DAT get.wet-PST3SG the hat-POSS ‘The boys’ hat got wet.’ b. A fiú-k-nak eláz-ott a kalap-j-uk. the boy-PL-DAT get.wet-PST.3SG the hat-POSS-3PL ‘The boys’ hat got wet.’ Given that lexical possessors do not trigger agreement, (54a) is expected but (54b) is to be explained. Den Dikken (1999) argues that (54a) involves possessor extraction, while in (54b) the external possessor is generated in its surface position and is co-indexed with a DP-internal (plural) pro possessor. Possessa always agree with their pronominal possessor, so the possessum in (54b) also agrees with the pro possessor, producing the regular plural inflection. This analysis allows one to maintain the generalization that possessa agree only with pronominal possessors. Den Dikken’s analysis derives the unexpected agreement with lexical possessors from a structure in which the lexical possessor is co-indexed with a pronoun. Building on this analysis, É. Kiss suggests that the unexpected 3SG agreement in (51a) and (51b) is also due to such a co-indexation structure. Recall that É. Kiss treats infinitives as nominal projections and the Dative subject is analyzed as a Dative possessor. She suggests that in the anti-agreeing cases in (51), the Dative DP is generated as a hanging topic (i.e. as an external possessor), and it is coindexed with a lower pro element. The agreement is triggered by this pro 17 This is an option only if the possessor is external. Lexical possessors that are internal to the possessum’s DP projection cannot trigger agreement.

Anti-agreeing infinitives in Old Hungarian

213

(which is presumably a 3SG pronoun, as it produces 3SG agreement). In essence, in this analysis there is no genuine infinitival anti-agreement: The relevant data are shown to instantiate possessive structures and exhibit the independently attested and motivated possessive anti-agreement. It is not my intention to argue for or against this analysis of the Modern Hungarian facts. What I would like to point out, however, is that this analysis cannot extend to the OH anti-agreement data (which, to be fair, it was never intended to cover). This analysis crucially rests on the fact that in Modern Hungarian anti-agreement is attested only with Dative noun phrases headed by a lexical 3PL noun in both possessive structures and infinitives. In OH, as we have seen, infinitival anti-agreement is not restricted to clauses that contain a Dative noun phrase. Subject, object, and Ablative control as well as ECM and raising structures also admit anti-agreeing infinitives. Furthermore, OH infinitival antiagreement is attested with pronominal subjects, but OH has no possessive antiagreement with pronominal possessors (possessa must show full agreement for pronominal possessors). Infinitival anti-agreement thus must be a different phenomenon in Modern and Old Hungarian: The former may plausibly be assimilated to possessive anti-agreement, while the latter cannot.18

5 3SG as default agreement Descriptive grammars are unable to analyze -a/e as the 3SG inflection with non3SG subjects because they are unaware of the existence of anti-agreement effects in other languages. In certain languages and under specific circumstances, 18 At first sight, the Slavonia and the Szigetköz dialects of Hungarian appear to have used OHstyle anti-agreeing infinitives at least up to the early 20th century. On closer inspection, however, it turns out that the data from these dialects are unlikely to involve anti-agreement. In the Slavonia dialect (spoken in Eastern Croatia) infinitives that may be suspected of featuring anti-agreement involve the -a ending. While in the standard dialect (the 3SG) -a combines with the infinitival suffix without further ado (i), the Slavonia dialect employs a hiatus-filling j (ii). (i) lát-ni-a see-INF-3SG for him to see (standard dialect)

(ii) lât-ni-jȧ see-INF-ja for him to see (Slavonia dialect) (Balassa 1894: 263)

Infinitives may be inflected with -nijȧ even when the subject is non-3SG (see Szarvas 1876: 62, Simonyi 1892: 290, Balassa 1894: 263, and Keresztes 1953: 341): (iii) a. Ȧzok tun-ȧk dȧnol-ni-jȧ. those could-3PL sing-INF-ja They could sing (it). (Balassa 1894: 263)

214

Éva Dékány

regular f-feature agreement on the verb is altered, and the verb bears an invariable inflection. This invariable inflection is either a special affix or the 3SG agreement marker across the board (i.e. with non-3SG subjects, too). Verbs in Kinande, for instance, cannot feature regular agreement with wh-subjects. Such subjects trigger the special agreement u- instead of the regular a-. (55) a. Kambale a-langIra Marya. AGR-saw Mary K. Kambale saw Mary.

(Schneider-Zioga 2000: ex. 1a)

b. *IyOndI yO a-langIra Marya? who that AGR-saw Mary (Schneider-Zioga 2000: ex. 1c) c. IyOndI yO u-langIra Marya? who that WH.AGR-saw Mary ‘Who saw Mary?’ (Schneider-Zioga 2000: ex. 1d)

As pointed out by Balassa (1894), however, the -nijȧ ending appears even in those cases in which vowel harmony would require the -nije allomorph of the 3SG person inflection (iv). He concludes that the a in (iii) and (iv) is not a person inflection, and claims that this a is the distal demonstrative, which fused into the infinitival ending. While he does not present any evidence for the latter conjecture, the lack of vowel harmony in (iv) convincingly shows that we are not dealing with a person inflection here. (iv) a. Gyere be ë-ni-jȧ. come.SBJ.2SG in eat-INF-ja Come in to eat. (Balassa 1894: 263) b. Mȧgunk szokâ-jok ȧz-tȧt të-ni-jȧ. ourselves habitually.do-1PL that-ACC do-INF-ja We (habitually) do that ourselves. (Balassa 1894: 263) In the Szigetköz dialect (spoken on the Szigetköz island in Northwestern Hungary) the n of the infinitival suffix is palatalized, so the suffix is used in the -nyi form. There are also infinitives ending in -nya, as in (v). The -a of these forms might be taken to be a 3SG ending. When infinitives like (v) have a non-3SG subject, we appear to have an anti-agreeing infinitive. (v) innya, ir-nya drink.inf write-inf to drink, to write (Szabó 1907: 22) However, the -a of (v) cannot be plausibly analyzed as a 3SG ending, as its appearance is conditioned by the verb. Specifically, -nya is possible only with a subset of monosyllabic verbs whose stem vowel is i/í (Szabó 1907: 22). We can thus conclude that the -nya forms with non-3SG subjects are uninflected infinitives with a special infinitival allomorph rather than anti-agreeing infinitives.

Anti-agreeing infinitives in Old Hungarian

215

Imbabura Quechua verbs can show regular agreement only with Nominative subjects. Experiencer subjects bear Accusative case, however, so they cannot induce regular agreement on the verb. In this case the verb bears 3SG agreement regardless of the f-features of the subject. (56) Juzi-ta puñu-naya-n. José-ACC sleep-DESID-3S ‘José wants to sleep; José is sleepy.’ (Baker 2008a: 241) (57) ñuka-ta puñu-naya-n (*púnu-naya-ni). sleep-DESID-3S sleep-DESID-1sS I-ACC ‘I want to sleep; I am sleepy.’ (Baker 2008a: 243) Once the existence of such phenomena, and especially the type of agreement in (57), is taken into consideration, there is no impediment to analyzing OH -nia/-nya/nÿa/-nie/-nye/-nÿe as two morphemes, the regular infinitival suffix -ni/-ny/-nÿ and the regular 3SG inflection -a/-e, with non-3SG subjects, too. I submit that this is indeed the correct analysis of OH infinitival agreement mismatches: The relevant examples feature a genuine 3SG inflection. This approach avoids both problems raised by the descriptivist alternative discussed above. It gives a natural account of the fact that -nia/-nya/nÿa/-nie/-nye/-nÿe is never followed by an inflection (it is already inflected), and does not require the implausible reanalysis of the 3SG ending from -a/e to f and back to -a/e. This gives the present analysis a significant advantage over the alternatives. I suggest that the 3SG inflection with non-3SG subjects in OH is a default ending. Among number values, singular is less marked than plural. Rooryck (2003); López (2008); Farkas & de Swart (2010) and Nevins (2011), among others, argue that plural is a privative syntactic feature, while singular is simply lack of number. This means that NumP is projected only in non-singular noun phrases. Among person features, third person is less marked than either first or second person. Ionin & Matushansky (2002), for example, argue that third person corresponds to the feature [+person], second person corresponds to the feature matrix [+person, +participant], and first person has the feature matrix [+person, +participant, +speaker]. In their analysis, the features of 3rd person are thus a proper subset of the features of both second and first person. A different line of research, in particular Benveniste (1971); Sigurðsson (2000); López (2008) and Harley & Ritter (2002a,b), argues that person features are associated only to second and first person, and third person is actually lack of person.19 Whether 19 See Nevins (2007, 2011) for arguments that third person has a feature specification, and Den Dikkken (2013) for a rebuttal.

216

Éva Dékány

third person is a person feature or not, everybody is agreed that third person is less marked than second or first person. 3SG is thus the least marked feature combination possible, and so it is the most suitable to surface as a default inflection (see also López 2008).

6 Speculations on the distribution of default agreement As we have seen in Section 2, all the anti-agreeing infinitives in my database instantiate control, raising, or ECM structures, and there are no anti-agreeing infinitives with a referentially independent (Dative) subject. The question that naturally arises here is whether this is a genuine or an accidental gap in the data. That is, were infinitives with a f-feature independent subject able to antiagree or not? This is a well-known problem of working with corpora, of course: The researcher does not necessarily know whether data that do not occur instantiate an accidental gap or they are missing from the corpus because they are ungrammatical. If the corpus is big enough and the structure in which the data of interest could potentially occur are frequent enough, then it is reasonable to conclude that the relevant data are missing because they are ungrammatical. This is not the case with anti-agreeing infinitives with a referentially independent subject, however. As already mentioned before, predicates that take an infinitive with a f-feature independent subject are much less frequent in the codices than predicates that take control, raising, or ECM infinitives. Furthermore, anti-agreement itself is optional and relatively infrequent even in those structures in which it definitely can occur: In control, raising, and ECM infinitives it is in free variation with uninflected and agreeing infinitives and is by far the least frequent of the three options. Since both infinitives with a referentially independent subject and anti-agreeing infinitives have a low frequency, their combination may simply constitute an accidental gap. On the other hand, if the gap in the data is real, and anti-agreement is not possible in OH with infinitives that have a referentially independent subject, then we must ask what the common property of control, raising, and ECM is that allows anti-agreement to arise. The analysis of control has for some time now been a battleground of the Agree based PRO analysis (Landau 2000, 2003, 2004, 2006, 2008) and the movement analysis (Boeckx & Hornstein 2003, 2004, 2006a,b; Boeckx et al. 2010; Hornstein & Polinsky 2010). The structure of ECM infinitives also remains controversial: The Accusative nominal is taken either to

Anti-agreeing infinitives in Old Hungarian

217

get case from the matrix verb in the infinitival subject position (Chomsky 1981), or to raise to the matrix object position (Postal 1974; Lasnik Boškovíc 2002; Runner 2006). Subject-to-subject raising, the movement analysis of control and the raising to object analysis of ECM all involve A-movement. The referentially independent Dative subject of the infinitive, on the other hand, certainly does not move out of the infinitival clause. It is possible, then, that the movement of the infinitive’s subject makes it possible for anti-agreement to arise. Anti-agreement in the Bantu languages is certainly related to movement. In these languages anti-agreement arises in subject interrogatives: Wh-subjects move out of the canonical subject position, which prevents the finite verb from agreeing with them (see (55)). This, however, is A-bar movement, and it remains to be seen if anti-agreement could also plausibly be caused by the A-movement of subjects. It also remains to be seen why anti-agreement is obligatory in the Bantu case but optional in OH infinitives.

7 Conclusion This paper has shown that Old Hungarian features anti-agreement on infinitives. In the data accessible to me, anti-agreement occurs in control, raising and ECM infinitives but not in infinitives with a referentially independent Dative subject. In those contexts where anti-agreement can occur, it is in free variation with agreeing and uninflected infinitives. Anti-agreement thus appears to be optional, and it is also less frequent than the other two alternatives. Previous treatments of OH infinitival anti-agreement suggest that the relevant data do not instantiate anti-agreement. They treat the -nia/nya/nÿa/nie/nye/nÿe ending that occurs with non-3SG subjects as a monomorphemic suffix, claiming that -a/e has lost its status as a 3SG agreement marker and fused into the infinitival ending. I have shown the weaknesses of this approach and argued that the relevant cases involve genuine 3SG inflection with a non-3SG subject as a result of default agreement. I also speculated that if OH infinitival anti-agreement is indeed only possible in control, raising and ECM infinitives, then the default agreement might be triggered by movement. Subject-to-subject raising infinitives definitely involve movement, control and ECM both have movement analyses, while referentially independent Dative subjects do not move out of the subject position of the infinitival clause.

218

Éva Dékány

8 Primary sources Cornides Codex. 1514–1519. András Bognár and Ferenc Levárdy (eds.), Cornides kódex. Facsimile, and critical edition. Budapest: Akadémiai Kiadó, 1967. Döbrentei Codex. 1508. Csilla Abaffy, Csilla T. Szabó, and Edit Madas (eds.), Döbrentei-kódex. Halábori Bertalan keze írásával. Facsimile, transcription of the original record, with introduction and notes. Budapest: Argumentum Kiadó–Magyar Nyelvtudományi Társaság, 1995. Festetics Codex. Before 1494. Csilla Abaffy (ed.), Festetics-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Argumentum, Magyar Nyelvtudományi Társaság. 1996. Funeral Sermon and Prayer. 1192–1195. In Régi magyar nyelvemlékek, ed. by Adrienne Dömötör. Budapest: Akadémiai Kiadó. 2006. 27. Guary Codex. Before 1495. Dénes Szabó (ed.), Guary-kódex. Budapest, 1944. Jókai Codex. C. 1440. János P. Balázs (ed.), Jókai-kódex. Transcription of the original record and its latin equivalent with introduction and notes. Budapest: Akadémiai Kiadó. 1981. Jordánszky Codex. 1516–1519. A Jordánszky-kódex bibliafordítása. Printed by Ferenc Toldy. Introduction by György Volf. 1888. Horvát Codex. 1522. Lea Haader and Zsuzsanna Papp (eds.), Horvát-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1994. Kazinczy Codex. 1526–1541. Zsuzsa Kovács (ed.), Kazinczy-kódex. Budapest: Magyar Nyelvtudományi Társaság, 2003. Könyvecse (= Könyvecse az szent apostoloknak méltóságokról). 1521. István Pusztai (ed.), Könyvecse az szent apostoloknak méltóságokról. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság. 1985. Miskolc Fragment. 1525. Zsuzsanna Papp and Zsuzsa Kovács (eds.), Vitkovics-kódex és Miskolci töredék. Facsimile, transcription of the original record, with notes. Budapest: MTA Nyelvtudományi Intézet, 1991. Munich Codex. 1466. Antal Nyíri (ed.), A Müncheni kódex 1466-ból. Transcription of the original record and its latin equivalent. Budapest: Akadémiai Kiadó. 1971. Soproni Virágének. around 1490. In Magyar nyelvemlékek, ed. by József Molnár and Györgyi Simon. Budapest, Tankönyvkiadó. 1977. 107. Székelyudvarhely Codex. 1526–1528. Csilla N. Abaffy (ed.), Székelyudvarhelyi kódex Facsimile and transcription of the original record with notes. Budapest: Magyar Nyelvtudományi Társaság. 1993. Teleki Codex. 1525–1531. György Volf (ed), Döbrentei codex. Teleki codex. Budapest: A Magyar Tudományos Akadémia Könyvkiadó Hivatala. 1884. Vienna Codex. Mid-15th c. Gedeon Mészöly (ed.), Bécsi kódex. Budapest: MTA. 1926. Weszprém Codex. around 1512. István Pusztai (ed.), Weszprémi-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Magyar Nyelvtudományi Társaság, 1988. Winkler Codex. 1506. István Pusztai (ed.), Winkler-kódex. Facsimile, transcription of the original record, with introduction and notes. Budapest: Akadémiai Kiadó, 1988.

Anti-agreeing infinitives in Old Hungarian

219

References A. Jászó, Anna. 1992. Az igenevek [Non-finite verbs]. In Loránd Benkő, E. Abaffy Erzsébet & Rácz Endre (eds.), A magyar nyelv történeti nyelvtana II/1. A kései ómagyar kor morfematikája. [A Historical Grammar of Hungarian. Volume II/1.], 411–454. Budapest: Akadémiai Kiadó. Baker, Mark. 2008a. The Syntax of Agreement and Concord. Cambridge: Cambridge University Press. Baker, Mark C. 2008b. On the nature of the anti-agreement effect: Evidence from wh-in-situ in Ibibio. Linguistic Inquiry 39(4). 615–632. Balassa, József. 1894. A szlavóniai nyelvjárás [The Slavonian dialect]. Magyar Nyelvőr 23. 259– 267. Bartos, Huba. 1999. Morfoszintaxis és interpretáció. A magyar inflexiós jelenségek szintaktikai háttere. [Morphosyntax and Interpretation. The Syntactic Background of Hungarian Inflexional Phenomena]. Budapest: Eötvös University dissertation. Bartos, Huba. 2000. Az inflexiós jelenségek szintaktikai háttere [The syntacic background of inflexional phenomena]. In Ferenc Kiefer (ed.), Strukturális magyar nyelvtan 3. Morfológia. [Hungarian Structural Grammar 3. Morphology], 653–761. Budapest: Akadémiai Kiadó. Benveniste, Émile. 1971. Problems in General Linguistics. Miami: University of Miami Press. Boeckx, Cedric & Norbert Hornstein. 2003. Reply to “Control is not movement”. Linguistic Inquiry 34(2). 269–280. Boeckx, Cedric & Norbert Hornstein. 2004. Movement under control. Linguistic Inquiry 35(3). 431–452. Boeckx, Cedric & Norbert Hornstein. 2006a. Control in Icelandic and theories of control. Linguistic Inquiry 37(4). 591–606. Boeckx, Cedric & Norbert Hornstein. 2006b. The virtues of control as movement. Syntax 9(2). 118–130. Boeckx, Cedric, Norbert Hornstein & Jairo Nunes. 2010. Icelandic control really is A-movement: Reply to Bobaljik and Landau. Linguistic Inquiry 41(1). 111–130. Bošković, Željko. 2002. A-movement and the EPP. Syntax 5(3). 167–218. Chomsky, Noam. 1981. Lectures on Government and Binding: The Pisa Lectures. Studies in Generative Grammar 9. Dordrecht: Foris Publications. Dékány, Éva. 2011. A Profile of the Hungarian DP. The Interaction of Lexicalization, Agreement and Linearization with the Functional Sequence. Tromsø: University of Tromsø dissertation. Dékány, Éva. 2015. The syntax of anaphoric possessives in Hungarian. Natural Language and Linguistic Theory 33(4). 1121–1168. Dikken, Marcel den. 1999. On the structural representation of possession and agreement. The case of (anti-)agreement in Hungarian possessed Nominal Phrases. In István Kenesei (ed.), Crossing Boundaries: Theoretical Advances in Central and Eastern European Languages, 137–178. Amsterdam: John Benjamins. Dikkken, Marcel den. 2013. Phi-features. Handout of the Morphology seminar held at the CUNY Graduate Center, Fall 2013. É. Kiss, Katalin. 2002. The Syntax of Hungarian. Cambridge: Cambridge University Press. Elouazizi, Noureddine. 2005. Anti-agreement effects as (anti)-connectivity. In John Alderete (ed.), Proceedings of 24th WCCFL, 120–128. Sommerville, MA: Cascadilla. Farkas, Donka F. & Henriëtte de Swart. 2010. The semantics and pragmatics of plurals. Semantics and Pragmatics 3. 1–54.

220

Éva Dékány

Harley, Heidi & Elizabeth Ritter. 2002a. Person and number in pronouns: A feature-geometric analysis. Language 78(3). 482–526. Harley, Heidi & Elizabeth Ritter. 2002b. Structuring the bundle: A universal morposyntactic feature geometry. In Heike Wiese & Horst J. Simon (eds.), Pronouns, 23–39. Amsterdam and Philadelphia: John Benjamins. Hornstein, Norber & Maria Polinsky (eds.). 2010. Movement Theory of Control. Amsterdam: John Benjamins. Ionin, Tania & Ora Matushansky. 2002. DPs with a twist: A unified analysis of Russian comitatives. In Browne Wayles, Ji-Yung Kim, Barbara H. Partee & Robert A. Rothstein (eds.), Proceedings of Formal Approaches to Slavic Linguistics 11: The Amherst meeting 2002, 255–274. Ann Arbor, MI: Michigan Slavic Publications. Jones, Michael Allan. 1993. Sardinian Syntax. London and New York: Routledge. Károly, Sándor. 1956. Igenévrendszerünk a kódexirodalom első szakaszában [The System of NonFinite Froms in the Early Codices] Nyelvtudományi értekezések 10. Budapest: Akadémiai Kiadó. Keresztes, Kálmán. 1953. A személyragos főnévi igenév használatáról [On the use of the inflected infinitive]. Magyar Nyelvőr 77. 340–352. Landau, Idan. 2000. Elements of Control: Structure and Meaning in Infinitival Constructions. Studies in Natural Language and Linguistic Theory 51. Dordrecht: Kluwer. Landau, Idan. 2003. Movement out of control. Linguistic Inquiry 34(3). 471–498. Landau, Idan. 2004. The scale of finiteness and the calculus of control. Natural Language and Linguistic Theory 22(4). 811–877. Landau, Idan. 2006. Severing the distribution of PRO from case. Syntax 9(2). 153–170. Landau, Idan. 2008. Two routes of control: Evidence from case transmission in Russian. Natural Language and Linguistic Theory 26(4). 877–924. Lasnik, Howard & Mamoru Saito. 1991. On the subject of infinitives. In Lise Dobrin, Lynn Nichols & Rosa Rodriguez (eds.), Papers from the Twenty-Seventh Regional Meeting of the Chicago Linguistic Society, 324–343. Chicago: Chicago Linguistic Society. Longa, Víctor Manuel. 1994. The Galician inflected infinitive and the theory of UG. Catalan Working Papers in Linguistics 4(1). 23–44. López, Luis. 2008. The [person] restriction: Why? and, most specially, why not? In Roberta D’Alessandro, Susan Fischer & Gunnar Hrafn Hrafnbjargarson (eds.), Agreement Restrictions. Interface Explorations 15, 129–158. Mouton de Gruyter. Miller, Gary D. 2003. Where do conjugated infinitives come from? Diachronica 20(1). 45–81. Miller, Gary D. 2004. The origin of the Welsh conjugated infinitive. Diachronica 21(2). 329–350. Modesto, Marcello. 2007. Inflected infinitives in Brazilian Portuguese as an argument both contra and in favor of a movement analysis of control. Revista Letras 72. 297–309. Nevins, Andrew. 2007. The represenation of third person and its consequences for person-case effects. Natural Language and Linguistic Theory 25(2). 273–313. Nevins, Andrew. 2011. Multiple agree with clitics: Person complementarity vs. omnivorous number. Natual Language and Linguistic Theory 29(4). 939–971. Ouali, Hamid. 2008. On C-toT transfer: The nature of agreement and anti-agreement in Berber. In Roberta D’Alessandro, Susann Fischer & Gunnar Hrafn Hrafnbjargarson (eds.), Agreement Restrictions, 159–180. Berlin and New York: Mouton de Gruyter. Ouhalla, Jamal. 1993. Subject-extraction, negation and the anti-agreement effect. Natural Language and Linguistic Theory 11(3). 477–518.

Anti-agreeing infinitives in Old Hungarian

221

Ouhalla, Jamal. 2005. Agreement features, agreement and anti-agreement. Natural Language and Linguistic Theory 23(3). 655–686. Pires, Acrisio. 2001. Clausal and TP-defective gerunds: Control without Tense. In Minjoo Kim and Uri Strauss (eds.), Proceedings of NELS 31, 386–406. Massachusetts, Amherst: GLSA. Pires, Acrisio. 2006. The Miminalist Syntax of Defective Domains: Gerunds and Infinitives Linguistik Aktuell/Linguistics Today 98. Amsterdam: John Benjamins. Postal, Paul. 1974. On Raising: An Inquiry into One Rule of English Grammar and its Theoretical Implications. Current Studies in Linguistics Series 5. Cambridge, MA: MIT Press. Rákosi, György & Tibor Laczkó. 2008. On the categorial status of agreement-marked infinitives in Hungarian. In Christopher Piñón & Szentgyörgyi Szilárd (eds.), Approaches to Hungarian 10, 149–172. Budapest: Akadémiai Kiadó. Raposo, Eduardo. 1987. Case theory and Infl-to-Comp: The inflected infinitive in European Portuguese. Linguistic Inquiry 18(1). 85–109. Richards, Norvin. 2001. Movement in Language. Interactions and Architectures Oxford Linguistics. Oxford: Oxford University Press. Rooryck, Johan. 2003. The morphosyntactic structure of articles and pronouns in Dutch. In Jan Koster & Henk van Riemsdijk (eds.), Germania et alia: A Linguistic Webschrift for Hans den Besten, 1–12. Groningen: University of Groningen. Runner, Jeffrey. 2006. Lingering challenges to the raising-to-object and object-control constructions. Syntax 9(2). 193–213. Schneider-Zioga, Patricia. 2000. Anti-agreement and the fine structure of the left edge. In Ruixi Ai, Francesca del Gobbo, Makie Irie & Hajime Ono (eds.), Working Papers in Linguistics 6, 94–114. Irvine: Department of Linguistics, University of California. Schneider-Zioga, Patricia. 2007. Anti-agreement, anti-locality and minimality: The syntax of dislocated subjects. Natural Language and Linguistic Theory 25(2). 403–446. Scida, Emily. 2004. The Inflected Infinitive in Romance Languages Outstanding Dissertations in Linguistics. New York: Routledge. Sigurðsson, H. Ármann. 2000. The locus of case and agreement. Working papers in Scandinavian syntax 65. 65–108. Simon, Eszter & Bálint Sass. 2012. Nyelvtechnológia és kulturális örökség, avagy korpuszépítés ómagyar kódexekből [Language technology and cultural heritage: Building corpora from Old Hungarian codices]. Általános Nyelvészeti Tanulmányok 24. 243–264. Simonyi, Zsigmond. 1892. A magyar határozók [Adverbs in Hungarian], vol. 2. Budapest: Magyar Tudományos Akadémia. Szabó, Sándor. 1907. A szigetközi nyelvjárás [The Szigetköz Dialect]. Budapest: Athaeneum. Szarvas, Gábor. 1876. A Slavoniai tájszólás II [The Slavonian dialect II]. Magyar Nyelvőr 5. 61–65. Tóth, Ildikó. 2000. Inflected Infinitives in Hungarian: University of Tilburg dissertation. Tóth, Ildikó. 2002. Can the Hungarian infinitive be possessed? In István Kenesei & Péter Siptár (eds.), Approaches to Hungarian 8, 134–160. Budapest: Akadémiai Kiadó. Tóth, Ildikó. 2011. A ragozott főnévi igenevek a kései ómagyar korban [Inflected infinitives in Late Old Hungarian]. In Marianne Bakró-Nagy & Forgács Tamás (eds.), A nyelvtörténeti kutatások legújabb eredményei VI [The Newest Results of Diachronic Research VI], 249– 265. Szeged: SZTE.

Rosemarie Lühr

9 The emergence of expressions for purpose relations in older Indo-European languages The present study discusses the development of purpose constructions in IndoEuropean, with a focus on the oldest Indo-European languages, as well as the reconstruction of strategies coding the function of purpose in Proto-IndoEuropean. The investigation includes Old Indic purpose constructions and some infinitive formations in Hittite. The focus is on Old Indic, because this language is the first to show competition between purpose infinitives and finite structures. It has to be asked to what degree these constructions compete exactly with each other. After a discussion of the theoretical background the primary data is presented. For the question of the origin of purpose constructions the oldest infinitive in Vedic, namely that on -dhyai, is of interest, for its original function highlights the primary function of infinitives. It is hypothesized that the earliest purpose constructions in Indo-European were infinitives, interpreted as being equivalent to constructions which could show subject control.1 As we can see from language history and from studies on child language, explicit expressions to verbalize the purpose of a certain action are not fundamental to language, and the relevant devices of Indo-European languages are not very old. Reconstructing some means for the expression of purpose in ProtoIndo-European is thus impossible. However, as the attested languages are genetically related, we would not expect the developments to differ completely. And indeed, one can easily discover two general strategies to create expressions for purpose in the languages, namely the functional extension of verbal nouns and verbal endings on the one hand, and the emergence of an elaborate system of subordinate clauses on the other hand. In most Indo-European languages both possibilities exist side by side, and it is worthwhile analyzing the differences. Crucial issues concern the competition between finite and non-finite structures

1 Clauses of the type With Mary [to talk to] . . . you won’t be bored are regarded as small clauses. But as Wilder (1991: 224) points out, a small clause analysis is ruled out by the following condition: XP can be the subject of a predicate of YP if no maximal projection dominating YP excludes XP. Rosemarie Lühr, HU Berlin DOI 10.1515/9783110520583-009

224

Rosemarie Lühr

and control phenomena. The data comes from the oldest Indo-European languages, from Hittite and Vedic, the oldest attested language stage of Old Indic. The question here is: Which of the strategies denoting purpose are the oldest and are potentially inherited from Indo-European? The study is organized as follows: First, a rough outline of the theoretical approach is presented. Second, we take a look at Hittite examples. Third, a discussion of the Vedic, especially the infinitive on Vedic -dhyai, follows. Central to this train of thought is the competition with finite purpose constructions. Fourth, the development of the ending -dhyai is pursued. It is assumed that this is originally a verbal ending. A parallel to this reanalysis can be found in the Romance languages. At the end, a comparison of the purposive strategies is made.

1 Theoretical background We discuss the theoretical background of our analysis of purpose constructions with English and Russian material, because the properties of English and Russian non-finite and finite purpose structures can be detected in the older IndoEuropean languages, too. According to Cristofaro (2003: 157f.) purpose relations generally link two states of affair one of which (the main one) is performed with the goal of obtaining the realization of another one (the dependent one). The prototypical purpose relation seems to be one in which the main and dependent states of affair are performed by the same entity, which can control the realization of the dependent state of affair. There may be a constituent (often phonetically empty) that evokes a result-state description. This constituent can be detected by means of a purpose infinitive construction (e.g. Max baked a chocolate cake for me to admire, cf. Nissenbaum 2005). However, most of the data on purpose relations concern the purpose of motion. As for semantics, purpose relations are quite similar to those of the complement relations established by desiderative predicates (modals, manipulatives). It is therefore assumed that purpose relations belong to the domain of deontic modality, where “deontic modality . . . concerns what is possible, necessary, permissible, or obligatory, given a body of law or a set or moral principles or the like” (von Fintel 2006: 2). But, as in this case, a special context forcing a deontic reading is needed, purposivity is also associated with teleological modality: “Teleological modality . . . concerns what means are possible or necessary for achieving a particular goal” (von Fintel 2006: 2). This means that the realization of the dependent state of affairs is presented as possible at a future point in time with respect to that at which the main state of affair is located. Concerning syntax it is generally agreed that

The emergence of expressions for purpose relations

225

purpose constructions have adjunct status (Jones 1991: 64), but they differ from typical adverbial relations, and represent a special case (Schmidtke-Bode 2009).

1.1 Non-finite purposive constructions In the following, we add the English translations of the Hittite and Vedic data to the corresponding English and Russian constructions to show which structures are documented in the old Indo-European languages, too, and which are not. Detailed analyses of the Hittite and Vedic examples are given in Sections 2 and and 3. In this section the relevant examples are rendered as English translations. To start with English non-finite purpose constructions, Jones (1991: 25f.) distinguishes three types: (1)

a.

Maryi brought Johnk along [in order ei to talk to himk].

in-order-to-clause2

b.

Maryi brought Johnk along [ei to talk to himk].

subject-gap purpose clause

c.

Maryi brought Johnk along [ei to talk to ek].

object-gap purpose clause

Type (1d) is usually referred to as a purpose clause: (1)

d. We’ve been hiring guardsi [ei to watch the children].

and (1e) as a rationale clause. (1)

e. Grassi is green [ei to promote photosynthesis]. (Landau 2013: 224)

Because of the conjunction čtoby the in-order-to-type is also assumed for Russian. But in each of the three purpose constructions čtoby is optional:3

2 In English, only in-order-to-constructions allow for wh-extraction: Whoi did you go to England [in order to meet ti] Furthermore, in-order-to-constructions are acceptable with perfective have inflection, but not the other purpose construction types (1b), (1c), (1d). 3 Cf. Faraci’s (1974) differentiation for rationale, objective and purpose clauses: (a) John trains the new recruits [to make a living for himself] rationale clause (b) John trains the new recruits [to make a living for themselves] objective clause (c) Carol bought a rack [to hang coats on] purpose clause

226

Rosemarie Lühr

(2) a. Maša vzjala ščetku [(čtoby) počistit’ plat’e] Mascha take: pret.fem. clothbrush: acc. that clean: inf. dress: acc. ‘Mascha took a clothbrush to clean the dress’ b. Muž dal žene den’gi [(čtoby) zaplatit’ za kvartiru] husband give: pret.masc. wife money that pay: inf. for lodging ‘The husband gave his wife money to pay for lodging’ c. Anton prines knigu [(čtoby) po-čitat’] Anton bring-pret book: acc. that deliminative-read: inf. ‘Anton brought a book to read for a while.’4 (Junghanns 1994: 107ff)

Vedic and Hittite rationale clauses are shown in (47) and (13), purpose clauses in (48) and (16): Vedic (47) Unyoke, o heroi, as at this journey’s end [ei to delight today in our Soma sacrifice] Hittite (13) And hei goes back down [ei to sleep] Vedic (48) Then, o Indra, lord of tawny coursers, these sistersi, goddesses, are praised, when you released the prisoned onesi with your help [ei to flow after a long time (i.e. captivity)] Hittite (16) To the kingi they give barley beerk [ei to drink ek] Control is the crucial factor for all these infinitive constructions. Williams’ (1980) distinction between obligatory control and non-obligatory control is essential here (cf. also Wurmbrand 2002). Obligatory control shows five properties: (3) 1. Lexical NP cannot appear in the position of PRO. 2. The antecedent precedes the controlled PRO. 3. The antecedent c-commands the controlled PRO. 4. The antecedent is thematically . . . or grammatically . . . uniquely determined. 5. There must be an antecedent. 4 For rationale and objective embedding in Russian cf. Junghanns (1994, 1994a).

The emergence of expressions for purpose relations

227

Checking the purpose construction types according to these properties, in-orderto-constructions show non-obligatory control, permitting context control (Jones 1991: 37): (4) The lights were turned off [in order earb to conserve electricity]. Non-obligatory control is also found with the subject of purpose clauses: (5) a. Bambik was brought [earb to read ek to the children]. b. I(i) brought this winek over [ei/arb to enjoy ek with our dinner].

In (5a) the referent of the subject is pragmatically determined. Characteristic of this purpose construction is that the sentence becomes ungrammatical if the object position in the infinitive clause is filled by an overt pronoun: (5) a’ *Bambik was brought [earb to read itk to the children].

On the contrary, in (5b) the matrix NP, I, could control the purpose clause subject as well as an arbitrary controller. Hittite and Vedic examples with a purpose clause containing arbitrary control are given in (18a) and (31), respectively. Hittite (18) a. one jug of winek [earb to libate ek] And with a verbal noun functioning like a purpose clause: Vedic (31) Indra, give usk wealth in brave men, good steeds and good cows [ for earb first thinking [of usk] like the priestk]. Purpose clauses (object-gap purpose clauses) can also take an exhaustive subject: (6) I brought this winek over for Johni [ei to enjoy ek with dinner]. (60) and (52) are relevant Vedic examples: Vedic (60) The rays bear Jātavedask up aloft, the god for alli [in order ei to look on ek] (52) When youi, Indra, take the club reeling with excitement in the arms for the snakek [ei to slay ek] (according to Keydana’s interpretation; but also confer with the discussion in Section 3.4.)

228

Rosemarie Lühr

The empty object position, however, must always be controlled by obligatory control. Obligatory control also occurs with the subject of rationale clauses. Subject control in these clauses can never be arbitrary control. Subject control in a rationale clause in Vedic is shown in (46): Vedic (46) To him then Ii offer this highest sun winning song of praise [ei to magnify with songs of invocation and with hymns the glorious] (49) is a Vedic example with object control in a purpose clause caused by a possessive pronoun of the matrix clause: Vedic (49) Make hisi ears hear [ei to show his vigor and (steer him) in the habitual direction ei to get excited] There are also syntactic differences between rationale clauses and purpose clauses. Purpose clauses are VP-internal, containing a gap bound by the matrix object, rationale clauses are external to the VP, and are not dependent on the matrix object. In the structure tree, purpose clauses are always attached lower than rationale clauses (cf. Faraci 1974, Huettner 1989). As for the semantics of the matrix predicate in English purpose clauses, Bach (1982: 38) identified three restrictions (cf. also Johnston 1998: 89): (7) a. have, be (in a place, on hand, available, at one’s disposal, in existence . . .): i. I have my mother [to look after e]. ii: Spoons are [to eat soup with e]. iii: Johni has an umbrellaj in the closet [ei to use ej when it rains]. iv. The umbrellaj is kept in the closet for youi [ei to use ej when it rains]. b. transitive verbs which involve continuance or change in the states of affairs indicated in (a) and are of a “positive” sort: i. I brought it [to build a fire with e]. ii. ??I destroyed it [to build a fire with e]. iii. Johni bought the umbrellaj [ei to use ej when it rains]. iv. I baked a cake [to eat with dinner]. c. verbs of choice and use: i. Mary chose John [to go to the dance with e]. ii. I use it [to keep my pencils in e]. iii. I chose a Jane Austen novel to read to the students. iv. Ii used itj [ei to slice the salami with ej].

The emergence of expressions for purpose relations

229

A Vedic example of type (7)(c)(iv) is given in (50). Vedic (50) O Vājas and Ṛbhukṣans, explore the paths to sacrifice for usi, masters, lauded [ei to press forward to each direction]. These restrictions are reminiscent of the predicate types purpose clauses are compatible with (Faraci 1974, Bach 1982, Jones 1985): (8) a. I bought that convertible for you to admire b. #I drove that convertible for you to admire

change of state non-change of state

c. I drove that convertible in order for you to admire me d. I planted that tree for my kids to play on

‘positive’ change

e. #I chopped it down to prevent my kids from playing on

‘negative’ change

f.

a pragmatic difference? (cf. Huettner 1989)

I chopped it down to use as firewood

The reason that the meanings differ only subtly in many cases (and are sometimes indistinguishable) is that when the result state is taken to be the direct, intended consequence of an action, the most salient goal that can be expressed about the result state is simply the one held by the agent of the causing event (cf. Nissenbaum 2005).

1.2 Finite purposive constructions Unlike the non-finite purpose constructions, in English finite purpose clauses are relatively rare. They are introduced by complementizers, for example that and lest, and can also be used if the subject of the matrix clause is identical with that of the subordinate clause. The important feature here is that with that the purpose clause takes may in the present and future, and might in the past, whereas with lest it takes should or may (cf. Schmidtke-Bode 2009 and the examples in (9)). We find a similar use with Old Indic purpose clauses, cf. the Vedic examples (23) and (24). (9) a. I play the violin that I may enjoy myself. b. I grabbed the rope lest I should fall.

230

Rosemarie Lühr

Vedic (23) Br ̥haspati gives us great [helps], o friends, that we may be guiltless for the bounteous (god). (24) Come readily to this mine invocation, that you may drink the juice lauded.

2 Hittite In Hittite, in main clauses and subordinate clauses a purpose meaning must be inferred from the context. Sometimes periphrasis yields a purpose sense: (10) CTH 81: Autobiography of Hattušilis III, III 9 f nu pa-a-un nu URUHa-wa-ar-ki-na-an and go: 1sg.pret.act. and Hawarkina: acc.sg.c. URUDi-el-mu-na-an-na

ú-e-da-ah-hu-un Delmuna: acc.sg.c.=and fortify: 1sg.pret.act.

‘I went and fortified Hawarkina and Delmuna’ ~ ‘I went to fortify Hawarkina and Delmuna’

Sometimes parataxis may have a purposive sub-sense: (11) KUB XXIX 1 (CTH 414.A: Foundation ritual), I 35 f. nu-wa-mu i-ni GIŠ-ru ma-ni-ya-ah and=quot=to me this: acc.s.gn. tree: acc.sg.n. transfer: 2sg.imp.act. na-at-kán kar-aš-mi and=it=ptcl fell: 1sg.pres.act. ‘Give me that tree and I’ll fell it’ ~ ‘Give me that tree so that I can fell it’

However, a consecutive meaning can also be proposed, since the result of an action may not only be an intended result but also accidental. This applies also to subordinated clauses which can be purposive as well as consecutive. (12) Myth of Telipinu, KBo III 7+ I 5ff. (CTH 257) ud-ni-wa ma-a-ú še-eš-du land: nom.sg.n.=quot prosper: 3sg.imp.act. flourish: 3sg.imp.act. nu-wa ud-ni-e pa-ah-ša-nu-wa-an e-eš-du and=quot land: nom.sg.n. protect: part.nom.sg.n. be: 3sg.imp.act.

The emergence of expressions for purpose relations

231

nu ma-a-an ma-a-i še-eš-zi and if/when prosper: 3sg.pres.act. florish: 3sg.pres.act. nu EZEN4 pu-ru-ul-li-ya-aš i-ya-an-zi now feast purulli: gen.sg. make: 3pl.pres.act. ‘May the land prosper (and) flourish, may the land be protected! So that it prospers (and) flourishes, they make the purulli-festival’

As one does not make a ritual feast if or when a land prospers, but to make it prosper, the conjunction mān can be purposive as well as consecutive. Yet, this is quite a rare case for the use of mān, and there is no exclusive conjunction for purposivity (Zeilfelder 2001). However, the infinitive, when it depends on either a finite verb or a noun, can express purpose. In (13) it depends on a finite verb: (13) KUB V 1 I 61 EGIR-pa =ya =aš =kan šešuanzi GAM DU-ri back =and =he: nom.sg. =pctl sleep: inf. down go: 3sg.pres.med. And hei goes back down [ei to sleep] (Holland 2011: 69, 71) ‘And he goes back down to sleep’

The enclitic pronoun -aš is the subject of the motion verb DU-ri and its intransitive infinitive šešuanzi. The infinitive is a rationale clause. Another purpose construction type is represented by (14): (14) KBo V 8 I 8 dUTU -ŠI =wa šumāš walaḫḫuwanzi uizzi majesty: nom.sg. =my =quot you: dat/acc smite: inf. come: 3sg.pres.act My majestyi will come to youk [ei to smite ek] My majestyi will come [ei to smite you] ‘My majesty will come to smite you’

šumāš can be either dative or accusative. If the underlying structure is ‘come to you (dat.) to smite (you)’, where the dative is construed with uizzi, the object position of the infinitive construction is an empty category. Otherwise, šumāš is the direct object of the infinitive construction (Holland 2011: 69ff.). In the first case, the construction is a purpose clause with object control, in the second, a rationale clause. A clear purpose clause is shown in (15):

232

Rosemarie Lühr

(15) KBo IV 4 ii 63-64 m.dLAMMA-ašš =a kue KARAŠ.ḪI.A INA KUR Lamma: nom.sg.c. =and acc.pl.c. troop: acc.pl. country: dat./loc. ḫalki ḪI.A-uš ḫarnikuwanzi pēḫudan Nuḫašši: dat.sg. grain: acc.pl.c. destroy: inf. brought: part

URUNuḫašši

ḫarta . . . has: 3sg.pret.act And the troopsi which Mr. LAMMA had brought to Nuḫašši [ei to destroy the grain . . .] ‘And the troops which Mr. LAMMA had brought to Nuḫašši to destroy the grain . . .’

The accusative object of the finite transitive matrix verb pēḫudan ḫarta (kue KARAŠ.ḪI.A, troops’) is the subject of the transitive infinitive ḫarnikuwanzi (ḫalk ḪI.A-uš). (16) is a clear purpose clause with object control: (16) KUB XXV 36 (CTH 647.6: Ritual for the prince), II 12’ LUGAL-i a-ku-wa-an-na mar-nu-an pí-an-zi king: dat.sg.c. drink: inf. II barley beer: acc.sg. give: 3pl.pres.act. To the kingi they give barley beerk [ei to drink ek] ‘To the king they give barley beer to drink.’

An active and passive reading of the infinitive seems possible in (17): (17) KBo IV 4 iv 20-21 BELI =NI =wa =nnaš ŠA URUAripšā iwar lord: voc.sg. =our =quot =us: acc.pl. gen Aripšā like URUḪattuši

šāruwauwanzi lē maniyaḫti Ḫattuša: dat./loc.sg. plunder: inf. not hand over: 2sg.pres.act.

Do not, our lord, hand usk over to Ḫattušai [ei to plunder ek like Aripšā] Do not, our lord, hand usi over to Ḫattuša [ei to be plundered like Aripšā] ‘Do not, our lord, hand us over to Ḫattuša to be plundered (= to plunder) like Aripšā.’

If the recipient in the matrix clause, Ḫattuši, functions as the controller of the transitive infinitive šāruwauwanzi, the infinitive construction is a purpose clause with object control, but with naš as controller it is a rationale clause with a passive reading.

The emergence of expressions for purpose relations

233

The infinitive may also depend on a noun, as in (18): (18) KUB 7.53 I 23 DUGḫaniššaš GEŠTIN šipanduwanzi 1 one jug: nom.sg.c. wine libate: inf. ‘one jug of wine for libating’ (Hoffner & Melchert 2008: 332)

The subject of the infinitive is arbitrary, the noun GEŠTIN is the controller of the empty object of šipanduwanzi: (18) a. one jug of winek [earb to libate ek] Therefore, a purpose clause should be assumed here.

3 Vedic In contrast to Hittite, in Vedic true finite purpose sentences are documented. They already appear in the oldest extant text, the R̥ gveda, being representative of an early Indo-Aryan language, and occur parallel to purpose infinitives. As mentioned above, the following discussion focuses on the infinitive on -dhyai, because it is almost completely restricted to the oldest text group and also appears in Old Iranian, and can therefore be traced back to its origin. In Vedic, in purpose clauses two conjunctions can be found: the polyfunctional conjunctions yád and yáthā. The indicator to distinguish the purpose use from other usages is mood: Purpose yád and yáthā demand the subjunctive, more seldom the optative.5 As language comparison shows, this is a well-known strategy. For example Latin, Greek and most Slavic languages can only form purpose clauses with subjunctive morphology in the embedded clause (Junghanns 1994a as well as Wiemer this volume for the role of mood in Slavic independent infinitives). Now, the question has to be answered whether finite and infinitival purpose constructions are competitors. In this context, Keydana’s recently published study on the infinitives in the R̥ gveda (2013) must be taken into account.

3.1 Complementary distribution With the exception of the following instances, Keydana (2013: 150–153) assumes free variation with the infinitive construction and finite sentences; Disterheft (1980: 65–69) even consistently argues in favor of free choice. 5 Hettrich (1988: 284f.). Hettrich’s (291) distinction between sprecherbezogen and neutral is not being observed here (cf. Keydana 2013: 144 note 119).

234

Rosemarie Lühr

First, finite purpose clauses, rather than infinitives, appear when the subjects of the matrix sentence and the subordinate clause are different6 and no rationale clause or purpose clause can be used; cf. (19): (19) RV 10,131,1 ápa prā́ ca índra víśvām̐ amítrān away eastern: acc.pl.m. Indra: voc.sg.m. all: acc.pl.m. enemy: acc.pl.m. ápā́ pāco bhibhūte nudasva . . . away=western: acc.pl.m. you superior: voc.sg.m. push away: 2sg.imp.pres.med. yáthā táva śárman mádema that you: gen.sg. shelter: loc.sg.n. be glad: 1pl.opt.pres.act. ‘Drive all our enemies away, Indra, the western, you superior one, and the eastern . . . that we in thy wide shelter may be joyful.’

As Keydana (2013: 148) points out, a rationale clause is impossible in this case, as the subject of the matrix sentence is not token identical with that of the subordinate sentence. A purpose clause (with arbitrary control) must also be excluded. The object would have to appear in an empty non-subject position in the purpose clause. Furthermore, a finite construction has to be chosen if the subject of the matrix clause and the subject of the infinitive construction stand in a wholepart relation, as in RV 7,104,3 duṣkr ̥taḥ ‘the wickets’: ékaś caná ‘one of them’ (Keydana 2013: 148).

3.2 Free variation? 3.2.1 Nominatives As for variation between finite and infinitival purpose structures, Keydana (2013: 149) assumes that in the next example, too, the finite structure is obligatory because for the verb as- ‘to be’ no infinitive is documented. Sentences like (20) with a predicative nominative are often documented in the R̥ gveda (cf. Hettrich 1988: 283):

6 For reasons of space in Lühr (1994) only the competition between finite purpose structures and infinitival purpose structures on ‑dhyai could be analyzed, which has been ignored by Keydana (2013).

The emergence of expressions for purpose relations

235

(20) RV 2,26,02 havíṣ kr ̥ṇuṣva subhágo oblation: acc.sg.n. prepare: 2sg.imp.pres.med. happy: nom.sg.m. yáthā́ sasi that=be: 2sg.subj.pres.act. Prepare [youi] oblation so that youi may get happyi ‘Prepare oblation so that you may get happy.’

But the lack of this infinitive could have been compensated for by the infinitive bhuvé of the verb bhavi ‘to become, to be’: (21) RV 10,88,10 tám ū akrṇ̥ van tredhā́ bhuvé káṃ he: acc.sg.m. ptcl make: 3pl.ind.impf.act. threefold become: inf. ptcl ‘They made him to appear in threefold essence’ (cf. Keydana 2013: 347)

Especially in deontic contexts, bhavi – and the verb as- ‘to be’ are used in the same way: (22) RV 7,35,8 śáṃ naḥ párvatā auspcious: nom.sg.n. us: dat.pl. mountain: nom.pl.m. dhruváyo bhavantu standing firm: nom.sg.m. be: 3pl.imp.pres.act. śáṃ naḥ síndhavaḥ śám u auspcious: nom.sg.n. us: dat.pl. river: nom.pl.m. auspicious: nom.sg.n. ptcl santv ā́ paḥ be: 3pl.imp.pres.act. water: nom.pl.f. ‘Auspicious be the firmly-seated mountains, auspicious be the rivers and the waters.’

It is therefore not surprising when a finite purpose clause displays a verbal form of bhav i ‘to become, to be’: (23) RV 7,97,2 bŕ ̥haspátir no maha ā́ sakhāyaḥ Br̥haspati: nom.sg.m. us: acc.pl. great: voc.pl.m. here friend: voc.pl.m.

236

Rosemarie Lühr

yáthā bhávema mīḷhúṣe ánāgā that be: 1pl.pres.opt.act. bounteous: dat.sg.m. guiltless: nom.pl.m. Br̥haspati gives usi great [helps], o friends, that wei may be guiltlessi for the bounteous (god) ‘Br̥haspati gives us great [helps], o friends, that we may be guiltless for the bounteous (god)’7

Thus, it is doubtful whether the missing infinitive of the verb as- is the reason for the use of the finite structure in (20). Rather, this structure could have been selected due to the predicative nominative which does not demand a covert subject PRO, but rather an overt subject noun or pronoun, which can be replaced by pro in pro-drop-languages, such as the oldest Indo-European languages. (The same holds for (24).) But if this is true, other finite purpose clauses with nouns and adjectives in the nominative must also be included. (24) is a further example: (24) RV 6,63,2 áram me gantaṁ hávanāyāsmaí readily my come: 2dual.imp.act. invocation: dat.sg.n.=this: dat.sg.n. gr ̥ṇānā yáthā píbātho ándhaḥ lauded: nom.dual.m. that drink: 2dual.subj.pres.act. juice: acc.sg.n. Come [youi] readily to this my invocation, that youi may drink the juice laudedi ‘Come readily to this my invocation, that you may drink the juice lauded.’ (Hettrich 1988: 281)

According to Keydana (2013: 54, 119), finite purpose constructions compete with infinitival ones here. Infinitive purpose constructions presumably show an appositive nominative reference to the covert subject of the infinitive constructions. One of the very seldom documented examples shows a rationale clause: (25) RV 4,2,1 yó mártyeṣv amŕ ̥ta r ̥tā́ vā who: nom.sg.m. mortal: loc.pl.m. immortal: nom.sg.m. law-abiding: nom.sg.m. devó devéṣv aratír nidhā́ yi | 8 god: nom.sg.m. god: loc.pl.m. aratí: nom.sg. appoint: 3sg.ind.aor.pass.

7 According to Keydana (2013: 149), neither a rationale clause nor a purpose clause can be construed, though the beneficient us is coindexed with the subject of the infinitive construction. In the case of a purpose clause with arbitrary control, an affected theme in the matrix sentence and its resumption in the infinitive construction would be required. 8 For the interpretation cf. Hayakawa (2014: 38ff).

The emergence of expressions for purpose relations

237

hótā yájiṣṭho mahnā́ śucádhyai sacrificer: nom.sg.m. best at worship: nom.sg.m. might: instr.sg.n. shine: inf. havyaír agnír mánuṣa oblation: instr.pl.m. Agni: nom.sg.m. human: gen.sg.m. īrayádhyai be raised: inf. (with passive interpretation)

But Keydana’s translation: ,Agni, der als Unsterblicher unter den Sterblichen, als Gesetzestreuer, als Gott unter den Göttern als Speichenkranz eingesetzt ist, um als der am besten opfernde Opferpriester mit Macht aufzuleuchten, um mit den Opferspenden des Menschen in Bewegung gesetzt zu werden‘

is not the only possibility. hótā yájiṣṭhaḥ could also be an apposition to agníḥ in the superordinate structure: (25) a. whoi as the immortali among mortals, law-abiding, the godi among the gods, appointed as aratí, as sacrificeri besti at worship in order to shine with might . . .9 The same applies to vāśrā́ ḥ, in a strucure with a rationale clause in (26). This adjective can also be an apposition to tyé sūnávaḥ ‘these sons’ in the superordinate structure: (26) RV 1,37,10 úd u tyé sūnávo gíraḥ upwards ptcl this: nom.pl.m. son: nom.pl.m. song: acc.pl.f. kā́ ṣṭhā ájmeṣv atnata goal: acc.pl.f. track: loc.pl.m. stretch: 3pl.ind.aor.med. vāśrā́ abhijñú yā́ tave roaring: nom.pl.m. knee-deep run: inf. These sonsi (strike) up praises. In their racings [theyi] have enlarged the goals, roaringi, in order to run knee-deep. ‘These sons (strike) up praises. In their racings they have enlarged the goals, roaring, in order to run knee-deep.’ (cf. Griffith)

9 Cf. ‘THE, Faithful, One, Immortal among mortals, a God among the Gods, appointed envoy, Priest, best at worship, must shine forth in glory.’ (Griffith).

238

Rosemarie Lühr

3.2.2 Comparisons in the nominative case According to Keydana, comparisons in the nominative occur in finite purpose constructions as well as infinitival ones: (27) RV 8,102,7f. agníṃ vo . . . | áchā náptre sáhasvate || Agni: acc.sg.m. you: gen.pl. here descendent: dat.sg.m powerful: dat.sg.m. ayáṃ yathā na ābhúvat this: nom.sg.m. that we: acc.pl. appear: 3sg.subj.pres.act. tvāṣṭā rūpéva tákṣyā Tvaṣṭar: nom.sg.m. shape: acc.pl.n.=like to be formed: acc.pl.n. ‘I [summon] your Agni . . . for the powerful descendant, so that this man enters into us as Tvaṣṭar into the shapes to be formed‘ (cf. Hettrich 1988: 288)

He gives the following examples of rationale clauses. They contain infinitives on -am and -é (2013: 56): (28) RV 9,82,1 punānó vā́ ram páry ety avyáyaṃ | purifying fleece: acc.sg.m./n pass: 3sg.ind.pres.act. sheep’s: acc.sg.m./n. śyenó ná yóniṃ ghr ̥távantam āsádam hawk: nom.sg.m. like womb: acc.sg.m. containing ghee: acc.sg.m. seat: inf. ‘Purifying he passes through the sheep’s fleece like a hawk to seat on the womb dropping with ghee’ (29) RV 6,29,3 vásāno átkaṃ surabhíṃ dr ̥śé káṃ | robed: nom.sg.m. garment: acc.sg.m. grossamery: acc.sg.m. look: inf ptcl svàr ṇá nr ̥tav iṣiró sun: nom.sg.m. like danser: voc.sg.m. vivacious: nom.sg.m. babhūtha become: 2sg.ind.pf.ac. ‘Robed in a grossamery garment to look on like the sun, you danser, you have become vivacious’

But the comparions can also refer to the subject of the matrix clause.

The emergence of expressions for purpose relations

239

(28) a. Purifyingi hei passes through the sheep’s fleece [like a hawk]i [ei to seat on the womb dropping with ghee] (29) a. [Robed in a grossamery garment]i [ei to look on like the sun], you danser, youi have become vivaciousi But while in sentences with rationale clauses the subject of the matrix clause, the covert subject of the infinitive construction and the comparison all exhibit nominative case, in a purpose clause the case does not match that of the controller in the matrix clause, i.e. it refers to a non-nominative antecedent (cf. Keydana 2013: 56f., 227). (30) RV 1,25,17 sáṃ nú vocāvahai púnar yáto me together ptcl speak: 1pl.subj.aor.med. again as soon I: dat.sg. mádhv ā́ bhr ̥tam | hóteva mead: nom.sg.n. brought: nom.sg.n. sacrificer: nom.sg.m.=like kṡádaṣe priyám eat: inf. dear: acc.sg.n. Once more together let us speak, because the mead is brought to mei [ei to eat the dear like the priesti] ‘Once more together let us speak, because the mead is brought to me that I eat the dear like the priest’

However, hótā iva ‘like a sacrificer’ may be formulaic, or perhaps the nominative case represents the default case,10 as in German comparisons;11 cf. the similar structure in (31): (31) RV 8,12,33 suvīŕ yaṃ sváśvyaṃ sugávyam brave men: acc.sg.n. good steeds: acc.sg.n. good cows: acc.sg.n. indra daddhi naḥ | hóteva Indra: voc.sg.m. give: 2sg.imp.pres.act. us: dat.sg. sacrificer: nom.sg.m.= like 10 Cf. kṛṣṇó rūpáṁ kṛtvā́ (TS.) ‘taking on a black form (i.e. making shape for himself as one that is black)’ (Whiteny 1896). 11 In German the nominative in comparison constructions can be assigned by default, as the nominative is unmarked with respect to all other cases (Hudson 1998; Sigurđsson 1991: 338; McFadden & Sundaresan 2011).

240

Rosemarie Lühr

pūrvácittaye prā́ dhvaré for the first thinking: dat.sg.f. pfx=sacrifice: loc.sg.m. Indra, give usk wealth in brave men, good steeds and good cows [for earb first thinking [of usk] like the priestk] ‘Indra, give us wealth in brave men, good steeds and good cows, that we may be remembered first like the priest at the sacrifice.’

pūrvácittaye ‘at the first thinking (of us)’ is an event nominalization with null elements as in purpose clauses. It comprises an arbitrary null subject and a null object controlled by the dative ‘us’ in the matrix clause, with the comparison in the nominative hótā iva referring to this object. If Keydana’s explanation for hótā iva in (30) were true, the accusative hótaram iva would have been expected. 3.2.3 Overt subjects Finite purpose clauses, rather than infinitives, are attested in sentences with the following characteristics: The finite subordinate clause contains an overt expression for the subject in the nominative. This construction is obligatory, for in infinitive constructions there is no overt subject in the nominative (cf. Keydana 2013: 57). (32) RV 5,6,4 ā́ te agna idhīmahi dyumántaṃ pfx your Agni: voc.sg.m. kindle: 1pl.opt.pres.act. resplendent: acc.sg.m. devājáram unfading: acc.sg.m yád dha syā́ te pánīyasī samíd that ptcl this: nom.sg.f. your very glorious: nom.sg.f. fuel: nom.sg.f. dīdáyati dyávi may shine: 3sg.pres.subj. heaven: loc.sg.m./f. ‘God, Agni, we will kindle your resplendent, unfading (fire), so that this glorious fuel may send forth by day its light for thee’ (cf. Hettrich 1988: 386)

The noun phrase syā́ te pánīyasī samíd ‘this your glorious fuel’ anaphorically refers to te dyumántaṃ devājáram ‘your resplendent unfading (fire)’, where the nominative expression in the subordinate clause is a nominal substitution of

The emergence of expressions for purpose relations

241

an expression in the superordinate structure. Thus, binding principle A12 in the version of Pollard & Sag (1992) applies: (33) An anaphor must be coindexed with a less oblique coargument if there is one. Another example is (34). Here, referential śakráḥ is coreferent with the preceding dative NP índrāya: (34) RV 1,10,5 ukthám índrāya śáṁsyaṁ said: nom.sg.n. Indra: dat.sg.m. laud: nom.sg.n. várdhanam puruniṣṣídhe strengthening: nom.sg.n. living many gifts: dat.sg.m śakró yáthā sutéṣu ṇo mighty: nom.sg.m. that pressed Soma drink: loc.pl.m. our rāráṇat sakhyéṣu ca take pleasure: 3sg.subj.pres.act. friendship: loc.pl.n. and ‘To Indra must a laud be said, strengthening him who gives many gifts, that the mighty one may take pleasure in our friendship and drink-offerings.’ (cf. Hettrich 1988: 279)

And with an overt subject in the subordinate purpose sentence, too: (35) RV 1,89, 5 tám īś́ ānaṃ jágatas tasthúṣas he: acc.sg.m. mighty: acc.sg.m. going: gen.sg.n. standing: gen.sg.n. pátiṃ dhiyaṃjinvám ávase lord: acc.sg.m. exciting devotion: acc.sg.m. favour: dat.sg.n. hūmahe vayám | pūṣā́ no yáthā invoke: 1pl.ind.pres.med. we: nom.pl. Pūṣan: nom.sg.m. our that védasām ásad vr ̥dhé wealth: gen.pl.n. be: 3sg.subj.pres.act. prosperity: dat.sg.f. rakṣitā́ pāyúr ádabdhaḥ svastáye keeper: nom.sg.m. guard: nom.sg.m. infallible: nom.sg.m. fortune: dat.sg.f. ‘Him we invoke for aid who reigns supreme, the Lord of all that stands or moves, inspirer of the soul, that Pūṣan may promote the increase of our wealth, our keeper and our guard infallible for our good’ (cf. Hettrich 1988: 283) 12 Cf. Keydana (2013: 147 note 123); Fanselow & Felix (1990).

242

Rosemarie Lühr

3.2.4 Number of words As Keydana (2013: 151f.) states, completely parallel examples, one with an infinitive on ‑ave, and one with a finite purpose sentence, are provided by (36) and (37): (36) RV 1,134,3 vāyúr yuṅkte róhitā vāyúr Vāyu: nom.sg.m. yoke: 3sg.ind.pres.med. red: acc.dual.m. Vāyu: nom.sg.m aruṇā́ reddish brown: acc.dual.m. vāyū́ ráthe ajirā́ dhurí Vāyu: nom.sg.m. chariot: loc.sg.m. swift-footed: acc.dual.m. pole: loc.sg.f. vóḷhave váhiṣṭhā dhurí vóḷhave | draw: inf. best: acc.dual.m. pole: loc.sg.f. draw: inf. ‘Two red steeds Vāyu yokes, Vāyu two purple steeds, Vāyu the swift-footed, to the chariot, to the pole to draw, most able, at the pole, to draw.’ (37) RV 3,35,2 úpājirā́

puruhūtā́ ya sáptī towards=swift: acc.dual.m. much invoked: dat.sg.m. joined: acc.dual.m. hárī ráthasya dhūrṣv ā́ dun: acc.dual.m. chariot: gen.sg.m. pole: loc.pl.m. pfx

yunajmi / dravád yáthā sámbhr ̥taṃ harness: 1sg.ind.pres.act quickly that completely arranged: acc.sg.m. viśvátaś cid / úpemáṃ yajñám ā́ from all sides ptcl towards=this: acc.sg.m. sacrifice: acc.sg.m. pfx índram vahāta13 carry: 3dual.subj.pres.act. Indra: acc.sg.m. ‘For him, who is invoked by many, the two swift bay steeds to the pole I harness, that they in fleet course may bring Indra hither, to this sacrifice arranged completely.’

In both constructions, an agent is harnessing horses, and the yáthā-sentence and the infinitive phrase denote the goal of the undertaking, namely that the horses shall carry the chariot or Indra in the chariot (Keydana 2013: 151). But there is 13 vahāta is unstressed (Hettrich 1988: 288 note 97).

The emergence of expressions for purpose relations

243

a fundamental difference between the purpose structures. The finite form has many more words than the infinitive one: dravád yáthā sámbhr ̥taṃ viśvátaś cid / úpemáṃ yajñám ā́ vahāta índram consists of five constituents. Moreover, the structure sámbhr ̥taṃ . . . úpemáṃ yajñám is a hyperbaton, a stylistic phenomenon which is due to its sequential information structure. In the competition of syntactic structures word number may not be neglected as can also be seen with causal hí- and yád-sentences in Vedic. The main sentence-like hí-clauses are mostly longer than the subordinate yád-clauses (cf. Lühr 2015). 3.2.5 Information structure Information structure may also be responsible for the choice of the finite structure in (38): (38) RV 1,138,2 prá hí tvā pūṣann ajiráṃ ná yā́ mani pfx ptcl you Pūṣan: voc.sg.m. swift: acc.sg.m. like way: loc.sg.n. stómebhiḥ krṇ̥ vá r ̥ṇávo laud: instr.pl.m. urge: 1.sg.ind.pres.med. make get going: 2sg.subj.pres.act. yathā mr ̥dhaḥ that contemner: acc.pl.m. ‘For I urge you, Pūṣan, like a swift one on his way, that you make the contemners get going.’

Insted of a possible purpose clause the poet used a finite structure. However, this structure allows for positioning the most stressed word of the new information focus in the finite sentence, mr ̥dhaḥ ‘contemner’, behind the conjunction at the end of the sentence, which is a special focus position.14 In describing the similarities between finite and infinitive purposive constructions Keydana (2013: 147, Note 124), and also elsewhere, does not consider such features relevant to stylistics or information structure. Thus, he also ignores the fact that the poet used an overt pronoun in the nominative in the subordinate clause in (39). He is, therefore, wrong in assuming that RV 10,159,6 is a counterexample to the competition between finite and infinitival purpose structures. The pronoun is highlighted and therefore bears stress. It functions as contrast focus in the triumph song of a woman: 14 For an older opinion cf. Lühr (1994: 214ff.).

244

Rosemarie Lühr

(39) RV 10,159,6 sám ajaiṣam imā́ aháṁ sapátnīr pfx subdue: 1sg.ind.aor.act. this: acc.pl.m. I: nom.sg. corrival: acc.pl.f abhibhū́ varī yáthāhám asyá vīrásya superior: nom.sg.f. that=I: nom.sg. this: gen.sg.m. man: gen.sg.m. virā́ jāni jánasya ca dominating: nom.sg.f. people: gen.sg.n. and ‘I have subdued these corrivals, I the superior, that I hold sway over this man and his people’.

Therefore, the infinitive structure (39’) is not an equivalent: (39’) sám ajaiṣam imā́ aháṁi sapátnīr abhibhū́ varī ei asyá vīrásya jánasya ca jánasya ca Ii have subdued these corrivals, I the superior [ei to sway over this man and his people] ‘I have subdued these corrivals, I the superior, to sway over this man and his people’. (Keydana 2013: 150).

3.2.6 Speech acts Sentences with speech act verbs and a finite subordinate structure require quite another interpretation, cf. (40) and (41): (40) RV 7,28,5 vocéméd índram maghávānam enam call: 1pl.opt.pres.act.=ptcl Indra: acc.sg.m. liberal: acc.sg.m. he: acc.sg.m. mahó rāyó rā́ dhaso yád great: gen.sg.m./f. wealth: gen.sg.m./f. gift: acc.pl.n. that dádan naḥ give: 3sg.subj.pres.act. us: dat.pl. ‘We will call him the liberal Indra, that he may grant us gifts of ample riches.’15

15 According to Keydana (2013: 152), in this example a purpose clause with object control would have been also possible.

The emergence of expressions for purpose relations

245

(41) RV 3,32,14 stávai purā́ pā́ ryād índram praise: 1sg.subj.pres.med. before decisive: abl.sg.n. Indra: acc.sg.m. áhnaḥ | áṃhaso yátra pīpárad yáthā no day: abl.sg.n. trouble: abl.sg.n. where save: 3sg.inj.aor.act. that us: dat.pl. ‘I will praise Indra before the decisive day, upon which / that he shall save us from all trouble’16

The yád/yáthā-clause can be considered as forming part of a speech act sequence consisting of dominant and subsidiary speech acts. Such sequences are complex with respect to their propositional as well as to their illocutionary potential (Lühr 2007: 289). (40) can be interpreted as follows: (40) a. ‘We will call him the liberal Indra, [for I wish] that he may grant us gifts of ample riches’ (cf. Hettrich 1988: 137, 389) The speakers justify the dominant speech act ‘praise’ with the subsidiary speech act ‘wish’. Praise represents a social strategy in that the speaker attempts to create rapport with the addressee by expressing approval. Moreover, it is a means to gain the hearer’s acceptance to fulfill the following wish. Such situations are recurrent in the R̥ gveda (Lühr 2004). Infinitive constructions with a similar content are also used in Vedic. In (42), the appeal to the Aśvins is motivated by the following wish. They shall bring treasure to men and avert misfortune and sickness. (42) RV 8,58,3 táṃ vāṃ huvé átiriktam he: acc.sg.m. you: gen.dual invoke: 1sg.ind.pres.med. extant: acc.sg.m. píbadhyai drink: inf. ‘I invoke your [chariot] that you drink the extant Soma’

Keydana (2013: 152) considers infinitive constructions as in (42) to be equivalent to the finite clauses in (40) and (41) and assumes a purpose clause for such cases. In example (42), the chariot of the Aśvins would be metonymically equivalent to the Aśvins themselves (Keydana 2013: 112). (42’) [táṃ vāṃ = aśvinā]i huvé [ei áti riktam pibadhyai] 16 The finite purpose clause is syntactically irregular, for it contains besides yáthā also yátra as conjunctions (Hettrich 1988: 305f. note 115). Keydana (2013: 150) does not consider that.

246

Rosemarie Lühr

But none of Bach’s (1982: 38) conditions permitting a purpose clause are met. His second condition, which seems to come into question here, is not met in the case of a speech act verb: A change of a state of affairs has to be denoted yielding a state of availability. Therefore, the object of the matrix clause is no “affected object”, which means that it is created, transferred or transformed such that “a change in the state of affairs . . . of a positive sort” occurs (Johnston 1998: 89). Consider furthermore (43) (Keydana 2013: 110f.) and (44) (Gathā Avestan): (43) RV 8,71,15 agníṃ dvéṣo yótavaí no Agni: acc.sg.m. hatred: abl.sg.n. keep away: inf. us: acc.pl. gr ̥ṇīmasy agníṃ śáṃ praise: 1pl.ind.pres.act. Agni: acc.sg.m. blessing: acc.sg.n. yóś ca dā́ tave happiness: acc.sg.n. and give: inf. We praise Agnii [ei to protect us against hatred, ei to give us happiness and blessings] ‘We praise Agni [for we wish] that he protects us against hatred, that he gives us happiness and blessings’ (44) Y 51,10 maibiiō zbaiiā aṣǝm vaŋuiiā I: dat.sg. call: 1sg.ind.pres.act truth: acc.sg.m. good: instr.sg.f. aṣī ga.t ̰ē reward: instr.sg.f. come: inf. I call truthi [ei to come to me with a good reward] ‘I call truth [for I wish] that he comes to me with a good reward.’ (Keydana 2013: 110: note 62)

Those constructions come close to sentences where the infinitive construction may function as an infinitive complement with exceptional case marking. Accordingly, in (45) the accusative vām is externally checked from outside its continaing IP.17 Hence, it is not an obvious instance of a purpose clause, either:

17 For infinitive constructions which can be a manipulative complement or a purpose clause cf. Keydana 2013: 314. Keydana (2013: 278) prefers an alternative solution here: vām would be the object of uśmahi, for dative complements are not documented with VAŚ otherwise.

The emergence of expressions for purpose relations

247

(45) RV 5,74,3 vayáṃ vām uśmasīṣṭáye we: nom.pl. you: acc.dual wish: 1.pl.ind.pres.act.=further: inf. We want [you to further us] ‘We long for you to further us’

To sum up so far: After the discussion of Keydana‘s counter examples against the assumed complementary distribution of finite and non-finite purposive constructions, it is doubtful whether his claim is right that finite purpose structures compete with infinitival purpose structures whenever a rationale or a purpose clause cannot be used. The crucial factors were, in the case of finite purpose structures, the number of words, aspects of information structure, overt subjects and adjectives in the nominative; in the case of infinitive constructions, the use of speech act verbs in the matrix sentence hardly allowing for purpose clauses. Furthermore, it was stated that the missing infinitive of as- ‘to be’ could have been substituted by bhuvé ‘to become, to be’ if the author of a Vedic hymn would have liked to provide an infinitive construction with an attributive nominative and if such a nominative would have been allowed in such constructions, but with an attributive nominative only finite purpose structures with as- appear. Finally, the very seldom use of such a nominative in rationale clauses is striking, while it is often documented in finite purpose sentences, and with a comparison the case nominative appears only once in a purpose clause. However, in rationale clauses the nominative could be an apposition to the subject in the superordinate structure and the nominative in a comparison in a purpose clause possibly is the default case. Leaving the nominatives aside, most of the alternating constructions clearly differ in their pragmatic functionality. However, the choice of finite purpose structures with overt subjects is obligatory and a matter of syntax.

3.3 Purpose infinitives With the infinitive on ‑dhyai and other infinitives different kinds of control are documented. 3.3.1 Subject control Subject control infinitives are often controlled by a covert subject in the first person singular. The infinitive construction is a rationale clause:

248

Rosemarie Lühr

(46) RV 1,61,3 asmā́

íd u tyam upamáṃ svarṣā́ m he: dat.sg.m. ptcl ptcl this: acc.sg.m. highest: acc.sg.m. sun winning: acc.sg.m.

bhárāmy āṅgūṣám āsyèna offer: 1sg.ind.pres.act. song of praise: acc.sg.m. mouth: instr.s.g.n. máṃhiṣṭham áchoktibhir matīnā́ ṃ most generous: acc.sg.m. invocation: instr.pl.f. devotional hymn: gen.pl.f. suvr ̥ktíbhiḥ sūríṃ vāvrd̥ hádhyai hymn: instr.pl.f. the glorious: acc.sg.m. magnify: inf. To him then Ii offer this highest sun winning song of praise [ei to magnify with songs of invocation and with hymns the glorious] ‘To him then I offer this highest sun winning song of praise, to magnify with songs of invocation and with hymns the glorious.’

The controller is the intentional subject of the matrix sentence and the matrix verb denotes an atelic activity (cf. Keydana 2013: 52, 61, 119). Subject control is also found in cases where the main clause is a request. The mood in the main clause is the imperative: (47) RV 4,16,2 śūrā́ dhvano nā́ nte áva 18 sya pfx this: voc.sg.m. hero: voc.sg.m.=journey: gen.sg.m. as=end: loc.sg.m. ’smín no adyá sávane mandádhyai this: lok.sg.m. of us today libation: loc.sg.n. delight: inf. Unyoke, o heroi, as at this journey’s end [ei to delight today in our Soma sacrifice] ‘Unyoke, o hero, as at this journey’s end, to delight today in our Soma sacrifice!’

3.3.2 Object control An example of object control and, therefore, of a purpose clause is provided by (48):

18 With verbal ellipsis of the verb sā-, to unyoke’.

The emergence of expressions for purpose relations

249

(48) RV 4,22,7 átrā́ ha te harivas tā́ u then=ptcl of you having tawny coursers: voc.sg.m. this: nom.pl.f. ptcl devīŕ ávobhir indra stavanta goddess: nom.pl.f. help: instr.pl.n. Indra: voc.sg.m praise: 3pl.inj.pres.med. svā́ sāraḥ yát sīm ánu prá mucó sister: nom.pl.f. when ptcl pfx pfx release: 2sg.inj.aor.act. badbadhānā́ dīrghā́ m ánu prásitiṃ syandayádhyai 19 prisoned: acc.pl.f. long: acc.sg.f. pfx duration: acc.sg.f. flow: inf. Then, o Indra, lord of tawny coursers, these sistersi, goddesses, are praised, when you released the prisoned onesi with your help [ei to flow after a long time (i.e. captivity)] ‘Then, o Indra, lord of tawny coursers, these sisters, goddesses, are praised, when you released the prisoned ones with your help, to flow after a long time (i.e. captivity).’

The subject of the infinitive construction is identified on the basis of semantic information and world knowledge. From mythology it is well known that the Indo-Aryan deity Indra defeats the huge serpent Vr̥tra and releases the waters fenced in by this dragon. Therefore, the infinitive ánu syandayádhyai ‘to flow’ can only refer to the object of the matrix clause, the waters, referred to here as ‘sisters’. A special use of object control exists if the subject of the infinitive construction is derived from a possessive pronoun of the matrix clause; cf. with reference to asya ‘his’:20 (49) RV 4,29,3 śrāváyéd asya kárṇā vājayádhyai make hear: 2sg.imp.pres.act. his ear: acc.dual.m. show vigor: inf. júṣṭām ánu prá díśam mandayádhyai habitual: acc.sg.f pfx pfx direction: acc.sg.f. to get excited: inf. Make hisi ears hear [ei to show his vigor and (steer him) in the habitual direction ei to get excited] ‘Make his ears hear, to show his vigor and (steer him) in the habitual direction to get excited.’

19 For tmesis with forms on –dhyai cf. Keydana (2013: 180) (against Benveniste 1935: 99). 20 Cf. Keydana (2013: 109).

250

Rosemarie Lühr

In (50), the controller is the beneficient nah ‘us’, the gap in the infinitive construction is the accusative object patháś. Keydana (2013: 113) convincingly compares such examples to English sentences like (7c)(iv): (7)

c.

(iv)

Ii used itj [ei to slice the salami with ej].

(50) RV 4,37,7 ví no vājā r ̥bhukṣaṇaḥ patháś pfx we: dat.pl. Vāja: voc.pl.m. Ṛbhukṣan: voc.pl.m. path: acc.pl.m. citana yáṣṭave | explore: 2pl.imp.aor.act. sacrifice: inf. asmábhyaṃ sūraya stutā́ víśvā we: dat.pl. master: voc.pl.m. lauded: nom.pl.m. all: acc.pl.f. ā́ śās tarīṣáṇi direction: acc.pl.f. press foreward: inf. O Vājas and Ṛbhukṣans, explore the paths to sacrifice for usi, masters, lauded [ei to press forward to each direction] ‘O Vājas and Ṛbhukṣans, explore the paths to sacrifice for us, masters, lauded, that we may press forward in each direction.’

3.3.3 Arbitrary control In the purpose clause in (51), arbitrary control could be assumed if the Hittite example (18a) were comparable to it: (18) a. one jug of winek [earb to libate ek] (51) RV 3,32,2 sékteva kóśaṃ sisice pourer: nom.sg.m. like vessel: acc.sg.m. pour out: 1sg.ind.perf.med. píbadhyai drink: inf. ‘Like a pourer I have poured out the vessel for drinking.’

But the context shows that the subject of the infinitive construction is Indra. (51) a. Indrai . . . Like a pourer I have poured out the vessel [ei to drink] This shows that in Vedic the null subject of a purpose clause can also be inferred from a preceding sentence.

The emergence of expressions for purpose relations

251

3.4 Constructions with double datives In Old Indic, constructions with double datives consisting of a noun and an infinitive are also documented. Gonda (1962: 145–150) assumes a combination of a dative of reference with a dative of purpose, other explanations are based on case attraction. By contrast, Keydana (2013: 133f.) assumes a purpose clause. A familiar construction is áhaye hántavā́ : (52) RV 8,96,5 ā́ yád vájram bāhvór indra pfx when club: acc.sg.m. arm: loc.dual.m. Indra: voc.sg.m. dhátse madacyútam áhaye lay: 2sg.ind.pres.med. reeling with excitement: acc.sg.m. snake: dat.sg.m. hántavā́ u slay: inf. ptcl ‘When you, Indra, take the club reeling with excitement in the arms to slay the snake’

The infinitive construction would contain a null object whose reference would be identical with that of the dative áhaye in the matrix clause (Keydana 2013: 128; but cf. 170). For the function of such an adjunct dative he suggests that this case denotes “Ding (im logischen Sinne), zu dem das vom Satz bezeichnete Ereignis in eine Relation gestellt wird . . .” (131). (52) When youi, Indra, take the club reeling with excitement in the arms for the snakek [ei to slay ek] Another suggestion was made by Lühr (1997): As predicative infinitives exist in copular sentences, she proposes a transfer to structures which could be understood as attributive constructions: (53) RV 5,62,9 yád báṃhiṣṭhaṃ nā́ tivídhe which: nom.sg.n. strongest: nom.sg.n. not= penetrate: inf. sudānū áchidraṃ śárma bounteous: voc.dual.m. undestroyable: nom.sg.n. shelter: nom.sg.n. bhuvanasya gopā world: gen.sg.n. shepherd: voc.dual.m. ‘Which shelter is strongest, not to be penetrated, undestroyable, bounteous gods, shepherds of the world’21 21 Gippert (1978: 89). According to Keydana (2013: 126, 156 note 142), this structure is likely to be predicative.

252

Rosemarie Lühr

(54) is an example with reference to an accusative: (54) RV 9,102,6 yám ī gárbham r ̥tā́ vr ̥dho who: acc.sg.m. as body fruit: acc.sg.m. increasing truth: nom.pl.m. drś̥ é cā́ rum ájījanan see: inf. lovely: acc.sg.m. generate: 3pl.ind.aor.act. ‘the babe whom they who strengthen law have generated fair to see’

The next step is the connection of a dative of purpose with such an attributive infinitive. Hettrich (1984 passim) calls this dative Patiensdativ. The result is the construction für die zu erschlagende Schlange ‘for the snake being slain’. Arguing against this proposal, Keydana (2013: 126) first notes that an appositive infinitive would also have been used with cases other than the dative. However, as Lühr considers the whole construction to be one of purpose, and whereas only the dative functions as a case of purpose, no other case is in line with that. The second objection is even more serious. Keydana (133) gives an example where an appositive accusative in an infinitive construction is coreferent with a noun in the dative in the superordinate structure: (55) RV 9,61,22 yá ā́ vithéndraṃ vrt̥ rā́ ya who: nom.sg.m. help: 2sg.ind.perf.act.=Indra: acc.sg.m. Vr̥tra: dat.sg.m. hántave | vavrivā́ ṃsam mahīŕ apáḥ slaughter: inf. compass: part.perf.act.acc.sg.m. large: acc.pl.f. water: acc.pl.f. ‘thou who has helped Indra to slaughter Vr̥tra who encompassed the mighty floods.’ yá ā́ vithéndraṃj vrt̥ rā́ yak ej hántave [DPek vavrivā́ ṃsam mahīŕ apáḥ] ,der du den Indra gegen Vr̥tra unterstützt hast, damit er ihn töte, der [die großen Wasser] eingeschlossen hielt.‘ (Keydana 2007)

The theme Indra of the relative clause would control the subject of the adjunct infinitive phrase, with the verb han- ‘to slay’ providing the null object with the object case accusative (the infinitive phrase could either be a purpose clause or a complement clause dependent on the verb av- AV ‘to help’). Appositive vavri-ạ̄́ ṃsam would then agree with the null object. An objection can be raised against this suggestion, too. In this regard Williams’ second and fifth controll properties play a role:

The emergence of expressions for purpose relations

253

(3) 2. The antecedent precedes the controlled PRO. 5. There must be an antecedent. Actually, there are sentences such as (56) showing a different word order: (56) RV 4,32,9 abhí tvā gótamā girā́ nūṣata pfx you: acc.sg. Gotama: nom.pl.m. song of praise: instr.sg.f.= prá dāváne | índra shout towards: 3pl.ind.aor.med. pfx give: inf. Indra: voc.sg. vā́ jāya ghŕ ̥ṣvaye benefit: dat.sg.m. pleasing: dat.sg.m. ‘The Gotamas have sung their song of praise to you that you may give, Indra, expected benefit.’

The dative vā́ jāya ghŕ ̥ṣvaye, the presumed controller of the null object of the infinitive construction, follows this construction: (56) a. The Gotamas have sung their song of praise to youi [ei to give ek] for the expected benefitk Similar for dr ̥śáye sū́ ryāya: (57) RV 10,14,12 tā́ v asmábhyaṃ dr ̦śáye sū́ ryāya púnar he: nom.dual.m. us: dat.pl. see: inf. sun: dat.sg.f. again dātām ásum adyéhá bhadrám || give: 3dual.imp.aor.act vigor: acc.sg.m. today=here auspicious: acc.sg.m. ‘May they (Yama’s envoys) restore to us a fair existence here and to-day, that we may see the sunlight.’ (Lühr 1997: 165f.)

(57) a. May they give to usi [ei to see ek] for the sunk a fair existence here and to-day Though Williams’ precedence requirement is not generally accepted today (cf. Stiebels 2007), the assumption of a topicalization of the infinitives prá dāváne or dr ̥śáye in front of the controller of the null object in the infinitive construction remains problematic. By contrast, as an attribute can appear before as well as behind its reference word in Old Indic, the assumption of a connection of a dative of purpose with an attributive infinitive seems more obvious. But if Keydana’s interpretation of the

254

Rosemarie Lühr

infinitive construction in (52) as a purpose clause cannot be maintained, an explanation for the use of the accusative vavrivā́ ṃsam is necessary. Here, it must be remembered that the accusative vavrivā́ ṃsam can be found three times in the R̥ gveda, also with reference to Vr̥tra: (58) a. RV 6,20,2 áhiṃ yád vr ̥trám apó dragon: acc.sg.m. when Vr̥tra: acc.sg.m. water: acc.pl.f. vavrivā́ ṃsaṃ hánn having enclosed: acc.sg.m. smite: 3sg.impf.act. ‘when you (Indra) slew Vr̥tra, the dragon who enclosed the waters’ b. RV 4,16,7 apó vr ̥tráṃ vavrivā́ ṃsaṃ water: acc.pl.v. Vr̥tra: acc.sg.m. having enclosed: acc.sg.m. párāhan smite away: 3sg.impf.act. ‘He smote away Vr̥tra, who enclosed the waters’ c. RV 2,14,1 yó apó vavrivā́ ṃsaṃ vtráṃ who: nom.sg.m. water: acc.pl.f. having enclosed: acc.sg.m. jaghā́ na slay: 3sg.pf.ind.act. ‘who has slain Vr̥tra who enclosed the waters’

Therefore, it is conceivable that vr ̥tráṃ vavrivā́ ṃsaṃ has become a stereotyped phrase which triggered an association with an attributive accusative vavrivā́ ṃsaṃ to yield the construction vr ̥trā́ ya hántave vavrivā́ ṃsaṃ.

3.5 Complementary distribution of infinitive purpose structures Passing now from the syntax and semantics of purpose constructions to the original functions of the infinitives it must be kept in mind that in the oldest Indo-European languages infinitives normally stem from verbal nouns. In Hittite, the infinitive in -anna is derived from the allative of the heteroclitic actions nouns in -ātar and verbs with verbal substantive in -war form their infinitive in -wanzi,

The emergence of expressions for purpose relations

255

originally the ablative-instrumental of the verbal substantive in -war (Hoffner and Melchert 2008: 76, 130), whereas in Old Indic the infinitival verbal nouns mostly have dative endings, but also accusative ones, as later in Sanskrit.22 However, there seems to be one exception, the infinitive on -dhyai. Here, it is essential that with this infinitive no overt expression of a subject or agent appears. Hence, Keydana’s (2013: 136–141) example RV 7,92,2 (Der flinke Presser ist bei den Opfern vorgetreten, damit Indra und Vāyu den Soma trinken), presumably containing an infinitive on -dhyai with an overt dative subject (índrāya vāyáve) and topicalization of the accusative object sóman, must be assessed in a different way: (59) a. RV 7,92,2 prá sótā jīró adhvaréṣv pfx presser: nom.sg.m. swift: nom.sg.m. sacrifice: loc.pl.m. asthāt sómam índrāya come forth: 3sg.ind.aor.act. Soma: acc.sg.m. Indra: dat.sg.m. vāyáve píbadhyai Vāyu: dat.sg.m. drink: inf. ‘Prompt at the holy rites the presser came forth (to press) the Soma for Indra and Vāyu to drink’

The sentence contains two infinitive constructions, a rationale clause and a purpose clause. The rationale clause exhibits an elliptical verb (sótave ‘to press’) being inferable from the agent noun sótar- ‘presser’, which is also the controller of the subject of this infinitive.23 By contrast, the controller of the subject of the purpose clause is the beneficient índrāya vāyáve and the empty accusative object refers to the accusative sóman of the matrix clause.24 (59) b. prá sótāi . . . asthāt [ei (sótave) sómamk índrāyaj vāyávej [ej píbadhyai ek] Thus, as there is no overt subject with infinitives on -dhyai, the author of a R̥ gvedian hymn had to choose another non-finite structure if he wanted to produce an infinitive purpose phrase with an overt subject, for example an infinitive on -é, which derives from a dative form (Keydana 2013: 222ff.). 22 For infinitives in the ablative and genitive cf. Keydana (2013: 76f.) 23 For ellipsis in Old Indic cf. Delbrück (1900: 122–27); Gonda (1960); Zeilfelder (2000). 24 Keydana’s (2013: 137) second example of an overt dative subject in an infinitive construction on -dhyai is not convincing (RV 1,183,3 = 6,49,5). Cf. Geldner’s translation.

256

Rosemarie Lühr

(60) RV 1,50,1 úd u tyáṃ jātávedasaṃ pfx ptcl this: acc.sg.m. Jātavedas: acc.sg.m. deváṃ vahanti ketávaḥ god: acc.sg.m bear up aloft: 3pl.ind.pres.act. ray: nom.pl.m. dr ̥śé víśvāya sū́ ryam look: inf. all: dat.sg.m. son god: acc.sg.m. The rays bear Jātavedask up aloft, the god for alli [in order ei to look on ek] ‘The rays bear Jātavedas up aloft, the god, that all may look on him.’

The overt subject of the infinitive clause is the dative víśvāya ‘everybody, the whole world’ (cf. also Lühr 1997; Keydena 2013: 210). In other words, the infinitive on -é and the infinitive on -dhyai show complementary distribution in this respect. The fact that with infinitives on -dhyai an overt expression of a subject or agent never occurs must be taken into consideration when discussing the possible pre-form.

3.6 The reason for a missing subject with the infinitive on -dhyai As for the history of the element -dhyai, Rix (1976) connects the Indo-Iranian infinitives on *-dhi ̭āi ̭ (Vedic -dhyai, Avestan -diiāi/-δiiāi) with the Sabellic passive infinitive ending /-fẹː/ (Umbrian –f(e)i, Oscan -fír with added mediopassive -r) and derived the Indo-Iranian and the Sabellic endings from the same pre-form *-dhi ̭ōi. However, García Ramón (1993), assuming that *-dhi ̭ōi ̭ would have given +-fiūi in Sabellic rather than /‑fẹː/, took an instrumental *‑dhi ̭eh1 as the basis, separating the Indo-Iranian and the Sabellic endings from each other. Examples of the use of the instrumental as an infinitive would be Vedic ūtā́ ‘with the help’ and svastí ‘with good luck’, occurring in the same contexts as the datives ávase or ūtáye ‘for help’ and svastáye ‘for good luck’. But as Fortson (2013: 50) rightly says, the dative is the most common case taken by verbal nouns functioning as infinitives. Fortson himself checks some other derivations of Indo-Iranian *-dhi ̭āi ̭ found in the literature, for example the old connection of the Greek mediopassive -σθαι, but he does not consider “the ultimate source of *-dhi ̭o- to be accessible under current knowledge.” (Fortson 2013: 57).

The emergence of expressions for purpose relations

257

However, forms on -dhyai often function as finite verbs,25 cf. (61). The first person dominates by far in this case (Delbrück 1888: 412). (61) RV 5,45,4 sūktébhir vo vácobhir well-chosen hymn: instr.pl.n. you: acc.dual. word: instr.pl.n. devájuṣṭair índrā nv àgnī ́ god-loved: instr.pl.n. Indra: voc.dual.m. ptcl Agni: voc.dual.m. ávase huvádhyai favor: dat.sg.n. invoke: inf. /1sg.subj.pres.med. ‘With well-chosen hymns and God-loved words I will invoke you, Indra and Agni, for your favor.’ (Lühr 1994a: 80)

In Avestan, even personal pronouns are found with a verb form on -diiāi: (62) Y 43,14 uzǝrǝdiiāi azǝ̄ m sarǝdānā̊ sǝ̄ ṇghahiiā win: 1sg.subj.pres.med. I: nom.sg. strength: acc.pl.f. annunciation: gen.sg.m. ‘I will draw force form your annunciation.’ (Lühr 1994a: 90)

This shows that -ai / -āi in -dhyai / -diiāi originally was the ending of a first singular subjunctive medium denoting a wish of the speaker.26 This ending must have been petrified in the form -ai / -āi; for this ending cf. the primary ending Old Persian maniyaiy (from man- ‘to think’). In Avestan, only primary endings 25 Keydana (2013: 172) uses the term „Matrixinfinitiv“. 26 Nevertheless, Keydana (2013: 173, 178) assumes another starting point: “Es ist also nicht unwahrscheinlich, dass der Matrixinfinitiv seinen Ursprung in Überblendungen von InfPs und finiten Sätzen mit kovertem Subjekt hat. Die Verteilung der Matrixinfinitive im RV legt nahe, dass es sich um Sätze mit Subjekt in der ersten Person gehandelt haben mag. Die Bevorzugung der ersten Person kann aber auch eine spätere Entwicklung sein, denn auch der Imperativ der zweiten Person ist ein naheliegender Ausgangspunkt für eine Überblendung.” [It is not unlikely that the matrix infinitive takes its origin in blendings of InfPs and finite sentences with covert subject. The distribution of the matrix infinitives in the RV suggests that it may have been sentences with subject in the first person. But the preference of the first person may also be a later development because also the imperative of the second person is an obvious starting point for blending.] Keydana sees the “Matrixinfinitiv” on -dhyai as an archaism (181), but the adjunct use would be oldest, for ‑dhyai would trace back to an oblique case form (183). An old optative has to be excluded, as in this case the negation ná should have been expected (179). But in the relevant infinitive constructions no negation is documented.

258

Rosemarie Lühr

are used in the subjunctive present (Hoffmann & Forssman 1996: 194f.). Concerning the medium, the starting point must be denotations of complex cognitive events, one of Kemmer’s (1993) middle voice categories; cf. Greek βούλομαι ‘wish, want’, oἴομαι ,‘believe’, Latin obliviscor ‘forget’. However, in specific contexts finite verbs with -dhyai / ‑diiāi were reanalyzed as infinitives. These contexts are sentences like (63), where a main clause with a verb in the first person singular precedes the structure with the form on -dhyai: (63) a. RV 8,39,1 agním astoṣy r ̥gmíyam Agni: acc.sg.m. praise: 1sg.ind.aor.med. glorious: acc.sg.m. agním īḷā́ yajádhyai Agni: acc.sg.m. ghee: instr.sg.f. I will worship: 1sg.subj.pres.med./inf. ‘I have the glorious Agni praised, I will worship him with ghee.’27

The phrase r ̥gmíyam agním īḷā́ yajádhyai can also be interpreted as a non-finite purpose construction with subject control, i.e. as a rationale clause. The result is (63b) (Lühr 1994a: 82; Keydana 2013: 108): (63) b. RV 8,39,1 agním astoṣyi r ̥gmíyam ei agním īḷā́ yajádhyai Ii praised the glorious Agni [ei to worship Agni with ghee] (Geldner; cf. Sgall 1958: 226)

After reanalysis as an infinitive ending this ending could also be used with other persons than the first singular. To answer the question of why infinitives on -dhyai are not connected with an overt subject expression,28 it has to be supposed that in speaker awareness verb forms on -dhyai were originally finite. This knowledge must have endured over time. Thus, used in purpose contexts, with reference to the first person singular an inherent subject was implied which resulted in subject control. This means that in the case of the infinitive on *-dhi ̭āi ̭ rationale clauses were the first purpose constructions. But being used in that way, in a next step purposes clauses could be built with this infinitive formants as well.

27 Cf. Sgall (1958: 226); Keydana (2013: 108). Cf. Griffith’s translation: ‘THE glorious Agni have I praised, and worshipped with the sacred food.’ 28 For a possible verbal form cf. Lühr (1994a).

The emergence of expressions for purpose relations

259

3.7 A parallel development of an original verbal ending to an infinitive ending The objection that the infinitive ending Indo-Iranian *-dhi ̭āi ̭ cannot be the ending of a first singular subjunctive medium, because there is no such ending known anywhere in Indo-European, must be rejected. A parallel for the change of a verbal ending into an infinitive ending can be seen in the development of the inflected infinitive in Romance languages (Scida 2004: 94ff.). This infinitive is founded on the Latin imperfect subjunctive:29 (64) Latin amārem amārēs amāret amārēmus amārētis amārent

Portuguese amar amares amar amarmos amardes amarem

Galician amar amares amar amáremos, amarmos amarédes, amardes amaren

The first parallel to the development of the first person singular Indo-Iranian *-dhi ̭āi ̭ into an infinitive ending in purpose constructions is the fact that the inflected infinitive in Portuguese and Galician often occurs in clauses expressing purpose, generally introduced by Portuguese para, Galician par ‘in order to’. The second parallel is that in Romance in these clauses the uninflected infinitive can also be used: (65) a. Para aquecermos um pouco, in-order-to warm-up: inf.1pl. a bit vamos fazer este pequeno exercício. we-go do this small exercise b. Para aquecer um pouco, in-order-to warm-up: inf. a bit vamos fazer este pequeno exercício. we-go do this small exercise ‘In order to warm up a bit, we are going to do this small exercise.’ (Vanderschueren & Diependaele 2013: 161)30 29 Conjugated infinitives can also have other sources. An example is Welsh where reanalysis of inflected prepositions to infinitives with agreement yielded conjugated infinitives (Miller 2003). 30 For the exact distribution cf. Vanderschueren & De Cuypere (2014); further Vandenschueren (2013); Mensching (2000).

260

Rosemarie Lühr

Hence, purpose constructions are the place where the transition of inflected verbal forms into uninflected ones can occur (see also Dékány this volume for anti-agreeing infinitives in Old Hungarian).

4 Conclusion The emergence of Hittite and Vedic structures denoting purposivity indicated that the oldest Indo-European languages partly overlap and partly differ from each other in their strategies. Hittite as well as Old Indic use infinitive purpose structures, rationale clauses and purpose clauses, allowing for subject, object and arbitrary control, where the infinitive could not only have active, but also passive meaning. But while Hittite displays no finite purpose clauses, Vedic has fully developed finite purpose sentences. Here, complementary distribution between finite purpose und infinitival purpose constructions is necessary if no rationale or purpose clause can be construed. But there are other cases as well where these structures compete with each other. This was shown to depend on the number of words, matters of information structure, overt subjects and adjectives in the nominative. On the other hand, speech act verbs display finite and infinitive constructions, but it is doubtful if these are really purpose constructions, for none of Bach’s conditions for purposes clauses are met. Furthermore, the missing infinitive of as- ‘to be’ could have been substituted by bhuvé ‘to become, to be’ if the author of a Vedic hymn would have liked to provide an infinitive construction with an attributive nominative, but only a great deal of finite purpose structures with as- ‘to be’ and attributive nominatives appear. All in all, attributive nominatives in rationale clauses are documented very seldom. They can be referred to the subject of the matrix clause. Only with a comparison does nominative case appear once in a purpose clause. Leaving this nominative aside, because it may represent the default case used in comparisons, most of the alternating constructions clearly differ in their pragmatic functionality. However, the choice of finite purpose structures, if there is an overt subject, is obligatory and a matter of syntax. The next item was constructions with double datives in Old Indic. While Lühr (1997) explained these structures as the connection of a dative of purpose with an attributive infinitive, Keydana (2013) suggested a purpose clause structure with null object controlled by the adjunct dative. But there are sentences where the infinitive must have been topicalized in front of the dative controller of the null object in the infinitive construction, which is hard to explain. Finally the origin of the infinitive ending Vedic -dhyai was discussed. It was stated that

The emergence of expressions for purpose relations

261

infinitives with this ending emerge from verb forms of the first person singular subjunctive denoting a wish. One can draw a parallel between this development and the use of subjunctives as infinitives in Romance languages. Neveretheless, being used in appropriate contexts, verbs on -dhyai could be reanalyzed as purpose infinitives. In this case, the rationale clause must be the oldest infinitive construction. But like the infinitives built of case forms, they then exhibit subject and object control. To conclude, it can therefore be said that constructions with subject and object control are the oldest devices in Indo-European for the expression of purpose.

Literature Bach, Emmon. 1982. Purpose clauses and control. In Pauline Jacobson & Geoffrey K. Pullum (eds.), The Nature of Syntactic Representation, 35–37. Dordrecht: Reidel. Benveniste, Émile. 1935. Les infinitifs avestiques. Paris: A. Maisonneuve. Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press. Delbrück, Bertold. 1888. Altindische Syntax. Halle: Verlag des Waisenhauses. Delbrück, Bertold. 1900. Vergleichende Syntax der indogermanischen Sprachen, Teil 3. Straßburg: Trübner. Disterheft, Dorothy. 1980. The Syntactic Development of the Infinitive in Indo-European. Columbus, Ohio: Slavica. Fanselow, Gisbert & Felix, Sascha W. 1990. Sprachtheorie. 2: Die Rektions- und Bindungstheorie. Tübingen: Francke. Faraci, Robert A. 1974. Aspects of the Grammar of Infinitives and For-Phrases. Massachusetts: University of Cambridge dissertation. Fintel, Kai von. 2006. Kinds of Modal Meaning http://web.mit.edu/fintel/modality.pdf (accessed 2 March 2015). Fortson, Benjamin. 2013. Pre-Italic *-d h ḙ̄ (*-d h ḙh1 ) versus Pre-Indo-Iranian *-dh ō̭i: Bridging the Gap. In Adam I. Cooper, Jeremy Rau & Michael Weiss (eds.), Multi Nominis Grammaticus: Studies in Classical and Indo-European Linguistics in Honor of Alan J. Nussbaum on the Occasion of his Sixty-Fifth Birthday, 50–60. Ann Arbor/New York: Beech Stave Press. García Ramón, José Luis. 1993. Zur Morphosyntax der passivischen Infinitive im OskischUmbrischen: u. -f(e)i, o. -fír und ursabell. *-fje (*‑d h ḙh1 ). In Helmut Rix (ed.), OskischUmbrisch: Texte und Grammatik, 106–124. Wiesbaden: Reichert. Geldner, Karl Friedrich. 1951. Rig-Veda: Das Heilige Wissen Indiens (Bd. 1–4. Harvard Oriental Series Vol. 33–36). Vollständige Übersetzung, neu hg. von Peter Michel. Wiesbaden: MarixVerlag 2008. Gippert, Jost. 1978. Zur Syntax der infinitivischen Bildungen in den indogermanischen Sprachen (Europäische Hochschulschriften, Reihe XXI: Linguistik 3). Frankfurt am Main: Peter Lang. Gonda, Jan. 1960. Ellipsis, Brachylogy and Other Forms of Brevity in Speech in the R̥gveda. (Verhandelingen / Koninklijke Nederlandse Akademie van Wetenschappen, Afd. Letterkunde; N.R., 67,4). Amsterdam: Noord-Hollandsche Uitg. Maatschappij. Gonda, Jan. 1962. The unity of the Vedic dative. Lingua 11. 141–150.

262

Rosemarie Lühr

Griffith, Ralph T.H. 1889. The Rig Veda (Eulogos 2007) http://www.intratext.com/ixt/ENG0039/_ index.htm (accessed 2 March 2015). Hayakawa, Atsushi. 2014. Circulation of Fire in the Veda (Nijmegen Buddhist and Asian Studies 2). Berlin: Lit Verlag. Hettrich, Heinrich. 1988. Untersuchungen zur Hypotaxe im Vedischen (Untersuchungen zur indogermanischen Sprach- und Kulturwissenschaft 4). Berlin/New York: de Gruyter. Hoffmann, Karl & Forssman, Bernhard 1994. Avestische Laut- und Formenlehre (Innsbrucker Beiträge zur Sprachwissenschaft 84). Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Hoffner, Harry A. & Melchert, H. Craig. 2008. A Grammar of the Hittite Language. Part. I. Reference Grammar (Languages of the Anceint Near East 1/1–2). Winona Lake: Eisenbrauns. Holland, Gary. 2011. Active and passive in Hittite infinitival constructions. In Stephanie W. Jamison, H. Craig Melchert & Brent Vine (eds.), Proceedings of the 22nd Annual UCLA Indo-European Conference. Los Angeles, November 5th and 6th, 2010, 69–81. Bremen: Hempen Verlag. Hudson, Richard. 1998. Functional control with and without structure sharing. In Anna Siewierska & Jae Jung Song (eds.), Case, Typology and Grammar, 151–169. Amsterdam: Benjamins. Huettner, Alison K. 1989. Adjunct Infinitives in English. University of Massachusetts, Amherst dissertation. Johnston, Michael J. R. 1999. A syntax and semantics for purpose adjuncts in HPSG. In Robert D. Levine & Georgia M. Green (eds.), Studies in Contemporary Phrase Structure Grammar, 80–118. Cambridge: Cambridge University Press. Jones, Charles. 1985. Syntax and Thematics of Infinitival Adjuncts. University of Massachusetts. Amherst, MA: GLSA dissertation. Jones, Charles. 1991. Purpose Clauses. Syntax, Thematics, and Semantics of English Purpose Constructions. Dordrecht/Boston/London: Kluwer Academic Publisher. Junghanns, Uwe. 1994. Satz und doch nicht Satz: Über die (doppelte) Einbettung finaler Infinitive des Russischen. In Anita Steube & Gerhild Zybatow (eds.), Zur Satzwertigkeit von Infinitiven und Small Clauses (Linguistische Arbeiten 315), 107–137. Tübingen: Niemeyer. Junghanns, Uwe. 1994a. Syntaktische und semantische Eigenschaften russischer finaler Infinitiveinbettungen. München: Sagner. Kemmer, Suzanne. 1993. The Middle Voice. Amsterdam: John Benjamins. Keydana, Götz. 2007. Latente Objekte und altindische Diskursgrammatik. Paper presented at the Workshop on Pragmatische Kategorien, University Marburg, 24 September 2007. Keydana, Götz. 2013. Infinitive im R̥gveda: Formen, Funktionen, Diachronie (Brill’s studies in Indo-European languages & linguistics 9). Leiden/Boston: Brill. Landau, Idan. 2013. Controll in Generative Grammar: A Research Companion. Cambridge: Cambridge University Press. Lühr, Rosemarie. 1994. Zu Konkurrenzformen von Infinitivkonstruktionen im Indogermanischen: Finale Infinitivkonstruktionen auf –dhyai und finale Adverbialsätze im Altindischen. In George E. Dunkel et al. (eds.): Früh-, Mittel-, Spätindogermanisch. Akten der IX. Fachtagung der Indogermanischen Gesellschaft vom 5. bis 9. Oktober in Zürich, 207–223. Wiesbaden: Reichert. Lühr, Rosemarie. 1994a. Zur Interdependenz der Methoden “Funktionsbestimmung” und “Rekonstruktion” in der Indogermanistik: das Infinitivmorphem indoiran. *-dhyāi. Münchener Studien zur Sprachwissenschaft 55, 69–97.

The emergence of expressions for purpose relations

263

Lühr, Rosemarie. 1997. Zur “Kasusattraktion” in altindischen dativischen Infinitivkonstruktionen. Ein Fall von syntaktischer Analogie? In Sasha Lubotsky (ed.): Sound Law and Analogy. Papers in honor of Robert S.P. Beekes on the Occasion of his 60th Birthday. (Leiden Studies in Indo-European 9), 155–170. Leiden: Brill. Lühr, Rosemarie. 2004. Sprechaktbegründungen im Altindischen. In Thomas Krisch, Thomas Lindner & Ulrich Müller (eds.), Analecta homini universali dicata. Arbeiten zur Indogermanistik, Linguistik, Philologie, Politik, Musik und Dichtung. Festschrift für O. Panagl zum 65. Geburtstag. Bd. 1, 130–144. Stuttgart: Akademischer Verlag. Lühr, Rosemarie. 2007. Information structure in Ancient Greek. In Anita Steube (ed.): The Discourse Potential of Underspecified Structures (Language, Context and Cognition 8), 487–512. Berlin: de Gruyter. Lühr, Rosemarie. 2015. Weil-Sätze in altindogermanischen Sprachen. Paper presented at the 37. DGfS-Tagung, University of Leipzig, 4 March Mensching, Guido. 2000. Infinitive Constructions with Specified Subjects: A Syntactic Analysis of the Romance Languages (Oxford Studies in Comparative Syntax): Oxford: Oxford University Press. Miller, Gary. 2003. Where do conjugated infinitives come from? Diachronica 20. 45–81. McFadden, Thomas & Sundaresan, Sandhya. 2011. Nominative case is independent of finiteness and agreement. http://www.hum.uit.no/a/mcfadden/downloads/bcgl_proc.pdf (accessed 2 March 2015). Nissenbaum, Jon. 2005. States, events and VP structure: Evidence from purposive adjuncts. North East Linguistic Society 47. University of Massachusetts, Amherst, October 28–30, 1–5. Pollard, Carl & Sag, Ivan A. 1992. Anaphors in English and the scope of Binding Theory. Linguistic Inquiry 23.2. 261–303. Rix, Helmut. 1976. Die umbrischen Infinitive auf -fi und die uridg. Infinitivendung *‑dh ōi ̥. In Anna Morpurgo Davies & Wolfgang Meid (eds.), Studies in Greek, Italic, and IndoEuropean Linguistics Offered to Leonard R Palmer on the Occasion of His Seventieth Birthday: June 5, I976, 319–331. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck. Schmidtke-Bode, Karsten. 2009. A Typology of Purpose Clauses (Typological Studies in Language 88). Amsterdam: John Benjamins. Stiebels, Barbara. 2007. Towards a typology of complement control. ZAS Papers in Linguistics 47. 1–80. Sigurđsson, Halldór Árman. 1991. Icelandic case-marked PRO and the licensing of lexical arguments. Natural Language and Linguistic Theory 9. 327–336. Sgall, Peter. 1958. Die Infinitive im R̥gveda. Acta Universitatis Carolina 2. 135–268. Scida, Emily. 2004. The Inflected Infinitive in Romance Languages. New York/London: Routledge. Vanderschueren, Clara. 2013. Infinitivo y sujeto en portugués y español. Un estudio empírico de losinfinitivos adverbiales con sujeto explícito. Berlin: De Gruyter. Vanderschueren, Clara & De Cuypere, Ludovic. 2014. The inflected/non-inflected infinitive alternation in Portuguese adverbial clauses. A corpus analysis. Language Sciences 41 B. 153– 174. Vanderschueren, Clara & Diependaele, Kevin. 2013. The Portuguese inflected infinitive. An empirical approach. Corpus Linguistics and Linguistic Theory, 9. 161–186. Whitney, William Dwight. 1896. A Sanskrit Grammar, Including both the Classical Language, and the Older Dialects, of Vedan and Brahmana. Leipzig: Breitkopf & Härtel.

264

Rosemarie Lühr

Wilder, Chris. 1991. Small clauses and related objetcs. Groninger Arbeiten zur Germanistischen Linguistik 34. 215–236. Williams, Edwin. 1980. Predication. Linguistic Inquiry 11. 203–338. Wurmbrand, Susi. 2002. Syntactic vs. semantic control. In C. Jan-Wouter Zwart & Werner Abraham (eds.), Studies in Comparative Germanic Syntax: Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27, 2000) (Linguistics Today 53), 93–128. Amsterdam: John Benjamins. Zeilfelder, Susanne. 2000. Präverbien ohne Verben im Rigveda. In Bernhard Forssman & Robert Plath (eds.), Indoiranisch, Iranisch und die Indogermanistik. Arbeitstagung der Indogermanischen Gesellschaft vom 2. bis 5. Oktober 1997 in Erlangen, 581–594. Wiesbaden: Reichert Verlag. Zeilfelder, Susanne. 2001. Zum Ausdruck der Finalität im Hethitischen. In Onofrio Carruba & Wolfgang Meid (eds.), Anatolisch und Indogermanisch. Akten des Kolloquiums der Indogermanischen Gesellschaft. Pavia, 22.–25. September 1998, 395–410. Innsbruck: Institut für Sprachen und Literaturen der Universität Innsbruck.

Björn Wiemer

10 Main clause infinitival predicates and their equivalents in Slavic: Why they are not instances of insubordination 1 Infinitives and insubordination It has become commonplace to regard prototypical finiteness as being couched in main (or independent) clauses, whilst non-finiteness (or reduced finiteness) is usually associated with subordinate clauses. Neither the notion of finiteness, nor that of subordination is trivial, but, as Evans (2007: 367) puts it, “the category ‘subordinate clause’, though not without its problems, is nonetheless cross-linguistically more robust than the category ‘nonfinite clause’”. I will adopt the definition of ‘insubordination’, conceptualized by Evans (2007: 367) as the “conventionalized main clause use of what, on prima facie grounds, appear to be formally subordinate clauses“. This concept “is arguably not a fully homogeneous phenomenon” (Malchukov 2013: 181) – of which Evans is aware –, and it involves some premises that have to be remembered (see below). First of all, as this definition shows, insubordination is not so much a matter of non-finiteness (or reduced finiteness), but rather a matter of when clause structures associated with subordinate clauses show up in syntactically independent clauses, i.e. clauses whose predicates are not headed by any other predicate (see Section 2). As a consequence, insubordination comprises not Acknowledgments: I am obliged to Veronika Kampf for a commented list of complementizers and CTP-groups in modern Bulgarian, from which I have profited (see 3.3), as well as for her consultations on Bulgarian data. Moreover, I thank Eleni Bužarovska and Liljana Mitkovska for their feedback on the Macedonian data as well as Dmitri Sitchinava for his support with the interpretation of examples from early East Slavic and Rainer Goldt for helping locate a quotation from Puškin’s. Furthermore, I am grateful to Markus Giger for concise consultations on modern Czech (see 3.1.3.2) and to Jarmila Panevová, Karolína Skwarska and Bohumil Vykypěl for their native judgments on Czech, as well as to a couple of Polish colleagues who “lent” me their native speaker knowledge: Magdalena Danielewiczowa, Rafał Górski, Maciej Grochowski, Mirosław Jankowiak, Izabela Kozera, Marek Łaziński, Norbert Ostrowski, Anna Socka, Marzena Stępień, and Magdalena Żabowska. Finally, two anonymous reviewers pointed out some weak or unclear places in a previous version and thereby helped me improve the coherence of the argument. Needless to say, all shortcomings and mistakes are exclusively mine. Björn Wiemer, JGU Mainz DOI 10.1515/9783110520583-010

266

Björn Wiemer

only clauses whose predicates demonstrate a reduced number of morphological properties assigned to finiteness (in a language-specific morphosyntax), but also features on the clausal level that are independent of the verb’s morphology (again, in a specific language). Regarding Slavic, we have to account not only for main clause uses of infinitives, but also for subjunctives (or whatever is treated as their equivalents), and subordinating connectives (complementizers, conjunctions) heading independent clauses. Other potential insubordination phenomena treated by Evans (and others) are, concerning Slavic, either hardly conclusive (word order), or non-existent (logophoric pronouns, affixes marking switch-reference), or are otherwise problematic (converbs). Against this background, this contribution pursues two aims. First, I want to present a variety of cases from different Slavic languages; in the first place from Russian, Polish, Czech, Macedonian, and Bulgarian, but I will also discuss some data from Ukrainian, Serbian-Croatian, and Old Church Slavonic (OCS.). These cases show that the infinitive as a predicate of syntactically independent clauses has been, and is, used in a majority of functions assigned to insubordination phenomena by Evans and other researchers who followed his insights. In fact, all three large functional domains for which Evans found examples in his world-wide survey can be exemplified with cases from Slavic languages (in earlier and contemporary stages), although to differing extents. These three domains are (Evans 2007: 368): (i) illocutionary distinctions (“expressions of interpersonal coercion”); (ii) propositional modality (in Palmer’s 22001 sense), otherwise ‘epistemicity’ in Boye’s (2012) treatment, i.e. epistemic modality and evidentiality. Actually, evaluative functions other than these two are treated in this rubric, too, namely: Those which logically presuppose factivity but evaluate observed facts from an emotional and/or cognitive point of view (e.g., It’s a pity / shame that p; see 2.2); (iii) usage to mark key moments in discourse, first of all narrative functions, but additionally also contrasting statements and reiteration. Actually, contrastive statements and similar phenomena, which Evans qualifies as being “high in [pragmatic; BW] presuppositionability”, can also be related to domain (ii). Second, I want to discuss syntactically independent clauses with finite verb forms (i.e. forms showing some tense and person-number distinctions) whose primary function falls into one of the aforementioned three domains and which, for some reason, can be subsumed under reduced finiteness. Among these, Balkan Slavic (Bulgarian, Macedonian) figures prominently, because these languages have lost the infinitive as a morphological form altogether (see in addition Nedelcu & Paraschiv this volume and Hill this volume for Romanian, as

Main clause infinitival predicates and their equivalents in Slavic

267

well as Jędrzejowski & Demske for a diachronic-typological context), and the question arises to what extent the “replacement strategies” which have become a firmly entrenched part of their grammar functionally coincide with the use of independent infinitives in other Slavic languages. Henceforth, ‘main clause use’ and ‘(syntactically) independent use’ will mean the same; ‘Balkan Slavic’ will be understood narrowly as restricted to the most “balkanized” part of South Slavic, which covers Bulgarian and Macedonian (beside the Torlakian dialects in southeast Serbia). For both types of reduced finiteness, I want to answer the question of whether insubordination phenomena are mainly (if not exclusively) motivated by non-factivity or suspension of factivity (for notional delimitations, see Section 2.2). A closely related issue is whether independent infinitives (or their functional equivalents in Balkan Slavic) are restricted to coding states-of-affairs and precluded as instantiations of propositions (for clarification see Section 2.1). Beyond that, since the whole issue of insubordination inherently implies diachronic directionality, we have to ask whether independent infinitives, or their equivalents in Balkan Slavic, are diachronically secondary (as insubordination would imply), or diachronically primary in the sense that – at least during the period of the earliest attested stages of Slavic languages (from the 9th century AD) and of the particular languages we will take into account – infinitives or structures which are considered to have replaced them existed prior to their use in dependent clauses. In the latter case, no insubordination would obtain. Just one further remark seems to be appropriate from the start. Although insubordination makes assumptions about the semantics and pragmatics of simple and conjoined clauses, the processes that it implies (see Figure 1) cannot be described without some more generally accepted syntactic notions. My concern is mainly with the semantic and discourse-pragmatic development, and I will not proceed within generative syntax (as, for instance, Madariaga 2015 does), even if I might use certain notions associated with this framework at times. In particular, I understand ‘reanalysis’ as a syntactic phenomenon (probably triggered by semantic shifts in the interpretation of speakers) in a widely accepted sense formulated, following Langacker (1977: 59), by Harris/Campbell (1995: 50, 61) as “a mechanism which changes the underlying structure of a syntactic pattern and which does not involve any [immediate or intrinsic] modification of its surface manifestations”. Among underlying structures, we may distinguish between (a) constituency, (b) hierarchical structure (argument relations etc.), (c) category labels (as in conversion), and (d) other grammatical relations. Correspondingly, reanalysis can show up in different manifestations. At a certain point we will have to deal with reanalysis in the sense of (a) and (b) (see Section

268

Björn Wiemer

4.1). This, however, does not necessarily imply that reanalysis is an integral part of insubordination itself, it might well prepare it. Evans himself seems to have used ‘reanalysis’ in a rather loose sense, as he subsumed the ellipsis of main clauses under this label (see the comments on Figure 1 and the remarks in Malchukov 2013: 181f.). Other, especially functional, linguists have also treated reanalysis in a rather loose way (see the discussion in Haspelmath 1998: 345). The main point, however, to be shown is that the syntactic changes which must have accompanied, or prepared, the meaning shifts discussed below are not indicative of insubordination, regardless of which syntactic theory one works with. Discussions about real or alleged instances of insubordination do not seem bound to any specific, or narrowly formal, theory of syntactic relations; they are flexible, in this sense, as they work with more broadly accepted notions of syntax. Apart from this, the phenomena analyzed here could be seen to yield a mirror-image to the issue dealt with by Madariaga (2015). Whereas Madariaga demonstrates, among other things, how infinitival clauses in East Slavic turned from independent (or ‘absolute’) into subordinate structures, my concern is to argue that, in fact, many types of clauses with independent infinitives have survived as diachronically primary structures from the earliest attested stages of Slavic. I will proceed as follows. Section 2 starts with a discussion of the relation between some criteria of (non)finiteness and insubordination, in particular of assertiveness and a very closely related notion, factivity. On this basis, I will specify the notional framework and the cognitive-communicative contrasts relevant for the rest of the paper. In Section 3, I provide a representative survey of phenomena from Slavic languages that, at first sight, might most likely be considered as candidates of insubordination. This survey is divided into independent infinitives (Section 3.1), other purported insubordination phenomena (Section 3.2), and equivalents in Balkan Slavic (Section 3.3). Then, in Section 4, I turn to an analysis of available data on the diachronic development of the structures presented in Section 3 before summarizing to answer the question of whether insubordination in these cases applies, and to formulate further conclusions (Section 5).

2 Notional distinctions It is important to realize that (the degree of) finiteness and the main (or independent) syntactic status of a clause are different things, primarily because finiteness probably “can be decomposed and derived from more basic notions”

Main clause infinitival predicates and their equivalents in Slavic

269

(Nikolaeva 2013: 103). One of these “more basic notions” is the distinction between main and dependent clause. Finiteness seems to be an epiphenomenon resulting from an agglomerate of diverse properties dispersed over (verbal) morphology, syntax (as a property of clauses), and semantics, or rather pragmatics (“finiteness related to assertion or some such property of an utterance”); cf. Nikolaeva (2013: 104) with reference to Sells (2007). For this study, morphology is of no real relevance insofar as the infinitive in Slavic neither carries any tense or agreement features, nor does it mark per se mood or illocutionary force. Morphological features are only of limited interest in Balkan Slavic equivalents of infinitival constructions since these languages have lost not only their infinitive, but also their morphological cases and, thus, syntactic relations are almost exclusively marked by means other than a morphological nominative – oblique distinction. For a similar reason, syntactic criteria are not very helpful, either. For instance, canonically finite clauses have often been characterized as being able to take referentially independent subjects, “while in canonical non‐finite clauses the subject is not licensed, which gives rise to subject‐related transparency/ opacity effects such as reflexive binding, control, and raising” (Nikolaeva 2013: 109). However, whether referentially independent subjects are licensed or not proves to be a matter of biclausal rather than monoclausal structure; at least it cannot really be tested for monoclausal structures, which, by definition, have only one predicate-argument structure and referentially independent subjects can be chosen freely. Compare, for example, the Macedonian sentence in (1a), which is a monoclausal structure (the single predicate is an auxiliary complex), and sentence (2a), which is biclausal (with a complement-taking predicate subordinating a clausal complement with its own core), with their Russian equivalents in (1b) and (2b): monoclausal (1) a. Mac. Moraš da (*ti) dojdaš. / Moraš (*ti) da dojdaš. must.PRS .2SG CON 2SG come:PFV. PRS .2SG b. Russ. Ty dolžen prijti. 2SG . NOM must.SG . M come:PFV. INF ‘You(SG ) must come.’ biclausal (2) a. Mac. Dime se PN

RM

plaši, gostite da ne afraid:IPFV. PRS .3SG guest.PL-ART. PL COMP NEG

zadocnat. come_late:PFV. PRS .3PL

270

Björn Wiemer

b. Russ. Dima

PN . NOM

boitsja, kak by gosti afraid:IPFV. PRS .3SG as if guest.NOM . PL

ne

opozdali. NEG come_late:PFV. PST. PL ‘Dima is afraid that the guests will be late.’ In the Macedonian case, to add an independent subject pronoun (ti ‘you.SG ’) before the inflected lexical verb (dojdaš ‘(that) you come’) would be ungrammatical since, in auxiliary or phasal verb complexes, the subject of both verbs must be identical for conceptual reasons. (See further in Section 4.3–4.4 on the diachronic development of da-clauses.) Thus, the issue of referentially independent subject-NPs would not make any sense here. Among the syntactic criteria of finiteness, we are left with the distinction between independent vs. dependent clauses (= C-8 in Nikolaeva 2013). This is actually the syntactic property with which (non)finiteness has most intimately been associated, or with which it even coincides in the practice of many linguists of diverse convictions, among them those dealing with South Slavic, or more specifically, Balkan Slavic. Curiously, it is some cognitive-typological approaches that come closest to an equation of (non)finiteness with the main–dependent distinction. For instance, Cristofaro (2003) introduced a purely functional notion of subordination based on cognitive-communicative asymmetries between clauses,1 but she also dismisses the traditional, morphology-based [± finite]-distinction altogether since it is crosslinguistically inapplicable (Cristofaro 2003: 53f.). However, if, alternatively, we understand the (non)finiteness distinction as a purely notional opposition, non-finiteness boils down to subordination. As Nikolaeva (2013: 108) states, “[f]initeness by definition has to do with subordination”, i.e. finite forms are characteristic of independent and non-finite forms of dependent clauses. Surely, this is not what we need, for at least two reasons: First, many of the clause patterns we will be analyzing as ‘main clause infinitival predicates’ do not need to occur with (or: be embedded in) longer stretches of discourse, or even be combined with (or opposed to) another clause; many of them are typically used in one-clause utterances. How should we speak of asymmetry in 1 This basic property occurs under her ‘Asymmetry Assumption’: “Under the Asymmetry Assumption, subordination is a cognitive situation corresponding to the non-assertion of one of the linked SoAs [states-of-affairs; BW]. (. . .) Non-assertions can be identified by means of any device available in the language to challenge the linked SoAs. If one of the linked SoAs is not open to challenge, that construction should count as an instance of subordination.” (Cristofaro 2003: 47)

Main clause infinitival predicates and their equivalents in Slavic

271

such cases (asymmetry in relation to what?). Second, not all independent infinitival clauses can reasonably be regarded as lacking assertiveness (although it often proves hard to test this in a natural way); ironically, this seems to apply particularly to those (admittedly minor) types of infinitival clauses which form parts of longer stretches of discourse (see Section 3.1.1).

2.1 Assertiveness and propositions vs. states-of-affairs Crucially, what is concealed behind the dependent = subordinate = non/less-finite equation, is another of the more basic notions which determines finiteness, namely: assertiveness. In Nikolaeva’s formulation, “[a]n assertive utterance makes a statement about a certain time span by identifying a point on the time line in which the respective proposition is true. Canonically finite clauses are temporally independent and assert a proposition located in the past, present, or future with respect to the moment of speech.” Among non-assertive speech acts, she mentions imperatives, hortatives, optatives, deliberative questions, and exclamations, saying that these “are often expressed by independently used non‐ finite forms” (Nikolaeva 2013: 113). These speech acts also supply good examples of assertiveness and tense (or, more properly, the relation between the interval of a denoted state-of-affairs and a reference interval) being independent of each other. “Most non‐assertive independent clauses denote an event which may be temporally located with respect to the time of speech, but they have no truth value at this time.” Therefore, “temporal anchoring does not reduce to assertion. Interrogatives are not assertive, although they are tensed” (Nikolaeva 2013: 114). As will become obvious shortly below, assertiveness is central to our concerns, whereas tense (or tensedness) plays a subsidiary role. Before going further with assertiveness, we should clarify another issue. Only propositions can be asserted, and directive speech acts, optatives and exclamatives do not have any proposition. Propositions are about (the degree of) certainty and other attitudes toward knowledge and belief from a speaker’s current point of view; propositions are the units (‘third-order objects’ in Lyons’ 1977 terms) which we can predicate as ‘true’ or ‘false’, and to which we can apply operators on a gradient of cognitive attitudes to something predicated. Non-assertive (e.g., volitional, deontic) speech acts do not contain propositions, they cannot be judged in terms of ‘true’ vs. ‘false’; they only code states-of-affairs (i.e. simple descriptions of situations, Lyons’ ‘second-order objects’, or what in formal semantics is often dubbed ‘eventualities’). Probably the most concise way to characterize the relation between proposition and state-of-affair (SoA) is a cognitive one provided by Boye (2012: 278, 281). It can be briefly paraphrased

272

Björn Wiemer

this way: A proposition anchors a SoA in a world W, it adds reference to a SoA so that it stipulates its existence (in W). Now, against this background, we could wonder whether there are cases in which an independent infinitive (henceforth: IndInf) codes a proposition (and is, thus, assertive), and, if such cases exist, we ought to pay attention to whether these are somehow different from those cases in which no proposition is encoded. Quite in line with Nikolaeva’s above-quoted characterization, Evans (2007: 429f.) suggested that the particular motivation behind both the finite–non-finite and the main–dependent distinction might be “assertativity”. He generalized that speakers “draw on nonfinite constructions to detach themselves from speechact or epistemic commitments”, i.e. to distance themselves from the propositional content of the message, to suspend its factivity status, or to conceal personal involvement in directive speech acts (which do not contain a proposition), e.g. “by presenting infinitives as impersonal alternatives that avoid making the command stance overt” (2007: 430). Under the heading ‘modal insubordination’ Evans (2007: 394) subsumed three different cognitive situations: (i) deontic, (ii) epistemic, (iii) mirative. Note that he did not mention dynamic (otherwise called ‘circumstantial’) modality, which is salient for Russian ‘modal infinitives’ (see Section 3.1.2), and that, in practice, his category ‘epistemic’ includes reportive meanings, which is a segment of the evidential domain. Two crucial points should be emphasized, following Evans. First, the deontic and the epistemic/evidential type are characterized by different source constructions: “[W]hereas epistemic insubordination involves ‘pure’ markers of subordinate status, implicating ellipsed main clauses of reporting, thinking, perceiving, or asserting, deontic insubordination frequently involves complementizers with additional semantic content, such as showing tense/mood relations between clauses” (2007: 394). This difference shows a striking parallel to Noonan’s (22007) distinction of independent–dependent time reference (to which analogous distinctions related to factivity and modality can be added), which often surfaces in restrictions of tense and/or mood choice after particular complement-taking predicates (CTPs). Time-dependence occurs only with CTPs related to volition and/or directive speech acts (commands, requests, desires, etc.), whereas CTPs related to knowledge and belief states (but also to hearsay) are neutral with respect to their temporal orientation. Thus, in principle, their complements are not time-dependent. Together with this, the difference of ‘deontic vs. epistemic subordination’ which Evans pointed out correlates with the status of the clausal complement in terms of a SoA–proposition contrast (cf. Boye 2012: Ch. 4–5). Second, although he did not use this term, Evans pointed out mirative statements, having those utterances in mind that express the speaker’s “reaction to

Main clause infinitival predicates and their equivalents in Slavic

273

the proposition, such as astonishment or disapproval” (Evans 2007: 394). Miratives cardinally differ from epistemically modified statements insofar as they logically presuppose the existence, or factivity, of the proposition commented on. In this respect, a mirative contrast (‘expected–unexpected’) is most similar to axiological contrasts inherent to predicates (CTPs) with an evaluative meaning (‘good–bad’, ‘desirable–undesirable’, ‘pretty–ugly’): The speaker cannot deny that P holds, although P runs counter to his/her expectation.2 As we will see below in several places, “mirative” CTPs often choose the same complementizer (and show identical morphosyntactic behavior) as do CTPs denoting cognitive attitudes. These parallels re-occur in main clauses with the same markers, i.e. in alleged cases of insubordination.

2.2 Assertiveness, (non)factivity and truth values With all of this taken into account, we must explore notions and research background that is connected to a differentiation between factive and non-factive utterances. We have to be careful since, in linguistic practice, we can discern (at least) two different ways in which [± factive] distinctions have been captured: In one understanding, [+ factive] means that the utterance can be assessed as ‘true’ or ‘false’, whereas [‒ factive] relates to utterances for which this distinction is either simply not applicable, or in which an assessment in terms of ‘true’ vs. ‘false’ is suspended. Truth-value judgments are not applicable to utterances that do not contain a proposition (e.g., directive speech acts), while they are suspended either if the proposition is logically presupposed and, thus, beyond assertion (see below), or if the proposition is couched within a modification intimating that the speaker (or another judging subject in narration) does not have enough access to knowledge and, thus, cannot (or does not want to) carry out a judgment in terms of ‘true’ or ‘false’. To my knowledge, these two treatments of [‒ factive] have never been clearly separated, which can partly be explained by the fact that utterances of both types in their empirical behavior share some properties, among others those which will be the topic of my analysis in sections 3 and 4. There is, however, another tradition, more firmly rooted in logic, according to which [+ factive] means ‘true’ (i.e. the underlying proposition holds true) and [‒ factive] means ‘false’. This narrower understanding of ‘factivity’ is exploited when linguists speak of factive predicates in the sense of Kiparsky/Kiparsky 2 For a more elaborate treatment cf., for instance, DeLancey (1997) and Aikhenvald (2004: 195– 209).

274

Björn Wiemer

(1970, abbreviated as K&K 1970). I will not employ the [± factive] distinction in this understanding, but rather in the sense circumscribed in the preceding paragraph. In either of the aforementioned traditions, factivity often coincides extensionally with assertiveness. In the logical tradition, factivity has been put into an opposition with (logical) presuppositions at least since K&K 1970: Factive predicates are defined as CTPs for which the clausal argument remains true even if the CTP itself is negated (e.g., It is / is not deplorable that you failed to pass the exam), i.e. it lies outside the scope of the utterance’s assertion.3 Assertiveness can be defined in a similar vein by tying this notion to truth-value conditions (see Nikolaeva’s definition quoted in Section 2.1). In other cases, emphasis is put less on cognitive certainty and more on communicative intent or back- vs. foreground (see Evans’ characterization of “assertativity” cited in Section 2.1, but also Nordström 2010: 17f.). Note that both Nikolaeva and Evans speak of ‘assertiveness/assertativity’ in the sense that an utterance is subject to truth-value conditions, whereas ‘non-assertive’ would mean that the utterance cannot adequately be assessed in such categories; their notion of ‘assertiveness/ assertativity’ thus correlates with ‘factivity’ in the first sense described above (and which I will follow in my analysis). Actually, from this perspective, an asserted (= foregrounded) part of an utterance is opposed to its backgrounded part, and factive predicates (in K&K 1970’s sense) happen to be foregrounded, but this is not one of their necessary properties. Thus, one can stress an opposition of clausal complements conjoined to the same factive predicate, e.g.: I do NOT regret that you forgot your umBRELla, but that I DIDn’t WARN you before the rain (capital letters indicate syllables under contrastive stress). After all, a speaker can assert that s/he does not guarantee for the proposition being true (i.e. being factive in the logical sense), e.g. by adding a marker of uncertainty. In other words: A speaker can epistemically modalize the utterance, as in You probably / You might have failed to pass the exam. Here, from a communicative point of view, the speaker is suspending factivity: S/He does not cancel (or deny) the propositional content, but just indicates that this content can neither be confirmed nor denied according to his/her actual knowledge (or belief) state. Therefore, the “unit” which underlies either reading of ‘factivity’ and which Nikolaeva referred to in her definition of ‘assertiveness’ is the proposition. In other words: A common denominator between asserted parts of utterances and factive predicates is that 3 This can be put the other way around by saying that the clausal complement must be true in order for the whole sentence to be true (e.g., I regret / I don’t regret that you failed to pass the exam, *although, in fact, you passed the exam).

Main clause infinitival predicates and their equivalents in Slavic

275

they build on propositions. Something cannot be asserted that cannot, in principle, be submitted to truth values.4 From now on I will use the term ‘factive’ only as meaning ‘submitted to a truth-value assessment’ (and not in the narrower sense of referring to propositions which are ‘true’). Accordingly, ‘non-factive’ will not mean ‘false’ but rather either ‘not submittable to a truth-value assessment’ or ‘factivity suspended’. Admittedly, this is a circumscription in negative terms, and ‘non-factivity’, strictly speaking, lumps together two usage types that should rather be distinguished more neatly. However, this treatment of ‘non-factivity’ will do at least for the purposes of this investigation, and it corresponds to the way this term is used in many typologically oriented works (e.g., in Boye 2012) or those that have dealt with insubordination (as far as I know). In this perspective, a factive utterance implies the propositional status of its content (Boye 2012), but this content must not be sent to the background or suspended. Under these conditions, the following clause-types are non-factive but nonetheless imply propositional content: (i) future indicative, (ii) conditionals (both protasis and apodosis), (iii) declarative and (true) interrogative clauses with epistemic or evidential modifiers. Evaluational modifiers operate on propositions as well, but they mark judgments of a moral or aesthetic nature (‘good–bad’, ‘pretty–ugly’, ‘(in)appropriate’). If these modifiers behave like CTPs, their clausal complements can be either factive in K&K 1970’s sense (i.e. logically presupposed) or non-factive, e.g. because they create some obligation. Compare, for instance, the difference between a temporally anchored statement I appreciate / It’s okay that you arrived in time (→ clausal complement is presupposed as being true) and its generic counterpart I appreciate / It’s okay when you arrive in time (→ clausal complement cannot be assessed as true or false); actually, the latter sentence bears resemblance to conditionals (see above). See also the discussion of ex. (53a–b) in Section 3.3. Regardless, evaluational utterances (or modifiers) are a slightly different species compared to epistemic and evidential modifiers. For a similar reason, it is debatable whether generic statements belong here, too; they are characterized by lowered referentiality, which eo ipso diminishes their assertiveness (see Nikolaeva’s clarifications discussed in Section 2.1). In summary, first, regarding the clause-types just mentioned under (i–iii), I will say that factivity (or assertiveness) is suspended because propositions are implied but get, so to say, suppressed or relativized in relation (a) to verifiability at the moment of speech, (b) to another proposition, (c) to the speaker’s certainty 4 There can, however, be propositions (= units underlying truth-conditional judgment) that are not asserted. This is what logical presuppositions (realized by clausal complements of factive CTPs in K&K 1970’s sense) demonstrate.

276

Björn Wiemer

or an information source. Propositional content is also implied in mirative statements or exclamations, but since such utterances mark unexpectedness, assertiveness, though for a different reason, gets suspended, too. Second, suspension of factivity/assertiveness has to be distinguished from non-factivity caused by the lack of propositional content. This is the case for utterances dealing with volition (desirable SoAs) or with non-epistemic possibility or necessity. Thus, all sorts of utterances characterized by deontic or dynamic/circumstantial modality belong here. Furthermore, imperatives (in their core function) and hortatives also belong to this group: all sorts of constructions specializing in directive speech acts and also optatives and other construction types expressing volitional attitudes. This includes deliberative questions. Rhetoric questions, in turn, behave like evaluational predicates (or modifiers) in that they are compatible with either factive or non-factive utterances.

2.3 Evans’ paradox: a caveat for diachronic accuracy Before getting started with the analysis of pertinent classes of data, it is useful to remember a paradox pointed out by Evans (2007: 370). The very notion of insubordination makes sense only if, in a given language, we can determine unequivocal clause structures which, on language-specific grounds, are assigned closely to what linguists consider to reflect a contrast of main–independent clauses. Provided this contrast is maintained with sufficient consistency, structures associated with dependent clauses are marked if they occur without any discernible dependency on (or asymmetry against) concomitant clauses, or in isolated utterances. However, when such marked uses become more frequent (or expectable) in well-defined contexts – and ellipsis is not an issue (any more) – this contrast begins to become faint and can eventually become blurred altogether. One can possibly interpret this process as markedness reversal; it takes place if the last stage of Evans’ diachronic four-stage scenario has been reached. The stages are as follows (Evans 2007: 370):

Figure 1: Historical scenario of the rise of insubordination (according to Evans 2007)

Main clause infinitival predicates and their equivalents in Slavic

277

The entire construct of insubordination is under the risk of becoming circular if we do not pay enough attention to this caveat, of which Evans himself has made us aware. I will take up this issue later. Regardless, the facts and analyses I am going to present in the following do not support the scenario given in Figure 1.

3 Survey of purported insubordination phenomena in Slavic In Slavic, independent infinitives and, in Balkan Slavic, their “replacements” with da+Vfin consistently demonstrate a stable array of functions with an impact on illocutions and factivity, which have often been ascribed to insubordination phenomena in various languages. Evans (2007: 387) listed the following more specific domains: request, desire, possibility predicator (for root and epistemic modality), purpose clauses (“with an implicit ‘I say this (in order that X)’”), conditional clauses (“with an implicit ‘It would be nice / You would make me happy / I would like it’ etc.”). These strategies are often exploited as indirect requests or warnings in order to mitigate potentially face-threatening speech acts. Excluded from the following analysis are clauses with the infinitive in the subject function (e.g. Russ. Zdes’ žit’ skučno ‘Living here is boring’, lit. ‘here to live (is) to be bored’), even if the property predicated by this infinitive is another infinitive, as in Old Russian zakonъ prestupitь, vo tmu postupitь ‘To violate the law (is the same as) to enter darkness’ (“Poslovicy XVII veka”; cited from Sabenina 1983: 62). Such “subject infinitives” are widespread not only in older East Slavic and modern Russian, but also in other languages.5 Clauses with both subject and predicate expressed by infinitives are typical for proverbs (see the sentence just quoted) and, in line with this, are normally deprived of reference to a specific time period or concrete agents. Thus, with other clauses headed by independent infinitives to be analyzed below, such sentences share a deranking of reference and assertiveness.

3.1 Independent infinitives From the contemporary perspective, it is most convenient to classify the types and functions of independent infinitives (IndInf) on the basis of modern Russian because, among all Slavic languages, this is the language with their broadest 5 For an analysis of clauses with infinitives in the function of the subject cf. Sabenina (1983) regarding Old Russian, Švedova (1964: 251–253) for 19th century Russian, Gawełko (2006) for a corpus-based study on eight European languages (see 3.1.3).

278

Björn Wiemer

array. Against the background of Russian, I will explore other contemporary Slavic languages. The survey does, however, not aim to be wholly comprehensive, rather it is meant as an introduction to the range of phenomena that have to be taken into account for constructions with IndInfs and their functions in Slavic. Before we come to the most prominent (and well studied) type of the so-called ‘modal infinitive’ (3.1.2), let me first investigate those types of the IndInf which cannot be considered modal. 3.1.1 Non-modal usage in contemporary Russian: narrativity and unexpectedness Example (3) presents the so-called ‘narrative infinitive’ in Russian (cited from Maurice 1996: 80): (3) Ona i govorit‘ so mnoj ne xotela, kak budto serdilas‘. Ja i nu tancevat’, čtoby razveselit‘ ee. SG . NOM and PTC dance: IPFV . INF in_order cheer_up: PFV . INF her. ACC 1 ‘She didn’t even want to speak with me, as though she was angry. Well, then, I started dancing to cheer her up.’ (F. Dostoevskij) The narrative infinitive is used as a discourse pragmatic device of highlighting some, often rather unexpected, event. This often causes an ingressive effect (‘started X-ing’). This effect should probably be considered as a default, which can be reinforced (e.g. by the incentive particle nu ‘well (then), come (up)’ in (3)) but also canceled. As in other languages,6 the Russian narrative IndInf is, as a rule, part of a sequence of clauses. It is preceded by finite (tensed) predicates, or at least I am not familiar with any example in which the narrative infinitive would have appeared as initial predicate of the chain.7 More importantly, the infinitival clause together with the preceding (and maybe following) clauses makes up a real sequence, not just a loose enumeration of events. In accordance with this, individual level predicates, e.g. state predicates like napominat’ 6 Narrative infinitives have been studied quite extensively, first and foremost, in Latin and Romance languages; they are particularly productive in French. Cf. Nikolavea (2007: 153–159) for references and extensive discussion. 7 This is one of the differences between the modern Russian ‘narrative infinitive’ and OCS. main clause uses of infinitives. According to Madariaga (2015), in OCS. such infinitives were found at the beginning of narrative stretches. Another, no less important, difference is that the highestranking argument of the infinitive did not occur in the nominative, but in the dative. It has to be left to future research as to whether there is any diachronic connection between these two different types of narrative use of main clause infinitives.

Main clause infinitival predicates and their equivalents in Slavic

279

‘resemble’, vygljadet’ ‘look like’ are disallowed (Avrutin 1999: 145–147); one may, thus, assume that the narrative infinitive implies controlled, or at least dynamic predicates. Moreover, narrative infinitives may occur “if the preceding verb expresses an habitual situation and the infinitive denotes events that follow each occurrence of the habitual situation in time” (Nikolaeva 2007: 153f.). Nikolaeva considered such cases as “marginally acceptable”, and they, in fact, are rare in Russian. At any rate, the nominatival subject has to denote a specific referent (Avrutin 1999: 139). As can further be seen from (3), narrative IndInfs show properties of finiteness: “[T]hey have overt referentially independent subjects, they express assertions, and they have time reference” (Nikolaeva 2007: 158f.). In this respect they crucially differ from all ‘modal’ uses discussed below, in fact for some properties they are diametrically opposed. From a functional viewpoint, narrative infinitives can be considered as manifestations of constructional economy (Nikolaeva 2007): Given a sequence of clauses (or predicates) which iconically reflects the real order of events described by these clauses, marking tense, agreement categories and anything related to realis-irrealis distinctions becomes superfluous; as long as the context does not supply indicators to the contrary, the interlocutors can understand that each clause contributes to the narrative chain of events. Under this perspective, the narrative infinitive resembles devices known from other parts of the world in which tense distinctions, or morphemes marking narrative past, are suspended for all predicates of such a sequence except the first one (cf., for instance, Longacre 1979: 120 on Ica, South America, and Dahl 1985: 114 on Bantu). In addition to Nikolaeva’s considerations, we should be aware of another connection to functions in which IndInfs seem to be prominent. Narrative infinitives can be used to highlight some (rather unexpected) event; this discourse function bears some similarity to miratives. A mirative connotation becomes more prominent with the adversative and the echoing type to be discussed below. We might, thus, be tempted to unite these three non-modal usage types under the heading of ‘unexpectedness’, which is a notion often covered by evidential or epistemic markers in different languages. The reason for this frequent meaning association certainly resides in the fact that all these domains (evidentiality, mirativity, epistemic modality) rest on propositions, or in other words: That markers with such functions semantically scope over propositions. However, as already stated, mirativity differs from epistemicity since it presupposes factivity: The speaker cannot but admit that some P holds true and expresses his/her surprise at this fact (see Section 2.1). Epistemic functions, to the contrary, lower the speaker’s commitment regarding P, and evidential functions suspend it insofar as they do not relate to truth functions or speaker’s commitment, but

280

Björn Wiemer

just indicate the cognitive-communicative basis of one’s judgment. Thus, the mirative function of the IndInf appears to be a link between ‘realis’-functions in narrative use and use for non-factive events or suspended factivity. In Slavic, the narrative infinitive is rare (and analyses are almost lacking). It is also a non-productive technique in, say, colloquial standard Russian and almost nonexistent altogether in contemporary Polish. Miklosich (1926: 851) mentioned assertive infinitives used in Slovene as “replacements” of OCS. participles within narrative clauses. See, for instance Slovene (4) Izraelci, to viditi, zapijejo. israel_people.PL DEM . N see.INF start_singing:IPFV. PRS .3PL ‘The people from Israel, when they see this, start singing.’ in which vidi-ti ‘see.INF ’ is used as a translational equivalent of OCS. vidě-vъš-e. PST PA.PL . M . Miklosich does not acknowledge this infinitive as narrative (“nicht mit dem historischen inf. zu vergleichen”), but it actually fulfils all the functions named above, except that it is used for background information. Apart from this, the philological question arises as to whether in Slovene these infinitives were not chosen just in order to find a more idiomatic way of copying the frequent use of participles in OCS. narrative texts. Let us now turn to the ‘adversative infinitive’; cf. Maurice (1996: 82), from which (5) is cited: Russian (5) Ja tebe rasskazyvaju veselye istorii, a ty spat‘. ADV 2SG . NOM sleep:IPFV. INF ‘I am telling you funny stories, and you (are going to) sleep.’ In (6) we have the infinitive in an echoing reply:8 (6) Ja, obmanyvat‘ svoego druga?! 1SG . NOM deceive:IPFV. INF REFLPOSS . ACC . SG . M friend.ACC . SG . M ‘Me and deceive my friend?!’

8 Maurice (1996: 81) called it an ‘infinitive of indignation’ (Germ. Infinitiv der Entrüstung) because of its pragmatic properties specified below. She evidently followed Veyrenc (1979: 11–13).

Main clause infinitival predicates and their equivalents in Slavic

281

Echoing eo ipso has something of a quotative function to it, but note that this does not imply a modal function: The speaker only refers to something literally said about (or suggested to him/her). Any modal (e.g., deontic) overtone can, at best, be regarded as a discourse-conditioned implicature. The echoing function comes close to the adversative type insofar as the speaker implies an opposing view to an opinion (or suggestion) uttered in a preceding speech act, and, as in the adversative type, the speaker gives way to his/her uncomfortable feeling. There are differences, though: The adversative type is not bound to dialogical replicae, it can be used in monologue and be part of a narrative plot, although it also requires some preceding linguistic context. The echo-type, to the contrary, is, as its name suggests, restricted to reactions in dialogue, the preceding context consists in another interlocutor’s utterance, whose verbalized “critical point” is reiterated (typically with a rising prosodic contour). In languages with restricted use or an absence of the infinitive as a morphological category, an echoic function can be fulfilled by nominalizations of verbs (cf. (7a)) or by structures with (morphologically) finite verbs and connectives that serve to defer factivity (cf. (7b)), such as Balkan Slavic da (see Section 3.3). See examples from Bulgarian: Bulgarian (7) a. Az i pluvane?! 1SG . NOM and spit:IPFV. NMLZ

‘Me and spit?!’

b. Az da pluvam?! ‘Me and spit?!’ 1SG . NOM CON spit:IPFV. PRS .1SG In a sense, the monological (or solipsistic) equivalent of echoic use is deliberative questions for which the IndInf is well documented, too. However, deliberative questions also include a certain modal, more precisely deontic, element absent in echoic use since they are usually meant as a mental self-check whether it is worth it (for the speaker him/herself) to undertake the action denoted by the infinitive. Deliberative questions have already brought us into the domain of modality proper, to which we now turn. 3.1.2 Modal independent infinitive: contemporary Russian The undoubtedly most widespread use of IndInfs usually associated with Russian is the so-called ‘modal infinitive’ (ModInf). From a discourse-pragmatic perspective it pronouncedly differs from the ‘narrative infinitive’ in that the ModInf can readily be used as an isolated utterance. In this respect, the ModInf also differs from the adversative and echo-type usages since it does not require preceding

282

Björn Wiemer

linguistic context. Lack of need for a preceding context correlates with the lack of an inherent argumentative function (for opposing viewpoints). A possible exception to both correlated properties are pseudo-questions or exclamations with negative polarity as in the following examples (cited from Veyrenc 1979: 27f.): (8) I kak tut bylo uderžat’sja ot slez! and how here be.PST.3SG . N refrain:PFV. INF from tears.GEN ‘And how one/I could withhold tears!’ (F.A. Abramov: “Prjasliny”. Moskva, 1974) (9) Mužiki suxodol’skie ničego ne rasskazyvali. Da čto im rasskazyvat‘-to bylo! ADV what.ACC 3PL . DAT tell:IPFV. INF- PTC be:PST.3SG . N ‘The peasants from Suxodol didn’t tell anything. Actually, what could they really tell!’ (I.A. Bunin: “Suxodol”) Such utterances do presuppose some preceding context since they only sound natural as reactions to some other utterance (or its non-verbal substitute) belonging to the same communicative event. As this type of reaction, they serve the purpose of convincing the interlocutor(s) of a point the speaker wants to make. These discourse properties may be inherent to some sort of construction, and the question is which role is played by the past tense copula (bylo). I will return to this question in Section 4.1.3. Regardless, the label ‘modal’ for this type of IndInf is justified insofar as the whole construction carries properly modal functions, i.e. in a domain defined on the basis of necessity (NEC) and possibility (POSS).9 The most common uses are deontic and dynamic.10 Importantly, modal functions of any kind of ModInf can be determined only for the construction as a whole, practically all of them consist of the infinitive and a potential agent coded in the dative (henceforth: potAgDAT ). I call it potential agent because of the modal and, thus, non-factive nature of the predication. The potential agent can remain implicit, though (see below). Remarkably, the ModInf only allows a genuinely epistemic reading to a very limited extent (cf. Maurice 1996). Even those mentioned, for instance, in Fortuin (2005: 39) seem to arise from the combination of a dynamic meaning with an explicit indication of the future, e.g., Russ. 9 For a justification of this restriction cf. van der Auwera/Plungian (1998), Hansen (2001: 54–81). 10 In the tradition of formal logic, ‘dynamic’ coincides more or less with ‘alethic’ modality (for which cf. Lyons 1977: 791–793). A nice overview of the different modal notions related to the ModInf is provided by Maurice (1996).

Main clause infinitival predicates and their equivalents in Slavic

283

(10) Ėto u tebja žizni ne ostalos’? Togda, polučaetsja, zavtra. mne umirat’ 1SG . DAT die:IPFV. INF tomorrow ‘It is you who hasn’t anything left of life? If so, this means I have to die tomorrow.’ (Ju. Oboroten’: “Komnata”) Examples like these could also be subsumed under what Maurice (1996: 30–32) dubbed ‘fatalistic modality’: The force that has the potential to evoke the event denoted by the infinitive is thought of as being beyond human control (nature, God). This, however, brings this subtype close to dynamic modality, and I suggest that it can be treated as intermediate between dynamic and deontic modality. Irrespective of this, an extension into the domain of epistemic judgments deserves attention since it correlates with a distinction of SoAs vs. propositions. Neither dynamically nor deontically modalized utterances code propositions, however, epistemic judgments presuppose them. By the same token, factive statements require propositions (see Section 2.2). Therefore, the reluctance with which the ModInf has come to be used in the sphere of epistemic modality seems to be indicative of its restrictions to utterances (predications) that code SoAs but not propositions, and which are, for this reason, non-factive. After all, even if ModInfs occasionally acquire epistemic readings, the implied proposition would be marked by a reduction or lack of speaker’s commitment regarding its factivity, since this is what epistemic markers (or constructions) convey. In modern Russian, the constructional frame [potAgDAT + INF] interacts with (external vs. internal) negation and the inevitable choice between imperfective and perfective aspect. We do not need to go deeply into the intricacies of these interactions because they do not relate to the main point, namely: Almost all kinds of the ModInf eo ipso go with non-factivity. Let me only briefly list the basic facts which seem to be agreed upon by researchers; cf., among others, Maurice (1996), Wiemer (2001), Fortuin (2005), which are the sources of most of the examples till the end of this subsection. A dynamic interpretation of the ModInf is usually bound to the external negation of a perfective verb, as in the following textbook example: (11) Emu ne sdat’ ėkzamen. he.DAT NEG pass:PFV. INF exam.ACC (On lenilsja učits’ja i ničego ne znaet.) ‘He won’t pass the exam. (He was too lazy to study and doesn’t know anything.)’

284

Björn Wiemer

A deontic interpretation of the ModInf correlates with the use of the imperfective infinitive externally negated sentences. This meaning is specified either as permission/prohibition or as unnecessariness (‘Why do SoA?’, ‘You need not care about SoA.’): (12) Emu ne sdavat’ ėkzamen. he.DAT NEG pass:IPFV. INF exam.ACC ‘He doesn’t have to / cannot pass the exam.’ (i) Zapretili ved‘. ‘For they forbid it.’

→ prohibition

(ii) On uže sdal ego v prošlom godu. ‘He already passed it last year.’

→ unnecessariness

Thus, the imperfective aspect (with and without negation) roughly correlates with controlled (or controllable) situations and volition, while (predominantly negated) the perfective aspect is associated with possible results of activities that may be controlled or not. This distribution is particularly clear-cut with telic, or event-bound, imperfective and perfective verbs united into so-called aspect pairs. Closeness to deonticity and volition is also evident for one of the crosslinguistically extremely widespread usage types of the IndInf, namely: for requests, which are usually treated as equivalents of imperatives (Evans 2007: 391). As far as affirmative IndInfs are concerned, the aspect opposition does not seem to exert any serious restrictions; compare the imperfective (13) a. Molčat’! (X Molči(te)!) ‘Be quiet!’ keep_silent:IPFV. INF keep_silent:IMP. SG (PL) b. Spat’! (X Spi(te)!) sleep:IPFV. INF sleep:IMP. SG (PL)

‘Sleep!’

with perfective (14) a. Slit’ vodu iz trjumov. pour_out:PFV. INF water.ACC from holds.GEN ‘Pour out the water from the holds.’ b. Milicija! Vzjat’ ee! police.NOM take:PFV. INF her.ACC ‘Police! Arrest her!’ (M.A. Bulgakov)

Main clause infinitival predicates and their equivalents in Slavic

285

With negation the situation changes in the same way as it does for negated imperatives: Negated imperfective IndInfs are used as prohibitives, e.g. (15) Dver‘ ne zakryvat‘! door:ACC NEG close:IPFV. INF ‘Don’t close the door!’ Negated perfective IndInfs are more rare, but, if they occur, they function as preventives, e.g. (16) I ne zabyt’ ix pozdravit’! and NEG forget:PFV. INF them.ACC congratulate: PFV. INF ‘And don’t forget to congratulate them!’

(Žvaneckij)

Cf. Maurice (1996: 166–171) for details. In line with this is the fact that, in many European languages, the IndInf is used as a default (or at least usual) choice in texts of instruction such as recipes (Glaser 2002). This includes not only Russian (Maurice 1996: 167f.) or German, but also Polish (Gębka-Wolak 2011: 52). The directive function of the IndInf is in harmony with the general “macro-illocutionary” predestination of texts of instruction. From the ModInf proper, they differ with respect to the potential agent both in grammatical and pragmatic terms: In recipes, the infinitive is not accompanied by a potAgDAT; in fact, it is prohibited, even in Russian, e.g. (17) Varenyj kartofel’ propustit’ (*vam) čerez boiled_potatoes.(SG .)ACC press:PFV. INF 2PL . DAT through mjasorubku! mincing-machine.ACC ‘Press (*you.DAT ) the boiled potatoes through the mincing-machine!’ This is naturally explained by the fact that recipes are directed to a generalized (or anonymous) addressee without any specific temporal reference. Let us continue with the Russian ModInf proper. Neither dynamic nor deontic utterances with the ModInf show restrictions regarding grammatical person. Deontic ModInfs are also quite usual in real questions: (18) Pape počitat‘ tebe iz knižki? daddy.DAT read:PFV. INF 2SG . DAT from book.GEN ‘Should daddy read for you from the book?’

286

Björn Wiemer

Sentences like (18) can be understood as suggestions or indirect requests to undertake the denoted action, while dynamically modalized utterances can hardly be so imagined. Instead, dynamically modalized questions are widespread with WH-words; for instance: (19) Kak / Gde / U kogo / Kogda mne zdes’ kupit’ bilet? how / where / at who.GEN / when 1SG . DAT here buy:PFV. INF ticket:ACC ‘How / Where / With whom / When can I buy a ticket here?’ Apart from this, dynamic questions can be formulated with the ModInf in Russian (but not in Polish) if the potAgDAT is not a person but some inanimate object or abstract entity; compare the following rhetoric question (quoted from Łaziński 2011: 147): (20) Da i otkuda emu vzjat’sja, ėtomu zdorov’ju, PTC and from_where he.DAT arise:PFV. INF DEM . DAT. SG . N health.DAT. SG . N posle vsex ėtix ėksperimentov, kotorye proizvodilis’ nad našim narodom? ‘And from where should it come, this health, after all these experiments which have been performed with our people?’ Above, we introduced the potAgDAT as a necessary component of the ModInf. It seems important to indicate certain types of the ModInf in which the potential agent is only implied, as it refers to the speaker of the utterance. Such cases are most characteristic of yes/no-questions in face-to-face communication when the speaker implies him/herself as the agent of the requested action. Because of this usual implicit reference, we may call this usage type ‘egocentric’. Compare textbook examples: (21) a. Nalit‘ tebe čaj? pour_in:PFV. INF 2SG . DAT tea.ACC ‘(Shall I) pour in tea for you?’ b. Nalit’? pour_in:PFV. INF ‘(Shall I) pour in (tea)?’ Note that, despite the rather directive, volition-based character of such requests, aspect choice differs from the deontic use of the ModInf: In egocentric use, ceteris paribus, perfective verbs are very usual.

Main clause infinitival predicates and their equivalents in Slavic

287

The egocentric viewpoint of such requests is a feature shared with deliberative questions, either in yes/no-questions (22) or in WH-questions (23). The 1SG pronoun is customary but may be “dropped”; moreover, the egocentric viewpoint can, as it were, be abandoned and be transferred to the addressee or a nonparticipant of the speech event: (22) Posovetovat’sja li (mne / tebe / emu) s bratom? counsel:PFV. INF Q 1SG . DAT / 2SG . DAT / he.DAT with brother.INS ‘Should I / you / he ask the brother for advice?’ (23) S kem (mne / tebe / emu) posovetovat’sja? with who.INS 1SG . DAT / 2SG . DAT / he.DAT counsel:PFV. INF ‘Whom should I / you / he ask for advice?’ In WH-questions, the subjunctive particle by may be inserted (S kem by mne posovetovat’sja? ‘Whom might I (probably) ask for advice?’). For more on this see Section 3.2. A difference between egocentric and deliberative questions lies in the fact that requests like (21a–b) simultaneously involve the hearer/addressee, whereas deliberative questions can be (and usually are) uttered without the presence of any interlocutors (as it were, in a solipsistic manner). In yes/no-questions, the speaker takes the initiative; the request need not be (and usually is not) a reaction to a preceding utterance. In the same vein, the illocutive purpose of such requests hardly renders counterparts imaginable with negation, and these seem, in fact, to be unattested, unless negation occurs with an effect of negative polarity, e.g. for politeness reasons. This can be also exploited as a hortative device: (24) A ne sxodit’ li (nam) v kino? and NEG go:PFV. INF Q 1PL . DAT in cinema (X Davaj,

HORT. SG

sxodim v kino!) go:PFV. FUT.1PL in cinema

‘Shouldn’t we go to the cinema?’ (‘Let’s go to the cinema!’) The temporal paradigm of both the ModInf proper and its “egocentric relative” in yes/no-questions is defective, insofar as the past and the future form of the copula (bylo, budet) occur rarely and only under very specific discourse conditions.11 For egocentric yes/no-questions, they are hardly encountered in natural 11 For some facts and deliberations cf. Maurice (1996: 288–292). Strangely, she defies bylo and budet a function as mere temporal markers (also 1996: 84) despite the fact that these forms paradigmatically correlate with zero realization in the present tense. Cf. Fortuin (2005: 41f.):

288

Björn Wiemer

speech, probably since their illocutionary function is tightly bound to the immediate speech situation. Thus, an utterance like ??Nalit‘ tebe bylo / budet čaj? ≈ ‘Was/Will it (for me) to pour in for you some tea?’ sounds very weird (Maurice 1996: 85f.); it could at best be interpreted not as an actual question (or suggestion), but as an echoic reaction. As for the ModInf proper, we do encounter utterances like (25)–(26), although examples like (26), with the future copula, are extremely rare:12 (25) Teper’ vyxoda uže ne bylo, domoj mne bylo ne projti, home(adverb) 1SG . DAT be.PST. N NEG go_through:PFV. INF a vpered idti smysla ne bylo. and forward go:IPFV. INF sense.GEN NEG be.PST. N ‘Now there wasn’t any escape, it was impossible for me to return home, and there was no sense in going forward.’ (26) Ty žar-pticu voz’mi, a zolotuju kletku ne trogaj: eželi kletku voz’meš’, to tebe ottuda ne ujti budet. PTC 2SG . DAT from_there NEG go_away:PFV. INF be.FUT.3SG ‘You take the Zhar-Ptitsa, and don’t touch the golden cage: if you take the cage, you won’t be able to go away.’ (from a fairy tale) As (25) demonstrates, the use of the past tense copula can be motivated by a transferal of the reference interval in a narrative context.13 3.1.3 Other modern Slavic languages Let us now compare the distribution of IndInfs with what we find in Polish, Czech and some varieties of western South Slavic. In his study based on parallel Bylo and budet should be considered as past and future tense markers of the ModInf, although they form part of the entire construction, since “they show idiosyncratic behaviour”, e.g. with negation, which has to be adjacent to the infinitive (see examples in this subsection) and cannot be used with the copula (Bylo ne najti ručku. vs. *Ne bylo najti ručku ‘It was impossible to find the pen’, with perfective infinitive). Maurice (1996: 288) admits that negation and infinitive have to be understood as one unit; thus, they can only jointly be put under the scope of a tense marker. The same holds true for aspectual modifications encountered in some dialects, like the inceptive Ne vorotit’sja stalo ‘It became impossible to return’ (Maurice 1996: 86). 12 In her corpus, Maurice (1996: 84) did not find a single example. Ex. (25–26) are quoted from Fortuin (2005: 42). 13 Many more examples of this kind are discussed in Veyrenc (1979: 27–31, 41).

Main clause infinitival predicates and their equivalents in Slavic

289

corpora of Polish, English, German and five Romance languages Gawełko (2006) provided some rough figures concerning basic grammatical functions of infinitives. He pointed out that, among the aforementioned languages, in Polish the role of the infinitive is in many respects much weaker, and its weight is apparently weakest as an independent clausal predicate (beside the infinitive used as a subject; Gawełko 2006: 269). 3.1.3.1 Polish ‘modal infinitives’ Unfortunately, I know of no comparative study with an equal coverage of Slavic languages, which included the independent infinitival predicate. The only reliable corpus-based study devoted to this matter is found in Łaziński (2011) who compared his findings on Polish with known facts from Russian (see Section 3.1.2). According to him, the most prominent differences between Polish and other Slavic languages can be observed in declarative sentences (against interrogative, imperative or exclamative ones), in the sense that declarative sentences are least “favorable” for the IndInf (Łaziński 2011: 143). In both Russian and Polish, IndInfs most often occur in questions and commands. The prohibitive function of the negated infinitive can be found also in public warnings and other inscriptions (e.g., Pol. Nie dotykać przewodów! ‘Don’t touch the wires!’, at railway stations).14 Łaziński’s (2011: 147f.) study showed that, in Polish, there were about twice as many instances of IndInfs in questions as occurred in commands or exclamations. However, contrary to Russian, the potAgDAT has practically been ousted from use, except for 1SG -forms in questions (often with a rhetorical purpose); cf. (2011: 145), which is the source of the following example: (27) (. . .) nie NEG

mnie krytykować rozstrzygnięcia, 1SG . DAT (EMPH ) criticize:IPFV. INF decision.ACC . PL

które ostatecznie zapadają m.in. w tej izbie. ‘It is not me who has to/can criticize the decisions which eventually are taken, among other places, in this chamber.’ (NKJP; stenogramm of a Sejm meeting) In Polish (as in Russian), IndInfs are quite usual in egocentric yes/no-questions as in utterances like 14 In general, the distribution of the negated infinitive vis-à-vis other devices of prohibition (negated imperatives or participles, explicit nominal formulae like Zakaz palenia lit. ‘Prohibition of litting (cigarettes)’) is particularly idiosyncratic, not only in Polish, and beyond the scope of the present paper.

290

Björn Wiemer

(28) a. Przynieść mu kawałek ciasta? bring:PFV. INF he.DAT piece.ACC cake.GEN ‘Shall I (really) bring him a piece of cake?’ uttered to oneself. Egocentric questions show up in WH-questions as well, e.g. Co robić? ‘What shall I/we do?’ (cf. also Gębka-Wolak 2011: 53, 75). However, contrary to Russian, such questions always imply the speaker, they cannot be used with reference to another person (either the addressee or a third person). Thus, questions in Russians such as (13), (15), (17–20), (21a–b), (24) no longer have equivalents in modern Polish. Moreover, this type of question appears most typically in requests for advice or help if the task to be performed is difficult, rather than simple questions for information that is easy to supply (Łaziński 2011: 146f.). Nonetheless, IndInfs do occur in such “simple” questions. For instance, it is quite normal to ask something like (28) b. Przynieść ci kawałek ciasta? bring:PFV. INF 2SG . DAT piece.ACC cake.GEN ‘Shall I bring you a piece of cake?’ where the speaker explicitly addresses an interlocutor (compare with ex. (28a)). However, WH-questions with dynamic modality – as illustrated in (11) for Russian – have really gone out of use. In sum: The Polish IndInf has largely become limited to deontic functions with the speaker as the implied potential agent. The IndInf used as a command (i.e. directed to another/other person(s)) seems to have been on its retreat, especially if it occurs without negation. In fact, unnegated IndInfs used in directive functions demonstrate a tendency toward idiosyncractic meaning shifts. For instance, Tak trzymać! ≈ ‘Great, continue doing it that way!’ (lit. ‘Hold (it) that way!’) is phraseologized. An anonymous reviewer pointed out that Jechać, jechać! (normally reduplicated like this), which literally means ‘Drive, drive!’, is appropriate not only “if you coach a team and your team is too slow or you are losing a game”, but also in a situation when “you are a leader of [a] group renovating a house and if the group is too slow” and you want your group “to be on time with the renovation”. In this last situation, the original meaning of directed movement has somewhat bleached. After all, unnegated directive IndInfs are no longer created freely from the majority of Polish verbs. Regardless of this, the distribution over sentence moods and illocutionary types makes one feel inclined to infer that, in Polish, IndInfs are heavily associated with utterances of suspended factivity (e.g., in exclamatives), or with

Main clause infinitival predicates and their equivalents in Slavic

291

utterances void of propositions (directives, deliberative questions). This fits well with the observation that the Polish IndInf is still productive in optative utterances, but usually only jointly with the subjunctive ‘particle’ by (see Section 3.2): by wszystko i jechać. (29) Ot, teraz. . . rzucić PTC now throw:PFV. INF IRR everything.ACC and drive:IPFV. INF ‘Well then, now. . . (it would be nice/good) to leave everything and (just) go.’ (M. Kuncewiczowa) Without this particle, optative use (as well as the IndInf to mark dynamic impossibility), again, tends to be severely restricted lexically (i.e. to turn in phraseologized units); cf. Łaziński (2011: 144f.). And even with this particle, the infinitive can only be used with the speaker as a potential agent; the addressee or a third person can be indicated only under the condition that the speaker of the utterance intimates to that person a desire and objects against that person’s attitude/behavior, e.g. (30) A tobie by tylko oglądać telewizję and 2SG . DAT (EMPH ) IRR only watch: IPFV. INF television.ACC (zamiast wziąć się do czegoś porządnego)! ‘And you would like to only watch TV (instead of dealing with something serious)!’ 3.1.3.2 Independent infinitives in Czech The use of IndInfs in contemporary Czech seems to be even more restricted than in Polish, but the narrow range of use fits well into the general picture that IndInfs are a favorable device to mark suspended factivity. Namely, the infinitive is used in conditional clauses, such as (31) Být tebou, tak to neudělám. be.INF 2SG . INS so DEM . N NEG .do:PFV. FUT.1SG ‘If I were you, I wouldn’t [lit. will not] do it.’ (32) Vědět to dřív, tak tam nejdu. know:IPFV. INF DEM . N earlier so there NEG .go:(I )PFV. PRS .1SG ‘If I knew that in advance, I would not go there.’

292

Björn Wiemer

In these examples, the conditionals are meant rather as relating to a potential event. The fact that this conditional construction may be used for indicating a counterfactual event (or state) in other contexts should not be excluded here, however. This topic is open to further research. More importantly, conditional clauses are not independent clauses. Really independent infinitives are clearly restricted in use. In the 19th century, there were attempts at revitalizing IndInfs with modal meanings on the model of Russian. These attempts did have some success, and in the writings of some authors who deliberately chose an archaic style, we find IndInfs in the modal functions known from Russian (e.g. dynamic necessity in Bylo jim odjeti:PFV. INF ‘They had to depart’). But such revitalization did not have a lasting effect, as nowadays they have become rare and sound obsolete (M. Giger, p.c., cf. also Grepl/Karlík 1998: 222). IndInfs are, however, still employed in a few types of directive utterances. They are used as quasi-imperatives, both unnegated (e.g., Mlčet! ‘Be quiet!’; cf. Němec 1977: §3.3) and negated ones; compare an example with a preventive function (negated perfective infinitive) from an internet blog (by courtesy of M. Giger): (33) Ještě je možnost ‒ ale s tím opatrně (nedotknout se kontaktu rukou) NEG .touch:PFV. INF RM contact.GEN hand.INS ‒ zkusit jeden ze startérů na chviličku vyzkratovat (. . .). ‘There is still a possibility – but be careful with this (don’t touch the contact with your hand) – of trying to short circuit one of starters for a short moment.’ (http://grower.cz/pestovani/marihuana/tema/t-854-p-5.html) Another type is the use of IndInfs in self-directed speech whereby the speaker admonishes him/herself for performing the denoted action. Wolfová (1971) calls this type ‘exhortative infinitive sentences’ and points out that this usage is salient in colloquial speech and prose with inner monologues (see ex. (34)). It is also typical in short notes that people take to remind themselves of something important (see ex. (35)). Compare two of Wolfová’s examples: (34) A honem se schovat před lidmi, and quickly RM hide:PFV. INF before people.INS abych se vyrovnal s tím, že jsem vlastně šťastný a jednoduchý člověk. ‘And quickly hide before the people, for me to become reconciled that I’m indeed a happy and simple person.’ (K. Čapek)

Main clause infinitival predicates and their equivalents in Slavic

293

(35) Odnést obuv do opravy. Přinést šaty bring_away:PFV. INF shoes:ACC to repair.GEN bring:PFV. INF clothes:ACC z čistírny. Důkladně prostudovat kap. 4. from dry-cleaning.GEN thoroughly study:PFV. INF ch. 4:ACC ‘Bring away the shoes for repair. Bring the clothes from the dry-cleaning. Study chapter 4 thoroughly.’ Such self-admonishing speech can acquire an echoing function, see for instance: (36) Neměknout v očích ani (v) tónu, NEG .bleat:PFV. INF in eyes.LOC NEG in sound.LOC umiňovala si Marta. ‘“Don’t let your eyes or tone of voice become weak”, Marta berated herself.’ (O. Scheinpflugová) Here we are reminded of echo replies discussed in Section 3.1.1 for Russian, although in Czech they appear to be fairly restricted to the egocentric domain, which we pointed out in Section 3.1.3.1 for Polish IndInfs. We can observe the following rule: If the egocentric utterance refers to an event that can be controlled by the speaker, the IndInf marks a directive speech act. If, however, the speaker has only limited or no control over the event named by the IndInf, the meaning becomes optative. Such utterances from Czech resemble Polish optative IndInfs discussed in Section 3.1.3.1 (see ex. (29)). The big difference is that, in Czech, optative IndInfs never occur with an “optative particle”, like by (or any other particle); cf. Wolfová (1971: 201) for a succinct discussion. Thus, we find examples like the following one: (37) Být zas děvče, běžet po lukách. be.INF again girl.NOM run:IPFV.INF over meadows.LOC ‘(If I could) be a girl again and run over the meadows.’ (B. Němcová) To summarize, in modern Czech the IndInf is restricted to directive utterances (as an alternative to the imperative) and to egocentric utterances which, depending on the controllability of the denoted possible event, can acquire directive or optative functions. Note that these functions are included in the range of IndInfs in Polish which still exhibits slightly more types of IndInf than Czech does. The few types still attested in Czech can certainly be regarded as a diachronic residual (after a stepwise retreat of the IndInf in different clause and

294

Björn Wiemer

utterance types), but there would be no point in regarding these usage types as diachronically secondary (see further Section 4.2). 3.1.3.3 Counterfactual clauses in Polish and Serbian-Croatian In Polish, beside directive and modal functions of the IndInf, which we discussed in Section 3.1.3.1, one can still encounter a construction with the infinitive and the past tense neuter form było of the copula być, which is used in a clearly definable counterfactual function and with the illocutive force of a reproach: (38) Było przyjść na czas (to by-ś się spotkał be.PST.3SG . N come:PFV. INF on time PTC IRR .2SG RM meet:PFV. PST.3SG . M z Jankiem). with PN . INS ‘You should have come on time (then you would have met Janek).’ (39) Było pracować (a nie gadać przez cały czas). be.PST.3SG . N work:IPFV. INF and NEG chat:IPFV. INF through entire time ‘You would have been better off to have worked (and not chat all the time).’ (40) Było nie dzwonić (tylko od razu przyjść). be.PST.3SG . N NEG call:IPFV. INF but at_once come:PFV. INF ‘You shouldn’t have called him (her/me/them) up (but come at once).’ Topolińska (2008a [1991]: 166f.) emphasizes that this construction is used exclusively in dialogue, as it presupposes the presence of a concrete (but unexpressed) addressee, that its “infinitive slot” can be filled by practically any verb of either aspect, and that two infinitives can be conjoined (see ex. (39–40)). The construction serves as “a verbal reaction in a situation when we want to criticize the interlocutor for something s/he missed to do” (2008a [1991]: 168).15 In modern Polish, the addressee can hardly be named; if expressed by a dative pronoun the utterance sounds awkward (??Było ci przyjść wcześniej. . . , to mean ‘You would/should have come earlier. . .’). This observation fits with Łaziński’s (2011) findings reported on in Section 3.1.3.1. The construction in (38)–(40) must have been productive (i.e. without serious restrictions of lexical input) until rather recently and is still used in colloquial 15 The original wording in Polish is: “(. . .) jest charakterystyczną dla współczesnej polszczyzny reakcją językową w sytuacji, kiedy chcemy naszego rozmówcę skrytykować za to, że czegoś nie zrobił”.

Main clause infinitival predicates and their equivalents in Slavic

295

speech, although its frequency might have started to decrease. As a small query showed, it is accepted and used rather freely even by educated speakers from different generations (between 25 and 65 years of age, see Acknowledgments). A search in the National Corpus of Polish (NKJP, http://www.nkjp.pl/) brought 14 hits to light from belletristic fiction and periodicals. The relevant pieces of discourse represent or somehow imitate dialogical speech; for instance: (41) (. . .) Co jej odstrzeliło, żeby odnaleźć pana Mietka? Było siedzieć cicho i czasami powspominać be.PST.3SG . N sit:IPFV. INF quietly and sometimes remember:PFV. INF te godziny, kiedy była srebrną dziewczyną do kochania i podziwiania. ‘Has she gone crazy to find Mietek again? She would have been better off to have sat at home silently and remembered from time to time those moments when she was a sweet girl there to be loved and admired.’ (E. Nowacka: „As w rękawie”; Warszawa 1998) This construction has also been included into a lexicographic compendium on colloquial Polish by Bogusławski/Danielewiczowa (2005: 49).16 As Topolińska (2008a [1991]) further remarked, we can only speculate about the provenance of this construction and its “age”, since historical data are practically lacking. The reason for the absence of pertinent attestations certainly resides in the dialogical character of this construction, which did not favor its appearance in historical documents based on written habits. However, there are some dialectal data from the Serbian-Croatian dialect continuum and from Eastern Ukrainian (see below), thus from regions unconnected to the Polishspeaking area. These scattered occurrences over different regions of West, East and South Slavic might be taken as an argument that this construction is quite old. Nonetheless, we cannot claim to be able to establish its original hotbed(s) or road(s) of diffusion, even less can we claim to be able to decide whether this

16 These authors registered (ibid.) another construction which is superficially identical (homonymous) with (38)–(40), but has a modal function, as (on the basis of negative polarity) it implies dynamic impossibility in rhetorical questions. See their example: Z kim było się spotkać? ‘With whom could I have met?’ (⊃ ‘There was nobody with whom I could have met.’). Probably, this construction has likewise started becoming slightly obsolete. More remarkably, the linking element between this construction and construction (38)–(40) is the counterfactual status of the event named by the infinitive, and as for either construction many contemporary speakers appear to be inclined to interpret it as elliptic, obviously on the background of constructions with modal auxiliaries (Z kim można było się spotkać?, with same translation). The latter ones, however, most probably developed later (parallel to what we observe for East Slavic, see Section 4.1).

296

Björn Wiemer

construction developed from complex sentences. Of course, from a contemporary perspective it is easy to treat the construction in (38)–(40) as a kind of protasis in a loosely connected conditional sentence (see ex. (38)) or as the first conjunct of an adversative sentence (see ex. (39)–(40)). However, first, it is speculative to assume that this biclausal structure was diachronically primary; second, even if it were it would not reflect subordination in the traditional sense (though maybe in the sense introduced by Cristofaro 2003); and third, both conjuncts can occur with an IndInf so that we would be back to our initial question of where these main clause infinitives might “derive” from. Probably, in terms of diachronic clause combining they derived from nowhere (compare with the remarks on Old Czech in Section 4.2). With the data at hand, we can only speculate as to whether the construction in (38)–(40) is diachronically connected to the dynamic use of the IndInf as we still find it in 18th century Polish (see Section 4.2) and in modern Russian (see Section 3.1.2). At least an immediate link does not seem plausible because the latter pattern marks negated dynamic possibility (¬ POSS p), while the construction in (38)–(40) implies the negation of a factive statement (i.e. just ¬ p). Counterfactivity can be conceived of as the “product” of the suspension of factivity carried by the infinitive, which gets combined with the past copula. This is in close parallel to known potentialis–irrealis oppositions in languages with a fullfledged system of perfect (or ‘anterior’) grams and a morphological subjunctive (such as the Romance languages, English, German, Baltic). In addition, the data which Topolińska (2008a [1991]: 169–171) furthermore discussed based on various Bosnian, Hercegovinian, Montenegrinian and Serbian dialects demonstrate that they do not align with the contemporary Polish ones in every detail. Namely, they allowed the addressee (in second person) to be coded in the nominative; the examples show that, in this case, the copula could be either in the neuter (as in (38)–(40)) or it agreed with the nominatival pronoun in gender and number (ex. (42) vs. (43), (45a) vs. (45b)); in other cases, the addressee could be elliptic but showed up in the finite ending of the copula (e.g., in the imperfect, ex. (44)). Both peculiarities made the construction look “more finite” than its purported “model” in (38)–(40). Last but not least: This construction is attested with reference to third-person subjects (ex. (45a–b)). Thus, the mentioned South Slavic dialects either lost a restriction which had been typical for it beforehand (and which would then have remained as an archaism in the Polish case), or the restriction to an implied second-person is an innovation of Polish (while the South Slavic dialects would have retained the more archaic features).

Main clause infinitival predicates and their equivalents in Slavic

(42) Nè

297

bìlo tî dòlazī! be.PST.3SG . N 2SG . DAT come:IPFV. INF

NEG

‘You’d better not come!’ (43) Bíla tî, Mîlena, dôći! be.PST.3SG . F 2SG . DAT PN . NOM come:PFV. INF ‘You’d better come, Milena!’ (44) Nè NEG

bijāše jèsti kìselīh jàbūkā! be:IMPF.3SG eat:IPFV. INF sour.GEN . PL apple.GEN . PL

‘You’d better not have eaten unripe apples!’ (to a child whose stomach is aching because it ate too many unripe apples) (45) a. Bílo òna dôć! be.PST.3SG . N she.NOM come:PFV. INF b. Bíla òna dôć! be.PST.3SG . F she.NOM come:PFV. INF ‘She’d better come!’ In sum, regardless of their restriction to an immediate addressee in face-to-face communication – and that this addressee had to be expressed by a pronoun – the really uniting properties of all sentences attested in Polish, dialects of western South Slavic and in Eastern Ukrainian are (i) the illocutive function (reproach) and (ii) counterfactual meaning. Again, we have arrived at non-factive utterances.

3.2 Other relevant phenomena: optative and apprehensive utterances In Russian, clauses in which the negated infinitival predicate is accompanied by the enclitic particle by yield apprehensive utterances, i.e. they show a tight correlation with warnings and admonitions (46a). This enclitic often cooccurs with the focus particle liš’ ‘only’ and, in this case, both units form one prosodic complex (46b): (46) a. Ne NEG

opozdat‘ by! come_late:PFV. INF IRR

b. Liš‘ by ne opozdat‘! only IRR NEG come_late:PFV. INF ‘If only I/we won’t come late!’

298

Björn Wiemer

By is a remnant 3SG -form of the aorist of byti ‘be’ left after the aorist as such vanished (a process which was finished by the 14th century). It is now the productive (and only) marker of the subjunctive. It also occurs as an integral part of many lexical items, such as, for instance, the complementizer and conjunction čtoby, used in various kinds of ‘irrealis’-clauses, or the complementizer kak by. The latter specializes as an ‘apprehensive complementizer’ used exclusively after CTPs with the meaning ‘be afraid’ (Russ. bojat’sja, opasat’sja). It is optional (instead the neutral complementizer čto can be used, see (47b)), but if kak by is used, the predicate takes the l-form of the general past and must be negated (47a). Note that regardless of this, the temporal reference of the complement clause is always posterior relative to the main clause; thus, in this respect, (47a) and (47b) are synonymous, they are identical also in their practically exclusive admissibility of the perfective aspect: (47) a. Katja

PN . NOM

ne NEG

b. Katja

bojalas‘, kak by gosti be_afraid:IPFV. PST. SG . F CONJ guest.NOM . PL

opozda-l-i. come_late:PFV. PST. PL

PN . NOM

bojalas‘, čto gosti be_afraid:IPFV. PST. SG . F CONJ guest.NOM . PL

opozda-jut. come_late:PFV. FUT.3PL ‘Katja was afraid that the guests will/might be late. / . . . lest the guests are late.’ The l-form requirement applies to all conjunctions and complementizers with agglutinated by, provided tense is distinguished on the verb. As (46a–b) show, by-units can also occur in infinitive clauses. In a sense, apprehensive utterances can be understood as the negative counterpart of optative utterances: In neither case does the speaker have control over the event, but the emotive attitude to this event is reversed. In addition, apprehensive utterances involve an epistemic component (the speaker forecasts that the undesirable event will happen), while for optatives such a component is absent. The affinity of optative and apprehensive utterances becomes manifest in the fact that independent unnegated infinitives with by are tightly associated with optative readings;17 compare: 17 Maurice (1996: 145f.), referring to other authors, assumes that IndInfs with by can also be interpreted deontically (as equivalents of imperatives), but she does not provide convincing examples.

Main clause infinitival predicates and their equivalents in Slavic

299

(48) S”est’ by čego-nibud’ vkusnen’kogo! eat:PFV. INF IRR anything.GEN tasty:DIM .GEN . SG . N ‘I’d like / It would be good to eat something tasty!’ (49) Kuda ugodno, liš’ by udrat’ ot ėtix wherever_to only IRR run_away:PFV. INF from DEM .GEN . PL psixov. mad_person.GEN . PL ‘It doesn’t matter where, if only (one/I could ran) away from these mad people!’ (cited after Maurice 1996: 153) Here, again, the predominant choice is perfective verbs. However, optative infinitive clauses with by can also be negated (‘I wish I need not X’, X = verb), and in this case imperfective verbs occur, e.g. (50) Ne NEG

vstavat’ by zavtra utrom tak rano! stand_up:IPFV. INF IRR tomorrow at_morning so early

‘If I could not stand up that early tomorrow morning!’ Thus, as elsewhere in the grammar of Russian (and other Slavic languages), aspect choice is sensitive to modal and illocutionary functions, as there is a covert, but neat distributional pattern distinguishing optative and apprehensive functions (in cooperation with negation).18 As the examples above demonstrate, optative clauses by default relate to the speaker (or the judging subject in free indirect discourse) as the potential agent, which, as a rule, remains implicit, but the wish may also refer to another person. In this case, the potential agent must be coded by a dative-NP (or pronoun), as in e.g. (cited from Maurice 1996: 149, 156) (51) A

tebe by, duraku, poprosit‘ da poklonit’sja. ADV 2SG . DAT IRR idiot.DAT ask:PFV. INF and bow:PFV. INF. RM ‘And you, idiot, would only ask and bow down.’ (Grigorovič 256)

(52) Tebe by otdoxnut’ kak sleduet! 2SG . DAT IRR take_rest:PFV. INF how suit:IPFV. PRS .3SG ‘You’d really take a sufficient rest!’

18 For some more details cf. Wiemer (2014; 2015).

300

Björn Wiemer

Here, the illocutionary point is either a reproach (51) or mild advice (52). Compare this with the Polish pattern było+INF discussed in Section 3.1.3.3. Let us now turn to the question about insubordination. Utterances like those in (46a–b) and in (48–52) do not have equivalents among dependent clauses, i.e., at least from a contemporary viewpoint, they cannot be understood as ellipses from complex sentences with “truncated” main clauses (see Section 4.1.2). Nor can they sensibly be “derived” from the protasis of conditional sentences (Esli by pogovorit‘ s nim, . . . ‘If one talked to him. . .’; ?Esli by ne opozdat‘, . . . ‘If I/we don’t be late. . .’); for an elaborate argument cf. Maurice (1996: 149–152). These observations run counter to the statement in Evans (2007: 393) concerning apprehensive utterances (admonitions): “[T]here is good comparative evidence that the subordinate use is historically prior”. This statement was based on Australian languages, Polish, and Basque. The only chance to ultimately solve these conflicting claims is to hope for sufficiently appropriate and reliable diachronic data going as far as possible back in the history of the respective language. Since both theoretical positions are reasonable, the problem turns out to be a purely empirical one: What was attested first? However, it is also essential to decide which period in the history of languages (however delimited from “sister idioms” and “ancestors”) can sensibly be taken as the point of departure. This issue will be taken up in Section 4.

3.3 Equivalents in Balkan Slavic (Bulgarian, Macedonian) The Balkan Slavic da+Vfin-construction, which has replaced the infinitive in practically all its earlier functions (see Sections 4.3–4.4), covers an identical range of functions in the same domains relevant for illocutions and (non)factivity that have been associated with insubordination (inside and outside Slavic): request, desire, possibility predicator (for root and epistemic modality), purpose clauses (“with an implicit ‘I say this (in order that X)’”), conditional clauses (“with an implicit ‘It would be nice / You would make me happy / I would like it’ etc.”; see beginning of this section) and others (see below). The grammatical and lexicographic status of da has been an issue of considerable debate. The reason is its morphosyntactic versatility: It occurs not only in biclausal structures (as a conjunction or complementizer), but also in monoclausal sentences, namely: in complex predicates as a kind of obligatory “link” between a phasal verb or auxiliary and a lexical verb (see ex. (1a) vs. (2a) in Section 2). These facts make it differ from conjunctions and complementizers with which it enters into functional [± factive] oppositions.19 It is certainly because of this behavior that 19 This was clearly articulated already by Gołąb (1954: 75f.). Cf. also Asenova (2002: 151f.) for Bulgarian.

Main clause infinitival predicates and their equivalents in Slavic

301

da has been considered as an analytic marker of mood or suspended assertiveness by many researchers. For the time being, the question of whether it is justified to regard da as an analytical inflection marker can be left open; the crucial thing is that, in clause combining and the formation of complex predicates, da plays an eminent and ubiquitous role in South Slavic morphosyntax. In Macedonian, da is used for clausal subordination as both a complementizer and a conjunction.20 As a complementizer, it contrasts with the epistemically neutral deka ‘that’ (more rarely oti); da occurs consistently after CTPs (or their nominalizations) with clausal complements that describe non-factive situations, including situations with suspended factivity. It also occurs after evaluative CTPs like nezgodno ‘inconvenient’, ubavo ‘beautiful’, if their clausal complement encodes an unreal (often generic) situation (53a); if, however, the clausal complement logically presupposes factivity of its content, the complementizer što is chosen (53b): (53) a. Dobre e da molčiš. good.N COP: PRS . SG COMP be_silent:IPFV. PRS .2SG ‘It would be good if you keep silent.’ (or, more literally: ‘It is good that you be / might be silent.’) b. Dobro e što molčeše. good.N COP: PRS . SG COMP be_silent:IPFV. IMPF.2SG ‘It is good that you kept silent.’ As an adverbial subordinator da introduces final, concessive, consecutive clauses and the protasis of conditional (irreal and potential) sentences. As mentioned above, da is also obligatory after phasal verbs and modal auxiliaries. In these cases, it cannot be qualified as a complementizer or clausal conjunction because these structures are monoclausal, i.e. da serves as an obligatory element in complex predicates connecting the auxiliary with the lexical verb (for which in other languages, among them Slavic ones, the infinitive would be used; see also Nedelcu & Paraschiv this volume and Hill this volume for a similar analysis of finite clauses in Romanian). The range of functions of da in Balkan Slavic can arguably be regarded as contemporary reflexes of the diachronic development of subordination patterns, which followed a cline in which the monoclausal usage appeared after the biclausal one (see Section 4.3.2). This cline can serve 20 Surveys on Mac. da can be found in Kramer (1986), Fridman [Friedman] (2011 [1987]), Topolinjska (2000: 90–104), Mišeska Tomić (2006: 416–456; 2012: 357–377), Georgievski (2009), and Wiemer (2014), where more references are given.

302

Björn Wiemer

as a nice illustration of syntagmatic tightening (“Today’s syntax is yesterday’s discourse”).21 Apart from subordinating functions, da occurs clause-initially in nonembedded clauses with a hortative (54) or optative function, or in suggestive questions (for which the speaker expects an affirmative answer, as in ex. (55)): (54) Da CON

gi prečekate! them.ACC wait:PFV. PRS .2PL

‘You should wait to welcome them!’ (55) Da

ne

si

CON NEG COP: PRS .2SG

nešto bolen? something sick.SG . M

‘Aren’t you somewhat sick?’ In questions, da is also used in combination with the interrogative clitic li: (56) a. Da CON

odam li? go:PFV. PRS .1SG Q

‘Should I go?’ b. Da CON

dojde li Ana? come:PFV. PRS .3SG Q PN

‘Does Ana have to come? / Would it be preferable for Ana to come?’ Note that such questions are not just about knowing, i.e. learning whether a proposition holds (or held) true or not, rather da + li implies “some modal background (Can I / Should I / Is it wise etc.)” (L. Mitkovska, p.c.). In any case, there is good reason to assume that these non-subordinating uses are historically primary, at least in the sense that they fit to the initial functions assigned to da in OCS. (see Section 4.3). In Bulgarian, the distribution over clause types and illocutionary functions is very similar to the picture that we sketched above for Macedonian.22 Regarding main clause structures, we may add usage in mirative statements (57), which 21 For Mac. da, though, this cline goes even a step further since its connection with the necessity modal mora has become so tight that both units might be argued to lexicalize into an epistemic particle. This can be considered, despite the fact that da in all its uses behaves like a proclitic which can be detached from mora, whereas it can be prosodically disconnected from the following finite verb only by other clitics (Wiemer 2014: 154–162). 22 For surveys cf. Genadieva-Mustafčieva (1970), Mišeska Tomić (2006: 457–476), Nicolova (2008: 409–429), among others. Gołąb (1954: 74f.) claimed that consecutive da-clauses are atypical for standard Bulgarian (in contrast to dialects and the Macedonian standard). His examples are,

Main clause infinitival predicates and their equivalents in Slavic

303

are functionally close to suggestive questions insofar as the speaker comments on a fact, or observation, which s/he does not want to trust or which s/he is afraid of (cited from Mišeska Tomić 2006: 474): (57) Da imaš pari, a da živeeš kato bednjak! CON have:IPFV. PRS .2SG money and CON live:IPFV. PRS .2SG like poor_man ‘To have money and to live as a poor man!’ Optative use is widespread, especially as counterfactuals in the perfect or past perfect: (58) Da

go bjax namerila tuk! CON he.ACC be.IMPF.1SG find:PFV. PRF. SG . F here

‘If only I would have found him here!’ (cited from Mišeska Tomić 2006: 474) Just like Macedonian, Bulgarian shows the use of main da-clauses in deliberative (59) or rhetoric questions (60): (59) Što23 da pravja sega? what.ACC CON do:PFV. PRS .1SG now ‘What should I do now? / What am I to do now?’ (60) Da

se

CON RM

zakoljam? stab_to_death:PFV. PRS .1SG

‘Shall I stab myself to death?’ Curiously, in Bulgarian we find da used in clause pairs with an adversative contrast. This use very much resembles the ‘adversative infinitive’, which we discussed for contemporary Russian. Compare the next example (cited from Gołąb 1954: 71) with (5) in Section 3.1.1:

however, not really convincing since they could also be interpreted as final clauses. In contrast to Bulgarian, Macedonian does not allow da to be used after unnegated perception CTPs (e.g., Bg. Viždam če / da idva edin čovek ‘I see a person coming’), unless these verbs are used as cognitive verbs (= ‘understand’); L. Mitkovska (p.c.). 23 Here što is not a complementizer (as it is in (53b)) but marks the focused constituent of a WH-question.

304

Björn Wiemer

(61) Az da ti prigotvja celija bagaž, 1SG . NOM CON 2SG . DAT prepare:PFV. PRS .1SG entire.DEF luggage da

ti podredja dori knigite, a ti CON 2SG . DAT arrange:PFV. PRS .1SG even books.DEF ADV 2SG . NOM tuk da si igraeš s tvoja Ferdinand. here CON RM . DAT play:IPFV. PRS .2SG with REFLPOSS . DEF PN ‘I prepare all of your luggage of you, I even bring your books into order for you, and/but you are just playing with your Ferdinand.’

This usage is also close to the ‘mirative’ one (see ex. (57)), insofar as the action or event described is considered to be factive, but the speaker distances him/ herself from it emotionally (thence the often encountered nuance of indignation). It can also be connected to the ‘narrative infinitive’. Compare the following examples (cited from Nicolova 2008: 428), in which da combines with present tense forms: (62) Az da mu pravja popara vseki den 1SG . NOM CON he.DAT do:IPFV. PRS .1SG sop(s) every day i mlin da mu toča, and mill CON he.DAT roll:IPFV. PRS .1SG a

toj da mi stoi kato pukăl nasrešta. ADV he.NOM CON 1SG . DAT stand:IPFV. PRS .3SG as_if jug opposite ‘I bake him sops every day and roll the mlin [a Bulgarian dish made of puff pastry], and/but he stands around as if he were a clay jug.’ (Čudomir) (63) Da

reša az onaja trudna zadača CON solve:PFV. PRS .1SG 1SG . NOM DEM . DEF. F difficult.SG . F task.SG . F

săs zabărkanata figura, a păk taja tolkova leka easy.SG . F with obscure.DEF figure ADV again DEM . SG . F very da CON

ja ostavja nedovăršena. . . her.ACC . F leave:PFV. PRS .1SG unfinished.SG . F

‘I solve that difficult task with the tricky figure, but that very easy one I leave unfinished. [⊃ How can I?!]’ (T.G. Vlajkov) Nicolova (2008: 428f.) does not explicitly connect this to narrative use; instead, she points out parallels with the so-called dramatic imperative in East Slavic.

Main clause infinitival predicates and their equivalents in Slavic

305

However, she explains that this usage is a transposition of the present tense to mark events that occurred in the past and are brought into some sharp contrast with each other: The speaker expresses his/her (negative) emotional attitude toward such unexpected (and undesired) contrast. If we sum up these observations, we are inclined to say that such transpositional use of da-clauses unites features that we know from the narrative, the adversative and the mirative use of IndInfs in Russian (see Section 3.1.1), or of other main clause uses of da in Balkan Slavic. All in all, differences between Bulgarian and Macedonian are minor,24 but they corroborate the two claims I want to make in order to pursue parallels with the IndInf in other Slavic languages. The first claim is that the main clause structures encountered with da in either language are not a result of insubordination; instead, they are diachronically primary patterns reflecting the usage types which we already find in the most ancient documents of Slavic, namely in OCS. The second claim is that all main clause phenomena with da are restricted to utterances which are non-factive because they either do not express propositions or suspend assertiveness; the latter includes emotional distance or lower referentiality (e.g., in generic statements). These uses have been labeled differently, not only with respect to Macedonian or Bulgarian, but also against a more general Balkan and South Slavic background. Most of them have been dubbed ‘analytical / bare subjunctive’ or ‘optative’ (cf. Topolińska 2008 [1994]: 175f., Noonan 22007: 105; Mišeska Tomić 2006: 439–444, 473–476; 2012: 370). However, these labels are rather a consequence of a mixture of notional distinctions related to finiteness, which we discussed in Section 2;25 among these notions, tense-aspect restrictions appear to be the most prominent. The crucial property relevant here is that da-clauses, regardless of any subordinate–main clause distinction (or its status in morphosyntax), are always associated with lowered or absent factivity. More recently, this has also been stated by Topolińska (2008 [1994]: 175f.) with regard to non-factivity. Gołąb (1954) should also be considered.

4 Diachronic development The Slavic (and Baltic) infinitive derives from the dative-locative form of a verbal noun (nomen actionis) in *tei (> Slav. -ti > Russ. -t’ / Pol. -ć etc.); cf. Ambrazas 24 Briefly, these differences are, first, that there are other neutral complementizers in contrast with da, namely če and deto. Second, Bulgarian da allows for a slightly wider range of CTPs than in Macedonian (see fn. 22). 25 For a still valid criticism of the treatment of da-clauses as an ‘analytical subjunctive’ in Bulgarian, cf. Genadieva-Mustafčieva (1970: 23–32).

306

Björn Wiemer

(1995), for a broader Indo-European background cf. Disterheft (1980; 1997). This also explains its diachronically primary usage as a form marking the purpose, or goal, of an action (Haspelmath 1989). However, as a diachronic point of departure, I assume the reconstructed situation in Common Slavic (approx. 4th–7th centuries AD), for which it is commonly assumed that this originally nominal formation had already been integrated into the verbal paradigm. At any rate, this can be observed for the oldest attested stages, such as OCS. (9th–10th centuries AD) and the oldest documents written in East Slavic varieties (from 11th c. AD onwards). For instance, the use of the infinitive after predicates (of verbal or nominal origin) denoting obligation, moral evaluation or the convenience of a given moment (e.g., OCS. nǫžda (estъ) ‘(it is) necessary’, dostoitъ ‘befits’, ugodьno (estъ) ‘suits, fits’) is already widespread at this stage; cf., for instance, Večerka (1996: 97–103). Collocations like these seem to be the most frequent focus of authors working on the diachronic syntax of East Slavic. The infinitive as a syntactically independent head of the clause has been studied to an astonishingly lesser extent. Before starting with an account of the relevant diachronic facts, please note that OCS. can be considered as a predecessor of South Slavic (in particular of Balkan Slavic) in a much more literal and direct sense than is the case for the earliest attested stages of East and West Slavic (which together will be called ‘North Slavic’). The influence of OCS. on written East Slavic (right through to standard Russian) is considerable but nonetheless more indirect than in the case of South Slavic. By now, South Slavic in many respects reflects a dialect continuum with Balkan Slavic as its southeastern part, which shows geographical integrity with a more or less uninterrupted succession of Slavic dialects that lay at the basis of OCS. The latter, in turn, was still very close to the structure of Common Slavic; this justifies regarding OCS. as the diachronic point of departure, or at least gauge value, for later Slavic varieties, even if North Slavic cannot be considered as its direct successor.

4.1 East Slavic, Russian On the one hand, IndInfs have been attested since the oldest documented times of East Slavic (11th century). Borkovskij (1968: 159) pointed out that, in early East Slavic (11th–14th centuries), IndInfs made up the absolute majority of clauses without a nominatival subject (so-called ‘impersonal sentences’), and that about two thirds of IndInfs occurred in main clauses (i.e. not in clausal arguments or adjuncts). According to him, the functional range of the IndInf in East Slavic has

Main clause infinitival predicates and their equivalents in Slavic

307

even increased. On the other hand, the IndInf (with a range of weakly differentiated modal functions) appeared in early East Slavic with the 3SG -form of the copula, mostly in the present tense (estъ), but past tenses (aorist bystъ etc.) were also quite common. Besters-Dilger (1999: 33f.) remarked that the copula played an even greater role than in OCS. However, a perusal of data presented and discussed, for instance, in Borkovskij (1968) does not corroborate this assertion since, in his material on early East Slavic present tense, estъ is almost absent (as for past bylo and future budet see Section 4.1.7). Apart from this equivocal picture of the role of the BE -verb,26 we have to be careful about a commonly assumed link between a certain type of IndInf and the MIHI EST LIBER-construction of predicative possession, which was characteristic of some ancient IE languages. In this construction, the possessor is expressed with a dative, the possessee with a nominative NP. Such a link has been assumed by quite a number of scholars (e.g., by Večerka 1996: 92; cf. Holvoet 2003 for a more critical view). For the time being, I prefer to remain agnostic as to whether the rise of IndInfs (with modal meanings) in Slavic (and Baltic) was a more or less direct consequence of changes which started with the MIHI EST LIBER-pattern, because an evaluation of this complicated diachronic relationship would require another chapter. Such a chapter would be outside the range of the present aim, which is to show that the use of infinitives as highest nodes (heads) of main (independent) clauses can, for the oldest retraceable periods of Slavic, be regarded as an original structure. For this purpose, it is of secondary concern as to whether this structure, in turn, arose from any kind of biclausal pattern in times prior to first written attestations of Slavic (which would be another diachronic point of departure). As a matter of fact, constructions with an infinitive and a dative NP denoting a potential agent (potAgDAT ) [1]

(ESSE –) INF – potAgDAT

are well-attested as early as in OCS. and the earliest written documents of East Slavic. As we will see, there is no reason to doubt that the widespread use of the dative for the potential agent in the IndInf (= [1]), which we discussed in Section 3.1.2 for modern Russian, can be considered as a conservative feature as old (at least) as the earliest documents written in some form of Slavic. The modal 26 To my knowledge, the status of byti ‘be’ has remained a topic of disputes, namely: Whether it should be characterized as an existential (“substantive”) verb, as a copula, or rather as a tense-mood marker. I thank one of the anonymous reviewers for having brought this issue to my attention.

308

Björn Wiemer

(deontic, dynamic, rarely dispositional) meaning of construction [1] can be explained as resulting from another conservatism, namely: the goal/purpose semantics of the infinitive which it inherited from the dative of the verbal noun and which has persisted into our days in Russian (as well as in Baltic). At first sight, a complication might arise from examples in which the infinitival clause contained a nominatival NP (which concomitantly might have given rise to the view that this construction developed out of predicative possession of the MIHI EST LIBER-type). Examples have been frequently adduced since Potebnja’s late 19th century works. Here is one of them originally given in Potebnja (1958: 406) and adduced by Holvoet (2003: 467) together with his translation (notice that a copula is lacking): (64) Takova pravda uzjati rusinu. such.NOM . SG . F right.NOM . SG . F take:PFV. INF Russian.DAT ‘These are the rights a Russian must (should, can?) enjoy.’ (64) is from early East Slavic, Večerka (1996: 93f.) gave analogous examples for OCS. From a semantic viewpoint, (64) represents a hybrid construction: In one interpretation, the nominatival NP looks like a possessee of the dative-NP rusinu (see [2a]), but it can also be interpreted as the undergoer (“semantic object”) of the infinitive (see [2b]). In either case (takova pravda uzjati ≈ ‘such rights are to take’ and takova pravda rusinu ≈ ‘such rights are for/belong to a Russian’) the result is a modalized utterance, with possibility or necessity determined only by context. In the latter case, the infinitive constitutes the predicate of the clause, in the former case it should be taken as a goal-adjunct of pravda ‘right’. Now, regardless of whether (64) can be understood as representing predicative possession, the relation between the infinitive and the nominatival NP as a predicateargument relation (with the nominatival NP expressing the undergoer) must have resulted from reanalysis. In other words: In order to assume reanalysis, one need not assume that (64) continued the IE. MIHI EST LIBER-pattern of predicative possession. [2a–b] pictures reanalysis in two variants: The dependency relations on the left side assume that no BE -verb need be postulated (or that it functioned only as a tense-mood marker; see fn. 26), the illustration on the right side implies such an existential BE -verb as the highest node (head) of the clause. Likewise, an explanation in terms of reanalysis does not depend on whether one wants to assume a BE -verb (often realized as zero) or not:

Main clause infinitival predicates and their equivalents in Slavic

309

[2a]

[2b]

Subsequently, this reanalysis became manifest in a case marking shift, which “actualized” the newly interpreted argument relation (as undergoer) of the nominatival NP to the verb (infinitive): NOM → ACC (rusinu uzjati takovu.ACC pravdu. ACC ). This shift happened except in those parts of the East Slavic territory, in which the Nominative Object established itself more firmly, i.e. the region of Pskov-Novgorod and north to it (cf. Mendoza 2008 for a recent comprehensive account). We will not consider it here anymore because, to the best of my knowledge, the Nominative Object has been of no importance to the history of the IndInf in the remaining parts of Slavic. We should realize that the existential verb (byti) – if one wants to postulate its reality –in the construction [1], or [2b] could still be considered as the syntactic head of the clause, with the infinitive as its dependent. With the disappearance of this verb, the infinitive acquired the status of the highest clausal node. Otherwise, if there was no such existential verb – or byti had to be regarded only as a tense-mood marker – the infinitive was the clausal head from the start, i.e. already in the earliest attested times. Whatever syntactic theory is followed, it is crucial to understand that nothing in the whole process can sensibly be qualified as insubordination. The reanalysis occurred in the confines of a simple sentence. As for the loss of the BE -verb, one has to ask whether a structure like OCS. in (65), cited from Miklosich (1926: 859) (65) ašte mi estъ sъ tobojǫ umьrěti . . . if 1SG . DAT be.PRS .3SG with 2SG . INS die:PFV. INF ‘if I have to die with you. . .’ (Mark 14, 31)

310

Björn Wiemer

which corresponds to [1], can be considered as biclausal. It can, in a wider sense, be treated as complementation if we see the BE -verb as taking a dativeNP and an infinitive as its arguments. This accepted, no reduction of a biclausal to a monoclausal structure occurred, only a former argument (for that matter) “advanced” to the head of the whole clause after its own assumed former head had vanished. Two other details remain to be accounted for. First, as Večerka (1996: 93– 105), among others, pointed out, in OCS. the pattern [1] with the copula was sustained by “impersonal” verbs in the 3SG -form (e.g., podobaetъ ‘befits’, dostoitъ ‘suits’) or by nominal predicates (e.g., podobьno ‘suitable’) which carried a modal (moral etc.) or other evaluative (‘good – bad’) meaning and thereby emphasized the deontic or dynamic character of the utterance.27 These modal predicates either combined with byti (see ex. (66), něstъ < ne+estъ ‘NEG + is’),28 or they replaced it (see ex. (67–68)). Second, the place of the copula byti ‘be’ was often occupied by other, semi-copular verbs; see ex. (69), which is a variant of ex. (65): (66) o mьně bo něstъ trěbě plakati sę. about 1SG . LOC since NEG .be.PRS .3SG need.GEN cry:IPFV. INF RM ni milovati nor love:IPFV. INF ‘there is no need to cry because of me, or to love (me)’ (from Su 102, 15–16) (67) ašte dostoitъ °člku pustiti ženǫ if suit:IPFV. PRS .3SG man.DAT abandon:PFV. INF wife.INS svojǫ REFLPOSS . INS . SG . F

‘if it suits a man to divorce with his wife’ (Matthew 19,3; from MarEv) (68) nužda dьnesь sъtvoriti obědъ need.NOM today create/make:PFV. INF meal.ACC ‘today there is need to prepare a meal’ (from Su 405, 17–18)

27 Neuter participles of speech act verbs (e.g., pověleno bě ‘(it) was ordered’) also belonged here. These “impersonal modal” predicates were already mentioned in the introduction to this section. The chronological relation between them and construction [1], based just on the BE -verb, remains to be clarified. 28 The genitive form trěbě (from trěba.NOM ) can be treated as “subject” argument of byti, but then the infinitive must be analyzed as an NP-internal modifier of trěbě.

Main clause infinitival predicates and their equivalents in Slavic

311

(69) ašte mi sę ključitъ s tobojǫ umьrěti. if 1SG . DAT RM happen:PFV. PRS .3SG with you(SG ).INS die:PFV. INF ne otvrъgǫ sę tebe ‘if I am bound to die with you, I won’t disavow you’ (Mark 14, 31; from MarEv) The same applies to early East Slavic. Some of these predicative items later developed into modal auxiliaries in different Slavic languages (cf. Hansen 2001; Besters-Dilger et al. 2009) and, if they did, began, in a sense, to compete with the IndInf in Russian. However, the rise of modal auxiliaries is a different diachronic process, which we need not dwell upon here. This is predominantly because auxiliarization consists in the evolution of complex predicates in monoclausal structures out of structures that should rather, in their inceptive stages, be analyzed as being biclausal. (This is what happened to da-clauses, analyzed in Section 4.3.2.) Of import to our present concern is the fact that predicative structures like those in (66–69) probably arose after a modal infinitive construction with a copula had appeared; in fact, the former presuppose the latter as a model to be built upon. Furthermore, the subsequent loss of byti ‘be’ (at least of its present tense form, estъ) with the IndInf does not testify to changes in constituency structure; more properly, this loss has to be seen in the larger context of the present tense copula’s general retreat in East Slavic (partially also elsewhere in Slavic). This loss is also indicative of byti being reinterpreted as a proper copula verb (not an existential predicate). 4.1.1 Continued development Let us now turn to the subsequent evolution of the reanalyzed construction [2a] in East Slavic. Clauses with the infinitive and a potAgDAT were widespread in directive utterances already at the earliest attested stages (examples cited from Stecenko 1972: 14f.): (70) a dvorjanomъ tvoimъ, knjaže, xoditi and noblemen.DAT your.DAT. PL duke.VOC walk:IPFV. INF po pošline over tradition.DAT ‘and your noblemen should behave as it used to be’ (Dog. gr. Novg. 1264 g.)

312

Björn Wiemer

(71) a knjazju velikomu dьržati Novъgorodъ bezъ without and grand_prince.DAT hold:IPFV. INF PN . ACC obidy violation_of_law.GEN ‘and the grand prince ought to hold Novgorod without law being violated’ (Dog. gr. Novg. 1316 g.) However, the potAgDAT was by no means a mandatory feature, although the clear majority of IndInfs included a potAgDAT (Borkovskij 1968: 159–164). It could be lacking particularly in deontic utterances without specific referents. This happened in lawbooks like the “Russkaja pravda” (Borkovskij 1968: 159–164; 1978: 279). Particularly in texts of a legal predestination, we find an infinitive clause as a kind of apodosis of a conditional sentence; the link to the would-be protasis is rather loose, e.g. because the latter came as a headless relative clause. Compare the following example from the “Russkaja pravda”: (72) ože li ne budetъ kto ego mьstę, NEG be.FUT.3SG anybody.NOM him.ACC revenge:PRS PA.NOM . SG . M if to položiti za golovu 80 grivenъ grivna.GEN . PL then put:PFV. INF for head.ACC ‘and if there won’t be anyone to revenge him, then one has to pay [lit. put] 80 grivnas for each person’ The IndInf with a meaning of obligation (= deontic necessity) was also prominent in juridical documents at later stages. It is well attested, for instance, in the 16th century. Furthermore, the IndInf was used, although probably much more rarely, to mark an imminent future, probably as an extension of the necessity meaning (for which deontic vs. dynamic use was only vaguely distinguishable). Compare an example from a testament cited in Bulaxovskij (51958: 351): (73) Se jaz (. . .) pišju siju gramotu duševuju v konce života. a bil mja Mixajla Skobel’cin bol’šoj s svoimi ljud’mi. . . a bil mja u svoego sela, a pojti mi s ix ruk, a dolgu and go:PFV. INF 1SG . DAT from their hands.GEN and debt.DAT mi dat’. . . 1SG . DAT give:PFV. INF

Main clause infinitival predicates and their equivalents in Slavic

313

‘Now I’m writing this testament at the end of my life, and Mixajla Skobel’cin the Great with his men beat me. . . and he beat me at his village, and I will die from their hands [more lit. ‘and I have to go from their hands’] and let (God) my debt’ (“Duxovnaja Pankrata Čeneja”, 1482). Moreover, IndInfs occurred in WH- and yes/no-questions. The material in textbooks is scarce and looks somewhat haphazard, but it shows that questions with IndInfs were also used rhetorically (or ironically), as in (75) (both examples cited from Stecenko 1972: 12f.): (74) Ot čego mi estь umreti? from what.GEN 1SG . DAT be.PRS .3SG die:PFV. INF ‘From what I will have to die?’ (Povest’ vremennyx let, 42) (75) (. . .) I posmejasja reče: otъ sego li l(ъ)ba smьrtь bylo vzjati mne? from this.GEN Q head.GEN death.ACC be.PST. SG . N take:PFV. INF 1SG . DAT i vъstupi nogoju na lobъ ‘And he said with laughter: I had to kill [lit. take death from] this head? And he stepped with his foot on the head.’ (Povest’ vremennyx let, 30) Note that such utterances do not need any further linguistic context (the second sentence in ex. (75) is inserted into direct speech and does not belong to the narrative thread). 4.1.2 The role of by Documentation is scarce on the usage of by (sometimes reduced to b) in infinitival clauses. It has been claimed, however, that this structure was much more frequent before the 19th century (cf. Bulaxovskij 51958: 350, among others). According to Borkovskij (1968: 164–166; 1978: 283), INF+by from the 15th century onwards established itself more firmly as a construction indicating obligation or an optative meaning. He eventually concluded that the optative meaning prevailed over the deontic one. A perusal of the examples leaves the impression that a distinction between deontic and optative use could be discerned only from knowledge of a larger (linguistic or situational) background. Compare one of Borkovskij’s examples:

314

Björn Wiemer

(76) i my totъ listъ zarazъ kъ vamъ otsylaemъ, a vamъ by ego poslomъ and 2PL . DAT IRR him.ACC envoy.DAT. PL Moskovskogo gospodarstva zarazъ že otdatь immediately PTC hand_over:PFV. INF Moscovian state.GEN ‘and we immediately send this letter to you, and you should give it (away) to the envoys of the Moskovian state’ (Orš. gr. 1615 g.) It is a requirement for further, corpus-based research to disclose the conditions under which deontic and optative uses of INF+by in earlier stages of East Slavic (and probably other Slavic varieties) became differentiated, in light of the contemporary use in Russian which shows by in a special optative construction (see Section 3.2). In the 19th c. apprehensive utterances (77) occur alongside optative usage (78):29 (77) Ne

raskajat’sja b potom. NEG regret:PFV. INF IRR afterwards ‘If only I/we don’t have to regret it later.’

(78) Grafine tol’ko by kupit’ čto ni popalo. duchess.DAT only IRR buy:PFV. INF whatever.ACC ‘The duchess would like to buy whatever she comes across.’ Apart from this issue, INF+by also occurred in the apodosis of conditional sentences: (79) Kaby ne ty, sidet’ by mne i teper’ tam. if_only NEG 2SG . NOM sit:IPFV. INF IRR 1SG . DAT even now there ‘If not you, I would (have to) sit there even now.’

29 Examples (77–78) are cited from Švedova (1964: 338). Her comments on examples like (78) do not clarify whether the subject of volition had to be the speaker, or whether it could also be the referent coded with the dative (= grafine ‘duchess’) and identical with the potential agent. It would be intriguing to clarify whether, and under which circumstances, it has been possible to use this optative construction if the desired event was not to happen with the speaker but rather the addressee or a non-participant of the actual speech act.

Main clause infinitival predicates and their equivalents in Slavic

315

As was argued in Section 3.2, there is no reason to assume that INF+by as optatives appeared as main clauses later than they did in complex sentences, i.e. as the apodosis in a kind of “truncated” conditional sentence. Rather, complex sentences as in (79) should be considered as a phenomenon diachronically posterior to sentences such as those in (77–78). 4.1.3 The fate of the copula In later periods, IndInfs gained even more territory. The present tense copula gradually retreated until, by the 19th century, it became practically extinct. Simultaneously, by the 19th century the potAgDAT became practically obligatory (Švedova 1964: 337–346). The result was the ‘modal infinitive’ analyzed at length in Section 3.1.2. As argued for above, this construction already existed from the earliest written period of East Slavic. On account of statements by different scholars, we may, however, surmise that its particular components had not been stable for many centuries. It would be elucidating to clarify the ups and downs of the potAgDAT during the history of East Slavic (and possibly other Slavic varieties), but whatever more reliable philological and corpus-based investigations of the available material might bring to light in this respect, it would not change the basic conclusion that the IndInf is diachronically primary and not an instance of insubordination. There are more points to make about the copula. As just mentioned, its present tense form vanished (as elsewhere in Russian), but the forms of the past (neuter, bylo) and the future (budet) with the IndInf have undergone specific restrictions, which, to my knowledge, have so far not been reconstructed and described in sufficiently reliable detail. I will restrict myself to some basic facts. To begin with, at least during the 18th century, the collocation of bylo with an IndInf is claimed to have been able to carry more than just a modal function; instead, its function could be “indicative” as well. Švedova (1964: 340) quotes Lomonosov30 who, in his grammar (from the mid-18th century), described bylo in combination with the infinitive of an imperfective verb as a means to express ingressivity. Thus, a sentence like Mne bylo pisat’ could be paraphrased as Ja stal pisat’ ‘I started writing’. Curiously, Lomonosov added that, if the order between bylo and the infinitive was changed (pisat’ bylo), the speaker expressed regret to not have performed the action named by the infinitive. This is a counterfactual function that is close to what was coined ‘antiresultative’ by Plungjan (2001). In this construction, the petrified copular form bylo plays a key role, too. Contrary to the construction INFI P F V +bylo, however, bylo combines with a verb 30 Cf. also Bulaxovskij (51958: 352f.) for further details.

316

Björn Wiemer

(usually perfective) in the past tense in the antiresultative, i.e. with a finite form of the opposite aspect. It would be interesting to learn whether there is an historical connection between both constructions with bylo, and whether the similarity between the construction described by Lomonosov and the counterfactual pattern bylo+INF discussed in Section 3.1.3.2 for Polish and Serbian-Croatian was accidental or derived from shared Common Slavic roots. These questions, however, go beyond the scope of this article. At present, let us be satisfied with the remark that the counterfactual function noticed by Lomonosov persisted into the 19th century, and has done so even to the present day, obviously including the distinctive function of word order mentioned by Lomonosov (see ex. (80–81)), according to Švedova. (80) Ne

vrat’ bylo mnogo, tak by pošel ne s takim nosom. NEG lie:IPFV. INF be.PST. N much

‘You shouldn’t have lied that much, (then) you would go now in a better mood’ [lit. not with such a nose]. (Lukin: „Ščepetil’nyj“, 8) (81) Xvat’ v kolčan. . . , an strel uže netu: Luk opuščen; stal ja v pen’. Ax, bereč’ bylo monetu Beluju na černyj den’! save:IPFV. INF be.PST. N money.ACC white.ACC . SG . F on black day.ACC ‘I seized into my quiver. . . , but there weren’t any arrows left: the bow put down, I stood up as if rooted to the ground. Oh, if only I had put money by for a rainy day!’ (Deržavin: “Strelok”) Švedova (1964: 342), the source of examples (80–83) from the late 18th–early 19th c., differentiated between this function – more precisely: the indication of the non-occurrence of a desirable event – and another one which she named ‘meaning of unsure decision’ (Russ. značenie neuverennogo rešenija), see ex. (82–83): (82) Popytat’sja bylo sprosit’ u nee? try:PFV. INF be.PST. N at:PFV. INF at her.GEN ‘Should I have tried to ask her?’

(A. Plavil’ščikov: „Bobyl‘“ I, 4)

(83) [Fineta:] Podi-t-ka lutče da poigraj v svajku. . . [Fatjuj:] Tak i vprjam’ poigrat‘ bylo ot skuki. thus straight_away play:PFV. INF be.PST. N from boredom.GEN ‘[Fineta:] Come better go and play svajka. [Fatjuj:] Well, I’d/we’d (indeed) better play out of boredom.’ (Sumarokov: “Ssora u muža s ženoj”, 5)

Main clause infinitival predicates and their equivalents in Slavic

317

This usage to mark unsure decisions comes, in turn, close to deliberative questions (see ex. (82)) or to its, as it were, assertive counterparts: Mild suggestions implying that the speaker (or another subject of judgment) is not quite certain as to the appropriateness of his/her advice. Especially in this “hedged” assertive usage, the IndInf in combination with the past neuter copula easily acquires an air of irony (see ex. (80)), among other things in echoic replies (compare with ex. (6) in Section 3.1.1). The non-factive implication borne by counterfactivity arises due to the past copula bylo; and I leave it open to further research as to whether or not the two functions distinguished by Švedova are subtypes under a common umbrella type. Regardless, we should realize that the common denominator of all the more specific usage types of INF+bylo mentioned above consists in the speakers distancing themselves epistemically and/or emotionally from the propositional content of the utterance: It is either denied, or the speaker mildly questions whether the action is worth undertaking. Note that, in the latter case, the utterance would not carry a proposition. Interestingly, a kind of mirror image of the counterfactual use of INF+bylo surfaces in a usage type which, following Švedova (1964: 340f.), should be characterized as properly modal. More precisely: Utterances of dynamic or deontic necessity are well attested based on a play on negative polarity; compare:31 (84) No delat’ bylo nečego,‒ ne načat‘ že bylo žalovat’sja na tarakanov. NEG begin:PFV. INF PTC be.PST. N complain:IPFV. INF on cockroach.ACC . PL ‘But there was nothing to be done about it – after all, there was no point in beginning to complain / should I have started complaining about the cockroaches’ (A. Gercen: „Byloe i dumy“ 2, XI). ⊃ the speaker did not complain

31 Here, Švedova remarks that, in the 19th century, INF+bylo could also be combined with the clitic by; compare one of her examples: Dva dni bylo by vam tol’ko podoždat’, i vy by polučili vaši leksikony ‘You would have had to wait only two days, and you would have received your dictionaries’ (N. Gogol‘ in a letter to P.P. Kosjarovskij, Sept., 9th, 1828). However, this is arguably not a combination of [INF+bylo]+by, but rather of INF+[bylo by], i.e. of the subjunctive of byt’ ‘be’ used as an auxiliary with the IndInf (podoždat’ ‘wait’) to simply mark mood. At least utterances like the quoted one are (without further knowledge of the context of speech) ambiguous since they can be read either way: as “ordinary” subjunctives (without a distinct potentialis–irrealis opposition), or as optatives with a reference interval prior to the moment of utterance (‘I wish / would be glad if you (can / accept to) wait’ + past reference). Obviously, Švedova wanted to have it the latter way.

318

Björn Wiemer

(85) Da PTC

neužto že mne ego zagubit’ bylo? really PTC 1SG . DAT he.ACC ruin:PFV. INF be.PST. N

‘After all, did I really have to ruin him?’ (F. Dostoevskij: „Idiot” 1, XVI). ⊃ the speaker did not ruin the other person’s life Here, bylo merely serves to transfer the reference interval to the past (prior to speech), it does not partake in the modal character of such utterance but is simply a tense marker (see also fn. 31). This interpretation is corroborated by the fact that sentence negation occurs with the infinitive, not with bylo, as in32 (86) Dlja čego bylo ne skazat’, for what.GEN be.PST. N NEG say:PFV. INF kotoryj iz carevičej plenil ee serdce? ‘Why should I not have said which one of the princes captived her heart?’ (N. Karamzin: „Prekrasnaja carevna i ščastlivyj karla“) (87) Kak bylo ne vspomnit’ i ne požalet’ how be.PST. N NEG remember:PFV. INF and NEG regret:PFV. INF o tex dnjax, kogda . . . ‘How should I not have remembered about and regretted those days when. . .’ (S. Glinka: “Zapiski o M.”, 6) This purely temporal function of the copula is attested with its future form budet, too. INF+budet already occurred in the oldest stages without any discernible functional differences to usage in the 18th–19th centuries (cf. Bulaxovskij 51958: 354f.). The examples given in Švedova (1964: 343f.), among other sources, make one inclined to assume that the IndInf has always occurred with budet in a compositional manner, i.e. budet did not, and does not, combine with the infinitive to create an holistic unit. Rephrasing this slightly: Budet did not, and

32 We observe the same illocutive effects arising from negative polarity such as in (80–81), but this effect has to be judged from the speaker’s point of view. Otherwise we would not be able to explain why (82–83), without context, do not allow the inference of whether the respective action was carried out or not: Did somebody tell who adored the princess? And: Have those days been remembered and felt sorry for? Illocutive negative polarity effects arise only if we consider what the speaker finds desirable; this wish may conflict with what might have happened (or not happened) in reality. In (82–83), this conflict does not arise because a firstperson narrator wrote these examples.

Main clause infinitival predicates and their equivalents in Slavic

319

does not, contribute to the modal (deontic, dynamic) character (or related conditional or “unreal” meanings) of clauses headed by infinitives. This is corroborated by a remark in Borkovskij/Kuznecov (21965: 419) that, in earlier stages of East Slavic, the dative NP denoting the potential agent with budet+INF had to be analyzed as an argument of the infinitive; this pattern survived even into the standard speech of the 19th century, although it was regarded as a feature of folk speech. Here are examples from Švedova‘s account: (88) [Leporello:] Prokljatoe žit’e. Da dolgo l’ budet mne s nim vozit’sja? PTC long.ADV Q be.FUT.3SG 1SG . DAT with he.INS toil:IPFV. INF ‘[Leporello:] Damned life. Shall I really have to toil/struggle with him for a long time?’ (A.S. Puškin: „Kamennyj gost’“, I) (89) Eželi, k neščastiju, otkroetsja vojna, istinno ne znaju, čto mne budet delat’. NEG know:IPFV. PRS .1SG what.ACC 1SG . DAT be.FUT.3SG do:IPFV. INF ‘If, unfortunately, war begins, I really don’t know what I (will) have to do.’ (A.M. Kutuzov in a letter to N.N. Trubeckoj, March, 22nd–April, 2nd, 1791) (90) Nevozvraten sčastlivyj mig. Ne doždat’sja vam budet, NEG wait:PFV. INF you(PL).DAT be.FUT.3SG poka povtoritsja blagoprijatnoe sočetanie obstojatel’stv. ‘The happy moment cannot be returned. You won’t succeed waiting until a favorable coincidence of circumstances will repeat.’ (N.G. Černyševskij: “Russkij čelovek na rendez-vous”)

4.2 West Slavic: Czech and Polish In early Czech (14th century), the construction [potAgDAT – COP – INF] was quite widespread; the copula as well as potAgDAT was often lacking, so that the construction was reduced just to the infinitive. There was no discernible functional distinction motivating the contrast of absence vs. presence of the copula (Porák 1967: 17). The construction predominantly marked necessity in an either dynamic or deontic reading. The fact that this reading was genuinely Slavic can be inferred from the IndInf occurring systematically in translations from German or Latin as an equivalent of modal auxiliaries (in a large sense; see ex. (91)), of

320

Björn Wiemer

the Latin gerundive (on -ndus), in direct speech as equivalent of the imperative (see ex. (92)), or, rendering it more explicit, as a modal implicature from a fixed collocation with a finite verb (see ex. (93)); cf. Porák (1967: 18–24): (91) tomu

DEM . DAT

nám jest kněžstva pomáhati 1PL . DAT be.PRS .3SG duchy.GEN . SG (?) help:IPFV. INF

‘we have to help him (to get) the duchy’ (Dal 98; L) MHG original: (rhymed) wir suln in zcum horczogtum entpurn (Jir. 11); (prose) dem wolle wir das lant eingeben (Jir. 278) (92) jemuž on odpovědě a řka: Chléb s vodú osoléc jiesti (. . .). bread.ACC with water.INS salt.ACC eat:IPFV. INF ‘He answered him by saying: one/you ought to eat bread with water and salt.’ Latin original: Cui ille respondens dixit: Comede panem tuum cum sale et aqua. (93) a dyž bieše čas přišel opatu Sizojemu, když mu bieše s tohoto světa jíti he.DAT be.IMPF.3SG from this.GEN world.GEN go:(I )PFV. INF ‘and then the time came for abbot Sizoj when he had to pass from this world’ Latin original: Abbati Sisoio cum tempus dormitionis eius advenisset The IndInf was already becoming increasingly obsolete by the 15th century (it reappeared in the 19th century due to the Czech national revival, obrození). A similar conclusion can be made for Old Polish. In West Slavic the space was more and more taken over by uprising modal auxiliaries (Porák 1967: 54f. and elsewhere). As Porák argued, mainly on the basis of a thorough analysis of his Old Czech corpus, infinitival clauses with interrogative or directive functions (questions, commands, prohibitions) can be treated as derived from the construction just discussed. The IndInf was much more frequent in WH-questions than in yes/no-questions; the frequency of the latter only increased in the 18th century (Porák 1967: 86, 93). Optative clauses were infrequent as well, but Porák considered them to be variants of the basic declarative construction modified by the conditional mood, an ‘optative particle’ or both. Svoboda (1988: 30) argued that optative uses appeared later than other illocutionary ones. However, the

Main clause infinitival predicates and their equivalents in Slavic

321

conditional itself consisted of an optative particle, namely by (as in East Slavic, see Sections 3.2 and 4.1.2), and the few examples provided by Porák (1967: 95) raise the question of whether the IndInf in optative use derived from conditional clauses or not. Compare, for instance, (94) Ké

kdy slyšeti, by řekli. IRR when hear:IPFV. INF IRR say:PFV. PST. PL ‘If somebody is able to/could hear, they would say.’ (ŠtitBrig, 1419; Výb. 714)

(95) Kéž vás pobrati na káry a vésti IRR you(PL).ACC take:PFV. INF on punishment and bring:(I )PFV. INF až na rozhraní, kde-ž by-ste byly vmetány until on border when.PTC IRR .2PL be.PST. PL thrown.PL na oheň nebo do vody, žádné by nebylo škody! ‘If one took you(PL) for punishment and brought you(PL) to the very border, if you(PL) were thrown into fire or into water, there would be no harm!’ (Tragedie, 1573; Stč. drama 243) (96) Ach, by IRR

mi bylo umřieti. 1SG . DAT be.PST. N die:PFV. INF

‘Ah, I’d be better off dead.’ (Baw 227; 1472) (97) Á, by

teď bylo ještě pobyti IRR then be.PST. N even be/remain:PFV. INF

živu sto let! alive.DAT. SG . M 100 years ‘Ah, then it would have been good to remain alive for (another) 100 years!’ (RokPost I, 82; B, 1503) The clauses beginning with ké(ž) in (94–95) can be read as the protasis of conditional sentences, whereas in (96–97) the clauses are isolated and used like exclamations. One thus wonders whether the usage in (96–97) should not be treated like conditionals with an ellipsis of their protasis. However, it is no less justified to turn this question on its head: How did conditionals arise? A probable answer might be that conditionals established themselves via the conventionalization of a pattern in which clauses with an optative particle

322

Björn Wiemer

(ké, by) cooccurred in juxtaposition with other clauses marked with by. Following this perspective, the rise of conditionals would be explained as the product of syntactic tightening of erstwhile loosely associated clauses for which the semantic conditions of conditionals hold. As the data stands, there is no reason to assume that isolated optative clauses with ké or by in Old Czech have to be treated as elliptic conditional sentences. The same, by analogy, applies to Polish oby, probably the closest equivalent of Czech ké(ž), for which, however, I am unaware of any conclusive diachronic investigations. In general, the historical material does not allow for any sound hypotheses concerning the chronological relation between the rise of non-declarative clause patterns with the infinitive and other illocutionary types in which the IndInf became prominent. Of course, we cannot assume any immediate link to the development in East Slavic, either. However, when it comes to answering the question of the IndInf’s historical primacy in Slavic as such, the West Slavic material, again, does not supply any indication of the IndInf being a secondary development. Non-modal IndInfs are investigated less thoroughly than modal ones. This holds true in particular for patterns associated with the narrative type discussed in Section 3.1.1. Svoboda (1988: 322, with reference to Porák 1967) pointed out that, in Czech, clauses built on the infinitive and a nominatival subject as in (98) pes jen štěkat dog.NOM PTC bark:IPFV. INF ? ‘the dog couldn’t / wouldn’t stop barking’ appeared before the 19th century; but he also admitted that they entered into the standard (written) language quite late. Unfortunately, he only presented isolated examples without a broader context so that, without a systematic investigation of diachronic text corpora, nothing conclusive can be said about an imaginable narrative function of such clauses. Obviously, this clause pattern has not found its way into any established usage. Informants do not readily, or in some cases at all, accept such a sentence (at least in isolation). If they marginally accept it, they tend to interpret it not with an ingressive meaning, but rather as indicated in the English translation of (98), i.e. as a continuous action that cannot (or does not want to) be stopped by its agent. This interpretation comes much closer to being modal (dynamic) or volitional rather than ingressive, which appears to be the predominant interpretation in Russian (see Section 3.1.1).

Main clause infinitival predicates and their equivalents in Slavic

323

Similar remarks can be made for Polish. Pisarkowa (1984: 37f.) supplied some relevant material from 16th–17th century Polish for which she claimed non-modal use of IndInf. Most of her examples, however, do not show this, as they are, with few exceptions, non-assertive. Some of them can be interpreted either as habitual-consecutive clauses (see ex. (99)), or they occur in the apodosis of temporal-conditional sentences (see ex. (100)): (99) Przyjdzie potrzeba jaka, dopiro sie wiercić. RM fidget:IPFV. INF come:PFV. FUT.3SG need.NOM some.NOM only ‘Some need will come, and you will fidget at last.’

(RejZwierc 61v)

(100) To szczęście nasze dobrze ktoś szkłu przyrownał. . . Kiedy sie z nami najbardziej pieści, to nam w te czasy po gębie da, kiedy nas najlepiej piastuje, dopiero się strzec. when 1PL . ACC best nurse:IPFV. PRS .3SG only RM beware:IPFV. INF ‘Somebody compared our fortune with glass. . . When it takes care of us most, at that time it gives us a smack. When it carresses us most, you particularly have to be on your guard.’ (Pasek, PisM 32) The common denominator is that the IndInf was tightly associated with nonfactive contexts or contexts with lowered temporal referentiality. In yet other cases, the IndInf carried a non-factive implication when it was introduced by że, as in (101) I dopieroż powstał tumult. . . że tylko oczy zamknąć, uczy [sic!] zatkać i PTC only eyes.ACC close:PFV. INF ears.ACC choke:PFV. INF and na dziesiątą uciekać miedzę. on tenth.ACC . SG . F run_away:IPFV. INF boundary.ACC . SG . F ‘And then such a turmoil arose. . . that you’d / you could better close your eyes, stop your ears and run far away.’ (ŁozSzar I 38) (102) płynęła rzeczka (. . .) żwirem wysypana, że tylko liczyć, tylko zbierać jej ziarna PTC only count:IPFV. INF only collect:IPFV. INF her grain.ACC . PL ‘a brook (. . .) filled with gravel, that you/one could only count and collect its grains’ (ŻmichPog 46)

324

Björn Wiemer

Is this a description of what the speaker (or another subject implied by the context) really performed (i.e. as sort of consecutive clause)? Or are the że+INFclauses to be understood as an encouragement that the described action was suitable or recommended to be performed? In the latter case, the reading is non-factive and comes close to dynamic or dispositional modality (‘X is able to SoA / depends on doing SoA’). This usage has persisted until today (on a colloquial level). Among the very few examples of IndInfs which really could be interpreted as parts of narrative chains, I have found only one with a nominative subject (see ex. (103)); in all other, presumably narrative cases, the subject remains implicit (see ex. (104)): (103) począłem rybę jeść, aż w niej okrutnie smak dobry, nuż ja ją jeść PTC 1SG . NOM her.ACC eat:IPFV. INF ‘I began to eat the fish, for its taste was so terribly good, so, well, I began to eat it’ (Pasek 84) (104) Więc że kiszka przecięta była, obaczył jeden czerwony złoty, dalej szukając znalazł więcej: dopieroż innych pruć. only other:ACC . PL rip_up:IPFV. INF ‘Thus, since the bowel was cut across, he saw a red zloty. When he continued searching, he found more: particularly (then) he started ripping up the other ones.’ (Pasek 57) IndInfs with a modal (usually dynamic) reading are attested as more or less widespread in Polish until at least the 18th century; afterwards, they gradually began to die out, albeit more slowly if under negation (Pisarkowa 1984: 34, 36), as in (105): (105) wolności nie kupić złotem freedom.GEN NEG buy:PFV. INF gold.INS ‘one cannot buy freedom with gold’ Concerning the issue of insubordination, again, the material on earlier stages of Polish does not sustain any claim that IndInfs are historically secondary. The structures we observe by the 18th century and in part even today (see Section 3.1.3.1) can arguably be considered as the continuation of a pattern that came into existence already in late Common Slavic times.

Main clause infinitival predicates and their equivalents in Slavic

325

4.3 Balkan Slavic: the spread of da-clauses The processes by which the infinitive as a morphological category was lost gradually in the eastern part of South Slavic, i.e. in Balkan Slavic proper, are well known and described; cf., among many others, Joseph (1983) and Asenova (2002: 141–148). The functions of infinitival clauses were, by and large, taken over by clauses which are introduced by some sort of connective, primarily by da (see below), and whose predicates show tense and agreement categories, thus the typical morphological features of finiteness. The choice of these features is, however, as a rule, restricted in comparison to indicative main clauses (see below).33 As the closest equivalents to infinitival clauses in general, and of IndInfs in particular, we have to, therefore, focus on the development of dependent and independent da-clauses in Bulgarian and Macedonian against the background of South Slavic in its entirety. I first try to give a brief, but comprehensive account of the history of da in Slavic (Section 4.3.1) before I come to the diachronic relationship of independent and main (or simple) da-clauses (Section 4.3.2). 4.3.1 Origin of da and its subsequent distribution over Slavic The etymology of da has been the object of some disputes. Most specialists seem to abide by the view that da had its roots in the ablative form of an originally IE demonstrative (*dōd.ABL), from which it developed into a kind of particle with the meaning ‘from then onwards’, assumed for Common Slavic (cf. GrkovićMajor 2004, among others). Večerka (1993: 81; 1996: 29) considers it most likely that da descended from an interjection, a consideration which would have parallels in other IE languages derived from IE demonstrative t-roots (e.g., ancient Greek δἠ, Lat. do-nec, quan-do); cf. also Genadieva-Mustafčieva (1970: 15). This particle or interjection, in turn, was the basis on which da started its career as a uniquely versatile connective in South Slavic (see below). An alternative view was expressed by Gołąb (1984: 171). On the basis of some authoritative etymological dictionaries, he argued that da descended from the imperative (singular) *dadjь! of the Proto-Slavic verb *dati ‘give’. This etymology would neatly explain why da, irrespective of its syntactic status, occurred first with an optative or hortative34 function. However, Gołąb (1984: 179) claimed that the verbal origin of da applied only to its use as a particle (“mood marker”), while 33 This fact has often served as a basis to dub da+Vfin-clause ‘analytical subjunctives’ (see Section 3.3). 34 Consider the generally well-attested path from ‘give’ to ‘let’ (e.g. in analytical causatives).

326

Björn Wiemer

the homonymous complementizer (his “conjunction”) derived from the demonstrative pronoun. We are thus faced with the problem not only of alternative etymological explanations, but even with the possibility of a merger from two different sources. After all, although we may not exclude that the pre-Slavic history of da had two different sources which happened to converge (via homonymy?) at some Proto- or Common Slavic stage, we can safely assume that, at the time of OCS., being the first Slavic variety with written documentation, da was used as one clitic morpheme for which any of its presumable etymological origins had become obliterated anyway. It seems, therefore, justified to treat the stage attested in OCS. and the earliest writings in East Slavic varieties (however influenced by OCS. or not) as a diachronic starting point. Judged from this starting point, a split between South Slavic and North (= East and West) Slavic becomes obvious. Through the entire history of South Slavic, da has been playing a pervasive role in the syntax and in the formation of analytical subparadigms of verbs, whereas in North Slavic its role in the grammar faded away quite early. As a free morpheme (i.e. not as an element in lexicalized units like, for instance, the Russian focus particle da-že ‘even’ or the now obsolete final conjunction da-by; for other scattered facts from North Slavic cf. Gołąb 1954: 70), da is only attested scarcely (and without certainty) in Old Czech and Old Polish (Večerka 1993: 81). In early East Slavic, da was used as a protasis-marker in conditional sentences (cf. Ptentsova 2013, among others) and over a number of centuries, occurred in different syntactic functions, which have been described for OCS. (see below).35 It eventually died out. In Russian, da has persisted as the affirmative marker (‘yes’) and in lexically restricted optative exclamations (Da zdravstvuet X! ‘May X live/thrive!’, and a few more), but even in these phraseologized units, da seems to have been borrowed from Church Slavonic. Admittedly, it can be encountered in folk speech as a copulative conjunction (e.g., Russ. dial. šči da kaša pišča naša ‘cabbage soup and porridge are our food’). This seems to indicate that da nonetheless had some footing in non-literary language, rather uninfluenced by OCS. However, the attested usage does not pertain to complementation or other types of clausal subordination. In sum: Da in (standard and non-standard) Russian has survived basically only as a sentence-initial particle (or conjunction) and, in standard Russian, predominantly with an adversative or kind of hortative function (cf. Mendoza 1996 for a comprehensive analysis; see also ex. (9), (15), (85), (88)). In OCS., da was used as a ‘particle’ (Gołąb 1954: 67–69) in different clause types, but all of them shared one feature: They were all non-factive. They had 35 Cf. Stecenko (1972: 181–184), Keršiene (1983, esp. p. 171), among others.

Main clause infinitival predicates and their equivalents in Slavic

327

either (i) a directive function (as equivalents of the imperative, also in optative or hortative function), or da (ii) occurred after CTPs of volition, command or with apprehensive semantics (verba voluntatis, dicendi, timendi), or (iii) introduced final or, rarely, consecutive clauses. From (ii), it could easily evolve into a complementizer, in (iii) it was easy to become a conjunction. Gołąb (1954: 70f.), following Vondrák (1908/II: 517), assumed that da first existed as a particle before it took part in the formation of tighter syntax (subordination). We turn now to this part of the story. 4.3.2 The spread of da in biclausal and monoclausal sentences The initial stages of a successive replacement of infinitival by da+VF I N -clauses can already be detected in the earliest writings of Slavic, i.e. in OCS. It would be difficult to draw distinctions between the functional ranges of infinitival and da-clauses, and it would be adequate to speak not of a simple replacement of the infinitive by da-clauses, but rather of a successive ousting by da-structures that had already existed beforehand (Asenova 2002: 143). We need not go into details of this long-lasting process here (for a comprehensive account cf. Mirčev 1937, Minčeva 1985). Unfortunately, despite quite impressive philological work on how the infinitive was incrementally ousted by da-clauses, the literature on Bulgarian or Macedonian da-clauses (or da-usage types) does not offer any clear stance regarding the chronology of the rise of different construction types, in particular of the relative order between “loose syntax” (juxtaposed clauses with da in the second conjunct) and the rise of structurally subordinate patterns. Articles dealing with Macedonian, like Fridman [Friedman] (2007 [1987]) and Topolińska (2008a–b), focus on the synchronic situation and the delimitation of da’s basic function(s). Bužarovska (1999) distinguishes between “primary” and “secondary” uses of main da-clauses, but she does not give any proof of a chronological order. With respect to Bulgarian, the same holds true for Genadieva-Mustafčieva (1970); Nicolova’s (2008) functional account, again, is merely synchronic. However, Asenova (2002: 182), based on Mirčev (1937), considers the main-clause da-uses in directive speech acts as diachronically primary. In sum, all we can deduce from the investigations hitherto is that the particle use is diachronically primary and that da took a vital part in all kinds of tightening of syntax. In view of this situation, let us turn to the northwestern neigbors of Macedonian and Bulgarian. Fortunately, the crucial processes which we need to comment on can be found in the comprehensive analyses presented by Grickat (1975) and, primarily, by Grković-Major (2004) on Old Serbian (for a summary

328

Björn Wiemer

cf. Wiemer/Hansen 2012: 80–83). The Croatian-Serbian dialect continuum has the advantage that, even up until the present day, it shows reflexes of practically all historical stages of development in a synchronic nutshell, whereas in Macedonian and Bulgarian, the earlier stages can no longer be “seen”. Grković-Major accepts the viewpoint that da derived from a demonstrative (see Section 4.3.1). In Old Serbian (12th–15th centuries), we encounter it as a paratactic, adjunctive connective, usually with an optative or hortative function, which it already had in OCS. (see Section 4.3.1). This initial stage can be illustrated with the following example from 13th century writings (“Stare srpske povelje i pisma”), quoted from Grković-Major (2004: 198): (106) Da CON

vi ni ste rekli da se stanemo. you.2PL us.DAT be.PRS .2PL say.PST. PL CON RM meet.PRS .1PL

‘Well, you told us. Let’s meet!’ The second part (da se stanemo ‘Let’s meet!’) could still be interpreted as direct speech, loosely attached to the preceding verb form ste rekli ‘you(PL) said’. The hortative potential of da was compatible with one of the possible readings of the speech act verb, namely: That it was used as a directive (and not in order to introduce a report on an accomplished fact). Initially, this semantic compatibility was established on mere discourse-based grounds, without tighter syntax. Subsequently, this loose bond – and the implicature between the possible non-factive reading of the speech act verb and the inherent non-factive semantics of da – was strengthened, the syntactic relation between both parts was reanalyzed. This step is schematized in Figure 2 with the corresponding clause pair in modern Serbian (which in principle still allows for both syntactic interpretations): (106) a.

Ta vi ste nam rekli | da se sastanemo. (juxtaposed clauses)

b. > Ta vi ste nam rekli da se sastanemo. (CTP + complement clause) ‘Well, you told us to meet / that we should meet.’ Figure 2: Reanalysis: juxtaposition > complementation

In parallel, analogous processes took place with da in adverbial subordination. This process probably started with final adjunct clauses before it spread to an increasing number of other clause types, until da eventually became the most ubiquitous connective in South Slavic (for details cf. the literature referred to

Main clause infinitival predicates and their equivalents in Slavic

329

above). After reanalysis took place, further steps followed with the newly born complementizer (but not the conjunction). From the early 16th century onwards, da-clauses after CTPs became more and more restricted to same-subject constructions. This turned into a strict rule in the contemporary stage of CroatianSerbian; compare: da skažui/k (107) xotešei. . . want.IMPF.3PL COMP say:PFV. PRS .3PL ‘theyi wanted that theyi/k say’ > ‘theyi wanted that theyi/*k say’ (= ‘they wanted to say’) After manipulative and desiderative verbs (as in ex. (107)), da also entered into tight correlation with modal auxiliaries and phasal verbs, i.e. it eventually appeared in other complementation patterns which rank highly in terms of semantic integration (in the sense of Givón 1980, Cristofaro 2003: 117–122, and others). Therefore, from the syntactic point of view, the diachronic path of expansion of da-clauses can be briefly characterized as an increase of syntactic tightness; in this process biclausal patterns (as in ex. (106)) were the first ones to appear (out of juxtaposition) before an expansion into monoclausal patterns (that is sentences consisting only of one predicative core) started. In all these complementation patterns, da has become obligatory in the south-eastern part of the South Slavic continuum (i.e. in Macedonian and Bulgarian), while its obligatoriness (against infinitival clauses) decreases to the north-west of this continuum.36 In Macedonian, we observe an even more far-reaching development. The modal mora ‘must’ can only be used in combination with da, not just as a verb inflected for tense and agreement-categories (see ex. (1a)), but also as a morphologically complex epistemic marker which has basically been detached from clausal syntax (dependency structure); cf. Wiemer (2014: 150–162), see also fn. 21. What we seem to observe, therefore, is an almost complete run through the entire scale of complementation: juxtaposition > complementation (biclausal > monoclausal) > (epistemic) particle (mora da). At the end of this cycle we have arrived again at loose syntax. 36 Simultaneously, da gets deprived of its non-factive semantics in the north-western part, as it is used as a neutral complementizer (‘that’) in Serbian, Croatian and Slovene. Paraphrasing this: In the latter languages, the contrast between da and an unmarked complementizer has been blurred since da itself has turned into the neutral complementizer. This, however, is a different part of the fascinating and complex story of complementation in South Slavic and not relevant for the issue of insubordination.

330

Björn Wiemer

We may summarize these observations saying that no insubordination occurred with da. Diachronic secondariness can be sensibly assumed only for the “dramatic” use of da-clauses in narration, which was discussed in Section 3.3 (see ex. (62–63)). Asenova (2002: 193f.) remarks that, in some Bulgarian dialects, the imperative in this use appears to be original; moreover, the imperative as a dramaticizing device of narratives has been attested in various other Slavic languages (folklore and colloquial speech). It may, thus, be argued that da-clauses entered into this domain of the imperative at a later stage than Common Slavic. Note, however, that even this assumption does not invalidate the claim that main-clause use of da is diachronically primary, since such clauses used in narration for dramatic purposes have to be considered as syntactically independent as well. The question should rather be as to whether main-clause use of da in this function has developed in parallel to the narrative infinitive in other languages, among others in Russian (or older stages of Polish). It seems that nobody has raised this issue so far, rendering it entirely open to future research.

5 Conclusions The data that have been consulted (mainly from the reference literature) strongly suggest that patterns of independent infinitives in Slavic are not a development to be judged as diachronically secondary to the rise of subordinative clause combining. Rather, they can be seen as a primary development, at least if looked at from the zero reference point of Common Slavic, i.e. from a period when the infinitive had already been sufficiently integrated into the verbal paradigm (though it still saliently preserved features testifying to its nominal origin and the relation to a goal-directed dative). The alternative assumption would require a model in which patterns of clause combining to be qualified as subordinative had established themselves by the earliest documented times (9th century AD), before clauses headed by infinitives would again have loosened their syntactic bonds with neighboring clauses. This assumption would be quite unrealistic even in terms of synchronic syntax since it appears to be impossible to draw a distinctive delimitation between coordinative and subordinative clause combining. This has been frequently reiterated by various specialists. To develop this further: Such an assumption would testify to an ahistoric attitude toward diachronic syntax, as, it actually results from a projection of our understanding of contemporarily tight patterns of clause combining into remote earlier stages. Instead, we should first ask to what extent discourse pragmatic principles were vital at those stages and nowadays, what might have changed in this regard, and what diagnostic tools can be used to ascertain this. In sum, what seems

Main clause infinitival predicates and their equivalents in Slavic

331

much more realistic (close to the observable facts) is that constructions with IndInfs represent a type of remnant of older clause patterns, and that they might resemble insubordination (i.e. the loss of subordination) only against the background of later (first of all today’s) stages. An analogous conclusion applies to equivalents of main clause phenomena with the connective da. The common denominator of all research done by generations of investigators consists in saying that, before da became a conjunction or complementizer, it was just an element loosely relating adjacent clauses to one another. Only when more distinctive patterns of subordination arose can subordinative functions be told apart from the particle or main clause connective da. Therefore, a scenario that would fit Evans’ original four-stage development into insubordination (see Figure 1) can neither be assumed for the infinitive, nor for equivalent clause structures of the da+Vfin-type. Rather, particle use of da as well as, correspondingly, the syntactically independent status of infinitive clauses must have co-existed with subordinative clause patterns from the time the latter appeared; the subordinative patterns were secondary. As I emphasized at the end of Section 2 while commenting on Evans’ wellconsidered paradox, insubordination needs a background of subordination against which, for a given synchronic period of a language (or speaker community), it can be perceived as a marked case. Apart from that, however, it must be possible (for the linguist) to trace the roots of the main clause phenomena in question. Thus, to Evans’ proviso concerning the notion of insubordination itself, we should add another condition, related to our empirical analysis of diachronic facts: At some moment, we must decide on an historic point of departure that comprises a sufficiently long period of development, and against the background of which ‘ellipsis’ may sensibly be stated (as Figure 1, following Evans, would require). If, we do not have enough reliable data at our disposal for considerably long time spans (e.g., 1,000 years of documented Slavic linguistic history or further back into Proto- or Common Slavic), this empirical question cannot be answered in a satisfying way. In addition, and finally, we must ask the question of how many generations of speakers, even in a continuous chain of transmission of linguistic patterns, can be considered to be conscious of earlier stages; after centuries, speakers cease to be aware of the origin of main clause phenomena that are assumed to derive from insubordination. The historical linguist may reconstruct this process, but the linguist’s knowledge will possibly differ from the metalinguistic awareness of “naïve” language users. Therefore, we should take a clear stance regarding whether insubordination is to be considered as a description of a process that can be “embraced” only by specialists with their tools of analysis, or whether insubordination is a concept anchored in linguistic awareness of ordinary speakers. In this respect, insubordination resembles grammaticalization: For

332

Björn Wiemer

generations of ordinary speakers, the processes evolve rather unconsciously and they are gradient.37 Linguistic reconstruction of the developmental steps certainly does not reflect the metalinguistic awareness of average language users. To conclude, let me stress that my study is exploratory and certainly far from exhaustive as, across Slavic languages, it surveys “highlights” of very complex changes that have to be assessed with more scrutiny against a background of more global changes in the history of particular languages, or only certain varieties, or in entire areas (including other languages than Slavic ones). Admittedly, there are lots of “holes” in the documentation of the diachronic continuum of Slavic languages, even in the descriptive work on particular Slavic languages for, say, the last 150 years;38 and what holds true for standard varieties applies a fortiori to dialects and other non-standard varieties.

Abbreviations CTP IE. IndInf MHG ModInf OCS. p SoA

complement taking predicate Indo-European independent infinitive Middle High German modal infinitive Old Church Slavonic proposition state-of-affairs

37 An anonymous reviewer criticized that “one must assume either that change is abrupt or that it is gradual”. This may be correct when we take into account changes from speaker to speaker (or from one instantiation of a pattern to the next in actual discourse; cf. Croft 2000), or if we speak about reanalysis (in any of its possible readings) or analogy, which are abrupt (Haspelmath 1998: 340f.). But gradience comes in as early as chains of (smaller) abrupt changes accumulated over many generations and looked at more globally ex post are considered. There is, therefore, no contradiction in assuming abruptness on a micro-level, while on a macro-level reconstructing a gradience of change (not to speak of gradience with respect to spread of innovations, after abrupt changes on a micro-level, within larger communities of speakers). 38 Actually, this general current stage concerning Slavic languages is not unusual. As Evans (2007: 429) pointed out, facts that might potentially be related to insubordination are often difficult to detect (even in the best-studied languages like English) since they “tend to get marginalized in linguistic analysis and description. As a result, it is hard to get a systematic picture across the world’s languages”. Obviously, this holds true not only for typological work on a global scale, but also even for small-scale comparisons within one (European) language group.

Main clause infinitival predicates and their equivalents in Slavic

333

Glosses, in addition to standard glosses (contained in the Leipzig Glossing Rules): ADV ART COMP CON DIM EMPH IRR

PRS PA PST PA Q REFLPOSS RM

adversative conjunction definite article complementizer connective (in complex predicates) diminutive suffix emphatic (non-clitic) form of pronoun particle with non-factive (irreal, optative, “subjunctive”) function present active participle past active participle interrogative particle reflexive possessive pronoun reflexive marker (clitic or agglutinated)

References Sources (mentioned in some of the examples) Dog. Gr. Novg. 1264 g. = Dogovornaja gramota Novgoroda s tverskim velikim knjazem Jaroslavom Jaroslavovičem from 1264 [Treaty between Novgorod and the Grand Prince of Novgorod Jaroslav Jaroslavovič]. In: Gramoty Velikogo Novgoroda i Pskova [Treaties of Novgorod the Great and Pskov]. Moskva Leningrad: Izdatel’stvo AN SSSR, 1949. Dog. Gr. Novg. 1316 g. = Dogovornaja gramota Novgoroda s tverskim velikim knjazem Mixailom Jaroslavovičem from 1316 [Treaty of peace between Novgorod and the Grand Prince of Novgorod Mixail Jaroslavovič]. In: Gramoty Velikogo Novgoroda i Pskova [Treaties of Novgorod the Great and Pskov]. Moskva, Leningrad: Izdatel’stvo AN SSSR, 1949. MarEv = Mariinskoe četveroevangelie s primečanijami i priloženijami. Ed. by V. Jagić. Berlin, St. Petersburg 1883. (Reprint Graz 1960.) Povest’ vremennyx let = Povest’ vremennyx let, č.1 [Primary Chronicle, part 1]. Мoskva, Leningrad: Izdatel’stvo AN SSSR, 1950. Orš. gr. 1615 g. = Oršanskaja gramota 1615 g. [Orša treaty from 1615]. In: Akty, otnosjaščiesja k istorii Zapadnoj Rossii, sobrannye i izdannye Arxeografičeskoju komissijeju, t. I–V [Documents relating to the history of Western Russia, collected and published by the Archeological Commission; vol. I–V]. Sankt Petersburg, 1846–1853. Su = Suprašalski ili Retkov sbornik 1, 2. Ed. by J. Zaimov and M. Capaldo. Sofija, 1982–1983.

334

Björn Wiemer

Literature of the subject referred to in the article Aikhenvald, Alexandra Y. 2004. Evidentiality. Oxford: Oxford University Press. Ambrazas, Vytautas. 1995. Lietuvių kalbos bendraties konstrukcijų raida. Lietuvių kalbotyros klausimai 33. 74–109. Asenova, Petja. 2002. Balkansko ezikoznanie (Osnovni problemi na balkanskija ezikov săjuz). Veliko Tărnovo: Faber. Avrutin, Sergey. 1999. Development of the Syntax-Discourse Interface. Dordrecht: Kluwer. Besters-Dilger, Juliane. 1999. Kirchenslavisches und Nicht-Kirchenslavisches in der Entwicklung der russischen Ausdrucksmittel für die Möglichkeits- und Notwendigkeitsmodalität. In Ernst Hansack, Walter Koschmal, Norbert Nübler & Radoslav Večerka (eds.): Festschrift für Klaus Trost zum 65. Gebutstag, 29–36. München: Sagner. Besters-Dilger, Juliane, Ana Drobnjaković & Björn Hansen. 2009. Modals in the Slavonic languages. In Björn Hansen & Ferdinand de Haan (eds.), Modals in the Languages of Europe (A Reference Work), 167–197. Berlin: de Gruyter. Bogusławski, Andrzej & Magdalena Danielewiczowa. 2005. Verba polona abscondita (Sonda słownikowa III). Warszawa: Elma Books. Borkovskij, Viktor I. 1968. Sravnitel’no-istoričeskij sintaksis vostočnoslavjanskix jazykov. Tipy prostogo predloženija. Moskva: Nauka. Borkovskij, Viktor I. (ed.). 1978. Istoričeskaja grammatika russkogo jazyka: Sintaksis. Prostoe predloženie. Moskva: Nauka. Borkovskij, Viktor I. & Petr S. Kuznecov. 21965. Istoričeskaja grammatika russkogo jazyka. Moskva. Nauka. Boye, Kasper. 2012. Epistemic Meaning. A Crosslinguistic and Functional-Cognitive Study. Bulaxovskij, L.A. 51958. Istoričeskij kommentarij k russkomu literaturnomu jazyku. Kiev: « Radjans’ka škola ». Bužarovska, Eleni. 1999. Nezavisnite da-konstrukcii vo makedonskiot jazik i nivnite korelati vo grčkiot jazik. Makedonski jazik LI-LII (Skopje: MANU), 217–236. Cristofaro, Sonia. 2003. Subordination. Oxford: Oxford University Press. Croft, William. 2000. Explaining Language Change: An Evolutionary Approach. Harlow: Longman. Dahl, Östen. 1985. Aspect and Tense Systems. Oxford: Blackwell. DeLancey, Scott. 1997. Mirativity: The grammatical marking of unexpected information. Linguistic Typology 1. 33–52. Disterheft, Dorothy. 1980. The Syntactic Development of the Infinitive in Indo-European. Columbus: Slavica Publishers. Disterheft, Dorothy. 1997. The evolution of Indo-European infinitives. In Dorothy Disterheft, Martin Hud & John Greppin (eds.), Ancient Languages and Philosophy. Studies in Honor of Jaan Puhvel, vol. 1., 101–122. Washington: Institute for the Study of Man. Evans, Nicholas. 2007. Insubordination and its uses. In Irina Nikolaeva (ed.), Finiteness: Theoretical and Empirical Foundations, 366–431. Oxford: Oxford University Press. Fortuin, Egbert. 2005. From necessity to possibility: The modal spectrum of the dative-infinitive construction in Russian. In Björn Hansen & Petr Karlík (eds.), Modality in Slavonic Languages (New Perspectives), 39–60. München: Sagner. Fridman, Viktor [= Friedman, Victor]. 2007. Tipologijata na upotrebata na da vo balkanskite jazici. In Viktor Fridman (ed.), Makedonistički studii, 43–52. Skopje: MANU. [Reprint from: Prilozi: Oddelenie za lingvistika i literaturna nauka, 1987.]

Main clause infinitival predicates and their equivalents in Slavic

335

Gawełko, Marek. 2006. O tendencjach rozwojowych polskiego bezokolicznika. Polonica XXVI– XXVII. 255–273. Genadieva-Mustafčieva, Zara. 1970. Podčinitelnijat săjuz da v săvremennija bălgarski ezik. Sofija: Izdatelstvo na Bălgarskata akademija na naukite. Georgievski, Georgi. 2009. Da-rečenicata vo makedonskiot jazik. Skopje: Institut za makedonski jazik „Krste Misirkov“. Gębka-Wolak, Małgorzata. 2011. Pozycje składniowe frazy bezokolicznikowej we współczesnym zdaniu polskim. Toruń: Wydawnictwo Naukowe UMK. Givón, Talmy. 1980. The binding hierarchy and the typology of complements. Studies in Language 4. 333–377. Glaser, Elvira. 2002. Fein geschmackte Pinienkerne zugeben! Zum Infinitiv in Kochrezepten. In David Restle & Dietmar Zaefferer (eds.), Sounds and Systems. Studies in Structure and Change (A Festschrift for Theo Vennemann), 165–183. Berlin: de Gruyter. Gołąb, Zbigniew. 1954. Funkcja syntaktyczna partykuły da w językach południowo-słowiańskich (bulgarskim, macedońskim i serbo-chorwackim). Biuletyn polskiego towarzystwa językoznawczego XIII. 67–92. Gołąb, Zbigniew. 1984. South Slavic da + indicative in conditional clauses and its general linguistic implications. In Kot K. Shangriladze & Erica W. Townsend (eds.), Papers for the V. Congress of Southeast European Studies (Belgrade, Sept. 1984), 170–198. Columbus, Ohio: Slavica Publishers. Grepl, Miroslav & Petr Karlík. 1998. Skladba češtiny. Olomouc: Votobia. Grickat, Irena. 1975. Studije iz istorije srpskohrvatskog jezika. Beograd: Narodna biblioteka SR Srbije. Grković-Mejdžor [= Grković-Major], Jasmina. 2004. Razvoj xipotaktičkog da u starosrpskom jeziku. Zbornik Matice Srpske za filologiju i lingvistiku 47(1–2). 185–203. Hansen, Björn. 2001. Das slavische Modalauxiliar (Semantik und Grammatikalisierung im Russischen, Polnischen, Serbischen/Kroatischen und Altkirchenslavischen). München: Sagner. Harris, Alice C. & Lyle Campbell. 1995. Historical Syntax in Cross-Linguistic Perspective. Cambridge: Cambridge University Press. Haspelmath, Martin. 1989. From purposive to infinitive – a universal path of grammaticalization. Folia Linguistica Historica X(1–2). 287–310. Haspelmath, Martin. 1998. Does grammaticalization need reanalysis? Studies in Language 22 (2). 315–351. Holvoet, Axel. 2003. Modal constructions with ‘be’ and the infinitive in Slavonic and Baltic. Zeitschrift für Slawistik 48(4): 465–480. Joseph, Brian. 1983. The Sychrony and Diachrony of the Balkan Infinitive. A Study in Areal, General, and Historical Linguistics. Cambridge: Cambridge University Press. Keršiene, Rufina B. 1983. Složnosočinennye predloženija. In Viktor I. Borkovskij (ed.), Struktura predloženija v istorii vostočnoslavjanskix jazykov, 141–184. Moskva: Nauka. Kiparsky, Paul & Carol Kiparsky. 1970. Fact. In Danny D. Steinberg & Leon A. Jakobovits (eds.), Semantics. An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, 345– 369. London: Cambridge University Press. Kramer, Christina Elizabeth. 1986. Analytic Modality in Macedonian. München: Sagner. Langacker, Ronald W. 1977. Syntactic reanalysis. In Charles N. Li (ed.), Mechanisms of Syntactic Change, 59–139. Austin: University of Texas Press. Longacre, Robert E. 1979. The paragraph as a grammatical unit. In Talmy Givón (ed.), Discourse and Syntax (Syntax and Semantics 12), 115–134. New York: Academic Press.

336

Björn Wiemer

Lyons, John. 1977. Semantics, vol. 2. Cambridge: Cambridge University Press. Łaziński, Marek. 2011. Kto imeet uši slyšat‘, da slyšit! Funkcje semantyczne bezokolicznika jako nadrzędnego predykatu w zdaniu polskim i rosyjskim. Mirosław Bańko & Dorota Kopcińska (eds.), Różne formy, różne treści (tom ofiarowany profesorowi Markowi Świdzińskiemu), 139–150. Warszawa: Wydawnictwo UW. Madariaga, Nerea. 2015. The decline of non-finiteness as a syntactic mechanism for embedding in East Slavic. Journal of Historical Linguistics 5. 139–174. Malchukov, Andrej. 2013. Verbalization and insubordination in Siberian languages. In Martine Robbeets & Hubert Cuyckens (eds.), Shared Grammaticalization (With Special Focus on the Transeurasian Languages), 177–208. Amsterdam: John Benjamins. Maurice, Florence. 1996. Der modale Infinitiv in der modernen russischen Standardsprache. München. Sagner. Mendoza, Imke. 1996. Zur Koordination im Russischen ( i, a und da als pragmatische Konnektoren). München: Sagner. Mendoza, Imke. 2008. Überlegungen zur Entstehung des Nominativobjekts im Altrussischen. In Peter Kosta & Daniel Weiss (eds.), Slavistische Linguistik 2006/2007 (Referate des XXXII. und des XXXIII. Konstanzer Slavistischen Arbeitstreffens), 299–317. München: Sagner. Miklosich, Franz. 1926. Vergleichende Grammatik der slavischen Sprachen, IV: Syntax. Heidelberg: Winter. [Reprint of first edition 1868–1874.] Minčeva, Angelinae. 1985. Za xaraktera na konkurencijata meždu infinitiv i da-izrečenija v starobălgarskite pametnici. Litterae Slavicae Mediaevi. München, 211–221. Mirčev, Kiril. 1937. Kăm istorijata na infinitivnata forma v bălgarskija ezik. Godišnik na Sofijskija universitet „Sv. Kliment Oxridski“ (GSU), Istoriko-filologičeski fakultet XXIII-12. 3–34. Mišeska Tomić, Olga. 2006. Balkan Sprachbund Morpho-Syntactic Features. Dordrecht: Springer. Mišeska Tomić, Olga. 2012. A Grammar of Macedonian. Bloomington, Indiana: Slavica. Němec, Igor. 1977. K vývoji funkcí infinitivu v češtině a v litevštině. Slovo a slovesnost 38–4. 275–280. Nicolova, Ruselina. 2008. Bălgarska gramatika. Morfologija. Sofija: Universitetsko izdatelstvo “Sv. Kliment Oxridski”. Nikolaeva, Irina. 2007. Constructional economy and nonfinite independent clauses. In Irina Nikolaeva (ed.), Finiteness: Theoretical and Empirical Foundations, 138–180. Oxford: Oxford University Press. Nikolaeva, Irina. 2013. Unpacking finiteness. In Dunstan Brown, Marina Chumakina & Greville G. Corbett (eds.), Canonical Morphology and Syntax, 99–122. Oxford: Oxford University Press. Noonan, Michael. 22007. Complementation. In Timothy Shopen (ed.), Language Typology and Syntactic Description, vol. II: Complex Constructions, 52–150. Cambridge: Cambridge University Press. Nordström, Jackie. 2010. Modality and Subordinators. Amsterdam: John Benjamins. Palmer, Frank R. 22001. Mood and Modality. Cambridge: Cambridge University Press. Pisarkowa, Krystyna. 1984. Historia składni języka polskiego. Wrocław: Ossolineum. Plungjan, Vladimir A. 2001. Antirezul’tativ: do i posle rezul’tata. In Vladimir A. Plungjan (ed.), Glagol’nye kategorii. (= Issledovanija po teorii grammatiki 1), 50–88. Moskva: « Russkie slovari ». Porák, Jaroslav. 1967. Vývoj infinitivních vět v češtině. Praha: Univerzita Karlova.

Main clause infinitival predicates and their equivalents in Slavic

337

Potebnja, Aleksandr A. 1958. Iz zapisok po russkoj grammatike, t. I–II (ed. by V.I. Borkovskij et al.). Moskva: Izd-vo Prosveščenie. Ptentsova, Anna. 2013. On a typological parallel in the expression of “imperativeness”, “optativeness”, and “condition” in Modern Russian and Old Russian. In Valentina Apresjan & Boris Iomdin (eds.), Meaning-Text Theory: Current Developments, 115–124. München: Sagner. Sabenina, A.M. 1983. Iz istorii dvukomponentnyx neglagol’nyx predloženij. In Viktor I. Borkovskij (ed.), Struktura predloženija v istorii vostočnoslavjanskix jazykov, 9–101. Moskva: Nauka. Sells, Pete. 2007. Finiteness in non‐transformational syntactic frameworks. In Irina Nikolaeva (ed.), Finiteness: Theoretical and Empirical Foundations, 59–89. Oxford: Oxford University Press. Stecenko, Anatolij N. 1972. Istoričeskij sintaksis russkogo jazyka. Moskva: Vysšaja škola. Svoboda, Karel. 1988. Kapitoly z vývoje české syntaxe hlavně souvětné. Praha: Univerzita Karlova. Švedova, Nina Ju. 1964. Izmenenija v sisteme prostogo predloženija. In Viktor V. Vinogradov & Nina Ju Švedova (eds.), Izmenenija v sisteme prostogo i osložnennogo predloženija v russkom literaturnom jayzke XIX veka, 20–368. Moskva: Nauka. Topolinjska [= Topolińska], Zuzanna. 2000. Polski – makedonski. Gramatička konfrontacija 3: Studii od morfosintaksata. Skopje: MANU. Topolińska, Zuzanna. 2008a. Było pomyśleć wcześniej. In Zuzanna Topolińska (ed.), Z Polski do Macedomii: Studia językoznawcze, problemy predykacji 1, 166–172. Kraków: Lexis. [Reprint from: Maciej Grochowski & Daniel Weiss (eds.), Words are Physicians for an Ailing Mind, 433–437. München: Sagner.] Topolińska, Zuzanna. 2008b. Factivity as a grammatical category in Balkan Slavic and Balkan Romance. In Zuzanna Topolińska (ed.), Z Polski do Macedomii: Studia językoznawcze, problemy predykacji 1, 173–184.Kraków: Lexis. [Reprint from: Slavia Meridionalis 1 (1994). 105– 121.] van der Auwera, Johan & Vladimir A. Plungian. 1998. Modality’s semantic map. Linguistic Typology 1–2. 79–124. Večerka, Radoslav. 1993. Altkirchenslavische (altbulgarische) Syntax, II: Die innere Satzstruktur. Freiburg/Br.: Weiher. Večerka, Radoslav. 1996. Altkirchenslavische (altbulgarische) Syntax, III: Die Satztypen: der einfache Satz. Freiburg/Br.: Weiher. Veyrenc, Jacques. 1979. Les propositions infinitives en russe. Paris: Institut d‘Études Slaves. Vondrák, Václav. 1908. Vergleichende slavische Grammatik, vol. II: Formenlehre und Syntax. Göttingen. Wiemer, Björn. 2001. Aspect choice in non-declarative and modalized utterances as extensions from assertive domains (Lexical semantics, scopes, and categorial distinctions in Russian and Polish). In Hauke Bartels, Nicole Störmer & Ewa Walusiak (eds.), Untersuchungen zur Morphologie und Syntax im Slavischen, 195–221. Oldenburg: BIS-Verlag. Wiemer, Björn. 2014. Mora da as a marker of modal meanings in Macedonian: On correlations between categorial restrictions and morphosyntactic behaviour. In Werner Abraham & Elisabeth Leiss (eds.), Modes of Modality. Modality, Typology and Universal Grammar, 127–166. Amsterdam: John Benjamins. Wiemer, Björn. 2015. O roli vida v oblasti kratnosti i pragmatičeskix funkcij (ėskiz s točki zrenija xronotopii). In Rosanna Benacchio [Bennak’o] (ed.), Glagol’nyj vid: grammatičeskoe značenie i kontekst / Verbal Aspect: Grammatical Meaning and Context, 585–609. München: Sagner.

338

Björn Wiemer

Wiemer, Björn & Björn Hansen. 2012. Assessing the range of contact-induced grammaticalization in Slavonic. In Björn Wiemer, Bernhard Wälchli & Björn Hansen (eds.), Grammatical Replication and Borrowability in Language Contact, 67–155. Berlin: de Gruyter. Wolfová, Jarmila. 1971. Infinitivní věty exhortativní v češtině. Sborník prací Filosofické fakulty Brněnské university. Studia menora facultatis philosophicae universitatis brunensis, A 19. 199–203.

Language index Afrikaans 1 Ancient Greek 18, 81, 181, 183, 325 Australian languages 300 Avestan 246, 256, 257 Balkan languages 1, 7, 19, 147, 155, 157, 166 Balkan Slavic 266–270, 277, 281, 300–306, 325 Bantu 195, 217, 279 Basque 300 Berber 195 Bosnian 296 Brazilian Portuguese 199 Bulgarian 6, 147, 155, 265–267, 281, 300– 305, 325–330 Byzantine Greek 104 Celtic languages 195 Classical Greek 18, 82, 91 Classical Latin 158 Common Romanian 158 Common Slavic 306, 316, 324–326, 330, 331 Czech 265, 266, 288–296, 319–326 Danish 11 Danubian dialects 158 Dutch 1, 8–16, 32–35, 45, 46, 82, 88 Early Modern Romanian 19, 35, 147–166 Early New High German 2–5, 14–16, 175, 183 East Slavic 265, 268, 277, 195, 304–326 English 1, 6–19, 33, 36, 48, 49, 82, 86, 116– 125, 132–135, 169, 173, 178–188, 224– 229, 250, 289, 296, 322, 332 European Portuguese 11 French 1, 11, 60, 61, 64, 150, 278 Frisian 1 Galician 199, 259 German 1–22, 31–35, 44–48, 169–189, 239, 285, 289, 296, 319 Germanic languages 7, 33, 182

Greek 6, 7, 18, 19, 36, 55, 81–106, 147, 155, 169, 181, 183, 233, 256, 258, 325 Hellenistic-Roman Greek 86, 87, 98, 103 Hercegovinian 296 Hittite 22, 223–233, 250, 254, 260 Homeric Greek 90 Hungarian 21, 193–217, 260 Imbabura Quechua 195, 215 Indo-Iranian 256, 259 Italian 6, 11, 16, 17, 31–61, 195 Italian dialects 195 Japanese 22, 32, 34, 35, 43, 45, 46 Ket 1 Kinande 214 Korean 36 Late Modern English 19, 116, 123, 142 Latin 98, 147, 151–159, 165–169, 181–183, 196, 200, 202, 233, 258, 259, 278, 319, 320 Macedonian 22, 265–270, 300–305, 325– 329 Mandarin Chinese 36 Middle English 12, 19, 116, 120, 123, 178, 182 Middle High German 14, 21, 175, 176, 183, 189 Middle Hungarian 210 Middle Polish 12 Modern British English 123 Modern Hungarian 193–196, 210, 211 Montenegrinian 296 Mycenaean Greek 90 North Slavic 306, 326 Norwegian 11, 14 Old Church Slavonic 266, 326 Old Czech 296, 320, 322

340

Language index

Old French 1 Old High German 7, 14, 16, 174–183 Old Hungarian 21, 193–217, 260 Old Indic 223, 224, 229, 251–260 Old Iranian 233 Old Leonese 199 Old Mirandese 199 Old Neapolitan 199 Old Persian 257 Old Polish 320, 326 Old Russian 277 Old Serbian 327, 328 Oscan 256

Salentino 6 Sanskrit 255 Sardinian 199 Serbian-Croatian 266, 294, 295, 316 Slavic languages 6, 15, 179, 233, 266–268, 277, 278, 288, 289, 299, 305, 311, 330, 332 Slovene 280, 329 South Slavic 270, 288, 295–306, 325–329 Spanish 11, 17, 31–35, 45–50, 60, 199 Swedish 21 Swiss German 13 Szigetköz dialect 213, 214

Pennsylvania German 13 Polish 9–22, 35, 265, 266, 280, 285–300, 316–330 Portuguese 11, 199, 259 Post-Classical Greek 18 Proto-Germanic 182 Proto-Indo-European 179, 181–183, 223

Torlakian dialects 267 Tzutujil 14

Romance languages 1, 6, 17–22, 33, 57–61, 147, 151, 155, 259, 261, 278, 289, 296 Romanian 17–20, 31, 35, 55–77, 147–166, 266, 301 Russian 1, 6, 22, 149, 224–226, 266–330

Ukrainian 266, 295, 297 Umbrian 256 Vedic 22, 223–230, 233, 243–250, 256, 260 Welsh 199, 259 West Slavic 306, 319, 320, 322 Wolof 31

Subject index We pursue two goals with the following collection of technical terms used in this volume. Firstly, it can be consulted as a traditional index to identify a concrete technical term. Secondly, it is meant to give an overview of infinitive types and approach components being used in individual papers. 1 Types of infinitives 1.1 Types of infinitives in individual articles Jędrzejowski / Demske

– non-stative infinitive 93 – stative infinitive 91, 93 present perfect infinitive 88 synthetic future infinitive 86, 103, 106

anti-agreeing infinitive 21 aorist infinitive 18 bare infinitive 2, 14, 15, 20, 21 complement infinitive 10 control infinitive 20 dependent infinitive 4, 6, 12 ECM infinitive 21 extraposed infinitive 4 inflected infinitive 21 long infinitive 20 perfective infinitive 22 root infinitive 22 short infinitive 20 to-infinitive 19 uninflected infinitive 21 wh‑infinitive 7, 9–13 zu‑infinitive 14

D’hoedt & Cuyckens

Nedelcu & Paraschiv

accusativus cum infinitivo (AcI) 169, 170– 174, 178, 182–185, 188, 189 bare infinitive 169, 171, 173, 175, 176, 178, 181, 182, 183, 185, 188 control infinitive 169, 175, 176, 184 nominalized infinitive 175 object control infinitive (OCIC) 169, 171–180, 183–185, 188, 189, 190 obligatorily coherent infinitive 172, 175, 188 obligatorily incoherent infinitive 172 subject control infinitive (SCIC) 169, 170, 171 subject raising infinitive 185 zu-infinitive 169, 172, 174, 175, 181–184

a-infinitive 62, 67, 72, 74 short bare infinitive 74 purpose adjunct infinitive 72 Kavčič active transitive perfect (ATP) 89–91, 96–97, 100, 101, 105, 107 passive aorist infinitive 93 ECM infinitive 82 future infinitive 81, 85–87, 89, 90, 95, 98, 103, 105, 106 passive perfect infinitive 89 present infinitive 81, 84, 85, 87, 90–95, 100, 102, 104–106

bare infinitive 118 for. . .to-infinitive 118 to-infinitive 115–119, 121–123, 125, 130, 141, 142 Hill a-infinitive 147, 150, 152, 154, 157, 158, 160– 166 long infinitive 148, 158 nominal infinitive 151, 153 nominalized infinitive 149, 150, 153 original infinitive 147, 153, 154 short infinitive 148, 158 verbal infinitive 149, 153, 166 Speyer

342

Subject index

Dékány Ablative control infinitive 200 agreeing infinitive 195, 196, 199–203, 207, 208, 210, 216 anti-agreeing infinitive 193, 195, 196, 198, 203–205, 207, 210, 213, 214, 216 bare infinitive 195 control infinitive 200, 201, 203 ECM infinitive 199, 205, 206, 216, 217 inflected infinitive 193–195, 198–200, 202, 203, 207, 208, 210, 214, 217 object infinitive 200, 203 raising infinitive 205, 217 subject infinitive 201, 203, 206 subject-to-subject raising infinitive 217 uninflected infinitive 195, 198–200, 202, 203, 207, 208, 214, 217 Lühr adjunct infinitive 252 attributive infinitive 252, 253, 260 conjugated infinitive 259 intransitive infinitive 231 predicative infinitive 251 purpose infinitive 223, 224, 233, 247, 261 transitive infinitive 231, 232 uninflected infinitive 259 Wiemer adversative infinitive 280, 303 assertive infinitive 280 exhortative infinitive 292 imperfective infinitive 284 independent infinitive (IndInf) 272, 277– 282, 284, 285, 288–294, 296, 298, 305–307, 309, 311–313, 315, 317–325, 330, 331 main clause infinitive 278, 296 modal independent infinitive 281 negated perfective infinitive 292 narrative infinitive 278–281, 304, 330 subject infinitive 277 1.2 Types of infinitives taken all together a-infinitive 62, 67, 72, 74, 147, 150, 152, 154, 157, 158, 160–166

Ablative control infinitive 200 accusativus cum infinitivo (AcI) 169, 170– 174, 178, 182–185, 188, 189 adjunct infinitive 72, 252 adversative infinitive 280, 303 agreeing infinitive 195, 196, 199–203, 207, 208, 210, 216 anti-agreeing infinitive 21, 193, 195, 196, 198, 203–205, 207, 210, 213, 214, 216 aorist infinitive 18 assertive infinitive 280 attributive infinitive 252, 253, 260 bare infinitive 2, 14, 15, 20, 21, 74, 118, 169, 171, 173, 175, 176, 178, 181, 182, 183, 185, 188, 195 complement infinitive 10 conjugated infinitive 259 control infinitive 20, 169, 175, 176, 184, 200, 201, 203 dependent infinitive 4, 6, 12 ECM infinitive 21, 82, 199, 205, 206, 216, 217 exhortative infinitive 292 extraposed infinitive 4 for. . .to-infinitive 118 future infinitive 81, 85–87, 89, 90, 95, 98, 103, 105, 106 imperfective infinitive 284 independent infinitive (IndInf) 272, 277– 282, 284, 285, 288–294, 296, 298, 305–307, 309, 311–313, 315, 317–325, 330, 331 inflected infinitive 21, 193–195, 198–200, 202, 203, 207, 208, 210, 214, 217 intransitive infinitive 231 long infinitive 20, 148, 158 main clause infinitive 278, 296 modal independent infinitive 281 narrative infinitive 278–281, 304, 330 negated perfective infinitive 292 nominal infinitive 151, 153 nominalized infinitive 149, 150, 153, 175 object control infinitive (OCIC) 169, 171–180, 183–185, 188, 189, 190 obligatorily coherent infinitive 172, 175, 188 obligatorily incoherent infinitive 172 original infinitive 147, 153, 154

Subject index

passive aorist infinitive 93 passive perfect infinitive 89 perfective infinitive 22 predicative infinitive 251 present infinitive 81, 84, 85, 87, 90–95, 100, 102, 104–106 – non-stative infinitive 93 – stative infinitive 91, 93 present perfect infinitive 88 purpose infinitive 223, 224, 233, 247, 261 raising infinitive 205, 217 root infinitive 22 short infinitive 20, 148, 158 subject control infinitive (SCIC) 169, 170, 171 subject infinitive 201, 203, 206, 277 subject raising infinitive 185 subject-to-subject raising infinitive 217 synthetic future infinitive 86, 103, 106 to-infinitive 19, 115–119, 121–123, 125, 130, 141, 142 transitive infinitive 231, 232 uninflected infinitive 21, 195, 198–200, 202, 203, 207, 208, 214, 217, 259 verbal infinitive 149, 153, 166 wh‑infinitive 7, 9–13 zu‑infinitive 14, 169, 172, 174, 175, 181–184 2 Explanation approaches 2.1 Explanation approaches in individual approaches Jędrzejowski / Demske AcI construction 9, 18, 20 clause union effect 17 clitic climbing 17 control 1, 14, 17, 20–21 – object control 2–3, 9, 21 – subject control 9, 21–22 extraposition 16 head 17 – aspectual head 17 – functional head 17 long inversion 16 long passive 17 negation operator 5 obviation effect 7

phrase 3–5, 12 – noun phrase 3–4 – verbal phrase 3 – wh‑phrase 12 pied piping 16 position 3–4 – final position 3 – prefield position 4 – second position 3 pronoun fronting 16 temporal mismatch 5 verb second rule 3 Wh-Infinitive-Generalization 10 Grano argument 31, 36, 39–40, 48–49 – experiencer argument 48 – event argument 49 – individual argument 40, 49 – proposition-denoting argument 39 – propositional argument 40 category 34, 45–46, 48 – aspectual category 48 – factive category 34 – functional category 45–46 – modal category 34 – motion category 34 – propositional category 34 Cinque-style hierarchy 33 clausal height 49 clause 31–32 – complement clause 31–32 – matrix clause 31 clause-boundedness 32 clause union 31 clitic 31, 34 clitic climbing 31, 34–35, 37, 41, 48 clustering 33 coherence 31, 33 complement 36, 42 – CP complement 42 – finite complement 36 construction 31–32 – infinitival construction 31–32 – long passive construction 32 control 37–38 – exhaustive control 37–38

343

344

Subject index

– partial control 37–38 dependent variable 42, 49 embedded tense 34 event 49 – speech event 49 – VP event 49 evidential mood 41 head 33, 35–38, 41–43, 46–47 – aspectual head 36, 42–43, 47 – functional head 33, 35–36, 38, 41–43, 46 – inflectional head 33, 47 infinitivus pro participio effect 34 inflectional layer 33, 35–36, 41–42, 45 long passive 32, 34 modal flavor 49 monoclausality 31, 33, 37 (long) passivization 32, 37 permutation invariance 46 position 31–32, 36, 38–39 – clause-local position 31 – matrix subject position 32, 39 – specifier position 36 – subject position 36, 38–39 predicate 31, 33, 36–42, 44–49 – control predicate 38–39, 48 – embedded predicate 37 – factive predicate 41, 46 – main predicate 33 – matrix predicate 31 – non-restructuring predicate 37 – one-place predicate 39 – raising predicate 38–39, 42, 48–49 – restructuring predicate 36–40, 44–45, 47 – subject-oriented predicate 42, 45 PRO 38, 42 projection 33, 36–37 – clausal projection 37 – extended projection 33, 36 – functional projection 33, 37 reduced non-restructuring 33 restructuring 33 – functional restructuring 33 – lexical restructuring 33 restructuring generalization 42 (cross-clausal) scrambling 35 structure 31, 34, 37, 42 – bi-clausal structure 31, 34, 42

– mono-clausal structure 37 subject 37, 39, 42 – collective subject 37 – expletive subject 39 – idiom chunk subject 39 – inanimate subject 39 – PRO subject 42 subject orientation 38–39, 42, 44, 50 thematic dependency 38–39, 42 transparency effect 31, 37–38 variable binding 38–40, 42–43, 45 verb 32–38, 41–43, 45, 47–49 – aspectual verb 32, 43, 47 – epistemic verb 41 – factive verb 34, 41 – implicative verb 47–48 – lexical verb 33, 42, 49 – modal verb 32 – motion verb 34, 47 – particle verb 45 – restructuring verb 33–38 – verb of speech 38, 41 verb raising 34, 49 Nedelcu & Paraschiv accusative clitic 72 adjacency 72–73 adjunct 62, 70–72, 77 – infinitival adjunct 77 – purpose adjunct 70–72, 77 – subjunctive adjunct 62 (propositional) argument 61, 64 aspectual adverbial 74 (subordinate) clause 72 clitic climbing 76–77 complement 64, 72 – finite complement 64 – non-finite complement 64, 72 connector 71 construction 55, 60, 62–63, 67, 70–74, 76– 77 – impersonal construction 70 – pseudo-relative construction 60 – reflexive-impersonal construction 61, 77 CP 64, 72, 74 DP 64 ECM 72

Subject index

event 61–62, 76–77 – factual event 61, 76 – intentional event 62, 77 – possible event 61, 76 – simultaneous event 61 functional domain 74 indirect interrogative 68 infinitival complement 57, 63, 74 infinitive subject 70 mono-clausal 74, 77 perception 57, 60–61, 64 – direct perception 57, 60–61, 64 – indirect perception 56, 61 pronominal clitic 74 sentential negation 74 (postverbal) subject 61–62, 67, 70, 72, 75, 77 verb 55–57, 60–63, 66–70, 72, 75–77 – cognition verb 55–56, 63, 66, 68–70, 72, 75, 77 – copula verb 67 – existential verb 67 – intentional verb 76 – motion verb 62 – non-obligatory control verb 55–56 – obligatory control verb 55 – (intentional) perception verb 55–57, 60– 62, 70, 72, 75–77 – verb of physical perception 56–57, 61, 63 Kavčič clause 81–106 – AcI clause 81–106 – declarative infinitive clause 83 – finite complement clause 82–84, 95, 105 – fully tensed complement clause 82 – hybrid complement clause 83, 92 – indicative complement clause 84 – infinitive complement clause 84 – nonfinite complement clause 83, 88 indicative 94–96 – aorist indicative 94–95 – present indicative 95 individual style 94, 97 perfect 87–90, 96, 101–102 – active perfect 87–88, 90, 101–102 – active perfect of actions and processes 89

345

– active transitive perfect 89, 96 – anterior perfect 87–88, 96 – resultative perfect 87–88 stativity 84, 88–89, 91–92, 100, 102–103, 105–106 tense morphology 82 verb 83–86, 88–89, 105–106 – intransitive verb 89 – stative verb 84–86, 88 – terminative verb 88 – verb of ordering 83 – verb of saying 83–84, 105 – verb of thinking 83–84, 105–106 D’hoedt & Cuyckens adjunct 127–128, 141 alternation 118–120, 123 alternation pair 116 alternation pattern 115, 117 anteriority 127, 137, 139–141 AP 119 binary logistic regression 117, 129–131, 135, 141 categorical 116, 119 CC 119, 122, 130, 132, 134, 139 – that-CC 119, 122, 130 – zero-CC 122, 130, 132, 134, 139 CC-specificity 122 CC-type 115–116, 118–121, 123–124, 136, 140–141 clause 115–119, 121–122, 124, 126–128, 137, 139–140 – complement clause 116, 118, 122, 126– 128, 137, 139–140 – gerundial -ing-clause 115–116 – matrix clause 122, 127 – non-finite complement clause 122 – passive ECM-clause 124 – reduced clause 119 – Small Clause 117–119, 141 – verbless clause 117 – zero-complementizer clause 117, 121 cognitive complexity principle 118 Cognitive Grammar 116 competition 119 complement 119, 121, 123, 125, 127, 130, 135–137, 139–142

346

Subject index

– SC-complement 119, 121, 123, 125, 127, 130, 135–136, 140–141 – that-complement 116, 136–137, 139–140, 142 – zero-complement 121, 134, 136, 138–142 complement choice 140 complement type 115–116, 118, 120, 130, 140–141 complement-taking predicate 115, 118, 130, 141 complementation 115, 117–118, 132–133, 138–139 – clausal verb complementation 115, 117– 118 – finite complementation 132, 139 – non-finite complementation 132–133, 138–139 – (non-)zero-complementation 132 complementizer 139, 141 conditional inference tree 133–134, 136–137 construal 116 conventional subject 117 CTP 115–117, 120, 122, 124–125, 127–128, 132, 138–142 degree of certainty 125 ECM construction 119, 141 epistemic value 116 grounded process 116 internal structure 120 intervening material 122, 132, 134, 136–139, 142 logistic regression model 122, 129–131, 141 matching problem 115 nominal predicate 127–128 NP 119–120, 128, 135 NP + XP-combination 127–128 online discourse 116 phrasal projection XP 119 PP 119 predicate geometry 117 principle 118, 122 – Cognitive Complexity principle 122 – horror aequi principle 118 probabilistic 116, 119–120, 123, 129 proposition’s validity 116 register 118, 123

relation 116–117, 120, 122, 124, 127–128, 135, 139–141 – anterior relation 132 – posterior temporal relation 124 – predicative relation 120, 128 – semantic predication relation 117 – stative relation 135 – temporal relation 116, 122, 124, 127, 139– 141 temporal path 116 semantic specialization 122, 140, 142 shifted vantage point 116 simultaneity 127, 135, 138–139, 141–142 speaker’s assessment 125 speaker’s epistemic stance 125, 132 structural complexity 122, 132 subject 134–135, 137, 139–140, 142 – CC-subject 135, 137, 139–140 – complex subject 134 – non-complex subject 134, 142 subject-predicate relation 119 time reference 140 variable 127–128, 130 – binary (response) variable 128, 130 – multi-level variable 127 – three-level variable 127 variation 116–117, 119–120, 122–123, 129– 131 – CC-variation 117, 119, 122, 129–131 – non-categorical variation 131 – probabilistic variation 120, 123 – probabilistic complement clause variation 116 verb 116, 119, 122–123, 126–128, 140–141 – copula verb 127–128 – emotive verb 122 – factual verb 122 – mental verb 126 – perception verb 140 – suasive verb 122 – volition verb 116, 119, 141 Hill ambiguous cue 154 Agree 156, 165 bi-clausal 161, 166 Case 150, 153

Subject index

cartographic approach 154 clausal negation 154 clause 150, 152, 154–155, 158, 162, 165 – adverbial clause 158, 164 – anaphoric indicative clause 152 – de-indicative clause 154 – embedded clause 155, 162 – infinitive adverbial clause 158 – non-finite clause 150, 154, 165 – non-finite relative clause 150 – relative clause 150 – să-subjunctive clause 147, 162 clause border transparency 155 clause segment 154 complementizer 149, 151–153 – infinitive complementizer 152–153 – lexical complementizer 151 – non-finite complementizer 149 complex tense 161 control 156, 163, 165 – obligatory control 156, 163, 165 – optional control 163, 165 cycle 165 distance Agree 156, 165 DP 150–151 – direct object DP 150 – regular DP 151 dual categorization 149 enclitic definite article 148–149 feature 155, 159 – finiteness feature 159 – modality feature 159 – non-finiteness feature 159 – phi-feature 155 feature transfer 155 field 154, 156–157, 159–162 – CP field 154, 156–157, 159–162 – inflectional field 154, 161 – TP field 154, 159 FinP 154–155, 161–165 Focus 154, 156–157, 159–160 ForceP 154, 156–157, 162–165 functional projection 162 grammatical mood 159–160, 162 head 159–161 infinitive replacement 147, 162 infinitive stem 149

left periphery 154–155 marker 148, 155, 157–158, 161 – infinitive marker 157 – infinitive mood marker 158 – mood marker 148, 155, 158, 161 – subjunctive marker 156 – subjunctive mood marker 155, 161 mono-clausal 161 movement 156, 161 – DP movement 156 – NP movement 156, 161 nominalization 149, 153 non-phasal 153, 155, 163–165 non-thematic 162 particle 155 phasal 154–156, 162, 165 phrase 154 – Complementizer Phrase 154 – inflectional phrase 154 position 153, 156, 161, 164 – A-position 161 – argumental position 153, 156, 161 – complement position 164 – pre-verbal argumental position 161 – subject position 153 – Topic position 161 PP 150 preposition 148–151, 153, 157–158 primary linguistic data 154 qualifying adjective 149 replacement process 147 Romanization 158 spell out 161, 165–166 subject 155, 161, 163–164 – controlled subject 155 – DP subject 155, 161 – lexical subject 164 – Nominative subject 155, 160 – non-controlled subject 163 – null uncontrolled subject 164 – pre-verbal subject 161 Topic 156, 159–161 truncation 161, 164 verb 150, 152, 154, 156, 161–164 – aspectual verb 150, 152, 163–164 – causative verb 152, 163–164 – control verb 156, 162

347

348 – – – –

Subject index

finite verb 154 non-finite verb 154 non-thematic verb 162 raising verb 156, 161

Speyer AspP 183 assignment 174, 181–183, 185, 188–189 – aspectual assignment 181–182 – status assignment 174, 182, 185, 188–189 – temporality assignment 182–183, 189 bi-constituental analysis 184 causation 184–186 clause 171, 173–174, 176, 178, 181–184, 188 – AcI clause 171 – complement clause 174, 182, 188 – dependent interrogative clause 174 – infinitival/infinitive clause 173, 176, 178, 181–184, 188 – matrix clause 171, 178, 188 – subordinate clause 173 – wie-clause 173 clause nucleus 172 clause union 171, 173, 188–189 coherence 171–173, 175, 188–189 competing grammar model 177–178 complement 169, 172–173 – PP-complement 172 – propositional complement 173 – sentential complement 169 complementizer 169, 174 construction 169–170, 174, 185 – AcI construction 169–170 – control (infinitive) construction 169–170 – infinitival/infinitive construction 169, 174, 185 – raising construction 169 external causer 184–186, 189 information density management 173 interrogative adverb of manner 174 IP-architecture 183 logical subject 170–171, 174 physical involvement 186–187 predicate 176, 184 – directive matrix predicate 176 – secondary predicate 184 PRO 170–171

proposition 174, 182, 188 realis modality 175, 182 Rheinische Verlaufsform 179 role 184–185 – agent role 184 – experiencer role 184 – semantic role 185 – stimulus role 184 – thematic role 185 sameness of events 188 status 171–172, 175–180, 182–183, 185–186, 188–190 – status I 171, 175–180, 182, 185–186, 188– 189 – status II 172, 175–179, 183, 185, 188–190 status system 183–184 structure 171, 189 – deep structure 171, 189 – surface structure 171 verb 172–174, 181–182, 184–185, 189 – atelic verb 184 – bivalent verb 184 – control verb 172, 174, 181–182 – matrix verb 173, 184–185 – object control verb 172, 182 – perception verb 173–174, 185, 189 – telic verb 184 – trivalent verb 184 verb phrase 170 verb-verb compound 172 verbal complex 173 Dékány Ablative 200 Agree based PRO analysis 216 agreement 193, 195–196, 209–210, 213, 215–217 – default agreement 195–196, 213, 216–217 – infinitival agreement 193, 209–210, 215 – regular agreement 214–215 agreement marker 210, 214, 217 anti-agreement 195–196, 207, 210–213, 216–217 – infinitival anti-agreement 210, 213, 217 – possessive anti-agreement 213 anti-agreement effect 212–213 argument 193–194, 199–200

Subject index

– Dative (noun phrase) argument 193 – DP argument 199 – noun phrase argument 194, 200 Dative case 199, 212 Dative experiencer 200, 206, 210 DP 199, 206, 210, 212 – Ablative DP 199 – Dative DP 206, 210, 212 ending 195, 198, 209–211, 214–215, 217 – default ending 215 – infinitival ending 198, 209–211, 214, 217 – person-number ending 195 feature specification 215 feature193–195, 198, 201, 212, 214–217 – φ-feature(s) 93, 195, 198, 201, 212, 214– 217 – person feature 215–216 – person and number feature 194 hanging topic 212 infinitival agreement mismatch 215 infinitival complementizer 195 infinitival paradigm 194 movement 217 – A-movement 217 – A-bar movement 217 movement analysis 216–217 number value 215 noun phrase 193–194, 200, 211, 213, 215 – Ablative noun phrase 200 – Dative noun phrase 193–194, 211, 213 – non-singular noun phrase 215 permissee 199 – Accusative permissee 199 – Dative permissee 199 position 217 – infinitival subject position 217 – matrix object position 217 possessor 212–213 – Dative possessor 212 – DP-internal (plural) pro possessor 212 – external possessor 212 – lexical possessor 212 – pronominal possessor 212–213 – pro possessor 212 possessor extraction 212 possessum 209, 212 predicate 193–194, 199–200, 204, 206–207

349

– evaluative predicate 199 – matrix predicate 193–194, 199–200, 204, 206–207 – monadic matrix predicate 199, 207 – nominal predicate 199, 206 – verbal predicate 206 privative syntactic feature 215 PRO 199–200, 210, 216 raising verb 199 structure 199–200, 212–213, 216 – Ablative control structure 200 – Dative control structure 200 – co-indexation structure 212 – control structure 199–200 – ECM structure 199, 216 – possessive structure 213 – raising structure 213 subject 193, 199–200, 206, 210, 213–217 – φ-feature independent subject 216 – Dative-marked lexical subject 206 – infinitival subject 193, 210, 217 – Nominative subject 215 – overt subject 206, 210 – pronominal subject 213 – pro subject 199 – PRO subject 199–200, 210 – referentially independent (Dative) subject 199, 216–217 – wh-subject 214 subject-to-subject raising 217 suffix 194–195, 199, 207–211, 213–215 – agreement suffix 199, 207, 209 – infinitival suffix 194–195, 207–210, 213– 215 – monomorphemic infinitival suffix 207 – nominalizing suffix 211 – zero suffix 210 trace 199 verb 199, 214 – monosyllabic verb 214 – permissive verb 199 Lühr anaphor 241 antecedent 226, 239, 253 appositive nominative reference 236 arbitrary controller 227

350

Subject index

atelic activity 248 binding principle A 241 case attraction 251 clause 223, 225–240, 243–249, 250–252, 254–256, 258, 260–261 – complement clause 252 – in-order-to-clause 225 – infinitive clause 227, 256 – main clause 230, 248, 258 – matrix clause 228–229, 232, 234, 238– 240, 249, 251, 255, 260 – object-gap purpose clause 225, 227 – objective clause 225 – purpose clause 225–229, 231–236, 239– 240, 243–248, 250–252, 254–255, 260 – relative clause 252 – rationale clause 225–226, 228, 231–232, 234, 236–239, 247, 255, 258, 260–261 – small clause 223 – subject-gap purpose clause 225 – subordinate clause 223, 229–230, 234, 240, 243 complementizer 229 construction 223–227, 229, 231–234, 236, 238–240, 245–247, 249–255, 257–261 – infinitive construction 226, 231–234, 236, 239–240, 245–247, 249–255, 257, 260– 261 – (finite) purpose construction 223–225, 227, 229, 231, 233, 236, 238, 254, 258– 260 contrast focus 243 control 223, 226–228, 234, 236, 247–248, 250, 258, 260 – arbitrary control 227–228, 234, 236, 250, 260 – non-obligatory control 226–227 – obligatory control 226, 228 – subject control 223, 228, 247–248, 258 double dative 251, 260 exceptional case marking 246 event nominalization 240 gap 228, 250 maximal projection 223 modality 224 – deontic modality 224 – teleological modality 224

nominative 234, 236, 238–240, 243, 247, 260 NP 226–227, 241 – dative NP 241 – lexical NP 226 – matrix NP 227 oblique co-argument 241 predicate 224 – desiderative predicate 224 – manipulative predicate 224 (controlled) PRO 226, 236, 253 pronoun 227–228, 231, 243, 249, 257 – enclitic pronoun 231 – overt pronoun 227, 243 – personal pronoun 257 – possessive pronoun 228, 249 purposive sub-sense 230 purposivity 231, 260 subject 227, 236, 239–241, 247–248, 255– 258, 258, 260 – arbitrary null subject 240 – covert subject 236, 239, 247, 257 – exhaustive subject 227 – inherent subject 258 – intentional subject 248 – overt subject 236, 240–241, 247, 255– 256, 258, 260 – purpose clause subject 227 verbal ending 223–224, 259 verbal noun 223, 227, 254–256 verbal substantive 254–255 verb 228, 231, 244, 246–247, 260 – motion verb 231 – speech act verb 244, 246–247, 260 – transitive verb 228 – verb of choice and use 228 VP-internal 228 – wh-extraction 225 Wiemer adverbial subordinator 301 agent 277, 282, 285–286, 290–291, 299, 307, 314, 319, 322 anti-resultative 315–316 argument 274, 278, 295, 306, 310, 319 assertiveness 268, 271, 273–277, 301, 305 attitude 271, 273, 276, 298

Subject index

– cognitive attitude 271, 273 – emotive attitude 298 – volitional attitude 276 clausal head 309 clausal node 309 clause 265–267, 269–271, 276–277, 291– 292, 294, 298–303, 305–306, 311–312, 315, 321, 323–325, 327–330 – adjunct clause 328 – concessive clause 301 – conditional clause 277, 291–292, 300, 321 – conjoined clause 267 – consecutive clause 301, 323–324, 327 – counterfactual clause 294 – da-clause 270, 302–303, 305, 311, 325, 327, 329–330 – dependent clause 267, 269–270, 276, 300 – final clause 301, 303 – finite clause 269, 271, 301 – habitual-consecutive clause 323 – headless relative clause 312 – independent clause 265–266, 271, 292 – infinitive clause 298–299, 312, 331 – (true) interrogative clause 275 – main clause 268, 272–273, 278, 296, 298, 300, 302, 305–306, 315 – non-embedded clause 302 – nonfinite clause 265 – purpose clause 277, 300 – subordinate clause 265 clause-type 275 complementizer 265–266, 272–273, 298, 300–301, 303, 305, 326–327, 329, 331 – apprehensive complementizer 298 – neutral complementizer 298, 305, 329 conditional mood 320 conjunction 266, 298, 300–301, 326–327, 329, 331 connective 266, 328 – adjunctive connective 328 – paratactic connective 328 – subordinating connective 266 construction 272, 307–308, 314 – hybrid construction 308 – mihi est liber-construction 307 – optative construction 314 – source construction 272

351

constructional economy 279 constructional frame 283 contrastive stress 274 control 269 deonticity 284 discourse function 279 ellipsis 268, 276, 321, 331 epistemic commitment 272 epistemicity 266, 279 event 280, 292 – non-factive event 280 – potential event 292 feature 269, 325 – agreement feature 269 – morphological feature 269, 325 – tense feature 269 finiteness 265–271, 279, 305, 325 function 266, 281–282, 285, 288–290, 292– 295, 297, 299, 302, 307, 315–316, 320, 322, 325–328 – adversative function 326 – counterfactual function 294, 315–316 – deontic function 290 – directive function 285, 290, 320, 327 – echoing function 281, 293 – illocutionary/illocutive function 288, 297, 299, 302 – hortative function 302, 325–328 – modal function 281–282, 292, 294–295, 307, 315 – narrative function 266, 322 – optative function 293, 302 – preventive function 292 – prohibitive function 289 – subordinating function 302 future indicative 275 gerundive 320 illocutionary force 269 implicature 281, 320 – discourse-conditioned implicature 281 – modal implicature 320 incentive particle 278 inflection marker 301 ingressive effect 278 ingressivity 315 insubordination 265–268, 272–273, 275– 277, 300, 305, 309, 315, 324, 329–332

352

Subject index

– deontic insubordination 272 – epistemic insubordination 272 – modal insubordination 272 interrogative clitic 302 main clause phenomenon 305, 331 markedness reversal 276 marker 279, 283, 307–309, 326, 329 – affirmative marker 326 – epistemic marker 279, 283, 329 – protasis-marker 326 – tense-mood marker 307–309 mirative statement 272, 276, 302 modal background 302 modifier 275–276 – evaluational modifier 275–276 – evidential modifier 275 modality 266, 272, 276–277, 279, 282–283, 290, 300, 308, 310, 312, 324 – alethic modality 282 – circumstantial modality 272, 276 – deontic modality 276, 283, 308, 310, 312 – dispositional modality 308, 324 – dynamic modality 276, 283, 290, 308, 310, 312 – epistemic modality 266, 277, 279, 283, 300 – fatalistic modality 283 – propositional modality 266 – root modality 277, 300 narrativity 278 negation 283–285, 287–288, 290, 296, 299, 318, 324 – external negation 283 – internal negation 283 negative polarity 282, 287, 295, 317–318 NP 299, 307–310, 319 – dative NP 299, 307–308, 310, 319 – nominative/nominatival NP 307–309 particle 287, 293, 297, 320–321, 326 – enclitic particle 297 – focus particle 297, 326 – optative particle 293, 320–321 – sentence-initial particle 326 – subjunctive particle 287 possessee 307–308 predicate 269–270, 272–274, 276, 278–279, 310–311

– complement-taking predicate 269, 272 – dynamic predicate 279 – existential predicate 311 – evaluational predicate 276 – factive predicate 273–274 – individual level predicate 278 – main clause infinitival predicate 270 – modal predicate 310 – nominal predicate 310 – state predicate 278 predicate-argument relation 308 predicative possession 307–308 pronoun 266, 270, 287, 294, 296–297, 299 – dative pronoun 294 – demonstrative pronoun 326 – logophoric pronoun 266 – nominatival pronoun 296 – (independent) subject pronoun 270 propositional content 272, 274–276, 317 proverb 277 question 271, 276, 281–282, 286–287, 289, 291, 302–303, 313, 317, 320 – deliberative question 271, 276, 281, 287, 291, 317 – dynamic question 286 – egocentric question 287, 289 – pseudo-question 282 – rhetoric question 286, 303 – suggestive question 302–303 – yes/no-question 286–287, 289, 313, 320 raising 269 reflexive binding 269 semantic integration 329 speech 292–293, 295, 319, 326, 330 – colloquial speech 292, 330 – dialogical speech 295 – folk(lore) speech 319, 326, 330 – self-admonishing speech 293 – self-directed speech 292 speech act 271–273, 276–277, 281, 293, 314, 327 – deontic speech act 271 – directive speech act 271–273, 276, 293, 327 – non-assertive speech act 271

Subject index

– volitional speech act 271 structure 269, 300, 310–311 – bi-clausal structure 296, 300 – mono-clausal structure 269, 310–311 – predicate-argument structure 269 – predicative structure 311 subject 269–270, 273, 277, 279, 289, 296, 306, 314, 317, 322, 324 – independent subject 269–270, 279 – nominative/nominatival subject 279, 306, 322, 324 – third-person subject 296 subjunctive 266, 287, 291, 296, 298, 305, 317, 325 time reference 272, 279 truth value 271, 273, 275 truth value condition 274 unexpectedness 276, 278–279 utterance 273, 275–276, 281, 285, 291–293, 297–298, 300, 311–312, 314 – apprehensive utterance 297–298, 300, 314 – deontic utterance 285, 312 – directive utterance 292–293, 311 – egocentric utterance 293 – evaluational utterance 275 – factive utterance 275 – isolated utterance 276, 281 – non-factive utterance 273, 276, 297 – optative utterance 291, 298 verb 270, 272, 276, 283–284, 286, 299– 301, 309–311, 314–315, 327, 329 – auxiliary verb 270, 300–301 – copula verb 311 – desiderative verb 329 – event-bound verb 284 – existential verb 309 – imperfective verb 284, 299, 315 – lexical verb 270, 300–301 – manipulative verb 329 – perfective verb 283–284, 286, 299 – phasal verb 270, 300–301, 329 – semi-copular verb 310 – telic verb 284 volition 272, 276, 284, 314, 327

353

2.2 Explanation approaches taken all together Ablative 199–203, 213 accusative clitic 72 adjacency 72–73 adjunct 62, 70–72, 76–77, 127–128, 141, 199–201, 203, 225, 251–252, 257, 260, 306, 308, 328 – infinitival adjunct 77 – purpose adjunct 70–72, 77 – subjunctive adjunct 62 adverbial subordinator 301 agent 184, 229, 242, 255–256, 277, 282, 285–286, 290–291, 299, 307, 314, 319, 322 Agree 156, 165, 216 Agree based PRO analysis 216 agreement 21, 75, 193, 195–196, 209–210, 213–217, 259, 325 – default agreement 21, 195–196, 213, 216– 217 – infinitival agreement 193, 209–210, 215 – regular agreement 214–215 agreement marker 210, 214, 217 alternation 39, 118–120, 123, 148, 152, 158, 160, 164 alternation pair 116 alternation pattern 115, 117 anteriority 18, 81, 86–91, 93–103, 105–106, 127, 131–133, 137, 139–141 ambiguous cue 154 anaphor 241 antecedent 41, 226, 239, 253 anti-agreement 21, 195–196, 207, 210–213, 217 – infinitival anti-agreement 207, 210, 213, 216–217 – possessive anti-agreement 213 anti-agreement effect 212–213 anti-resultative 315–316 AP 119 appositive nominative reference 236 argument 30–40, 48–49, 64, 193–194, 199– 200, 274, 306 – clausal argument 274, 306

354

Subject index

– event argument 49 – experiencer argument 48 – Dative (noun phrase) argument 193 – DP argument 199 – individual argument 40, 49 – noun phrase argument 193–194, 200 – proposition-denoting argument 39 – propositional argument 40, 64 aspectual adverbial 74 AspP 183 assertiveness 268, 271, 274–277, 301, 305 assignment 174, 181–183, 185, 188–189 – aspectual assignment 181–182 – status assignment 174, 182, 185, 188–189 – temporality assignment 182–183, 189 atelic activity 248 attitude 271, 273, 276, 298 – cognitive attitude 271, 273 – emotive attitude 298 – volitional attitude 276 bi-clausal 16–17, 21, 31, 34, 42, 161, 166, 269, 296, 300–301, 307, 311, 327, 329 bi-constituental analysis 184 binary logistic regression 117, 129–131, 135, 141 binding principle A 241 cartographic approach 35, 154 Case 150 case attraction 251 categorical 116, 119, 175, 177, 189 category 34, 45–46, 48 – aspectual category 48 – factive category 34 – functional category 45–46 – modal category 34 – motion category 34 – propositional category 34 causation 184–186 CC 119, 122, 130, 132, 134, 139 – that-CC 119, 122, 130 – zero-CC 122, 130, 132, 134, 139 Cinque-style hierarchy 33 clausal head 309 clausal height 49 clausal negation 154 clausal node 309 clause 31–32

– – – – –

– – – – – – – – – – – – – – – – – – – – – – – – – – – –

– – – – – – –

AcI clause 81–106, 171 adjunct clause 328 adverbial clause 158, 164 anaphoric indicative clause 152 complement clause 31–32, 116, 118, 122, 126–128, 137, 139–140, 174, 182, 188, 252 concessive clause 301 conditional clause 277, 291–292, 300, 321 conjoined clause 267 consecutive clause 301, 323–324, 327 counterfactual clause 294 da-clause 270, 302–303, 305, 311, 325, 327, 329–330 de-indicative clause 154 declarative infinitive clause 83 dependent clause 267, 269–270, 276, 300 dependent interrogative clause 174 embedded clause 155, 162 final clause 301, 303 finite clause 269, 271, 301 finite complement clause 82–84, 95, 105 fully tensed complement clause 82 gerundial -ing-clause 115–116 habitual-consecutive clause 323 headless relative clause 312 hybrid complement clause 83, 92 in-order-to-clause 225 independent clause 265–266, 271, 292 indicative complement clause 84 infinitival/infinitive clause 173, 176, 178, 181–184, 188 infinitive adverbial clause 158 infinitive complement clause 84 (true) interrogative clause 275 main clause 230, 248, 258, 268, 272–273, 278, 296, 298, 300, 302, 305–306, 315 matrix clause 31, 122, 127, 171, 178, 188, 228–229, 232, 234, 238–240, 249, 251, 255, 260 non-embedded clause 302 non-finite clause 150, 154, 165 non-finite complement clause 83, 88, 122 non-finite relative clause 150 object-gap purpose clause 225, 227 objective clause 225 passive ECM-clause 124

Subject index

– purpose clause 225–229, 231–236, 239– 240, 243–248, 250–252, 254–255, 260, 277, 300 – rationale clause 225–226, 228, 231–232, 234, 236–239, 247, 255, 258, 260–261 – reduced clause 119 – relative clause 150, 252 – să-subjunctive clause 147, 162 – Small Clause 117–119, 141, 223 – subject-gap purpose clause 225 – subordinate clause 173, 223, 229–230, 234, 240, 243, 265 – verbless clause 117 – wie-clause 173 – zero-complementizer clause 117, 121 clause-boundedness 32 clause-type 275 clause border transparency 155 clause nucleus 172 clause segment 154 clause union 31, 171, 173, 188–189 clause union effect 17 clitic climbing 17, 31, 34–35, 37, 41, 48, 76– 77 clustering 33 Cognitive Grammar 116 coherence 31, 33, 171–173, 175, 188–189 competition 119 competing grammar model 177–178 complement 36, 42, 64, 72, 116, 119, 121, 123, 125, 127, 130, 134–142, 169, 172– 173 – CP complement 42 – finite complement 36, 64 – non-finite complement 64, 72 – PP-complement 172 – propositional complement 173 – SC-complement 119, 121, 123, 125, 127, 130, 135–136, 140–141 – sentential complement 169 – that-complement 116, 136–137, 139–140, 142 – zero-complement 121, 134, 136, 138–142 complement choice 140 complement type 115–116, 118, 120, 130, 140–141

355

complementation 115, 117–118, 132–133, 138–139 – clausal verb complementation 115, 117– 118 – finite complementation 132, 139 – non-finite complementation 132–133, 138–139 – non-zero-complementation 132 complementizer 149, 151–153, 298, 305, 329 – apprehensive complementizer 298 – infinitive complementizer 152–153 – lexical complementizer 151 – neutral complementizer 298, 305, 329 – non-finite complementizer 149 complex tense 161 conditional inference tree 133–134, 136–137 conjunction 266, 298, 300–301, 326–327, 329, 331 connective 328 – adjunctive connective 328 – paratactic connective 328 subordinating connective 266 connector 71 construal 116 construction – AcI construction 9, 18, 20, 169–170 – control (infinitive) construction 169–170 – ECM construction 119, 141 – (finite) purpose construction 223–225, 227, 229, 231, 233, 236, 238, 254, 258– 260 – hybrid construction 308 – impersonal construction 70 – infinitival/infinitive construction 169, 174, 185, 226, 231–234, 236, 239–240, 245– 247, 249–255, 257, 260–261 – long passive construction 32 – mihi est liber-construction 307 – optative construction 314 – pseudo-relative construction 60 – raising construction 169 – reflexive-impersonal construction 61, 77 – source construction 272 constructional economy 279

356

Subject index

constructional frame 283 contrast focus 243 contrastive stress 274 control 1–3, 9, 14, 17, 20–22, 37–38, 156, 163, 165, 223, 226–228, 234, 236, 247– 248, 250, 260, 269 – arbitrary control 227–228, 234, 236, 250, 260 – exhaustive control 37–38 – non-obligatory control 226–227 – object control 2–3, 9, 21 – obligatory control 156, 163, 165, 226, 228 – optional control 163, 165 – partial control 37–38 – subject control 9, 21–22, 223, 228, 247– 248, 258 controller 227 conventional subject 117 CP 64, 72, 74 CTP 115–117, 120, 122, 124–125, 127–128, 132, 138–142 cycle 165 Dative case 199, 212 Dative experiencer 200, 206, 210 degree of certainty 125 deonticity 284 dependent variable 42, 49 distance Agree 156, 165 double dative 251, 260 DP 150–151, 199, 206, 210, 212 – Ablative DP 199 – Dative DP 206, 210, 212 – direct object DP 150 – regular DP 151 dual categorization 149 ECM 72 ellipsis 268, 276, 321, 331 embedded tense 34 enclitic definite article 148–149 ending 195, 198, 209–211, 214–215, 217 – default ending 215 – infinitival ending 198, 209–211, 214, 217 – person-number ending 195 – verbal ending 223–224, 259 epistemic commitment 272 epistemic value 116 epistemicity 266, 279

event 49, 61–62, 76–77, 280, 292 – factual event 61, 76 – intentional event 62, 77 – non-factive event 280 – possible event 61, 76 – potential event 292 – simultaneous event 61 – speech event 49 – VP event 49 event nominalization 240 exceptional case marking 246 external causer 184–186, 189 extraposition 16 feature 155, 159, 193–195, 198, 201, 212, 214–217, 269, 325 – φ-feature(s) / phi-feature(s) 55, 193, 195, 198, 201, 212, 214–217 – agreement feature 269 – finiteness feature 159 – modality feature 159 – morphological feature 269, 325 – person feature 215–216 – person and number feature 194 – non-finiteness feature 159 – tense feature 269 feature specification 215 feature transfer 155 field 154, 156–157, 159–162 – CP field 154, 156–157, 159–162 – inflectional field 154, 161 – TP field 154, 159 finiteness 265–271, 279, 305, 325 FinP 154–155, 161–165 Focus 154, 156–157, 159–160 ForceP 154, 156–157, 162–165 function – adversative function 326 – counterfactual function 294, 315–316 – deontic function 290 – directive function 285, 290, 320, 327 – discourse function 279 – echoing function 281, 293 – hortative function 302, 325–328 – illocutionary/illocutive function 288, 297, 299, 302 – modal function 281–282, 292, 294–295, 307, 315

Subject index

– narrative function 266, 322 – optative function 293, 302 – preventive function 292 – prohibitive function 289 – subordinating function 302 functional domain 74 future indicative 275 gap 228, 250 gerundive 320 grounded process 116 hanging topic 212 head 17, 33, 35–38, 41–43, 46–47 – aspectual head 17, 36, 42–43, 47 – functional head 17, 33, 35–36, 38, 41–43, 46 – inflectional head 33, 47 illocutionary force 269 implicature 281, 320 – discourse-conditioned implicature 281 – modal implicature 320 incentive particle 278 indicative 94–95 – aorist indicative 94–95 – present indicative 95 indirect interrogative 68 individual style 94, 97 infinitival agreement mismatch 215 infinitival complement 57, 63, 74 infinitival complementizer 195 infinitival paradigm 194 infinitival subject position 217 infinitive replacement 147, 162 infinitive stem 149 infinitivus pro participio effect 34 inflectional layer 33, 35–36, 41–42, 45 information density management 173 ingressive effect 278 ingressivity 315 insubordination 265–268, 272–273, 275– 277, 300, 305, 309, 315, 324, 329–332 – deontic insubordination 272 – epistemic insubordination 272 – modal insubordination 272 internal structure 120 interrogative adverb of manner 174 interrogative clitic 302

357

intervening material 122, 132, 134, 136–139, 142 IP-architecture 183 left periphery 154–155 logistic regression model 122, 129–131, 141 long inversion 16 long passive 17, 32, 34 main clause phenomenon 305, 331 markedness reversal 276 marker 148, 155–158, 161, 279, 283, 301, 307–309, 326, 329 – affirmative marker 326 – epistemic marker 279, 283, 329 – inflection marker 301 – infinitive marker 157 – mood marker 148, 155, 158, 161 – protasis-marker 326 – subjunctive marker 156 – tense-mood marker 307–309 matching problem 115 matrix object position 217 mirative statement 272, 276, 302 modal background 302 modal flavor 49 modality 224, 266, 272, 276–277, 279, 282– 283, 290, 300, 308, 310, 312, 324 – alethic modality 282 – circumstantial modality 272, 276 – deontic modality 224, 276, 283, 308, 310, 312 – dispositional modality 308, 324 – dynamic modality 276, 283, 290, 308, 310, 312 – epistemic modality 266, 277, 279, 283, 300 – fatalistic modality 283 – propositional modality 266 – realis modality 175, 182 – root modality 277, 300 – teleological modality 224 modifier 275–276 – evaluational modifier 275–276 – evidential modifier 275 mono-clausal 161 monoclausality 31, 33, 37 mood 41, 159–160, 162, 320

358

Subject index

– conditional mood 320 – evidential mood 41 – grammatical mood 159–160, 162 mood marker 155, 158, 161, 307–309 – infinitive mood marker 158 – subjunctive mood marker 155, 161 – tense-mood marker 307–309 movement 10, 21, 156, 161, 165, 217 – A-bar movement 217 – A-movement 21, 217 – DP movement 156 – NP movement 156, 161 movement analysis 216–217 narrativity 278 negation 5, 15–16, 22, 74, 154, 156–157, 159–160, 257, 283–285, 287–288, 290, 296, 299, 318, 324 – external negation 283 – internal negation 283 negation operator 5 negative polarity 282, 287, 295, 317–318 nominalization 149, 153 nominative 234, 236, 238–240, 243, 247, 260 non-phasal 153, 155, 163–165 non-thematic 162 noun phrase 3–4, 128, 169, 193–194, 200, 211, 213, 215, 240 – Ablative noun phrase 200 – Dative noun phrase 193–194, 211, 213 – non-singular noun phrase 215 NP 119–120, 128, 135, 226–227, 241, 299, 307–308, 310, 319 – dative NP 241, 299, 307–308, 310, 319 – lexical NP 226 – matrix NP 227 NP + XP-combination 127–128 number value 215 oblique co-argument 241 obviation effect 7 online discourse 116 particle 287, 293, 297, 320–321, 326 – enclitic particle 297 – focus particle 297, 326 – optative particle 293, 320–321 – sentence-initial particle 326 – subjunctive particle 287

(long) passivization 32, 37 perception 57, 60–61, 64 – direct perception 57, 60–61, 64 – indirect perception 56, 61 perfect 87–90, 96–97, 100–102, 105, 107 – active perfect 87–88, 90, 101–102 – active perfect of actions and processes 89 – active transitive perfect (ATP) 9–91, 96– 97, 100, 101, 105, 107 – anterior perfect 87–88, 96 – resultative perfect 87–88 permissee 199 – Accusative permissee 199 – Dative permissee 199 permutation invariance 46 phasal 154–156, 162, 165 phrase 3–5, 12, 154, 193–194, 200, 211, 213, 215 – Complementizer Phrase 154 – inflectional phrase 154 – noun phrase 3–4, 193–194, 200, 211, 213, 215 – verb(al) phrase 3, 170 – wh‑phrase 12 physical involvement 186–187 pied piping 16 probabilistic 116, 119–120, 123, 129 position 3–4, 31–32, 36, 38–39, 153, 156, 161, 164 – A-position 161 – argumental position 153, 156, 161 – clause-local position 31 – complement position 164 – final position 3 – matrix subject position 32, 39 – pre-verbal argumental position 161 – prefield position 4 – second position 3 – specifier position 36 – subject position 36, 38–39, 153 – Topic position 161 possessee 307–308 possessor 212–213 – Dative possessor 212 – DP-internal (plural) pro possessor 212 – external possessor 212 – lexical possessor 212

Subject index

– pronominal possessor 212–213 – pro possessor 212 possessor extraction 212 possessum 209, 212 PP 150 predicate-argument relation 308 predicate geometry 117 predicate 31, 33, 36–42, 44–49, 115, 118, 130, 141, 176, 184, 193–194, 199–200, 204, 206–207, 224, 269–270, 272–274, 276, 278–279, 310–311 – complement-taking predicate 115, 118, 130, 141, 269, 272 – control predicate 38–39, 48 – desiderative predicate 224 – directive matrix predicate 176 – dynamic predicate 279 – embedded predicate 37 – evaluational predicate 276 – existential predicate 311 – factive predicate 41, 46, 273–274 – individual level predicate 278 – main predicate 33 – main clause infinitival predicate 270 – manipulative predicate 224 – matrix predicate 31, 176, 193–194, 199– 200, 204, 206–207 – modal predicate 310 – nominal predicate 127–128, 199, 206, 310 – non-restructuring predicate 37 – one-place predicate 39 – raising predicate 38–39, 42, 48–49 – restructuring predicate 36–40, 44–45, 47 – secondary predicate 184 – state predicate 278 – subject-oriented predicate 42, 45 predicative possession 307–308 preposition 148–151, 153, 157–158 primary linguistic data 154 principle 118, 122 – Cognitive Complexity principle 118, 122 – horror aequi principle 118 privative syntactic feature 215 (controlled) PRO 38, 42, 170, 171, 199–200, 210, 216, 226, 236, 253 probabilistic variation 120, 123 projection 33, 36–37, 119, 162, 223

359

– clausal projection 37 – extended projection 33, 36 – functional projection 33, 37, 162 – maximal projection 223 – phrasal projection XP 119 pronominal clitic 74 pronoun 227–228, 231, 243, 249, 257, 266, 270, 294, 296, 326 – dative pronoun 294 – demonstrative pronoun 326 – enclitic pronoun 231 – logophoric pronoun 266 – nominatival pronoun 296 – overt pronoun 227, 243 – personal pronoun 257 – possessive pronoun 228, 249 – (independent) subject pronoun 270 pronoun fronting 16 proposition 174, 182, 188 propositional content 272, 274–276, 317 proposition’s validity 116 proverb 277 purposive sub-sense 230 purposivity 231, 260 qualifying adjective 149 question 271, 276, 281–282, 286–287, 289, 291, 302–303, 313, 317, 320 – deliberative question 271, 276, 281, 287, 291, 317 – dynamic question 286 – egocentric question 287, 289 – pseudo-question 282 – rhetoric question 286, 303 – suggestive question 302–303 – yes/no-question 286–287, 289, 313, 320 raising 269 reduced non-restructuring 33 reflexive binding 269 register 118, 123 relation 116–117, 120, 122, 124, 127–128, 135, 139–141 – anterior relation 132 – posterior temporal relation 124 – predicative relation 120, 128 – semantic predication relation 117 – stative relation 135

360

Subject index

– temporal relation 116, 122, 124, 127, 139– 141 replacement process 147 restructuring 33 – functional restructuring 33 – lexical restructuring 33 restructuring generalization 42 Rheinische Verlaufsform 179 role 184–185 – agent role 184 – experiencer role 184 – semantic role 185 – stimulus role 184 – thematic role 185 Romanization 158 sameness of events 188 (cross-clausal) scrambling 35 semantic integration 329 semantic specialization 122, 140, 142 sentential negation 74 shifted vantage point 116 simultaneity 127, 135, 138–139, 141–142 speaker’s assessment 125 speaker’s epistemic stance 125, 132 speech 292–293, 295, 319, 326, 330 – colloquial speech 292, 330 – dialogical speech 295 – folk(lore) speech 319, 326, 330 – self-admonishing speech 293 – self-directed speech 292 speech act 271–273, 276–277, 281, 293, 314, 327 – deontic speech act 271 – directive speech act 271–273, 276, 293, 327 – non-assertive speech act 271 – volitional speech act 271 spell out 161, 165–166 stativity 84, 88–89, 91–92, 100, 102–103, 105–106 status 171–172, 175–180, 182–183, 185–186, 188–190 – status I 171, 175–180, 182, 185–186, 188– 189 – status II 172, 175–179, 183, 185, 188–190 status system 183–184 structural complexity 122, 132

structure 31, 34, 37, 42, 171, 189, 199–200, 212–213, 216, 269, 300, 310–311 – Ablative control structure 200 – bi-clausal structure 31, 34, 42 – co-indexation structure 212 – control structure 199–200 – Dative control structure 200 – deep structure 171, 189 – ECM structure 199, 216 – mono-clausal structure 37 – possessive structure 213 – predicate-argument structure 269 – predicative structure 311 – raising structure 213 – surface structure 171 subject – φ-feature independent subject 216 – arbitrary null subject 240 – CC-subject 135, 137, 139–140 – collective subject 37 – complex subject 134 – controlled subject 155 – covert subject 236, 239, 247, 257 – Dative-marked lexical subject 206 – DP subject 155, 161 – exhaustive subject 227 – expletive subject 39 – idiom chunk subject 39 – inanimate subject 39 – independent subject 269–270, 279 – infinitival/infinitive subject 70 – inherent subject 258 – intentional subject 248 – lexical subject 164 – logical subject 170–171, 174 – Nominative subject 155, 160, 215 – non-complex subject 134, 142 – non-controlled subject 163 – null uncontrolled subject 164 – overt subject 206, 210 – pre-verbal subject 161 – pronominal subject 213 – pro subject 199 – PRO subject 42, 199–200, 210 – purpose clause subject 227 – referentially independent (Dative) subject 199, 216–217

Subject index

– third-person subject 296 – wh-subject 214 subject-predicate relation 119 subject-to-subject raising 217 subject orientation 38–39, 42, 44, 50 subjunctive 266, 287, 291, 296, 298, 305, 317, 325 suffix – agreement suffix 199, 207, 209 – infinitival suffix 194–195, 207–210, 213– 215 – monomorphemic infinitival suffix 207 – nominalizing suffix 211 – zero suffix 210 temporal mismatch 5 temporal path 116 tense morphology 82 thematic dependency 38–39, 42 time reference 140, 272, 279 Topic 156, 159–161 trace 199 transparency effect 31, 37–38 truncation 161, 164 truth value 271, 273, 275 truth value condition 274 unexpectedness 276, 278–279 utterance 273, 275–276, 281, 285, 291–293, 297–298, 300, 311–312, 314 – apprehensive utterance 297–298, 300, 314 – deontic utterance 285, 312 – directive utterance 292–293, 311 – egocentric utterance 293 – evaluational utterance 275 – factive utterance 275 – isolated utterance 276, 281 – non-factive utterance 273, 276, 297 – optative utterance 291, 298 variable 127–128, 130 – binary (response) variable 128, 130 – multi-level variable 127 – three-level variable 127 variable binding 38–40, 42–43, 45 variation 116–117, 119–120, 122–123, 129– 131 – CC-variation 117, 119, 122, 129–131 – non-categorical variation 131

361

– probabilistic complement clause variation verb 116 – aspectual verb 32, 43, 47, 150, 152, 163– 164 – atelic verb 184 – auxiliary verb 270, 300–301 – bivalent verb 184 – control verb 156, 162, 172, 174, 181–182 – copula verb 67, 127–128, 311 – desiderative verb 329 – emotive verb 122 – epistemic verb 41 – event-bound verb 284 – existential verb 67, 309 – factive verb 34, 41 – factual verb 122 – imperfective verb 284, 299, 315 – implicative verb 47–48 – intentional verb 76 – intransitive verb 89 – lexical verb 33, 42, 49, 270, 300–301 – manipulative verb 329 – matrix verb 173, 184–185 – mental verb 126 – modal verb 32 – non-obligatory control verb 55–56 – object control verb 172, 182 – obligatory control verb 55 – particle verb 45 – (intentional) perception verb 55–57, 60– 62, 70, 72, 75–77, 140, 173–174, 185, 189 – perfective verb 283–284, 286, 299 – phasal verb 270, 300–301, 329 – raising verb 156, 161, 199 – restructuring verb 33–38 – semi-copular verb 310 – speech act verb 244, 246–247, 260 – stative verb 84–86, 88 – suasive verb 122 – telic verb 184, 284 – terminative verb 88 – transitive verb 228 – trivalent verb 184 – verb of choice and use 228 – verb of ordering 83 – verb of physical perception 56–57, 61, 63

362

Subject index

– verb of saying 83–84, 105 – verb of speech 38, 41 – verb of thinking 83–84, 105–106 – volition verb 116, 119, 141 verb-verb compound 172 verbal complex 173 verbal noun 223, 227, 254–256

verbal substantive 254–255 verb raising 34, 49 verb second rule 3 volition 272, 276, 284, 314, 327 VP-internal 228 wh-extraction 225 Wh-Infinitive-Generalization 10